Andronov Bifurcations

NASA
T.-l
? i
"-<Sh
A. A. ANDRONOV, E.A. LEONTOVICH, q.1
1. I. GORDON, and A.G. MAIER
THEORY OF BIFURCATIONS
OF DYNAMIC SYSTEMS
ON A PLA N E
J '~
, . ' ,- . - .
Translated from Russian 2 .
.
:..
,u
~
-- >?"
Published for t h e National Aeronautics and Space Administratign , t '

and t h e National Science Foundation, Washington, D.C. -:-.2--/
by the Israel Program for Scientific Translations
TECH LlSRARY KAFB, NM
1I111llillll00b9000
llllliIIililI I
A.A. Andronov, E.A. Leontovich,
1.1. Gordon, and A.G. Maier
THEORY OF BIFURCATIONS
OF DYNAMIC SYSTEMS
ON A PLANE
(Teoriya bifurkatsii dinamicheskikh sistem
na ploskosti)
Izdatel stvo Nauka

Glavnaya Redaktsiya
Fiziko-Matematicheskoi Literatury
Moskva 1967
Translated from Russian
Israel ?rogram for Scientific Translations

Jerusalem 1971
I TT 69-55019
NASA TT F-556
I
Published P u r s u a n t to an A g r e e m e n t with
THE NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
and
T H E NATIONAL SCIENCE FOUNDATION, WASHINGTON, D. C.
Copyright 8 1971
Iaael Rogram for Scimtific Trandadonr L?d#
IPST Cat. NO. 5438
T r a n s l a t e d by IPST staff
Printed in Jerusalem by Katcz Pre8a

Binding: Wiener Hndery Ud., Jerusaiem
Available f r o m the
u. s. DEPARTMENT OF COMMERCE
National Technical Information Service
Springfield, Va. 22151
I XII/lB/S I
PREFACE
The present monograph is a d i r e c t continuation of o u r Q u a1 i t a t i v e

T h e o r y o f S e c o n d - O r d e r D y n a m i c S y s t e m s (QT) published
in 1966. It may a l s o be considered as a second volume of the definitive
t r e a t i s e on dynamic s y s t e m s and t h e i r applications planned by A. -4.
Andronov back in the 1040s. All t h i s notwithstanding, however, T h e o r y
o f B i f u r c a t i o n s can be t r e a t e d a s an independent volume and the
r e a d e r is only expected t o be acquainted with the basic concepts of the
qualitative theory of differential equations on a plane.
In distinction f r o m QT, the g r e a t e r p a r t of which is devoted to the
c l a s s i c a l theory (of Poincar; and Bendixson), T h e o r y o f B i f u r c a t i o n s
p r e s e n t s relatively recent r e s u l t s vihich w e r e obtained during the l a s t
t h r e e decades and published - in part or completely - in a number of
l e t t e r s and p a p e r s in scientific journals. T h e s e r e s u l t s a r e closely linked
t o the theory of oscillations and have by now found many important u s e s in
physics and engineering.*
The present book, like QT, w a s begun by A. A. Andronov, E. A.
Leontovich, and A. G . Maier and completed by E. A. Leontovich and
I. I. Gordon. K. A. Gubar' and R. R.I. Mints a l s o took part in the preparation
of the monograph, the f o r m e r being responsible f o r Chapter VI11 and the
l a t t e r f o r part of Chapter X N . The final version was prepared by
I. I. Gordon.
The main r e s u l t s presented in Chapters I11 through VI1 were derived by
A. A. Andronov and L. S. Pontryagin, and those in Chapters IX through XI1
by A. A. Andronov and E. A. Leontovich. Chapter VI11 is based on the work
of N. A. Gubar' and t h e r e s u l t s of Chapter XI11 are due t o E. A. Leontovich,
A. G. Maier, and L. S. E'ontryagin. The g e n e r a l editing of t h e book was
undertaken by Yu. M. Romanovskii.
T h e book naturally falls into two p a r t s - the theory of s t r u c t u r a l l y
stable s y s t e m s (Chapters I through VII) and the theory of bifurcations
(Chapters VI11 through XIV). The second part is l a r g e l y independent of the
f i r s t , and t h e r e a d e r will only r e q u i r e s o m e basic information f r o m
Chapters I, 11, IV, and V.
-Although the book contains numerous r e f e r e n c e s t o QT, many of t h e s e
refer t o proofs of w e l l known or relatively s i m p l e and obvious propositions
contained in QT, and the r e a d e r may safely ignore t h e s e r e f e r e n c e s .
Each chapter includes a brief introductory s u m m a r y . T h e s e chapter
introductions w e r e written in such a way as t o enable the r e a d e r t o f o r m
a c l e a r idea of the contents of each chapter and t o decide what c h a p t e r s
d e s e r v e detailed study and what can be skipped.
* Same data on structurally stable dynamic systems and bifurcations (without exhaustive pro00 will be
found in the second edition of A. Andronov, A. Vitt, and S. Khaikin, T h e o r y of O s c i l l a t i o n r
( h s c o w , 1959).
The book contains numerous drawings and worked-out examples
illustrating the various mathematical propositions. Unfortunately, s p a c e
limitations prevented u s f r o m including many m o r e r e m a r k a b l e examples
which a r i s e f r o m applications (see, e.g., f 2,3/ 1.
The sections, t h e o r e m s , definitions, figures, and examples are
numbered continuously through the book. The numbering of l e m m a s and
equations is r e s t r i c t e d t o each section. In t h e Appendix at the end of the
book, the equations and l e m m a s are numbered according t o the sub-
sect ions.
The r e f e r e n c e S21.2, (5) is t o equation (5) in subsection 2 of SZl. The
r e f e r e n c e ( 7 ) is t o equation (7) of the c u r r e n t section. The r e f e r e n c e QT,
S8.5, L e m m a 4 is to L e m m a 4 in subsection 5 of $ 8 in QT.
A list of bibliographical r e f e r e n c e s directly related t o t h e subject
m a t t e r of the present volume will be found at t h e end of the book.
R e f e r e n c e s t o the s o u r c e s in t h i s bibliography a r e indicated by n u m b e r s
between s l a s h e s .
E. A. Leontovich
I. I. Gordon
Gor'kii, 1 9 6 6
iv
Table of Contents
Preface ........................................ iii

Introduction ..................................... xi
Chapter I. LIULTIPLICITY O F ROOTS O F FUNCTIOKS AKD

LICLTIPLICITY OF INYERSECTIQN POINTS O F 'TWO CURVES. .. 1
Introduction ..................................... 1
$1. Xlultiplicity of a root of a function . . . . . . . . . . . . . . . . . . . 1

1. F -closeness to rank I (1). 2. The rheorem of a small increment of implicit functions (3).
Roof multiplicity of a functior of a single variable (7). 4. Ilultiplicity of a root relative
to a given class of functions (13).
s2. T h e multiplicity of a common point of t w o c u r v e s ......... 14

I . Definirton of multiplicity (14). 2 . Condition of simplicity for a n intersection point of
two curves (IS). 3. Condition of duplicity for a n intersection point of two curves (17).
Chapter 11. DYNXhIIC SYSTEAIS CLOSE TO A GIVEN SYSTEILI

A N D PROPERTIES O F THEIR PHASE PORTRAITS ........... 23
introduction .........................................
9 3. C l o s e n e s s of solutions. Regular t r a n s f o r m a t i o n of close
systems ........................................ 23
1. Theorems of closeness of solutions (23). 2. E -closeness of regions. Lemmas of regular
transformation (27).
54. Intersection of paths of c l o s e s y s t e m s with a r c s and c y c l e s

without contact ................................... 30
1. Intersection with one a r c without contact (30). 2. Paths of close systems between two arcs
birhout contact (33).
Chapter 111. THE S P A C E O F DYKAMIC SYSTEMS AKD STRUC-

TURALLY STABLE SYSTEMS ........................ 50
Introduction ..................................... 50
55. T h e s p a c e of dynamic s y s t e m s . ..................... 51

1. [hespace of dynamic systems i n a plane region (51). 2. The space of dynamic systems on a
iph?rr (59.
$6. Definition of a s t r u c t u r a l l y stable dynamic s y s t e m . .... . . . 55
1. Dynamic systemt on a plane (55). 2. Structurally stable systems on a rplere (58).
3. Structural stabLlky of dynamic systems i n R(:) and R $ ) (59).
7. Structurally stable and s t r u c t u r a l l y unstable paths.

N e c e s s a r y condition of s t r u c t u r a l stability of a n equilibrium .... 62
1. Structurally stable and structurally unstable paths (62). 2. Nnfte number of equilibrium
states in a structurally stable system (63). 3. Multiplicity of an equilibrium state (65).
Chapter IV. EQUILIBRIUM STATES O F STRUCTURALLY STABLE

SYSTEMS. SADDLE-TO-SADDLE SEPARATRM . . .. ... . . ..
Introduction . '. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5 8. S t r u c t u r a l stability of a node and a s i m p l e focus. . . . . . . . .

1. Canonical system (68). 2. Structural stabiliry of a simple node and a focus ( 7 0 .
Si 9. S t r u c t u r a l stability of a saddle point .. .. . . . . . . .... . . 78

1. Reduction of the system t o canonical form by a nearly identical transformation (78).
2. Proof of the structural stability of a saddle point (80).
10. S t r u c t u r a l instability of an equilibrium s t a t e with pure

imaginary c h a r a c t e r i s t i c r o o t s . ...... . .. . ... .. .. .. .... 89
1. Investigation of a n equilibrium state with complex characteristic roots (a review) (89).
2. Calculation of the first focal value (92). 3. The theorem of the creation of a closed path
from a multiple focus (93). 4. Proof of structural instability (95).
s 11. A saddle-to- saddle s e p a r a t r i x . . . . . . . . . . . .. . . . . . . . . 97

1. The behavior of the separatrix under vector field rotation (97). 2. Proof of structural
instability (100).
Chapter V. CLOSED PATHS I N STRUCTURALLY STABLE

SYSTEMS ...................................... 103
Introduction ..................................... 103
$12. A closed path and its neighborhood. Succession function . .. 104

1. Introduction of the succession function (104). 2. The configurarion of paths i n the neighbor-
hood of a closed path (105). 3. ?he case of an analytical dynamic system (108). 4. The case
of a nonanalytical dynamic system (109).
S 13. Curvilinear coordinates in the neighborhood of a closed path.

Succession function on a n o r m a l t o a path . . . ...... . . . . . . . . 110
1. Curvilinear coordinates in the neighborhood of a closed path (110). 2. T r a m f a m a t i o n to the
variables S. II in a dynamic system (113). 3. Succession function on a n m a l to a closed
path (116).
$14. Proof of s t r u c t u r a l stability of a s i m p l e l i m i t cycle . ... . . 118
vi
i 15. S t r u c t u r a l l y unstable c l o s e d paths - ---. .- - - - 124
1. The fundamental l e m m a (124). 2. The thcorem of the creation of d closed path from a
multiple limit c y c l e (127). 3. Structural instability of a closed path with a zero characterlstic
index (133).
C h a p t e r VI. NECESSARY AND SUFFICIENT CONDITIQXS O F

STRUCTURAL STABILITY OF SYSTEbIS . ... . .. ... . .. .. . . 136
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , 136
16. Singular paths ant3 s e m i p a t h s of dynamic s y s t e m s . . . . . . . 137

1. Finite number of closed p a & for structurally stable systems (137). 2. Regions with normal
boundary (134).
s 17. -4 r e g u l a r s y s t e m of neighborhoods and t h e partition of F into

canonical neighborhoods and e l e m e n t a r y quadrangles . .. . . . . . . 142
1. i regular system of canonical neighborhoods for itructurdlly stable systems (142). 2. The
partition of the region 8. into canonical neighborhoods and elementary quadrangles (146).
18. The fundamental t h e o r e m of s t r u c t u r a l stability of a dynamic

system ....................................... 150
I. Lemmd, (150). 2. The fundamental theorem for a plane region (156). 3. The fundamental
theorem for a sphere (162). 4. Remarks and supplements (166).
C h a p t e r VII. C E L L S OF STRUCTURALLY STABLE SYSTEMS.

A N ADDITION T O THE THEORY OF STRUCTURALLY STABLE
SYSTEMS.. ..................................... 174
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
19. C e l l s of s t r u c t u r a l l y s t a b l e dynamic s y s t e m s . . . . . . . . . . 175

1. Generdl considerations pertaining to cells of dynamic systems (175). 2. Doubly connected
cells of suucturally stable systems f 176). 3. Interior cells of structurally stable sysrcms. S m p l y
connected interior cells (179).
20. Examples of s t r u c t u r a l l y stable s y s t e m s . . . . . .. . . . . . . 190
21. A definition of s t r u c t u r a l stability foregoing t h e r e q u i r e m e n t

of E -identity . . . . . . . . . . . . .. . .. . .. .. . . . . . . . . .. .. .. 195
C h a p t e r VIII. BIFURCATIONS O F DYNAMIC SYSTEMS. DECObT-

POSITIOX O F A hTULTIPLE EQUILIBRIUM STATE INTO
STRUCTURALLY STABLE EQUILIBRIUM STATES. .. . ... . . 202
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
. S22.The d e g r e e of s t r u c t u r a l instability and bifurcations of
dynamic s y s t e m s . . ............................... 203
23. Decomposition of a multiple equilibrium s t a t e into s t r u c -

t u r a l l y stable equilibrium s t a t e s . ...................... 218
1. The number of structurally stable equilibrium states obtained from a multiple equilibrium
state (218). 2. The character of the structurally stable equilibrium states obtalned from a
-
multiple equilibrium state with u # o (222). 3. The character of the structurally stable
equilibrium states obtained from a multiple equilibrium state with a 0 (227).
Chapter M. CREATION OF LIMIT CYCLES FROM A MULTIPLE

FOCUS ....................................... 238
Introduction .................................... 2 38
124. F o c a l v a l u e s . . .............................. 239

1. Some properties of the succession function (239). 2. Multiplicity of a multiple focus. Focal
values (242). 3. Calculation of rhe focal values of a multiple focus (244). 4. The case of a n
analytical system (249).
25. Creation of l i m i t cycles f r o m a multiple focus ......... 2 54

1. The fundamental theorem (254). 2. Bifurcations of a dynamic system i n the neighborhood of
a multiple focus (259). 3. Bifurcations in the neighborhood of a multiple focus of
multiplicity 1 (261).
Chapter X.. CREATION OF CLOSED PATHS FROM A MULTIPLE

LIMIT C Y C L E . . ................................. 266
Introduction .................................... 266
26. Expressions for the derivatives of the succession function.

Multiplicity of a limit cycle ......................... 267
1. Expressions for the derivatives of succession functions (267). 2. Multiplicity of a lfmit
cycle (272).
S 27. Creation of l i m i t cycles f r o m a multiple limit cycle ..... 277

1. The fundamental theorem (277). 2. Supplements (282).
Chapter XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF

A SADDLE-POINT SEPARATRM ...................... 286
Introduction .................................... 286
28. Auxiliary m a t e r i a l . ........................... 287

1. Correspondence function and succession function (287). 2. Some properties of a saddle point
and its separatrices (295).
...
Vlll
5 29. Creation of limit c y c l e s f r o m t h e s e p a r a t r i x loop of a
s i m p l e saddle point ................................ 299
1. s o m e properties of the separamx loop (299). 2. Theorems of the creation of a closed path
loop (311). 4. The case P i (x.. u d Q& (a,u.) -

froin a separatrix loop (307). 3. The uniquenes, of the closed path created from a separatrrx
+ 0 1317).
Chapter XII. CREATION OF A LIMIT CYCLE FRON THE LOOP

O F X SADDLE-NODE SEPARATRIX. SYSTERIS O F F I R S T DEGREE
O F STRUCTURAL INSTABILITY AiYD THEIR BIFURCATIONS ... 322
Introduction .................................... 322
s 30. Creation of a limit cycle f r o m the loop of a saddle-node

separatrix ..................................... 323
1. The existence theorem (323). 2. The u n i q u e n r s theorem (326).
s 31. Dynamic s y s t e m s of the f i r s t d e g r e e of s t r u c t u r a l instability

and t h e i r bifurcations. .............................. 330
1. The definition of d system of h e first degree of structural inscability (330). 2. Equilibrium
stdtrs of systems of the first degree of structural instability (331). 3. Closed paths of systems of
rhe firs! degree of structural instability (340). 4. A saddle-point separatrix forming a loop (349).
5. The simplest structurally unst'ible paths (351). 6. The properties of the separatrices of a
snddle-node i n systems of the f i r s r degree of structural instability (364). 7. Properties of
sep.xatrlces of saddle points of syrtenis of the first degree of structural instability (367). 8. The
fundamental theorem (the necessary and sufficient conditions for systems of the first degree of
structural instabiltty) (374). 9. &furcations of systems of the first degree of structural
instability (075).
Chapter XIII. LIMIT CYCLES OF SOME DYNAMIC SYSTEMS

DEPENDING ON A PARAMETER. ...................... 377
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
0 32. The behavior of limit c y c l e s of some dynamic s y s t e m s

following s m a l l changes in the p a r a m e t e r ................. 378
1. The succession function in the neighborhood of a closed path (378). 2. Statement of the
problem (385). 3. Newton's p l y c o n and solution of the equation F (ut. z ) 0 (387). 4. The
behavior of limit cycles of w m e dynamic systems following small changes in the parameter (356).
s 33. Creation of a limit cycle f r o m a closed path of a conservative

system ........................................ 402
1. The integral invariant and comervative system. Statement of the problem. The method of
the small parameter (402). 2. Syxems close to a linear conservative system (409). 3. The
general case of a system close to a conservative system (414). 4. Systems close to a Hamiltonian
system (418).
iY
Chapter XIV. THE APPLICATION O F THE THEORY O F
BIFURCATIONS T O THE INVESTIGATION O F PARTICULAR
DYNAMIC SYSTEMS . . . . . . . .. . . .. .. . . . . . . .. .. . . . . . . 423
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
34. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
Appendix . . . . . . . .. . . . . . . . . . . . . .. . . .. . . . . . . .. .. . . 462
1. Theorems of the continuous dependence of the solutions of a system of differential equations
on the right-hand sides and of the differentiability of solutions (462). 2. A proposition regarding
functions of many variables (468). 3. The l e m m a about the normals of a simple smooth closed
curve (469). 4. Proof of the differentiability of the function R (P. e) with respect to P (471).
5. A remark concerning the definition of a structurally stable dynamic system (477).
Bibliography. .. .... .. . . . .. . .. . . .... . . . ....... .. .. 478

Subject Index . . . . . . . . . . . . . . . . .. . .. . . . .. .. .. . .. . . . 481
X
INTR 0D UC TION
The s u b j e c t - m a t t e r of the qualitative theory of dynamic s y s t e m s is

formulated in QT. The theory is concerned with t h e topological s t r u c t u r e
of the partition into paths of the domain of definition of the dynamic s y s t e m .
12 number of topics relating t o second-order s y s t e m s (defined in a plane
region or on a s p h e r e ) a r e t r e a t e d in QT. In p a r t i c u l a r , the different kinds
of paths of different s y s t e m s a r e identified, the limit s e t s 3f t h e s e paths
a r e d e s c r i b e d , and methods f o r investigating the configuration of paths in
the neighborhood of an equilibrium s t a t e are given. A l a r g e part of QT is
devoted t o establishing t h e minimum information about the paths of a
dynamic s y s t e m needed in o r d e r t o d e t e r m i n e i t s topological s t r u c t u r e in
a region. T h i s problern is completely solved in QT under c e r t a i n limitations
an the c l a s s of dynamic s y s t e m s being considered: it is established that
t h e topological s t r u c t u r e of a dynamic s y s t e m is determined by the
c h a r a c t e r and the configuration of the so-called s i n g u l a r p a t h s
(equilibrium s t a t e s , limit c y c l e s , and s e p a r a t r i c e s ) . ill1 the relevant
information can be presented in the f o r m of a c e r t a i n finite scheme. T h e
basic problem in elucidating t h e topological s t r u c t u r e of a dynamic s y s t e m
i s thus t o find t h e p a r t i c u l a r s c h e m e of the system. So f a r , however, no
r e g u l a r methods have been devised enabling u s t o establish t h e existence
of limit cycles of a dynamic s y s t e m , t h e i r configuration, and the configurk-
tion of the s e p a r a t r i c e s . Only individual p a r t i c u l a r techniques are
available which permit solving - and s o m e t i m e s quite successfully - a
number of p a r t i c u l a r problems related t o the existence and behavior of
limit cycles and s e p a r a t r i c e s . The most useful of t h e s e techniques are
described in QT, together with examples of t h e i r application.
In Q T the approach t o t h e qualitative s t r u c t u r e of dynamic s y s t e m s is
s t a t i c - it i s assumed that the s y s t e m does not change. On t h e other
hand, the main problems treated in the present volume are concerned with
the c h a n g e s i n t h e t o p o l o g i c a l s t r u c t u r e of a d y n a m i c
s y s t e m when the s y s t e m itself changes. A s in QT, w e a r e dealing with
autonomous s y s t e m s on a plane, i.e., s y s t e m s of the f o r m
Let a s y s t e m of t h i s kind be defined in s o m e region G. What happens

t o the topological s t r u c t u r e of the partition of t h i s region into paths when
the s y s t e m - i.e., the functions P and Q on t h e right - changes?
T h i s question is obviously of independent mathematical significance.
It is a l s o highly importznt for various applications. The point is that
dynamic s y s t e m s corresponding - under c e r t a i n idealizations - t o physical
or technical problems invariably contain a certain number of p a r a m e t e r s ,
and we a r e generally interested in t h e changes in t h e topological s t r u c t u r e
of the s y s t e m when t h e p a r a m e t e r s a r e varied. In p a r t i c u l a r , it i s impor-
tant t o find the partition of the p a r a m e t e r s p a c e into regions each c o r r e -
sponding t o identical topological s t r u c t u r e and t o d e t e r m i n e the change in
I
the topological s t r u c t u r e when the s y s t e m moves a c r o s s the boundary of
two such regions in the p a r a m e t e r space.
The changes in topological s t r u c t u r e evidently need be considered only
for the c a s e of s m a l l c h a n g e s in t h e s y s t e m . One of the most
important c l a s s e s of dynamic s y s t e m s c o m p r i s e s t h o s e s y s t e m s whose
topological s t r u c t u r e in a given region d o e s n o t c h a n g e under s m a l l
modifications of the right-hand s i d e s of the s y s t e m . Such s y s t e m s ,
generally' known a s s t r u c t u r a l l y s t a b l e , w e r e f i r s t introduced by
A. Andronov and L. Pontryagin i n 1 4J under the name of c o a r s e
s y s t e m s or s y s t k m e s g r o s s i e r s .
Structural stability of a dynamic s y s t e m is particularly important in
applications, e.g., in various physical problems. The values of the
p a r a m e t e r s entering the right-hand s i d e s of the s y s t e m a r e generally linked
with t h e p a r t i c u l a r physical problem being considered and a r e only known
t o some approximation. If s m a l l changes in t h e s e p a r a m e t e r s - within the
experimental margin of e r r o r - lead t o a change in t h e topological s t r u c t u r e
o f t h e dynamic s y s t e m , Le., i f t h e s y s t e m i s s t r u c t u r a l l y u n s t a b l e ,
the topological s t r u c t u r e of the s y s t e m is c l e a r l y not a suitable c r i t e r i o n f o r
analyzing the physical phenomenon. Conversely, if t h e s y s t e m i s s t r u c t u r a l l y
s t a b l e , i t s s t r u c t u r e may be directly r e l a t e d t o the p r o p e r t i e s of the
physical phenomena. It i s interesting t o r e m a r k in t h i s connection that
Andronov and Pontryagin's t e r m , c o a r s e s y s t e m s , was originally
proposed in contradistinction f r o m f i n e s y s t e m s whose topological
s t r u c t u r e would break under t h e action of a r b i t r a r i l y s m a l l external
.
d i s t u r b an c e s
The f i r s t problem t o be considered i s that of the distinctive o r identifying
c h a r a c t e r i s t i c s of s t r u c t u r a l l y stable s y s t e m s . F o r s y s t e m s defined in a
bounded plane region, t h i s problem was essentially solved in A. Andronov
and L. Pontryagin's original paper 141, and subsequently elaborated in
/ S I and 161.
The n e c e s s a r y and sufficient conditions of s t r u c t u r a l stability for plane
regions a r e relatively simple ( s e e Introduction t o Chapter VI), but a
rigorous derivation of t h e s e conditions n e c e s s i t a t e s detailed scrutiny of
a whole r a n g e of important concepts and scrupulous proofs. The first half
of the present volume (Chapters I through VII) is entirely devoted t o the
theory of s t r u c t u r a l l y s t a b l e s y s t e m s and, in p a r t i c u l a r , t o the derivation
of t h e s e conditions of s t r u c t u r a l stability. The book d e a l s only with
s t r u c t u r a l stability of s y s t e m s in a bounded plane region and on a sphere.
Note, however, that the concept of s t r u c t u r a l stability h a s been extended
and investigated, especially in t h e l a s t decade, for a number of other
objects also. Peixoto considered t h e conditions of s t r u c t u r a l stability of
dynamic s y s t e m s on an a r b i t r a r y closed s u r f a c e ( s e e 1 7 1 ) . Gudkov 181
introduced t h e concept of s t r u c t u r a l stability of algebraic curves.
Structural stability of many-dimensional dynamic s y s t e m s is t r e a t e d in
/32,33,36/.
xii
Structurally stable s y s t e m s a r e the rule, so t o say, in the m e t r i c s p a c e
whose points a r e dynamic s y s t e m s defined in s o m e region. It is shown
in Chapter VI that s t r u c t u r a l l y stable s y s t e m s f o r m a n open everywhere
dense s e t in this space. The s e t of s t r u c t u r a l l y stable s y s t e m s is
partitioned into components, each consisting of s t r u c t u r a l l y stable s y s t e m s
of identical topological s t r u c t u r e . The "partitions" between these
components consist of s t r u c t u r a l l y unstable dynamic s y s t e m s . When a
dynamic s y s t e m is a l t e r e d or modified, i t s topological s t r u c t u r e will
change only i f the s y s t e m p a s s e s through an intermediate s t a g e of
s t r u c t u r a l instability. The t h e o r y o f b i f u r c a t i o n s, which is con-
c e r n e d with changes in the topological s t r u c t u r e of dynamic s y s t e m s ,
t h e r e f o r e appropriately concentrates on s t r u c t u r a l l y unstable s y s t e m s .
St tucturally unstable s;ystems a r e a l s o of i n t e r e s t in applications: the so-
called conservative s y s t e m s (see Chapter XIII), often encountered in
physics, a r e s t r u c t u r a l l y unstable. We a r e thus naturally led t o a detailed
examination of s t r u c t u r a l l y unstable s y s t e m s .
T h e f i r s t s t e p in t h i s direction evidently involves a classification of
s t r u c t u r a l l y unstable s,ystems. Structurally unstable s y s t e m s can be
divided into ''less s t r u c t u r a l l y unstable" and " m o r e s t r u c t u r a l l y unstable."
T h i s l e a d s to a classification according to the d e g r e e s o f s t r u c t u r a 1
i n s t a b i 1 i t y origin;illy introduced in / 9 / . The least structurally
unstable s y s t e m s in t h i s classification a r e the s y s t e m s o f t h e f i r s t
d e g r e e o f s t r u c t u r a l i n s t a b i l i t y : under s m a l l changes, t h e s e
s y s t e m s e i t h e r go to a s t r u c t u r a l l y unstable s y s t e m or retain t h e i r
topological s t r u c t u r e . The complete conditions f o r a s y s t e m t o be of the
f i r s t d e g r e e of s t r u c t u r a l instability w e r e derived f o r plane s y s t e m s (see
/9, IO/; t h e s e conditions are derived in Chapter XII). A dynamic s y s t e m
of the f i r s t d e g r e e of s t r u c t u r a l instability was found t o have one and only
one s t r u c t u r a l l y unstable singular path, i.e., it h a s either a multiple
equilibrium state, or i multiple l i m i t cycle, o r a saddle-to-saddle
separatrix." T o establish the bifurcations of a s y s t e m of the f i r s t d e g r e e
of s t r u c t u r a l instability, it suffices t o consider the changes in i t s topological
s t r u c t u r e in the neighborhood of i t s s t r u c t u r a l l y unstable singular path.
Bifurcations involving a change in the number of l i m i t cycles - i.e.,
bifurcations in which k m i t c y c l e s a r e c r e a t e d or d e s t r o y e d - a r e
of p a r t i c u l a r i n t e r e s t in theory and in applications. Dynamic s y s t e m s of the
f i r s t degree of s t r u c t u r a l instability m a y only exhibit the following
instances of creation of limit cycles: f r o m the multiple focus of the s y s t e m ,
f r o m the multiple cycle_.,f r o m a loop of a saddle-point s e p a r a t r i x , f r o m a
loop of a saddle-node s e p a r a t r i x . The v a r i o u s cases a r e t r e a t e d s e p a r a t e l y
in Chapters IX through XII. Note that the m a t e r i a l presented in t h i s chapter
applies t o bifurcations of s y s t e m s of both f i r s t and higher d e g r e e s of
s t r u c t u r a l inst ability.
T h e concepts of s t r u c t u r a l instability, d e g r e e s of s t r u c t u r a l instability,
and especially examination of the s i m p l e s t bifurcations l e a d s t o a number
of techniques f o r the investigation of p a r t i c u l a r differential equations.
T h e s e techniques w e r e successfully applied t o a number of equations of
s p e c i a l physical i n t e r e s t (see, e.g., /2, 3, 20, 2 5 - 2 8 1 ) .
* Certain additionai conditions !hould also be rausfied; they are formulated in Chapter XII.
xtii
Chapter XIV is e n t i r e l y devoted to examples of dynamic s y s t e m s ,
analyzed by the tools of the t h e o r y of bifurcations.
Chapter XI11 occupies a somewhat s p e c i a l position in t h e book: it
t r e a t s the c r e a t i o n of limit cycles f r o m closed paths of conservative
systems.
xiv
Chapter I
JIULTIPLICITY OF ROOTS OF FUNCTIONS AND
.ZIULTIPLICITY OF INTERSECTION POINTS OF
TWO CUR Y E S
IhTRODUCTION
In this chapter He consider s o m e fairly elementary concepts relating to

the r o o t o f a f u n c t i o n and the i n t e r s e c t i o n p o i n t o f t w o
c u r v e s . In $ 1 we define the 6 - c l o s e n e s s o f t w o f u n c t i o n s t o
r a n k r and the m u 1 t i p 1i c i t y of a root of a function. Roughly speaking,
a root .ro of a function f ( 5 ) is said to be of multiplicity r > 1 if functions 7 ( E )
sufficiently c l o s e to f ( 5 ) cannot have m o r e than r r o o t s in a sufficiently
s m a l l neighborhood of xor but t h e r e is any number of functions sufficiently
c l o s e to f ( 5 ) which have exactly r r o o t s in any a r b i t r a r i l y s m a l l neighbor-
hood of s o . The n e c e s s a r y and sufficient condition of r-multiplicity of a
root i s derived (Theorem 5), which s t a t e s that f (.r,])= f (.ro)= . . . f ( r - * ~ ( r o )= 0,
p) (so) 0. It follows .irom this condition that f o r analytical functions (and
in particular, f o r polynomials), the root multiplicity defined in this chapter
coincides with the usucl concept of multiplicity.
The m u l t i p l i c i t y o f a c o m m o n p o i n t o f t w o c u r v e s ,
analogous to the concept of root multiplicity of a function, is introduced
in S 2 , and the n e c e s s a r y and sufficient conditions a r e established f o r r -
multiplicity of a common point ( x o , yo) of two c u r v e s F i( x , y) = 0 and F? (s.y) =
= 0, when r = 1 and r = 2 . ff r = 1, the common point is s a i d to be s i m p l e
or s t r u c t u r a l l y s t a b l e . The n e c e s s a r y and sufficient condition of s t r u c t u r a l
stability of a point ( x o . yo) is simply
The conditions f o r a common point with r = 2 are m o r e complicated

( s e e Theorem 7).
J 1. h 1 U L T l P L I C I T Y OF A ROOT OF A F U N C T I O N
1. 6 - c l o s e n e s s to r a n k : r
\Ve w i l l consider functions defined a t all points :VI (r,,I$. ..., z,,) of s o m e
open (or c l o s e d ) region G (or 4 ) i n n-dimensional euclidean s p a c e E , . In
I
applications we w i l l be mainly concerned with the c a s e s n = 1 o r n = 2.
Here it is assumed, however, that n is any natural number.
A function is said to be a function of c l a s s k in G (or E,), where k is a
natural number, if it is continuous and continuously differentiable up to
o r d e r k , inclusive, in i t s domain of definition; a function is said to be a
function of the analytical c l a s s in s o m e region if it is analytical in that
region."
.
Let Fo (xi, x 2 , . ., a)be a function of c l a s s k or a n analytical function in
G (or C , ) > 8 is some positive number, r a natural number such that r < k if
Fo is a function of c l a s s k .
D e f i n i t i o n 1 . Afunction F(xlrx2,...,x,,) of class k , > r oranalytical
in G (8,) i s said to be 8 -close to rank r to the function Fo(x,,xZr . ., x,,) in the .
region G(C,) if at any point of the region
where 1
... + % = I .
- . . ., r , all
1,2, ai are non-negative numbers and a,+ a2 +- ...
Clearly i f two functions a r e 8-close to rank r in some region G, they a r e
8 - c l o s e to any r a n k r, < r in that region; moreover, f o r any 81, 8 , , 8 , they
a r e b1-close to rank r in any subregion of G.
If everywhere in the region we only have the one inequality
i . e . , only the functions a s such a r e 8-close, but not t h e i r derivatives, the

functions F and Fo a r e said to be 8-c 1o s e t o r a n k 0. In what follows,
with r a r e exceptions, we will always consider 8-closeness at l e a s t to
r a n k 1. Therefore the expression "two 8-close functions" is to be u n d e r -
stood a s qualifying two functions which a r e 8 -close to r a n k r > 1. The most
interesting c a s e is that of a function depending on one o r s e v e r a l p a r a -
m e t e r s which f o r any ( " a r b i t r a r i l y s m a l l " ) 8 > O can be made 8-close to
any r e q u i r e d r a n k to a given function Fo (5)by a n a p p r o p r i a t e choice of the
parameters.
Let u s consider simple examples of functions which a r e 8-close to r a n k
r to a given function Fo ( x ) when n = 1 and Fo ( 5 ) = 0. To be specific, we
w i l l only consider functions defined on the segment [-1, + 11.
E x a m p l e 1. Let f ( x ) b e a f u n c t i o n o f c l a s s k o n j - 1 , + 1 ] , 6 s o m e
positive number. Then the function pf ( x ) f o r any sufficiently s m a l l (in
magnitude) p is 8-close to 0 to rank k. If f (z)is an analytical function,
then f o r any n a t u r a l number r and sufficiently s m a l l p, the function pf ( x ) is
8 -close to r a n k r to z e r o (Le., to the function F ( x ) 0 ) .
E x a m p 1e 2. Consider the function
F, (2) = p sin 5
I('
( p > 0). For any given 6 > 0, an appropriate choice of a sufficiently small p
* If 5,is closed and Mo is a boundary point, partial derivatives (of any order and type) do not necessarily
exist at Mo. In this case, a partial derivative at M o is defined as the limit value of the corresponding
partial derivative at an inner point M when M tends to d d o (see /ll/, Vol. I, Sec. 258, p. 589). An
analytical function in a closed region is evidently defined in some larger open domain.
will c l e a r l y make this function 8-close to z e r o , but only to rank 0.
Indeed,
I
F ; (x)= -COS
P
<P
and as p d e c r e a s e s , the upper bound of F; (5) i n c r e a s e s to infinity. T h e r e -
fore, if 6 is sufficiently s m a l l , w e c a n never choose a p that w i l l e n s u r e
6 - c l o s e n e s s of F, (I) to z e r o to rank higher than 0. It is a l s o clear that,
by an appropriate choice of a sufficiently s m a l l p, the function
can be made a r b i t r a r i l y c l o s e to z e r o to rank ni, but never to rank g r e a t e r

than in.
\ V e will now give without proof two t h e o r e m s R-hich a r e repeatedly used
in what follows. One is W e i e r s t r a s s s classical theorem on polynomial
approximation to functions ( s e e / 2 3 / , Sec. 1 0 9 , Theorem l), reformulated
in t e r m s of o u r new ccmcept of 6-closeness to r a n k r , and the other readily
follows f r o m the f0rmi.r.
T h e o r e n t 1. Let F ( x l r x2 . . . . . x,)beafunction of class kdefined i n a
closed bounded region C of the space E,. Fot- m y t a and r -i.k there =-
exists a polynotiiinl @ (x,,x2, . . ., z,) which i s e-close to rank r to the function
.
F ( z ~ , . ., G ) in G .
T e o r e wt 2. Let Ma,(J;,xi. . . . . J:) be a point in a closed bounded
region G , and F (z,,x 2 . . . ., xn) = F (,%I)a fiinction of class k defined in that
region. For any e > O cmd r < k there exists a polynomial @ k t , z 2 ,. . .. x n ) =
= (1) (Jf) ichich i s e-close to rank r to the function F (JI) in G such that
io.ime 1 = 1 , 2 , . . ., r , all 3, are non-negatire zohole numbers, and a i f a2-

4 . .A - c & = l .
2. The t h e o r e m of a small increment of implicit functions
\Ve now proceed to ,Jrove a theorem which c a n be called the t h e o r e m

of a s m a l l i n c r e m e n t o f i m p l i c i t f u n c t i o n s . Let F ( z , y , z ) b e
a function of c l a s s k defined in the parallelepiped A
21 ,< x Q z,, y, 5 y .c yz, 21 %

< 2 Q -12 (1)
of the three-dimensional s p a c e E S I such that a t any point of A

F i ( r , y. z)+O.
Let f u r t h e r
z=F(5, y)
be a function defined everywhere in the rectangle R
3
Ch. 1. MULTIPLICITY OF ROOTS AND INTERSECTION POINTS
in the plane (5, y) , such that inside this rectangle
and
F r o m (l), (2), and ( 3 ) i t follows that rp(x, y) is the unique solution of the
equation
F ( x , y, 2) -0 (4)
in A, and the surface
z-'p(+, Y)
h a s no common points either with the top o r the bottom of A . Moreover,
in virtue of the theorem of implicit functions, rp ( 5 , y) is a function of c l a s s k .
T h e o r e m 3. Forany e>Othereexists 6>0suchthatfmany
function p ( x , y, z) defined in A which is 6-close to rank r ( I < r < k ) to
F ( x , y, 2) the equation
F ( x , y, z ) = O
has a solution
defined in the rectangle R , such that (a) @(x, y) is the only solution of
equation (5 1 in A; (b) @(x, y) is a function of class 2 r ana' i s e - close to the
function ' p ( x , y) to rank r .
P r o o f . According to o u r assumptions, inequality ( 2 ) is satisfied a t
any point of the rectangle R:
21 < 'p ( 5 , Y) <z2.
Therefore, if e, >0 is sufficiently s m a l l , we have
zi < q@, Y) * 8, <zz. (6)
W e choose e,>O which satisfies condition (6), such that e i < e . By (1) in A
F ; ( x , y, z ) # O .
To fix ideas, let in A
F;(x, Y, 4>0. (7)
F r o m ( 3 ) , ( 6 ) , and (7) i t follows that in R
F ( G Y, ~ ( x Y. ) - ~ I ) < 0, F ( x , Y, y)+ed>
(~(2. 0. (8)
Now since F and cp a r e continuous, w e conclude f r o m (8) that a t any point in
H
F ( s , Y, ~ ( x y)-el)<-c,
, F ( x , Y, cp(x, y ) + e J > c , (9)
4
ii. -.I( L I P L L C ~ T YOF A ROL)I' OF .A F L K C T W L
w h e r e c is a p o s i t i v e n u m b e r . LIoreover, f r o m i n e q u a l i t y ( 7 ) we c o n c l u d e
that i n 1
F; ( x . y. z ) m, (10)
x.rhere m i s a p o s i t i v e n u m b e r .
L.et b b e s u c h t h a t
h>O, b<$, 8<f. (11)
C o n s i d e r a n y functicm F ( x . y, ; ) d e f i n e d i n A which is 8 - c l o s e to r a n k r to
t h e function F ( s . y. z ) i n A . F r o m t h e d e f i n i t i o n of 8 - c l o s e n e s s a n d f r o m
re1ation.s ( S i , ( l o ) , a n d (11) i t f o l l o w s t h a t i n 1
a n d that
rhen c l e a r l y the equation

-F ( x , y. z)=O (5)
h a s a u n i q u e s o l u t i o n z = i ( x . y) i n A , w h i c h is d e f i n e d i n R a n d which e v e r y -
* - h e r e in R s a t i s f i e s t h c i n e q u a l i t y
IS(., Y)--(P(". Y)I<F!CE. (14)
By t h e t h e o r e m of i m p l i c i t f u n c t i o n s w e now c o n c l u d e t h a t t h e s o l u t i o n
z - = (F I.X. y) of e q u a t i o n ( 5 ) i s a f u n c t i o n of c l a s s > r .
I t r e m a i n s to show that 6 > 0 c a n b e c h o s e n s o t h a t c o n d i t i o n ( b ) of t h e
t h e o r e m is s a t i s f i e d . To t h i s end we n o t e t h a t t h e p a r t i a l d e r i v a t i v e s of
'p (,s. y) are o b t a i n e d s u c c e s s i v e l y f r o m t h e s e t of e q u a t i o n s
F; -7 FIT,) =.:1
F; 1- F;vq = U, (15)
F:x C 2F;,v; + F;, (9;)' C F&iX = 0.
....................
a n d t h e p a r t i a l d e r i v a t i i . e s of &-,y) a r e c a l c u l a t e d f r o m analogous equations:
S i n c e F ; ( x . y , z ) # 0 n o w h e r e i n A, t h e p a r t i a l d e r i v a t i v e s of 'p (s,y) t o o r d e r r
i n c l u s i v e a r e c o n t i n u o u s f u n c t i o n s of t h e a r g u m e n t s F;. F I , FI, F I X , F&. .... F:;',
a n d t h e r a n g e of t h e s e a r g u m e n t s c a n b e r e g a r d e d as a c l o s e d r e g i o n .
Hence i t c l e a r l y f o l l o w s t h a t if 8 > 0 is s u f f i c i e n t l y s m a l l , e.g., 6 < 6!, a n d
t h e function 7 is 6 - c l o s e t o r a n k r t o F , t h e f u n c t i o n (p ( 5 , y) is &-closet o
r a n k r to 'p (I. y). Thus any positive number 6 s m a l l e r than 5- , ,;- a n d 6,
satisfies the theorem. This c o m p l e t e s the proof.
5
I Ch. I. MIJLTIPLICITY OF ROOTS A N D INTERbECTlUN POINTS
The previous t h e o r e m is naturally generalized to any number of functixie

and equations. W e will formulate the corresponding t h e o r e m f o r two
functions and two equations.
Let the functions
Y, u , u) and Fz@, Y, u, 4
of c l a s s k > l be defined in some parallelepiped II
XiGXQxn yiQy<ya, ui,<uQuz. ui,<V,<vz
in E,. Moreover, suppose that

aF
1 ) La V a n d I = - -
D(F,, F2)-
D(u, 4 IF& Fi,, I
F L Fi. do not vanish in ll;
2 ) the equation
F I ( ~Y,, u , v)=O
h a s in IT the solution
u = e (z,Y, u),
defined a t any point x , y, u, xI<x<xz, y,<y<y,, ul<u<u2, such that
u, < 0 ( 5 , y, u) < u2. Since %# 0, this solution is unique;
3 ) the equation
F~ (2,Y, U, e (5,Y, U N =o
h a s a solution u=cp(x, y) in the rectangle R,
XI < 5 << X2, yi 6 I/ Q yz,

such that u1<cp (x. y) <u2.
It follows f r o m condition 1 that the derivative aFZ(3. *

au
*
(I )) does not
vanish in the relevant region, and cp(x, y) is the unique solution of the
equation F2(z,y, u, e@, y, u)) = 0 . Therefore the functions
constitute the unique solution of the s y s t e m
in n, which is defined f o r all ( x , y)E R. By the theorem of implicit functions,

p and @I are both functions of class k .
T h e o r e m 4 . Under the above assumptions, f o r a n y e > O andr,
1 < r < k , there exists 8> 0 such that if the functions
F I ( ~Yt, u, 0) and Fz(z9 91 u, V)
are defined in II and are 6-close to yank r to the functions Fl and F 2 ,

respectively, the equations
-FI(x, Y, ut u)=O,
-
F ~ ( xyt, u, V) = O
6
51. \ICLTIPL[CiTY OF A R O O T OF A FUliCTIOZ
IiaLte a unique solution in n,

-
u='p(x, Y), v=*(2, Y),
which is defined f o r all ( x . y ) R,

~ and the functions 6 and 6 are 6-close to
rank p to the respectirte functions q and tp.
T h e o r e m 4 is proved by applying T h e o r e m 3 f i r s t to the function
Ft (2,y, u. r,) ' and then to the function F z ( s , y , u,8 ( s , y, u ) ) . where 0 (s. y , u) is
the solution of the equation F i(3, y , u. c ) = 0 f o r r .
R e m a r k . Theorem 4 c l e a r l y r e m a i n s valid if the functions F 1and F2
a r e independent of x arid y, i.e., i f w e are concerned with equations of the
f o r m F! (u,v ) = 0 and F, (u.c ) = 0. The wording of the theorem can be
changed without difficulty to conform t o t h e s e new conditions. This a l s o
applies to Theorem 3 .
3. Root multiplicity of a function of a single v a r i a b l e
The concept of r o o t m u 1t i p 1i c i t y is generally applied to the r o o t s

of analytical functions (in particular, polynomials) in connection with the
calculation of derivatives and factorization of polynomials. In t h i s sub-
section w e will advanc- a g e n e r a l definition of root multiplicity for a
function of a single v a r i a b l e i n a f o r m that will readily link up a t a l a t e r
stage with a number of other, m o r e complex concepts in the theory of
dynamic s y s t e m s . For analytical functions, our definition w i l l naturally
coincide with the usual definition of multiplicity.
Let
Y = Fo (4
be a function defined on s o m e segment It!,z21, where it is a function of c l a s s
k > 1 or a n analytical function, and let xo be a root of the equation
Fo (2)= 0,
in lxl, x21. For simplicity let zo = 0 (this can always be accomplished by
changing o v e r t o a new v a r i a b l e ;= z - X O ) and suppose that Fo ( x ) is defined
f o r I I 1 Q U , where a is s o m e positive number. Let r be a natural number
D e f i n i t i o n 2. The root 0 of the equation
Fo (2)= 0
i s called a root of multiplicity r (or an r-tuple root) of this equation and
also a root of multiplicity r of the function Fo ( x ) , Eo ( s ) i s a function of
class k > r and the following conditions are satisfied:
(a) there exist > 0 , 6o > 0 such that any equation F (x) = 0, where
F (2)is afunction of class r which is 6, - close to rank r to the function F , ( x ) , has
at most r roots f o r I x I e o ;: ,
(b) fmany positive E < E , and 6 there exists a function F (x),s-closeto

rank r to the function Fo (c) such that the equation F (I)= 0 has precisely r
r o o t s f mIx I< E .
* Theorem 3 applies to functions of three variables. Nevertheless, it c a n be formulated and proved without
modifications for a function of any number of variables.
7
Ch. I. MULTIPLICITY OF ROOTS AND INTERSECTION POINTS
A root of multiplicity 1 is called a s i m p l e or a s t r u c t u r a l l y

s t a b 1e root of the equation.
A root of multiplicity r a t the s a m e t i m e may be a root of a lower o r a
higher multiplicity.
L e m m a 1 . Let F ( x ) be a function defined f o r 1 x / < a such that
F (2)= z"Q, ( Z ) ,
where n is a natural number, d, ( 2 ) i s a continuous function, and Q, (0)f 0.
Then f o r any e > 0 and 6 > 0 there exist numbers a , , a2,. . ., such that
/aiI<6 ( i = i , 2, ..., n-i),
and the function
F (z)= a l t +azza+ . . . +U,,-~Z%~+z~ (z)
has at least n different roots for I x I < E .
P r o o f . By assumption, 4, (0) # 0 . Let Q, (0) > 0, .and choose E > 0,
6 > 0. Since Q, (5)is continuous, t h e r e e x i s t s q, > 0 , q1 < e such that f o r
all positive z, z < q ~ ,
F (5)= z"CD (5)>0.
We choose one of t h e s e z, s a y zlr and keep i t fixed. Then 0 < z, < q l and
F ( 5 1 ) = 2: Q, (51) > 0. If a , - i is of sufficiently s m a l l magnitude, the function
F1 (I) = + z"Q, ( z ) a t the point x , h a s the s a m e sign a s the function
F ( x ) . The number a,-,t h e r e f o r e can b e chosen so that
I a,,+ 1 < 6, <0 and F1(q)

a,,-$ z?R ( z , )> 0 .
= CZ,,-~XT' +
Now consider the function
+ JCD (z)).
Fl (z)= z"-l (a,,-,
For all sufficiently s m a l l z, the sign of Fl (5)coincides with the sign of

a,-l, i.e., i t is negative. Thus, t h e r e e x i s t s a number qr < z1 such that f o r
all z, 0 < z < qz, Fl (z)< 0. We choose one of t h e s e z, s a y 5 2 , and keep
it fixed. Then
0 <Z Z < qz <xl <ql<e
and
Fl (zl)=U,,-lz~-'+z~CD(z~<o.
A s the next step, consider the function
+
Fz (zz)= c%-,zn-* F1 (z)= zn-' (G-2 + + 20 (z)).
The number is chosen so that
an-, >0, I %-, I <6, F2(zl)>0, <0.
Fz (zz)
8
$1. .:C'LTIPBICITY OF .A ROOI' OF A FLihCflON
For s m a l l positive z, F 2 ( x ) > 0, and w e can choose a number q3<x2and x 3 ,

0 <13<tl3 such that F ~ ( z 3 ) >0. Continuing along the s a m e lines, we obtain
a function
-
F ( z ) = I ' n - l ( x ) = a , x i a 2 z a + .. . + a , . l z n - ~ + z n ~ D ( x )
and a s e t of n u m b e r s xl. x 2 , . .., I,-!. x,, such that
and
Z ( I I ) > @ .F ( X , ) < O . . . .. F(X") I> O f o r odd n ,

( 0 f o r even n.
Sow, each of the i n t e r v a l s ( x , + ~ ,z,)> i = 1, 2,, . ., n -1, contains a t l e a s t

one root E, of the function P (x), and
0 <En-) < En-2 <. . .< E3 < E? < 5 1 < E .

But F ( 0 ) = 0. Thus, for [ I [ <F , F (z) h a s a t l e a s t n different r o o t s . T h i s
completes the proof of the lemma.
R e m a r k 1. Lemma 1 c a n be generalized a s follows. Let
(D (z: al,a2.. . ., a,,-l) be ,a continuous function of all i t s arguments for I z I S a ,
l a , l < ~ ( ~ > O , i = l , ,2. . . , n - l ) , a n d l e t @ ( O ; O , O ,..,, 0 ) Z O . The
proposition of Lemma 1 is then a l s o ti-ue, i . e . , we can choose numbers
a l , a * , . . ., a n - , such tha: their magnitudes a r e a r b i t r a r i l y s m a l l and in any
a r b i t r a r i l y s m a l l neignborhood of the point x = 0 the function
-F @ ) = a t r + a 2 z * + . . . +a,-ld"'l+znO(x; al,. . . , a n - l )
h a s at least n different roots.
An even m o r e g e n e r a l proposition is the following: let c i i be any real
numbers, and the function d, (z: a,.. . ., as before. The above proposition
is then a l s o t r u e for the function
-
F (4= alx +(az+ cZlal)z* + ( a 3+ c32% + c31a1)33 +. ..
... . fc,-,
+ ~ , - ~ , , - ~ a , - ~. + . lal)x"-lfx"@(z; a l , . . ., a,,-l).
Both t h e s e general propositions are proved along the s a m e lines a s the

lemma .
R e m a r k 2 . Sometimes an additional requirement is introduced into
the conditions of Lemma 1, namely that s o m e (fixed) coefficients ai are
z e r o . The function F :z) in this c a s e h a s the f o r m
-F (I) =a& + ak+ + . . . +a k , - l z k s - l + (I),
where 1 < k l < k2 < . . . < k,+i s n - 1, and s < n. In t h i s case w e can choose
(sufficiently s m a l l ) n u m b e r s aki such that in any a r b i t r a r i l y s m a l l neighbor-
hood of the point z = Cl the function F @) h a s a t least s different r o o t s . This
proposition is proved like the l e m m a . A s i m i l a r r e m a r k applies to
functions considered i n Remark 1.
9
Ch.1. MULTIPLICITY OF ROOTS AND INTERSECTION POINTS
T h e o r e m 5 . The number x = 0 is a root of multiplicity r of the func-

tion Fo ( x ) if and only if
FO(O)= 0. F;(0) = 0, ..., e-"

(0) = 0, Fg' (0) # 0 (17)
( Fo (t)is naturally assumed to be a function of c l a s s k > r ) .

P r o o f . Let u s f i r s t establish that conditions ( 1 7 ) a r e sufficient.
Suppose that t h e s e conditions a r e satisfied. Let I F p (0) I = m (m > 0). We
choose e, > 0 sufficiently s m a l l s o that for a l l I z I < E,, we have
IFbP' (x)I >$ .

For 6, we take a positive number s m a l l e r than $ (e.g., 6,= f ). Then f o r
any function F ( x ) defined on the s a m e segment as Fo (x)and 80 -close to r a n k r
to F,(z), we c l e a r l y have f o r I z I < E, the inequality
I F"'(z)I >E
2 - E.=
4 E>
4 0. (18)
Now suppose that s o m e function F ( I ) , which is 6,-close to r a n k r to Po (I),

h a s a t l e a s t r + 1 r o o t s in the i n t e r v a l I z I < E,. Then, by Rolle's theorem,
the function F' (I) has a t l e a s t r r o o t s in this interval, F" (I)h a s a t l e a s t
r - 1 r o o t s , etc., and finally F(r) (z)has no r o o t s altogether, a t v a r i a n c e with
inequality (18). Thus, i f condition (17) is satisfied, any function F (I), which
is 6,-close to rank r to Fo (z), cannot have m o r e than r r o o t s f o r 1 z I < e o ,
i.e., condition ( a ) of Definition 2 holds t r u e .
Let 11snow prove that condition ( b ) of Definition 2 is a l s o satisfied. The
function Fo (5)c a n be written in the f o r m
= zro(I),
F, (I) (19)
where Q, (5)is a continuous function and Q (0) # 0. Indeed, by T a y l o r ' s

theorem and conditions (17), we have in the neighborhood of x = 0
~ ) ( e x ) (0 < e
F o ( 4 = z r F7 < 1).
Setting
Q , ( I ) = ~f o r z.fO and a ( 0 ) = l i m @ ( z ) = TFP)
, (0)
-0
we conclude that r e p r e s e n t a t i o n (19) is applicable a t any point of the segment

I z I < a , (0 (x) is continuous i n t h i s segment, and Q, (0) # 0.
F r o m (19) and L e m m a 1 it follows that for any e > 0 and 8 > 0 t h e r e
e x i s t s a function (I)of the f o r m
-P(z)=aiz+azzl+. . . +a,-lz'-i +Z'CD(Z),
which is 8 -close to r a n k r to F (I) and h a s a t l e a s t r different r o o t s f o r

1 % I < E . This m e a n s that condition (b) of Definition 2 a l s o holds t r u e , and
the sufficiency of (17) h a s been fully established.
10
51. E.IULT[PLICITY OF A ROOT OF A FCNCTION
It now r e m a i n s t o prove that conditions (17) are n e c e s s a r y . Suppose

that t h e s e conditions a r e not satisfied, i.e., e i t h e r
1 ) t h e r e e x i s t s r , , 1 Q r , < r , such that
Fo (0)== Fb (0)= . . . = pi1-')

(0)= 0, (0)# 0,
or
2) F,(O)=Fi(O)= ... =P!'(O)=O. (20)
U'e will show that in ei.ther c a s e z = 0 is not a root of multiplicity r . This

w i l l establish the necessity of conditions (17).
In case 1, the proposition is self-evident: because of sufficiency z = 0
is a root of multiplicity r, -=r and by definition i t cannot be a root of
multiplicity r .
Consider case 2. Let e and 6 be any positive n u m b e r s . By T h e o r e m 2,
t h e r e e x i s t s a polynomial P (z) which is 6 / 2 - c l o s e to r a n k r to the function
Fo (z) and such that
P ( 0 ) = P'(0) = . .. = P(') (0) =o
(if the function Fo (I) is a polynomial, Fo (I) itself can be chosen a s P (2)).
P (3)h a s the f o r m zr+W ( x ) , where cf, (I) is s o m e polynomial. We may take
Q(0) # 0 (otherwise, we may take f o r P (z) the function xr+i (Q (I) + y ) , where
y is a sufficiently s m a l l number, y # 0). By L e m m a 1 t h e r e e x i s t s a
polynomial P ( 5 ) which is 6/2-close t o r a n k r to P ( I ) , and t h e r e f o r e 6 - c l o s e
to r a n k r to F , (z)~ and which h a s a t l e a s t r + 1 r o o t s for I z I < e . Hence,'
2' = 0 is not a root of multiplicty r f o r Fo (s). T h i s completes the proof.
It follows f r o m Theorem 5 that f o r analytical functions, and in p a r t i c u l a r
for polynomials, root multiplicity in the s e n s e of Definition 2 coincides with
the n o r m a l concept of multiplicity.
R e m a r k . Let z = 0 be a s i m p l e root of the function Fo (z), i.e., r = 1,
Fo (0)= 0 , Fi (0)# 0. Then t h e r e exist e, > 0 and 6o > 0 such that any function
F (.r) which is 6,-close to FO (I) h a s precisely one root f o r I z I < eo, which is
m o r e o v e r a s i m p l e root. F u r t h e r m o r e , f o r any e < e,, e > 0, we s a n find
6 > 0 such that any function F (z), 6-close to Fo (I), h a s p r e c i s e l y o n e ( s i m p l e )
root f o r \ z I < E .
Indeed, f o r eo we may take any number such that the derivative F; (3)does
not vanish on the segment I x I < e O . Then on t h i s segment I F; (z) I ,m > 0
and for 6, we may take, say, any positive number which s a t i s f i e s the
inequality
F o r E < e,, 6 is found IJY the s a m e method.

We have shown that if F o (2) is of class r a n d condition (20) is satisfied,
t h e r e e x i s t s a function F (z) a r b i t r a r i l y c l o s e to rank r to Fo (I) which h a s a t
l e a s t r + 1 r o o t s in any s m a l l neighborhood of the point 0. T h i s proposition
c a n be strengthened by showing that under the s a m e conditions t h e r e e x i s t s
a function F (z) a r b i t r a r i l y c l o s e to rank r to Fo (2)which h a s a t l e a s t 1 r o o t s
ch.I. MULTIPLICITY OF ROOTS AND INTERSECTION POINTS
in any s m a l l neighborhood of 0, where 1 is a n a t u r a l number. This proposi-

tion is formulated and proved in the following lemma.
L e m m a 2 . If
F o ( 0 ) = F i ( O ) = . . . =F:'(O)=O, (20)
then f o r any positive e and 6 and any natural number 1 , there exists a
function F ( 2 ) which i s 6 -close to rank r to the function Fo ( 5 ) and has at
least 1 roots f o r I x I < e (we r e c a l l that the e n t i r e t r e a t m e n t i s confined t o
s o m e segment I I I < u).
P r o o f . The function F ( 2 ) whose existence is to b e proved can be
constructed by s e v e r a l different methods. We w i l l d e s c r i b e h e r e one of
the possible constructions. Let e > 0, 6 > 0, and 1 b e given.
1) We choose el < e so s m a l l that the function Fo ( 5 ) is 8 / 2 - c l o s e to
r a n k r t o z e r o (i.e., to the z e r o function) f o r 0 <=<et, and the polynomial
Fo(el)+~(z-el, . . .f
)+, ~ 6 '(et)
) (5-
is 6 -close t o z e r o t o r a n k r f o r 0 < x g u . For el we may naturally take any

sufficiently s m a l l positive number; this follows f r o m conditions (20) and
f r o m the continuity of the function Fo ( 5 ) and i t s d e r i v a t i v e s .
2 ) We choose a l a r g e n a t u r a l number N , s o that each of the n u m b e r s
N , N , ..., 4 is
1 l e s s than e,.
3 ) We choose p > 0 s m a l l enough f o r the function
'p (2)= p ~'+ ~

(I- sin nNx .
to be 6 / 2 - c l o s e to z e r o to r a n k r f o r 0 <%<e,. Note that the n u m b e r s
i 2
N , T ,
- .. .. 1
a r e r o o t s of the function ' p ( x ) which lie i n the i n t e r v a l 0 < z < c i ,
and that the functions ' p ( x ) , $(I), .. ., qdr)(2)vanish a t the points x = 0 and
x i
We now define the function F ( z ) as follows:
F ( x ) = F o ( z ) for - - a < x < O ;

F(x)='p(x) for 0 <%<el;
F ( X ) = F ~ ( ~ ) - - [ F ~ ( L ~ ) + ~ (.~f~
-F .
- ~( "~( e) ~+) .( x - e l pf] o r e l < x < u .
The function F ( I ) constructed in this way is r e a d i l y s e e n to be a function
of c l a s s r which is & - c l o s e to r a n k r to the function Fo ( x ) and h a s a t least
I roots in the i n t e r v a l ( 0 , e). This completes the proof of the l e m m a .
The function F ( I ) constructed in the above proof is of c l a s s r, but in
g e n e r a l i t s c l a s s is not higher than r. It is r e a d i l y seen, however, that
t h e r e e x i s t s a P O l y n o m i a l P ( 2 ) which satisfies all t h e conditions of the
l e m m a (i.e., a polynomial which is 6 -close to r a n k r to Fo (3) and h a s a t
l e a s t I r o o t s in the e - neighborhood of 0). Indeed, i t follows f r o m
Theorem 5 that the r o o t s k, x , .. .,-$ are s i m p l e ( s t r u c t u r a l l y s t a b l e ) r o o t s
2
of the function F (I). But then, in v i r t u e of the r e m a r k to T h e o r e m 5, any

polynomial P ( I ) which is sufficiently c l o s e to F ( x ) h a s a t least 1 r o o t s f o r
12 I < e, and thus s a t i s f i e s the r e q u i r e m e n t s of the lemma.
12
$1. \!L'LTLPLLCITY OF A ROOT OF A FCKCTION
Note that Lemma 2 is a l s o valid f o r r = 0. This means, in particular,

that if the function F,, (.r) of c l a s s k > 1 h a s a simple root, x = 0 say, then
for any e > 0 and 6 > 0 t h e r e e x i s t s a function F (I) (possibly a polynomial
P ( 3 ) ) u hich is 6-close to rank 0 to Fo (r)and whose r o o t s in the c-neighbor-
hood of the point 0 a r e m o r e numerous than any chosen number. However,
i f c l o s e n e s s to rank 1 is required, F ( E ) will have only a single root in the
neighborhood of 0 f o r s m a l l 6 ( s e e r e m a r k to T h e o r e m 5). This shows
that the requirement of c l o s e n e s s to a c e r t a i n rank is highly s i g n s i c a n t in
the definition of multiple or s i m p l e r o o t s .
D e f i n i t i o n 3. A root x = O of the function F , ( x ) i s said to be of
infinite multiplicity or not of finite multiplicity if either
1) F , (x) i s a functicm oj. class r but not of class r + 1 (r>O) and
Fo(o)=F;(0)=. ..= F b " ( O ) ,
OY
2) Fo (x) is differeniiable to any order over the releL-ant segment, and
all deritvatires ranish .at the point 0.
A root s = 0 of the-function Fo (I) i s said to have a multiplicity higher
than r (or to be a root of multiplicity > r , where r i s a natural number) if
it has a finite nru1tiplii:ity r' > r o r i f it has infinite multiplicity.
F r o m Definitions 2 and 3 and Theorem 5 i t follows that the root of each
continuous function h a s a definite, finite or infinite, multiplicity. Each
root of a n analytical function which is not identically z e r o h a s a finite
multiplicity.
4. Multiplicity of a root r e l a t i v e to a given class

of functions
In o u r definition of a root of multiplicity r w e a s s u m e d that a l l the

functions ( F , (x). F ( s ) , e t c . ) w e r e functions of class r , without imposing
any f u r t h e r r e s t r i c t i o n s . For c e r t a i n purposes, however, i t is worthwhile
considering n a r r o w e r = l a s s e s of functions.
Let m q r ' b ethe s e t of all functions of class r (where r is a non-negative
i n t e g e r ) defined on the segment I x I <a. W e will particularly concentrate
on the following groups:
1 ) the s e t 211tk' of all functions of class k ;
2 ) the s e t YllA of all analytical functions;
3 ) the s e t %of all polynomials;
4 ) the s e t W, of all polynomials of d e g r e e Q n .
Clearly, f o r any n and k,
W, c mp c W A c
hroreover, if R, e n2 and ki < k 2 , we have

W,,c Wn2 and Fm'") c r"**'.
W e will now define r o o t m u l t i p l i c i t y r e l a t i v e t o a g i v e n c l a s s
m.
13
D e f i n i t i o n 4. A root z = O of a function F , ( x ) o f class Fm is said to

be a root of multiplicity r relative to the class w if (m c 1137~') and the
following conditions are satisfied:
(a) there exist eo > 0 and 6o > 0 such that any function F ( x ) of the class W
which is &-close to rank r to Fo ( x ) has at most r roots in the interval
I5 I < eo;
(b) f o r any positive F < e , and 6 there exists a function F ( x ) of the class
which is 6-close to rank r to F , ( x ) and has exactly r roots in the interval
IZl<&.
If 1137 is one of the four classes above (Ijn'*), Sn,, IIJlp, and !Illx),
conditions
( 1 7 ) of Theorem 4,
Fo (0) = F; (0)= . . . =FV-" (0) = 0, F t ' ( 0 ) # 0,
a r e both n e c e s s a r y and sufficient for the root z = 0 of the function F o (z) to

b e a r o o t o f m u l t i p l i c i t y r r e l a t i v e t o t h e c l a s s 1137.
Indeed, i t is readily s e e n that the proof of sufficiency in t h i s c a s e
(i.e., f o r multiplicity relative to the class a) follows l e t t e r by l e t t e r the
proof of Theorem 5.
The necessity of conditions ( 1 7 ) is a l s o proved a s in Theorem 5, except
for the case IIJl = (m, and r = n. In t h i s case, the equalities
Fo(O)=F;(O)= ... = F ( ' ) - O

0 -
show that F o (I) 5 0, which c l e a r l y contradicts condition (a) of Definition 4 .

It follows f r o m the above that if F , ( E ) is a function of c l a s s 912, where W
is one of the four classes J'31,. % I p , %In, and x = 0 is a root of
multiplicity r of t h i s function in the s e n s e of Definition 2 (i.e.> relative to
the c l a s s W'"),then z = 0 is a root of F o (z) of multiplicity ' r in the s e n s e of
Definition 4 (i.e., relative to the c l a s s (m), and vice v e r s a . Therefore, we
do not have to consider multiplicity relative to any of the above particular
c l a s s e s of functions, and in all that follows root multiplicities are under-
stood in the s e n s e of Definition 2 .
5 2 . THE lLlULTIPLICITY OF A COMMON POINT OF

TWO CURVES
1. Definition of multiplicity
Let F , ( x , y) and F , ( x , y) be functions of class k, 1 defined in some closed

region of the plane (z,y).
Consider the s e t of equations
F , ( x . y) = 0, Fz ( x , y) = 0. (11
Let (C,)and (C,)be the c u r v e s described by these two equations,

respectively. Let equations (1) have a simultaneous solution zO,uo in z,
i.e.,
Fi ( 2 0 , yo) = 0, Fz (Zo, Yo) = 0.
14
$2. hIL'LTPLLCITY OF A CO!.l~.lOS POIKT OF T W O CCRVES
The point M o ( q , , yo), which is a c o m m o n p o i n t o r a p o i n t o f

i n t e r s e c t i o n of the c u r v e s ( C , ) and ( C z ) , is invariably a s s u m e d t o lie
inside the region b. R[oreover, we m a y take rO=yo= 0. T h i s a s s u m p t i o n
c l e a r l y d o e s not d e t r a c t f r o m the generality of o u r a r g u m e n t .
Let r b e a n a t u r a l niimher not g r e a t e r than the c l a s s of F , and F P ,
i.e., 1 g r d k (if Fi a n d Fz a r e analytical functions, r i s any n a t u r a l n u m b e r ) .
De f.in i t i o n 5. A conztnon point M o ( 0 , 0 ) of the ciirres ( C , ) and (C,) is
Saul to be a cotnmon point of ?nultiplicity r of these curz'es a~ a solution of
t)iiiltiplicity r of equations (1 i i f the folloichzg conditions are satisfied:
:.
(a) There exist eo 0 , 6,, > 0 such that any two curres B,(I, y ) = U and
F?(x,y) = 0 , ic!iere P, and F2 are functions of class r , bo- close to rank r to
FI arid F 2 , respectirely, hare at tnost r co))it)zotipoints in U,, (Mo).
(b) Fov any 6 and E < ( 6 > 0 , E > 0) there exist fiinctim2s @, (I. y ) and
cD2 ( x . y), 6-close to rank r to F , (x.y) aizd F? ( E . Y), such that the ciirres
@I ( x , Y ) = 0 a d Q2 ( x , 51) = 0 hare exactly r co))it)zonpoints in ue(&Io).
An i n t e r s e c t i o n point of multiplicity 1 is called a s i m p 1e or
s t r u c t u r a 1 l y s t a b 1 e point of i n t e r s e c t i o n of c u r v e s (C,) and (C?).
A common point MOof two c u r v e s (1) is s a i d t o be m u 1t i p 1e or
s t r u c t u r a 1 l y u n s t a b 1e if i t is not a s i m p l e i n t e r s e c t i o n point. Note
that a common point of two c u r v e s is not n e c e s s a r i l y of finite multiplicity.
Consider the following example.
Let FO (z) be a function of class r which is not a function of class r + 1 and
let F,(O)=F;(O)= ... =C'(O).
d i s chosen a s s o m e bounded closed region with O ( 0 , 0 ) a s i t s i n t e r i o r
point.
Consider t\vo curve:;:
y=O, ~-FO(Z)=O. (2 1
O.(O, 0 ) is a coninion pcmint of t h e s e c u r v e s . L e t E and 6 b e any two positive

n u m b e r s , 1 s o m e n a t u r a l n u m b e r , and F (z) a function c o n s t r u c t e d a s i n
Leninia 2 (F1. 3 ) , i.e., a function which is 6-close t o r a n k r to Fo ( x ) and h a s
a t l e a s t 1 r o o t s in t h e i n t e r v a l I x I < E . The c u r v e s
y =o, y-F(z) =0 (3)
a r e then 6 -close t o r a n k r to the r e s p e c t i v e c u r v e s in ( 2 ) and have a t l e a s t

1 conimon points in Lf, (0).The i n t e r s e c t i o n point O ( 0 , 0 ) of c u r v e s ( 2 ) thus
d o e s not ha\-e a finite riuitiplicity.
If a conimon point 0''two c u r v e s d o e s not have a finite multiplicity, it
i s s a i d t o b e a n i n t e r s e c t i o n point of infinite multiplicity. Finally, a
coninion point O ( 0 , 0 ) of two c u r v e s is s a i d t o b e of multiplicity higher than
r e i t h e r i f it h a s finite multiplicity r' > r o r if i t h a s infinite multiplicity
( s e e Definition 3, 5 1).
2. Condition of s i m p l i c i t y f o r a n i n t e r s e c t i o n point
of two c u r v e s
LVe Nil1 now c o n s i d e r the n e c e s s a r y and sufficient condition for a n

i n t e r s e c t i o n point of two c u r v e s t o b e a s i m p l e ( s t r u c t u r a l l y s t a b l e ) i n t e r -
section point, i.e., a coninion point of multiplicity 1.
15
T h e o r e m 6 . The intersection point 0 (0, 0 ) of two curves

Fl (J,Y) = 0 (Cl)
FZ (5,Y) = 0 (CZ)
i s of multiplicity 1, i.e., simple (structurally stable), if and only if

. = I Fi,(O, 0)
Fi, (0, 0)
Fi,(O, 0)
F;,(0, 0) l+o. (4)
P r o o f . We will f i r s t show that A. # 0 constitutes a n e c e s s a r y condition.

Suppose A,,= 0. Let E and 6 be s o m e positive numbers. Let further PI (x,y)
and Pz (5,y) be polynomials which satisfy the following conditions:
1) P I @ ,y) and Pz(z,y) a r e 6/2-close to rank 1 to the functions F , ( z , y) and
F , ( J , y), respectively;
2) P i ( 0 , o ) = 0 , Pi,(O, o ) = G ( 0 , o), Pi,(O, o ) = FI(0, 0 ) ( i = 1,2).
These polynomials exist by Theorem 2, S1. Two cases are possible:
( a ) a t l e a s t one of the n u m b e r s P L ( 0 , 0), P ; , ( O , 0 ) , P L ( 0 , O ) , P i u ( O , 0 )
does not vanish; ( b ) all these numbers a r e z e r o .
F i r s t let u s consider case (a). The polynomials P , and P2 in this case
have the f o r m
+
Pi (z,Y) = A i s + B , y C l i . . . .+
Pz (5,Y) = A z J + Bzy+ C ~ Z . . . + .
where A -1
O-
4 BBiz
A,
I = 0, but a t l e a s t one of the coefficients A I , B,, A,, B, does
not vanish. For example, let B , # 0. Then either B z # 0 o r A 2 = B 2 = 0.

Consider the polynomials
Fl (x,y) = AIJ +B I Y+ aif + C1s2+ . . . = PI (5,v) +alz,

FZ(x, Y) = 4 s + BZY+ a@+ CZJ + . . . =PZ,I( v) +a z x
and the point M , (5,.-&z,), where s,# 0. We choose the coefficients a, and
B,
a2 so that the two c u r v e s
p,(x,y) = 0 and & (J,y) = 0

p a s s through the point MI. Clearly a, and a2 should satisfy the equations
a,q + c,z; + ...= 0,

azq + czz;+ ...= 0.
Dividing through by J, (xi # 0), we get
a, = -C,q - . . ., a2= -C ,2,-. ..

Thus a, and a , a r e uniquely determinable and can be a s s m a l l as d e s i r e d
f o r sufficiently s m a l l 2,. We choose 5, s m a l l enough f o r the point MI to lie
inside U , (0)and f o r the polynomials PI (J,I/)and PZ (J,g) t o be ?6- c l o s e to
PI (2,y) and Pz (5,y). respectively (and hence, 6 -close to the original functions
16
F , (s,y) and F 2 ( x , y), re:.pectively). Then the c u r v e s pt ( r ,y) = 0 and PZ( x , y) =
= 0 have a t l e a s t two common points in U e (0), namely the point 0 and the
point MI. This c l e a r l y signifies that the intersection point 0 of the c u r v e s
F I (2. y) = 0 and F Z ( x , y)= 0 in case ( a ) is not a s i m p l e ( s t r u c t u r a l l y
s t a b l e ) intersection point of t h e s e c u r v e s .
In case (b), when
Pi, (0, 0) = P i v (0, 0 ) =Pi, (0, 0 ) = P;v (0, 0 ) = 0,
w e r e a s o n along the s a m e lines a s in case (a), but the previous point is

now chosen a s MI ( x , , 0). The number 3 , is chosen so that the polynomials
6
are -close to the polynomials P I and Pf, respectively, and the c u r v e s
pi (5. Y) = 0 and PZ( x , y)= 0 p a s s through the point kf. Thus again 0 ( 0 , O ) is
not a simple point ot c u r v e s (C,) and (C2). We have proved that the condition
S = 0 is n e c e s s a r y . Tne sufficiency of this condition follows directly f r o m
Theorem 4, if we a s s u m e that the functions F , , F 2 and F,, FZof that t h e o r e m
are dependent on x and y only ( s e e r e m a r k to T h e o r e m 4). This completes
the proof.
R e m a r k 1. If the original functions F, (z,y) and Fz ( x , y) are polynomials,
w e may take them as the polynomials Pi and P?. Then the polynomials
P , ( x , y) and P, ( r ,y) constructed in o u r proof of the necessity of condition ( 4 )
are of the s a m e d e g r e e as the polynomials F, and F Z respectively.
R e m a r k 2. It follows f r o m condition ( 4 ) that if O ( 0 , 0 ) is a simple
intersection point of c u r v e s ( C , ) and (C2),the angle between these c u r v e s
a t the point 0 is o t h e r than z e r o , i.e., the c u r v e s a r e not tangent to each
other a t this point.
R e m a r k 3 . A s i m p l e (structurally s t a b l e ) intersection point 0 ( 0 , 0 )
of the c u r v e s ( C , )and (C:!)h a s the following property: t h e r e exist E O > 0 and
60 > 0 such that if the functions cD1 ( x , y) and ~ D (z r ,y) a r e &,-close t o the
functions F , ( r , y) and F2 (s,y). respectively, the c u r v e s
CD, ( 5 , y) = u and (D2 (z,y) =0
have exactly one intersection point in U, (0) and this intersection point is
a l s o simple. hloreovei-, r e g a r d l e s s of how s m a l l E < E~ is, So can be made
sufficiently s m a l l s o that t h i s intersection point f a l l s inside C', (0).
The validity of t h i s r e m a r k follows directly f r o m T h e o r e m 4 ($1. 2 ) and
Theorem 6.
3. Condition of duplicity for a n intersection point

of t w o c u r v e s
We have established the n e c e s s a r y and sufficient condition f o r a root of

a function to be of multiplicity r (Theorem 5). The derivation of the
n e c e s s a r y and sufficier-t condition f o r a n r-tuple intersection point of two
c u r v e s in the g e n e r a l c a s e (any natural r ) is far f r o m being so elementary,
17
Ch. I. MULTIPLICITY OF ROOTS AND INTERSECTION POINTS
and it is t h e r e f o r e not considered h e r e . For r = 1 the corresponding

condition is given by Theorem 6. We w i l l now derive the condition for a
d o u b l e i n t e r s e c t i o n p o i n t o f t w o c u r v e s , or a n i n t e r s e c t i o n
p o i n t o f m u l t i p l i c i t y 2 , which is often used in what follows.
As before, w e consider two c u r v e s
Fl(2, Y) = 0, Fz ( x , Y) = 0
with a common point O ( 0 , 0). F, and F 2 a r e now a s s u m e d to be functions of
c l a s s 2 in C.
T h e o r e m 7. A common point 0 (0,0 ) of the two curves F1 (5,y) .= 0
and F 2 ( x , y) = 0 i s a double intersection point if and only if the followtng
conditions are satisfied:
(a)
I
F i x (0,0) Pi, (0,0)
"= Fix(0,0) F;, (0,0)
1 = o;
(b) at least one of the elements in the determinant A. i s other than zero;
(c) the number x = 0 i s a double root of the function Fz (2,cp ( x ) ) , where
y = cp ( x ) i s the solution of the equation F , (x,y) = 0 fmy in some sufficiently
small rectangle I x I g a , I y (this solution exists and is unique in virtue
of condition (b) and the themem of implicit functions; also cp (0) = 0).
If F ; , (0, 0 ) = 0, but s o m e other element of A,, does not vanish, condition
( c ) should be appropriately reworded.
P r o o f . 1) Necessity. Condition (a), A. = 0, is c l e a r l y n e c e s s a r y by
Theorem 6. We will now show that condition ( b ) is n e c e s s a r y . Suppose
that t h i s condition is not satisfied, i.e.,
F i x (0, 0) =Fi, (0,0)= F i x (0, 0) = F i , (0,0)= 0.
Let e and 6 be s o m e positive numbers. By Theorem 2, t h e r e e x i s t

6
polynomials Pi(z, y) and Pz(2,y), %-close to rank 2 to the functions F , and F a ,
respectively, such that
Pi (0, 0) = P i x (0, 0) =Piv (0, 0) =0 ( i = 1. 2)
(if F , and F , a r e polynomials, P, and PZare identified with F1 and F z ) . The

polynomials Piand P, a r e written in the f o r m
Pi(z, y) = 4.9+2Blry + Cly* + . ...

Pz ( x , Y) = A d +2B,q/+ Cay' + ...,
where the omitted t e r m s are a l l of higher than second o r d e r .
The r e s t of the proof proceeds along the s a m e l i n e s a s in Theorem 6.
Consider the two polynomials
Pl(G Y) ==%E +Blll +P1(.. Y).

(.z, Y) =aaz + +Pz (r. Y).
8261
5438 18
cr'. ;,:rLr:PL[c[rY C ~ .FI COLI:.:OK T TW
P C ~ OF CORVES
Lye choose aI.PI, u2, 8 2 s o that the c u r v e s PI (2. y) = 0 and F2 (I, y ) = 0 p a s s

through two fixed points Aft (11, 0) and NI ( 0 , YI) (21# 0,yl f 0). To this end,
the following equalities should b e satisfied:
and
Uividing through by 11 and yt, respectively, we e x p r e s s a; and as

polynomials in zI and y, without a free t e r m . The n u m b e r s a,and P i t h e r e -
fore go to z e r o for 3, -+ 0 and y, 0, and we c a n choose x1 and yI so s m a l l
&
0
that the polynomials and Fz are - c l o s e t o r a n k 2 to the polynomials Pl
and P, and t h e r e f o r e 6 - c l o s e t o r a n k 2 to the functions FI (I, y) and F Z (I. y).
I f , m o r e o v e r , I z1 I < E , [ yl I < E , the c u r v e s Bl (I. y) = 0 and P, (I, y) = 0 have
at l e a s t 3 common points i n c', ( 0 ) : t h e s e a r e 0 (0, 0), MI (I,, 0), and N1 ( ( 1 . y,).
This c l e a r l y p r o v e s that the common point 0 af c u r v e s (C,) and (C,) i s of
multiplicity higher than 2. &.e have thus established that condition ( b ) is
also necessary.
Let u s now proceed with condition (c). Let y = q (I) b e the unique solution
of the equation
Fl(3, Y) = 0
in s o m e sufficiently s m a l l r e c t a n g l e / s l < a , \ylL<fi, such that I(p(z)J<p. Let
e (2)= F? (". 'F ( I ) ) . (5)
Since ~ ( 0 =) 0, w e have 0 ( O ) = 0. N o w , u s i n g t h e equality
w e readily see that
i . e . , e ' ( O ) = 0 in v i r t u e o f condition (a).

Suppose that condition ( c ) is not satisfied, i.e., z = 0 i s not a double root
of the function 0 (2). Since 6 ( O ) = e ' ( O ) = 0, t h i s a s s u m p t i o n is equivalent to
the equality 0" ( 0 ) = 0 . Note that the d e r i v a t i v e s of the function rp (I) at the
point 2 = 0 are e x p r e s s e d in t e r m s of the p a r t i a l d e r i v a t i v e s (up to
corresponding o r d e r s , Lnclusive) of the function F 1 ( I , y) a t the point 0 (0, 0 )
and the d e r i v a t i v e s of the function 0 (2) a t the point 0 are e x p r e s s e d in t e r m s
of the p a r t i a l d e r i v a t i v e s of F1 a n d F? a t O ( 0 , 0).
-3
Thus, let 6" ( 0 , 0 ) = 0 . \t.e choose s o m e E > 0 and 6 > 0. L e t T b e a
positive number smalle.: than 61 on which additional r e s t r i c t i o n s w i l l now
b e iniposed. Consider the polynomials P I (2, y) and P P (2. y),b,-close t o
r a n k 2 to the functions Fl (2, y) and F? (2. y ) , r e s p e c t i v e l y , s u c h that t h e i r
values and the v a l u e s of t h e i r d e r i v a t i v e s to second o r d e r inclusive at the
19
(3.1. MULTIPLICITY OF ROOTS AND INTERSECTION POINTS
point O ( 0 , 0 ) coincide with the r e s p e c t i v e values of the functions F f ( x 9 g) and

F z (2,y) and their d e r i v a t i v e s a t that point ( s e e T h e o r e m 2). By T h e o r e m 3
(the t h e o r e m of a s m a l l i n c r e m e n t of implicit functions), if &i1 is sufficiently
s m a l l , t h e r e e x i s t s a function y = $ ( x ) which is the solution of the equation
Pl (2,Y) =o
f o r y in the rectangle 1 % I d a , I y We choose 81 so s m a l l that the
function 9 (2)b e c o m e s sufficiently c l o s e to r a n k 2 to 'p (z) f o r I z I g a , and
the analog of 0 (x), the function
Y (4= P2 ( 5 , tp (4)*
b e c o m e s sufficiently c l o s e to r a n k 2 to the function
e (2)=F~ ( x , MI.
Clearly I# (0) = y (0) = 0. The d e r i v a t i v e s $' (O), I#" (0), y' (0) and y" (0) are
e x p r e s s e d in t e r m s of the p a r t i a l d e r i v a t i v e s of the polynomials Pi and P E
(to second o r d e r inclusive) a t the point O ( 0 , 0). Seeing that the v a l u e s of
t h e s e d e r i v a t i v e s coincide with the corresponding values of the d e r i v a t i v e s
of F , and FZ, we r e a d i l y conclude that
9'(0) = 9' (0) 7 I#" (0) = cp" (O),

y' (0) = e' (0) = 0, y (0)= e" (0) = 0.
y ( x ) is c l e a r l y an analytical function. If y ( x ) E 0, all the points (%,I# (I))f o r
I x \ < a a r e common points of the c u r v e s Pi ( x , y) = 0,PI (z, y) = 0, i.e., in
any neighborhood of 0 (0,O) t h e s e c u r v e s have a n infinite number of common
points.
Now suppose that y (5) is not identically z e r o . Then i t can be written in
the f o r m (since y (0) = y' (0) = y e (0) = 0)
y (z)= 2?a) (4.
where k, 3, and Q, (0) # 0 ( s e e S1, (19)).

Let
7 (z)= B2 ( 5 , $ ( x ) ) =a1r+ a$' +d ~(3.

,
F ( x ) is a l s o an analytical function. By L e m m a 1, S1, and R e m a r k 2 to the
l e m m a w e know that for any c, > 0 and 8 , > 0 we can choose two n u m b e r s
al# 0 and a2 # 0 such that the polynomial pz(3, y) is 6-close to r a n k 2 to the
polynomial P a (z,y) and the equation
Y"(x)=O,
a p a r t f r o m the r o o t x = 0, h a s a t least two m o r e r o o t s z1 and x 2 , 1 xt I < ti,

I z2 I < 8,. This choice of a,,a2e n s u r e s that the polynomials B, and pzare
20
$2. h'L1LTIPLIC17-t' OF A CO?.JhJOh POIhT OF T W O CCRVES
8-close to rank 2 to the functions F1 and F Z , respectively, and the

curves
have a t l e a s t t h r e e common points,
such that if E , > 0 is sufficiently s m a l l , -MI and -VI2 lie in U e(0). T h i s means,
however, that 0 (0, 0) is not a double intersection point of the c u r v e s (C,)and
(C?). The necessity of condition (c) is thus established.
Note that if F , and F? are polynomials, they can be used a s PI and P?.
The polynomial P2 in t h i s case is of the s a m e d e g r e e a s the polynomial F : .
2 ) Sufficiency. Let conditions (a), (b), and ( c ) of T h e o r e m 7 be
satisfied. Since Fi,(O. O ) # o , by Theorem 3 f o r any 6,>0 we c a n find b o > O
such that i f the functions Fl(r,y) and P,(s,y) are 6,-close t o r a n k 2 to the
function F1 and Fz in c, the following holds true;
1) the equation Fl (3, y) = O h a s a unique solution y=<(z) in the rectangle
Izl,<a, Iyl-<p, and @(z)is&-close to rank 2 t o c p ( t ) f o r l z l < a ;
2 ) the function 6(z)=Fz(z,<(z)) is 6 - c l o s e t o r a n k 2 to the function
O ( z ) = F ? ( r ,~ ( 3 )f)o r Izl<o:.
In virtue of condition (c), x = O is a double root of the function 0 ( 1 ) .
Therefore, t h e r e e x i s t s E , > 0 such that f o r sufficiently s m a l l 6, the equation
6(z\=0, i.e., ZZ(z, ( p ( r ) ) = O
h a s a t most two r o o t s which a r e s m a l l e r than F, in magnitude. Let eo be a

positive number s m a l l e r than e, such that L f e , ( 0 ) is entirely contained in the
rectangle ( z 1 S a . I y l S < B . Then, if do is sufficiently s m a l l and the functions
k', and k'? are 6,yclose t o rank 2 to F , and Fz, respectively, the c u r v e s
F,(z, y)=O and Fz(z,y)=O
cannot have m o r e than two common points in U , , ( O ) , i.e., condition ( a ) of

Definition 5 of the duplicity of point 0 is satisfied.
It now r e m a i n s to show that condition (b) of Definition 5 is a l s o satisfied,
i.e., for any positive & < e , and 6 t h e r e e x i s t functions PI and F2, 6-close to
r a n k 2 to the functions PI and F 2 , such that the c u r v e s
F,=O and F 2 = 0
have two common points in C T e ( 0 ) .

Consider the functions
and
For these functions c l e a r l y 6 = cp (z) and
21
22
C h a p t e r II
DYNA.ZIIC SYSTEiZlS C L O S E TO A GIS'EN SYSTEJI
AND PROPERTIES OF THEIR PHASE PORTRAITS
I NT RGDU C T ION
All the r e s u l t s of t h i s chapter directly follow f r o m the t h e o r e m s of

continuous dependence on the initial conditions and the right -hand s i d e s .
Although of no intrinsic significance and almost trivial, these r e s u l t s a r e
absolutely e s s e n t i a l for a r i g o r o u s treatment of the main m a t e r i a l . This
background chapter con.;ists of two sections.
In 8 3 , 6 - c l o s e s y :s t e m s are defined and the relevant t h e o r e m s of
continuous dependence a r e stated ( T h e o r e m s 8 and 9). Some p r o p e r t i e s of
r e g u l a r mappings are considered. The next section, 4, c o n s i d e r s i n t e r -
sections of paths of c l o s e s y s t e m s with arcs and cycles without contact; it
is established that the behavior of the paths of s y s t e m (A) which is
sufficiently c l o s e to s y s t e m (A) relative to a r c s (or c y c l e s ) without contact
is on the whole s i m i l a r :o the behavior of the paths of the original s y s t e m
( A ) . The r e a d e r who is i n t e r e s t e d in following the main line of argument
can s k i m through this chapter, omitting the proof of the v a r i o u s l e m m a s
and concentrating only on the relevant statements. The r e a d e r must
acquaint himself, however, with the concepts of e - c 1o s e r e g i o n s
(Definition 7 ) , ? - t r a n s l a t i o n (Definition 8 ) , and e - i d e n t i c a l
p a r t i t i o n s o f t \ \ o r e g i o n s i n t o p a t h s (Definition S), which a r e
introduced in this chapter. The l a s t of these definitions - e-identical
partition into paths - is the most significant: i t is u s e d in the definition
of s t r u c t u r a l stability, which is the main subject of the book.
3 . CLOSENESS OF SOLUTIONS. REGULAR

TRANSFORRIATION OF CLOSE SYSTERIS
1. T h e o r e m s of c l o s e n e s s of solutions
We will consider syst'ems of differential equations (dynamic s y s t e m s ) of

the form
These s y s t e m s are defined in a bounded region G of the plane ( x , y); they

are often considered, however. only in s o m e closed subregion G* of G.
23
Ch.11. CLOSE DYNAMIC S Y S T E M S
System (A) is said to be a s y s t e m of c l a s s k o r of analytical c l a s s i f P and

Q a r e functions of class k or analytical in G .
Consider two s y s t e m s defined in G,
and
which both belong to s o m e cltss k o r are both analytical.

De f i n i t i o n 6 . System (A) is said to be 8 -close to rank T to system (A)
in G ( m C*) ifthefunctions P and 0 in G (mE *) are 8 -close to rank r to P
and Q , respectively ( s e e Definition 1, Sl).
Let
O ( G Y ) - - Q ( G Y ) = ~ ( ~v).
P(,Y)-~(,~Y)=P(I,Y), ,
The functions p (2,y) and q (z,y) a r e the i n c r e m e n t s of the right-hand
s i d e s of s y s t e m (A). If t h e s e functions a r e 8-close to r3nk r to zero, w e
w i l l c a l l them 6 - i n c r e m e n t s of r a n k T . System (A) considered in
conjunction with (A) is called m o d i f i e d ( r e l a t i v e to s y s t e m (A)), and is
s o m e t i m e s written in the f o r m
I n w h a t follows, i f s y s t e m (A) i s 6 - c l o s e - t o r a n k 1 t o s y s t e m
(A), w e s h a l l s i m p l y s a y t h a t s y s t e m (A) i s 6 - c l o s e t o
s y s t e m (A), omitting the qualification "to r a n k 1."
Consider two vector fields defined by s y s t e m (A) and a 8-close s y s t e m
(A). At each point M (I,y) of G two v e c t o r s a r e defined, v ( P , Q) and @, 0).
Let v # 0, # 0, and let f3 be the angle between v and ( s e e QT, Appendix,
95.1). It is readily s e e n that this angle is infinitesimal f o r sufficiently
s m a l l 6 . Indeed,
and f o r and 0 which a r e c l o s e to P and Q sin 0 is close to z e r o w h i l e c o s 13

is positive (close to + I).
Let u s now formulate for t h e s e s y s t e m s the t h e o r e m of continuous
dependence of solutions on the i n c r e m e n t s of_the right-hand s i d e s and on
the initial values. Let the s y s t e m s (A) and (A) be defined in G. Let
z=cp(t---to, 20, Yo), y=$(t--to, so, Yo) (1)

be the solution of s y s t e m (A) corresponding to initial conditions to, x o , yo.
Solution (1) is defined f o r a l l t in some interval T < t < T . Let T, and TZ
be two numbers, such that 7 < T, < to < r 2 < T. Let L be the path c o r r e -
sponding to solution (1). Let the segment of L corresponding to the values
24
43. REG( L A R TRAE;SFORXI.-\TIOS OF CLOSE SYSTElfS
of t , 7, < f g r ? be completely contained in the closed region G* (c*c G), and

let
be the solution of systeni

- -yo.
(x)corresponding to the initial conditions to, q,
T h e o r e m 8. F a r m y e > 0 , there exist q > 0 a n d 6>0, such that if
I ) ~ ~ o - ~ o ~ l&-yol-=f;
< q , and 2) system (A) is b-close to s y s t e m ( a ) i n G*,
solution ( 2 ) of system (8)is defined fmall t . ~ l , < t , < ~ ~ , and in this time
interral
- c -
(cp(!-to,xoo,yo)-cp(t--to, Xoryo)[<e,
$@--to, x; d-9(1--10, Zot yo)I<e.
Theorem 8 is a particular case of Theorem 2 f r o m subsection 1 in

the Appendix.
R e m a r k . Since the functions cp(t-to, .to,yo) and 9 ( t - t o , io. yo) are
continuous, and are therefore uniformly continuous in t o v e r the segment
q < t , < T z r Theorem 8 c l e a r l y c a n be strengthened a s follows: f o r any
e > 0 , w e can choose q > 0 and 6 > 0 such that i f in addition to conditions 1
and 2 above w e a l s o have condition 3 ) I t - t I < q (t and t are any two
values f r o m Irl, T ~ I ) , ther.
17(t-to, 20, yo)-cp (t--to. Io7 Yo) I<,

JC(t--to, &yo)-9rp(t-to, 2-0, YO)I<.
Theorem 8 c a n a l s o b e formulated in geometrical t e r m s . T h i s f o r m u l a -

tion is particularly convenient f o r what follows.
Let L be a path of s y s t e m (A), M 0 and MI a r e the points of this path
corresponding to t o and f l ; the a r c MoM1 of L is completely contained in a
closed region e*
(a* cG). M ( t ) r e p r e s e n t s the point of L which c o r r e s p o n d s
to time t . Similarly, a
(e is the point of the path d o f s y s t e m (xi)
corresponding to t i m e t .
G e o m e t r i c a l f o r m o f T h e o r e m 8 . F o r m y e > O , there exist
q > 0 and 6 > 0, such that if system (A) is 6 -close to system ( A ) i n C * and
the path 2 at f = topasses through the point ,q0E U , ( M a ) , the corresponding
motion along E is uniquely determinable f o r any t , t o < t , < t l , and f a r these
t , d ( t ) E U, ( M ( t ) ) . In particular 3,= h? (fl) E U, (MI)(Figure 1).
The next theorem is a generalization of Theorem 8. Let F be a closed
bounded region, F c G. Consider the solutions
2=cp(t--to- 207 Yo), y=*(t-to, ZO, Yo)
of s y s t e m (a)
corresponding to v a r i o u s points MO (xo,yo) of region F .
Suppose that e a c h solution ( 1 ) is certainly defined in a closed interval
25
26
91.. REGLL.4R TR.LNSFOfihlATION OF CLOSE SYSTELIS
2. E -Closeness of regions. L e m m a s of regular

transformation
In this subsection we formulate two propositions concerning the

substitution of v a r i a b l e s in s y s t e m s of differential equations. The proof
is elementary and i s t h e r e f o r e omitted. ,
F i r s t w e have to def me the concept of E - c l o s e n e s s of regions.
D ef i n i t i o n 7. The closed regions GI and 4 are said to be E - close
I) each point in GI is distant less than Efro)ti G,, and cont.ersely each
point in G2 i s distant less thun E f r o m G,;
2) each boundary point of G, is distant less than e f r o m the boundary
of C2, and cont~erselyetzch boundary point of 6?is distant less than E froin
the boundary of 8,. *
Clearly if G2 is the image of under a topological mapping f and f o r any
Mf))< e, regions
-11 E G,, p ( M , f and c2are e-close.
Consider a r e g u l a r transformation of v a r i a b l e s of c l a s s k + 1
u --F (X. y). c=*(x, y). (3)
which is defined in s o m e region of the plane (.r, y) ( s e e QT, Appendix, S 6 . l ) .

We s a y that this transfcirmation is (E-close to rank r ( r , < k f 1) to the identity
transformation
u = x, u=y. (4)
if the functions q (I, y) and $ ( x , y) in this region are & - c l o s eto rank r to the
functions 3: and y. respectively.
Suppose that transformation ( 3 ) is considered in a n op?n domain G, and
GI is a closed bounded region, 6 , c G. Let
x=f(u, r), y = g ( u , c) (5 1
be the i n v e r s e of transformation ( 3 ) . 8: is the (u. c) image of GI under

transformation ( 3 ) .
Theorem 4 (the theorem of a s m a l l increment of a n implicit function)
and the compactness of c, c l e a r l y show that if transformation ( 3 ) is
sufficiently c l o s e to r a n k r to the identity transformation (4), t r a n s f o r -
mation (5) in is a r b i t r a r i l y c l o s e to rank r to the identity transformation
x =u, y = L.
If the plane ( u , L.) coincides with the plane (z, y) and the a x e s x and y coin-
cide with the a x e s u and c , respectively, i.e., if the point .\I* (u,c) is in fact
a point in the plane (+, y) with the coordinates u , u (relative to the s y s t e m of
coordinates defined in t nat plane), then, transformation ( 3 ) being 6-close
to the identity transformation, we immediately conclude that a:
is 6,-close
to z,, where 61 = v/z6.
Consider a dynamic system of class k
SimpIe examples will show that conditions 1 and 2 are independent.
27
I defined i n domain G of plane (2, y).
Ln.11. CLOSE DYNAMIC SYSTEMS
We apply to t h i s s y s t e m a r e g u l a r
transformation ( 3 ) of c l a s s k f 1:
p* (% 4 = 9;(f (% % g v)) P (f (44,g ( u , 4)+

+ 9;(f (u. z9)v g( ~ 9 v)) Q (f (u, v), g (u, 4)-
Q* (ut v) = 9; (f (u, v), g (a,u ) ) P (f (ut v), (u, v)) +
+ qb (f (u, g (u, 4)Q (f (u,v) 8 0)).
(A*), l i k e (A), is c l e a r l y a s y s t e m of c l a s s k.
Treating u , u as coordinates in the plane (x, y) and r e v e r t i n g to previous
notation, we write s y s t e m (A*) in the f o r m
I Let
dx
-
dt = P* (5, y), -$-= Q* (x, g).
F, c G be a closed bounded region, and GT i t s image under (3).
(A* )
L e m m a 1. For any e > 0, there exists 6 > 0 such that if transfm-

m a t i m (3 is 6 -close to rank k -+ i to the identity t r a n s f m m a t i a in G, then
-
G : c G and system (A*) in 9 i s 6-close to rdnk k to system (A).
Now consider two s y s t e m s " o f c l a s s k defined in G (at t h i s stage, it does
not m a t t e r whether t h i s region is closed o r open):
dv
2%
dt = P (x, Y), = Q ( x , y). (A)
d.t du -
-dt= P ( x , Y), -;ii-=Q(x, Y). (A)
and l e t
u = v ( x , y), v=*(x. El) (3)

be a regula: transformation of class k + 1. Applying t h i s transformation
to (A) and (A), we get
-=
du
dt P* (4 4, x
dv = P(u, 4 (A* )
and
du
-=
d* p* (u, VI, dv =
dt P (u, v) (X*1
which a r e both s y s t e m s of c l a s s k defined in G* (G*is the image of G under
(3)).
L e m m a 2. F m any e > 0 there exists 6 > 0 such that if system (A) is
8-close to rank k to system ( A ) i n G, then the system(A*)i s e-close to
rank k to system (A*)in G*.
28
$3. REGL'L.AR TRA?iSFORhl.4TION OF CLOSE SYSTEMS
In conclusion of this section, we consider s o m e s i m p l e p a r t i c u l a r cases

of i n c r e m e n t s to right-hand s i d e s of a given s y s t e m (A), which are often
encountered in what follows.
Consider two s y s t e m s defined in G:
The s i n e of the angle 0:' between the direction of the-field described by

s y s t e m (A) and the direztion of the field of s y s t e m (A) a t any point i n G is
e x p r e s s e d by
The angle 0 is positive a t any point where PQ-QB>O, i t is negative if

PG-Qp<O; wherever PG-QP=O the f i e l d s of s y s t e m s (A) and (A) a r e either
p a r a l l e l (0=0) or antiparallel (e=n), so that a t the relevant points the paths
of s y s t e m s (A) and (A) a r e tangent to one another.
Consider i n c r e m e n t s of the f o r m
P=-pfQy q=+pfP, (6)
where f = f (z, y) is some function of the s a m e class as the functions

Q (z, y), P (z,y), and p is a p a r a m e t e r .
If the relevant region is closed and
bounded, then f o r sufficiently s m a l l p, p and q-are evidently a r b i t r a r i l y
s m a l l (to s o m e rank r ) increments. System (A) then h a s the f o r m
In particular, consider the case f = 1. Then the modified s y s t e m is

xd =
r P-pQ, -$$=Q+pP,
The sine of the angle-@ betureen the directions of the field of s y s t e m ( A ) and
the field of s y s t e m (A) is e x p r e s s e d by
In o u r p a r t i c u l a r case ( f E l), the angle between the fields of (A) and (A)
is constant everywhere in the region,
We say in this case that t h e f i e l d o f s y s t e m (A) i s r o t a t e d

t h r o u g h a c o n s t a n t a n g l e r e l a t i v e t o t h e f i e l d of
'The angle between two ordered vectors is defined as the angle not exceeding 180' through which the first
vector should be rotated so as to coincide with the second vector.
29
Ch.11. CLOSE D Y N A M K SYSTEMS
s y s t e m (A), o r alternatively, that i n c r e m e n t s of the f o r m ( 6 ) f o r f =1

produce a rotation of the field through a constant angle.'::
L e m m a 3 . The states of equilibrium of system (71 are those of
system (A) and vice versa.
P r o o f . System ( 7 ) is in equilibrium a t any point where
P - p f Q = 0, pfP + Q = 0. (11)
Since (1 1) is a s y s t e m of linear homogeneous equations for P and Q and the

determinant of t h i s s y s t e m
-;fI=i+pf.
does not vanish f o r any x , y, equations (11) can be satisfied if and only if
p (5,Y) = Q (5,v) = 0.
i.e., if and only if the point x , y is a s t a t e of equilibrium of s y s t e m (A).
This completes the proof.
S4. INTERSECTION O F PATHS OF CLOSE SYSTEMS

WITH ARCS A N D CYCLES WITHOUT CONTACT
1. Intersection with one a r c without contact
This section p r e s e n t s a number of elementary, almost self -evident

propositions which a r e analogous to those discussed in QT, 53. The main
difference-is that together with s y s t e m (A) we w i l l a l s o consider modified
s y s t e m s (A). Since t h e s e propositions a r e repeatedly used in what follows,
they will be given detailed proof.
Consider a s y s t e m of class k
dt
~ = P ( - GY), +=Q(G Y), (A)
defined in G, and let c*be a closed region, QcG. The modified s y s t e m

dr
-=
dt p"(GY), -$=B(x.y) (m
is a l s o defined in G .
Consider an a r c I o r a cycle C , which a r e without contact for a path of
s y s t e m (A) completely contained in g * . The following self -evident
proposition is given without proof.
L e m m a 1 . There exists 6o > 0 such that if system is 8,-close to (A)
system @ ) i n 8*, the arc I (the cycle C ) i s without contact for paths of
system(A), and these paths make with arc 1 (cycle c) an angle of the same
sign ( s e e QT, Appendix, 55.5) as the paths of system(AX
* With increments of this form, the field vectors of system (A) are not only turned through a constant angle
but also stretched in a ratio of J/v.
are concerned only with directions, however.
We
30
$4. LNTERSECTIOh {IF PATHS WITH ARCS AivD CYCLES \VITtiOCT CONT.ACT
To simplify f u r t h e r formulations, w e will now introduce the concept of

e-translation.
D ef i n i t i o n 8. The inapjing f oj' a set E in a tnetric space R into the
sattie space is called E-translation* i f f i s a topologkal mapping and i f f o r
each point J I E E the dislance beticeen the original and the image i s less
than P , p (M, f (31))< E .
lvow consider s y s t e m s ( A )and (A) in regions H 5nd k (closed or open),
respectively, H c G , C G (in particular, ( 4 ) and ( A ) may coincide, and
then the s a m e s y s t e m is considered in two different subregions of G ;
alternatively, I and may coincide).
D e j i n i t i o n 9. A partition of H b y the paths of system ( A )and a
partition oj. i 7 b y the paths of systetti (A) are said to be E - identical, or in
sytribols
(If, A ) & @ , .T),
i j there exists a ttiappiiig of H onto 2 which is an E - translatz'pz and tohich

transforms the paths 0,'.systein (A)intothe paths of systein(A).**
Thus, the partitions of H and 3 by the paths of ( A )and a r e -identical (x)
i f these partitions have the s a m e topological s t r u c t u r e and are "distorted"
o r "translated" one relative to the other by a n amount less than E . T o
e n s u r e -identity
( H , A)&((r7, 1).
H and should be horr.omorphic and E -close to each o t h e r . N o r e o v e r , i t
i s n e c e s s a r y that the partitions of t h e s e regions by the paths of the c o r r e -
sponding s y s t e m s have the s a m e topological s t r u c t u r e (see QT, 15,
Definition V). T h e s e n e c e s s a r y conditions, however, a r e not sufficient
in general, since even if they a r e satisfied, t h e r e may prove to be no
topological mapping of H onto a
which c o n s e r v e s paths and is a t the s a m e
t i m e a n -translation.
Now suppose that the arc 1 , which is a n a r c withoat contact f o r the paths
of s y s t e m ( A ) , is defined by the p a r a m e t r i c equations
z=i(s), Y==g(S)*
w h e r e a s s 4 b. According to the definition of a n arc without contact, f (s)

and g (s) a r e continuous1.y differentiable continuous functions. Let 1 3 > ~ 0 be
the number introduced in L e m m a 1, i.e., such that f o r a l l modified s y s t e m s
(-a do-close to s y s t e m ( A ) in 8* (8* c G ) , th_e arc 1 is a n a r c without contact.
We will only consider modified s y s t e m s ( A )which are &-close to s y s t e m ( A ) .
* The term e-translation is generally understood in a wider iense. Indeed. 8-translation is defined as a
continuou.? mapping (not n e e isarity topological) which translates e a c h point of the set by less than e.
However, we will have opportunity to use this concept only in relation t o topological mapping;, and
therefore in what folIows e-translation is understood in the restricted iense of Definition 8 .
** A mapping which transforms paths into paths is given in QT (15. Definition V). This is a topological
mappinz. such that any two points of a path of system (A) are mapped into points of o n e path of system (x),
and any t w o points of a path of system (A) are mapped by the reverse mapping into two points of o n e path
of system (A). In QT this m t p p i n g i s c a l l e d an i d e n t i f y i n g m a p p i n g . We wilI refer to it as a
m a p p i n g w h i c h m a p s p a t h s i n t o p a t h s or more briefly a p a t h - c o n s e r v i n g m a p p i n g .
31
Let
x=cp ( t - t o , 501 Yo), Y = 9 (t--o, SO? Yo)
be a g e n e r a l solution of s y s t e m (A) and

-
x=cp(t--to. to, Yo).
-
I/=Q(f--lo, 20, Yo)
a g e n e r a l solution of s y s t e m (A). Then the equations
+=cp(t-to, f(s), g(s))=@U, 4, Y=$(t--to. f (4, g ( s ) ) = Y ( t , 4

and respectively
-
s=cp((t-fo), f (s), g ( s ) ) = 6 ( t , s), u=$((t-to), f (4.g ( s ) ) = Q ( t t 4
for any fixed sL a < s < b , a r e equations of that path of s y s t e m (A) (and
respectively (A)) which for t = to c r o s s e s the a r c I a t the point (j(s), g (s))
( i . e . , a t the point of the a r c I which c o r r e s p o n d s to t h e given value of the
p a r a m e t e r s). According to the definition of the functions cp, 9, @, w e see 6
that
-
Q, ( t o , s) = Q, ( t o , s) = f (SI, Y (to, s) = f ( t o , s) = g ( s ) . (5)
Now suppose that a t any point M o(zo, YO) of the a r c 1 (to= f (s), go = g (s). a d s < b )
solution (1) is defined for a l l t , to<f,<T (s), * where T (s) is a continuous
function (in p a r t i c u l a r , it may b e a constant), and that the corresponding
a r c of the path is completely contained in G* and h a s no common points with
the a r c 1 , except M o . Evidently, under t h e s e conditions, the functions
CD ( f , s), 'Y ( f , s) a r e a p r i o r i defined everywhere in the closed region
a<s<b, t,<t,<r(s). (6)
By QT, 5 1 . 3 , L e m m a 5, the functions Q, and '4 have continuous f i r s t - o r d e r

p a r t i a l derivatives in region (6). which are e x p r e s s e d by
@;v, s ) = d ( t - - t o , f(49 g ( s ) ) ,
( t , s) = ,cp;
Q,:
%(4
- t o , f (4,g (s)) f' (4+ 'pio( t - to* f (4.8 (4) d (s),
S)=4Go(1--o,
(t
'Yi( t , s) = 9; (t -tor f (4, g (s)),
f(4,g ( S ) ) f ' ( S ) + $ P ; o ( ~ - - f of , (4,g ( s ) ) g ' ( s ) .
F u r t h e r , by QT, 5 3 . 5 . Lemma 8, equations ( 3 )
t (7)
x = Q, ( t , s), I/ = Y ( t , s)
define a r e g u l a r mapping of region ( 6 ) onto s o m e closed region K i n the

plane ( x , g) with the a r c I a s p a r t of i t s boundary ( s e e Figure 2; region (6)
is shown in Figure 3). In the light of our assumptions, region K is
contained in @.
* Or for all t , t o > t > 7 (z). This case is analyzed precisely in the same way.
32
44. [NTERSECTION OF PATHS WITH ARCS AND CYCLES WITHOL'T CONTACT
gE
54
FiCCrRE 2. FIGL'RE 3
L e m m a 2. For any e > 0, there exists 6 2 0 such that if the system (A)
is 6-clo:e t o ( A )in G*, the functions 6 (f, s) and Y ( t , s) cmresponding to
system (A) are defined in region (6), hace continuous f i r s t - order partial
derittatices in this region, and are &-closethere to the functions Q, ( t , s) and
Y ( t ,s).
The proof of t h i s l e m m a follows directly f r o m Theorem 9 ( 3 ) and f r o m
Appendix, subsection 1, T h e o r e m 3.
L e m m a 3. (a) There exist 6 > 0 and h > 0 such that the system is
6 -close in b to systew(A!, the mapping
x = 6 (t. s), y = F ( t , s) (4
is a regular mapping of the rectangle
a<s,<b, It--t,l,<lr (8 )
in the plane ( t , s) onto the closed region K" in the plane ( x , y), and R i s entirely
contained in @.
(b) Let K be a closed region which is the image of rectangle ( 8 ) under
the transformation
x=D(t, 4, Y=Y(t, s), (3)
system(A). For any e > 0 , there exists

corresponding to the rrinitiglrr
8* > 0 such that if system <A)is 6*-close to ( A )in @, I? i s E -close to K .
P r o o f . Let u s fir s t prove (a). According t o the definition of a regular
mapping ( s e e QT, Appendix, S6.1), we have to prove that a n appropriate
choice of6>Oand h > Owill make ( 4 ) a one-to-one mapping in rectangle ( 8 )
and that everywhere in this rectangle
33
Ch.11. CLOSE DYNAMIC SYSTEMS
By QT, 53.4, Lemma 3, t h e r e e x i s t s ho > 0 such that the mapping (3)

is r e g u l a r in the rectangle
a,<s,( b, I t - to I<ho. (10)
Hence, a t any point ( t , s) of the rectangle ( l o ) , the determinant
h a s the s a m e sign. Suppose that t h i s sign is positive. Then a t any point
r e m e m b e r i n g that the rectangle is compact and the e l e m e n t s of the

determinant A are continuous, we conclude that t h e r e exist q > 0 and u >0
with the following property:
if
I F r o m the l a s t relation and L e m m a s 1 and 2 i t follows that t h e r e e x i s t s
I and s i ( i = 1, 2, 3, 4) satisfy (11), 2 is a n a r c without contact for the paths
f o r which ( 4 ) is not a r e g u l a r mapping of the rectangle (8) into the plane

(z,y). As 6 and h we choose any two numbers satisfying the inequalities
6<80, h<ho, h<%. (13)
This_ choice c l e a r l y does not detract f r o m the generality of our argument.

@ and Y a r e single-valued functions. Now, by (12).
in rectangle (10) and hence a l s o i n rectangle (8). Therefore mapping (4)

is not r e g u l a r only if i t is not one-to-one, i.e., i f i n rectangle (8) t h e r e is
a t l e a s t one pair of different points (t', SI) and (t", which are mapped
SI)
under (4) into the s a m e point of the plane (z,y). i.e., such that
34
Xpplying T a y l o r ' s expansion to these equalities, w e see that
Gt ( t l , SI) ( t ' - t " ) + 8):( t l , SI) (s'-s*) =o,
Ti(!?. SZ) ( 1 - t " ) + R ( t 2 * s*)(s'-sJ')=O. (15)
whsre t, and t2 a r e numbers lying between t' and t', and si and s, are
n u m b e r s lying between s'and s". We will now show that I s ' - s" I S G . Indeed,
i f I s' - s" I <,.T, w e have I sl - .s2 I < u. F u r t h e r m o r e , I f l - t 2 I < I t - t' I < q
and I tt - to I < ko.l t 2 - t3 I < ho by (13). Hence, inequalities (11) are s a t i s -
fied, and by ( 1 2 )
The l a s t equality sh,3ws that s y s t e m (15) considered with r e s p e c t to

t' - t". s'
- s" h a s a nowanishing determinant, i.e., t' = t". .s' = s*. T h i s is
at variance with the s t a r t i n g assumption that (t', s') and ( t " . s " ) are two
different points in the rectangle (8). Hence I s' - s" J > u .
Let u s consider two sequences of positive n u m b e r s ht and h i , i = 1,2,. . ,,
satisfying conditions (13), such that lim&,=limki=@. We have s e e n that
I-= i+=
if proposition ( a ) ,of the l e m m a is not t r u e , then for each p a i r 6 i and k, t h e r e
e x i s t s a s y s t e m ( A i )which is &;-close to s y s t e m (A) and a pair of points
(t:. si). (t:. s;) such thatIsI-s;I,oand
;Di ( t L ,Si)
- (f:.s;),
=ac(ti, s:), vi ( t i , s;) = Y, (16)
u-here and ?La r e the functions corresponding to s y s t e m Since (xi).

\t;-fo(<hi -0 andiff-), ( h i - 0 , we have limtl=limtl=to. If n e c e s s a r y w e
..-z ,-a
c a n pick o u t suitable srbsequences, so that si and s: c a n always be made

to converge. Let lims;:=sA, lims?=s;. C l e a r l y a < s ; g t , a < s ; u ,
&-I ,*=
F r o m the definition of ib, and qi and by T h e o r e m 8 ( 8 3 ) together with the
r e m a r k to that theorem. we obtain the equalities lim&i ( t i , s;) -4, (to, s;) and
I-*
lim&,, (t;, s ; ) = ( D ( t , , s;) and s i m i l a r l y f o r 'Yi. T h e r e f o r e , taking the limit in

2-*
(16), w e get
or, f r o m (5),
T h e s e relations a r e a t variance with the inequality Is:-& 120. P r o p o s i -

tion ( a ) of the l e m m a is thus proved.
The validity of proposition ( b ) follows directly f r o m T h e o r e m 9 ( 3 ) ,
which completes the proof of the l e m m a .
R e m a r k . It follo.w_s directly f r o m L e m m a 3 that i f 6 > 0 is sufficiently
s m a l l arud the s y s t e m (A) is 6-close in g* to the s y s t e m (A), the paths of
s y s t e m ( A ) which meet a t t = to the arc 1 ( a r c without contact) have no
common points with the a r c I f o r a l l other values of f . t i t o , I f - co I<h.
35
01.11. CLOSE DYNAMIC SYSTEMS I
Let now M O( X O , YO) be an i n t e r i o r point of the a r c 1 , corresponding to the
value SO of the p a r a m e t e r s (a < so(b).
L e m m a 4 . For any e > 0 , hO> 0, there exist q > 0 , 8 > 0 which satisfj!
the following condition: system (A) i s 8 -close to_systeim ( A )in *, M is
some point in U, ( M o ) , and L" i s the path of system ( A )which at t = to passes
through the point M , then f o r some t * , I t* - to I < ho, the path intersects
the arc 1 at the point M 4 , such that the arc MM* of the path f; is entirely
contained in U. (M,) (Figure 4 ) .
FIGURE 4 FIGURE 5
P r o of. Fix e > 0 and ho > 0. Without l o s s of generality w e may take

C G*. Choose u > 0 such that the p a r t of the a r c I corresponding to
U. (Mo)
the values of the p a r a m e t e r s,so - crgs,<so +
a, is completely contained in
U ,( M o ) . By L e m m a 3 (a), using the continuity of a l l the functions, we s e e
that t h e r e e x i s t 8 > 0 and h, O < h < h o , such that the mapping
s=tD(t, s), y=Y((t, s), (3)
corresponding to s y s t e m (A), and a l s o the mapping

-
z=a,(t, s), y=f#((t, s), (4)
corresponding to any s y s t e m (A) which is 6-close to s y s t e m (A) in @, a r e

r e g u l a r in the rectangle
in the onto a. closed region H (or g)in the

plisne , is evj.dently an i n t e r i o r point of H
and 17. Let r be the distance of M o f r o m the boundary of H ( F i g u r e 5). By
Lemma 3 (b), 8 c a n b e cosen so s m a l l that a is +-close to H. Then the
boundary of is $-close to the boundary of H . For '1 we can take any
positive number s m a l l e r than G. Indeed, if q < 5, it is r e a d i l y s e e n that
U, (Mo)is completely contained in each region I?. But then f r o m the
36
54. INTERSECROX OF PATHS W I T H ARCS A N D CYCLES WITHOCT CONTACT
definition of r? and the inequality h < ho we conclude that the p a r t i c u l a r q and

S that w e have chosen satisfy the condition of the l e m m a . Q. E. D.
R e m a r k . Lemma 4 ban be generalized as follows. Let i.be a p a r t of
the arc without contact I which is contained entirely in 1 ( i . e . s the end-points
of the a r c i.do not coincide with the end points of I ) . Then for any E > 0 and
h 7 0 theEe exist q > 0 and 6 > 0 which satisfy the following condition: if
s y s t e m ( A ) is 8 -close to s y s t e m ( A ) in E*, N is any point whose distance
f r o m the_arc 1 is less tnan q (i.e., Jf E L', ( I ) ) > and E is the path of the
s y s t e m (A) which a t t = to p a s s e s through the point .V3 then f o r s o m e f * ,
IF - t o ) < h , the path i i t e r s e c t s the arc 1 a t the point M*, and the arc
X M * of the path E is en:irely contained in U , (A).
rhe validity of this proposition c a n be established in the usual way by
reductio ad absurdum, using L e m m a 4 and the compactness of the arc 1.
A s before, let the a r c without contact 1 lie inside the region 6*. Con-
s i d e r s o m e i n t e r i o r point .If0 (so, yo) of this region which does not belong to 1 .
Let the path L of s y s t e m (A) p a s s through the point -)fa)a t t = t o . and a t
t = T i- t o i t c r o s s e s the arc 1 a t a point M , which is not a n end-point of the
a r c 1 . Moreover, let the a r c M o X of L be entirely contained in G*
( F i g u r e 6).
L e m i n a 5 . F m any e>O, h>O, there exist q > O and d > O such that if
systein ($1 is a - close to system (A) in E*, .goE U,, (.wd, and is the path of
systein (A)zuhichat t = t o passes through the point;M,, then at some t=G the
path L crosses the a r c I at the point M so that (a) I T - z I c h ; (b) z % U c ( ~ v f ) ;
(d the a r c ROL% of the path Z is Contained entirely in G+; (d) i j I.r'-tol<
-tol,
>17--t01, then . @ ( t ) U , ( M ( t ) for
) all t c [ z , to]and the arc of the path L coyre-
sponding to t E ,I; T ~ i ]s entirely contained in U. ( M (z)) * (Figure 6 ) .
L e m m a 5 is a n obvious c o r o l l a r y f r o m Theorem 8 ( 8 3 ) and Lemma 4.
R e m a r k . Lemma 5 c a n be generalized like the preceding l e m m a {see
r e m a r k t o Lemma 4). Indeed, consider a compact s e t F i n G* such that
each path which a t t = to p a s s e s through the point Mo (q, yo) of t h i s s e t i n t e r -
s e c t s a t s o m e t = 7 ( 3 0 , y,) the a r c without contact 1 a t the point M (zo,yo), always
remaining inside G until the intersection point is reached; then for given
* Here, as i n the following. M ( t ) IS the point of the path of the relevant system which corresponds to the
time t for the particular motion chosen along the path.
37
(3.11. CLOSE DYNAMIC SYSTEMS
e > 0 and h > 0, the n u m b e r s q > 0 and 8 > 0 entering Lemma 5 can be chosen
independently of the point M o E F , and w i l l thus satisfy the lemma i r r e s p e c -
tive of the p a r t i c u l a r point Mo E F that is taken. In what follows, the compact
s e t F is generally identified with s o m e a r c without contact 1 which has no
common points with the a r c 1 .
In Lemma 3 the functions Q, ( t , s),Y ( t , s), and a l s o (t, s) and q ( t ,s), were
considered only f o r t such that t - to I<ho, where ho > 0 is some sufficiently
s m a l l number. Let u s now again consider the c a s e when the functions
Q ( t , s) and Y ( t , s) a r e determined f o r a l l t and s in the closed region
a,<s,<b, to<t,<7(s), * (6)
where 7 (s) is a continuous function. W e a s s u m e , as before, that for all

t and s f r o m t h i s region, (0( t , s), Y (f,s)) E G * and that if to < ~ Q T(s) and s, d
a r e any two n u m b e r s f r o m [ a , bl, a t l e a s t one of the following two inequali-
t i e s is satisfied:
Q ( t ? s) # Q, (lo, s), YJ ( t , s) # y (to, s),
i.e., any path of s y s t e m (A) defined by the equations
x=cp(t--to. f(sh g(s))=Q,(t, 4, y=$(t-t0, f(s), g ( s ) ) = y u , SI, (3)
h a s no common points with the a r c 1 for t going f r o m to to T (s), except the

one point a t t = t o .
W e have noted above that under t h e s e conditions equations ( 3 ) define a
r e g u l a r mapping of the region ( 6 ) in the plane ( t , s) onto s o m e closed region K
in the plane ( x , y) ( F i g u r e 2).
Consider the following lemma.
L e m m a 6 . F m any e > 0 there exists 8 > 0 such that if system (A)is
a
8 -close in G* to system(A), the functions ( t , S) and f ( t . s) are defined in the
region ( 6 ) and the equations
x = & ( ( t , s), y=Q(t, s) (41
describe a regular mapping of (6 1 onto some closed region R , 4 c G*, such

that the regions K and K are e-close.
P r o o f . It suffices to show that for s m a l l 6 the mapping ( 4 ) of the
region ( 6 ) is r e g u l a r , since the other propositions of the l e m m a a r e
contained in Lemma 2. To establish the r e g u l z r i t y of the mapping (6), it
suffices to show that the paths of aqy s y s t e m (A) which i s 8-close t o the
s y s t e m (A) have no common points with the a r c 1 f o r f f r o m the i n t e r v a l
to < t ,<C (S).
By Le_mma 3, t h e r e e x i s t 8%> 0 and h > 0 such that the paths of any
s y s t e m (A) 6,-close to s y s t e m (A) which i n t e r s e c t the arc I a t t = to have
no common points with t h i s a r c f o r t f r o m the i n t e r v a l to < t Q t o h . +
+
Here h can be taken a r b i t r a r i l y small. Let to h < 7 (s) f o r all s, aQs<b.
6, > 0 is chosen so s m a l l that a l l the conditions of the lemma, p o p i b l y
except the regularity of the mapping, a r e satisfied f o r s y s t e m s (A) which
a r e b,-close to (A).
* Or t o t > F (J). Condition (6) is used without loss of generality.
35
SI
Consider the region
aSs,<b, t0ih4tSz(s) (17)
in the plane ( t . s) (the cross-hatched part of Figure 7). Because of the

particular choice of 6, a.nd h , mapping ( 3 ) m a p s t h i s region onto s o m e closed
region H in the plane (.r.y) a t a positive distance from-the a r c 1 ( F i g u r e 8).
If 6 > 0 . 6 < 6, is sufficiently s m a l l , and the s y s t e m ( A ) is 6-close t o the
s y s t e m (A), mapping ( 4 ) m a p s the region (17) onto region ?i which is suf-
ficiently c l o s e to H and therefore h a s no common points with the a r c 1 .
Since b < SI, 6 c l e a r l y s a t i s f i e s all the conditions of the l e m m a . Q. E. D.
2. P a t h s of c l o s e s y s t e m s between two arcs

without contact
Let u s n o w consider the intersection of the paths of s y s t e m (A) and of

s y s t e m (A) c l o s e to (A) with two a r c s without contact. We will prove two
l e m m a s.
Let I , and 1, be two arcs without contact for the paths of s y s t e m ( A )
which lie in G* and have no common points. Let
x=fl(S), Y=i?t(S)
be the p a r a m e t r i c equa.tions of the arc I , . Suppose that the path of

s y s t e m (A) which a t t = f~ p a s s e s through the point (f,(s), g, (5)) of the arc I ,
i n t e r s e c t s the arc l 2 a t s o m e f = T (s); the part of t h i s path corresponding
to t f r o m the interval t o . < t t , < T ($1 iS entirely contained in G* and h a s no
conimon points with l i 01' 12> except i t s two end-points (to fix ideas, we
take T (s) > t o ) .
In QT, 5 3 . 6 , L e m m a 9 i t h a s been shown that T (s) is a continuous and
therefore bounded function of s. Let MI and M z be any two points of the
a r c I , , other than i t s end-points, corresponding to the values siand s2 of the
p a r a m e t e r s, a < sI < s z < b . Let f u r t h e r the paths L , and L2 that p a s s through
the points Mi and M 2 a t t = to meet the arc l2 a t the points N , and N 2 f o r
T I = 'F (SI) and T Z = T (s?), respectively; IV, and N z do not coincide with the
end-points of 1 2 , e i t h e r . Let r designate a n e l e m e n t a r y quadrangle ( s e e QT,
5 3 . 6 , R e m a r k 2 to L e m m a 10) limited by the s e g m e n t s MIMz and Nl.Vz of the
a r c s I , and 1 2 and by the s e g m e n t s MINl and M2:Vnof the paths. Because of
o u r assumptions, I' c G* ( F i g u r e 9).
39
I Ch.11. CLOSE DYNAMIC-SYGTEMS
FIGURE 9
L e m p a 7. Foranye>Othereexist q>Oanda>Osuch thatif

system (A) i s a-close to system (A)then I , and I , are arcs without contact
for the paths of system(A), and m m e m e r
(a) if a, and are kwo points of the arc I t whichJie in U, ( M , ) and U , (hi2),
respectively, and 2, and z,
are the paths of system ( A )which pass through
these points at t = to, the paths z, and E2at t > to intersect the m c 1, at points
m,and H2which lie in U,(N,)and U, (Nz),respectively, and the s e c t i p s
i@ffliand 2,zzof these paths, together with the sections Hi&2 and N,R2of
the arcs 1%and I ~ delineate
, a region Iwhich constitutes an elementary
quadrangle f o r system(A);
(a) the elementary quadrangle F i s entirely contained in ? I and i s e-close
to the elementary quadrangle r.
Pr oof . Consider the transformation corresponding to s y s t e m (A):
I By assumption, (P
r = c o ( t , s), y = Y (f,8 ) .
and Y a r e defined f o r a l l f and I satisfying the

(3)
inequalities
I to< t < T (4, aQ 8 Q b,
and f o r all t h e s e t and s the points ((P ( t . s), Y ( t . s)) belong t o G*. By QT, 53.4,
(18)
Lemma 3, it is readily seen, however, that the t i n t e r v a l in (18) always

can be somewhat increased; t h e r e f o r e , t h e r e e x i s t s a c e r t a i n ho > 0 such
that the functions and Y a r e a l s o defined f o r
I u,<s,<~, to<t,<~(~)+ho
and for a l l t h e s e values of the p a r a m e t e r s the points ((P(f,s),Y ( t , s)) lie i n G*.
(19)
The points ((P ( t , s), Y ( t , s)) f o r which s i Q s 4 s 2 , T (s) < t< T (s) + ho evidently lie
outside the e l e m e n t a r y quadrangle r, Le., t h e s e points and the i n t e r i o r
points of l 2 c l o s e to I lie on the two s i d e s of the a r c La.
By L e m m a s 1,2,5 and r e m a r k to L e m m a 5, we see that f o r any t > O and
h>O, h < h o , t h e r e e x i s t 6 > 0 and q > O such that l i and la are a r c s without
contact f o r any s y s t e m (A) which is a-close to s y s t e m (A) in c?*, and
moreover:
1) the functions & i t , s), ( t , s) a r e defined f o r all t and I ( a < s g b , to<
. - < t < ~ ( ~ ) + and
h ~ ) f o r t h e s e values of the p a r a m e t e r s they are e-close to the
fuqctions (P and Y, respectively, the points ( g ( t ,s). q ( t , 8 ) ) lying in G*;
40
44. ~ T E R S E C T ~ OI N~ FPATHS w t r t i ARCS A N D CYCLES w t r m . x r CONTACT
2 ) if .\I(j1(s), gl (s)), sl,<s,<s2 is a point of the section .M11Z12of the a r c I ,

L is the path of s y s t e m (A) passing through t h i s point a t f = t o r .V is the point
of the path L which bel.sngs to the section .V1.l; of the a r c It. and c o r r e s p o n d s
to the t i E e 7 ( s ) , .$(;) is a point of the a r c I , , .@C,(M), is the path of
s y s t e m { A ) which a t t = : t o p a s s e s- through
- - ^
the point .<I (Figure- 9), then the
path c r o s s e s the arc I? a t f = T ( s ) , I ? ( S ) - - Z ( S ) ( ( h a t a point .S which l i e s
in L;,(.V) ( F i g u r e 9).
Let gl(T2) be the value of the p a r a m e t e r s corresponding to the point
.r?, (&).
It follows f r o m ( I ) , in particular, that the paths t,and of s y s t e m (A)
passing through the po:ints Aol E c', (M,) and .G-EDT, (.M2), respectively, meet
the a r c l2 a t the points . ~ l U z ( . Sand l ) .T2CLre(-S2). Moreover, arguing a s
before in Lemma 62 we c a n show that if 8 > 0 is sufficiently s m a l l , e a c h
path E of s y s t e m (A), which. a t t = t o i n t e r s e c t s -the - a r c I , a t the point
$I ( I ( ; ) , g G ) ) , :I,<;,<&, and m e e t s the a r c I? a t f -T(s), h a s nq common points
with I , and I? f o r intermediate values of t , f o < t < G ( s ) . But then the region
limited by the sections . ~ l . ~ z ,of- .the i l a. r~c st 1, and I? and by the section2
. ~ , . T l ,.o#? of the paths L , and L? is a n e l e m e n t a r y quadrangle of s y s t e m (A).
P a r t ( a ) of the l e m m a is thus proved. That f o r the given choice of 8>0,
is E-close to the e l e m e n t a r y quadrangle r and T c G+ follows directly f r o m
Theorem 9 ( 3 . 1 ) and f r o m proposition 2 above. T h i s completes the proof
of the lemma.
R e m a r k . Let -\I; and be any two points of the a r c t1 which lie
between .Ift and .%Il. By Lemma 7 i t is readily s e e n that if,6 > 0 is
sufficiently s m a l l and s y s t e m (A) is S-close t,o
s y s t e m (A), the arcs of the paths of s y s t e m ( A )
lying between the p a r t M;M;of the arc II and the
corresponding section of the arc l 2 belong to the
e l e m e n t a r y quadrangle I' of the original s y s t e m
(A) ( F i g u r e IO).
"r' W e will prove another l e m m a which p e r t a i n s
4 to elementary quadrangles r and F made up of
a, r, paths of t,"o sufficiently c l o s e dynamic s y s t e m s
FIGI'RE 10 (A) and (A). This l e m m a is repeatedly used in
what follows. We r e t a i n the s a m e notation as
before. Let s l a n d s2, a < s l < s 2 < b , be the values
of the p a r a m e t e r s along the a r c I t , corresponding to the points MI and M 2 r
2, and 7, the values of s corresponding to the points fi,and %?(al E c', (.+Il),
.Q, E U, ( M ? ) , q > O is sufficiently small). Then the points of the section M i N i
of the path Li ( i = 1,2) which is p a r t of the boundary of the quadrangle r
correspond to t values f r o m the interval t,,<t,<r (st), and the points of the sec -
tion *@iJTiof the path 2, ( i = 1,2) which is p a r t of the boundary of the quadrangle
correspond to t values f r o m the interval t ~ < < f( s<i ) ~ (Figure 9).
Consider a topological mapping 'p of the section M i N i of the a r c II onto
the section f i I ~such * that 'p (MI) = fiTl, 'p ( M 2 ) = &,. A suitable mapping 'p is
the l i n e a r mapping defined by the equation
- - 5-5
s = s ,sz-sl
d + q
-- s-s2
s,-s2'
where s is the value of the p a r a m e t e r in the equations of the arc 1, c o r r e -

sponding to s o m e point M f r o m M I M z , ?nd 2 is the value of the s a m e
p a r a m e t e r corresponding to the point M which is the image of the point M .
41
Ch. 11. CLOSE DYNAMIC SYSTEMS
The mappings cp of the a r c M1M2 onto the a r c M , M z are not r e s t r i c t e d to

linear mappings of the f o r m (ZO), and in what follows other c a s e s w i l l a l s o
have to be considered. The g e n e r a l mapping can obviously be e x p r e s s e d in
the form
-s = 0 (s),
with equation ( 2 0 ) considered a s a p a r t i c u l a r c a s e . If the explicit e x p r e s -

sion of the mapping cp is not given, we will always define it by equation (20).
however.
L e m,m a 8. For any e > 0 there exist 6 > 0 and 9 > o such that if
system (A) i s 6 -close in G* to system(A), and the distance between any
point M of the section M , M 2 of the arc I , and its image 'p ( M ) is less than 9,
Le., p ( M , 'p ( M I ) < '1, then there exists a path-conserving topdogical mapping
T of the elementary quadrangle r onto P (also conserving the direction of
motion along the paths) which coincides along the section M i M z of the arc L,
with the mapping 'p and is in fact its &-translation.
P r o o f . Every point P (z,y) of the elementary quadrangle r lies on the
path L of s y s t e m (A)which a t t = to p a s s e s through the point M Cf, (s), g, (5)) of
the a r c 1,. Let the point P (5, y) correspond to s o m e t i m e t . The n u m b e r s
t , s can be r e g a r d e d a s curvilinear coordinates of the 2oint P E I?. Here s
v a r i e s f r o m st to s2. If s is fixed, t v a r i e s f r o m toto 7 (s), and the point P
t r a c e s the section M N of the path L ( F i g u r e 11). The Cartesian coordinates
of the point P a r e x , g, and
x=a,t, s)='p(t--to, f ( 4 ,g ( s ) ) ,
?/='Y(t. S ) = S ( t - - t o , f (SA g ( s ) ) .
Similarly the point P"(z,
3 of the elementary quadrangle can be made to
correspond to the curvilinear coordinates 7, G, ;i<G<& to<;<<;G), and
c I - -
r=cD(t,s),
- - c
v = ' P ( t , s).
"/---I:
'I
FIGURE 11
We w i l l now define new curvilinear coordinates A, I in the elementary

quadrangle r :
S=S, t=to+k('C(S)-fo).
If s is fixed (st<s<s2) and h v a r i e s f r o m 0 to 1, the point P t r a v e l s along the

e n t i r e section M N of L . In the elementary quadrangle F w e introduce new
42
coordinates i, by the analogous r e l a t i o n s
-4st g o e s f r o m to to G G ) , X a l s o v a r i e s f r o m o to 1.
Let u s nou find the inlapping T of t h e q u a d r a n g l e r , a s s u m m g t h a t e a c h
point P (k, s) of t h e quadrangle is mapped onto a new point T ( P ) = p ( c where 6,
-s.=u)(s). -
h-i..
n o t h e r words, a point P with the Cartesian c o o r d i n a t e s
2 =@ (to 3- ( 5 (S)- t o ) , S), y = 'Y ( t o j-k (T (9) - t o ) , S)
s nlapped u n d e r T into a point P with the Cartesian coordinates

- -
z=@(Co+k(T
- (0( ~ ) ) - L O ) , 0(S)), ;=
I
(r (a(s))--to)t a (s)).
The mapping T defined in this way is c l e a r l y a one-to-one mapping of

the quadr_angle r onto 1;; it m a p s the paths of s y s t e m (A) into paths of
s y s t e m ( A ) conserving the direction of motion, and coincides with the
mapping 'p along the section .1f,.112 of the arc 1 , . It is r e a d i l y s e e n that T is
a l s o a continuous mapping; t h i s follows f r o m the continuity of the functions
'D. If', 6,(p,a, T, and (,the l a s t two functions a r e continuous by QT, 53.6,
L e m m a 9 ) . T is t h e r e f o r e a topological mapping.
Finally f r o m L e m m a 2 and a l s o f r o m L e m m a 5 and the c o r r e s p o n d i n g
r e m a r k i t follows that i.f q > 0 and 6 > 0 a r e sufficiently s m a l l , then
p ( P , 7' ( P ) ) < e f o r any P E r. T h i s c o m p l e t e s the proof of the l e m m a .
R e m a r k . It follows f r o m L e m m a 8 that i f q > 0 and 6 > 0 a r e
sufficiently s m a l l , t h e p a r t i t i o n s of the e l e m e n t a r y q u a d r a n g l e s r and by
the paths of the c o r r e s p o n d i n g s y s t e m s a r e &-identical(see Definition 9).
The following l e m m a is a s t r o n g e r v e r s i o n of L e m m a 8. We r e t a i n t h e
s a m e notation and f u r t h e r a s s u m e that, a p a r t f r o m the mapping 'p of the
section M , J f ~of the a r c 1, onto the section dgl.fi2 of the s a m e a r c , t h e r e i s
a l s o given a mapping a of the section ~ll,.V, of the path L , onto the section
-~,.~, of the path z,, s u c h that
a (W,)
-
= MI, a (.VI) =.Yi,
-
and a mapping $ of the section Jf2S2 of the path L2 onto the s e c t i o n .c2.v2
of
the path z2,
such that
-
p (.)I?) = -112, $ (.VVc)
-
=N 2
(Figure 9).
L e t n pi a 9. F m at!g e > 0, there exist 6 > 0 arid q > 0 such that If
systetn (A) is 6 -close to system (A) and the mappings 'p, a, a d $ are
q - translatioris (Definition 8, 0 4 . 4 , there exists a path - conserving topo-
logical wrapping T of'the eleiiientary quadrangle r anto lr"(iohich also
c o n s e w e s the directiov! o j rktotion ) that coincides on the sections J 1 , X 2 ,.$I,.v,,
arid M2SL. of the bounda,ry of the eletnentary quadrangle r with the wiappings
'p. c*, and $, respectirely, and i s an &-translation.
43
Pr oof . The argument constitutes a minor modification of the line of

reasoning which established the proof of Lemma 8. As in the previous
lemma, we introduce a coordinate A varying f r o m 0 to 1 along each section
of the pa? of s y s t e m (A) in r and along each se_cti_onof the path of
s y s t e m (A) in T. Along the sections M I M z and M t M z of the a r c li we m o r e -
over introduce the coordinates p and .;
respectively, which v a r y f r o m 0 to l
and a r e linearly r e l a t e d to the p a r a m e t e r s: s=s, (1 - p) s2p along the a r c +
M I M z and s =';I(1 --p) +;',> along the a r c iI?,&,. The mappings (p, a , and $
can be defined by the appropriate r e l a t i o n s F=rp (p), = a (I), x,
= $ ( I ) . (We
u s e h e r e the s a m e notation (p'a. $. This is not quite rigorous, but need not
c a u s e any confusion.) Here I1 is the value of the p a r a m e t e r h on the path E ,
at the point corresponding to the point I , on the path Lt (i= 1,2). Clearly
a (0)= p (O)=O, a (1) = p (1) = 1, rp(0) = 0. rp (1) = 1.
For T we take a mapping which maps the point ,P (x,!) E r defined by the
-
p a r a m e t e r s p, I ( O & p & l , O & h < l ) onto the point P ( z , y) with the p a r a m e t e r s
p, 4 r e l a t e d to p and I by the equalities
i; = rp (PIv h"= a (4(1-P) B ( I ) P' +

In other words, the point P (z,y) E r with the coordinates
= Q, ( t o + (Si (1--P) + SZP).), 8, (1--PI +SZIL),
Y= ($0 +Is (Si (1-I4 + SZP), si (1-PI +sap)
is mapped onto a point p ( z , 3 with the coordinates

-~ = ( t6o + ~ a ((1-p) + B(11) PI; 61(1-q
- +%
(PI)+sa9 (PI). si (1-q (PI)
~
Y = q ( ( t o + ~ a ( h ) ( ~ - P ) + B ( ~ ) P I ; 6 i (1-q
-
( C L ) ) + ~ + ( P ) ) , (~-vw)+&w)-
st
The mapping T defined in t h i s way is r e a d i l y s e e n to be a path-conserving
topological mapping of the elementary quadrangle I? onto ?? which coincides
with the mappings q , a,B on the corresponding s e c t i o n s of the boundary. If
now 6 and q a r e sufficiently s m a l l , and Y can b e made a s %lose 5 s is
a
needed to Q, and Y, a ( I ) and $ ( I ) to 1,; and to p and I , and sl and 8 2 to s,and
s2, respectively. But then f o r sufficiently s m a l l 6 and q the mapping T is
an e-translation. Q. E. D.
Retaining the previous notation, let u s consider the e l e m e n t a r y quad-
r a n g l e I? formed by the a r c s of the paths of s y s t e m (A), Here, with
s y s t e m (A), we will only consider modified s y s t e m s (A) of one p a r t i c u l a r
f o r m , namely s y s t e m s
d"=P
dt
- (S, Y), 2 = B (z,Y). (A1
such that a t any point of 61 which is not a s t a t e of equilibrium of s y s t e m (A)
we have
P"QQP#O. (21)
Thus, a t any nonequilibrium point in 6 we have e i t h e r P@-QF>O or

PQ-QF (0.
44
$4. IkTERSECTIW OF P.4THS WITfi .IRCS A N D CYCLES Lv'ITHOL-T COST.4CT
An example of such s y s t e m s is provided by 9 3 . 2 , where we described

s y s t e m s of the f o r m
f ( 5 . y) being a function which d o e s not vanish in E'.

Condition ( 2 1 ) c l e a r l y indicates that a t those points in c*which are not
equilibrium s t a t e s of s,ystem (A), the acgle between the direction of the
field of s y s t e m ( A ) and that of s y s t e m (A) r e t a i n s a constant sign.
Together with the p a r a m e t r i c equations of the arc I ,
x = / I (s), Y =RI(S)
w e will consider p a r a m e t r i c equations of the a r c L?
z = f 2 ( Z ) , y=gz(S), Z,<546.
The p a r a m e t e r s along the a r c I , is s o chosen that the paths of s y s t e m ( A )

intersecting the arcs I , and L2 make with t h e s e arcs a n g l e s of the s a m e
sign. This c l e a r l y i m p l i e s that the determinants
both have the s a m e sign. Suppose that D l > 0 and D z> 0 (this c o r r e s p o n d s
to the case schematica.lly shown in Figure 1 2 ) .
FIGL'RE 12
Let and < b e the values of the p a r a m e t e r 7 corresponding to the points

.Vi and .V~,ofthe a r c 1 2 . Also let No be s o m e point of the arc 1, which lies
between and i V 2 , so the corresponding value of the p a r a m e t e r s (sI < so < s a ) ,
Lo the path of s y s t e m (A) which p a s s e s through the point- -1.1, for t = t o , N othe
point a t which the path L o m e e t s the a r c Zz for t = T (s,), so the value of the
p a r a m e t e r s corresponding to the point NO (F, < so < g). The section M w V o of
the path Lo c l e a r l y partitions the e l e m e n t a r y quadrangle F into two e l e m e n -
t a r y quadrangles rl and r2 and is p a r t of the boundary of the two quad-
r a n g l e s ( F i g u r e 1 2 ) . The points of the quadrangle rl (or I+,)are d e s c r i b e d
by the coordinates
x=@(t. s), y=Y(t, S),
45
where
si ,<S Q So. to Q t 45 (s)
(or sogs(sz, respectively). The boundary of the quadrangle

t o < t < ~ (s),
t'I (r,)c l e a r l y
includes the section N i N 0 ( N o N 1 )of the a r c 1, corresponding
to the values g<<sgz (<gsgs2)of the p a r a m e t e r s. The sections M a 2 and
N f 1 2 of the a r c s I , and 12, respectively, l i e on the positive side of the path
Lo and the s e c t i o n s M a , and N,&, fall on its negative side.
Let S be s o m e point of the section M O M 2of the a r c I , , other than M 2 ,
and let S* be the corresponding value of the p a r a m e t e r s. Clearly, so<sL < sa.
L e m m a 10. There exists 6 > 0 such that if system (A)is 6 - close to
system(A)in @ and at any point in G*
PQ"-QP> 0, (23)
then any path L" of this system which at t = t o meets the section MoS of the
ayc 1, at the point A7 will cross the sectionNJf,of the arc l2 at the point
at t = ; , so that the section I@@of Z has no common points with the arcs 1,
and I , , except its two end points, and is completely contained in the
elementary quadrangle r2. A similar proposition holds true if the point s
lies on the section M O M l of the arc I, and Pd-QF<O. In this case, the
section f i f i of E i s entirely contained in the quadrangle P,.
P r 00:. By the r e m a r k to Lemma 7, t h e r e e x i s t s 6 > 0 such that i f
s y s t e m ( A )is 6-close to s y s t e m (A) and i t s path 2; p a s s e s through s o m e
point 2 of the a r c MQS at t = t o , then at t the path =<
m e e t s the a r c l 2 at
s o m e point 8 s o that the section M f i of Z h a s no common points with the
a r c s 1, and 1 2 , except i t s end points, and is entirely contained in the
s t a r t i n g quadrangle r. W e will now show that if inequality (23) is addition-
ally satisfied, the section M f i of is entirely contained in rz.
Let T denote the p a r a m e t e r ( t i m e ) along the path (so as to distinguish
it f r o m the p a r a m e t e r t of the paths of s y s t e m ( A ) ) . The corresponding
motion along the path L" is described by the equations
.=cp(T), y=$(O. (24)
The section ~@fi of z

is generated a s T goes f r o m to to ?G), where_ 2 is
the value of the p a r a m e t e r on the arc 1, corresponding to the point M.
Because of the p a r t i c u l a r choice of 6 , t h i s section of L i s completely
contained in the s t a r t i n g quadrangle I'. Indeed, f o r any T , t , g T g G ( r ) ,
x=aI((t,s), y=Y((t,s).
where s and t a r e s o m e n u m b e r s satisfying the inequalities
si Qs<sz, to<t<T (s).
This evidently m e a n s that the equations
( D ( t ,s ) = G ( T ) , Y(t, s ) = $ ( T )
46
I
-
for any T f r o m the inte.-val [ t o , T G ) ]have a unique solution for f and s in the
region ( 2 5 ) . L V e w r i t e this solution in the form
t-t(T), s=s(T).
ds
Let u s calculate z. Differentiation of ( 2 6 ) with r e s p e c t to T g i v e s
The determinant of t h i s s y s t e m is
where
Since by assumption D,>O, w e have D<CJ and ( 2 7 ) is a C r a m e r s y s t e m .

Therefore,
But
Therefore
Using ( 2 3 ) and the inequality D = -Ill[ (0 we conclude that a t any point of

the section .G.y of the path L , i . e . , for all T,
_- dr
f , i T d T ( s ) , ,T>O, i.e.,
s ( T ) is an i n c r e a s i n g function of T . Since T = t , , s = s > s 0 , we conclude that
for t o < T - i ; ( y ) a l s o s ( T ) > s o . T h i s evidently m e a n s that the e n t i r e section
.h"T of t i e s inside the e l e m e n t a r y quadrangle r2, T h i s c o m p l e t e s the
proof.
The next l e m m a is concerned with a single a r c without contact, but the
paths a r e a s s u m e d t o meet it twice. Thus, l e t MI and 3 f Z be two i n t e r i o r
points of the a r c I ; s u p p c s e that any path of the s y s t e m ( A ) that m e e t s the
47
section M I M z of the a r c I a t t = to c r o s s e s the a r c 1 again a t s o m e

7 =7 (s) > to. At the intermediate values of t , to < t < 7 (s), these paths,
however, have no common points with the
arc 1.
Let the paths L, and Lz passing through the
points M 1and M z s respectively, c r o s s the
a r c 1 f o r the second time a t the points N l and
Nz, which a r e again different f r o m the end-
points of the a r c 1 (if & or Lo is a closed
path, M 1 and N l , or respectively M Z and N 2 ,
coincide). W e a s s u m e that the sections
M 1 M 2 and N I N Zof the a r c 1 have common
points, i.e., they i n t e r s e c t , since otherwise
w e could have reduced the t r e a t m e n t to the
FIGURE 13
previous c a s e of two distinct arcs without
contact I , and l a .
Let Ct ( i = 1,2) be a simple closed c u r v e
which coincides with the path L , if t h i s path is closed; otherwise, this
simple c u r v e is made up of a section M f N l of the path L f and a section M I N ,
of the a r c 1 ( F i g u r e 13; compared QT, 5 3 . 9 ) .
Let W be a region whose boundary is f o r m e d by two s i m p l e closed
c u r v e s C1 and GI. This region, together _with its boundary, is e n t i r e l y
contained in G*. As before, let s y s t e m (A) be 8-close in z* to s y s t e m (A)
and l e t 8 > 0 b, so s m a l l that the a r c 1 r e m a i n s an a r c without contact f o r
the paths of (A).
Let g1 be a point of the a r c I , sufficiently c l o s e to Ml,and li?, a point of
the a r c 1, sufficiently close to M a ; l e t z, and La be the paths of s y s t e m (A)
which a t t = t o p a s s through the points Ml and-hz, respectively. By L e m m a 5
we readily see that when the points I%?l and M a are sufficieGly c l o s e to Ml
and Mz, and 8 is sufficiently small, the paths L, and & a t zi and y2 (&>to
and 7 2 > t O ) , respectively, will again c r o s s the a r c 1 at points %, and w,.
The sections I%?,zl and gZH2 of the paths zland E2 have no common points
with the a r c I , except the crud points. L e t E, ( 1 = 1,2) b e a simple closed
curve which coincices with L; if t h i s is a closed path and otherwise is made
up f r o m a section M i x i of the path Et and a section @at of the a r c 1 . Let W
b e the region formed by the simple closed c u r v e s Cland E$.
L e m p a 1 1 . F m any e>O there exist 8 > 0 and q > O such that if
system (A) i s 6 -close in @ to system (A) and the points fii and I%, lie in
u, ( M , ) and u, ( M A respectively, then
(a) the points N,and N zlie in U. (N,) mLd U. (N2),respectively;
(b) each path passing through a point in fi crosses the m c 1 both fm
increasing and decreasing t ;
(c) iV i s e-close to W .
The validity of t h i s l e m m a is readily established by drawing an auxiliary
a r c without contact which partitions w into two elementary quadrangles,
to which Lemma 7 is then applied.
Note that f o r W and the propositions of L e m m a 8 i n g e n e r a l do not
hold t r u e , i.e., without additional assumptions regarding the exact nature
of the paths in W we cannot maintain that for any s m a l l q 5 n d 6 the p a r t i -
tions of W and @ by the t r a j e c t o r i e s of s y s t e m s (A) and (A) are e -identical.
48
In conclusion of this chapter w e will prove another lemma which i s
concerned with regions delimited by two c y c l e s without contact.
Let C, and C2 be two s i m p l e cl_osed c u r v e s , which a r e c y c l e s without
contact f o r the paths of s y s t e m ( A ) ; one of t h e s e c u r v e s , Cp say, lies
inside the other c u r v e . Suppose that e a c h path L of s y s t e m ( A ) passing a t
t = t,, through the point 111 of the cycle C, i n t e r s e c t s f o r s o m e T = T (W) > t o
the cycle CSa t the point .V, s o that the section J1.V of the path L h a s no
common points with the c y c l e s C1 and C2, other than i t s end-points
( F i g u r e 14). Let W be a closed annular region between the c y c l e s Cl and
C?, s o that 11. c C*.
L e v i wi a 12. For a,ay E > U there exists 6 > 0 such that if SyStepTi ( K )is
b-close to system ( A ) i n C*, then
(a) cycles C, and L'? are cycles without contact for the trajectories of
sgstevi (A);
(b) the partition of V by paths of system (A) i s E -identical to the
partitian 01this region $ 5 paths
~ of systerii(Al.
P r o o f . Draw any two paths which i n t e r s e c t the cycle without contact
C',.They c l e a r l y partition into two e l e m e n t a r y quadrangles. Applying
Lemma 8 to each of t h e s e quadrangles, w e readily verify Lemma 1 2 .
49
C h a p t e r III
THE SPACE OF DYNAMIC SYSTEMS AND

STRUCTURALLY STABLE SYSTEMS
INTRODUCTION
In t h i s chapter w e define the concepts of a s t r u c t u r a l l y s t a b l e

s y s t e m and a s t r u c t u r a l l y s t a b l e p a t h and establish s o m e of
t h e i r e l e m e n t a r y p r o p e r t i e s . The concept of a s t r u c t u r a l l y stable dynamic
s y s t e m is in fact the c o r n e r s t o n e of t h i s book. Exact definitions a r e given
in 56 (Definitions 10 and 12). Roughly speaking, we say that s y s t e m (A) is
structuraJly stable in s o m e two-dimensional r e g i o c W if a sufficiently c l o s e
s y s t e m (A) partitions W ( o r s o m e c l o s e region w) into paths in a manner
which is topologically identical to the partition of W by s y s t e m (A), and an
infinitesimal translation suffices to change o v e r f r o m one partition to the
o t h e r . It can be shown that s t r u c t u r a l l y stable s y s t e m s constitute, so to
say, a majority i n the s e t of all dynamic s y s t e m s . Indeed, a given
dynamic s y s t e m is s t r u c t u r a l l y stable a s a r u l e , and s t r u c t u r a l l y unstable
s y s t e m s are a n exception. Structurally stable s y s t e m s a r e of considerable
importance in the a n a l y s i s of physical problems.
Chapter 111, 55 is of introductory nature. I t s a i m is to define a m e t r i c
in the s e t of dynamic s y s t e m s in a plane region o r on a s p h e r e , s o as to
convert t h i s s e t into a m e t r i c space. The m e t r i c is introduced in the
natural, and apparently the s i m p l e s t , way. The a i m of the m e t r i c - s p a c e
approach to the s e t of dynamic s y s t e m s is to p e r m i t operating with
geometrical concepts, which a r e intrinsically l e s s a b s t r a c t in discussion.
In 6 s o m e basic definitions a r e introduced: the definition of a
s t r u c t u r a l l y stable s y s t e m in a plane region (56.1) and on a s p h e r e (56.2).
and the definition of relative s t r u c t u r a l stability.
In 57, s t r u c t u r a l l y s t a b l e and s t r u c t u r a l l y u n s t a b l e
p a t h s a r e defined. A path L of s y s t e m (A) is s a i d to be s t r u c t u r a l l y stable
i f s y s t e m (A) is s t r u c t u r a l l y stable in some neighborhood of L . It is
proved that if a s y s t e m is s t r u c t u r a l l y stable in s o m e region, a l l i t s paths
in that region a r e s t r u c t u r a l l y stable (Lemma 1). Therefore, if t h e r e is
a t l e a s t one s t r u c t u r a l l y unstable path in some region, the s y s t e m is
s t r u c t u r a l l y unstable in that region. It is proved (Theorem 10) that a
s t r u c t u r a l l y stable s y s t e m may have only a finite number of s t a t e s of
equilibrium in a closed region. Finally, the m u 1t i p 1i c i t y of a n
e q u i l i b r i u m s t a t e is defined and it is shown that a s t r u c t u r a l l y stable
equilibrium s t a t e M u ( z O yo) , is of necessity simple, i . e . ,
5438 50
$5. T H E SPACE OF DYNAMIC SYSTElIS
1. The s p a c e of dynam.ic s y s t e m s in a plane region
In this section we will o p e r a t e with m e t r i c s p a c e s whose points are

identified with dynamic s y s t e m s . The introduction of t h e s e s p a c e s lends
a m o r e graphic g e o m e t r i c f o r m to the fundamental concepts and a r g u m e n t s
of o u r theory.
We a r e concerned with dynamic s y s t e m s in a plane region o r on a
s p h e r e . In what follows, when dealing with dynamic s y s t e m s on a plane,
w e invariably a s s u m e that a l l the relevant s y s t e m s are defined in the s a m e
c
closed region of the plane. If n e c e s s a r y , w e %willa s s u m e , without
mentioning i t explicitly, that the s y s t e m s a r e a l s o defined in a l a r g e r open
domain (which contains E ) , but the a n a l y s i s w i l l always be confined to G.
IVe will often have to c o n s i d e r dynamic s y s t e m s (defined in a ) in s o m e
closed or open subregions of G . In t h i s case, w e always a s s u m e that the
c l o s u r e s of these subregions are entirely contained in G, i.e., they lie a t
a finite distance f r o m the boundary of 8.
Ali dynamic s y s t e m s of a given c l a s s k ( k is a fixed n a t u r a l n u m b e r ) or
of the analytical c l a s s i n will be t r e a t e d a s points of s o m e s p a c e . Let r
be a given n a t u r a l nunib,-.r, r ; : k . Let f u r t h e r .If, and Sfr be two points in
o u r space, i.e., two dynamic s y s t e m s
dr
dPI(^,
t Y). % = Q I ( ~ , Y), (&I I)
Consider the maximum of the absolute value of the dlfference

between the functions P , and P 2 in i.e., c,
mal t Pi (2.Y)-P3 (I, Y) I.
(S. uj;c
and also the maxima of the absolute values of the differences between
the corresponding derivatives of these functions to o r d e r r , inclusive, i.e.,
and s i m i l a r e x p r e s s i o n s f o r the functions Q, and Q2 and their derivatives,

i.e.,
mar_ I QI-Q?!
(x. Y E G (3)
and
The l a r g e s t of the n u m b e r s (1)--(4) is taken a s the distance between

the points MI and M Z in the s p a c e of dynamic s y s t e m s . All the fundamental
axioms of m e t r i c s p a c e s are readily verified.
The s p a c e of dynamic s y s t e m s of class k (or of analytical c l a s s ) with
the above maximum m e t r i c is designated R? (or R z ) , respectively). Clearly
51
Ch.111. T H E SPACE OF DYNAMIC SYSTFMS AND STRUCTURALLY STABLE SYSTEMS
the s e t of all dynamic s y s t e m s of c l a s s k which a r e 8 -close to rank r in

to the s y s t e m of c l a s s k
dX du
-
dt = Po (I. Y) 7 = Qo ( x , Y)
constitutes a 8-neighborhood of the point M O i n the space R r ) . A s i m i l a r

proposition is t r u e f o r points in the space Rh. F o r a fixed k , w e may
c l e a r l y consider the s p a c e s
RL, RL, ..., RLk.

They
- consist of the s a m e e l e m e n t s (dynamic s y s t e m s of c l a s s k defined in
G ) , but have different m e t r i c s . The 8 -neighborhoods defined by t h e s e
m e t r i c s a r e thus different. Indeed, let Ub (MI R ) denote the 8 -neighborhood
of the point M i n m e t r i c space R , and let 1<rl < r z < k . Consider a dynamic
s y s t e m Mo of c l a s s k and the neighborhoods
U.3 (Mo I R P ) ) and ub ( M o I RP)).

Clearly U a ( M o I R F ) ) c Ua ( M , I R F ) ) . but the r e v e r s e inclusion is not
always t r u e . The space RIi) wil b e designated Ri (without s u p e r s c r i p t ) . The
s p a c e Rl c l e a r l y contains each of the s p a c e s RV) ( k = 1, 2, ...
). In the c a s e
of dynamic s y s t e m s of the analytical c l a s s , we can consider an infinite
sequence of s p a c e s
El:), R:, ..., RZ), . . .

A s before, a l l t h e s e s p a c e s consist of the s a m e e l e m e n t s but have different
m e t r i c s . Let now k l < k z and choose a fixed r . Since any s y s t e m of c l a s s k.
is a l s o a s y s t e m of c l a s s k , , and an analytical s y s t e m is a s y s t e m of any
c l a s s , we c l e a r l y have
RP? 2 RC 2 RZ). (5 1
It is readily s e e n that any space in (5) is a subspace of all the preceding

s p a c e s i n the s e n s e that the distance between any two e l e m e n t s defined in
t h i s space coincides with the distance between t h e s e e l e m e n t s in the
enclosing space.
2. The s p a c e of dynamic s y s t e m s on a s p h e r e
Dynamic s y s t e m s on a s p h e r e a r e generally defined i n t e r m s of o p e n

c o v e r i n g s of the s p h e r e ( s e e QT, Appendix, 57.3). However, for
purposes of defining a m e t r i c in the s e t of dynamic s y s t e m s on a s p h e r e ,
it is b e t t e r to consider c 1 o s e d c o v e r i n g s .
W e will define these
closed coverings a s follows: let S be a s p h e r e (e.g., a s p h e r e in the
three-dimensional s p a c e la d e s c r i b e d by the equation a? + ys + z* = l), and
X = { G l , Gz, ..., GNJ
52
$5. THE SPACE OF DYNAh!IC SYSTEhIS
s o m e o p e n covering of the s p h e r e . If the c l o s u r e 5, of each G; (i=

= 1,2,. . ., N ) is homeomorphic to a closed region Hi of the plane ( ui, v i ) ,
we s a y that the covering
constitutes a c 1 o s e d covering of the s p h e r e S. Under t h i s definition,

each closed covering of the s p h e r e is related t o s o m e open covering, but
not e v e r y open covering c o r r e s p o n d s t o a closed covering. Thus, none of
the G , c a n be a pierced s p h e r e (i.e.* a s p h e r e with a single point removed).
Note, however, that t h i s r e s t r i c t i o n of the concept of a closed covering
does not r e s t r i c t the class of dynamic s y s t e m s being considered.
A dynamic s y s t e m on a s p h e r e is defined in t e r m s of closed coverings
precisely in the s a m e wa.y as i t is defined in t e r m s of open coverings
( s e e QT, 52.2). Consider a closed covering
- - -
= {GI, G?, . . ., &}
of the s p h e r e S . In each of the c,
w e define a local s y s t e m of coordinates
( I , . r c . To this end, w e consider a mapping of s o m e region p cof the plane
( u t , r L )onto the region G, of the s p h e r e S , defined by the equalities
2 = qt ( 4 ,%), y = *' ( U ' . r , ) , J = x , (u,,U J . (6)
The mapping ( 6 ) should satisfy the following conditions :

1 ) This is a topological mapping of the plane region E , onto z,;
2 ) the functions q', *,. x c are functions of class k + 1, if the dynamic
s y s t e m is of class k, and analytical, if the dynamic s y s t e m is analytical;
3 ) the functional determinants
do not vanish simultaneously anywhere i n E : .

A dynamic s y s t e m (A; of class k (or of the analytical class) on a s p h e r e
.
is defined by specifying in e a c h E?i (i = 1 , 2,. ., *V) a dynamic s y s t e m of
c l a s s k (or of the analytical class, respectively)
Moreover, in each region
the dynamic s y s t e m s ( A j )and (A,) are t r a n s f o r m e d into one another by the

s a m e transformation which t r a n s f o r m s the coordinates u,, v j into uk. vk (see
QT, 8 1 . 1 0 , and a l s o 2.2, Definition 1 ) .
A dynamic s y s t e m (A) defined in t h i s way is d e s c r i b e d in t e r m s of a
s y s t e m A' of local coordinates on a s p h e r e , which in i t s t u r n is d e s c r i b e d
by the covering and equations ( 6 ) . The s a m e dynamic s y s t e m (A) c a n
be specified using any o t h e r s y s t e m K* of local coordinates defined by the
53
Ch.111. THE SPACE OF DYNAMIC SYSTEMS AND STRUCTURALLY STABLE SYSTEMS
closed covering
-z*= {GF, e,.. ., Gif..)
and equations analogous to ( 6 ) ( s e e QT, 52.2).
To define a m e t r i c in the s e t of dynamic s y s t e m s on a s p h e r e and thus
convert this s e t into a m e t r i c space, we will d e s c r i b e a l l the dynamic
s y s t e m s using one fixed s y s t e m K of local coordinates on a s p h e r e .
Suppose t h i s s y s t e m is defined by a closed covering = . . ., 8,) and a {c,,
s e t of equations ( 6 ) . Consider two dynamic s y s t e m s (A) and (A) of c l a s s k
(or of the analytical c l a s s ) . Let s y s t e m (A) be described by the equations
au(= du
dt ut ( w r 01). +=vi (ut. ut) (A,)
( i = 1, 2, . . ., N ) and s y s t e m (A) by the equations
(A,)
Choose a natural number r > l , such that r < k , i f k is the c l a s s of the
s y s t e m (for analytical s y s t e m s , any r can be chosen). Consider the
numbers
max lUi(U1, ~ t ) - B i ( U i . U 1 ) I (6=& 2, ..., N ) ,

(Ui. V i ) ii,
and a l s o the n u m b e r s
as w e l l a s s i m i l a r e x p r e s s i o n s f o r the functions V i and 8, and t h e i r p a r t i a l

derivatives to o r d e r r , inclusive. The l a r g e s t of all these n z m b e r s is
taken a s the distance between the dynamic s y s t e m s (A) and (A). All the
axioms of a m e t r i c space a r e readily s e e n to hold t r u e .
The space of dynamic s y s t e m s of c l a s s k (or of the analytical c l a s s ) on
a s p h e r e with t h i s maximum m e t r i c is designated RI" (or I?:)). The r e m a r k s
a t the end of 95.1 pertaining to analogous s p a c e s of dynamic s y s t e m s on a
plane are applicable to the s p a c e s R f ) and R r ) of dynamic s y s t e m s on a
sphere.
Two dynamic s y s t e m s (A) and (A) of c l a s s k (or the analytical c l a s s ) on
a s p h e r e a r e said to be 8-close to r a n k r ( r g k ) , if the distance between
them in space R g ) (R:)) is l e s s than 6 .
Remark .
The m e t r i c defined in the set of dynamic s y s t e m s essentially
depends on the fixed s y s t e m K of local coordinates-on the s p h e r e . The
distances between the dynamic s y s t e m s (A) and (A) defined in this way
using different s y s t e m s of local coordinates K and X* a r e in g e n e r a l
different. It can b e seen, however, that the m e t r i c s defined by v a r i o u s
s y s t e m s of local coordinates a r e all equivalent, i.e., they induce the s a m e
topology i n the space of dynamic systems.* This is obvious f r o m the
following proposition: f o r any S > 0, t h e r e e x i s t s 8* > 0 such that any
dynamic s y s t e m (A) which is 8*-close to s y s t e m (A) i n the m e t r i c defined
I * See A l e k s a n d r o v . P . S . C o m b i n a t o r i a l T o p o l o ~ .Chapter 1, $2:3.

(.6. STRC'CTL'RALLY STABLE DYK.4hlLC SYSTEhlS
by the local coordinate s y s t e m K* is 6-close to (A) in the m e t r i c defined

by the coordinate s y s t e m K.
This proposition can be proved without difficulty, and is left a s a n
e x e r c i s e to the r e a d e r .
$6.D E F I N I T I O N OF A S T R U C T U R A L L Y STABLE
DYNAhIIC SYSTE3.I
1. Dynamic s y s t e m s on a plane
Let
be a dynamic s y s t e m defined i n a bounded closed region 6, and W a closed

or open subregion of G."
D e f i n it i o n 10. il dynamic system ( 4 ) i s said to be structurally
stable in 11- c G if there exists an open domain H containing lr,
-
wcH C R c G,
which satisfies the fo_lloulingcondition: for any E > a , there exists 6 ==. 0
such that if system (A) is 6-close to system (A) in 8, one can find a region
f o r which
( s e e $4.1, Definition 9).

i f s y s t e m ( A ) is not s t r u c t u r a l l y stable in region W , i t is said to be
s t r u c t u r a l l y u n s t a b l e i n t h a t region.
Evidently a dynamic s y s t e m (A) is s t r u c t u r a l l y unstable in W i f f o r any
H . 1V c H c R c G , t h e r e e x i s t s E,! > 0 with the following property: f o r any
6 > 0 and any H , t h e r e e x i s t s a-system (A) 6 -close to (A) such that the
partition of by the paths of (A) is not e,-identical to the partition of H by
the paths of (A).
It follows f r o m Definition 10 that if s y s t e m (A) is s t r u c t u r a l l y stable in
iL', the topological s t r u c t u r e of the partition of s o m e neighborhood fl of W' by
the paths of s y s t e m (A) doe2 not change in a c e r t a i n s e n s e on passing to a
sufficiently c l o s e s y s t e m (A), or m o r e precisely, a n infinitesimal t r a n s l a -
tion-will t r a n s f o r m H into s o that the paths of (A) coincide with the paths
of (A). This property explains the t e r m s t r u c t u r a l l y s t a b l e s y s t e m .
An alternative t e r m used in the Russian l i t e r a t u r e is a c o a r s e s y s t e m ,
which i m p l i e s that the topological s t r u c t u r e of the partition of a given
region by paths is not affected by s m a l l changes in s y s t e m (A) or, i n other
words, the s t r u c t u r e can r e s i s t s m a l l disturbances i n s y s t e m (A).
Examples of s t r u c t u r a l l y stable s y s t e m s are considered a t a l a t e r stage.
H e r e we will analyze a s t r u c t u r a l l y unstable s y s t e m .
' As we have noted in the previsu; section, it is implicitly assumed that the closures of the relevant sub-
regionj a r e entirely contained i n G, i.e., they 3re at a finite Jistance from t h e boundary of C.
55
Ch.111. THE SPACE OF DYNAMIC SYSTEMS AND STRUCTURALLY STABLE SYSTFMS
E x a m p 1e 3 . Consider the s y s t e m
It is defined on the e n t i r e plane, so that any closed region on the plane

can be taken a s 6. Let b e the region defined by the inequality
xe + y* 100.
W is defined by the inequality
ts + y* Q 16. (3)
We w i l l now show that s y s t e m ( 2 ) is s t r u c t u r a l l y unstable in region (3).

Define H a s
2 + g 25. (4)
Any positive number can b e taken a s e,. The paths of s y s t e m ( 2 ) a r e the

point 0 (0, 0) and the c i r c l e s
For s m a l l y, s y s t e m (5) is a r b i t r a r i l y c l o s e in t o s y s t e m (2), the

point 0 (0,O) is the focus of s y s t e m (5), and all the other paths of this system
a r e s p i r a l s ( s e e QT- $1.14,
R e g a r d l e s s of what H we choose, i t s partition
by the paths of s y s t e m (5) c l e a r l y cannot be
e,-identical to the partition of region (4) by the
paths of s y s t e m (2). Indeed, all the paths of
s y s t e m ( 2 ) i n region (4), except the equilibrium
point 0, a r e closed, w h e r e a s s y s t e m (5) h a s no
.-, - " , . _
l i s h e s the s t r u c t u r a l instability of s y s t e m ( 2 )
in region (3).
Before proceeding any f u r t h e r with our
analysis, we would like to offer s o m e back-
FIGURE 15 ground information on s t r u c t u r a l l y stable
s y s t e m s . Structurally stable s y s t e m s w e r e
f i r s t considered i n 1937 by A. A. Andronov and
L. S. Pontryagin 141, who originally called them s y s t B m e s g r o s s i e r
or c o a r s e s y s t e m s .
They considered, however, dynamic s y s t e m s
in a particular region W lying inside a cycle without contact. The definition
56
$6. 5fRCiCTURALLY ST.ABLE DYh.4.L.K SYS I ELIS
o f a s t r u c t u r a l l y stable s y s t e m in such a region is v e r y s i m p l e . Indeed,

s y s t e m (A) defined in a region It lying inside a cycle without contact ?! is
said to be s t r u c t u r a l l y stable in_that region if f o r any E > 0 t h e r e e x i s t s
5 > t) such that f o r any s y s t e m ( A ) 6 -close to s y s t e m (A)
e I
(if-, d ) E (W,
A)
Substantial simplificati2n is achieved here because u-e <o not have to

consider e i t h e r the region H 3 I V or the corresponding H : the e n t i r e
a n a l y s i s is confined to If. It can be shown that in region If- lying inside a
cycle without contact r both definitions - Definition 10 and the definition
in 1 4 j - a r e equivalent. Unfortunately, in c a s e of a g e n e r a l region It., a
s t r u c t u r a l l y stable s y s t e m cannot be defined without introducing a n
auxiliary region II ( s e e Appendix, subsection 5).
Structurally stable s y s t e m s (in a given region) are in a s e n s e the
simplest dynamic s y s t e m s , just as the s i m p l e r o o t s of a function can be
regarded a s the most e l e m e n t a r y among a l l the r o o t s or s i m p l e (nonmulti-
ple) intersection points of two c u r v e s a r e the most e l e m e n t a r y among a l l
the intersection points. T h e s e t h r e e concepts - a simple root of a function,
a simple intersection point of two c u r v e s , and a s t r u c t u r a l l y stable s y s t e m
in a c e r t a i n region - are analogous in the s e n s e that u n d e r s m a 11
d i s t u r b a n c e s (of the function, the p a i r of c u r v e s , or the dynamic
s y s t e m ) the object r e m a i n s intrinsically unaffected and only a s m a l l t r a n s -
lation or shift is obserced ( s e e r e m a r k to Theorem 5, I > R e m a r k 3 to
Theorem 6 , S2, and the definition of a s t r u c t u r a l l y stable s y s t e m ) .
In what follows (Chapter VI, 18.2, Theorem 2 3 ) w e w i l l d e r i v e the
n e c e s s a r y and sufficient conditions for a dynamic s y s t e m to be s t r u c t u r a l l y
stable in a given region It. T h e s e conditions, like the conditions of
simplicity of a root or simplicity of a n intersection point of two c u r v e s ,
a r e analytically e x p r e s s e d in the form of inequalities between c e r t a i n
quantities which a r e continuous functions of the right-hand s i d e s of the
dynamic s y s t e m . Hence i t follows that the s y s t e m s which a r e structurally
stable in 1). ( s e e $5) f o r m a n open s e t i n the m e t r i c s p a c e R,.
The definition of a s t r u c t u r a l l y s t a b l e s y s t e m using the m e t r i c s p a c e R 1
can naturally be formulated as follows:
The systern (A) cmesponding to a point Jf E R 1is said to be stricctzivaliy
stable in W i f there exists a region H , fi: c H c R c G , such that the
Jolloicing condition is satisfied: for any E > 0 we can choose 6 > 0 such that
if 0E U6 (;%I),the follouvng velation holds true f o r fhe system (A) c m r e -
sponding to the point A? ~ n some
d -region2 E G:
, , e
= ( H , A).
( H , .i)
Note that i t does not follow directly f r o m the definition of a s t r u c t u r a l l y

stable s y s t e m that s t r u c t u r a l l y stable s y s t e m s (in W ) f o r m a n open subset in
R I . T h i s proposition follows, as u-e have r e m a r k e d above, f r o m the
analytical conditions of s t r u c t u r a l stability.
W e conclude this section with two simple, but highly important l e m m a s .
The proof is self-evident and is therefore omitted.
L e m >ti a 1. If systmz (A) i s structurally stable in I!, it is structtrvally
stable in any subregion it-, of it.
'3.111. THE SPACE OF DYNAMIC SYSTEMS AND STRUCTURALLY STABLE SYSTEMS
The next lemma deals with substitution of v a r i a b l e s . Consider a r e g u l a r

mapping of c l a s s 2,
u='p("* Y), u=$,(+, Y). (6)
defined in 6,which maps this region into s o m e region @ i n the plane (u,u).
is t r a n s f o r m e d by this mapping into some s y s t e m
defined in c* ( s e e 13.2). Let W be some subregion of G and W* the image
Lemma 2 follows f r o m the uniform continuity of the functions cp and q.

Equations ( 6 ) can be considered a s defining a c e r t a i n substitution of
v a r i a b l e s . L e m m a 2 thus signifies that the s t r u c t u r a l stability of a dynamic
s y s t e m is a property which is invariant under a substitution of v a r i a b l e s of
c l a s s 2.
2. Structurally stable s y s t e m s on a s p h e r e
The only c a s e of i n t e r e s t among dynamic s y s t e m s on a s p h e r e is that

when the s y s t e m is defined on the e n t i r e s p h e r e . Indeed, a dynamic
s y s t e m defined i n a region whose c l o s u r e does not coincide with the e n t i r e
s p h e r e evidently can be considered a s a s y s t e m defined in a plane region
( s e e QT, 1 2 . 2 , r e m a r k following equation (10)).
The definition of s t r u c t u r a l stability of a dynamic s y s t e m on a s p h e r e ,
on the one hand, is substantially s i m p l e r than the analogous definition on
a plane, since one does not have to consider the auxiliary region H.': On
the other hand, the a n a l y s i s on a s p h e r e involves c e r t a i n difficulties,
since the m e t r i c defined in the s p a c e of dynamic s y s t e m s on a s p h e r e
depends on the p a r t i c u l a r s y s t e m of local coordinates chosen on the s p h e r e
( s e e r e m a r k a t the end of 55.2). This, however, does not constitute a
fundamental difficulty.
F i r s t let u s define the concept of e-identity on a sphere, analogous to
Definition 9.
D e f i n i t i o n 1 1 . Let ( D ) a n d ( 6 ) b e t w o d y n a m i c s y s t e m s d e f i n e d o n a
sphere S . The partition of the sphere S b y the paths of system (D) i s said
to be e-identical to the partition b y the paths of system (6), m in symbols
Also see the remark in the previous subsection concerning the definition of structural stability of a system
inside a cycle wirhout contacr.
When speaking of a n &-translation, w e naturally a s s u m e that s o m e
m e t r i c is defined on the s p h e r e , e i t h e r internal or induced by the enclosing
euclidean space.
To define a s t r u c t u r a l l y stable s y s t e m , w e introduce s o m e s y s t e m of
local coordinates on the s p h e r e and m e a s u r e the distances between two
dynamic s y s t e m s r e l a t i v e to this coordinate s y s t e m ( s e e 15.2).
D e f i n i t i o n 12. A dynamic system ( D ) on a sphere S is said to be
structzcrally stable if f o r any e > 0 there exists 6 > 0 such that, for any
system ( 6 ) ,6 -close to system ( D ) , we hace
(S, D ) (S, 5)
Otherwise s y s t e m (D) is said to be s t r u c t u r a l l y u n s t a b l e .
If a dynamic s y s t e m (D) on a s p h e r e is s t r u c t u r a l l y unstable, t h e r e
e x i s t s eo >,0 with the fo:llowing property: f o r any 6 > 0 t h e r e e x i s t s a
s y s t e m (D) 6-close to (D), such that partitions of the s p h e r e ( S )by the
paths of (D) and ( 6 )are n o t eo-identical.
Definition 1 2 m a k e s u s e of a particular s y s t e m of local coordinates K on
the s p h e r e . W e now have to establish that the concept of a s t r u c t u r a l l y
stable s y s t e m is in fact independent of the choice of the local s y s t e m of
coordinates on the s p h e r e , or in other words, we have to show that i f
s y s t e m (D) is s t r u c t u r a l l y stable when the m e t r i c (in the s p a c e of dynamic
s y s t e m s ) is defined in t e r m s of s o m e s y s t e m of local coordinates K , it is
a l s o s t r u c t u r a l l y stable in the m e t r i c defined using any other s y s t e m of
local coordinates K *. This proposition c l e a r l y follows f r o m a previous
proposition formulated a t the v e r y eqd of 1 5 (in the r e m a r k to 55.2).
Definition 1 2 is thus meaningful, and s t r u c t u r a l l y stable s y s t e m s on a
s p h e r e c a n be analyzed in a m e t r i c introduced using a n a r b i t r a r y fixed
s y s t e m of local coordinates.
Let (A) be a dynamic: s y s t e m on a s p h e r e and W s o m e subregion of the
s p h e r e whose c l o s u r e F: d o e s not coincide with the e n t i r e s p h e r e . We
choose a s y s t e m K of local coordinates so that a t l e a s t one of the compo-
nents of the closed covering corresponding to this s y s t e m , Gl say,
entirely contains ff', i.e., rf'c GI. F r o m the definition of the s y s t e m of
local coordinates and or' a dynamic s y s t e m on a s p h e r e ( s e e 15.2) w e see
that c o r r e s p o n d s to s o m e region Fl on the plane ( u l , cI). Let W -c o r r e -
spond to s o m e W* on the plane, iT* c Hi. System (A) on the s p h e r e thus
c o r r e s p o n d s to s o m e dynamic s y s t e m (Al) in gl.
L e m m a 3 . If ( A ) I S a structurally stable system on a sphere, the
dynainic system ( A l ) is shctzcrally stable in w*.
This lemma follows directly f r o m the definitions of s t r u c t u r a l stability
(Definitions 10 and 1 2 ) and the uniform continuity of the mapping of G , onto
Hi.
In s i m p l e r language, Lemma 3 c a n be formulated a s follows: if
s y s t e m ( D ) is s t r u c t u r a l l y stable on a s p h e r e , i t is s t r u c t u r a l l y stable in
any subregion of the s p h e r e .
3. S t r u c t u r a l stability of dynamic s y s t e m s in Rg' and Rg)
In the above definitions of s t r u c t u r a l l y s t a b l e and s t r u c t u r a l l y unstable

dynamic s y s t e m s , ( A ) and (D) w e r e a s s u m e d t o be s y s t e m s of c l a s s 1, and
59
we have considered all s y s t e m s 8-close to (A) and (D) to r a n k 1. In other

words, we considered points in the space R 1with t h e i r r e s p e c t i v e 8-neigh-
borhoods ( s e e 5). In s o m e c a s e s , however, w e are not i n t e r e s t e d in a l l
the possible s y s t e m s of c l a s s 1, but in the s y s t e m s of s o m e n a r r o w e r
c l a s s , e.g., analytical s y s t e m s o r s y s t e m s defined on a plane whose right-
hand s i d e s a r e polynomials.
Moreover, s o m e t i m e s w e have to consider 8 - c l o s e n e s s to s o m e higher
r a n k , and not only to rank 1. This evidently m e a n s that the a n a l y s i s is not
confined to R l , but to some other space RI" o r RY'. We thus naturally a r r i v e
a t the concept of r e l a t i v e s t r u c t u r a l s t a b i l i t y of a dynamic
s y s t e m , i.e., s t r u c t u r a l stability r e l a t i v e to s o m e space i n which the given
dynamic s y s t e m is a point.* The corresponding definitions a r e entirely
analogous to Definitions 10 and 1 2 , and we will t h e r e f o r e give only the
definition of a s t r u c t u r a l l y stable dynamic s y s t e m on a plane r e l a t i v e to RY).
As before, we consider s y s t e m s defined i n a bounded plane region.
De f i n i t i o n 13. A dynamic system ( A )of class k is said to be
structurally stable in W relative to R r ) (wc '@ c G ; r 6 k ) i f there exists an
open domain H containing w, a
c H c c G , which satisfies the
following condition: f o r any e > O one can choose 8 > 0 such that if (A) is
a system of class k 8-close to rank r to ( A )in there exists B f m which
- - . e
(H,A ) = ( H . A )
(also s e e Definition 10).
Otherwise, s y s t e m (A) of c l a s s k i s said to be s t r u c t u r a l l y u n s t a -
b l e i n W r e l a t i v e t o Rr).
Using geometrical terminology, we s a y that a dynamic s y s t e m (A)
corresponding to point M i n Rr' is s t r u c t u r a l l y stable in W r e l a t i v e to t h i s
space if t h e r e e x i s t s a domain H , wc
X c g c G, satisfying the following
condition: f o r any e 2 0 we can choose 8 > 0 such that if a
E Ua (M I Rr)),
then f o r the s y s t e m (A) corresponding to the point fi and s o m e H" c G we
have
(B,A) (H, A).
"Simple" s t r u c t u r a l stability (in the s e n s e of Definition 1 0 ) is c l e a r l y

s t r u c t u r a l stability r e l a t i v e to the space R l .
Let kt and kz be n a t u r a l numbers, ki < k z , (A) a dynamic s y s t e m of c l a s s
kz (and thus a l s o of class k , ) , and r a n a t u r a l number, r < k , . System (A)
then belongs to both RK) and RG). We r e c a l l that RG) c RK). It c l e a r l y
follows f r o m Definition 13 that if s y s t e m (A) is s t r u c t u r a l l y stable in W
r e l a t i v e to RK), it is a l s o s t r u c t u r a l l y stable r e l a t i v e to RK). Similarly, if
(A) is an analytical s y s t e m , then f o r any n a t u r a l k and r 6 k s y s t e m (A)
belongs both to RP)and to R!? c RJ". If (A) is s t r u c t u r a l l y stable in W
relative to RY), it is a l s o s t r u c t u r a l l y stable r e l a t i v e to R?. All t h e s e are
p a r t i c u l a r c a s e s of the following g e n e r a l proposition, which follows directly
f r o m the definition of s t r u c t u r a l stability: i f a dynamic s y s t e m (A) belongs
to two s p a c e s one of which is a subspace of the other and if the s y s t e m is
s t r u c t u r a l l y stable i n W r e l a t i v e to the enclosing space, it is s t r u c t u r a l l y
stable in W r e l a t i v e to the enclosed space.
* Instead we can speak of structural stability relative to a given class of dynamic systems. Ext then it should
be explicitly stared what w e mean by &closeness (i.e., to what rank).
60
56. SLRLCTLRALLY STABLE DYhALIIC SYSTmIS
Let f u r t h e r r, c= r2 ::h
~ (A) is a dynamic s y s t e m of c l a s s k, 111 is the
point corresponding to (A) in E: (or i n RP; w e r e c a l l that both t h e s e
s p a c e s a r e made up of the s a m e points, but the corresponding m e t r i c s a r e
different, see 15.1). Then, i f s y s t e m (A) i s s t r u c t u r a l l y s t a b l e
i n I i r e l a t i v e t o Ryl), i t i s a l s o s t r u c t u r a l l y s t a b l e i n W
r e 1 a t i v e t o E;?. T h i s follows directly f r o m Definition 13 and f r o m the
relation
1.6 (.If 1 RF) c Lb ( L\f1 R:)
( s e e $5.1).
I s the r e v e r s e a l s o t r u e ? In other words, c a n we maintain that i f
s y s t e m (A) is s t r u c t u r d l y stable in If- relative t o R;* i t is a l s o s t r u c t u r a l l y
-=
stable r e l a t i v e to Rh: ( r g k , k2)or, alternatively, if s y s t e m (A) is s t r u c -
turally stable in ti relative to Rh,i t i s a l s o s t r u c t u r a l l y s t a b l e r e l a t i v e to
,;Ill ( r , < r? .<k) ? T h e s e propositions c l e a r l y do not follow directly f r o m the
definition of relative s t r u c t u r a l stability. Moreover, i f s y s t e m (A) belongs
to two s p a c e s R a n d R*, s u c h that R* c R, then in g e n e r a l s y s t e m (A) may
prove s t r u c t u r a l l y s t a b l e in W relative to the enclosed s p a c e R*, w h e r e a s i t
i s s t r u c t u r a l l y unstable r e l a t i v e to the enclosing s p a c e R .
However, i f a t the c o s t of generality we concentrate on the s p a c e s RY
and Ra, which are of the main i n t e r e s t in the a n a l y s i s of dynamic s y s t e m s ,
the situation is considerably simplified. It c a n be shown that i f s y s t e m (A)
belongs to one of these s p a c e s , the n e c e s s a r y and sufficient conditions f o r
i t s s t r u c t u r a l stability (in W )relative to these s p a c e s are a t the s a m e time
the n e c e s s a r y and sufficient conditions of i t s s i m p l e s t r u c t u r a l stability
(i.e.> s t r u c t u r a l stability r e l a t i v e to R , ) . T h i s will be established in the
derivation of the n e c e s s a r y and sufficient conditions of s t r u c t u r a l stability
(518.4, R e m a r k d); additional proof (in connection with RF or R!:) is
naturally r e q u i r e d only for the n e c e s s a r y conditions of s t r u c t u r a l stability,
Thus, if s y s t e m (A) belongs t o RF or Rr and is s t r u c t u r a l l y s t a b l e in IF
relative to the corresponding space, i t is simply s t r u c t u r a l l y stable.
Therefore, w e do not have to consider s t r u c t u r a l stability r e l a t i v e to these
s p a c e s and i n what follows w e can concentrate on s t r u c t u r a l stability in the
s e n s e of Definition 10.
The situation is c l e a r l y the s a m e for dynamic s y s t e m s on a s p h e r e .
The above a r g u m e n t s pertaining to dynamic s y s t e m s are analogous to the
v a r i o u s a r g u m e n t s offered in 51.4 in connection with r o o t s of functions and
a t the end of $2 in the a n a l y s i s of intersection points of two c u r v e s . T h i s
analogy between dynamic s y s t e m s , on the one hand, and functions or p a i r s
of c u r v e s , on the o t h e r , unfortunately b r e a k s down a t one significant point.
Let u s d i s c u s s t h i s a s p e c t in s o m e detail.
W e will consider the analogy between dynamic s y s t e m s i n a plane region
G and p a i r s of c u r v e s F 1(z, y ) = O , F? (z, y) =O. At the end of S2 we considered
the multiplicity of the i i t e r s e c t i o n point of two c u r v e s r e l a t i v e to a given
c l a s s Bl of functions. %R w a s identified in p a r t i c u l a r with the class a,,of
all polynomials in two \,ariables not higher than of d e g r e e n and the n e c e s s a r y
and sufficient condition of s t r u c t u r a l stability of the intersection point of two
c u r v e s relative to t h i s ,class was established t o coincide (A # 0) with the
condition of s t r u c t u r a l stability in the s e n s e of Definition 5 (S2.1).
61
ch.111. THE SPACE OF D Y N A M I C SYSTEMS A N D STRUCTURALLY STABLE SYSTEMS
By analogy with the c l a s s of functions Sn, we can consider the c l a s s of

dynamic s y s t e m s (on a plane) whose right-hand s i d e s a r e polynomials not
higher than of d e g r e e n . We w i l l designate this c l a s s of s y s t e m s by 8,.
The s t r u c t u r a l stability r e l a t i v e to PI, is defined in the u s u a l way (6-close-
n e s s should be considered to r a n k 1 or to higher rank). The analogy
between the p a i r s of c u r v e s in \a,, and the dynamic s y s t e m s in %, is unfor-
tunately incomplete. The point is that the a t t e m p t s to d e r i v e the n e c e s s a r y
and sufficient conditions of s t r u c t u r a l stability of dynamic s y s t e m s r e l a t i v e
to the c l a s s 8, so far have remained unsuccessful. On the other hand, we
do not know of any p a r t i c u l a r example of a s y s t e m
( Pand Q a r e polynomials of not higher than n-th d e g r e e ) which is s t r u c t u r -

ally stable in s o m e region r e l a t i v e to the c l a s s B1, and s t r u c t u r a l l y unstable
in the s e n s e of Definition 10. The question of the possible existence of
these s y s t e m s and of the n e c e s s a r y and sufficient conditions of s t r u c t u r a l
stability r e l a t i v e to the c l a s s et, thus r e m a i n s open a t t h i s stage.
17. STRUCTURALLY STABLE AND STRUCTURALLY

UNSTABLE PATHS. NECESSARY CONDITION OF
STRUCTURAL STABILITY OF AN EQUILIBRIUM
1. Structurally s t a b l e and s t r u c t u r a l l y unstable paths
O u r immediate problem is the derivation of n e c e s s a r y and sufficient

conditions of s t r u c t u r a l stability of dynamic s y s t e m s on a plane and on a
s p h e r e . This problem is discussed in t h i s section and a l s o in Chapters IV,
V, and VI. The concepts of s t r u c t u r a l l y s t a b l e and s t r u c t u r a l l y
u n s t a b 1e p a t h s introduced in this section considerably simplify the
approach to our problem. It is a s s u m e d throughout t h i s chapter that a l l
the s y s t e m s a r e dynamic s y s t e m s of f i r s t c l a s s defined in a fixed plane
region 8, and 6 - c l o s e n e s s is always interpreted as c l o s e n e s s to r a n k 1.
In other words, the analysis is confined to R i . When dealing with dynamic
s y s t e m s on a s p h e r e , t h i s is not stated explicitly.
Let (A) be a dynamic s y s t e m which is s t r u c t u r a l l y stable i n a closed
o r an open region W , and L s o m e complete path of s y s t e m (A). F r o m the
definition of s t r u c t u r a l stability and L e m m a 1, 6 we s e e that if L is
entirely contained in W , then t h e r e is a c e r t a i n neighborhood V of L where
the s y s t e m (A) is s t r u c t u r a l l y stable. Moreover, c V.*
The r e v e r s e , naturally, is not always t r u e : s y s t e m (A) can be s t r u c -
turally stable in some neighborhood V of L , V c W , while it is s t r u c t u r a l l y
unstable in W.
These considerations lead to the concept of s t r u c t u r a 1 l y s t a b 1e
and s t r u c t u r a l l y u n s t a b l e p a t h s .
D e f i n i t i o n 1 4 . A path L of a dynamic system (A) is said to be
structurally stable if there exists a certain neighborhood V , Z c V c 7 c G,
where system ( A ) is structurally stable. Otherwise, L i s said to be
structurally unstable.
requirements.
!7. S r R l . C r l ~ l U L L Y ST.-\RLE .%hU CLSI.4dLE PATHS
According to Definition 14, to establish s t r u c t u r a l stability of a path L it

suffices to show that systeni (A) is s t r u c t u r a l l y s t a b l e in s o m e neighbor-
hood t-, c V . To establish s t r u c t u r a l instability, on the other hand, w e
have to show that s y s t e m (-4) is s t r u c t u r a l l y unstable in any region which
contains L . For this it is n e c e s s a r y and sufficient that s y s t e m (A) be
s t r u c t u r a l l y unstable in any a r b i t r a r i l y s m a l l neighborhood of L . LVe thus
a r r i v e a t the following n e c e s s a r y a n d s u f f i c i e n t c o n d i t i o n o f
s t r u c t u r a l i n s t a b i l i t y of a p a t h .
A path L is structur&lly unstable i f f o r any E > 0 t h e r e is a neighborhood
i. zc I c i= c c,(L) where s y s t e m (A) is s t r u c t u r a l l y unstable.
L e rn ni a 1. 1. system ( A ) i s stmctuvally stable in If, any path of this
system which is erttise1,y contained in It- i s stviicturally stablc.
T h i s follows directly f r o m Lemma 1, $6 and f r o m Definitions 10 and 14.
-4s w e have s e e n above, the proposition contained in this l e m m a w a s in fact
the r e a s o n f o r introdacing the concept o f a s t r u c t u r a 1 l y s t a b 1 e path.
By Lemma 1, i f t h e r e is a t least one s t r u c t u r a l l y unstable path of
s y s t e m ( A ) in W, the s y s t e m is s t r u c t u r a l l y unstable in It. The next
question to a s k is w h e t h e r a b s e n c e o f s t r u c t u r a l l y u n s t a b l e
p a t h s i n It n e c e s s a r i l y l e a d s t o s t r u c t u r a l s t a b i l i t y o f
t h e s y s t e m a s a w h o 1e i n It. Only the potentially limiting paths
are of importance in connection with this problem. A s is known (QT, 4.6
and 1 5 . 6 ) , these paths include
1 ) equilibrium s t a t e s , 2 ) closed paths, 3 ) paths which are a t the
s a m e t i m e a - and o - s e p a r a t r i c e s , i.e., a - and w-orbitally-unstable paths
approaching the equilibrium s t a t e s .
We will now establish which of t h e s e t h r e e types of paths a r e s t r u c t u r a l l y
stable.
2. Finite number of equilibrium s t a t e s i n a s t r u c t u r a l l y

stable s y s t e m
Let
-
be a dynamic s y s t e m defined in G , and l e t be a closed region, tV c G .
T h e o Y e ni 10. I f systeni ( A ) i s structrivallg stable in F, it has only
a finite nunrbev of equilibrirm states in SS;.
P r o o f . \Ve will f i r s t show that f o r any ( A ) and any 6 > 0, t h e r e e x i s t s
a dynamic s y s t e m (A) 3 - c l o s e t o ( A ) which h a s only a finite number of
equilibrium s t a t e s . C i o o s e s o m e S > 0. F r o m W e i e r s t r a s s s theorem
&
( 1.1, Theorem 1) t h e r e exist two polynomials P* (z, y). Q* (z, y). which are z-
c l o s e in 6 to the f u n c t i m s P (z, y),-Q (2,y).
If P* and Q* are irreducible, (A) c a n be chosen a s the s y s t e m
63
6
Indeed, t h i s s y s t e m is Z-close, and hence a l s o 6-close, to s y s t e m (A).
Its equilibrium s t a t e s a r e determined f r o m the s e t of equations
P* ( x , y) = 0, Q* (I,Y) = 0,
and since the l a r g e s t common divisor (P*,Q*)= 1, t h e s e equations have only
a finite number of solutions according t o the Bkzout theorem (see /12/,
Chapter 111, 53.1).
Now suppose that P* and Q* a r e not i r r e d u c i b l e . They can be written in
the f o r m
P'(5, Y) =Pi (x, SI)R (19 4,

Q' (5, Y) = Qi (2s SI)R (2, Y),
where R ( x , y) are polynomials of higher than z e r o degree, and Pi and Q, are

i r r e d u c i b l e , ( P I , Qj)= 1.
(I,Y) = Pi (G Y) [ R(z, Y) + a
19
Si (5, Y) = Qi ( ~ 7Y) tR (2, Y) +PI*

where a and j3 a r e r e a l n u m b e r s satisfying the following conditions:
1) a and j3 a r e sufficiently s m a l l , 2) aZj3. 3 ) the polynomial P,(z,g)
is i r r e d u c i b l e with R + B , and the polynomial Q, is i r r e d u c i b l e with R+a.
F i r s t w e have to show that such real n u m b e r s always e x i s t . Let
..
PI(G y) = P I ( x , Y) PZ (G y) . p . ( x , y) be the factorization of the polynomial Pi
into i r r e d u c i b l e f a c t o r s . Consider the polynomials
R (2,Y) +Si$ R ( G Y)+BZ~ * * *) R(G v ) + LR(G V) + & + i t (1)

where B r a r e any sufficiently s m a l l different numbers. Suppose that none
of the polynomials in (1) is i r r e d u c i b l e with P, (5, y). Then each of these
polynomials is reducible a t least by one of the polynomials p , ( x , y), ..., p . ( x , g).
Since the number of polynomials in (1) is one higher than the number of
polynomials p I ( 5 , y). a t l e a s t two polynomials i n (l), R (2.y) f l k and +
R ( 5 , y) + P I , k # 1 , say, are reducible by the s a m e polynomial p t ( x , g). Then
t h e i r difference - f I t is reducible by p i (2,g), which is a b s u r d s i n c e p h # fl,.
Thus at least one of the polynomials in (1) is i r r e d u c i b l e with Pi (2,y).
We designate t h i s polynomial a s R ( 5 , y) B. +
Exactly in the s a m e way we can show that t h e r e e x i s t s a number a such
that the polynomials R + a and Qi are irreducible.
they can be taken different. Conditions 1, 2, 3 are thus all satisfied.

W e choose a and 0 sufficiently s m a l l so that the polynomials P and are
d
3 - c l o s e to the polynomials P* and Q*, respectively, and thus 8 -close to the
* functions P and Q.
Since a + p, the polynomials R a and R + +
fl a r e i r r e d u c i b l e (otherwise,
the difference a - $ would be reducible by a z e r o d e g r e e poiynomial, which
64
57. STfiC'CT[lRALLY ST;\BLE AhD C'hS T;\BLE PATHS
i s impossible). F r o m ( E f a, R +
p) = 1, ( P t , Q,)= 1 and condition 3 it
follows, using a well-known algebraic theorem, that ( p , 0)= 1. Then the
system
dr
x=P(z,
- dy
id, x=az,
v)
i s 6 -close to s y s t e m (A) and h a s only a finite number of equilibrium s k t e s .

W e have thus established that f o r any 6 > 0 t h e r e e x i s t s a s y s t e m (A)
6 -close to (A) which h a s only a finite number of equilibrium s t a t e s . Note
that t h i s l e m m a is t r u e f o r any s y s t e m (A), whether s t r u c t u r a l l y stable or
unstable.
Let now ( A ) be a s t r u c t u r a l l y stable s y s t e m in @ with a n infinite number
of equilibrium s t a t e s in w. F r o m thedefinitionof s t r u c t u r a l stability, f o r any
E > O t h e r e e x i s t s 6 > O such that i f (A) is 6-close to (A) then
( H , A) 2 (p,.2), (2)
where N and H" are s o m e regions, H 3 F .

We have s e e n that a systerll (A) 6-close to (A) c a n be chosen so that i t
h a s only a finite number of equilibria in a plane. Then, in v i r t u e of (2),
If and thus a l s o should contain only a finite number of equilibrium s t a t e s
of (A), which contradicTs the s t a r t i n g assumption. Q. E. D.
C o r o l l a r y . If s y s t e m (A) is s t r u c t u r a l l y s t a b l e in s o m e region F>
i t h a s only isolated equilibrium s t a t e s in that region.
Indeed, by Theorem 10 w e conclude that a n equilibrium s t a t e which is a n
inside point of tl' is n e c e s s a r i l y i s o 1a t e d . A boundary point of lvis
a l s o inevitably a n isolated equilibrium s t a t e . This follows f r o m Theorem 10,
since a s y s t e m which is s t r u c t u r a l l y stable in IT is evidently s t r u c t u r a l l y
stable in s o m e region containing r.
F r o m Theorem 10 and Definition 14 i t follows that only isolated equili-
b r i u m s t a t e s c a n be s t r u c t u r a l l y stable.
Therefore, as we a r e concerned with s t r u c t u r a l l y s t a b l e equilibrium
s t a t e s , w e need only cc'nsider isolated points.
3. Multiplicity of an equilibrium s t a t e
Consider a dynamic s y s t e m
defined in z. Let Mlo(z0, yo) be a n equilibrium s t a t e of t h i s s y s t e m , .Ifo E C

T h i s Mo is an intersection point of two c u r v e s
P ( 2 . y)=O, Q(z, y)=O. (3)
The m u l t i p l i c i t y o f t h e e q u i l i b r i u m s t a t e M o is defined a s
the multiplicity of the intersection point .%Io of the two c u r v e s ( 3 ) (see $2.1).
D e f i n i t i o n 15. An equilibrium state M o (xo,yo) of a dynamic system is
said to be of multiplicity r (OY r -tuple) if M u i s a common point of multipli-
city r of cumes ( 3 ) .
65
An equilibrium s t a t e of multiplicity 1 is s a i d to be s i m p 1e .
An equilibrium s t a t e M o is s a i d to be of infinite multiplicity if M o is a n
intersection point of infinite multiplicity of c u r v e s ( 3 ) .
An equilibrium s t a t e M O is said to be of multiplicity higher than r if it
is of finite multiplicity r' > r or h a s infinite multiplicity.
Finally, an equilibrium s t a t e is s a i d to be multiple if i t s multiplicity is
> 1.
F r o m Definition 5 (2.1) and Definition 15 we s e e that if an equilibrium
s t a t e M O is of multiplicity r , then s y s t e m (A) is a s y s t e m of c l a s s k > r and
the follow_ingare satisfied: a ) t h e r e e x i s t E, > 0 and 60 > O such that any
s y s t e m (A) 6,-close to rank r to s y s t e m (A) h a s a t most r equilibrium s t a t e s
in Ueo(Mo); b ) for any e < eo and 6 > 0 t h e r e e x i s t s a s y s t e m (A) 6-close to
r a n k r to s y s t e m (A) which h a s a t l e a s t r equilibrium s t a t e s in U , ( M o ) .
The following theorem e s t a b l i s h e s the n e c e s s a r y condition of s t r u c t u r a l
stability of an isolated equilibrium s t a t e .
T h e o r e m 11. An isolated equilibrium state M o ( x 0 , yo) is structurally
stable only if it is simple (of multiplicity l ) , i.e., a necessary condition
of structural stability of an equilibrium state i s the inequality
P r o of . Let an isolated equilibrium s t a t e M Obe s t r u c t u r a l l y stable

and multiple. By the definition of s t r u c t u r a l stability of an equilibrium
s t a t e , t h e r e e x i s t s a region H containing Mo with fhe following property:
f o r any e > 0 t h e r e is 6 7 0, such that if s y s t e m (A) is 6-close to s y s t e m ( A )
then
where % is some region. Clearly H can be identified with any sufficiently

s m a l l neighborhood of M o . We take Ueo(M0)as H, where eo > 0 is s o s m a l l
that M O is the only equilibrium s t a t e of (A) inside U, (Mo)
(this exists
since M o is an isolated equilibrium state). Relation ( 5 ) is now written in
the f o r m
FIGURE 16
Let W b e a neighborhood Ue0/z(Mo)

(Figure 16). Some positive number
s m a l l e r than 2 is chosen a s E. System (A)is taken to be sufficiently c l o s e
66
57. 5 F R I ' C T I ' R 4 L L Y S T E L E .AND C;NSI'.ABLE PATtlS
to (A), so that relation ( 6 ) is satisfied; m o r e o v e r (z)

h a s a t l e a s t two
equilibrium s t a t e s in 11'. Such s y s t e m (x)
e x i s t s because A?+ is a multiple
(and not a s i m p l e ) equilibrium s t a t e of (A) (see 52.2, Definition 5 and
Theorem 6 ) .
& is generated from (,%Io)by a n e -translation, where
CEO E < ?. There-
fore, as is readily see.?, W = Cto,q(Mo) c &.',: Hence, in there are a t
least two equilibrium s t a t e s of (A). This contradicts relation ( 6 ) , s i n c e
in (X<,) t h e r e is only one equilibrium s t a t e of (A). Thus, the assumption
that .Ilo is a s t r u c t u r a l l y stable equilibrium s t a t e which is not a s i m p l e
isolated equilibrium s t a t e leads to a contradiction. T h i s completes the
proof.
T h e o r e m 11 c a n be alternatively s t a t e d a s follows: an isolated equili-
brium s t a t e &Io (z0. yo) f o r which
(i.e., a multiple s t a t e ) is not s t r u c t u r a l l y stable.

Thus, in o u r a n a l y s i s of s t r u c t u r a l l y stable equilibrium s t a t e s , we need
consider only simple equilibrium s t a t e s , which is the subject of the next
chapter.
' Indeed, take some point M W outside HI. Let f denote the e-translation which transforms CLo into 2,
and let I'be the boundary of the neighbarhood U E o ( M ) , Fthe boundary of 2 (f'=f(r)), andM=f(.M).
6y sssuniption p(M, %) <$. T h e sesment .MS contains at least one point s o f the boundary
-r. There-
fore p (Jf. 5)< %. Let s"=f (,S), where S ( r . Then p (S. 3)< +. This leads fo the inequality
p (S,M)Q p (AI, s ) + p (g,S) < 2, which is impossible since p (S,Fi')=?.
C h a p t e r IV
EQUILIBRIUM S T A T E S OF S T R U C T U R A L L Y S T A B L E
S Y S T E M S . S A D D L E - T 0-SADDLE S E P A R A T RIX
INTRODUCTION
We have shown in Chapter I11 ( $ 7 . 3 , Theorem 11) that if a dynamic

s y s t e m is s t r u c t u r a l l y stable in s o m e bounded region, it h a s only s i m p l e
equilibrium s t a t e s in that region. In t h i s chapter we will establish which
of t h e s e simple equilibrium s t a t e s a r e s t r u c t u r a l l y stable equilibria. The
chapter is divided into four sections, S 8 through %11. In $ 8 it is proved
that a s i m p 1e n o d e (ordinary, d i c r i t i c a l , o r confluent) and a s i m p 1e
f o c u s * a r e s t r u c t u r a l l y stable s t a t e s of equilibrium. The proof f o r
these different c a s e s is exactly the s a m e , and we t h e r e f o r e consider only
the o r d i n a r y node.
In 9 w e p r o v e t h a t a s i m p l e s a d d l e p o i n t (i.e., a s i m p l e e q u i l i b r i u m
s t a t e which is a saddle point) is a s t r u c t u r a l l y stable s t a t e of equilibrium.
In l o we consider a simple equilibrium s t a t e with p u r e imaginary
c h a r a c t e r i s t i c roots, and it is proved that t h e s e equilibria a r e s t r u c t u r a l l y
unstable. Incidentally L highly important t h e o r e m is proved concerning
the creation of a closed path f r o m a multiple focus (Theorem 14). Ac-
cording to t h i s theorem, if 0 is a m u 1t i p 1 e f o c u s of s y s t e m (A) (i.e.,
a point with pure imaginary c h a r a c t e r i s t i c values which is neither a c e n t e r
n o r a c e n t e r - f o c u s ) , infinitesimally s m a l l i n c r e m e n t s will convert the
s y s t e m (A) into a modified s y s t e m which h a s a closed path in any a r b i t r a r i l y
s m a l l neighborhood of 0.
Equilibrium s t a t e s do not f i g u r e in 511. T h i s section, however, is
closely linked with l o , and hence its place in the p r e s e n t c h a p t e r . Indeed,
l l d e a l s with a s a d d l e - t o - s a d d l e s e p a r a t r i x , and it is proved
that such a s e p a r a t r i x extending between two saddle points is a s t r u c t u r a l l y
+
unstable path of the dynamic s y s t e m . F o r t + 00 and t + - DJ the
s e p a r a t r i x goes to saddle points which may be different or coincident.
8. STRUCTURAL STABILITY OF A NODE AND

A SIMPLE FOCUS
1. Canonical s y s t e m
In t h i s subsection we will review the basic propositions concerning
simple s t a t e s of equilibrium. Detailed proof will be found in QT, Chapter IV.
* A simple node is a simple equilibrium state which is a node (the characteristic values are real numbers of
*
equal sign). A simple focus is an equilibrium state with complex characteristic values, which are not pure
imaginary.
68
SP. 5TRLCTI'RAL ST.iBCLCTY OF KODE ;\ND SIllPLE FOCVS
15.ithout loss of generality, we c o n s i d e r a s i m p l e equilibrium s t a t e a t the

origin, i.e., a t the point 0 (0.0). T h e dynamic s y s t e m i n t h i s case c a n be
written in the f o r m
-d=r a a z + b y + q (z, y), $=cz-;dy+q((r, y),
dt (1 1
where the functions 9 ( 5 , y) and I$ (I, y) a r e continuous and continuously dif-

ferentiable to first o r d e r in I and y in c;
a t the point 0 (0. 0), the functions q
and $I and t h e i r p a r t i a l d e r i v a t i v e s all vanish:
'F('?, U ) = $ ( U , O ) = ' F ; ( O , O ) = q ; ( o , O)=I&(@, O)=$.k(O. 0)=O. (2)
Since 0 is a s i m p l e equilibrium,
Applying a non-singular l i n e a r transformation to s y s t e m (1) w e c a n

r e d u c e it to a canonical f o r m d e s c r i b e d by a m a t r i x (z i) in a normal
Jordan f o r m . Let ri., and A.2 be t h e c h a r a c t e r i s t i c values, Le., the r o o t s of the
c h a r a c t e r i s t i c equation
1 a--E. .Lo (4)

c d4.1
or
i.?--aE.+l=O, (5)
where
We should distinguish between the following cases:

I. The c h a r a c t e r i s t i c r o o t s ;.! and 1.2 are real, different, and of the s a m e
sign. System (1) is then reduced t o the canonical f o r m
where ilk2> 0. The equilibrium point 0 (0,0) is then called a n o d e (an

o r d i n a r y node).
11. E., and ).,are equal, i.e., ? . , = h = h . System (1) r e d u c e s e i t h e r t o the
form
or
where p + O .
69
Ch.IV. EQUILIBRIUM S T A T E S OF STRUCTURALLY STABLE SYSTEMS
In c a s e ( 8 ) , the equilibrium 0 is c a l l e c a d i c r i t i c a 1 n o d e and in

~
case (9) a c o n f l u e n t n o d e .
111. h, and & a r e real, different, and of opposite signs. In t h i s c a s e ,
the canonic f o r m of the s y s t e m is a s in c a s e I, i.e., (7). but A i l z = A < 0.
The point 0 is then called a s a d d l e p o i n t .
IV. hi and hz a r e complex numbers, which a r e not pure imaginary. The
canonical f o r m of s y s t e m (1)is then
I
where h i , 2 = a f pi, a + O p>O. The equilibrium s t a t e 0 is then a f o c u s
(a s i m p l e f o c u s ) .
V . hi and & a r e pure imaginary numbers, Li=pi, & = - p i , f5=#=0. The
canonical f o r m of s y s t e m ( 1 ) is then
In t h i s c a s e 0 is called a n equilibrium s t a t e with pure imaginary I

characteristic roots.
W e will show that in c a s e s I through IV the point 0 is a s t r u c t u r a l l y
s t a b l e equilibrium s t a t e , and in c a s e V i t is s t r u c t u r a l l y unstable.
The behavior of the paths of the dynamic s y s t e m i n the neighborhood of
0 (0,O)in each of t h e s e five c a s e s is investigated in QT, 57 and S8.
We will now review some of the p r o p e r t i e s which a r e used in establishing
the s t r u c t u r a l stability (for detailed proof, see QT, $7).
In c a s e I (a node), all the c i r c l e s
P +ya = rs (12)
of sufficiently s m a l l r a d i u s r a r e contact-free c y c l e s f o r the t r a j e c t o r i e s
of the s y s t e m . Let 0 be a s t a b 1 e n o d e ,* i.e., hl < 0, hz < 0. Then any
path with 0 as i t s limiting point goes to 0 f o r t -+ +
00. Let
be m e of these paths, M ( t ) a point of t h i s path with the coordinates x ( t ) ,y ( t ) >

and let p ( t ) = 1/z (t)P+ Y (t)*be the distance of M f r o m the origin. For
t -t + 00, p ( t )goes monotonically to z e r o . In the c a s e of an u n s t a b l e
n o d e (Ai > 0, hz > 0 ) , p ( t ) + 0 f o r t --t- 00.
In the case of a d i c r i t i c a 1 n o d e ( c a s e 11, canonical s y s t e m ( 8 ) ) or a
s i m p 1e f o c u s ( c a s e IV), the situation is p r e c i s e l y the s a m e as f o r an
o r d i n a r y node. Indeed, c i r c l e s ( 1 2 ) of sufficiently s m a l l r a d i u s r a r e
without contact cycles, and any path (13) which f o r t = t o c r o s s e s one of
+
t h e s e c i r c l e s goes to 0 f o r t + 00 if the node or the focus is stable ( h (0,
or a < 0, respecitvely) and to t -+ - 00 if the node or the focus is unstable
( h > 0, or a > 0, respectively). In e i t h e r c a s e , p ( t )-+ 0 monotonically.
In the c a s e of a c o n f l u e n t n o d e ( c a s e 11, canonical s y s t e m ( 9 ) ) t h e 321
c i r c l e s (12) are replaced by e l l i p s e s
$8. STRCrCTCR.AL S L'hYILlTY OF NODE ihD SI!.LPLE FOCkS
where k is s o m e positive number." F o r sufficiently s m a l l r , a l l t h e s e

e l l i p s e s a r e without contact c y c l e s f o r the paths of s y s t e m (9). F o r a
stable (unstable) node, i.e., f o r k < 0 (A > 0), each path initially c r o s s i n g
such a cycle without contact approaches the point 0 f o r t -+ +- 00 ( t -+ - 0 0 ) .
2. S t r u c t u r a l stability of a s i m p l e node and a focus
\Le w i l l prove that a simple s t a t e of equilibrium which is a n o d e

(ordinary, dicritical, 0:- confluent) o r a s i m p 1e f o c u s is a s t r u c t u r a l l y
stable path of the dynarr.ic s y s t e m . The main part of the proof is contained
in L e m m a 1 below.
Let
be a dynamic s y s t e m defined in s o m e region a,,

and 0 (0, 0) a n equilibrium
s t a t e of this s y s t e m which is a s i m p l e node o r a s i m p l e focus (0c GI).
s y s t e m (A) is given in c i n o n i c a l f o r m .
L e tn ttt a 1 . There exists a neighboyhood {r of the equilibuititn state 0
with the following property: f o r any E > 0 there exists > 0 such that <f
systeiti !B) dejyned in E , is u -close in E , to system ( A ) and 0 (0. 0 ) is an
equilibrium state of system ( B l , then
e
( H , . 4 ) = ( H , B).
P r o o f . F i r s t consider the case when 0 is a n ordinary node (case I).

System (A) h a s the f o r m ( 7 ) :
-dd=
z
t ) . * f f q ( s , Y). $==.2Y,*(X, Y). (Aj
We take 3.t < 0, J.2 < 0, i . e . , a stable node.

Let u s s u m m a r i z e s o m e of the r e s u l t s f r o m QT, Chapter IV.
In QT, 9 6 . 3 i t is shown that the functions 'p and $ c a n be r e p r e s e n t e d in
a c e r t a i n neighborhood cf 0 in the f o r m
'F (J.Y) =*a", \I* Y) 4-Y g 2 (z, Y)t Ip (3, Y) = x i ! (f, Y) 5 Y/? (I,Y), (15)
where
i 1
gl (5,Y) = 1'fk (tr, CY) d t ,
0
g? ( 5 , Y\ = j
a
(FY ( t ~ t, ~d t), (16)
and i , and f 2 a r e s i m i l a r l y e x p r e s s e d in t e r m s of the derivatives of $(z, y).

The functions g , , g2, f,. fz are continuous and
a"! (0, 0) = g: (0, 0)= f , (0, 0 ) = f 2 (0, 0 ) = 0. (17)
Let
s=s(t), Y=Y(t) (18)
For k we can take any positive number, satisfying the inequality k C 4 >.
CI
See QT, 57.1.
71
c;h.IV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
be a path of s y s t e m (A). Its equation in polar coordinates is
Since
w e have f r o m (A) and (15), using the standard r e l a t i o n s s=pcos8, y=psin8,
9 = 2ps ( t ) [It, cos2 e + & s i n g e+ cos) eg, + cos e sin e ( g 2 + f , ) + sins ef,j, (21)
where f l , f2, g,, g, are functions of p cos e and p sin 0. The e x p r e s s i o n

+
A, cosp 0 Az sin' 8 is negative f o r a l l r e a l 0 and is periodic in 0. Its maximum
value is t h e r e f o r e - m , where m > 0. Since g,. g,, f , . f 2 are continuous
functions which vanish a t the origin 0 (0, O), t h e r e e x i s t s s o m e ro > 0 such
that if p ( t ) < r o , then
I cosa Og, +cos e sin e (gz+ f,)+sin* etz I <f .

The expression in s q u a r e b r a c k e t s i n (21) is thus definitely less than
_ _2m <-T,
m so that
F r o m ( 2 2 ) it follows immediately that all the c i r c l e s
x2+ys=rs, (23)
where r < r o a a r e cycles without contact f o r paths of s y s t e m (A). Indeed,

for a path (18) to b e tangent to one of t h e s e c i r c l e s at the point @ ( t o ) , ~ ( 1 , ) )
we should have z(to)x'(to)+y(to)y' (to)=Os i. e., dpiF)= 0 , which c o n t r a d i c t s
(22). Now separating the v a r i a b l e s in (22) and integrating f r o m to to t > t o ,
we s e e that
pa ( I ) < pz (to)e-m(l-to),
whence i t follows that f o r f+ +

w , p ( t ) goes to z e r o (monotonically, as we
s e e f r o m (22)). We thus obtain, a s required, that any path (18) which for
t = t o p a s s e s through the c i r c l e
c r o s s e s a l l the concentric c i r c l e s of s m a l l e r r a d i i as t i n c r e a s e s and

approaches the point 0 for t -+ co.+
Let Co be the c i r c l e ( 2 4 ) and H the region inside t h i s c i r c l e (i.e., Ur0 (0).
We will show that the H defined in t h i s way s a t i s f i e s the lemma.
Let (B) be a dynamic s y s t e m defined i n c,,
which is sufficiently c l o s e
to s y s t e m (A) and f o r which 0 is an equilibrium s t a t e . System (B)can be
72
98. STRCCTVR4L STABILITY OF &ODE .WD SIhlPLE FOCL'S
where ai are sufficiently s m a l l numbers, and the functions (p and 3 are

sufficiently c l o s e to the functions g, and $, respectively; m o r e o v e r , (p and 7
and t h e i r f i r s t derivatives vanish a t 0.
(p and 7 can be r e p r e s e n t e d , like 'p and J., in the f o r m
where zii_and TLare e x p r e s s e d in the f o r m (16) i n t e r m s of the f i r s t d e r i v a -

t i v e s of 'p and'$.
Let
p=i;(t). e=G(t)
be a path of s y s t e m ( B ) defined by the equations in polar coordinates.

Using ( B ) and (25) to calculate q, we obtain
--
d;;"dt ( I ) - 2 3 [(Ai cos25 + sin2 ir, + (atc o s ~e'+ (a2 as)cos tt s i n e +
1.2
# #
f Z;
- - 6 + (g, f f , ) cos 8sin tl+
sine e) C g ,
COS~
# - - -f 2 sin? m. (26)
If s y s t e m (B) is sufficiently c l o s e t o s y s t e m ( A ) in 4 , the e x p r e s s i o n s

in b r a c k e t s in the right-hand s i d e s of (21) and (26) are sufficiently c l o s e to
each o t h e r in G I , and hence a l s o in B." In virtue of the p a r t i c u l a r choice of
-
f?, the expression in b r a c k e t s i n ( 2 6 ) is less than - 4 m < - f in H . T h e r e -
fore, if s y s t e m ( B ) is sifficiently c l o s e t o s y s t e m (A), the expression in
b r a c k e t s in (26) is a l s o l e s s than -7 in 8. Then relation (22) with a l l the
consequences is satisfied f o r s y s t e m (B) in g . Hence i t follows that t h e r e
e x i s t s ui > 0 with the following property: if s y s t e m ( B ) is u,-close t o
s y s t e m ( A ) i n G,, all the circles (23) are c y c l e s without contact f o r the
paths of s y s t e m ( B ) and any path of t h i s s y s t e m which f o r t = to p a s s e s
through the c i r c l e (24) i n t e r s e c t s all the concentric circles of s m a l l e r r a d i i
a s t i n c r e a s e s , approaching the point 0 f o r t -f + OO:PT .L*
Let ri be a positive number such that r, < ro, r1 < $. C1 is a c i r c l e of .

I
r a d i u s r , centered a t 0, HI is the region inside t h i s circle, and G' the r i n g

between the c i r c l e s Co and C1 (Figure 17).
We c a n now apply Lemma 1 2 (14.2). By this l e m m a , t h e r e e x i s t s
> 0 such that i f s y s t e m (B) is 02-close to s y s t e m (A) the partition of ST; by
the paths of s y s t e m (B) is E -identical t o the partition of by the paths of
- 1
We compare the values of thesc expressions in the same point (z, y), i.e., w e take p = p , 8=8. The
5
closeness of the Functions and & to f i andqi, respectively, follows from equations (16) and the analogous
relations for 'fi and T i . Here closeness is to be understood to rank 0.
* * Hence it follows, in particular, that the point 0 is a node or a focus of system (6). It c a c be shown with
much less effort, using the continuous dependence of the characteristic roots on the system coefficients,
that for sufficiently close systems (B), 0 is a node. This is not enough for our purposes, however. It is
clear from the proof that, among other things, system (8) has no h i t cycles in p.
73
Ch. IV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
s y s t e m (A). The mapping T of w o n itself corresponding to t h i s e -identity

can be chosen so that each point M E CO is mapped into itself:
T(M)=M.
Let a be any number satisfying the inequalities a > 0, a < ul, a < ua. Let
s y s t e m (E) be o-close to s y s t e m (A) in cl. Since u < u2, t h e r e exists, a s
we have just pointed out, a mapping of the ring
onto itself which is an e-translation, which t r a n s -
f o r m s the paths of s y s t e m (A) into paths of
@ s y s t e m (E), and which l e a v e s unchanged all the

points of the boundary c i r c l e CO. Let this mapping
b e T. It is defined on FT>and hence on the c i r c l e
C,. Let N E C , , L a path of s y s t e m (A) passing
through N, M an intersection point Of L with Co, z a
Y
path of s y s t e m (E) through M , and N a n intersection
point of with the c i r c l e C , ( F i g u r e 1 7 ) . Clearly
T (N)= 3.
The mapping T i s originally defined in the ring w.
FIGURE 17 W e will now continue it to the e n t i r e & i n the
following way. Let T (0) = 0. Now let S be any point
in Hi,which is not 0. The path L of s y s t e m (A)
passing through S a t t = ti c l e a r l y c r o s s e s the c i r c l e Ci a t the point N f o r
t = t o < tl. Let T ( S ) = s', where 3 is the point which c o r r e s p o n d s to the time
tt on the path L"of s y s t e m (B), if a t t = to the path z p a s s e s through the point
N = T ( N ) ("time reflection," s e e Figure 17). Since the r a d i u s of Cl is l e s s
than;, w e have p (S.3)< e..
The mapping T continued in t h i s way is now defined in the e n t i r e H . It
is evidently an e-translation which t r a n s f o r m s the paths of (A) into the
paths of (E). We have thus established that if ( E ) is 6 - c l o s e to (A), then
(a, A) 2 (a, B ) , and a l s o ( H , A ) ( H , B ) . This completes the proof of the
l e m m a f o r the c a s e when 0 is a n o r d i n a r y node.
If 0 is a d i c r i t i c a l node o r a focus, the l e m m a is proved p r e c i s e l y in
the s a m e way. If 0 is a confluent node, the proof proceeds along the s a m e
lines, and the only change is that the concentric c i r c l e s ( 1 2 ) a r e replaced
by a family of e l l i p s e s (14), and H i s taken as the region lying inside one
of these e l l i p s e s . The proof of the l e m m a is thus complete.
We now proceed with the fundamental theorem of s t r u c t u r a l stability of
a simple node or focus.
T h e o r e m 12. The equilibrium state M,,( x 0 , yo) of the system
dx
~ = P ( s g),
, %=Q(x, g) (A)
f o r which A =- 0, a # 0 (Le., a node o r a focus) i s structurally stable.

P r o o f . Without loss of generality, we a s s u m e that the s t a t e of equili-
b r i u m is a t the origin 0 (0.0). and s y s t e m (A) is given in canonical form.
This evidently can be achieved with the aid of a l i n e a r transformation. By
Lemma 2, 6.1 and f r o m the definition of a s t r u c t u r a l l y stable s y s t e m
(Definition 14, 1 7 . 1 ) M o is a s t r u c t u r a l l y stable equilibrium s t a t e of the
original s y s t e m i f and only if 0 is a s t r u c t u r a l l y stable equilibrium s t a t e
of the t r a n s f o r m e d system.
14
The proof will be c a r r i e d out f o r the c a s e when O i s a n ordinary node.
In the other cases (confluent and d i c r i t i c a l node, focus) the proof is exactly
the s a m e . System (A) is taken in canonical f o r m (7).
Let H be the region considered in Lemma 1, i.e., the i n t e r i o r of a
circle Co of sufficiently s m a l l r a d i u s ro centered a t 0 . Let GI be s o m e
region satisfying the condition H c Gt, gt c G ( F i g u r e 18; c
is the original
region used to define the c l o s e n e s s of s y s t e m s ) . c,
is a t a positive distance
f r o m the boundary of G. Let this distance be d .
FIC L'RE 18
Let E be s o m e positive number. By L e m m a 1, t h e r e e x i s t s o > O such

that if s y s t e m ( B ) is defined in where it is 0 - c l o s e t o s y s t e m (A) and h a s
a n equilibrium s t a t e a t 0 (0. 0), then
2
( H , -4) = (H, B).
Let 6 be a positive number, 6<:, with the following property: if

system (x) is 6-close t o s y s t e m (A) in e, then (A)h a s only one equilibrium
s t a t e Ot ( E o , qo) in and
P (0,0,)
= m< Po. (28)
where po is a fixed number satisfying the inequalities
PO<%, po<d (29)
and a n additional condition which will be formulated a t a l a t e r stage. This

6 e x i s t s in virtue of R e m a r k 3 to Theorem 6 (52.2).
75
ch. IV. EQUILrBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
Consider the transformation
x=u+Eo* y=v+qo, (30)
where O,&, qo) is the previously mentioned equilibrium s t a t e of s y s t e m (a)

6-close to s y s t e m (A). Let s y s t e m (A) be
Transformation (30) r e d u c e s it to the s y s t e m

dv
$==Pu+&09 u+qo). ;ir"iZ(U+EO, V+%),
which, changing over f r o m u and u to x and 1, respectively, takes the f o r m
$=P(,+Eo: Y+qo). $=w+&O* Y+(lO). (B)

System (B) is defined in G*, which is obtained when a is translated by a
vector v ( - E o , -qo). Since by a_ssumption v m <
po<d, G* contains zi,
i . e . , ct c G * . Therefore both (A) and (B) a r e defined in 8,. Clearly i f po is
sufficiently s m a l l , (A) and (B) a r e sufficiently c l o s e in F,. The third
condition imposed on po in addition to ( 2 9 ) is the following: po i s s u f f i -
c i e n t l y s m a l l s o t h a t i f v-<po, s y s t e m (B)i s : - c l o s e
t o s y s t e m (Xi) i n 4 .
Note that 0 (0,0) is the equilibrium s t a t e of s y s t e m (B).
Let K be the inside of a c i r c l e of r a d i u s ro centered a t 0,(Figure 18).
K is obtained when H i s t r a n s l a t e d by a vector v (i.e., a vector with the
coordinates Eo, qo). Hence K c 8, and s y s t e m (A) is t h e r e f o r e defined in K.
Transformation (30) t r a n s f o r m s regio_n K into region H, s y s t e m (A) into
s y s t e m (B), and the paths of s y s t e m (A) i n K a r e t r a n s f o r m e d into the
paths of s y s t e m (B) in H . Since v m < po < G, and (30) is a topological
mapping, we conclude that
-8
(K,X)2 ( H , B ) . (31)
Let now (?i)be some s y s t e m 6-close to (A) in E. Then, f r o m the

definition of 6 and in virtue of the conditions imposed on pa. s y s t e m (B)is
;-close to s y s t e m (A) in (?,
and relation ( 3 1 ) is satisfied. Now, since 8 < %s
s y s t e m (B) is a - c l o s e t o s y s t e m (A) i n a,.
Moreover, the point 0 (0.0) is
an equilibirum s t a t e of s y s t e m (B). Therefore relation ( 2 7 ) is a l s o satisfied,
and f r o m (31) and (27) we c l e a r l y have
( H , A ) 2 (K, x). (32)
We have thus shown that for any e > 0 t h e r e e x i s t s 6 > 0 such that if
system (A) is & - c l o s e to (A) in 6, r e l a t i o n (32) applies. This i m p l i e s i n
i t s t u r n that s y s t e m (A) is s t r u c t u r a l l y stable i n H, i.e., 0 is a s t r u c t u r a l l y
stable s t a t e of equilibrium of s y s t e m (A).
76
?e. STRC'CTL'R4L ST.4BILITY OF SODE AKD SIhlPLE FOCL'S
A s w e have noted above, the s t r u c t u r a l stability of a d i c r i t i c a l or a

confluent node and of a simple focus is proved exactly in the s a m e way; in
the case of a confluent node, H is the inside of the corresponding ellipse.
T h i s completes the proof of the theorem.
R e m a r k 1. Lemma 1 can be strengthened by omitting condition 2 f r o m
i t s statement. It will be used in t h i s f o r m in what follows, and we therefore
give h e r e the a l t e r e d formulation and the corresponding proof.
There exists a neighborhood H* of a simple node m a focus 0 enclosed
in a cycle without contact tchich has the following property: fmany E > 0
there exists 6 > 0 such that if system (A) is 6-close to system ( A ) , then
(Z* A ) G (P**I ) .
Note that the s t r u c t u r a l stability of the equilibrium s t a t e 0 follows
immediately f r o m t h i s proposition. W e w i l l , conversely, prove this l e m m a
proceeding f r o m the previously established s t r u c t u r a l stability of the
point 0.
P r o o f . Consider the case when 0 is a n o r d i n a r y node and s y s t e m (A) is
given in canonical f o r m . Take a sufficiently s m a l l circle H of r a d i u s r
centered a t 0, which contains no o t h e r equilibrium s t a t e s except 0 and no
closed paths of the s y s t e m . System (A) is s t r u c t u r a l l y s t a b l e i n t h i s circle.
Let H* be a c i r c l e of r a d i u s f centered a t 0. We take a sufficiently s m a l l
E* > 0, e.g., e * < 6. Because of s t r u c t u r a l stability t h e r e e x i s t s 8* > 0 such
that i f s y s t e m (A) is 6*-close to (A), then
e*
( H , A ) E(&',
- 2). (33)
Since E*<,; w e conclude that H* c H'. Hence, and by (33), i t follows that
K* contains no closed paths of s y s t e m (A). The boundary circle of H* is
a r l y a cycle without contact f o r s y s t e m (A).
c l e-
H* is thus enclosed by a cycle without contact of s y s t e m (A) and a l l
s y s t e m s sufficiently c l c s e to (A) have no closed paths in F*. The r e s t of
the proof proceeds along the s a m e lines a s th? proof of L e m m a 1, but the
paths of s y s t e m (A) and of the c l o s e s y s t e m (A) in g e n e r a l tend to different
points 0 and 6.
R e m a r k 2. A s i m p l e s t a t e of equilibrium M o (q, yo) of the s y s t e m
is a node or a focus i f
p ; ( 2 0 , Yo) PI (20, YO)

20, YO) Q; (20, YO)
' = / Q ' x(
+
6 = p; (zo, YO) Q&(20, YO) # 0.
If s y s t e m (6)is sufficiently c l o s e t," s y s t e m (A), t h e r e is precisely one
equilibrium s t a t e 0-(& iTo) of s y s t e m (A) in a sufficiently s m a l l neighborhood
of the equilibrium s t a t e M o . The quantities 3 and corresponding to the
point d a r e not marked1,y different f r o m A and 6, so that A > 0 , & # O , and the
point 0" is a simple node or focus. It is readily s e e n that if 0 is an ordinary
node, b is a l s o an ordinary node, and if 0 is a s i m p l e focus, d is a l s o a
simple focus. If, however, 0 i s a confluent node, d i s e i t h e r a n o r d i n a r y
59. STRUCTURAL STABILITY O F A SADDLE POINT
1. Reduction of t h e s y s t e m to canonical f o r m by a
n e a r l y identical t r a n s f o r m a t i o n
In our proof of the s t r u c t u r a l stability of a s i m p l e equilibrium s t a t e

which is a saddle point, we a s s u m e a s before that the saddle point coincides
with 0 (0,O) and the dynamic s y s t e m h a s the canonical f o r m
where hi& < 0. Without loss of generality, we may take
%>0, ;L,<o. (1)
We w i l l show that any s y s t e m sufficiently c l o s e to s y s t e m (A) can be

reduced to canonical f o r m by a t r a n s f o r m a t i o n which is as c l o s e to the
identity t r a n s f o r m a t i o n as d e s i r e d .
L e m m a 1 . i f t h e system
i s sufficiently close in E t o system (A), a nearly identical linear transfm-

mation will reduce system (A) to the f m m
where e , and e2 are infinitesimal, and the functions and 5 together with
tlteir f i r s t derivatives vanish at the point 0 (0.0) and are arbitrarily close
to the cmresponding functions g, and 9.
PI- o o f . Let eo and B2 be two positive n u m b e r s with the following pro-
perty: f o r any s y s t e m (A) Bo-close in E to s y s t e m (A), t h e r e is p r e c i s e l y
one equilibrium s t a t e a (Eo, qo) in U,, (O),and this equilibrium s t a t e is a
saddle point. The existence of t h e s e eo and Bo follows f r o m Theorem 6 and
Definition 5 (S2.1), and a l s o f r o m the fact that s m a l l changes in s y s t e m (A)
leave the determ-inant A = liX2 negative.
Let s y s t e m (A) be 6 -cl_ose to s y s t e m (A), where 8 5 6,, and let (to, a qo) be
the equilibrium s t a t e of (A) lying in U, (0). System (A) c l e a r l y c a n be
78
F. 5-RLCTLRAL S T A B I L I T Y OF .A SADDLE POIST
and i f 6 > 0 is sufficiently s m a l l , the n u m b e r s a i , P I , EO. q o c a n be made a s

s m a l l a s Cesired and 'pl and sl
w i l l be a r b i t r a r i l y c l o s e to the corresponding
functions 'F and 5, vanishing together with their derivatives a t the point
a (Eo, qo). Applying the transformation
2=.Y+Eo, y=Yfqo, (3)
we obtain the s y s t e m
where v2 and $2 vanish a.tO(0, 0) together with their derivatives.

If the n u m b e r s ai anti pi are sufficiently s m a l l , the m a t r i x
is sufficiently c l o s e to the m a t r i x
and the c h a r a c t e r i s t i c r o o t s of the m a t r i x ( 5 ) are t h e r e f o r e c l o s e to h, and

j.2. W e w r i t e t h e s e c h a r a c t e r i s t i c r o o t s in the f o r m A . , + E ~ , A2+e2.
\ V e will now establish the existence of a non-singular m a t r i x S which is
c l o s e to the unit matrix and s a t i s f i e s the relation
W e s e e k S i n the f o r m
Right-multiplying (7) by S, we write i t in the f o r m
(rP qs ) (:h 12 f P?) = ( ' . l S e l

a? ?.2!ez)(r t). (8)
Comparing the elem'ents in the left- and the right-hand s i d e s of (8), we

obtain four equations:
p (A, i a 1 ) 4-Q P l = (11 f1) p . P.rt Q ( b f B2) = (Ai + E l ) Q (9)
and
79
ch. IV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
The determinant of t h i s s y s t e m of homogeneous equations i n p and q is

equal to the c h a r a c t e r i s t i c polynomial of the m a t r i x (5) for b = h l e l . But +
+
since k1 e, is a c h a r a c t e r i s t i c r o o t of the m a t r i x (5), t h i s determinant
vanishes. Hence, equations (11) a r e equivalent, and i t suffices to find a
solution satisfying only one of these equations. Let p = 1, and the second
equation in (11) gives q =
hl--hz+el-fh~
Similarly, seeing that 1, + e2 is a c h a r a c t e r i s t i c root of the m a t r i x (5),
we take S= 1 in (10) and find r =
A2-h -k %--ai
These n u m b e r s p , q, r , s s a t i s f y equations ( 9 ) and (lo), and therefore
the m a t r i x
s a t i s f i e s the m a t r i x equation (7). Since h l Z h 2 , the m a t r i x S is a r b i t r a r i l y

c l o s e to t_he unit m a t r i x f o r sufficiently s m a l l a t , pi, et ( i = 1, 2), i.e., when
s y s t e m (A) is sufficiently c l o s e to s y s t e m (A), and it is t h e r e f o r e a non-
singular m a t r i x . Applying the transformation
u=pX+ qY, u = rX+ sY, (13)
where p , q. r , s a r e the m a t r i x e l e m e n t s f r o m (12), to s y s t e m (4)and writing

x and y f o r u and u , respectively, w e obtain the s y s t e m
( h +~ e,) Y + ~ l r(11.
dx
= (kit- ei) X + 'pa (x, v), = r Y) (14)
( s e e QT, 6.2), i.e., s y s t e m (B). If s y s t e m (A), i.e., s y s t e m (Z), is

sufficiently close to s y s t e m (A), transformation ( 3 ) and (13), as we have
seen, a r e a r b i t r a r i l y c l o s e to the identity transformation. But then the
i n v e r s e transformations a r e a l s o a r b i t r a r i l y c l o s e to the identity t r a n s f o r -
mation, and the-same applies to t h e i r product. Now, t h i s product trans;
f o r m s s y s t e m (A) into s y s t e m (14). We have thus shown that if s y s t e m (A)
is sufficiently c l o s e in to s y s t e m (A), a l i n e a r t r a n s f o r s a t i o n a r b i t r a r i l y
c l o s e to the identity t r a n s f o r m a t i o n will reduce s y s t e m (A) to the f o r m (14),
i . e . , to the f o r m (B). The n u m b e r s el and ez in t h i s transformation a r e
a r b i t r a r i l y s m a l l , and the functions 'ps and $ s r a s is readily seen, are
a r b i t r a r i l y c l o s e to the functions cp and $, respectively. T h i s completes
the proof of the lemma.*
2. Proof of the s t r u c t u r a l stability of a saddle point
Before proceeding with the actual proof, we briefly review the p r o c e -

dure for investigating the pattern of the paths of s y s t e m (A) near a saddle
point 0 (0, 0). For m o r e details, s e e QT, S7.3.
'
In our proof of Lemma 1 we only made use of the fact that At # A,, without resorting to the different s i g a
of L, and Lr. The lemma therefore remains valid when 0 (0,0) is an ordinary node, and not only a saddle
point. This proposition, however, was not needed in our proof of the structural stability of the node.
80
5". SI'KI'CTl'RAL ST.4BILITY OF .4 SADDLE P O I N T
System (A), in accordance with relations (15) f r o m S8, is written in

the form
where f , , f ? , g,. g, a r e continuous functions which vanish a t the point 0 (0, 0).
LVithout loss of generality, we again take
1.1 > O , izZ<O. (1 1
The e n t i r e a n a l y s i s is confined to a neighborhood of the point 0 where 0

is the only equilibrium s t a t e of s y s t e m (A).
Let k be a fixed positive number. We draw the s t r a i g h t lines y =
and focus o u r attention on the rectangle with i t s v e r t i c e s a t .4 (q.kq),
* ks
(-- ZO. k z d , E , (-- 20,-- kzd, A I ( S O . - k z d . T h e s e s t r a i g h t l i n e s are the

diagonals of this rectartgle ( F i g u r e 19). Let E correspond to the e n t i r e
rectangle, and R to i t s i n t e r i o r . zo is understood to b e a sufficiently s m a l l
positive number. Consider the i n t e r c e p t s of horizontal and v e r t i c a l lines
between the diagonals of the rectangle R and the diagonals a l s o .
A ff
Y Y
4 I 4
FIGURE 19
Let L be a path of s y s t e m (A) corresponding to the solution
z=z(t), y=y(t).
If a t t = t o > the path L c r o s s e s the diagonal y = kx ( o r y = - kz) a t s o m e

point o t h e r than the origin, we have
If a t f = t o , the path L c r o s s e s a n intercept of the v e r t i c a l line between

the diagonals of the rectangle R, w e have
where
I-/<k.
81
ch.IV. EQUILIRRIUM S T A T E S OF STRUCTURALLY STABLE SYSTEMS
Finally, if a t t = t o , the path L c r o s s e s the i n t e r c e p t of the horizontal

line between the diagonals of the r e c t a n g l e R, we have
(19)
and
z (to) i
m G T . (20)
Since the functions f l and gi a r e continuous and vanish a t the point 0 (0, 0),
r e l a t i o n s (161, (17), and (19) f o r sufficiently s m a l l s o l e a d to the equalities
where xl, x2, x, are s o m e n u m b e r s which satisfy the inequalities, say,
( i = 1, 2, 3. The sign + in ( 2 1 ) c o r r e s p o n d s to the diagonal A B , , and the

sign - to the diagonal B A , ) .
It follows f r o m (21)-(23) ( s e e QT, 17.3) that the paths of s y s t e m (A) in
the r e c t a n g l e R make the p a t t e r n shown in Figure 20. On each of the s i d e s
A B and A , B , of the r e c t a n g l e R t h e r e is one point - D and D 1 , r e s p e c t i v e l y -
through which an a - s e p a r a t r i x of the saddle 0 p a s s e s , and on each of the
s i d e s A A , and B B , t h e r e a r e the points C and C, through which p a s s a-
s e p a r a t r i c e s , which are the continuation of the o - s e p a r a t r i c e s . The s i d e s
of the rectangle a a r e s e g m e n t s without contact for the paths of s y s t e m (A).
Through any point inside E , which is not the origin 0 and does not belong
t o one of the above s e p a r a t r i c e s , p a s s e s a path of the s y s t e m (A) which
e m e r g e s f r o m the r e c t a n g l e R through on2 of the s i d e s A B , A & as t d e c r e a s e s
o r through one of the s i d e s A A , , B B , a s t i n c r e a s e s .
FIGURE 20 I
5438 82
$,I. 5 LKL~CI'L'WL Sl'ABILITY OF .4 S.4DDLE POINT
-A11 the above relations and propositions c l e a r l y r e m a i n valid if a is

replaced by any s m a l l e r "concentric" rectangle with the s a m e diagonals
t~ = 5 k r , whose s i d e s a r e p a r a l l e l to the coordinate a x e s (the rectangle
.4'B'B,.4; in F i g u r e 20).
It is a s s u m e d in what follows that the number xo > 0 is sufficiently s m a l l ,
so that equalities (21)--(23) are applicable. We hence consider the
rectangle R corresponcing to t h i s x 0 .
L e tn n i a 2. Ij a dynainic systern ( A )has the canonical 30rtii
arid i s silff'iciently close to systerii ( A ) , the patter?, of paths of systeui (A) in

rectangle a i s sitnilar to the pattevn of paths of system ( A ) . Specifically,
on each of the sides AL: and d l B , of the rectangle R there is one point - b -
atid B,- through which passes an a-separatrix of the_ saddle 0 of systerniA),
and on the sides A-4 , and B B , there are points avid C, through tchich pass
cz - separatvices that cottstitute the contiriuatiun of the O - separatrices. The
sides of the rectangle 2 are segments without contact for the paths oj.
system (A), and each bath of systetn (A)passingthrough an interim point
of'the rectangle R which i s neither a separatrix nor the point 0 eitierges
frot)i R through one of [he sides A B , A ~ B a, s t decreases and through one of'
the sides A.4,, B B , a s t increases.
P r o o f . If s y s t e m (A) is sufficiently c l o s e to s y s t e m (A), h", and are x2
sufficiently c l o s e to kt .and I C , Cespectively. T h e r e f o r e E., > 0, x?
< 0 , and
0 is a saddle point of s,ystem (A). The point 0 (0.0) is a simple equilibrium
s t a t e of s y s t e m (A) (47.3, Definitio? 1 5 and T h e o r e m 11). Hence i t readily
follows that a l l dynami,z s y s t e m s (A) which are sufficiently c l o s e to (A)
have precisely one equ.ilibrium s t a t e in a,
namely the point 0 (0, 0). F u r -
ther a n a l y s i s of (16)-(20) shows that relations ana_logous t o (21)-(24)
apply to the paths x = 3 ( t ) , y = ( t ) of any s y s t e m ( A ) which is sufficiently
c l o s e to,@).:': The situation with r e g a r d to the pattern of paths of
s y s t e m (A) in rectangle is therefore exactly the s a m e a s for s y s t e m ( A )
\QT, J7.3), and the proof of the l e m m a is complete.
In the following two l e m m a s , (A) is a dynamic s y s t e m in canonical f o r m
which is sufficiently c l o s e to (A) f o r L e m m a 2 to apply. Let L,, L,,, L,, L , , be
the s e p a r a t r i c e s of s y s t e m (A) passing through the points D , D I , C, C i s
respectively (Figure 20), and t,,Ll,, E , , z,,
the s e p a r a t r i c e s of s y s t e m (A)
passing through the p o i n t s b , 6,. 2. c1, respecitvely, which lie on the s i d e s
of the rectangle fi ( E i t s on the side A B , 8,on the s i d e s A,B,, etc.).-
L e )?I ?ti a 3. For aiiy e > 0 , there i s 6 ==- 0 such that if systein ( A )i s
6 -close to system (-4) (2nd the separatrices L , and Z, pass through the
points D and d at t = t o , m y two points of these separatrices cmresponding
to the sattie t a t oaye distant less than e f r o m each other. A sitnilar
proposition applies to (2n-v of the separatrices L,,, L,. L , , .
P r o o f . The proof is c a r r i e d out f o r the s e p a r a t r i x L,. The s a m e
proof applies to the other s e p a r a t r i c e s .
Let e > 0 be given. Consider a rectangle R', "concentric" with R, which
is entirely contained i n U , , l ( 0 ) . L e t d ' . B', B;, and A ; be the v e r t i c e s of t h i s
* This is so because if system (x)is sufficiently close t o system (A), the functions 7, and 2~are arbitraril)
close to f i and g i , respectivel) (I= 1.21, as we see from equations (151, 95 and the corresponding
reiations for f i , 7,. gi.
83
rectangle; Dis the intersection point of the s e p a r a t r i x L, with the side AB

of the rectangle (Figure 2 1 ) . Suppose the s e p a r a t r i x L, p a s s e s through the
point D f o r t = to and through the point Dfor t = T > to. Let T , be some
number, T,> T.
By TheoTem 8 (53) t h e r e e x i s t s a segment K I K z of the side A B of the
rectangle R containing the point D and a number >6 , 0 with the following
property: if s y s t e m (A) is 8,-close to
s y s t e m (A), M,,is a point-of the segment K I K z ,
is the path of s y s t e m (A) which p a s s e s
through fi, a t t = t o , M ( t ) , 3 ( t ) the points of the
paths L, and L corresponding to the t i m e t , then
p (M (t), ( t ) ) < e f o r a l l f , f o < f < T i and a l l the
points M (T,)lie inside the rectangle R.
By L e m m a 5, $ 4 t h e r e e x i s t s a segment
K ; K ; of the side AB of the rectangle R con-
taining the point D ( F i g u r e 21) and a number
62 > 0 with the following property: a s Lde-
A, c r e a s e s , all the paths of any s y s t e m (A) which
FIGURE 21
is 6,-close to (A) passing through the points of
the segment K ; K ; c r o s s the side A B of the
rectangle R at the points of the segment K I K z .
Consider the paths L, and La of s y s t e m (A)
which a t s o m e t p a s s through the points K ; and K ; , respectively. Since t h e s e
points lie on different s i d e s of the s e p a r a t r i x L,, with the i n c r e a s e in f the
paths L, and Lz will c r o s s the diagonals OB and O A o f the rectangle R ,
will a l s o meet the diagonals OB and O A , respectively, with increasing t . Then

by Lemma 2 the intersection point Br of the s e p a r a t r i x &, with the side AB
of the rectangle R l i e s between K ; and K ; .
Let 6 be the s m a l l e s t of the t h r e e n u m b e r s 6,r82183. If s y s t e m (B)is
6 -close to s y s t e m (A), i t s s e p a r a t r i x L, evidently c r o s s e s the segment K ; K ;
(since 6 < I!$), and hence a l s o the segment K I K z (since 8 < 6 , ) . Let b be the
intersection point of the s e p a r a t r i x E, with K,K,. Suppose that the s e p a r a -
t r i c e s L, and &, p a s s through the points D and b , respectively, a t the s a m e
t i m e t o . Then, since 6<a5 for any t , to< t < Ti,we have p ( M ( t ) , lii ( t ) )< e.
All the points M ( t ) , n? I t ) of t h e s e s e p a r a t r i c e s corresponding to t > T,lie i n -
side the rectangle R, i.e., in U.lz (0). The distance between any p a i r of sEch
points is t h e r e f o r e l e s s than E . We have thus established that if s y s t e m (A)
is 6 -close to s y s t e m (A), p ( M i t ) , fi ( t ) )< e f o r any t , r o . T h i s completes the
proof of the l e m m a .
R e m a r k . L e m m a 3 can be generalized. Let I be any a r c without
contact that m e e t s the s e p a r a t r i x L, at a single point S, which does not
coincide with the endzpoints of 1 . Then f o r any e > 0 t h e r e e x i s t t 8 > 0,
such that-if s y s t e m (A) is 6-close to s y s t e m (A), the s e p a r a t r i x L, of
s y s t e m (A) m e e t s the-arc I a t a single point ge and p (S, 3) <e; if in addition
to the above, L, and L, p a s s through S and a t t = to, any two points of
these s e p a r a t r i c e s corresponding to the s a m e t i m e i > t o a r e a l s o distant
l e s s than e f r o m each other. This proposition c a n be proved without
difficulty. In what follows we w i l l only make u s e of the fact that if 8 is
sufficiently s m a l l , the points S and 3 (or D and a) are a r b i t r a r i l y c l o s e .
84
$3. STRK'CTL'RAL STABILITY OF A SADDLE POIKT
W e now proceed to the next le-mma. Retaining the notation of Lemma 3,

w e again consider the rectangle R and the s e p a r a t r i c e s L,, L t , , Z,,. L , , of
the saddle point 0.
On the s i d e A B of the rectangle R we choose two points S andS,lying on
the two s i d e s of the point D, and on AIBl w e choose two points S2 and S 3 lying
on the two s i d e s of D 1( F i g u r e 2 2 ) . Let
I.. L,, L 2 . L 3 be the paths passing through
t h e s e points. A s the p a r a m e t e r t i n -
creases, these paths meet the s i d e s A.4,
and BB1 of the rectangle R a t the points
T , T I , T Z , T B , respectively. Let H be the
region delimited by the s e g m e n t s
S S , , S2S3,T T 3 , T I T 1 and the a r c s ST and
&a
4s S i T I ( i = 1, 2, 3 ) of the paths L and L i .
H is a canonical neighborhood of the
equilibrium s t a t e 0 (QT, S19.2). The
s e p a r a t r i c e s L,,L,,. La, and L,, of
&& s y s t e m (A) partition the canonical neigh-
borhood H into four regular saddle-point
FIGURE 22
regions, which a r e designated u, U I . 0 2 , 0 3
( F i g u r e 2 2 ) . The region u is delimited
by the arcs OD and OC of the s e p a r a -
t r i c e s , the point 0, the a r c S T of the path L , and the s e g m e n t s DS and C T ;
the boundaries of the _other regions ui are s i m i l a r to the boundary of (5.
Consider s y s t e m ( A ) which is 6-close t o s y s t e m (A) and is a l s o given
in canonical f o r m . Let 6 > 0 be s o s m a l l that, f i r s t , Lemma 2 holds
t r u e and, second, the points b and B, [the intersection points of the
separatrices z, and I,, with the s e g m e n t s A B and A t B , ) lie inside the
segments-S,S and S2S3,respectively. Let z, z3
&, E,. be the paths of
s y s t e m (A) which p a s s through the points Sc S1, S t , S1, re2pectively (in
Figure 22, these paths and the s e p a r a t r i c e s of s y s t e m (A) a r e m a r k e d by
dashed lines).
Let 3 be the canonical neighborhood of th_e saddle point 0 of s y s t e m (A)
analogous to H . Like H , the neighborhood His partitioned by the s e p a r a t r i c e s
of the saddle point 0 of s y s t e m (8) into four r e g u l a r saddle-point regions a
aqd 01( i = 1,2,3), which a r e analogous to the r e g i o n s (5 and ut. All the
r e g i o n s u and u are a s s u m e d to be closed.
L e ) ) ?i)i a 4 . For any F > 0 , there exists 6 > 0, such that if systerit (A)
is 6 -close to system (A), we hatie
P r o o f . We will show that if 6 is sufficiently s m a l l , t h e r e e x i s t s a

mapping 8 (or Bi, i = 1, 2, 3, respectively) which m a p s u (and 0;)onto (gi);
the mappings 0 and et (a) are e-translations, (b) map paths into paths, and
( c ) e and e,, Bi and e2, 8: and e3# e 3 and 8 coincide on the arcs of the
s e p a r a t r i c e s OD, OCl. ODl, OC, respecitvely.
L e m m a 4 c l e a r l y follows directly f r o m t h i s proposition.
To fix ideas, let u s consider the region 5 and the corresponding
( F i g u r e 23).
85
a.IV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
I, IF \c A
I I
FIGURE 23 FIGURE 24
Let e > 0 be given. A s in Lemma 3, we consider the rectangle R' with

the v e r t i c e s A ' , B ' , B ; , A ; , which is "concentric'' with R and is entirely
contained in U e l z (0). In (5 we draw the path EF which m e e t s the s i d e s
A B , A A t , A'B', A'A; of the r e c t a n g l e s R and R' a t the points E, F, E ' , F',
respectively ( F i g u r e 24). Let I, 11, 111 be the elementary quadrangles
shown in F i g u r e 2 4 {their v e r t i c e s a r e TSEF, E'EDD', FF'C'C, respectively).
By L e m m a 9, 54.2, f o r any given e , t h e r e e x i s t s a p a i r of n u m b e r s
qt, 6, ( q ~ 6,,; 23.6,) f o r the quadrangle I (11, 111) with the Lollowing property:
if s y s t e m (A) is 6, (b2, g3)-close to s z s t e m (A), I (11, 111) is the corresponding
elementary quadrangle of s y s t e m (A), and mappings a r e given of the side?-
FE, E S , and ST of_the quadrangle I onto the r e s p e c t i v e s i d e s FE, &s, and ST
of the quadrangle I (or of the s i d e s DD'.DE, and EE' of the quadrangle I1 and
the s i d e s CC', C'F', and F'Fof the quadrangle III), which a r e qi-translations
(or qz-, q3-translations, respectiv-ely), t h e r z e x i s t s a Lath-conserving
mapping of the quadrangle I onto I (I1 onto 11, I11 onto 111) which is an
e-translation and coincides with the given mapping on the t h r e e s i d e s of the
quadrangle.
Let 6, > 0 be s o s m a l l that if system (x) is 6,-close to (A), the segments
EE', E'F', and F'F of the path of s y s t e m (A) can be mapped onto the r e s p e c t i v e
s e g m e n t s E2,%'p, and pF of the path of s y s t e m (A) through the point E, and
the segment ST can be mapped onto the segment ST by mappings which a r e
q4-translations, where q, < min {q,, q2,q3). The existence of such 8 , is seU-
evident.
Draw the path KL' of s y s t e m (A) which c r o s s e s the side A B of the
rectangle R a t the point K lying between D and E ( F i g u r e 25), and the
segment C'F' a t the point L'. F r o m QT, S7.3 it follows that i f point K is
sufficiently close to D, L ' i s a r b i t r a r i l y c l o s e to C'. We choose the point K
s o that i f the length of the segment DK is l e s s than <%, the length of C'L'
is l e s s than <%. Let 65 > 0 be s o s m a l l that the following condition is
satisfied: i f s y s t e m (&) is 6,-close to s y s t e m (A), we have 100"I 3,
I C'? I <$, the a r c DD' of the s e p a r a t r i x L, can be mapped onto the a r c ad'
of the s e p a r a t r i x by an q,-translation, and the a r c C'C of the s e p a r a t r i x La
can be mapped onto the a r c p c of the s e p a r a t r i x E , by an q,-translation.
86
5'. STRL'CTCRAL STASILITY OF A SADDLE P O I N T
The existence of such 6> follows f r o m Lemma 3 and f r o m g e n e r a l c o n s i d e r a -

tions concerning the phase p o r t r a i t s of erose s y s t e m s ( s e e Lemma 5 , 14.1).
Let finally & > O be such ?hat i f s y s t e m (A) is 6,-close to s y s t e m (A), the
paths of (A) and (A) passing through any point .lI of the segment KE w i l l
intercept with increasing t the s i d e A'A; of the rectangle R' a t two points
which are distant less th2.n q 3 f r o m e a c h other.
t:
FIGURE 2 5
We will now show thct if 6 is a positive number l e s s than any of o u r

8 , ( I g i 4 6 ) and s y s t e m (-4) is 6-close to s y s t e m (A), then
To prove t h i s we will establish the existence of a mapping 13 which

r e a l i z e s the relation (26). The mapping e is constructed in s e v e r a l stages.
\Ve will d e s c r i b e t h e s e s u c c e s s i v e s t a g e s by indicating what is mapped
onto what and to what closeness. When the mappings are extended, care
is taken to e n s u r e that paths are s t i l l mapped into paths.
1 ) S T onto S T . EE' onto EE', E'F' onto E'F', F'F onto PF, all t h e s e to
c l o s e n e s s q, < min {qt. q p . ql}.
-
2 ) E S onto ES identically (to z e r o c l o s e n e s s ) .
3 ) !Mappings 1 and 2 a r e extended to I: I onto I to c l o s e n e s s E .
4) D K onto 6 K ( c l o s e n e s s automatically < 2$ < q?), KEonto K E identi-
cally, DD' onto dB' to c l o s e n e s s q-.
5 ) Mappings 1 and 4 a r e extended t o 11: I1 onto I"I to c l o s e n e s s E .
6 ) K'L'onto I?Z' a r b i t r a r i l y , K'E'onto R'6' using the mapping which is
induced by the identical mnpping of KE onto itself.
7 ) Mappings 1 and 6 a r e extended to a mapping of the e l e m e n t a r y
quadrangle K'E'F'L' onto the quadrangle d'8'k'p. Since both quadrangles
lie i n L;, 2 (0), the c l o s e n e s s is automatically less than E .
8 ) 0 into itself, D'O onto 5'0.OC' onto O?, D'K' onto 3 3 b y the mapping
induced by mapping 4, DK onto dS.
87
ch. LV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
9 ) Mappings 6 and 8 a r e extended to a mapping of the r e g u l a r saddle-

point region L'K'D'OV"' onto the analogous region fiI?&OC?, This is feasible
by QT, 518.4. L e m m a 11. This mapping is an e-translation, since both
saddle-point regions a r e i n U.,2 (0).
1 0 ) C'C onto ?E to c l o s e n e s s q 3 , C'F'onto p p using the mappings
induced by the mapping of D'K' in 8 and the mapping of K'E in 6 . The
c l o s e n e s s is <q3.
11) Mappings 1 aii d 10 a r e extended to 111: I11 onto 111 to closenes
Mappings 3, 5, 7, 9, and 1_1 jointly define a mapping 8 of the r e g u l a r
saddle-point region u onto u . It is readily s e e n that 8 is an e-translation
and m a p s paths into paths, i.e., it h a s the p r o p e r t i e s ( a ) and (b). The
mappings ( i = 1 , 2 , 3 ) of u i onto riare constructed i n the s a m e way.
Note that on the a r c OD of the s e p a r a t r i x L , the mapping 0, should coin-
cide with_the given mapping 8. Therefore, i n constructing the mapping of
DD'onto DD' in 4, c l o s e n e s s t o q2 is not enough: we should e n s u r e c l o s e -
n e s s to q: < q ~ , and correspondingly r e p l a c e 62 by 8: < 8:. This is c l e a r l y
always feasible. A s i m i l a r r e m a r k applies to the mappings I2and e3. -The
s e t of four mappings 8, 8,, 02. can be t r e a t e d a s a mapping of H onto H .
This mapping r e a l i z e s the relation (25). The proof of the lemma is
completed.
Xh e o r e m 1 3 , An equilibrium state M o ( x o , yo) of the system
f o r which A < 0 (a saddle point) is structurally stable.

P r o o f . Theorem 1 3 is proved i n the s a m e way a s Theorem 1 2 ($8).
However, instead of L e m m a 1 and transformation (30), used in 5 8 in our
proof of Theorem 1 2 , w e should u s e L e m m a 4 and a l i n e a r transformation
which r e d u c e s the modified s y s t e m
to the canonical f o r m
(x)
If s y s t e m is sufficiently close to (A), t h i s transformation, by
Lemma 1, is a r b i t r a r i l y close to the identity transformation. The a r g u -
m e n t s used i n the proof of Lemma 1 2 t h e r e f o r e still apply. Q. E. D.
R e m a r k 1. F r o m Lemma 4 and the proof of Theorem 1 3 we s e e that
if H is a sufficiently s m a l l canonical neighborhood of the saddle point 0, then
for any E* > Othere e x i s t s 6* > 0 with the following property: if s y s t e m (A*)
is 6'-close to s y s t e m (A), then
e*
( H , A) = (H*, A*), (27)
where H* is the canonical neighborhood of the saddle point O* of s y s t e m (A").

Indeed, r e l a t i o a ( 2 7 ) follows f r o m r e l a t i o n ( 2 5 ) of L e m m a 4 and f r o m
-
the relation ( g ,3 ) ( H * , A * ) which is obtained in the proof of Theorem 13.
But x" is a canonical neighborhood, and H* and A* are obtained f r o m and A",
88
respectively, by a linear transformation (Tpecifically, by the transformation
which r e d u c e s ( M )to the canonical f o r m (A)). T h e r e f o r e H* is a l s o a
canonical neighborhood of the saddle point O*.
R e m a r k 2. A s we have a l r e a d y noted, i f Mo ( r o ,yo) is a s i m p l e saddle
point of s y s t e m (A), any s y s t e m sufficiently c l o s e to (A) h a s precisely one
s t a t e of equilibrium in a sufficiently c l o s e neighborhood of M0,which is
a l s o a saddle point.
10. STRUCTURAL INSTABILITY OF AN EQUILIBRIUhI

STATE W I T H PURE IRIAGINARY CHARACTERISTIC
ROOTS
1. investigation of a n ecpilibrium s t a t e with complex

c h a r a c t e r i s t i c roots (a r e v i e w )
\.+'ewill show in t h i s section that a n equilibrium s t a t e with p u r e

imaginary c h a r a c t e r i s t i c r o o t s is s t r u c t u r a l l y unstable. A s in the previous
sections, w e will study, ,without l o s s of generality, canonical s y s t e m s of
the form
-dr--:
0's --By-;-'F(r,y). -dty
-=pIT*(x,y),
(1)
dl
where $ li 9. W e always a s s u m e that $ > 0. and $ are functions of

c l a s s k ', 1 or analytical in c; they vanish a t the point 0 (0, U) together with
their f i r s t - o r d e r p a r t i a l derivatives.
System (1) is a particc.lar case of the s y s t e m
The phase p o r t r a i t of s y s t e m ( 2 ) n e a r the point 0 (0. 0) is studied in

detail in QT, 58. We will only s u m m a r i z e the corresponding r e s u l t s ,
which a r e needed in what follows. \Ve u s e the s a m e notation a s i n QT.
System ( 2 ) is investigated in polar coordinates, introduced by the
relations I = p cos 8, y = p sin 8. The transformation to polar coordinates
gives the s e t of equations
where
~ ( p e, ) = a p i g (pcose, psin8)cose+Ip(pcose, psine)sine,
tp (p cos 6, p sin 6) case- p ( p c 0 ~ 6 p. s i o 6 ) sin 8.
0 (p, 6)= - (4)
P P
It is f u r t h e r a s s u m e d that
a (0, e) = o (5)
for --a3 cB<+m. This condition e n s u r e s the continuity of the function 0.
89
I
Ch.IV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
System ( 3 ) c a n be reduced to a single equation
which is obtained by dividing the f i r s t equation in (3)through the second

equation. The right-hand side of equation ( 6 ) is written in abbreviated
f o r m as R ( p , 0):
I
System ( 3 ) and equation (6), and hence the function R ( p , e), a r e consi-
d e r e d in the s t r i p I
-p* <p <p* (8)
of the (p, 0) plane, where p* is a sufficiently s m a l l positive number.

E l e m e n t a r y calculations show that the function R (p, 0) h a s a continuous
partial derivative with r e s p e c t to p in the e n t i r e s t r i p ( 8 ) , and
for any 8. The existence of a continuous derivative of R with r e s p e c t to p

indicates that the existence and uniqueness t h e o r e m and the t h e o r e m of
continuous dependence on the initial conditions both apply to equation ( 6 )
in the s t r i p (8) (QT, Appendix, 48.2). T h e r e f o r e , for any eo and por
I po I < p* t h e r e e x i s t s a unique solution of equation (6)
satisfying the condition
f (eopeo, pol = p0. (11)
Solution (10) is defined i n s o m e (maximum) i n t e r v a l (e,, e,) which contains

the point Bo. Moreover,
f (0; eo, 0) = 0, (12)
s o that p = 0 (the a x i s 0 in the (0, p) plane) is a solution of equation (6). This

solution is defined f o r all 6, - w < 8 < w.
The family of paths of s y s t e m ( 3 ) i n s t r i p (8) coincides with the family
of i n t e g r a l c u r v e s of equation ( 6 ) . If
P=P(Q, e=ew
is a solution of s y s t e m (3), 2 is the corresponding path, and (po, e0)is a point
on this path,
P = f (0, eo, p0) (13)
is the equation of this path.
90
910. iTRC C I I R.:.L LYSr.ABlLITY c T AX EQLrIL[SRIK.\! STATE
The relation between the paths of s y s t e m ( 2 ) in the (I.y) plane, on the

one hand, and the paths o f s y s t e m ( 3 ) in the ( p . 0) plane (or, equivalently,
the integral c u r v e s of equation (ti)), on the other, amounts to the following:
the path p = 0 of s y s t e m ( 3 ) in the (p. 8 ) plane c o r r e s p o n d s to the equilibrium
s t a t e 0 ( 0 . 0 ) of s y s t e m ( 2 ) in the (s.y) plane. Let now t be a path of s y s t e m ( 3 ) ,
other than the a x i s 8, which lies in the s t r i p (8)and c o r r e s p o n d s t o the
solutionp = p ( t ) . 8 = 8 ( t ) ; let ( 1 3 ) be the equation of t h i s path. In the (z,y)
plane, t c o r r e s p o n d s to a path L of s y s t e m ( 2 ) lying inside a c i r c l e of
r a d i u s p* centered a t the point 0, which d e s c r i b e s the solution
z:-=p(t)cose(t), y=p(t)sine<t) (14)
of t h i s s y s t e m . Equation ( 1 3 ) c a n be considered as the equation of the

path L i n polar coordinates. Note that if L is a closed path, t h e r e is only
one path of s y s t e m ( 3 ) zorresponding to it; i f , however, L is not closed,
t h e r e is a n infinity of such paths corresponding to i t , with the p a r a m e t r i c
equations
p=p(t), tl=e(t)-+2kn ( k = 0 . 5 1, & 2 , ...;see QT, S 8.3).
The investigation of the paths of s y s t e m ( 2 ) in the neighborhood of 0 ( 0 , O )

is based on the following proposition:
For any E > 0 t h e r e e x i s t s > 0 , such that any path of s y s t e m ( 2 ) which
at t = t., p a s s e s through s o m e point .%Io i n C,, (O), o t h e r than the point 0 itself,
will c r o s s with the i n c r e a s e and d e c r e a s e in t e v e r y half -line 8 = const
without leaving U ,(0)( F i g u r e 2 6 ) (QT, 8.4, L e m m a 3 ) .
FIGLrRE 26.
T a k e any r a y 8 = Bo. p > 0 (or p < 0 ) . By the l a s t proposition, i f po is

sufficiently s m a l l (e.g., I po I < bo), solution ( 1 3 )
P = / (87 00. Po)
i s defined for a l l 8 , Flog8 .c Bo $. 2n.

T h e function
Pi = i (eo 3-k eo. Po) = fen (Po)
91
is known a s the s u c c e s s i o n f u n c t i o n on the r a y e = e O , since the

points M o (po, 0,) and M i(pi. eo) are two s u c c e s s i v e intersection points of a
path of s y s t e m ( 2 ) with t h i s r a y obtained for i n c r e a s i n g t (for p > 0; f o r
dB
< 0, < 0 and i n c r e a s e in e corresponds to a d e c r e a s e in t ) . Since eo is
any number, w e may take without loss of generality eo = 0, and f o r the
corresponding succession function fo (po) we simply write f (PO). Thus,
f (Po) = f (2n; 0, Po). (15)
Since f (e, eo, 0) E 0, we have
To determine the behavior of the path L of s y s t e m ( 2 ) which p a s s e s

through the point M o with the polar coordinates po, 0 (po > 0). consider the
function
If d (PO) = 0, the path L through M o is a closed path. If, however,

d (po) < 0 (d (po) > 0), L is a s p i r a l which f o r t + + 00 (or t -+ - 00 ) e i t h e r
approaches the equilibrium s t a t e 0 or tends to a closed path enclosing the
point 0.
The equilibrium s t a t e 0 of s y s t e m ( 2 ) is a stable (unstable) focus if and
only if f o r a l l sufficiently s m a l l po > 0, d (po) < 0 (d bo)> 0).
2. Calculation of the first focal value
Since the right-hand side R (p, e) of equation ( 6 ) h a s a continuous p a r t i a l

derivative with r e s p e c t to p. the solution of t h i s equation f (0; eo, po) is
continuously differentiable with r e s p e c t to po (QT, Appendix, 58.3). The
succession function f (po) t h e r e f o r e h a s a continuous f i r s t derivative, whose
value a t po = 0 we will now calculate.
Note that by definition f (e; 0, po) satisfies the differential equation
Differentiation with r e s p e c t to po gives
As we know (1131, 524, Theorem 16), the mixed p a r t i a l derivative in

the left-hand side is continuous and is thus independent of the o r d e r of
differ entiation. Theref o r e
92
5 IO. STRL'CTVFUL IPST.4BILITY OF AN EQCILIBRICrhl STATE
This is a v a r i a t i o n a 1 e q u a t i o n r e l a t i v e to the initial value po. It is

c l e a r l y a l i n e a r differential equation with the unknown $. F r o m ( 1 2 ) and
( 9 ) we see that f o r po=O equation ( 2 0 ) t a k e s the f o r m
Let u s d e t e r m i n e the corresponding initial condition. By (111, f(0; 0, po) = po.

Therefore
d i (e:0. p0) =
apo 1e=o (22)
Integrating (21) with the initial condition (22), w e obtain
Hence, using (17) and (15), we find

2n 2
d'(O)=e 6-1.
The number d ' ( O ) i s called the f i r s t f o c a l v a l u e of the

equilibrium s t a t e 0." WE. s e e f r o m ( 2 4 ) that for an equilibrium s t a t e with
p u r e i m a g i n a r y c h a r a c t e r i s t i c r o o t s , in p a r t i c u l a r f o r the
equilibrium s t a t e 0 ( 0 , O ) cif s y s t e m ( l ) , t h e f i r s t f o c a 1 v a 1 u e
is z e r o .
3. T h e t h e o r e m of the c r e a t i o n of a closed path f r o m a

multiple focus
An equilibrium s t a t e with p u r e imaginary c h a r a c t e r i s t i c r o o t s is called

a f o c u s (either s t a b l e o r unstable), a c e n t e r , or a c e n t e r - f o c u s
(QT, 5 8 . 6 ) .
D e f i n i t i o n 16. An equilibrium state of a dynamic system which has
pure imaginary characte:vistic roots and is a focus will be called a
multiple focus.
T h e o r e In 14 (theorem of the creation of a closed path f r o m a
multiple focus). If the equilibrium state Ma (io,yo) of a dynamic system (A)
is a multiple focus, then for any e > 0 and 6 > 0, there exists a system (A)
6 -close to ( A )which has at least m e closed path in the e -neighborhood
of l W 0 .
W e w i l l u s e this theorem i n o u r proof of the s t r u c t u r a l instability of an
equilibrium s t a t e with p u r e imaginary r o o t s . It is, however, a l s o of
c o n s i d e r a b l e independent significance.
P r o o f . As before, i t suffices to c o n s i d e r the equilibrium s t a t e 0 (0,O)
of the canonical s y s t e m
The i-th focal value is the number d i )(0) (if it exists. of course).
93
Ch. IV. EQUILIBKIUM STATES OF STKUCTUKALLY STABLE SYSTEMS
Let @ > O a n d s u p p o s e t h a t O ( 0 , O ) i s a s t a b l e f o c u s . A s w e h a v e n o t e d
a t the end of $10.1, t h e r e e x i s t s ro > 0, such that f o r a l l po. 0 < P O < ro, the
function d (p0) is defined and d (po) < 0 .
Let ri be some fixed number, 0 < r, < ro. Then
d (rl) <0.
Consider the modified s y s t e m
dx -
==ax-@By+cp(x, By),
(A)
~ = b + ~ B y + w Z Y)
,
with the corresponding functions F, etc., which a r e the analogs of the

functions F, a, etc., relating to s y s t e m (A). If a" is sufficiently s m a l l in
magnitude, s y s t e m (A) is a r b i t r a r i l y c l o s e to s y s t e m (A). It is readily
s e e n that the functions 3 and 8 , and t h e r e f o r e a l s o R", a r e a r b i t r a r i l y
close to the r e s p e c t i v e functions F, 0,and R ($10.1, (4)and (6)), a t l e a s t
to r a n k 0. Then by Theorem 1, Appendix, subsection 1, the solutions
f (0, eo, po) and r ( 6 , Bo, pa) of the equations $=R (p. e) and $=% (p, 6) a r e
a r b i t r a r i l y c l o s e to each other over a finite range of values of 8, and hence
the n u m b e r s d ( r , ) and z ( r , ) a r e a l s o a r b i t r a r i l y close. Using (25) we thus
conclude that i f ;is sufficiently s m a l l , we have
d"(rl) < 0. (26)
Let % be so s m a l l that inequality ( 2 6 ) is satisfied; we choose t h i s

number p o s i t i v e , a">O. Then, by (24)
-
$(O)=e 2n a6-1>0. (27)
Since d' (pa) is continuous,

-d' (po) > O
for a l l sufficiently s m a l l pa; in other
words, dc(po) is a n increasing function i n a c e r t a i n range. Hence, using
the equality di(0) = 0 ( s e e (15), (16), (17)), we conclude that f o r all suffi-
ciently s m a l l positive pOa B(pO)>o. In p a r t i c u l a r , f o r s o m e rzr O<rz<ri,
d"(rz)> 0. (28)
F r o m inequalities ( 2 6 ) and ( 2 8 ) and the continuity of 3 it follows that

t h e r e e x i s t s a t l e a s t one r3,_r2< r3 < r,, such that 2 (rs) = 0. This signifies
that the path L", of s y s t e m (A) passing through the point with polar coordi-
nates ( r 3 , 0) is a closed path. -
It is readily seen that i f a and rl are sufficiently s m a l l , the closed path
Eois entirely contained in U, (0). Indeed, l e t p = f (0; 0, p) be the g e n e r a l
solution of the equation $ = R" (p, e) corresponding to system (A). The
solution corresponding to the closed path L", is p = 7(e; 0, r3). The g e n e r a l
solution of the equation = R" (p, e) is p =f (e; 0, po), and f (8; 0, 0) 0
( s e e (12)). If a and rl a r e sufficiently s m a l l , (A) is a r b i t r a r i l y c l o s e to
94
$10. STRC CTCrR.-\L C\ST;\R[LLTY OF AS EQK'ILISRILrhl ST.-\TE
s y s t e m (A), the function R (p, 0) is a r b i t r a r i l y c l o s e to R (p, e), and r3 is

a r b i t r a r i l y c l o s e to zero. Therefore, by the t h e o r e m of continuous
dependence on the right-hand s i d e and the initial conditions (Appendix,
subsection 1, Theorem 2), f o r a l l '3, 0,<8,<2n, the difference 7(e; 0, r J ) -
f (e; 0 , I)), equal to 7 (e; 0, r 3 ) . is less than e, i . e . ,
U< ?(e; 0 , r3)< E .
T h i s indicates that the closed path zo

is entirely contained in U, (0). Q. E. D.
R e m a r k 1. Since by assumption 0 ( 0 , O ) is a f o c u s of s y s t e m (A),
t h e r e are no closed paths in a sufficiently s m a l l neighborhood Uof 0. On
a
the other hand, any s y s t e m (A) sufficiently c l o s e to (A) with > 0 h a s a t
l e a s t one closed path in c'. We w i l l s a y in what follows that this closed
path is c r e a t e d f r o m the multiple focus.
R e m a r k 2. Clearly, if r 3 is the l e a s t of n u m b e r s satisfying the
conditions r2 < r3 < r , , 2 ( r 3 ) = 0, the closed path Eo is a p r i o r i a l i m i t -
continuum f r o m the i m i d e (it may, however, a l s o be a limit-continuum
f r o m the outside, and thus a limit cycle). Similarly, i f r 3 is the l a r g e s t
of these numbers, Eo i s a p r i o r i a limit continuum from_ the outside.
R e m a r k 3 . In our proof of Theorem 14, s y s t e m (A) c a n be replaced
by the s y s t e m
whose vector field is obtained f r o m the field of s y s t e m (A) by rotating

through a n angle tan-" p (_seeend of 3). I @ > 0, the point 0 (0,O)is a
s i m p l e focus of s y s t e m (A), which is stable fo_r p > 0 and unstable f o r p < 0.
I 0 (0.0) is g stable focus of s y s t e m (A), and 2 is the function corresponding
to system_(x), we %an again find rl and r 2 , such that 0c r2 < rl < ro and
d Vi) < 0, d (rt) < 0, d ( r ? ) > 0.
The last inequality is satisfied f o r sufficiently s m a l l r? if p < 0. We thus
have the following prcposition: if 0 (0.0) is a multiple focus of s y s t e m (A),
a sufficiently s m a l l rotation of the vector field through a positive angle or
through a negative angle c r e a t e s a closed path in a n a r b i t r a r i l y s m a l l
neighborhood of 0.
4. Proof of s t r u c t u r a l instability
TI2 e o r e V I 15. The state of eqtrilibuiuiu ?\Io

( x o , yo) of the system
95
Ch.lV. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
Pr oo f . We will c a r r y out the proof f o r an equilibrium s t a t e 0 (0,O)of

s y s t e m (A) i n canonical f o r m (1) (by L e m m a s 1 and 2, S6.1, t h i s can be
done without l o s s of generality). Let 0 be a s t r u c t u r a l l y stable equilibrium
s t a t e of s y s t e m (A). According to Definition 14 (57.1) t h i s implies that
s y s t e m (A) is s t r u c t u r a l l y unstable i n some neighborhood H of the point 0.
where H can be made a r b i t r a r i l y s m a l l .
F r o m the definitiznof s t r u c t u r a l stability, for any E > 0 t h e r e is 6 > 0
such that f o r each (A) which is 8-close to (A) we have
(If,A ) A (2,4, (29)
where H i s s o m e region. The neighborhood H i s made so s m a l l that 0 is the

only equilibrium s t a t e of system (A) in it. The next s t e p is to make 8 s o
s m a l l that the region r?, obtained f r o m H by ,E-
translation, contains the point 0. Finally, (A) is
identified with s y s t e m ( 2 ) with sufficiently s m a l l
@
non-zero a.
Under these Zonditions 0 E H" and i t is an equili-
b r i u m s t a t e of (A). It follows f r o m (29) that, a s
H c o n t a i t s only one equilibrium s t a t e of (A),
s y s t e m (A) h a s only one equilibrium s t a t e in B,
namely the point 9. Since a # 0, the point 0 is a
I
focus of s y s t e m (A). - H e n c ~ ,in a sufficiently s m a l l
neighborhood of 0 s y s t e m (A) h a s no closed paths.
Under the e-translation realizing relation (29),
FIGURE 27 point 0 is mapped into itself (since t h i s is a path-
conserving translation). T h e r e f o r e s y s t e m (A) does
not have any closed paths in a sufficiently s m a l l
neighborhood of 0, e i t h e r . This indicates that 0 is neither a c e n t e r nor a
center-focus for (A), i.e., it is inevitably a focus.
We have shown so far that if 0 (0, 0 ) is a s t r u c t u r a l l y stable s t a t e of
equilibrium of s y s t e m (A), 0 is a multiple focus.
Let U be the neighborhood UT(O),where r is so s m a l l that s y s t e m (A) has
no closed paths in U and the only equilibrium s t a t e in t h i s neighborhood is 0;
moreover, s y s t e m (A) is s t r u c t u r a l l y stable in U. For a s m a l l r, t h e s e
conditions a r e satisfied since 0 is a s t r u c t u r a l l y stable multiple focus.
I
Let W denote U,,z (0)( F i g u r e 27). Fix E > 0, e (2, and let 8 > 0 be s o s m a l l
that if s y s t e m (A) is 8-close to (A), w e have
e - -
(u,A ) = (U,A ) ,
where @ i ssome region. System (z)
is a s y s t e m 8-close to (A) which h a s
a closed path & in W ; t h i s s y s t e m (A) e x i s t s i n v i r t u e of Theorem 14
(i.e., the theorem of creation of a limit-cycle f r o m a multiple focus).
Note that W c b.* Therefore @contains the closed path of s y s t e m (A),
and then, by (30), U contains a closed path of s y s t e m (A). This c o n t r a d i c t s
our choice of U. The contradiction e s t a b l i s h e s the fallacy of our original
* All the points of W are distant more than I

2
from the boundary of U. Therefore a I--translation cannot
4
move any point of W outside the "translated" region f. Also see foomote on p.61.
96
4 11. A SADDLE- TO-SADDLE SEP.IP..ATRLX
assumption, namely that the equilibrium s t a t e 0 of s y s t e m (A) is s t r u c -

turally stable. W e have thus proved that this equilibrium s t a t e is
s t r u c t u r a l l y unstable. Q. E. D.
R e m a r k 1. If ( A ) is a s y s t e m of c l a s s S, but not of class iV - 1 (dV> 1).
it can be t r e a t e d as a point in any of the s p a c e s R:, l,<lS,<.V, l , < r , < k
(see $5.1). If (A) is a:? analytical s y s t e m , i t c a n be t r e a t e d as a point of
any s p a c e R Z ( k is any natural number, I , < r , < k ) or of any s p a c e R:(ris
any natural number). Theorem 1 5 of s t r u c t u r a l instability r e m a i n s valid
r e l a t i v e to any of thesz s p a c e s containing s y s t e m (A). T h i s i s s o because
the modified s y s t e m (A) u s e d in o u r proof of s t r u c t u r a l instability differs
f r o m s y s t e m (A) by the analytical i n c r e m e n t s & and a y , and is t h e r e f o r e
p a r t of any space containing (A).
R e m a r k 2 . T h e c r e m s 11, 1 2 , 13, and 15 show that s t r u c t u r a l l y stable
equilibrium s t a t e s a r e simple nodes, foci, and saddle points. F r o m
R e m a r k 2 to Theorem 1 2 and R e m a r k 2 to T h e o r e m 1 3 i t follows that i f
point 0 is_a s t r u c t u r a l l y s t a b l e equilibrium s t a t e of s y s t e m (A), any
s y s t e m (A) sufficiently c l o s e to (A) h a s p r e c i s e l y one equilibrium s t a t e in
a sufficiently s m a l l neighborhood of the point 0, which i s a l s o s t r u c t u r a l l y
stable and is a point of the s a m e type as that of s y s t e m (A) (i.e., a node,
a focus, or a saddle point, respectively).
$11. A SADDLE-TO-SADDLE SEPXRATRTX
A s is known (QT, f 4 . 6 and 23.1), i f a dynamic s y s t e m considered in

a bounded closed region h a s a finite number of equilibrium s t a t e s , the a -
or o - l i m i t - s e t of any path of this s y s t e m is e i t h e r (a) a n equilibrium state,
(b) a closed path, o r ( c ) a limit continuum comprising a finite number of
s e p a r a t r i c e s , which a r e continuations of one another f r o m the s a m e
direction, and a finite number of equilibrium s t a t e s .
We would like to establish what limit s e t s the paths of s t r u c t u r a l l y
stable s y s t e m s have. In the preceding sections we have identified the
s t r u c t u r a l l y stable equilibrium s t a t e s . The s t r u c t u r a l stability of a closed
path is discussed in the next c h a p t e r . At t h i s stage, consider a structurally
stable s y s t e m with a limit continuum of the f o r m (c). This continuum
c o m p r i s e s the s e p a r a 7 r i c e s of s t r u c t u r a l l y s t a b l e saddle points (since other
s t r u c t u r a l l y s t a b l e equilibrium points have no s e p a r a t r i c e s ) , and each of
-
these s e p a r a t r i c e s tends to a saddle point f o r both f + 00 and t - t + 00.
Such a s e p a r a t r i x is s a i d t o g o f r o m o n e s a d d l e p o i n t t o
a n o t h e r , and f o r brevity we will refer to it a s a s a d d l e - t o - s a d d l e
s e p a r a t r i x . The two saddle points of such a s e p a r a t r i x a r e e i t h e r
distinct or coincident. We w i l l show that a s a d d l e - t o - s a d d l e
s e p a r a t r i x i s s t r u c t u r a l l y u n s t a b l e (Theorem 16). Hence i t
follows that the limit s e t of the paths of a s t r u c t u r a l l y stable s y s t e m is
e i t h e r a n equilibrium s t a t e or a closed path.
1. T h e behavior of the s e p a r a t r i x under v e c t o r field rotation

Let
97
FIGURE 28
Consider a modified s y s t e m (A:) of the f o r m
where p is a p a r a m e t e r . Clearly, i f p is sufficiently s m a l l in magnitude,

(A:) is a r b i t r a r i l y c l o s e to (A). System (A*) was considered at the end of
1 3 ( 2 3 . 2 , Lemma 3). It was shown that the s e t s of the equilibrium s t a t e s of
( A ) and (A*) coincide and that a t any point which is not a s t a t e of equilibrium
the field of s y s t e m (A) m a k e s a constant angle f3 with the field of s y s t e m ( A * ) ,
such that
F o r p > 0 this angle is positive. According to S9 ( s e e 5 9 , L e m m a s 2

and 3 and r e m a r k to Lemma 3), if (A*) is sufficiently c l o s e to s y s t e m (A),
i.e., if p is sufficiently s m a l l , 0 is a saddle point of s y s t e m (A")), w h e r e a s
s y s t e m (A) h a s a single s e p a r a t r i x L: of the saddle point 0; this s e p a r a t r i x
m e e t s the arc 1, and t h e r e is only one i n t e r s e c t i o n point C*. If L, is a n
a (o-)-separatrix of the saddle point 0 of s y s t e m (A), L: is a n a ( a - ) - s e p a r a -
t r i x of the saddle point 0 of s y s t e m (A*). Let s: be the value of the
p a r a m e t e r s corresponding to the point C*.
L e m m a 1 . If p > o and Lo is an a-separatrix of the saddle point 0,
then s,' > s,; i f , however, p > 0 and Lo i s a o - separatrix, then :s < so.*
\!'e naturally assume that the above condition is satisfied, namely that the paths of system (A) make
positive angles with the arc without contact 1.
98
S 11. .-3.4
I DDLE- TO-SADDLE SEP.4RA TRLX
P r 0 0 f . Kithout l o s s of generality we may take the saddle point 0 of

sylsteni ( A ) to coincide v;ith the origin and w r i t e ( A ) in the canonical f o r m
d t = > . , s i 9: ( 5 , = P (x,y),
(1)
2
_ = &g -!-'p (5, y) = Q (5. y).
w h e r e i.,> 0. A? < 0.
A s in J5, consider a sufficiently s m a l l rectangle with v e r t i c e s
-4, D . B1.A I ( F i g u r e 20). T h e paths of s y s t e m ( A ) i n this rectangle are shown
in Fi-gure 20. A s the s e p a r a t r i x L,, we take L,. W e will f i r s t prove a
p a r t i c u l a r case, when the a r c without contact 1 is the s i d e .4Ac of the
rectangle E.':: The positive direction along 1 is defined as the direction f r o m
A , t o A, and the p a r a m e t e r s is identified with y . Under t h e s e conditions,
the paths of syst-em (A) f o r m positive a n g l e s with A i . 4 .
Note that i f LA) is a s y s t e m sufficiently close to s y s t e m (l), then
1 ) s y s t e m (A) h a s a single equilibrium s t a t e 6 in the rectangle B, which
is a l s o a saddle point a n 3 is a r b i t r a r i l y c l o s e t o 0 (I),@);
2 ) the-sides of the rectangle a r e a r c s without contact f o r the paths of
s y s t e m (A), the paths le.iving R through the s i d e s A A I and BB, and entering
the rectangle R through A B and ?&;
3 ) the s e p a r a t r i c e s I;, and LI, of s y s t e m ( E ) c r o s s the s i d e s A.4, and
CHI a t the points and c,, respectively, which are a r b i t r a r i l y close to C and
C,, and the s e p a r a t r i c e s 2, and z,,
cross the s i d e s A B and d l B 1a t the points
fi and d,, respectively, .,vhich a r e a r b i t r a r i l y close t o D and D,.
The validity of propositions 1 and 2 above is self-evident.
T h e validity of proposition 3 follows f r o m propositions 1 and 2, the
theorem of continuous dependence on the right-hand s i d e s , and f r o m the
fact that ,a closed path cannot e n c l o s e only one equilibrium s t a t e of
s y s t e m ( A ) if it is a saddle point (see QT, 5 1 1 . 2 , C o r o l l a r y 1 of Theore_m 29 and
$11.4, TheoreniJO), so that R contains no closed paths of s y s t e m (A}.""
For s y s t e m (A) we t a k e s a s y s t e m (A'::) of the f o r m
Let 11 > 0 ; this number i s a s s u m e d t o be so s m a l l that conditions 1 , 2 , 3

above a r e satisfied. Then s y s t e m (A"') h a s a single equilibrium s t a t e in
a, v i z . , the point 0, which is a saddle point, and the s e p a r a t r i x L; of this
saddle point c r o s s e s the segment -41.4 a t a point C*. Let yo be the ordinate
of the point C, and y; the ordinate of the point C * . To prove the l e m m a ,
w e have to establish that y:7y,. Consider the path L * o f s y s t e m (A )
which p a s s e s a t t = t, through the point C. Since the s e p a r a t r i x L , m a k e s
a positive angle with L*, L*enters into t h e region W , limited by
the s i m p l e closed line UD,d,CO, a s t i n c r e a s e s ( F i g u r e 29). Take some
point .\I*f r o m I f ' which l i e s on the path L*. A s t i n c r e a s e s , the path L of
systeni ( A ) passing through X* c r o s s e s the segment .lt:l at the point A- which
is below the point C, and a s t d e c r e a s e s it crosses the segment f l i . l l a t the
p i n t .Y. Consider the region II-, limited by the s i m p l e closed c u r v e .Y.-ltK.V.
* It LZ resdiil;: seen that i t the rect..neIe is rufficicntly sm211, each cf its siJes, ZnJ in particular d d l ,
. c prev!ouc condinom Imposed on the arc !>ithout contiict t.
. ~ r u i ~ cthe
*. -Ihc reader ii a d ; w d tfi w o r k ou: the complete x o o f of p r o p e i r i o r i ?.
99
I
+z=&
Ch.1V. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTENS
The path L makes a positive angle with L* a t the point M*. and t h e r e f o r e a s
t d e c r e a s e s , L* p e n e t r a t e s into W,. A s t d e c r e a s e s f u r t h e r , the path L*
should e m e r g e f r o m t h i s region. The path
L*, however, may c r o s s neither the segment
K A 1 nor the a r c N K of the path L . T h e r e -
f o r e , a s t d e c r e a s e s , L* will c r o s s the
segment N A I . But then a l l the paths of
s y s t e m (A*) passing through v a r i o u s points
of the segment A J w i l l c r o s s the segment
4s
N A I a s t d e c r e a s e s , i.e., none of these paths
W' is the s e p a r a t r i x L: of s y s t e m (A*). Hence
N it follows that the s e p a r a t r i x L; c r o s s e s the
4w segment A I A a t a point C*above C, a s o r i g i -
nally stated. The corresponding proposition
FIGURE 29
f o r the o - s e p a r a t r i x L a i s proved along the
s a m e lines. Our l e m m a is thus proved for
a p a r t i c u l a r c a s e , when the a r c without
contact is the segment A I A (or B A ) . Now, by Lemma 10, 44.2, we s e e that
the l e m m a also holds t r u e for any a r c without contact 1 , which s a t i s f i e s
the relevant conditions. Q. E. D.
R e m a r k . It is c l e a r f r o m the above proof that the l e m m a r e m a i n s
valid i f s y s t e m (A*) is given in the f o r m
_
dr -
dt-'-PLf(5> Y)Q. $=Q+Pf(3tY)',
with f ( 5 , y) >0 everywhere in &, except the saddle point, where it may
vanish.
2. Proof of s t r u c t u r a l instability
T h e o r e m 16. A saddle-to-saddleseparatrix is a structurally

unstable path.
P r o o f . Let the path Lo be a l s o an a - s e p a r a t r i x of the saddle point 0 and
a a - s e p a r a t r i x of the saddle point 0'. To fix our ideas, let 0 and 0' be two
distinct points (the s a m e proof will apply i f t h e s e two points coincide).
Let the s e p a r a t r i x Lo be s t r u c t u r a l l y stable, i.e., s y s t e m (A) is s t r u c -
t u r a l l y stable in any sufficiently s m a l l neighborhood H of th_e s e p a r a t r i x Lo.
Then for any e > 0 t h e r e e x i s t s 6 > 0, such that i f s y s t e m (A) is 8-close
to s y s t e m (A), we have
( H , A) (3,
where gis s o m e region. Let d b e the distance f r o m the boundary of H to
Lo and choose e < d. Let T be the mapping which r e a l i z e s relation ( 2 ) (i. e.,
T is an e-translation which t r a n s f o r m s H into I? and maps paths into paths).
Under the mapping T, the s e p a r a t r i x Lo of system-(A) is c l e a r l y mapped
into a saddle-to-saddle s e p a r a t r i x L", of s y s t e m (A) which l i e s in H (the
-=
l a t t e r follows f r o m the inequality E d , so that the e-translation does not
move the path Lo out of H ) .
100
511. A S.4fiDLE-TO-SADDLE SEPAR4TRIX
LVe thus see that if L, is a s t r u c t u r a l l y s t a b l e p a t k o f s y s t e m (A), then,

for any neighborhood If of t h i s path, e v e r y s y s t e m (A) sufficiently c l o s e t o
( A ) h a s a saddle-to-sad?.le s e p a r a t r i x which is entirely contained in E l . W e
will show that the l a s t condition is not satisfied in a c e r t a i n neighborhood
H . T h i s contradiction a i l 1 evidently establish the validity of our t h e o r e m .
Let L , , L,, L 3 be s e p a r a t r i c e s of the saddle point 0, and L;,L ; , L; s e p a r a -
t r i c e s of the saddle point 0', none of them coinciding with Lo. On the
s e p a r a t r i x L, (.L'z) we cho3se a point dl, (Mi) and draw through t h i s point an
a r c without contact 1, (Z;)> whose end points d o not coincide with ill, (N;) ( I =
= 1 , 2, 3 ) . Let f u r t h e r the point M , (X)be the only common point of the
arc 1 ) ( 1 ; ) with the separa:rices of the saddle points 0 and 0' ( F i g u r e 30).
6 r,
FIGURE 30
H is taken to be a sufficiently s m a l l neighborhood of the path Lo; i t is

so s m a l l that 0 and 0' a r e the only equilibrium s t a t e s of s y s t e m ( A ) in t h i s
neighborhood, and the ar-s without contact Zi and 1; ( i = 1,2, 3 ) lie outside H.
Let -If0 be a point on the s e p a r a t r i x Lo. Through this point w e d r a w a n
a r c without contact I , , which i s . e n t i r e l y contained in H and which h a s no
common points with the s e p a r a t r i c e s of the saddle points 0 and 0'. other
than M0.
Consider the s y s t e m
-dd=tz P - p Q , 2d t= Q + p P .
Let p>U. If p is sufficiently s m a l l , the only equilibrium states of

s y s t e m (A) in H are 0 a n c 0; t h e s e equilibrium s t a t e s are saddle points,
and their s e p a r a t r i c e s E, and 1;( i = 1, 2 , 3 ) c r o s s the r e s p e c t i v e a r c s without
contact I , and 1; (in v i r t u e %f the r e m a r k t o L e m m a 3, S9). hloreover, the
saddle point 0 of s y s t e m ( A ) h a s a s e p a r a t r i x %, which c r o s s e s the a r c lo a t
the point .Go, and the saddle point O ' h a s a s e p a r a t r i x 2; which c r o s s e s the
a r c 1, a t the point .GA. By the l e m m a of the previous subsection, .q,and
lie on the arc-& on different s i d e s of the point .Ifo ( F i g u r e 30). We will prove
that s y s t e m (A) does not have in H a s e p a r a t r i x going f r o m saddle point 0 to
saddle point 0'.
Indeed, suppose that such a s e p a r a t r i x e x i s t s ( c a l l i t y). Clearly Y cannot
be one of the s e p a r a t r i c e s E, or z;
( 1 = I , 2 , 3), s i n c e t h e s e s e p a r a t r i c e s
Ch.1V. EQUILIBRIUM STATES OF STRUCTURALLY STABLE SYSTEMS
c r o s s the a r c s l t ( I ; ) and e m e r g e f r o m H . Hence y should coincide both with

z;,
and Le., a t s o m e t =&, it should p a s s through the point &,,and a t some
t =Ti it should p a s s through the point I@:. Clearly, G<?;. Let 5 @ ) b e the
intersection point of one of the s e p a r a t r i c e s Zi (Zi), the s e p a r a t r i x Z, (2:),
say, with the b o u n t a r y of region H . o We choose 3 ( g ) a s the f i r s t i n t e r -
section point of Zs(L,) with the boundary, i.e., a l l the points of this s e p a r a -
t r i x between Oand S (between Oand 2 , ) a r e e n t i r e l y contained in N. Con-
s_ider a simple closed line OMoM;OS7$0 made up of a r c s of the paths z3 and
Lj, the section a0M; of the arcl,, and the sections O a 0 and GiOof the
s e p a r a t r i x y. This line delimits a c e r t a i n subregion N*of H .
Let &,be the point of y corresponding to the t i m e &+T, and B i t h e point
corresponding to the time X-T, whece T is-a sufficiently s m a l l positive
number and %+.<<-.. The points Miand M , , a s i s readily seen, lie on
two different s i d e s of th_e a r c without contact lo. Therefore, one of t h e s e
points l i e s outside H* (M,in Figure 30), and the _other l i e s inside this
region. But then the path y, moving f r o m 2, to Miwith the i n c r e a s e i n t ,
should c r o s s the boundary of H*. This is impossible, however, s i n c e f o r
&+z<t ~ 2 i - r the s e p a r a t r i x cannot c r o s s itself somewhere alongOBOor
MAO. nor can it c o r s s the a r c without contact lo. Moreover, for any t , the
s e p a r a t r i x y cannot have common points with the a r c s Os and OF of the
separatrices z3and z,, nor with the boundary of region H . We have thus
proved that t h e r e is no s e p a r a t r i x in H going f r o m 0 to 0. Q. E . D.
C o r o 11 a r y. If a dynamic s y s t e m (A) is s t r u c t u r a l l y stable i n a
bounded region, the a- a n d o - l i m i t s e t s of any path of this s y s t e m are
either equilibrium s t a t e s or closed paths.
Indeed, i f s y s t e m (A) is s t r u c t u r a l l y stable in a bounded region, it only
has a finite number of equilibrium s t a t e s (Theorem 10, 07.2). Therefore,
by QT, S4.6 and S 2 3 . 1 , each l i m i t s e t of the s y s t e m is e i t h e r a n equilibrium
state, or a closed path, o r a limit continuum comprising saddle-to-saddle
paths. The l a s t possibility, however, is ruled out in virtue of Theorem 16.
R e m a r k 1. System (A)
-=P-pQ,
dz $=Q+pP
dt
i s a s y s t e m of the s a m e c l a s s a s the original s y s t e m (A), and for suffi-

ciently s m a l l p it is a r b i t r a r i l y c l o s e to (A) to any rank. Hence and f r o m
the proof of Theorem 16 it follows that a saddle-to-saddle s e p a r a t r i x of
s y s t e m (A) is a s t r u c t u r a l l y unstable path relative to any space R!?, R!,
containing (A) a s a point.
R e m a r k 2. In the proof of Theorem 16, s y s t e m (A) can be replaced
with
where the function f ( x , y) maintains a constant sign everywhere in(% except

the saddle points, where i t may vanish. This follows f r o m the r e m a r k to
L e m m a 1, S 11.1. Thus, i f p # 0 is sufficiently-small, the saddle-to-saddle
s e p a r a t r i x d i s a p p e a r s on passing f r o m ( A ) t o ( A f ) .
It is assumed, without loss of generality, that the boundary of B is a simple smooth closed curve.
I02
C h a p t e r I7
C L O S E D P A T H S IN S T R U C T U R A L L Y
S T A B L E SYSTEiVIS
INTRODUCTION
In the p r e s e n t chapter w e consider closed paths and classify them into

s t r u c t u r a l l y stable and s t r u c t u r a l l y unstable. The chapter is made up of
f o u r sections (412 through 415). In 512 we investigate the configuration of
paths in the neighborhood of a closed path Lo. T o this end, we draw a n arc
without contact 1 which i n t e r s e c t s L, and on this arc consider the s u c c e s s i o n
function f (n) and the function d (n) = f (n) - n, where n is the p a r a m e t e r
defined o n 1. It is es:ablished that e i t h e r I ) a c e r t a i n neighborhood of Lo
contains no closed pa.:hs other than L , i t s e l f , i.e., Lo is a limit cycle, or
21 a l l the paths passing through the points of s o m e neighborhood of Lo are
closed paths, or finally 3 ) any a r b i t r a r i l y s m a l l neighborhood of Lo con-
tains both open and closed paths, other than Lo itself. If the s y s t e m is
analytical, only cases 1 and 2 a r e possible. The p a r t i c u l a r topological
s t r u c t u r e of the dynamic s y s t e m i n a neighborhood of the path Lo depends
on the p r o p e r t i e s of the function d (n); particularly significant is the
quantity d' (n,,),,where no is the value of the p a r a m e t e r on the a r c 1 c o r r e -
sponding to the intersection of i with L o .
In 413, a relatively simple s y s t e m of curvilinear coordinates is i n t r o -
duced in the neighborhood of the closed path Lo which m a k e s i t possible t o
calculate d' ( n o ) . This s y s t e m is introduced a s follows: through e v e r y point
.I1 (s) of Lo. corresponding to the time s, a segment of the normal is passed,
and to the points on this segment w e a s s i g n the coordinates r and n, where
n is the value of the p a r a m e t e r on the normal. The arc I is identified with
one of these n o r m a l segments. It is established that i f the equations of the
c l o s e path L, a r e s = q ( t ) , y = $ ( t ) , where and 9 are periodic functions of
period T , then
T
i [r;tocti. t ( i ) k Q k t W s ~ .W ~ ) CdsI
d' (i)=e'' -l=eJ-1.
Lf the integral J in this equality d o e s not vanish, Lo is a n isolated closed

path, i.e., a limit cycle, and i t is called a s i m p 1e 1i m i t c y c 1 e . It is
proved that if J < Lo is a stable limit cycle, and i f J > 0, Lo is a n unstable
limit cycle.
If L , i s a l i n i i t c y c l e a n d J = u , L , i s c a l l e d a m u l t i p l e l i m i t c y c l e .
In 414 i t is proved that any simple limit cycle is a s t r u c t u r a l l y stable
path of a dynamic s y s t e m (Theorem 18). Finally, in 515 we consider a
I03
ch.v. (ZLOSED I'ATHS IN STRUCTURALLY STABLE SYSTEMS
closed path Lo with J = 0 and prove that t h i s path is s t r u c t u r a l l y unstable.

The proof u s e s Theorem 19, which is of considerable independent i n t e r e s t
(this is the theorem of c r e a t i o n of a closed path f r o m a multiple limit
cycle).
Note that the investigation of a s t r u c t u r a l l y unstable closed path h a s
much i n common with the investigation of a n equilibrium s t a t e with p u r e
imaginary c h a r a c t e r i s t i c roots, presented in 10.
512. A CLOSED PATH AND ITS NEIGHBORHOOD.

SUCCESSION FUNC TION
1. Xntroduction of the s u c c e s s i o n function
Let
be a dynamic s y s t e m of c l a s s N or a n analytical system, defined in z.

Suppose that (A) h a s a closed path Lo in G. Let
z=(P(t), Y = c P W (1)
be the motion corresponding to t h i s path. 'p (t)and @I ( t )a r e periodic

functions of the s a m e period, which we denote z (r> 0). W e are i n t e r e s t e d
in the configuration of the paths of s y s t e m (A) in the neighborhood of Lo.
Let e, > 0 be so s m a l l that U. (Lo)does not contain any equilibrium s t a t e s of
(A) (this eo e x i s t s since Lo is a closed path).
On L o r we choose some point M o and p a s s through t h i s point a n a r c
without contact I which is contained i n U, (Lo),ensuring that M o r e m a i n s
a n inner point of 1. On the a r c I , we define s o m e p a r a m e t e r n . T h i s c a n
be done, say, by specifying the p a r a m e t r i c equations of the a r c 1 :
5 = gl(n), Y =gz (n). (2)
The p a r a m e t e r n = no c o r r e s p o n d s to the point Mo on the a r c 1 .

Let q > 0 be so s m a l l that all the paths c r o s s i n g a t t = to the p a r t of the
a r c 1 corresponding to
no-q 4n 4no +q (3)
c r o s s the a r c 1 again f o r t g r e a t e r than tor without leaving U,, (Lo)

p r i o r to
the second intersection. T h i s q e x i s t s by QT, 53.8, L e m m a 13. In t h i s
way, on the p a r t of the a r c 1 corresponding to the values (3) of the p a r a -
m e t e r n , w e have defined a s u c c e s s i o n f u n c t i o n
G = f (n), (4 )
constructed i n the direction of increasing t (QT, 53.8).
I 04
f; 12. hE[GhBOP.HOOO LF .A CLOSEU P-II'H. SL'CCESSLOh F C h C r l O N
Let M' be the poinl. of 1 corresponding to n = no - q , L' the path of (A)

passing through .\I,, ;7 the point a t which L c r o s s e s the a r c 1 the second time
with i n c r e a s i n g t . The point .V'evidently
c o r r e s p o n d s to n = f (no - q). .Y is e i t h e r
I different f r o m M'or coincides with it. If
the two points coincide, L' is a closed path,
and w e denote by r' the annular region b e -
tween the paths Lo and L'. If M' and h ' a r e
different points, L ' i s not a closed path, and
r' is the annular region limited by the path L o
and the closed c u r v e consisting of the turn
hft.v#of the path L'and the p a r t .M',Vof the
a r c I (Figure 31).
Similarly, taking M"to be the point of the
+
a r c t corresponding to n = no q, w e i n t r o -
FIGURE i I duce a path La, a point N", and a region r".
analogous to r'( F i g u r e 31). Regions of this
type w e r e considered in QT, 5 3 . 9 . F r o m
QT, , 5 3 . 9 , L e m m a 14 and Remark 1 to the l e m m a , we c l e a r l y see that
1 ) regions r ' a n d L-" together with t h e i r boundaries a r e contained in
('eo (Lo);
2 ) e v e r y path passing through a point of r ' c r o s s e s the segment MoN'(or
No.V', i f S'lies on I between Mo and M ) of the a r c 1 for both i n c r e a s i n g and
d e c r e a s i n g time t . Similarly, e v e r y path passing through a point of I'"
c r o s s e s the segment Mo4VR(or M O W of ) the a r c I for both i n c r e a s i n g and
decreasing t .
2. The configuration of paths i n the neighborhood of a

closed path
Let u s first e s t a b l i s h the possible configuration of a single path passing

n e a r a closed path. To this end, w e take a path L*which for t = to p a s s e s
through a point M * o f the a r c I corresponding to the value of the p a r a m e t e r
n*. I n* - no I < 1 ' . Let N** be the succeeding intersection point (the
s u c c e s s o r of M * )of the path L* (with the a r c I ) . Ai** c o r r e s p o n d s to n = f (n*).
We introduce a n auxiliary function
J (n)= f (4--n, (5)
which is analogous to the function d (p) used in the investigation of a multiple

focus ( S l O . 1 , (17)). The function d (n) is a p r i o r i defined for a l l the values
of n satisfying inequality ( 3 ) , I n - no )<q. A s we show i n QT, S3.8, Remark I
to L e m m a 13, if s y s t e m (A) and function ( 2 ) belong to c l a s s N ( o r to the
analytical c l a s s ) , the function f (n), and hence d (n), a r e functions of the c l a s s
.V (or analytical).
The following case:; should be considered:
1 ) d (n*) = 0 , i.e., n* = f (n*). In this c a s e M* and iM** coincide, and L* is
a closed path completely contained in I" and r".
-=
2 ) n* > 0,d (n*) 0 or n* < 0, d (n*) > 0 . T o fix i d e a s , l e t u s consider
:.
the c a s e n* > 0, d (n*) 0. In t h i s c a s e , L* is not a closed path, M** is a point
105
$12. .hE[Gfih~Ri:OC~D
CIF .A CLOSED P.4TS. SL'CCESSI<>X F C l C TIQX
path contained completely in r"(r'), or to L o . A s i n c r e a s e s ( t > t o ) , the

path L* e i t h e r l e a v e s '?I (r')or g o e s to a closed path in f" (I?) which is
different f r o m Lo.
C a s e s 1,2, 3 above c l e a r l y c o v e r all the types of paths which c r o s s the
section WM' of the arc without contact I . Every path passing through a
point of r' and r" cro:;ses t h i s section M'M" of the a r c 1 e i t h e r with increasing
or with decreasing t . We have thus established the possible configuration
of the paths which p a s s through points of r' and I".
Let u s now c o n s i d e r the topological s t r u c t u r e of the dynamic s y s t e m (A)
in the neighborhood cf a closed path L o - To this end, w e again introduce
the function d ( n ) = f ( 1 2 ) - n. Since n = no c o r r e s p o n d s to the point M oof the
closed path Lo on the a r c without contact, w e nave d (no) = 0. The following
two cases should be considered:
a ) t h e r e e x i s t s m > U, m < q , such that d ( n ) $ 0 f o r all n , 0 0 , t h e r e e x i s t s n, 0 < [ n - no I < m , such that d ( n o ) = 0 .
Let u s consider case a in s o m e detail. A11 the paths passing through the
points of the arc 1 which correspond to the values of the p a r a m e t e r n ,
0 < I n - no[<<m, are not closed.
The closed path L i in this case is isolated, i.e., i t is a limit cycle (QT,
C4.9). Because of the continuity of d ( n ) , we conclude that in case a, d ( n ) h a s
the s a m e sign f o r all positive n and the s a m e sign for a l l negative n (I n 1 < m ) .
The following four s u b c a s e s are therefore possible :
aLjd(n)<O for n>no, d ( n ) > O for n<no;

a 2 ) d ( n ) > O for n>no, d ( n ) < O for n < n o ;
a s ) d ( n ) < O f o r n>no, d ( n ) < ' J f o r n<no;
a 4 J d ( n ) > O for n>no, d ( n ) > O for n<no.
In c a s e al, all the paths passing f o r a sufficiently s m a l l 6 through the

points of Ua ( L o )which do not coincide with Lo go to Lo f o r t -+ + 00, w h e r e a s
for decreasing f, they leave Ug ( L o ) . The limit cycle Lo is s t a b l e , and w e
s h a l l s a y that all the paths sufficiently c l o s e to L o , which do not coincide
with L o , w i c d o n t o L o .
In case a 2 , all the paths sufficiently close to Lo (which do not coincide
with L o ) go to Lo for t - + - 03, and f o r increasing t , they leave U6( L o ) . In
this case, the limit cycle Lo is u n s t a b 1e . We shall s a y that all these
paths paths u n w i n d f r o m L o .
In cases a3 and a 4 > all the paths which p a s s through points of U.,( L o ) and
lie on one side of Lo go to Lo for t -+ f00 and leave Lr6 ( L o )a s t d e c r e a s e s ; the
paths lying on the o t h e r side of L o , go to Lo f o r t -+--co and leave Ua(Lo)as
t increases.
The l i m i t cycle i n cases a3 and a4 is s a i d to be s e m i s t a b 1e .
The
paths c l o s e to the sernistable cycle L o ,xind onto L,, on one side and
unwind from Lo on the other s i d e of the cycle.
Let u s now consid,sr case b. Any neighborhood of Lo contains a n infinite
number of closed p a t i s in this case. Without going into a detailed
analysis, we will only note that this case, like the analogous case of a n
equilibrium s t a t e with p u r e imaginary c h a r a c t e r i s t i c values ( s e e QT, S8.6),
c o v e r s a n infinity of .iarious possible topological s t r u c t u r e s of the neighbor -
hood of the closed path L o . Any p a r t i c u l a r topological s t r u c t u r e is c o m -
pletely determined by the p r o p e r t i e s of the function d ( n ) and depends on the
I07
Ch.V. CLOSED PATHS IN STRUCTURALLY STABLE SYSTFMS
s t r u c t u r e of the s e t of i t s r o o t s and on the signs of the function d (n)when-

e v e r it does not vanish. In p a r t i c u l a r , if f o r sufficiently s m a l l R , d (n) = 0,
a l l the paths through the points of a sufficiently s m a l l neighborhood of Lo a r e
closed.
3. The case of a n analytical dynamic s y s t e m
W e w i l l now consider in g r e a t e r detail the c a s e of an analytical s y s t e m

(A) and establish a l l the possible topological s t r u c t u r e s of the neighborhood
of a closed path. We will derive the conditions to be satisfied by the
s u c c e s s i o n function i n the neighborhood of a closed path Lo i n o r d e r f o r t h i s
closed path to be a l i m i t cycle of a c e r t a i n type (stable, unstable, or s e m i -
stable). The arc I (Le., the functions (2)) is assumed analytical. Then,
i f (A) is analytical, the succession function n = f (n) is a l s o analytical
(QT, 53.8, Remark 1 to L e m m a 13). A s before, we a s s i g n to the closed
path L, the p a r a m e t e r n = no. so that
=no
f (no) for d (no)=0. (6)
In the c a s e of an analytical dynamic s y s t e m (A), we have to consider two

possibilities:
1) At l e a s t one of the derivatives d(') (nn)does not vanish, i.e., t h e r e
e x i s t s a natural number k > l such that
d' (no)= d" (no)= = dtk-" (no).

I0, dk)(no)# 0. (7)
In t h i s c a s e , expanding the function d (n) around n = no in powers of n - nor

we get
d (n)=(n--no)k[d'k)(nO)+(n-n~) d'k"'(no)+. ..I=
= (n-no)k [d(')(no)+(n-no) CD (n)], (8)
where Q, (n) is some analytical function. It follows f r o m ( 8 ) that the sign of

d (n)in the neighborhood of no coincides with the sign of the number
(n - no)kd ( k )(no). Hence it readily follows that f o r a n even k, we have c a s e as
o r a 4 (according as dch)(no)is negative or positive), i.e., a s e m i s t a b l e l i m i t
cycle is obtained. F o r a n odd k and d(k) (no) 0, we have c a s e al, i.e., a
stable l i m i t cycle, and finally f o r odd k and d' (no)> 0 , we have c a s e az, i.e.,
a n unstable limit cycle.
Let u s consider s e p a r a t e l y the c a s e k = 1, i.e., when
d'(no)=f'(no)-1#O. (9)
The closed path Lo in this c a s e is called a s i m p l e l i m i t c y c l e . A

simple l i m i t cycle is evidently either stable o r unstable, according as d' (no)
is positive o r negative, or, equivalently, according a s f' (no)< 1 o r f' (no)> I.
If k > 1, the closed path Lo is called a m u l t i p l e l i m i t c y c l e o f
m u 1 t i p 1i c i t y k .* The multiplicity of the cycle evidently coincides with
the multiplicity of the root n = no of the function d (n).
These definitions of a s i m p l e l i m i t c y c l e and a m u l t i p l e l i m i t c y c l e o f m u l t i p l i c i t y k
are linked with the choice of the a m without contact 1. For these definition t o be meaningful, we have
t o show that they are independent of t h e particular choice of t h e am. No proof of this w i l l be given here,
but later (913) a n invariant definition of a simple limit cycle will be formulated.
108
1.12. \E[GI,BORPOOD OF .1 CLOSED P.l T t l . SC'CCESSION FClKCTIOh
2 ) All the derivati.vesd(') (no)a r e z e r o ,
d' (no)= d" (no)=: ... = d'" (no)= ... = 0.

Then, by analyticity,
d (n)= 0,
i.e., a l l the paths through the points of a sufficiently s m a l l neighborhood of

the path L o are closed.
We have thus shown that any closed path of an analytical dynamic s y s t e m
is a s i m p l e or a multiple l i m i t cycle, or e l s e all the paths through the
points of a sufficiently s m a l l neighborhood of the path L o a r e closed. In the
c a s e of a n analytical dynamic s y s t e m , the n e c e s s a r y and sufficient condition
f o r a closed path Lo t,2 be a l i m i t cycle is that a t least one of the d e r i v a t i v e s
dtcl (no) (i>; 1) vanishes.
4. The case of a nonanalytical dynamic s y s t e m
Let u s now c o n s i d e r the case when a dynamic s y s t e m (A) is not analytical

and belongs to the c l a s s .V. T h e a r c I (i.e., the functions ( 2 ) ) is a l s o
a s s u m e d to belong to the c l a s s Z. Then, by QT, 53.8, R e m a r k 1 to
L e m m a 13, the s u c c e s s i o n function n = f (n) is a l s o a function of class N .
In this c a s e , proceeding along the s a m e l i n e s a s before, we c a n find the
sufficient conditions .lor a closed path to be a l i m i t cycle. Indeed, suppose
that not all the d e r i v a t i v e s of the function d (n)
d' (no), .. ., d'" (no)
vanish at n = n o , and t h e r e is a number k, l,<k,<.V, such that
d' (no)= . .. =dtk-" (no)=0, dk'(no)# 0. (10)
F r o m this condition z.nd f r o m the equality d (no)= 0 , we obtain using Taylor

expansion
where O<e<l.
Since the k-th derivative is continuous, the s i g n of the coefficient of
(n - no)k in (11) f o r n - no of sufficiently s m a l l absolute value coincides with
the sign of dk)(no). Then it follows f r o m ( l l ) , as in the previous section
f r o m (8), that we obtain one of the c a s e s al, a z r a3, and a4 according as k is
even or odd and d k )(no)is positive or negative.
A s f o r a n analytical dynamic s y s t e m , we c o n s i d e r the case k = I
separately.
In this c a s e , the path is called a s i m p 1 e 1i m i t c y c 1e , as before.*
T h e condition that a t l e a s t one of the n u m b e r s d ( k ) (no),k = 4 , 2, ..., N d o e s
not vanish is sufficient, but not n e c e s s a r y , for the closed path Lo of a
s y s t e m (A) of c l a s s .V to be a l i m i t cycle.
' Sec footnote t o the definition of simple and multiple limit cycles in 512.3.
109
Ch.V. CLOSED PATHS I N STRllCTIlRAI
If for a s y s t e m of c l a s s N
d (no)=d (no)= 6 . . =
d
additional information r e g a r d i n g the function

draw definite conclusions about the topologic
s y s t e m n e a r the closed path Lo.
113. CURVILINEAR COORDINATES IN THE NEIGH-

BORHOOD O F A CLOSED PATH. SUCCESSION FUNCTION
O N A NORIbIAL TO A PATH
1. Curvilinear coordinates i n the neighborhood of a

closed path
In the previous section we demonstrated that the s u c c e s s i o n function f (n)

on an a r c without contact 1 (or the function d ( n ) = f (n) - n ) and the v a l u e s of
i t s d e r i v a t i v e s a t the point no corresponding to a closed path Lo are of the
g r e a t e s t importance f o r investigating the topological s t r u c t u r e of the
dynamic s y s t e m in the neighborhood of Lo. To calculate the d e r i v a t i v e s of
the s u c c e s s i o n function, we introduce an a u x i l i a r y s y s t e m of c u r v i l i n e a r
coordinates in the neighborhood of L o . This s y s t e m is analogous to the
p o l a r s y s t e m of coordinates, and the t r e a t m e n t that follows is not unlike
the a n a l y s i s of the paths n e a r a focus (110).
Let
be a dynamic s y s t e m (of c l a s s N o r analytical), Lo a closed path of this

system,
z=(P(t)t Y=lp(t) (1)
the motion corresponding to the path L o , T > 0 the period of the functions p
I and Ip. Note that the functions cp and ~p, being solutions of a s y s t e m (A) of
c l a s s N (or a n analytical s y s t e m ) , are functions of c l a s s N + 1 (or
analytical). In p a r t i c u l a r , they a r e a p r i o r i known to have continuous
second derivatives. In what follows, the p a r t i c u l a r choice of the a r c without
contact f o r the construction of the s u c c e s s i o n function n e a r the closed
path i s i m m a t e r i a l . We t h e r e f o r e choose 1 as the s i m p l e s t of these
a r c s , namely a segment of the n o r m a l to the path Lo. We w i l l c o n s t r u c t a
s y s t e m of c u r v i l i n e a r coordinates i n the neighborhood of Lo which is the
most convenient for o u r purpose.
Through e v e r y point M (cp (s), Ip (s)) of the path Lo we draw a n o r m a l to the
path a t that point and lay off s e g m e n t s of length 6 Vcp (s)I + 9 (s)* on e i t h e r
side of Lo along the n o r m a l .
L e m m a 1 . If 6 > 0 i s sufficiently small, no two segments of normals
drawn through different points of L~ have any points in comman.
The proof of this l e m m a using the usual compactness considerations and
the existence and continuity of the second derivatives of the functions q and
'1 is given in the Appendix, subsection -1:.
m
W e will take 6 > (J to be so s m a l l that L e m m a 1
is satisfied. Then the ends of the s e g m e n t s of
length 6 1 q' ( s ) - ~ q' (s)'-laid off the n o r m a l on the
positive (negative) side of the path L,, f o r m a s i m p l e
closed c u r v e I?, (r2). T h e c u r v e s F, and r- are
"concentric" with the path Lo and enclose a region 0
of the plane (2,y). Q is c l e a r l y homomorphic to a n
open c i r c u l a r ring (Figure 3 6 ) .
I'IGi,R .n For a sufficiently s m a l l 6 , s y s t e m ( A ) h a s no
t:quilibrium s t a t e s in SI. We w i l l a s s u m e that this
condition is a l s o satisfied.
M'e introduce the functions
, s h e r e s stands f o r the time t . These functions are defined on the e n t i r e

plane (s, n), but w e w i l l only consider them i n the s t r i p
- - ~ < ( s < ~ x . -6<n<8. (3)
The functions (F and 3; c l e a r l y have the following p r o p e r t i e s :

I ) i f ( A ) is a n a n a l y t x a l s y s t e m , (F and 3 a r e analytical functions; i f
[A) is a s y s t e m of c l a s s .\, and a r e a l s o functions of class .V, and
m o r e o v e r q a n d T have continuous derivatives with r e s p e c t t o n of all o r d e r s
.b..l - dk+l -
and continuous mixed derivatives q
en -
4 for any
, ,.SI &I,751
-
k and s.<-\.
2 ) The functions (p and $ with their partial derivatives a r e periodic
functions
- of s of period- T.
3 ) q (s. 11) q (s). 11 (s, 0 ) = J' (s), i.e., f o r n = i j the equations . r = c ( s , n ) ,
Z-$(s. n ) a r e the p a r a m e t r i c equations of the path Lo where the p a r a m e t e r s
coincides with t .
4) The functional determinant
for n = 0 is equal to
i.e., for n=O i t does not vanish for any s , - - s o < s < + m .
F r o m property 4 and the compactness of the segment u,<s,<T, using the
periodicity of the functions 9, and $ in s, we see that for all sufficiently
sniall 1 1 ,
1 ( S . n) =+u.
(5)
Finally, i t is readily s e e n that if 6 is sufficiently s m a l l , all the n o r m a l

s e g m e n t s a r e a r c s without contact for the paths of (A).
I
Ch.V. CLOSED P A T H S IN STRUCTURALLY STABLE SYSTEMS
Let 6 be s o s m a l l that a l l the above conditions a r e satisfied. To help

the r e a d e r , we l i s t them h e r e again.
1) The segments of n o r m a l s of length 26 vq' (s)I f q ' ( s ) ~ I
different points of the path Lo do not i n t e r s e c t . Hence it follows, i n
p a r t i c u l a r , that the region 8 is homomorphic to a c i r c u l a r ring.
2 ) All these n o r m a l segments a r e a r c s without contact f o r the paths of .
(A).
3 ) T h e r e a r e no equilibrium s t a t e s of (A) i n 9.
4 ) The determinant A (s, n) does not vanish in the s t r i p (3).
Consider the mapping defined by the equalities
z=G(s, = 'P (s) +nrp' (SI,

(6)
g=$(s, n) =Q(s)-nq'(s).
I
It h a s the following p r o p e r t i e s :
--
a ) Mapping ( 6 ) maps the s t r i p ( 3 ) of the (s. n) plane onto the annular
, . -, *
b ) The a x i s n = 0 is mapped by ( 6 ) into the path L o , the l i n e s n =
= C, 0 < I c 1 < 6 p a r a l l e l to the axis s are mapped into nonintersecting s i m p l e
closed c u r v e s which lie one inside the other i n 8 (these c u r v e s a r e
''concentric" with L o ) .
c ) The segments s = const, - 8 < n < 6 , of the s t r i p ( 3 ) a r e mapped into
n o r m a l segments to & ( F i g u r e 37).
FIGURE 37
d ) All the points (s, n) with the s a m e value of n and s differing by

multiples of the period T, i.e., all the points of the f o r m (s k T , n), +
k = 0 , f 1, f 2 , ..., a r e mapped into one point of the plane (5, u).
e ) The mapping ( 6 ) is now one-to-one. However, f o r any fixed so, i t is
one-to-one on the "half -open" rectangle in the (s, n) plane, defined by the
inequalities
s~(s<s~++, -8<ncb. (w)
and i t maps each of these r e c t a n g l e s onto 9.

It follows f r o m property e that (6) is a locally one-to-one mapping, i.e.,
i t is one-to-one in any sufficiently s m a l l region of the s t r i p (3). Since
moreover A (s, n) # O and the mapping ( 6 ) is continuous, it is r e g u 1a r i n
e v e r y s m a l l region of t h i s kind. In other words, the mapping (6) is
1o c a 11y r e g u 1a r . s and n can be considered as curvilinear coordinates
in 8. To e v e r y point in 9 c o r r e s p o n d s one value of the coordinate n and a n
I12
513. CI.RVILI'.E.W COORDINATES
infinity of v a l u e s of s, d.iffering by multiples of t h e period T (the situation is

analogous in this r e s p e c t to the polar coordinates on a plane).
2. T r a n s f o r m a t i o n to the v a r i a b l e s s, n i n a
dynamic s y s t e m
T a k e the s y s t e m ( A }
_
d2 -
d t - p (I,Y). $= Q (I, Y)
and change o v e r - quite formally a t this s t a g e - to the new coordinates s and

n . Differentiating
-
z=cF(s,n), y=T(s,n)
with r e s p e c t to 1 and using (A), w e find

dz - ds - dn
~ = ' F ; ~ + f c p ; , ~ = P ( g ,n),
( s T. ( S . n)),
(7)
*d=t i j ; $ + f g = Q ( v ( S , n), i j ( s , n)).
By condition 4, the d e t e r m i n a n t
does not vanish in the s t r i p ( 3 ) . Equations ( 7 ) a r e t h e r e f o r e solvable for

d? dn
Since (see ( 2 ) )
is a solution of s y s t e m (A), we have
(g, (s. O), i (s. O)), a$ = Q (9, (s. 0 ) u (s. 0 ) )

d6 (S. G) (s, 0 )
--us -p 7
f o r all s, O , < S < T . Hence i t follows that
P (G (s, i(st 1))) G (S,0 )-Q (%(s, Oh lP (s, 0 ) )& (s, 0) =

=~:(s,O)~~(S,~)-$(S.O~$,(S,O)=A(S,O). (10)
Bycondition(5), A ( s , 0 ) g O f o r all s, O,<s,(r. The left-hand s i d e of the l a s t

equality t h e r e f o r e d o e s not vanish for all s, O,<s<z. But then, f r o m the
I I3
R (s, 0) 3 0.
T h e r e f o r e n = 0 is the solution, and the a x i s s i n the plane (s, n) is a n

i n t e g r a l curve of equation (12).
All the integral c u r v e s of equation ( 1 2 ) lying i n the s t r i p ( 3 ) evidently
coincide with the paths of (8). If
s = s (t), n =n ( t ) (16)
is a path e
of ( 8 ) , and so, no is a point on this path, the solution n = f (s; so. no)
of equation (8) is a n equation of 2 i n the coordinates s. n. The mapping ( 6 )
moves the path t into the line
2 = 6 (s ( t ) ,n ( t ) ) , Y =3 (s( t ) , n ( t ) ) , (17)
* The treatment of these subjects is analogous to the treatment of paths near a focus. See $10.1 and also
QT, 68.3.
rr138 114
$1':. P \ I L [ T E i R COORDlS.-\TES
<;I
Ixhich, i n virtue of l o c a l regularity of \ 6 ) , is a path of s y s t e m (A) lying in

the ring I!. Let t h i s pzth be L . It is readily s e e n that e v e r y path L of (A)
in $2 is a n image (under ( 6 ) ) of a t l e a s t one path of (8) in the s t r i p ( 3 ) ,
i.e., i t is a n image of ,it l e a s t one integral c u r v e of equation (12).'::
The q u a t i o n
n = r (s: so. no)
(13)
of the path i in the plane !s. n ) may be treated as a n equation i n curvilinear

coordinates s. n of the ?ath L on the plane (I, y ) . U'e will make u s e of this
fact in o u r a n a l y s i s of the succession function. Let I be the segment of the
n o r m a l tij Lo which l i e s i n and p a s s e s through the point . I f o of the path L o
corresponding to s = 0 , A s we know, 1 is a n a r c without contact for (A).
F r o m the theorettis of existence and continuous dependence of the
solution on the initial conditions, and a l s o f r o m the fact that - 0 solves
equation ( l 2 ) , we directly have the following propositions :
I. Any solution n = f (s; so, no)of equation ( 1 2 ) f o r all possible so. O < S ~ < T
and sufficiently s m a l l ,7,, is defined for all s, v,<s,<T and c a n be written i n the
form
n = f (s; 0, n ; ) .
For s y s t e m (A) t h i s m e a n s , in geometrical t e r m s , that e v e r y path of

the s y s t e m passing through a point i n a sufficiently s m a l l neighborhood of
I.,, c r o s s e s the n o r m a l 1 (for s = 0 ) and a l s o all the o t h e r n o r m a l s to Lo in the
ring i2> and then crosses the n o r m a l 1 again (for s = T ) (Figure 38). N o t e
that for n - 0, f = 1 (sq?e ( l o ) and ( 8 ) ) , and since the n u m e r a t o r and the
(lenominator of the e x p r e s s i o n for $ i n (8) do not r e v e r s e their sign i n the
s t r i p ( 3 J 2 we have :;> 0 in t h i s s t r i p , i.e., a n i n c r e a s e in s c o r r e s p o n d s
to a n i n c r e a s e in t , an11 vice v e r s a .
11. For any e > O , there exists q > 3. q = q (E), such that if no I < q, then
I f (s. 0, no)I < e

for a l l s. O < S , < T .
T h i s i m p l i e s that the p a r t of a path of s y s t e m \A) lying between two
s u c c e s s i v e intersection points of the path with the arc without contact I is
' Thc p t h 1, of (.A) is a n h a s < o f o n e (if it is claed) or infinitely many (if it is o p e d parhs of system ( 8 ) .
t k ' ~ ,w i l l not ci> into this problen. SeeQT. 9E.3.
115
I16
$12. CC'RI I L I 1 E . W COCRDLKATES
Setting n,=O in equation (23) and integrating with the initial condition (24),
we find
Since f ( s ; 0, O ) = O (see (19)), the integrand in (25) is equal to v. Let

u s calculate i t explicit:.y.
Differentiation of ( 1 2 ) with r e s p e c t to R gives
where P = P ( G . $ ) , P k = P k ( F , q), etc. -_- -_-

F r o m ( 9 ) w e see that f o r n = 0 , P ( ' p , 9 ) $i-Q('p. ip) 9 ; = 0 . T h e r e f o r e , for
n = U , only the f i r s t fraztion r e m a i n s in the right-hand side of ( 2 6 , .
We have
Z = ~ ( ' F ( S ) *$-=Q(T(S),
1C'(s)),
The left-hand s i d e of the l a s t relation is the integrand in (25). Therefore,

s
rif 0. no)
1 <W;+Q&[in N'P'!s)ll"tV ( ~ ) P ) I ds
')
(t:
= ,o -
-
on,, l-=o I
I'F, tc))lzC [ ~(0j12

,
$, [pi('P (sf. * (*i)+Q; (P (s), 0 (s))Id*
-
- IT' (SPLI+'(S)P
(29)
F r o m (18) and (29). setting S=T and r e m e m b e r i n g that the functions 'p and
rb and their derivatives a r e periodic, we finally obtain a n e x p r e s s i o n
I I7
I Ch. V. CLOSED PATHS I N S'I'RUC'I'URALLY S l A B L E S Y S I'EMS I
for f' (0):
%
D e f i n i t i o n 17. The numbev
is called the characteristic index of a closed path Lo.

Direct calculations show that \r.
0
+ 9 (4)l cis, and t h e r e -
[ P ' ('P (SI,$ (4) Q; ('P (4,
f o r e the c h a r a c t e r i s t i c index x . a r e invariant under a transforniation of
coordinates, i.e., they do not depend on the p a r t i c u l a r s y s t e m of coordinates
i n which the dynamic s y s t e m is d e s c r i b e d . T h i s i n t e g r a l is t h e r e f o r e
completely determined by the closed path L o . ff the c h a r a c t e r i s t i c index
x # O , thend' (0) + O , and the closed path Lo is a l i m i t cycle (112.3 and 512.4).
D ef i n i t i o n 18. A closed path Lo is called a simple limit cycle if
IK ((P ( 4 9 9 (s)) +4% ((P (4, 9 (411d s f 0 (33)
w, equivalently, i f the characteristic index x does not vanish. If Lo i s a

limit cycle and x = 0, i.e., condition (33) is not met, the path L , is said to
be a multiple limit cycle.
Note that i f x = 0, the closed path Lo is not n e c e s s a r i l y a limit cycle
( s e e 112.4).
F r o m 912.3, 112.4, and equation (31), we have the following theorem.
T h e o r e m 17. A simple limit cycle Lo is stable if
ana' unstable i f this integral is positive.
$14. PROOF OF STRUCTURAL STABILITY OF A

SIMPLE LIMIT CYCLE
I18
be a dynamic s y s t e m of class S',. 1 or a n analytical s y s t e m , defined i n G.
Let L., be a closed path 3f ( A ) in G , which is a s i m p l e limit cycle. Let a l s o
r=q(t). y=*(t), (1 1
where 9, and q- a r e periodic functions of period T, d e s c r i b e the motion

corresponding to the cyzle L o .
T h e p r e s e n t section IS concerned with the proof of the following t h e o r e m .
T h e o Y e )vi 18. An) simple limit cycle L , of a dj*numic system ( X i is a
structurally stable path.
P r o o f , We draw a n o r m a l through s o m e point bfo of the path L o and
define a p a r a m e t e r n on this normal, as in the previous section.'' The
point Moc o r r e s p o n d s to n = V on the normal. To fix o u r ideas, l e t the
negative v a l u e s of n correspond to the points of the n o r m a l which lie inside
L o , and positive n to the points of the n o r m a l which lie outside L o ( F i g u r e 3 9 ) .
Consider the part of the n o r m a l corresponding to n. 0 6 I n l<n*. Jf n* > 0 is
sufficiently s m a l l , the corresponding p a r t of the n o r m a l is a n a r c without
contact, which w e denote 1 .
Let n = f ( n ) be a s u c c e s s i o n function constructed in the direction of

i n c r e a s i n g t . W e may take i t t o have been defined everywhere along the
a r c 1 (this can be accomplished by using a sufficiently s m a l l n*). Consider
the function
d (n)= f (n)-n.
Clearly d (0) = 0, and f r o m the definition of a simple limit cycle, d' (0) fi).
Let d' (u) < 0 , i.e., the limit cycle Lo is stable. If L o is a n unstable limit
cycle, the proof is completely analogous.
* I f the m i n t YOcorresponds t o t =- 0 o n the path L o , the equations of the normal are (see $ 1 ~ 1 .115,)
.
I =~~Iu-~z$'(O), y=t(O)-nq'(O).
I I9
ch.V. CLOSED PATHS IN STRUCTURALLY STABLE SYSTEMS
Since 0 is a s i m p l e root of the function d (n), this function h a s no o t h e r

r o o t s i n a sufficiently s m a l l neighborhood of 0. We a s s u m e that d (n)h a s no
r o o t s f o r 0 < I n [(a*. Under this assumption, all the paths c r o s s i n g the
a r c 1 a r e not closed, going to the l i m i t cycle Lo f o r f-+ 00. +
Let L, be a path through s o m e point P,of the a r c 1 corresponding to
n = n,, 0 < nI < n* (i.e.> a path through s o m e point outside Lo)>and Lz a path
through some point Q, corresponding to n = nz,nz<O, nz I < n* (Le., through
a point inside L o ) . We have thus generated a sequence of points
P , , Pz. .. ., P,. .. .
of the path L , lying on the p a r t of the a r c I outside Lo which evidently go to
the point M o . and another sequence of points
QjY Qzt *. .* Qn, . . .
of the path L z lying on the p a r t of the a r c 1 inside L a , which a l s o go to the
point M o ( s e e QT, 9 3 . 7 , Corollary 2 of L e m m a 11). Evidently, any path
c r o s s i n g the segment P i p z (QIQz)of the a r c I a t any of the points o t h e r than
the ends successively c r o s s e s e v e r y other segment PkPktI ( Q k Q k + I ) a t one
and only one point which is not an end point.
Together with 1 , l e t u s consider another a r c without contact 1 passing
through the point M i of the path L o , which h a s no common points with 1. A
section of the n o r m a l through the point M ; can be chosen as such a n a r c .
W e take n* > 0 to be s o s m a l l that e a c h path c r o s s i n g the a r c 1 f o r t = to
c r o s s e s the a r c 1 f o r t ) t o r and h a s no other common points with e i t h e r 1 or
1. Let the paths passing through the points P , and Q, (i.e., the paths L, and
L2) c r o s s the a r c 1 a t the points P; and Q ; , r e s p e c t i v e l y ( F i g u r e 3 9 ) .
Let CI be a s i m p l e closed c u r v e formed by the arc P I P z of the path L, and
the segment P 2 P I of the a r c 1 , and Cz a s i m p l e closed c u r v e formed by the
a r c QIQzof the path L z and the segment QzQl of the a r c 1. Let f u r t h e r H be
the region limited by the c u r v e s C, and C,. All the paths of s y s t e m (A)
passing through the points of H evidently c r o s s the a r c 1 a t points which fall
between QI and P,, and Lo is the only closed path between t h e s e paths. The
segments PIQl and P;Q; of the a r c s without contact I and 1 divide the region
into two e l e m e n t a r y quadrangles P,P;Q;Q, and P;PzQ,Q;. We denote the
f i r s t quadrangle by A,, and the second by Az,
Let u s now c o n s i d e r modified s y s t e m s (A) sufficiently close to (A). By
L e m m a s 1,2, and 11 of 94 and the remark_ to T h e o r e m 5 ( 1 . 3 ) , t h e r e
e x i s t s a number 61 > 0 such that for any (A) which i s &-close to (A)
1 ) the a r c s 1 and 1 a r e a r c s without contact;
2 ) on the a r c 1 for all n, I n l g n * , a s u c c e s s i o n function n = f ( n ) is defined;
3) the equation b ( n ) = f ( n ) - n = O h a s a single r o o t n such that l Z l < n * , and
this r o o t satisfies the conditions nz <E < n,and ?L (G)(0;
4) all the paths of s y s t e m (A) which f o r t = t , c r o s s the segment P , Q , of the
a r c 1 , c r o s s for t > t o the a r c l, without mee_ting again the a r c 1 before this
t i m e . Let & and L, be the paths of s y s t e m (A) 6,-close to (A) which p a s s
through the points P , and Q, of the a r c I , respectively, and P; and Q; the
i n t e r s e c t i o n points of t h e s e paths with the a r c 1.
Condition 3 signifies that among the paths of s y s t e m (A)&-close to (A)
which c r o s s the a r c 1 , t h e r e is only one closed path E,, and this path is a
stable l i m i t cycle c r o s s i n g the a r c 1 a t a point M o between the points P , and
Q, on this a r c . Evidently, the path L, l i e s outside the closed path Eo, and
I20
the path z2 inside Lo. The segment Pl.G7, of the a r c / c o n t a i n s a sequence of
points.
- - - ..
PI. P , . Pa, .
of the path <

converging to the point -Go( h e r e PI coincides with P I ) , and the
segment Ql.@, of the a r c 1 contains a sequence of points
of the path L2convergin: to the point .Go( h e r e ol

coincides with Q,). Consi-
tfer s i m p l e closed c u r v e s analogous to the c u r v e s C , and C2, namely the
c u r v e 2, consisting of the a r c PIP, of the path z,
and the segment p2Flof the
arc 1 , and the c u r v e ?, consisting of the a r c old2of the path z2
and the
segment C,o, of the a r c !. L e t u s f u r t h e r c o n s i g e r the region k e n c l o s e d by
the c u r v e s ??, and c2. All the paths of s y s t e m ( A ) through the points of
region evidently cross the arc 1 at point? lying between Ql and P I . T h e r e -
f o r e , not a single closed path of s y s t e m (A), except the path &, p a s s e s
through any of the point:; i n ti. The segmen_ts P,o, and of the a r c s
--
without..-
contact 1 and 1' partition the region 17 into two e l e m e n t a r y quadrangles
/ ~ F;P202G;,which a r e designated A, and &, respectively (these
P , P , ~ ; I and
quadrangles are analogcus to the quadrangles & a n d A? of s y s t e m (A)).
iie will show that for any E > 0, t h e r e e x i s t s 8>0 s u c h that i f s y s t e m (A)
i s 6-close to s y s t e m (A), then
( H , A)
e - -
= ( H , .A).
In this way w e s h a l l have proved that L o is a s t r u c t u r a l l y s t a b l e path (see
1 7 . 1 and Definition 10, $6.1).
Let n be any positive number. W e s h a l l f i r s t show that i f 6?, 0 < 6? < b 1
is sufficiently s m a l l , and s y s t e m (A) is &-close to s y s t e m (A), w e c a n
c o n s t r u c t a topological mapping 'p of the segment P,Q, of the a r c 1 onto itself
.which h a s the following p r o p e r t i e s :
( a ) The points P , and QI a r e mapped into themselves, i.e., p (PI)= P I ,
'p rQ,)= 0,. Points lying on one path of s y s t e m ( A ) are mapped into points
of one path of s y s t e m (A) (and vice v e r s a ) .
( b ) The mapping 'p is a n-translation ( i . e . , f o r any point ,If of the
segment P I Q l ,.o fJf. p (ill)) -=cn I.
\ V e w i l l first c o n s t r u c t a topological mapping of the a r c P I Q I onto itself
which s a t i s f i e s p r o p e r t y ( a ) above, without bothering with p r o p e r t y (b) for
the time being. T h i s mapping is constructed s e p a r a t e l y on the segment
P,.\I% and then o n the s e g m e n t QI.lfo. The segment P , M 0 is mapped onto the
segment F,,e,, i n the following m a n n e r :
1) T a k e any topological mapping T~ of the segment P I P t onto the
segment such that
TI (PI)=PI. TI (P2)= P , .
21 Assuming that the mapping TFR-,of the segment P&$k onto ph-,P, (k =
= 2. 3. 4 , ..
. ) h a s been c o n s t r u c t e d , we define q Aa s the mapping of P 4 P G Lonto ,
ii$,,, i n d u c e d by the mapping 9 k - I i n the fol1owing s e n s e : i f .Me_,and . ! I k
a r e two s u c c e s s i v e intersection points a t which s o m e path L of s y s t e m ( A )
m e e t s the path l a s t i n c r e a s e s , which lie i n the s e g m e n t s P k - l P k and P I P h . , ,
c3.v. CLOSED PATHS I N s I'RUC I'URALLY STABLE sysrms I
respectively, and M k W l an? iGk a r e two s u c c e s s i v e intersection points a t
which path 2 of s y s t e m (A) m e e t s the arc 1 a s t i n c r e a s e s , which lie in the
I
I
segments &.$A and p.&+fr respectively, and if ' p k - l ( h f k - 1 ) = -2fh-1, then

I
' P h (ivh)= i v k
(Figure 40). In this figure, the intersections of the arc I with the path L and
5
with the path are shown s e p a r a t e l y f o r c l a r i t y . Since e v e r y a r c L (Z)of
s y s t e m (A) ((A)) passing through points of the segment P I P , (Pip,)m e e t s each
of the s e g m e n t s PkPI14,($k&), k = 2 , 3, 4 , . . ., p r e c i s e l y in one point, the
mapping ' p h is single-valued. Clearly,
3 ) Let cp((Mo)=Moand suppose that 'p coincides with 'pR on the segment
P,$Pk+g(k 1, 2, . . .).
E
FIGURE 40
W e obtain a mapping tp defined on the segment P , M , . Exactly the s a m e

mapping 'p is constructed on the segment QIMo(i.e., we f i r s t take a n
a r b i t r a r y mapping 'pT of QiQz on ol&,
and then the induced mappings (PA of
Q ~ Q Ron + ~& a + , ) . onto p,ol. The
A s a r e s u l t , we obtain a mapping 'p of P,Ql
mapping 'p constructed in t h i s way s a t i s f i e s condition ( a ) above and is a
topological mapping.':.
* W e can readily write an analytical expression defining the mapping 'p. Let f(n) be the succession function
on the arc I defined by system (A), and r(n) (x).
an analogous function for system Let g(n) be a function
which defines a topological mapping of the segment P , P , onto Ripz. Clearly, the iteration
[ f ( n ) j k - ' = f ( k - ' ) (n) defines a mapping of P I P z on PkPa+l. This is a topological mapping, and the
inverse mapping therefore exists. The corresponding function for the inverse mapping w i l l be designated
(n). The functions T("-') and f - ( k - l ) are treated similarly. Then if M is a point of the
segment PkP&l, n the value of the parameter corresponding t o this point M , and ;the value of the
parameter corresponding to q ( M ) , we have
\"-'I gf-\"-"(n).
An analogous expression is obtained for the mapping cp on the segment QkQk+l. I(*) (n)= [f (n)f is not a
..
symbol of a derivative here: this is the k-th iteration of the function f. i.e., f (f (f (. f (n) .))). ..
122
!Ve will now show that f o r a sufficiently s m a l l & 2 and a n appropriate
choice of q r and g,,*. the mapping g, constructed according to the above
formula jvill satisfy con,jition (b), i.e., i t will be a 0-translation. To prove
this, w e s t a r t with a fix,-.d u > 0 and choose a sufficiently l a r g e natural
number I, so that the segment P,Qlof the a r c I is entirely contained in
I', ? (.llo)(the point Mo evidently lies on this segment).
A11 the paths passing f o r t = t o through points of the segment P I P 2 ( Q 1 Q 2 ,
respectively) c r o s s all the s e g m e n t s Pipitl (Q,Q,+,,respectively) f o r
i - - 2, 3, .... I - 1 in a fin:.te time (see QT, $ 3 . 8 , L e m m a 13). It follows from
this and f r o m S4.2, L e m m a s 7 through 11, that f o r sufficiently s m a l l 6 2 and
a n appropriate choice of the mapping 'p, and 'pi*
1 J the section p,G, of the a r c 1 (containing the point .\?d) is entirely con-
tained in C,-arz(.If,,);
2 ) the mapping qh of the segment PAPAAIonto the segment Pk& (k =
- 1 , 2, .... I - 1 ) i s a o-translation;
3 ) the mapping q: of the segment QhQk+, onto the segment q& ( k =
- 1. 3. . . ., I - 1) is a u-translation.
Since the two s e g m e n t s P,Q,and p,@,lie in U,.2 and the mapping g: moves
the f o r m e r into the l a t t e r , the mapping 'p is a u-translation on the
segment P,Q,. Hence and f r o m conditions 2 and 3 i t follows that under
these conditions the mapping 'p of P,Q, is a a-translation.
We have thus shown that if a2 is sufficiently s m a l l , and s y s t e m (g)is
a:-close to s y s t e m (A), we c a n construct (for any 0 ) 9 ) a mapping q of the
segment P,Q, of the arc 1 onto itself which s a t i s f i e s conditions ( a ) and (b).
LVe will now show tha: f o r any E > 0 and for s y s t e m (A) sufficiently c l o s e
to [,A), we have
(H, A) & ( E , d).
Let e > O be fixed. By L e m m a 8, S4.2, t h e r e exist u'>O and & > O , & ( b 2
which satisfy the following condition: if q ' is a mapping of the segment P;Q;
of the arc I' onto the szgment p;o; of the s a m e arc, which i s a o'-translation,
and s y s t e m s ( A ) and ( A ) a r e d,-close, t h e r e e x i s t s a mapping T, of the
e l e m e n t a r y quadrangle -I! onto the quadrangle ?12 which coincides with q'on
P,Q;, moves paths into paths, and is a n &-translation. Let & > O , d , < h 2 be
sit s m a l l that if s y s t e m ( E ) is 6,-close t o s y s t e m (A), t h e r e e x i s t s a
mapping T of the e l e m e n t a r y quadrangle A, on the quadrangle 6, satisfying
the following conditions:
1 ) TI coincides on PI(),with the previously described mappinq q~of this
segment onto itself;
2 1 T,moves paths into paths;
3 ) TI is a n -translation;
31 on P@;, the mapping T is a 0'-translation (moving this segment into
P ; Q , ).
'The existence of a number 6, > 0 with the above p r o p e r t i e s follows f r o m
Leninia 8, S4.2, and f r o m the fact that f o r c l o s e s y s t e m s the mapping cp is
a n a r b i t r a r i l y s m a l l translation (property b). A s the number J w e choose-
a n a r b i t r a r y positive number, s m a l l e r than e i t h e r 6 ; o r 6 , . Let s y s t e m ( A )
be &-close to s y s t e m ( A ) . X'e construct a mapping 'p of the segment P,Q,,
then a mapping TI of the cuadrangle A , onto SI with p r o p e r t i e s 1 through 4 .
The mapping q' is identified with the mapping T,on the segment P;Q;(which
123
Ch.V. CLOSED PATHS IN STRUCTURALLY STABLE SYSTEMS
moves this segment into p;G), and a s the l a s t s t e p we c o n s t r u c t a mapping Tz

of the e l e m e n t a r y quadrangle A z onto a,,
which coincides with cp'on P;Q;,
moves paths into paths, and is an e-translation.
It is readily s e e n that the mapping T which coincides with T ion A, and
with T z on A 2 moves the region I into r?, c o n s e r v e s paths, and is an E-
translation. This indicates that
( H , A ) & (ii, A",.
The proof of the t h e o r e m is thus complete.

Rema r k . It follows f r o m o u r proof that i f Lo is a s i m p l e l i m i t cycle
of a d y n a p i c s y s t e m (A), t h e r e e x i s t E*> 0 and 6* 2 0 such that any
s y s t e m (A) 6*-close to (A) h a s a single l i m i t cycle Lo in the e*-neighborhood
of the paths L o , and the c h a r a c t e r i s t i c indices of the c y c l e s Lo and 2, have
the s a m e sign ( i , e , , the c y c l e s Lo and 2, a r e e i t h e r both s t a b l e or both
unstable).
J15. STRUCTURALLY UNSTABLE CLOSED PATHS
1. The fundamental l e m m a
Let, as before,
be a dynamic s y s t e m of c l a s s N > I o r a n analytical s y s t e m in region G ,

Lo a closed path of this s y s t e m ( L O c G ) ,
the motion corresponding to this path, cp ( t ) , q ( I ) are periodic functions of

period T > 0. rp, I), being a solution of s y s t e m (A) of a c l a s s 2 1, a r e a
p r i o r i functions of c l a s s 2 .
L e m m a 1 . There exists a function of class 2, z = F ( x , g), defined in
B such that fm all S, - 00 < s < 0 0 , +
(4 F (9 (s), $ (s!) = 0;
(b) (F; ('P +
(4, +
(s)))* (FI ('P (s), $ W)*z 0.
p r o o f of (a). We w i l l f i r s t c o n s t r u c t a function F (3,v) of c l a s s 2
satisfying conditions a and b not in the e n t i r e region but only in some
neighborhood of the path Lo.
To this end, consider the s e t of equations
where a ( s ) and b ( s ) a r e periodic functions of period 7 which belong to c l a s s 2,

such that the functional determinant
124
915. C TRr.CTC RALLY I N S T A B L E CLc)SED P.A7klS
does not vanish for any s:
For a ( s ) and b (s) w e may take the functions $' (s) and - cp' ( s ) , respectively,
provided they are functi.ons of c l a s s 2 (this will be s o i f ( A ) is a s y s t e m of
c l a s s 2, say). If, however, $' ($)and - cp' (,s)arefunctions of c l a s s 1, but
not of c l a s s 2, a (s) and t t (s) c a n be identified with trigonometric polynomials
of period T which e n s u r e a n adequate approximation to 9' (s) and - cp' (s).
I'hese polynomials exist in v i r t u e of W'eierstrass's theorem ( s e e /ll,/,
L'ol. 3 . Sec. 734, p. 5 8 0 ) evidently meet condition (4).
Let 6 be s o m e sufficiently s m a l l positive number. W e will consider ( 2 )
as equations which define the mapping of the s t r i p
-CO (s(fos, -6,<n,<d (5)
in the plane (s, n) into the plane (I, y ) .

Under this mappinq, the a x i s s is c l e a r l y moved into the path Lo> and
e v e r y v e r t i c a l segment
s-const, --6,<n,<6
of the s t r i p (5) is moved to a straight segment I , through the point .TI (s) of
the path L o corresponding to the value s of the p a r a m e t e r . in virtue of
condition (4), the segment Isdoes not touch the path L o at the point ,I/ (s).
3loreover, from the the3rem of implicit functions and from condition ( 4 )
it follows that i n a sufficiently s m a l l neighborhood of any point (s, 0) of
the a x i s s, the mapping (i2)is one-to-one. But then i t c a n be shown,
p r e c i s e l y a s in L e m m a 1, 513.1, that if 6 is sufficiently s m a l l , no two
s e g m e n t s I , corresponding to different points of the path Lo (e.g., f o r
OKs < T) i n t e r s e c t . W e w i l l a s s u m e that this condition is indeed satisfied.
The s t r i p ( 5 ) is then mo7;ed by mapping ( 2 ) into s o m e closed ring in the
plane (I. y) enclosed by simple closed c u r v e s f,and rn ( F i g u r e 41). If 6 is
sufficiently s m a l l , then a t any point of the s t r i p ( 5 )
Ne will a s s u m e that this condition is a l s o satisfied.
Mapping ( 2 ) of s t r i p (5) onto the region is e n t i r e l y analogous to the

mapping ( 6 ) considered i:i 513.1 and i t h a s s i m i l a r p r o p e r t i e s . In
Ch. V. CLOSED P A T M IN STRUCTURALLY SIABLE SYSIEMS
p a r t i c u l a r , the mapping ( 2 ) is locally one-to-one in s t r i p (5), and by v i r t u e

I
of condition ( 6 ) i t is a l s o r e g u l a r , i.e., this mapping is locally r e g u l a r .
Let u s c o n s i d e r ( 2 ) as a s e t of equations in s and n, a s s u m i n g that
(I, y) E Q. In v i r t u e of the above, to e v e r y point ( 5 , y) E P t h e r e c o r r e s p o n d s
p r e c i s e l y one value of n and infinitely many values of s, such that the
n u m b e r s s, n satisfy equations ( 2 ) (the v a r i o u s s differ by a multiple of T).
The p a r t i c u l a r n is t h e r e f o r e a single-valued function of x , y. We will
designate i t by F (I,y):
n = F ( z , y). (7)
If we c o n s i d e r only the l o c a l situation, s may a l s o be r e g a r d e d as a single-
valued function of x and y. Moreover, n and s are functions of c l a s s 2 (in
v i r t u e of the theorem of implicit functions).
We will show that the function of c l a s s 2 defined by relation (7), F (I, y),
s a t i s f i e s conditions ( a ) and (b) of the l e m m a .
Condition ( a ) follows d i r e c t l y f r o m the fact that a t e v e r y point (I,y), i.e.,
(cp (s), q (s)), the value of n corresponding to the path Lo is z e r o in v i r t u e of
equations (2).
The proof of (b) will be based on l o c a l considerations, and both n and a
w i l l be r e g a r d e d a s single-valued functions of x and y .
P r o o f o f (b). Differentiationof ( 2 ) w i t h r e s p e c t t o x and y r e s p e c t i v e l y
gives
1= [cp (4+ nu (41 g + a (4 t
(8)
O=[$(s)+nb(s)]~+b(s)~
and
o = [cp (s) + na (s)] 2
aY
+ a (s) *
av
,
l = [ q ( s ) + n b ( s ) j zs+b ( s an) .~
ay
I
For n = O , each of the s y s t e m s ( 8 ) and (9), considered a s a l i n e a r s y s t e m
of equations i n the p a r t i a l derivatives, h a s a determinant A (s) which d o e s not
vanish in v i r t u e of (4). Solving these s y s t e m s and r e m e m b e r i n g that
n = F ( x , y) and that for n = O , z=cp(s), y = Q ( s ) , we find
Hence,
(F; (cp(s), q ( s ) ) ) 2 + ( F ; ( c p ( s ) , l$(s)))2= [~(s)a+[*(s)P

A
The l a s t e x p r e s s i o n does not vanish, since i f cp (s) = Q (s) = Ofor s o m e s,

I
(cp (s), $ (s)) is a n equilibrium s t a t e of s y s t e m (A), and this c o n t r a d i c t s the
fact that the point (cp (s), Ip (s)) belongs to the closed path Lo. Condition (b) is
thus proved.
W e have thus shown that t h e r e e x i s t s a function F (I,g)of c l a s s 2 in the
ring which s a t i s f i e s conditions ( a ) and (b) of the l e m m a .
126
The boundary of the- ring 5 c o m p r i s e s the two simple c u r v e s I", and r2
.with t h e p a r a m e t r i c ecquations
.r = 9 (s) f 6a (S), y = 11' (s) + &5 (S)
and
.r = cp (s) - 6a (Y), y = $ (Y) -6b ( s ) .
The right-hand s i d e s of t h e s e equations a r e functions of c l a s s 2 . There-

fore, by Lihitney's theorem ( s e e I l l / , Vol. 1, p a r a . 260, p . 594), the
function f' (I, y)can be continued to the e n t i r e region without changing i t s
c l a s s . The continuation of the function evidently r e t a i n s the s a m e
p r o p e r t i e s ( a ) and (b). This completes the proof.
R e m a r k 1 . If ( A ) is a s y s t e m of class S > 1, t h e r e e x i s t s a function
F (I. y ) o f class IV $- 1 which s a t i s f i e s conditions ( a ) and (b) of the l e m m a .
rhe proof is the s a m e as before, but a (s) and b rs) are functions of class
.v - 1.
R e m a r k 2. Let the plane (I. y) in which s y s t e m (A) is defined be a
coordinate plane of the three-dimensional s p a c e (z, y. z ) . The equation
I - F (I. y), where F is the function d i s c u s s e d in L e m m a 1, is the equation
of a surface through the path L o . It follows f r o m condition (b) that a t the
points of this path the s u r f a c e does not touch the plane ( I , y).
2. The theorem of the c r e a t i o n of a closed path f r o m a

multiple l i m i t cycle
Ne r e c a l l that a closed path L,, is said to be a l i m i t cycle if i t is

i s o 1 a t e d i.e., i f any neighborhood of the path d o e s not contain any closed
~
path except Lo itself. In this c a s e , a s we have s e e n in 512.2, a l l the paths

which p a s s through points sufficiently c l o s e to the path L , e i t h e r w i n d
o n t o Lo (a stable l i m i t a x i s ) or u n w i n d f r o m Lo (unstable l i m i t cycle),
o r else paths on one s i d e of Lo unwind f r o m i t and paths on the o t h e r s i d e
urind onto it (a s e m i s t a b l e cycle). A limit cycle is called multiple i f i t s
c h a r a c t e r i s i t i c index is not z e r o (Definition 18, $ 1 3 . 3 ) .
W e w i l l now prove ;L theorem which is the analog of the theorem of the
c r e a t i o n of a closed path f r o m a multiple focus (Theorem 14).
T h e o r e ttz 19 (theoretn of the creation of a closed path f r o m a
multiple littiit cycle). Let ( A ) be a dynattiic systetn of class IV ,.' i (or
analytical), L o a multiple titnit-cycle of the system. For any t' > 0 and
,i o,, there exists a system ( A ) of the sattie class which is 6-close to (A)
+-
to rank r<,V ( r < a )and which has at least two closed paths in the E -
neighborhood of L , .
P r o o f . The proof of Theorem 19 is based on L e m m a 1, and in all
other r e s p e c t s follows the proof of T h e o r e m 14. We w i l l consider, a s in
the previous sections, a s u c c e s s i o n function f (n)and a function
d (n) = f (n)- n on s o m e n o r m a l t o the path L o . Let n be the p a r a m e t e r along
the normal, defined by equations (2) in 912.4. Since by assumption L o is a
limit cycle, t h e r e e x i s t s n* > 0 such that f o r all n, I n I -== n*. n # 0, a
s u c c e s s i o n function f ( ! I ) and f (n) - n = d In) # 0 (d (0) = 0) are defined. Let 1
I27
Ch.V. CLOSED PATHS IN STRUCTURALLY STABLE SYSTEMS
be the segment of the n o r m a l corresponding to the values of n such that

1 n I<n*. To fix ideas, l e t Lo be a stable l i m i t cycle (the proof for a n
unstable or a s e m i s t a b l e cycle proceeds along the s a m e l i n e s ) . Then f o r
.all n, I n I c n*, n # 0 ,
d ( n ) = f (n)-n<O, if n>0;
d(n)=f(n)-n>O, if n<O.
We w i l l f i r s t prove the theorem for s y s t e m (A) of c l a s s 1.

Let E > 0 and 6 > 0 be given. Let n* > 0 be s o s m a l l that each a r c of a
path of s y s t e m (A) whose ends l i e on our normal and correspond to the
values n and f (n)of the p a r a m e t e r ( i . e . , an arc between two successive
intersection points with the normal), I n I<<*, is contained in U, ( L o ) .
Take any nl, 0 < nl < n*. By (12),
d (nt) <0. (13)
Let f u r t h e r a,, 0<8,<6, be so s m a l l that if s y s t e m (a)
is 6,-close to
system (A) then (5) the normal I r e m a i n s a n a r c without contact for the
paths of s y s t e m (A); (b) for a l l n, Inl<n*, the functions 7 ( n ) and d"(n)a r e
defined on the normal; ( c ) the a r c s of the paths of s y s t e m (A) between two
s u c c e s s i v e intersection points n and 7 (n)with the n o r m a l (In I Q n *) a r e
contained in U,(Lo), and ( d ) the following inequality is satisfied:
J(nl) < 0. (14)
The existence of a number 6 satisfying the above conditions follows
f r o m L e m m a s 1,2, and 1 1 of 84.
We will cGnsider modified s y s t e m s of a special f o r m , a s follows.
Let F (z, y) be afunction of c l a s s 2 satisfying the conditions of L e m m a 1,
and p any r e a l number. We take a modified s y s t e m (A) of the f o r m
Clearly, (A) is a s y s t e m of c l a s s 1, and if p is sufficiently s m a l l in

absolute value, (A) is a r b i t r a r i l y c l o s e to (A).
) a solution of s y s t e m (A) corresponding to the
Since z=cp(t), ~ = g ( t is
path L o r and the function F ( r , y) has been defined so that it s a t i s f i e s the
equality
the functions cp((t), g ( t ) a r e a l s o a solution of s y s t e m (A), i . e . , the path Lo of

s y s t e m (A) is a l s o a path of s y s t e m (A).
Let p be so s_mall that (A) is 6,-close to (A). Inequality (14) is then
satisfied, i . e . , d (nl)<O.
Let u s compute s(0). By definition d'(O)=O, i.e.,
128
915. STRVCTLRALLY CKST.ABLE CLOSED P.ATHS
Differentiating and Q" with r e s p e c t to I and y , respectively, and

using (15) and (16), we find
.(
\ IPk((cp(4,+ ~ s ~ ) + ' ? L ( ~ ( s ) , $ ( s ) ) l d s =
i
The l e t t e r J identifies the integral in the right-hand side of the equality.

F r o m property (b) of function F i t follows that J > 0. Finally, f r o m (17)
we see that
Let p)OL Then 2(0):>0, i.e., the path L o i s a n unstable l i m i t cycle of

s y s t e m (A). Hence, f ' 3 r all sufficiently s m a l l n > O , a(,) > O .
In p a r t i c u l a r for s o m e n,, O < n Z < n , ,
B(n,) >o. (19)
The continuity of the function ;and inequalities (14) and ( 1 9 ) show that
between n, and ng t h e r e is a t least one value ,R*>O such that a(,*)= 0. This
number c o r r e s p o n d s to a closed path L* of (A) which d o e s not coincide with
Lo. By condition (c) imposed on 8 , . L * c U , ( L o ) . The second closed path of
(A) lying in L'e(Lo) is Lo itself. The theorem is thus proved f o r the case
when (A) is a s y s t e m of class 1.
If (A) is a s y s t e m of c l a s s S > l . r g S , and the c l o s e n e s s is considered
to rank r, the proof proceeds along the s a m e lines, but F ( r , y) is a function
-'+
of c l a s s 1 ( s e e R e m a r k 1 to L e m m a 1 ) and p is sufficiently s m a l l for
( A ) to be &close to (A) to rank r .
Let now (A) be a n analytical s y s t e m . In this case, L e m m a 1 does not
apply i n i t s original f o r m . Indeed, w e c a n c o n s t r u c t in the neighborhood
of Lo a n analytical function F (I, y) satisfying conditions ( a ) and (b); this
may be accomplished a s f o r s y s t e m s of c l a s s A'. This function, however,
cannot be continued to the e n t i r e region i n g e n e r a l . W e will therefore
adopt a slightly different c o u r s e .
To fix ideas, l e t u s suppose, as before, that L o is a stable (multiple)
l i m i t cycle. W e choose s o m e n,. 0 < nl < n * . r h e n d (n,)< 0. Taking a
fixed h > 0 , w e construct a function F (z. y) of class ( r + 1) which s a t i s f i e s
the conditions of L e m m a 1 and consider the s y s t e m (x),
p is chosen as a
a
positive number, Sufficiently s m a l l f o r (A)to be close to (A) to r a n k r and
f o r the following inequality to be satisfied:
d(n,) -==0.
Since f o r p > 0 , 2 (0)> 0 and a l s o d"(0)= 0, we have f o r sufficiently s m a l l
nz>Oand n 3 < 0
2 (n.) >o (19)
129
Ch.V. CLOSED PATHS IN SI'I1UCTUKALLY STABLE SYSlEhlS
and
2 (n.3)<0
(Figure 42a).
FIGLIRE 42
Let u s now c o n s i d e r an analytical s y s t e m ( A ,) q-close to (A) to rank r,
where q :-. q is taken to be sufficiently s m a l l s o that a s u c c e s s i o n

function f * ( n ) i s defined on the a r c 1 for s y s t e m (A'") and the following
inequalities a r e satisfied:
d* (nt) <0, d* (nz)>0, d* (n3)<U. (21 1
Clearly, ( A ,) c a n be chosen a s any s y s t e m whose right-hand s i d e s a f e

polynomials providing a sufficiently c l o s e fit to the right -hand s i d e s of (A).
Since q <+, s y s t e m (A) is 6-close to (A) to r a n k r . It follows f r o m (21)
that t h e r e e x i s t a t l e a s t two v a l u e s of n (which lie between n, and n2 and
between nz and nSr r e s p e c t i v e l y ) f o r which the function d* v a n i s h e s . These
values of the p a r a m e t e r c o r r e s p o n d to two closed paths of system (A':.).
For sufficiently s m a l l 6, these paths c l e a r l y l i e in U, ( L o ) ) .
The c a s e of unstable o r s e m i s t a b l e Lo is t r e a t e d along the s a m e lines.
T h e s e i n s t a n c e s a r e illustrated in F i g u r e 42, b and c. The proof of the
theorem is thus complete.
R e m a r k 1. It is readily s e e n that i f LA is a stable o r a n unstable
multiple l i m i t cycle, t h e r e exist s y s t e m s (A) a r b i t r a r i l y c l o s e to (A) which
have a t l e a s t t h r e e closed paths in any a r b i t r a r i l y small neighborhood of
Lo. The validity of this proposition can be established reasoning along the
s a m e l i n e s as in the proof of the theorem ( s e e F i g u r e 42a illustrating the
stable c a s e ) . F o r a s e m i s t a b l e cycle L o , however, the l a s t proposition
( r e g a r d i n g the existence of a t l e a s t t h r e e closed p a t h s ) is in g e n e r a l
inapplicable.
R e m a r k 2 . It can be readily s e e n that i f Lo is a multiple limit cycle
of s y s t e m (A), t h e r e e x i s t s a s y s t e m (A;:?)a r b i t r a r i l y c l o s e to (A) which h a s
any a r b i t r a r i l y s m a l l neighborhood of the cycle L o . We w i l l prove this
I30
proposition f o r the c a s e of analytical s y s t e m s . The f i r s t , sf'p i s to
c o n s t r u c t , as before, .I s y s t e m ( A I ) of c l a s s 1 of the type ( A ) for which
L,, i s a s i m p l e liniit cy-le and which h a s a n o t h e r c l o s e d path L , close to Lo.
If L , is a s i m p l e liniit cycle, w e take a polynomial (and t h e r e f o r e
a n a l y t i c a l ) s y s t e m (A. I sufficiently c l o s e to ( A I ) . Because of the
s t r u c t u r a l stability of the c y c l e s L , and L , of s y s t e m ( A i ) , s y s t e m (A' )
w i l l have sLructurally :;table c y c l e s f.: and L; in a c l o s e neighborhood of
I.., a n d L t L If L , i s not a simple l i m i t c y c l e of (Ai), we c o n s t r u c t a
s y s t e m \ A , ) r e l a t e d to the c y c l e L , in the s a m e manner a s s y s t e m (.K) is
r e l a t e d to c y c l e L,>n the proof of J 3 e o r e m 15. For a n a p p r o p r i a t e choice
or' the number p , ( A t ) will be so-close to (A,) that in any a r b i t r a r i l y s m a l l
neighborhood of L o the system (Al) will _have a s i m p l e l i m i t c y c l e I%,. The
path L , will be a s i m p l e l i m i t cyc_le of (AI) ( s e e proof of Theorem 15). I t
noiv r e m a i n s to approx-mate to (AI) with a sufficiently c l o s e analytical
sy s te m .
Let u s prove a n o t h e r lemnia ;vhich will be needed in the following. Its
proof is based d i r e c t l y on Leninia 1.
L c )ti ttt n 2. Let
he a dytiattiic spstetti 41'class .I-: 1 dgfined iti region C, and x = q ( t ) ,y = IF (1)

ci c~loscclpath L qf tlie s.ysteni ( L c idiich is riot a stnrcturallp stable
(i.c., simple) l i t t i i t cycle. For a)ig 6 > o, there exists an nrbitunrilg sttiall
tieiqhboi-hood i' of the path L atul n dynatriic sgstet)? i B i of class .Y iritlt tlic
w/bLloic.itiqproperties:
(a, Sy.sterti (3) is 8 -close to ratik .V to sj.qterir (-4) in E .
(b) Systett? (I31 coiricides with spstetti (-4) outside tlie neighbat-hood r.
(c9 Tlic! patlt L ?f systeitz ( A ) is a stvrrctiirallg stuble l i l t l i t cycle of
sy.ctetti iR i .
P r o o f . Since ( A ) i.; a s y s t e m of class .V, 'p and 11 a r e functions of
class S f 1. In the neighborhood of the path L , we introduce c u r v i l i n e a r
c o o r d i n a t e s s and ti defined by the r e l a t i o n s
X = ( F ( S ) ~ ~ $ ' ( S ) , y=$(s)--nq'(~). (22i
In L e m m a 1 w e have s e e n that to e v e r y point (I.y ) o f a sufficiently s m a l l

neighborhood of the path L c o r r e s p o n d s p r e c i s e l y one value of the
parameter n ,
R = F (5. Y),
and infinitely many v a l u e s of s, s u c h that the n u m b e r s s and n s a t i s f y

equations ( 2 2 ) . By the theorem of implicit functions, F (2. y) is a function of
c l a s s .Y,
Let R , be a sufficient1,v s m a l l positive number, and nZ and n 3 s o m e positive
n u m b e r s s u c h that n3 < n 3 < n,. L e t iL.. (I, and V be the neighborhoods of L
(iefined by the r e s p e c t i v e inequalities
lnl<n3. I-nl<nz, Inl<n,.
131
Ch.V. CLOSED PATHS IN STRUCTURALLY STABLE SYSTEh4S 1
Clearly,
WCUCV
( s e e F i g u r e 43a, where the neighborhood W i s c r o s s - h a t c h e d ) .
FIGURE 43
Let y (n) be a function satisfying the following conditions:

( a ) y (n)is a function of c l a s s N defined for all n, I n I < n,;
b) y(n)=l for Jnlgn,,
y(n)=O for n2<JnJ<n,,
O<y(n)<i for n l g ( n l < n ,
( F i g u r e 43b). T h e r e evidently e x i s t functions y (n)of any c l a s s satisfying
condition (b).
In our proof of Theorem 1 9 we established that if L is not a s t r u c t u r a l l y
s t a b l e l i m i t cycle of (A), t h e r e e x i s t s a s y s t e m
of c l a s s N a r b i t r a r i l y close to (A) to r a n k N for which L is a s t r u c t u r a l l y

stable c y c l e .
132
515. S T R I C K R . I L L Y L%ST.-\BLE CLOSED P.XTHS
Consider a s y s t e m ( B ) which coincides with s y s t e m (A) in c-Vand is

e x p r e s s e d by the equations
dr
; 1 T = = P ( r Y. ) i l p " ( I , Y)--p(.r, Y)lY(F(Z,P)),
iB)
g = Q (2.Y) + rd (2,Y) -Q (I,Y)l T ( F (z,Y))
in the neighborhood 1. of the closed path. It is readil_y s e e n that in the

neighborhood W , s y s t e m ( B ) coincides with s y s t e m (A), i.e., L is a
s t r u c t u r a l l y stable l i m i t cycle f o r (B). hIoreover, in V - U s y s t e m ( B )
coincides with s y s t e m (A). Hence, (B) coincides with (A) in - C: too.
Finally, ( B ) is evident1;y a s y s t e m of c-lass iV and is a r b i t r a r i l y c l o s e to
rank .V to s y s t e m (A), provided that (A) is sufficiently c l o s e to (A).
System ( B ) thus s a t i s f i e s all the propositions of the l e m m a . This c o m -
pletes o u r proof.
3. Structural instability of a closed path with a z e r o

c h a r a c t e r i s t i c index
T h e o r e H I 20. A closed path Lo of system ( A )f o u which
5 l-pL(T(4? $ ( S ) ) + Q l ( c p ( S ) , $(s))lds=O
is structurally irnstable (in relatimi to any of the spaces R!;', r<LV, if ( A )

is a spstevn of c b s s N , ard Rr', if ( A ) is an analytical system).
P r o o f . We s h a l l f i r s t prove m e o r e m 2 0 assuming that (A) is a
s y s t e m of c l a s s -V and s t r u c t u r a l instability is treated in relation to the
s p a c e R';', where i , < r < X .
Suppose, c o n t r a r y to the proposition of the theorem, that the path L o is
s t r u c t u r a l l y stable in s a m e neighborhood H of L o . Then f o r any e > 0,
t h e r e e x i s t s 6 > 0 such that i f s y s t e m ( A ) of class N is 6-close to rank r to
s y s t e m (A), we have
( H , A ) E ( H * , A*),
where H* is s o m e region. We will now show that L , in t h i s case is

n e c e s s a r i l y a l i m i t cycle of (A).
Let 1 be the arc without contact for the paths of s y s t e m (A) considered
above (which is a n o r m a l to L o ) , R a neighborhood of the path Lo with the
property that e a c h path 3f s y s t e m (A) passing through a point of R m e e t s
the arc without contact I both f o r i n c r e a s i n g and d e c r e a s i n g f. Any sufficiently
s m a l l canonical neighborhood of the path L o may be taken as 52 ( s e e QT, 5 2 4 . 3 ) .
Regardmg the neighborhood H , w e a s s u m e that i t lies inside R a t a
positive distance f r o m the boundaries of this region. L e t e > 0 be s o
s m a l l that a region H* generated by an e-translation of H is contained i n R,
i.e., H* c '2. Let 6 > 0 be a number corresponding to this E i n a s e n s e
that i f (A) and (A:) a r e C-close, relation ( 2 3 ) is satisfied. Let ai, 0 -=
6i < 6,
be s o s m a l l that if s y s t e m ( A ) is &-close to (A), e v e r y path of (A*)
I33
134
51s. m w c r c p i i LI- [ Z 5 l - i W . E C L ~ ~ S EP.-\I~!s
D
C o r o l l a r y . X d o s e d path Lo is s t r u c t u r a l l y stable .nd only i f it s

a s i m p l e l i m i t c y (:le, i.e.. the c h a r a c t e r i s t i c index
r
i r
.- 7 (9( S ) . # ( S ) ) T vi ( q ( d ) , #($))I d S
3
(loes not vanish.

r h e valklity of this proposition f o l l o w s d i r e c t l y f r o m nil:ion 18 a n
rheorerns 18 and 20.
I35
C h a p t e r VI
NECESSARY AND SUFFICIENT CONDITIONS OF

STRUCTURAL STABILITY OF SYSTEMS
INTRODUCTION
In the previous c h a p t e r s w e derived a number of n e c e s s a r y conditions

of s t r u c t u r a l stability of a dynamic s y s t e m . We established, in p a r t i c u l a r ,
that i f a dynamic s y s t e m (A) is s t r u c t u r a l l y stable i n a closed bounded
region c*, then:
I. It may have only a finite number of equilibrium s t a t e s in &*, which
a r e of necessity simple nodes, saddle points, or foci ( A < 0 o r A > 0 and
a # 0; s e e 17, T h e o r e m s 1 0 and 11, and 110, Theorem 15).
11. All the closed paths of s y s t e m (A) are simple l i m i t cycles ( i . e . ,
paths 2= 'P (4,I/ = II,(4f o r which J = ( P ; ('P (4.
'P (s)) + Q; (cp (s), 'P (4))ds # 0,
where z is the period of the functions'q and cp; see 915, Theorem 20).
111. System (A) does not have saddle-to-saddle s e p a r a t r i c e s in c*((s11,
Theorem 16).
In the present chapter, we will prove that conditions I through I11 a r e
both n e c e s s a r y and s u f f i c i e n t f o r s t r u c t u r a l stability of a s y s t e m
i n G*. A r i g o r o u s proof of this proposition, although essentially simple,
r e q u i r e s a fairly lengthy and tedious analysis. It is analogous to the proof
of Theorem 76 i n QT, 529.4.
Chapter VI c o n s i s t s of t h r e e sections (516, 17, 118).
In $16 we prove that i f s y s t e m (A) is s t r u c t u r a l l y stable in G*, it may
have only a finite number of closed paths i n 6* (Theorem 21). and hence
only a finite number of orbitally unstable paths and semipaths (Theorem 22).
The concept of a r e g i o n w i t h a n o r m a l b o u n d a r y is a l s o i n t r o -
duced i n $16. A normal boundary is made up of a finite number of simple
closed c u r v e s , each of which is e i t h e r a cycle without contact or c o n s i s t s
of a n even number of alternating a r c s without contact and a r c s of paths.
The fundamental theorem of s t r u c t u r a l stability i s proved for regions with
a normal boundary, since t h i s assumption greatly simplifies the proof
without imposing a significant r e s t r i c t i o n .
517 is completely devoted to supplementary m a t e r i a l . It is proved in
t h i s section that a region e* with a normal boundary can be partitioned into
canonical neighborhoods of equilibrium s t a t e s and l i m i t c y c l e s and into
elementary quadrangles. T h i s partition is actively used i n the proof of the
fundamental theorem.
In 518, we give a complete proof of the fact that conditions I through 111
a r e the n e c e s s a r y and sufficient conditions of s t r u c t u r a l stability for a
136
5,ystem [%.A) c*
in (The,2r'em 2 3 ) . In 518.3 we prove that these are a l s o
necsessary and sufficient conditions f o r s t r u c t u r a l stability on a s p h e r e
( T h e o r e m 2 4 i . At the end of the section, $18.4 offe1.s a nuniber, of signifi-
c*anf r e m a r k s and supplements. T h e s e include a t h e o r e m which states that
the s t r u c t u r a l l y stable s y s t e m s forni a n open set in the s p a c e of a l l dynamic
systems in a plane retgion (Theorem 2 5 ) and o n a s p h e r e ( T h e o r e m 2 5 ' ) and
ti rheorcni a c c o r d i n g to Lvhich the s t r u c t u r a l l y stable s y s t e m s a r e e v e r y -
Lvhere dense in the s p a c e of the dynamic s y s t e n i s \Theor.em 2 6 ) . T h e s e
t h e o r e m s indicate !hat "almost all" the dynamic s y s t e m s a r e structut-ally
s t a b l e , and s t r u c t u r a l l y unstable s y s t e m s a r e a n exception.
2. r e a d e r ivishing to get h i s teeth into the nieat of the theory without
further. delay may s k i p the proof of the fundamental t h e o r e m s 2 3 ( $ 1 8 . 2 1
ami 2 4 (S18.3j and onl:.- p e r u s e $ 1 6 , 1 , the s t a t e m e n t of T h e o r e m 2 3 i n S 1 8 . 2 ,
iiml a l s o S 1 8 . 3 and S l E . 4 .
1. Finite number of closed p a t h s for s t r u c t u r a l l y

stable systems
B e f o r e proceeding with the proof of the sufficiency of conditions I

through 111 ( s e e introduction to this c h a p t e r ) for s t r u c t u r a l stability of a
s g s t e n i , we ;vi11 show that these conditions allow only a finite number of
c*losa(I paths in a s t r u c t u r a l l y s t a b l e s y s t e m . Note that in v i r t u e of
c:onciition 11, e a c h closed path in a s t r u c t u r a l l y s t a b l e s y s t e m is isolated.
rhis conclusion in i t s e l f , h o s e v e r , d o e s not e s t a b l i s h that the number of
closed paths is finite, s i n c e a condensation point f o r the points of c l o s e d
paths evidently need not belong to a closed path.
T h e n y e tti 21. If syslerri ( A ) is st~-ctctiil-ally stable iri G*, it may hace
otilly u finite niotibel- OJ' closed pntlzs i)t G*.
P i . o o f . Suppose t.?at this proposition i s not t r u e , i.e., a s y s t e m ( A )
h a s a n infinite number of closed paths i n c*. T a k e a sequence L , , L 2 , L a ,
of t h e s e paths and choose a n a r b i t r a r y point on e a c h path. Let Xi, M2,.If3. , . .
be the sequence of t h e s e points, .It, E L i . Since G* is compact, the sequence
{.If. i has a t l e a s t one condensation point, and without l o s s of generality we
may a s s u m e that { - ) I r }:.sa convergent sequence ( i f this is not so, we c a n
always choose a conve:,gent subsequence). L e t the sequence converge to a
point . T I * . T h u s closec. paths completely contained i n c* p a s s through any
neighborhood of .I/*. \Te will noLc- show that i f ( A ) i s a s t r u c t u r a l l y s t a b l e
s y s t e m , no such point .If* may exist in z*. ff N * i s not a state of equili-
b r i u m , w e \vi11 designate by L* the path through Ma. Ii the a - and a - l i m i t
s e t s of the path L* lie i n 6*,w e will denote them by K,z and Kd, r e s p e c t i v e l y .
\C~ehave to consider t h e following a l t e r n a t i v e s :
I ) ,I/* is a s i m p l e focus o r a s i m p l e node.
2 ) .If* i s a s i m p l e saddle point.
3 ) T h e path I.* l e a v e s G* a s t i n c r e a s e s or d e c r e a s e s .
4) K , o r Ku i s a node o r a focus.
.?,I K , or Ku is a closed path.
I
Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS OF STRUCTURAL STABILITY
6 ) K , o r K , is a saddle point.
7 ) K , or K , is a limit continuum, comprising a saddle point and
saddle -to-saddle s e p a r a t r i c e s which are continuations of one another.
Let u s consider each of the seven c a s e s separately.
C a s e 1 is inapplicable, since all the paths passing sufficiently close to
a node o r a focus go to that singular point, and a r e therefore not closed.
Case 3 is inapplicable, since i f L* leaves G*, any closed path L Apassing
sufficiently close to M * a l s o l e a v e s G*. This contradicts o u r assumption
that all the closed paths are completely contained in G*.
Case 7 is inapplicable, since a structurally stable s y s t e m h a s no
saddle -to-saddle s e p a r a t r i c e s .
We can now concentrate on the remaining c a s e s .
Case 4. Let K , be a node o r a focus 0. Let U, (0)be a sufficiently
s m a l l neighborhood of the point 0, s o that all the paths through this neigh-
borhood go to the point 0 for t + 00 and +
a r e thus not closed. Since the path L* h a s
- points in U, (0), the theorem of the con-
c l o s e to M* a l s o p a s s e s inside U , (0), and is

thus not closed, c o n t r a r y to the definition
Case 5. Let K, be a closed path. W e

choose a n a r b i t r a r y point S on this path.
Any neighborhood U, ( S ) contains points o
the path L*, and by the theorem of con-
tinuous dependence on the initial values, i t
a l s o contains points of any closed path L k
passing sufficiently close to M * . But then
FIGURE 44
K , is not an isolated closed path, which
c l a s h e s with the s t r u c t u r a l stability of (A).
Case 6 . Let K , be a saddle point, s o
that L* is a s e p a r a t r i x of one of the saddle points of the s y s t e m . Since a
s t r u c t u r a l l y stable s y s t e m h a s no saddle-to-saddle s e p a r a t r i c e s , t h e r e are
two alternatives: L*leaves c* as t i n c r e a s e s (case 3), o r KO is a node, a
focus, o r a closed path i n 8* ( c a s e s 4 and 5). None of these cases is
applicable, as we have shown above.
Case 2 . Let M* be a saddle point. Consider a sufficiently s m a l l
canonical neighborhood U of t h i s point, limited by a simple closed c u r v e C ,
which is made up of four a r c s without contact and four a r c s of paths
(Figure 44). W e take U to be s o s m a l l that the saddle point M* is the only
equilibrium state of (A) contained in U . By assumption, U contains a n
infinite number of points M I which belong to the paths Lk. However, none
of the closed paths of s y s t e m (A) may be completely contained in U , since
such a closed path, i f i t existed, would not enclose any equilibrium s t a t e s
o r would enclose a t most one equilibrium state, which is a saddle point.
This is forbidden by QT, J11.2, Theorem 30 and Corolldry 1 f r o m
Theorem 2 9 . Therefore e a c h closed path L&h a s a t l e a s t one point ii?k lying
on the c u r v e C . Since t h i s curve is compact, w e can s e l e c t a sequence of
points { a h } converging to s o m e point M * E C . n?* is not a state of equi-
librium, and i t c a n be taken as the initial point M*. W e have thus reduced
our problem to one of the cases 3 - 7, which are inapplicable.
138
None of the c a s e s 1 throuqh 7 is thus applicable. T h i s e s t a b l i s h e s the
validity o i the t h e o r e m .
In QT, $15, we inti-ocluced the concept of o r b i t a l stability and d i s -
t i n p i s h e d between orbitally s t a b l e and orbitally unstable paths. \\e xi11
not repeat the previous r e s u l t s h e r e . It sufiices to s a y that if s y s t e m :A]
is s t r u c t u r a l l y s t a b l e in r e g i o n ti*, i t s orbitally unstable paths in t h i s r e g i o n
a r e a l l the equilibriuni s t a t e s , limit c y c l e s , and saddle-point s e p a r a t r i c e s .
T h e o r e tri 22. [f systetti ( A ) is st,tictusally stable in ti*. it r)iut*h u i v
only a finite nurnber of orbitally tinstable paths and seuiipn!hs in Z;*.
The validity of this theoretn follows d i r e c t l y f r o m Theorem 10,
T h e o r e m 11 ( 7 ) , and T h e o r e m 2 1 ,
Rema r k . T h e o r e m s 2 1 and 22 c l e a r l y r e m a i n valid ..vhen the r e q u i r e -
ment of s t r u c t u r a l stability of s y s t e m (-4)is replaced by a r e o u i r e m e n t that
s y s t e m \.A] satisfies conditions I through 111 (these r e u u i r e m e n t s a r e
actually equivalent, but this still r e m a i n s to be provecl I .
2. Regions w i t h n o r m a l boundary
The proof of sufficiency of conditions I- I11 (see Introduction to this

L*haptcr)for s t r u c t u r a l stability of a s y s t e m in c*
will be crirrieil out f o r a
p a r t i c u l a r case, a s s u m i n g that G'* h a s a so-called n o r n i a 1 b o u n d a r y ,
Regions with a normal boundary w e r e defined in QT, S 1 6 . 2 , ani1 '.IC ,vi11
r e p e a t h e r e the c o r r e s p o n d i n q definition i n full. Note that the r e q u i r e m e n t
o f a n o r m a l boundary does not i m p o s e a significant r e s t r i c t i o n , but i t helps
u s to avoid v a r i o u s complications i n the proof of sufficiency. .
D c j ' i n i t i o n 19. The boiitdary of a cotnpact connected wgion is culled
nortnal jos a gicen dyimtnic sjstern \ A j if the jollozring cowlitions are
sat is.f icd:
1) The boundary i s made i i p of a finite number qf siiriplc closed ciit*<*es.
2) Each of these cr'osed ciirc'es is either a crcle Lc,itlioiit contact os
consists of a jinite nurnber of altetwiting arcs without confact arid a ) ~ qf s
paths. The cotntnon point of an arc of a path and an arc icitliorit conlact
making the boundary rq?illbe called a corner point: a setuipath lying in ti*
ana' tertninating at a c m n e r point will be called a corner settiipnth: an
arc of. a path which is completely contained in G*, except for its end points
which lie on boiindary arcs without contact so that at least one of therii is
a corncr point, will bt? called a comer arc; the cm-esponding arc of c:
path -for ichich neither end point is a c o m e r point will be called a rt,hole
non - s ingzclar arc.
3 F o r any corner arc, only one end point is a cornei- point.
4) None of the corner setnipaths is a saddle-point sepamtris.
5 ) None o j the boundary arcs o j paths belong to a closed path cotti-
pletely contained in 6*.
Conditions 4 and 5 2learly indicate that the boundary a r c s of parhs d o
not belong t o orbitally unstable paths or s e m i p a t h s lying completely in c*.
Sz.!i;c-point separdtrices ma!: oe whole paths or stml:athr. !e:ending on whethi,; the, d r c
xxitdir!rd i i i ??or not.
* * l i t h e boundary of a rcipicn.@ consists of a f i n i t e numbr: c ~simple
t closed non-inr;:crc [ I ~ I pSi t , -
m w t h <:uncs. i t can be f i t t e d by an arhitrdrili close no:mil boundary.
f .ioun,iar) arcs of paths w i l l sometimes be calkt;l for b r e i i t ) hoiinrlar) arcs.
I39
A boundary a r c and a c o r n e r a r c of a path with a common end point,

which thus f o r m a single a r c of a path of s y s t e m (A), a r e said to be a
c o n t i n u a t i o n o f e a c h o t h e r (in the direction of i n c r e a s i n g o r
d e c r e a s i n g t ) . Similarly, a boundary a r c and a c o r n e r path with a
common end point a r e r e g a r d e d as a c o n t i n u a t i o n of e a c h o t h e r .
F i g u r e 4 5 shows a triply connected region 6 with a n o r m a l boundary.
The boundary c o n s i s t s of t h r e e s i m p l e closed c u r v e s , one of which i s a
cycle without contact A, and each of the other two is made up of an even
number of a r c s of paths bi and a r c s without contact I , (1Si,< 5 f o r the e x t e r i o r
boundary curve and 6 - S i < 7 for the i n t e r i o r boundary curve). The a r c s with-
out contact I t a r e m a r k e d i n the figure by s t r a i g h t s e g m e n t s .
The c o r n e r points a r e A i (1 <,cis 1 0 ) and B, (1 ,<j,<4). The figure a l s o
shows the c o r n e r a r c s C3A4, C 6 A 6 ,A,C,, B,Cs; a c o r n e r semipath Ll with an end
point A i going to a stable focus 0,; a c o r n e r semipath L: going to a s t a b l e
l i m i t cycle 2, which e n c l o s e s an unstable node o r focus 02; a saddle point O3
with four s e p a r a t r i c e s .
FIGURE 45 FIGURE 46
The boundary a r c s bz and b, have no continuation in 6*. Each of the other

boundary a r c s h a s continuations i n two directions. The continuation of the
boundary a r c bs i n the direction of d e c r e a s i n g t is the c o r n e r a r c A&,, and
the continuation in the direction of i n c r e a s i n g t is the c o r n e r semipath L i .
Figure 46 shows a boundary a r c h with a continuation i n one direction only
(in the direction of i n c r e a s i n g t ) .
D ef i n i t i o n 20. Orbitally unstable paths, saddle-point separatrices
and corner semipaths, boundary and corner arcs of paths, boundary arcs
without contact and boundary cycles without contact in e* will be r e -
spectively called singular paths, semipaths, arcs (of paths), arcs without
contact, and cycles without contact. A l l the other paths, semipaths, and
arcs of paths will be called non-singular. Singular paths, semipaths,
etc., will also be referred to as singular elements.
140
!16. SLNGULAR PATHS A N D SEhlIPATHS
LVe will now give a complete l i s t of the various paths, semipaths, and
a r c s of a s t r u c t u r a l l y stable s y s t e m (A) in a region c* with a n o r m a l
boundary.
A) S i n g u l a r o r b i t a l l y u n s t a b l e p a t h s a n d s e m i p a t h s :
1 ) An equilibrium s t a t e (a stable node or focus, a n unstable node or
focus, a saddle point).
2 ) A limit cycle (stable or unstable).
3 ) A s e p a r a t r i x which goes to a saddle point f o r t + + m (- OJ) and to a n
unstable (stable) focus, node, or limit cycle f o r t + - 00 ( - a),or which
leaves c* through a boundary a r c or through a cycle without contact a s t
decreases (increases).
B) S i n g u l a r o r b i t a l l y s t a b l e s e m i p a t h s .
4) A c o r n e r semipath going for t + - 00 (" m ) to a n unstable (stable)
node, focus, or limit cycle.
C) S i n g u l a r a r c s a n d c y c l e s w i t h o u t c o n t a c t :
5) A corner arc.
61 X boundary a r c of a path.
7 ) A boundary a r c without contact.
8 ) A boundary cycle without contact.
D) N o n - s i n g u l a r w h o l e p a t h s a n d s e m i p a t h s :
9 ) P a t h s going for I -+ - m to a n unstable and for t + -i- m to a stable
node, focus, or limit cycle ( 9 different possibilities).
1 0 ) Semipaths goin,g f o r t+ - 00 (+ m) to a n unstable (stable) node,
focus, or limit cycle and emerging f r o m c*through a boundary arc or cycle
without contact as t i n c r e a s e s ( d e c r e a s e s ) .
E) N o n - s i n g u l a r w h o l e a r c s :
1 1 j An a r c of a path which is neither a c o r n e r a r c nor a boundary arc,
whose end points lie on boundary a r c s o r c y c l e s without contact, and all
the other points are in G*.
In what follows, stable nodes, foci, and limit c y c l e s of s y s t e m ( A ) lying
in c*(or i n fact only i n G*) will be called a t t r a c t i o n e l e m e n t s or
s i n k s , and unstable Lodes, foci, and limit c y c l e s w i l l be called r e p u 1 -
s i o n e l e m e n t s or s o u r c e s . Boundary a r c s or c y c l e s without con-
tact through which paths e m e r g e f r o m c* a s t i n c r e a s e s c a n a l s o be i n t e r -
preted in a s e n s e a s attraction e l e m e n t s (sinks), w h e r e a s boundary a r c s
and c y c l e s without contact through which paths e n t e r c* may be r e g a r d e d
a s repulsion e l e m e n t s ',sources).
Since a s t r u c t u r a l l y stable s y s t e m i nc* h a s only a finite number of
singular elements, all the propositions regarding the partition of E* into
c e l l s , formulated i n QY, 16> r e m a i n valid. In particular, the s e t E of all
the points of c* which telong t o singular e l e m e n t s is a closed s e t . Its
complement, i.e., the open s e t C * - E , c o n s i s t s of a finite number of c o m -
ponents, called c e 11 s . Each cell is filled with non-singular paths, s e m i -
paths, or a r c s of paths which show "identical" behavior i n a c e r t a i n s e n s e
(see QT, 516, T h e o r e m s 46-48, 5 3 , 5 7 ) . The cells are e i t h e r singly
connected o r doubly connected. In the next chapter w e w i l l consider the
various types of cells of s t r u c t u r a l l y stable s y s t e m s .
Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS OF STRUCTURAL S'iABILITY
I
s17. A REGULAR SYSTEM OF NEIGHBORHOODS AND
THE PARTITION OF -6* INTO CANONICAL NEIGHBOR-
HOODS AND ELEMENTARY QUADRANGLES
1. A r e g u l a r s y s t e m of canonical neighborhoods f o r
structurally stable systems
A s we know, a cycle without contact c a n be drawn around any node or

focus, s o that i t e n c l o s e s no equilibrium s t a t e s o t h e r than that node o r
f o c u s and no closed paths. The cycle can
be drawn i n any a r b i t r a r i l y s m a l l neighbor -
hood of the node o r the focus.':' We w i l l s a y
that t h i s c y c l e w i t h o u t c o n t a c t
belongs to the given node or
f o c u s , and conversely, that t h e n o d e o r
the focus belongs to the particu-
l a r cycle without contact.
The closed region comprising the points
inside this cycle without contact and the
points of the cycle itself w i l l be called a
c l o s e d c a n o n i c a l n e i g h b o r h o o d of
the node o r the focus. All the paths passing
through the points of a canonical neighbor-
hood of a node o r a focus 0, which do not
FIGURE 41 coincide with the point 0, evidently do not
leave that neighborhood, going to 0 f o r
t + -i- 00 (if 0 is a stable node o r focus)
o r for t + - 00 (in the unstable c a s e ) .
Let u s consider a l i m i t cycle Lo of s y s t e m (A). In QT, '324.3, it is
shown that in any neighborhood of the cycle Lo we c a n p a s s two c y c l e s with-
out contact C' and C", one lying inside L O . and the other enclosing L o , so
that the annular region T between the c y c l e s C' and C" contains no equilibrium
s t a t e s and no closed paths other than Lo ( F i g u r e 4 7 j . The closed annular
region T is a union of two closed u n i l a t e r a l canonical neighborhoods T' and
T", limited by the closed c u r v e s Lo and C' and Lo and C", respectively. We
w i l l s a y that t h e c y c l e s w i t h o u t c o n t a c t C ' a n d C* b e l o n g t o t h e
l i m i t c y c l e Lo. T w i l l be called a b i l a t e r a l c l o s e d c a n o n i c a l
n e i g h b o r h o o d , o r simply a canonical neighborhood of the limit cycle Lo.
Any path, o t h e r than L o , p a s s i n g through a point of the canonical neigh-
borhood T goes, without leaving T, to the l i m i t cycle LOf o r t + 00 i f Lo is +
s t a b l e and for t+ - 00 if i t is unstable. A s t d e c r e a s e s or i n c r e a s e s ,
respectively, this path l e a v e s T through one of the c y c l e s without contact C'
o r C". Clearly, f o r any E > 0, the c y c l e s C' and C" c a n be always drawn s o
that T i s e n t i r e l y contained in U, (Lo).
In what follows, we w i l l not d e a l s e p a r a t e l y with the canonical neighbor-
hoods of nodes, foci, o r l i m i t cycles: we will speak invariably of the
canonical neighborhoods of s i n k s or s o u r c e s (attraction or repulsion
elements). We w i l l a l s o s a y that a cycle without contact belongs to a given
sink or s o u r c e , and conversely, that a sink or a s o u r c e belongs to a given
cycle without contact.
I See OT. 18.1. Lemma 3 and the remark to this lemma.
$77. REGULAR SYSTEhl OF C;\NONICAL KFIGHRCXFrOODs
Xpart f r o m canonica; neighborhoods of nodes, foci, and limit cycles,

w e will a l s o consider c m o n i c a l neighborhoods of a saddle point. This
neighborhood is delimited by four a r c s without contact, e a c h c r o s s i n g one
of the saddle-point s e p a r a t r i c e s , and by four arcs of paths ( F i g u r e 44; the
canonical neighborhoods of a saddle point a r e defined i n QT, 519.2). For
any e > 0, we c a n choose a canonical neighborhood of the saddle poini 0
i,vhich is entirely contained in c', (0).
In what follows, we invariably a s s u m e that all the dynamic s y s t e m s
satisfy conditions I through I11 i n the Introduction to Chapter 1-1and consider
regions with a normal bmndar?;.
L e tti tti a 1 . If the a r c s without contact fmtriing the boiinclarj of a
canonical neighborhood gf a saddle point 0 a r e siifficientljj small, each of
these avcs has onlv one coiti)tion point with the corresponding separatvix
of the saddle point 0, atld has no corn))ianpoints with any of the othev
singular paths or setnipaths (i.e., limit cycles, separatrices, a d covneY
soiiipatiis) OY zritli any of the corner a r c s .
P r o o f . The validit:{ of the lenima follows directly f r o m the finite
number of singular path:; and semipaths in the relevant dynamic s y s t e m
( S l S , T h e o r e m s 21 and 32; r e m a r k to T h e o r e m 22 and Definitions 19 and
2 0 ) and f r o m the fact tha.t, i n virtue of condition 111, the saddle-point
s e p a r a t r i c e s of these s y s t e n i s cannot be limit paths (i.e., they do not
e n t e r the limit a - or o-continuum of any path).
L e )ti tn a 2. There exists F,, > 0 such that if the canonical neighbodzod
oj. each saddle Point 0:lies inside L', (Oc), none of the paths throicgh the
poitzts oj. the canonical neighbarhood of one saddle point has points in the
ranonical neighborhood of any other saddle point.
Proof . We f i r s t cht.ose the canonical neighborhoods of s i n k s and
s o u r c e s (nodes, foci, and limit c y c l e s ) i n such a way that they have no
common points. All the saddle points of s y s t e m (A) evidently lie outside
these neighborhoods. Let u s consider the s e t 0 comprising all the c y c l e s
:vithout contact which en':er the boundaries of the canonical neighborhoods of
the s i n k s and s o u r c e s , all the boundary c y c l e s without contact, and all
the open boundary arcs vcithout contact (i.e., the boundary arcs without
t h e i r end points). By condition 111, e v e r y a - s e p a r a t r i x (w-separatrix) of
each saddle point i n t e r s e c t s with i n c r e a s i n g (decreasing) t one and only one
element of the s e t R a t a single point.
Let 0, t i = 1,2, . . ., m ) be the saddle points of s y s t e m (A) lying in G*,
and Lik' ( k = 1, 2, 3, 4) the s e p a r a t r i c e s of the saddle point 0;. Consider a
seniipath which is p a r t of the s e p a r a t r i x LIk), ending a t the point of i t s
intersection with the corresponding cycle or arc without contact of the
s e t ! I . To avoid introducing new symbols, Lik' w i l l denote this semipath.
T h e s e t comprising the saddle point 0;and all the points of the semipaths
.
L?' ( k = 1, 2, 3,4) will be denoted Fi ( i = I, 2 , . . a m). Each of these Fi is
evidently a closed s e t , and they have no common points a s t h e r e are no
saddle-to-saddle s e p a r a t r i c e s . The d i s t a n c e s between any two s e t s Fi are
t h e r e f o r e positive, and t h e r e exist nonintersecting closed neighborhoods tTt
.
o f these s e t s . By ffi ( i = 1 , 2 , . ., k ) we mean a closed neighborhood of
the s e t Pi i n the s e t comprising the points of c* which lie outside or on the
boundary of the canonical. neighborhoods of the sinks o r s o u r c e s . The
neighborhoods i?' are typically "cross-shaped'' (Figure 48).
143
Ch. VI. NECESSARY A N D SUFFlCIENT CONDITIONS OF STRUCTURAL STABILITY I
FIGURE 48
Let u s consider the canonical neighborhoods U , (0,) completely contained

in W , . Let the boundary of U i c o m p r i s e the a r c s without contact I?' ( k =
= 1,2, 3 , 4 ) . If these a r c s a r e sufficiently s m a l l , then f o r both d e c r e a s i n g
and increasing t , e v e r y path passing through the points of these a r c s w i l l
e i t h e r leave the region z* or e n t e r one of the canonical-neighborhoods of the
s o u r c e s or sinks, remaining until that time in the s e t W i . But then the
s a m e property is c h a r a c t e r i s t i c of e v e r y path through any point of the
neighborhood U i .
Let eo be a sufficiently s m a l l positive number, so that for any i =
= 1, 2, . . ., m, U,, (0,) c U i . T h i s number c l e a r l y s a t i s f i e s the proposition
of the lemma, which completes o u r proof.
D e f i n i t i o n 21. Let ( A )be a dynamic system satisfying conditions I
through 111 in the introduction of Chapter Vi. Consider a region with a
normal boundarv. A svstem of canonical neighborhoods of the svstem -, (A)
is sa&d to be Y :Jar.* if the fokowing conditcms are sati .sfied:
1) The can:onitcal neighbmhoods of various s c t ztes of equilibvium are all
disjoint and do not intersect with any of the canonical neighborhoods of the
limit cycles; the canonical neighbmhoods of different limit cycles are
also nonintersectinn;
2) Non:e of the !ths of system ( A )passes thy the cano:nical
neiaht)mhioods of I different saddle boints .
2) Every a& without contact contained in the boundary of the canonical
neighbmhood of a saddle point satisfies the conditions of Lemma 1 , Le., it
has precisely one common point with the cmresponding sebaratrix and has
no common points with any of the other singular paths and semipaths o r
c m n e r arcs.
L e m m a s 1 and 2 establish the existence of r e g u l a r s y s t e m s of canonical
neighborhoods. In what follows, we will only deal with r e g u l a r s y s t e m s of
canonical neighborhoods.
* In QT, 427, the concept of a regular system of canonical neighborhoods was defined for dynamic systems
of a more general type. T h e dynamic systems considered in this chapter satisfy conditions 1-111, and
therefore their regular system of canonical neighborhoods meets the condition formulated below (these condi-
rions are stronger than in the general case).
144
$17. FEGULAR SYFTEN OF (7.4KOhIC;IL KEIGkfBORtlOODS
T h e following 0bviou:j propositions d e r i v e f r o m the above l i s t of paths,

semipaths, a r c s of paths, etc., and f r o m the p r o p e r t i e s of canonical
neighbor hoods :
A ) Any non-singular path completely contained i n E* c r o s s e s one cycle
without contact which belongs to a s o u r c e and one cycle without contact
which belongs to a sink.
B j Any non-singular semipath c r o s s e s one cycle without contact which
belongs to a s o u r c c or a sink, and one boundary a r c or boundary cycle
without contact.
C ) Any a - s e p a r a t r i x (w-separatrix) e i t h e r c r o s s e s one cycle without
contact which belongs tc a sink ( s o u r c e ) or l e a v e s c* with increasing
(decreasing) t through a boundary cycle o r a boundary a r c without contact.
U ) Any positive (negative) c o r n e r semipath c r o s s e s one cycle without
contact which belongs tc a sink (source).
We will now present the terminology and s o m e f a c t s pertaining to c y c l e s
and a r c s without contact comprising the s e t 0 and the division of these a r c s
and c y c l e s into p a r t s by singular paths and semipaths (the s e t 0 is defined
in the proof of L e m m a 2).
Let C be a cycle withqxt contact contained in the boundary of s o m e
canonical neighborhood m in the boundary of the region c*. If this cycle
h a s no common points with singular paths, semipaths, and c o r n e r arcs of
the s y s t e m , i t is said to be f r e e . We s a y that C is a free --cycle (a-
cycle), i f C belongs to a sink ( s o u r c e ) or i f C is a boundary cycle through
which the paths of the s y s t e n i leave E* ( e n t e r E*).
-4 cycle without contact C is said to be n o n - f r e e i f i t h a s a t l e a s t one
coninion point with singular paths, semipaths, o r c o r n e r a r c s (a c o r n e r a r c
niay have a common point with C only i f C is a boundary cycle without
contact). If a non-free cycle without contact C h a s m o r e than one point in
common with singular semipaths, paths, or c o r n e r arcs, i t is divided by
t h e s e points into a finite number of simple arcs, wh.ich have no common
points with the singular elements, except the end points. T h e s e arcs a r e
called s i m p l e e l e m e n t a r y a r c s . If a non-free cycle L'has only one
point JI in common with singular paths, semipaths, and c o r n e r arcs, C is
c a l l e d a c y c l i c e l e m e n t a r y a r c , and JIis the e n d p o i n t o f a
c y c 1i c a r c An elementary arc - whether simple or cyclic - is called
a n e l e m e n t a r y w - a r z ( a - a r c ) o r s i m p l y a w - a r c ( a - a r c ) i f the
cycle without contact C containing this a r c belongs to a sink (a s o u r c e ) or
i f C is a liniit cycle withfmt contact through which the paths of the s y s t e m
leave @ ( e n t e r c").
Boundary a r c s without contact are a l s o divided by points which belong
to singular e l e m e n t s - specifically to s e p a r a t r i c e s or c o r n e r a r c s - into
s i ni p I e e 1 e ni e n t a r y a r c s , which have no common points with singular
elements, except their end points. In particular, a simple e l e m e n t a r y a r c
niay coihcide with a boundary a r c without contact. T h e s e e l e m e n t a r y a r c s
a r e a l s o w - or a-arcs according as the paths of the s y s t e m leave or e n t e r the
region C* through t h e s e z r c s .
We will now formulate without proof s o m e auxiliary proposition which
will be helpful in the following. The proof c a n be found in QT, $27.4 f o r a
m o r e g e n e r a l class of dynamic s y s t e m s (namely, s y s t e m s which do not
n e c e s s a r i l y satisfy conditions I- 111). F r o m the l i s t of paths, semipaths,
etc., i n S l S . 2 and f r o m the definition of free c y c l e s without contact and
elementary a r c s , i t follows directly that e v e r y non-singular path of a

dynamic s y s t e m (A) lying in c*, whether a whole path, a semipath, or a
whole a r c , c r o s s e s one a - a r c or f r e e a -cycle and one a - a r c or
f r e e a-cycle.
L e m m a 3 . All the paths passing through the points of one f r e e a-cycle
(a-cycle) cross the same f r e e a-cycle (a-cycle), one of the two cycles lying
inside the other.
R e m a r k . The paths of this l e m m a a r e non-singular (from the definition
of a f r e e cycle). They a r e e i t h e r a l l whole paths, or a l l semipaths, or a l l
whole a r c s . F r e e a - and a - c y c l e s through which p a s s the s a m e paths a r e
called c o n j u g a t e .
L e m m a 4 . All the paths passing through the (inner) points of one a - a r c
(w-arc) C Y O S S the same a - a r c ( a - a r c ) .
Remark . H e r e again we a;-e dealing with non-singular paths. Two
elementary a r c s without contact through which p a s s the s a m e paths a r e
called conjugate e l e m e n t a r y a r c s . Two conjugate elementary a r c s of a
s t r u c t u r a l l y stable s y s t e m a r e both simple elementary a r c s ; alternatively
one of them is simple and the other is cyclic.
If (A) is a s t r u c t u r a l l y unstable s y s t e m , i t may have two conjugate cyclic
a r c s . However, s t r u c t u r a l l y unstable s y s t e m s have no conjugate cyclic
a r c s . T h i s w i l l be Droved i n ChaDter VII. 819.3. L e m m a 3.
FIGURE 49
In Figure 49, the simple a - a r c s AiBl and BzCl a r e the conjugates of the
simple a - a r c s A B and BC, and the simple a - a r c B,B2 is the conjugate of the
cyclic a - a r c I' with M o a s i t s end point. All these a r c s a r e contained i n
the boundary of the doubly connected region 6*.
2. The partition of the region c*

into canonical
neighborhoods and e l e m e n t a r y quadrangles
We again r e g a r d (A) a s a dynamic s y s t e m which s a t i s f i e s conditions I-

111 of the Introduction i n region 6*with a n o r m a l boundary. Let some
r e g u l a r s y s t e m of canonical neighborhoods be defined in e*. Let K be the
s e t of all the i n t e r i o r points of all the canonical neighborhoods. The points
of G* which do not belong to K constitute the closed s e t G * - K .
5438 I46
017. REGULAR SYSTEhI OF C l N O N l C A L NEIGHBORHOODS
L e ) n ni a 5. The closed set C* - h' can be partitioned into a finite nuinber

of closed elementary quadrangles, s o that ecery arc withoiit contact con-
tained in the boundary cf any of these quadrangles is either a part of a
boundary cycle without ,contact, w a part of a cycle without contact which
belongs to some sink OY source, or a part of a boundary arc without contacl,
or an arc without contact contained in the boundary of the canonical
neighborhood of a saddle point.
.\io reove r,
a) any two qruzdrangdes o j the partition either hace no coinmon points or
their common points form an arc of a path which is part of the boundary
of each quadrangle;
b) an arc without colitact forming part of the boundary o j a quadrangle
of the partition either hirs one point, othe,' than the end point, in cornwon
with the saddle-point separatrix and no other comtiion points icith singular
elements; or one of its end points belongs to a c o m e r semipath OY a
corner arc or is a COYIZCY point of the boundary, and all the other points
belong to non-singular paths; w it entirely consists of points which belong
to non-singular paths.
Re ma r k . A partition of c* into canonical neighborhoods forming a
r e g u l a r s y s t e m and into e l e m e n t a r y quadrangles satisfying the conditions
of L e m m a 5 will be called a r e g u l a r p a r t i t i o n o f c*.
P r o o f . F i r s t l e t u s c o n s i d e r the canonical neighborhoods of the
saddle points of s y s t e m (A). Let 0 be s o m e saddle point, H i t s canonical
neighborhood, y a n a r c without contact
e n t e r i n g the boundary of H , d and B the
end points of this a r c , L a s e p a r a t r i x of
the saddle point O c r o s s i n g the a r c y,
D the i n t e r s e c t i o n point of L and y
( F i g u r e 5 0 ) . To fix o u r ideas, l e t L be
L, ff' a n a - s e p a r a t r i x . Then, f o r i n c r e a s i n g
t , L l e a v e s the neighborhood H through
FIGCRE 50
the point D and c r o s s e s e i t h e r a cycle
Without contact C ( t h i s is a boundary
cycle or a cycle belonging t o a s i n k ) or
a boundary arc without contact. L e t the s e p a r a t r i x L c r o s s the cycle with-
out contact C a t point D' (,if L c r o s s e s a boundary arc without contact, the
argument r e m a i n s the s a m e ) . F r o m the definition of a regular s y s t e m of
canonical neighborhoods we see that all the paths through points of the arc 'J
will c r o s s with increasir-g t a n a r c without contact y' which is p a r t of the
cycle C. The end points of y' ( .4'and B ' ) l i e respectively o n the paths L A and
L B through the end points A and B of the arc y, and y' h a s no points which
belong to singular paths, except the point D'.
T h e arcs y and y' will be called c o n j u g a t e a r c s . The quadrangle 1
delimited by y and y' will be called a n e l e m e n t a r y quadrangle.
Thus, to e a c h a r c without contact y e n t e r i n g the boundary of a canonical
neighborhood of a saddle point c o r r e s p o n d s a n e l e m e n t a r y quadrangle. Let
A , ( i = 1, 2, ..., .Ir) be all :such quadrangles, y , and y; the corresponding a r c s
without contact. The quadrangles A i are evidently all disjoint (see
Definition 21, 2 ) .
Let u s now c o n s i d e r a c o r n e r path f. which e n d s at a c o r n e r point R of
the boundary. Let be a negative semipath. Then h h a s a common point A
I47
I
with a cycle without contact C which belPngs to a s o u r c e (Figure 51). On
c
the cycle we choose two points A and B on the two s i d e s of R and sufficiently
c l o s e to i t so that the a r c A B of the cycle
contains no points of the previously defined
I
arcs y; and no o t h e r points of any c o r n e r
paths, except the point R. Let u s consider
the a r c s RA and RB s e p a r a t e l y , denoting
them i , and x2.
A s t i n c r e a s e s , the paths through the
points of one of t h e s e a r c s (in F i g u r e 51,
through the points of x 2 ) w i l l c r o s s the a r c
x 2 which is p a r t of a boundary a r c without
contact and h a s R as one of i t s end points
(all the other points of xz belong to non-
singular paths). The paths through the
FIGURE 5 1 points of the second a r c , xi, will c r o s s with
i n c r e a s i n g t the a r c xi, which is a l s o p a r t
of a boundary a r c o r cycle without contact,
o r (as in F i g u r e 51) is p a r t of a cycle without contact which belongs to a sink.
One of the end points of xi - w e denote it by R, - belongs to a c o r n e r a r c o r
semipath o r is a c o r n e r point of the boundary, and all the other points of the
a r c xi belong to non-singular paths. The quadrangle delimited by the a r c s
without contact xi and it (i = 1, 2) and the a r c s of the paths through t h e i r end
points is the e l e m e n t a r y quadrangle A,. The a r c s xi and xi will be called
conjugate, as before.
W e take all the c o r n e r s e m i p a t h s k i , whether positive o r negative, and
c o n s t r u c t the corresponding e l e m e n t a r y quadrangles A, in the s a m e way.
Note that i f t h e r e a r e two different semipaths, a positive and a negative one,
which are a continuation of the s a m e boundary a r c of a path, the two c o r r e -
sponding arcs x and x w i l l be chosen "compatibly." This is b e s t illustrated
with a n example: in F i g u r e 51, the semipaths 2 and L1 with the end points R
and R,, respectively, a r e continuations of the s a m e boundaryAarcR,k,. We
may t h e r e f o r e choose a r b i t r a r i l y one of the two arcs xi and xi, and the other
a r c is automatically determined by this choice. Both t h e s e a r c s define
the s a m e e l e m e n t a r y quadrangle A,.
Let b j ( j = 1, 2, ...,A) be all the f i f f e r e n t e l e m e n t a r y quadrangles constructed
in this way. All the a r c s xt (and x,) are taken sufficiently s m a l l , so that they
have no common i n n e r points between themselves and no common points
(whether i n n e r points o r end points) with the previously defined a r c s y , .
Under t h e s e conditions no two e l e m e n t a r y quadrangles A , ( i = 1, 2, ...) and
6 , ( j = 1, 2, ...) have any i n n e r points i n common (but and d, may have
common a r c s of paths i n their boundaries or p a r t s o,f boundary a r c s . T h e
second c a s e is observed, e.g., for the quadrangles A i and A2in F i g u r e 51).
The e l e m e n t a r y quadrangles A, adjoin c o r n e r paths (and possibly their
continuations also; for example, the quFdrangle A2 i n F i g u r e 51 adjoins the
semipath L , and 6, adjoins the semipath L , i t s continuation - the boundary
a r c RR, - and the continuation of this a r c , the semipath Li).
Following the s a m e p r o c e d u r e a s f o r the quadrangles A j , we define the
e l e m e n t a r y quadrangles A i (k = 1, 2, ..., N*)adjoining c o r n e r a r c s . The a r c s
without contact forming the boundaries of the quadrangle A i will be d e -
signated hk and h f ; t h e s e a r c s will a l s o be called conjugate. Note that a
148
51 7. REGIJLAR SYSTEM OF CANONICAL NEIGHdORHOODS
quadrangle adjoining a ,corner semipath may a t the s a m e time adjoin a

in F i g u r e 52). Therefore, A i ( k = 1, 2. ..., S * ) a r e
c o r n e r a r c a l s o (e.g., ;i2
defined specifically a s e l e m e n t a r y quadrangles adjoining c o r n e r arcs, but
not adjoining c o r n e r semipaths. We moreover take the arcs hR and i.: to be
sufficiently s m a l l . Under t h e s e conditions, the quadrangles A i , A j , and A:
a r e a l l different and no two of t h e s e quadrangles have i n t e r i o r points in common
(but and A; may have common a r c s of paths in their boundaries or p a r t s
of boundary arcs; such are, e.g., the p a i r s of quadrangles 1; and &, 3, and
A?, A; and A; in F i g u r e 52).
FIGLIRE 5 2
Let u s now c o n s i d e r boundary a r c s of paths. Those which are the

continuations of c o r n e r arcs O r c o r n e r semipaths f o r m the boundary of the
quadrangles A; and A, (such are, e.g., the arcs S,Tland S ? T , i n F i g u r e 52,
which f o r m the boundaries of & and A t a respectively). &ow suppose that
a boundary arc of a path ST h a s no continuation i n G*. It is readily s e e n
that the c o r n e r points of this arc belong to two different boundary arcs
without contact. Take a sufficiently s m a l l section p of one of these arcs,
adjoining a c o r n e r point (e.g., the section SB1 i n Figure 5 2 ) and draw
through the points of p paths of the s y s t e m until they e m e r g e through a
section c( of the second boundary a r c without contact. We have thus f o r m e d
a n e l e m e n t a r y quadrangle adjoining a boundary a r c of a path ST (in
Figure 52, this is the quadrangle XI). Let ( I = 1, 2, ..., .V) be all such
quadrangles. ' T h e i r boundaries contain the conjugate arcs without contact
p,,, andFm.
Let u s now consider all the e l e m e n t a r y quadrangles A t , ij,A:, and 5, and
all the arcs without contkct 91, x, and xi, >.A and Xt, p, and FL entering t h e i r
boundaries (except the arcs yi. entering the boundaries of the canonical
neighborhoods of saddle points). For simplicity, w e will designate all these
quadrangles by Ai (i = 1, 2 , ..., s, where s = -V f . S 4- ,V* + .V) and the relevant
arcs without contact by y'? (y(m0)) ( m = 1, 2, ..., k) if they are p a r t s of c y c l e s
without contact which belong to s o u r c e s (sinks) or i f these arcs belong to
I49
Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS O F STRUCTURAL STABILITY
I
the boundary of e*, the paths of s y s t e m (A) entering z* (leaving c*)through
them."
Let u s f u r t h e r consider all the non-free cycles without contact &Or) which
belong to s o u r c e s , and a l l the boundary a r c s without contact and the
non-free cycles without contact Z(")through which paths of s y s t e m (A) ente?
G'*. All the a r c s y $ ) (m = 1, 2 , ..., r)belong to these cycles without contact and
boundary a r c s C(nJ,C ( a l , and Z(Or). Removing f r o m each cycle CCa)and Z(Or)and
f r o m each boundary a r c c(") the points of the a r c s y e ) which belong to the
corresponding elements, we obtain a finite number of open a r c s without
contact which have no common points, The c l o s u r e s of t h e s e a r c s a r e
designated ai"), u p ) , ..., a?). Note that the end points of each of t h e s e a r c s
belong to non-singular paths.
Similarly t r e a t i n g a l l the non-free cycles without contact C(") which belong
to sinks and the boundary a r c s without contact do)and the c y c l e s Z(O), w e
remove f r o m them the points of the a r c s yg) to obtain a r c s without contact
whose c l o s u r e s a r e designated aio), u p ) , ... .
Let u s now consider a l l the paths which a t s o m e t p a s s through points of
the a r c aia) (i = 1, 2, ..., p ) . It is readily s e e n that a s t i n c r e a s e s , they a l l
c r o s s one of the a r c s a(") (by a n appropriate choice of our
notation, we c a n e n s u r e that t h i s is the a r c a$*)). Hence
&<@/ it follows directly that the number of a r c s a$@)is exactly
(23
equal to the number of a r c s aiR), Le., p i n both c a s e s .
The a r c s of paths extending between the a r c s without
contact a{") and a!") f o r m an e l e m e n t a r y quadrangle, which
we designate A : ( i = 1, 2, . . , , p ) . All the quadrangles A;,
like the quadrangles A i , a r e e l e m e n t s of the s e t G* - K .
Finally, l e t u s consider the f r e e a - c y c l e s of s y s t e m (A)
FIGURE 53 i n G* (if any). They a r e designated Bia) (i = 1, 2, ..., q). A
cycle BI") is conjugate, a s we know ( s e e 818.1, L e m m a 3),
with the free a-cycle EloJ, and both t h e s e c y c l e s delimit
an annular region filled with sections of non-singular paths (Figure 53).
Drawing t h r e e path sections i n each of these annuli, we partition them into
e l e m e n t a r y quadrangles A; (i = 1, 2, ..., 3q). They a r e all evidently e l e m e n t s
of the s e t E* - K .
It is readily s e e n that each point of the s e t E* - K belongs a t l e a s t to one
of the quadrangles A; ( j = 1, 2, ..., p ) , A i (i = 1, 2, ..., s), and A i ( k = 1, 2, .... 3q),
and that a l l these quadrangles satisfy the conditions of L e m m a 5. T h i s
completes the proof of the l e m m a .
F i g u r e 52 shows a doubly connected region c* with a n o r m a l boundary.
It c o m p r i s e s two canonical neighborhoods, that of a n unstable node Oi and
that of a saddle point 02. The complement c*- K is partitioned into 1 5
elementary quadrangles. T h i s partition m e e t s a l l the r e q u i r e m e n t s of
the lemma.
S18. T H E FUNDAMENTAL THEOREM O F STRUCTURAL

STABILITY O F A DYNAMIC SYSTEM
1. Lemmas
We w i l l give h e r e a number of l e m m a s that will be needed i n
connection with the proof of the fundamental theorem of s t r u c t u r a l
* ~2)
It is readily seen that the number of the arcs y!,? is equal to the number of (we designate this common
number by r ) . If there are no saddle points in @, then a = r . If 6 has k saddle points, then r = - 2k.
31s. THF FLiiNr)AhlFNTAL THEOFEL1 OF j T R I ' C T ( R \ L ST.IGIL1 r Y
stability of a dynamic s y s t e m . Some of these l e m m a s a r e presented

u.ithout proof.
L e i n m a 1 . Let (A! be a dgtuzitiic system defined in and let c,
?i t @ c G) be a region with a troriti:l borindar_v. For an?; F > 0, there exists
e
a region with n normal borindary G'* such that CG'*C@* c G , and all the
points of e* which are not points of C* ure contained in an e-neighborhood
OJ'tire boundary of@; r.uoreoL3es,the boiinclaries of C* and z**have
identical sclrerties .*
Lemma 1 is g e o m e t r i c a l l y self -evident, and i t s proof is omitted.
Region c** which satisfies the proposition of the l e m m a w i l l be called a n
e s t e n s i o n or a n E - e >: t e n s i o n of e.
F i g u r e 54 shows a doubly c o n -
nected r e g i o n @ and i t s extension. The r e g i o n @-@ is c r o s s - h a t c h e d .
R e m a r k . Lt' conditions I - I11 formulated in the Introduction to t h i s

c h a p t e r are s a t i s f i e d i n @ a n d i f E > O is sufficiently s m a l l and z**
is a n
?-extension of 0 , systet-1 (A) h a s no equilibrium states and no c l o s e d paths in
c** o t h e r than those that i t h a s in c*, i.e., conditions I- I11 are a l s o s a t i s f i e d
i n G**,
The truth of the r e m a r k r e g a r d i n g the equilibrium s t a t e s follows f r o m
t h e fact that the equilibrium s t a t e s of a s y s t e m cannot lie a r b i t r a r i l y close
t o the boundary of @, s i x e t h e r e a r e no equilibrium s t a t e s on the boundary.
T h e t r u t h of t h e r e m a r k r e g a r d i n g c l o s e d paths is proved by the s a m e
argument that we have used in the proof of T h e o r e m 21.'";
L e H I tn a 2. Let 1 be a sitnple a s c , -11, and .It2 two points on this a r c .
I.t' r) > 0 is sujficiently siiiall, and -It; and -!I; ase trco points of the a r c 1
lying respectively in the q-neighborhoods of M , and M z , the direction of

the arc 1 defined by the motion f r o m M , to M z coincides with the direction
defined b y the motion f r o m M ; to M ; (Figure 5a.
L e m m a 3. Let C be a simple closed curve, Mi (i = 1, 2, ..., n ; n>3)
points lying on this curve. Let these points be ordered M , , M z , ..., M,, fm
the motion in a certain direction (one of the two possible directions) along
the curve C. Then i f '1 > 0 is sufficiently smull, and MI (i = 1, 2, ..., n) are
points of the curve C so that M ; E U, ( M i ) , the motion in the same direction
along the curve C will find the points M ; in the order M ; , M i ,... , M ; (Figure56).
The proof of L e m m a s 2 and 3 is omitted.
Together with a given s y s t e m (A), we will consider modified s y s t e m s
(x) sufficiently c l o s e to (A).
FIGURE 5 5 FIGURE 57
L e m m u 4 . Let L be a structurally stable limit cycle of system ( A ) ,

i.e., a limit cycle with a son-zero characteristic index), and y a
canonical neighborhood (ring, of the limit cycle bounded b y cycles without
contact C , and C2. Then for any B > 0, there exists 6 > 0 such that i f
system (a) is 6-close to ( A ) , then
(a) C, and C z are cycles without contact for the paths of system (A), and
these paths cross each of the cycles C, and C z in the same direction as the
paths of the original system ( A ) ;
(6) the ring y contains a single closed path t" of system (A), and this Z is
a stable structurally stable limit cycle i f L is stable, and an unstable
structurally stable limit cycle if L is unstable;
P r o o f . To fix ideas, l e t u s consider a s t a b l e l i m i t cycle L. W e choose

some point M o on this cycle and draw a n o r m a l to the path L through t h i s
point. Let 1 be a segment of t h i s normal, P,, Q, its end points, Liand Lz
the paths of s y s t e m (A) passing through P1 and 9,. respectively. We a s s u m e
that M o is an i n t e r i o r point of the segment I . Moreover, the segment I is taken
to be so s m a l l that 1 is an a r c without contact for the paths of s y s t e m (A)
152
$18. THE FUNDAMENTAL 1HEOREXL OF STRUCTURAL STABILITY
and the paths L, and L2 with i n c r e a s i n g t w i l l again c r o s s the segment I a t

points P 2 and Q2> the a r c s P i p 2 and QIQz of t h e s e paths lying completely
inside y ( F i g u r e 57; s e e 112.1).
Let I', be a simple closed c u r v e made up of the turn P I P Iof the path Lt and
the segment P I P pof the n o r m a l 1 ; rZ is the analogous c u r v e consisting of the
t u r n Q,Qz of the path L2 and the segment QtQz of the n o r m a l 1 ( F i g u r e 57). We
u s e Wto designate the region between the c u r v e s rl and rn. Clearly,
L c N' c y .
F r o m the definition of a canonical neighborhood, y and therefore Wdo
not contain any closed paths of (A), except L. T h e r e f o r e , a s t d e c r e a s e s ,
the paths L , and L2 passing through the points P , and Q1 will e m e r g e f r o m 7
through the points P o and Qo lying on the c y c l e s without contact C, and CzI
respectively ( F i g u r e 57).
Take s o m e point S on the segment P I P 2of the n o r m a l I and s o m e point R
on the segment Q1Q2. The paths L; and L; of s y s t e m (A) through these
points e m e r g e with d e c r e a s i n g t
f r o m the neighborhood y through the
respective points So and R, of the
c y c l e s C , and C2 ( F i g u r e 57). The
a r c P O P tof path L , and the a r c SoS of
path L; partition the region between
the closed c u r v e s C,and rr into two
e l e m e n t a r y quadrangles, which we
denote A I and A; ( F i g u r e s 57, 58).
Also the a r c QoQl of path Laand the
a r c RoR of path Li partition the r e -
gion between the c u r v e s C2and r2
into two e l e m e n t a r y quadrangles A,
and A;.
Let (6)be a dynamic s y s t e m
sufficiently c l o s e to (A), and &, &.,
FIGURE 55 g, g, L;, E;, &, e t c . , the e l e m e n t s
of (A) corresponding to the e l e m e n t s
P,, Q2, S, R. L,, L;, A,* e t c . , of (A)
(it is a s s u m e d that the points P , , Q,, the nor-mal I, and the closed c u r v e s C ,
and C , do not change on passing to s y s t e m (A)).
In Chapter V , 14, in our proof of T h e o r e m 18 (on the s t r u c t u r a l
stability of a simple l i m i t c y c l e ) w e established that_for any E , > O , t h e r e
e x i s t s drp>O with the following property: if s y s t e m (A) is a,-close to
s y s t e m (A), then
(E,
A)
e1
EE
-
(P,Z),
and the mapping Tiwhich r e a l i z e s this relation i s defined i n and can be

chosen so that
( P I )= P I , Tt (QJ =Q, and Ti( I ) = 1. (3)
We will now u s e L e m m a 9 of J4 (Chapter 11, 54.2). By t h i s l e m m a , for

any e,>O, W e c a n find two n u m b e r s d,>O, q > O with the following property:
i f s y s t e m (A) is a,-close to s y s t e m (A) and a topological mapping 'p is given,
153
Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS O F STRUCTURAL STABILITY
t r a n s f o r m i n g the a r c s P0P,P, SSo, Qo&Qz and R R o of the paths of s y s t e m (A)

into the a r c s PaP,Pz, ss0,QoQ,Qz and RR,of the paths of s y s t e m (A), r e s p e c -
tively, and the a r c s without contact SP2 and RQ2 into the a r c s gp, and I?&,
and 'p is an q - translation, then the mapping 'p c a n be continued to a
mapping T , which moves the quadrangles A; and A; into and a; a;,
respec-
tively, c o n s e r v e s paths, is a n &,-translation, and coincides with the
mapping 'p w h e r e v e r the l a t t e r is defined.
By L e m m a 8, $4 (ChapJer 11, 4.2), f o r any ~ 3 > 0 , we c a n find such 6,>0
and q * > O that if s y s t e m (A) is 8, -close to (A) and a topological mapping 'p* is
given moving the a r c s without contact S P , a%d RQ, of s y s t e m (A) into the
a r c s without contact gPi and RQ, of s y s t e m (A), respectively, and this 'p* is
an q* - t r a n s l a t i o n , the mapping 'p* c a n be continued to a mapping T 3 which
moves the e l e m e n t a r y quadrangles A, and A, into &, and A,, r e s p e c t i v e l y ,
c o n s e r v e s paths, is an translation, and coincides with the mapping 'p* on
the s e g m e n t s SP, and RQ,.
Finally, l e t 8 , > O be s o s m a l l that i f s y s t e m (A)
is a4-close to s y s t e m (A),
the c y c l e s without contact C , and C, and the a r c without contact 1 of s y s t e m
(A) a r e r e s p e c t i v e l y cycle_s without contact and an a r c without contact of
s y s t e m (A), the paths of (A) c r o s s i n g each of the c y c l e s C, and C, and the
a r c 1 i n the s a m e d i r e c t i o n s a s the paths of (A).
Let E > O . We take & , = E and for this e, find the n u m b e r s 6, and q . We
take & , < E , e 3 < q , and for this E~ we find the n u m b e r s q* and 6,.
We take E , < E , E, <q, E, < q*, and f o r this E , find 8 , .
Finally, we take
6 < min {dl, a,, 8,, h,}.
The number 8 > 0 obtained i n this way m e e t s all the r e q u i r e m e n t s of the

lemma.
Indeed, l e t s y s t e m (A) be 6 -close to s y s t e m (A). Xhen, according to
the choice of 8 , , t h e r e e x i s t s a mapping Tiof onto w which s a t i s f i e s
conditions ( 2 ) and ( 3 ) .
F u r t h e r , since 6<6, and E , (q', E, < E , w e c a n c o n s t r u c t a mapping T3
which moves the e l e m e n t a r y quadrangles A, and A2 into 3, and a,, r e s p e c -
tively, and h a s the r e q u i r e d p r o p e r t y ('p+ is identified with the mapping T,
previously defined on the s e g m e n t s P , S and Q , H ) .
Finally, since e z = e , e , < q . e3<q2 and we c a n c o n s t r u c t a mapping
Tzof the quadrangles A; and A; onto A; and a;,
respectively, which h a s the
r e q u i r e d p r o p e r t i e s ('p is identified with the mappings T 3 and T, previously
defined on the corresponding s e g m e n t s of the boundaries of A; and A i ) .
It is r e a d i l y s e e n that the mappings T , , T 2 , T 3 jointly define a mapping T
of the canunical neighborhood y ozto itself which is path-conserving and is
a n & - t r a n s l a t i o n . Therefore, if (A) is 6-close to (A), we have
(y. A ) :ty, A",.

i.e., proposition (c) of the l e m m a is satisfied. Proposition (b) follows
d i r e c t l y f r o m ( c ) . Proposition (a) is satisfied in v i r t u e of the peculiar
choice of the number 8, and the r e l a t i o n 8 <6,. T h i s completes the proof
of the lemma.':
'
Figure 58 shows the numbers e, 8 , q corresponding to t h e regions W , A,, A;,A,, A;. The fint step in the
construction of the mapping is the construction of T 1(in W),then the construction oi T3(in the quadrangles
Aland Az), and finally t h e construction of TZ (in t h e quadrangles A; and Ai).
154
The following l e m m a is a s t r o n g e r version of Leninia 4.
L E i n ttia 5. Let L be a structiwally stable liinit cycle of' systetri (ti),
its canonical neighborhood limited by cycles withoilt contact
y and C2.
Thcn, for an! F > 0 there exist q > 0 am1 6 > 0 with the following property;
~
ij' system (8) is 8 -close to systetn ( A ) , and 'p is a topological ntapping

tnoving each of the cycles C1 and C2 into itselj. which is an 11-translation,
theti C , and C: are cycles nithoiit contact f o r system ( K ) and there exists a
riiapping T qf the neighborhood y into i t s e u rc8hichis at1 e-tYWSldiOn,
which conserces paths, and which coincides with the mapping 'p on the
houmiary of the neighbovhood y ( i . e . , on the cycles C, and C,).
P r o (3 f of Lemma 5 is analogous to the proof of L e m m a 4, differing only
in a n obvious modification of the argument. Specifically, w e f i r s t construct
the mapping I' of the e l e m e n t a r y quadrangles
! . S2. 11,
AI, A ; , ~ , , ~ ; o n t o I 3;. respectively,
s o that i t coincides with the mapping q on
the boundary of ;* . T h i s mapping induces a
mapping of the s e g m e n t s PIP2and QlQ2 onto
the s e g m e n t s PIP22nd ol&,
respectively
(the points P , and P I need not coincide in this
c a s e , whereas in L e m m a 4 this was the s a m e
point. T h i s a l s o applies to the points Q,and
0,). T h i s incluced mapping is continued f i r s t
to a mapping of P,Q,pnto plgl, and then to a
mapping of If' onto t i ' by the technique deve-
loped in the proof of T h e o r e m 18 (see 5i4).
N e will r e q u i r e two f u r t h e r l e m m a s . The
'IC.r.PF ;;
f i r s t d e a l s with the neighborhood of a saddle
point.
Let 0 be a s t r u c t u r a l l y stable saddle point
of s y s t e m (A), y i t s canonical neighborhood
limited by a r c s without contact i\%', l y ' , l!?'. i?)'': and arcs of paths C,C,, C 2 B J ,
HIC;, and B 2 E , (Figure 59). T h e s e p a r a t r i c e s of saddle point 0 c r o s s i n g the
a r c s without contact will be designated L?', LL-). t!,'".L:"', respectively.
Let j.':"' (i-3. 4 ) be th? "elongation" of the a r c without contact IF), i . e . ,
a n a r c without contact incorporating liu' whose end points a r e not the
end points of li"'(in this way, all the points of 1:" a r e the inner points of L ! ~ ' ) .
Let be a dynamic s y s t e m which is sufficiently c l o s e to (A). A s t d e -
c r e a s e s , the paths of (11) passing through the end points Ci and B i of the arcs
without contact l?' ( i = 1. 2 ) will c r o s s the a r c s >d')(t = 3 , 4 ) at four points
designated Ea. &. c*,
E., respectively. The sections T,o, and ?,Ei of the
a r c s ).:'"'and ?.\"'will be designated TLw) and Tia', respectively. Let denote
the region limited by the a r c s I , , I,. G , and lT and the sections C,E3. C2g3. B,C,,
and H& of the paths of s y s t e m is).
Finally, 'p is a topological mapping
moving each of the arcs I , and I? onto itself, so that the end points of i, and
l 2 r e m a i n fixed, and the i n t e r s e c t i o n points of t h e s e a r c s with the s e p a r a -
t r i c e s L'?) and LY', resFectively, move into intersection points with the
s e p a r a t r i c e s 27'and i'?"'.
L e t n t n a 6 . F m an,y E > 0, there exist 6 > 0 and q > 0 with the folloiring
property: if systetn (A; is 6 -close to spsterri (A) and the mapping 'p is an
q-translation, the arcs 1,. IP, , and ?,. are a r c s without contact for the
. A'+ w i l l s w i e t i m e s use rhc snorter notatioi! I , , I?, I,. I,.
I55
Ch. VI. NECESSARY A N D SUFFICIENT CONDITIONS OF STRUCTURAL STABILITY
paths of (A) and there exists a mapping T of region y onto which coincides
with cp on the arcs LY)and @', conserves paths, and is an e-translation.
Remark . It evidectly follows f r o m L e m m a 6 that 7 contains a single
I
equilibrium s t a t e 0" of (A), a
being a saddle point and 5 i t s canonical neigh-
borhood. Also I
e , -
(Y?A) = (Y7 A).
P r o o f of L e m m a 6 is analogous to the proof of L e m m a 4, S9.2, f r o m

which i t is obtained by obvious modifications.
L e m m a 7. Let 0 be a structurally stable state of equilibrium of
system ( A ) , which i s either a node or a focus, y is its canonical neighbm-
hood limited by a cycle without contact C . For any_e > 0 , there exist q > o
and 6 > 0 with the following property: if system ( A )i s 6 -close to system ( A )
and cp is a topological mapping of cycle C into itself, which is an q -
translatian, then C is a cycle without contact f w system ( A ) and there exists
a mapping T of v*intoitself which is a path-conserving e-translation that
coincides with the mapping cp on the boundary of y (Le., on C).
P r o of of L e m m a 7 is analogous i n a l l r e s p e c t s to the proof given in
the r e m a r k to Theorem 1 2 (58.2).
2. The fundamental t h e o r e m f o r a plane region
T h e o r e m 23. F m a dynamic system ( A )defined in a plane region to

be structurally stable in a region C*with a normal boundary (C*c G ) , it is
necessary and sufficient that Conditions I through III in the Introduction to
Chapter VI be satisfied:
I . System ( A )has in G* only a finite number of equilibrium states, which
are simple nodes, saddle points, m foci.
II. The closed paths of system ( A )in G* are simple limit cycles.
III. System ( A )has no saddle-to-saddle separatrices in G*.
Proof . The necessity of conditions I- 111 for s t r u c t u r a l stability of a
s y s t e m (A) i n c*
h a s been proved i n previous c h a p t e r s (S7, T h e o r e m s 1 0 and
11; l o , Theorem 15; 815, T h e o r e m 20; $11, T h e o r e m 16). We thus have
to prove the sufficiency of conditions I- I11 for the s t r u c t u r a l stability of a
system.
Suppose conditions 1-111 a r e satisfied. By L e m m a 1, 818.1, and the
r e m a r k to that l e m m a , for a sufficiently s m a l l (T > 0, any o-extension of 6
contains no equilibzium s t a t e s and no closed paths of system (A) other than
those contained i n G*. Let be such a n extension of G*. We will show that
h a s the-following property: f o r any e > 0, t h e r e e x i s t s d > O such that if
s y s t e m (A) is 8 -close to s y s t e m (A), then
= -
(e,A ) = (H,
8
A), ( 41
where B is a region. Since c**c H , t h i s implies, by definition, that

s y s t e m (A) is s t r u c t u r a l l y stable i n G*.
Clearly, B h a s a n o r m a l boundary and s y s t e m (A) s a t i s f i e s conditions I-
w
I11 in t h i s region. Thus t h e r e e x i s t r e g u l a r partitions of into canonical
I56
IS. THE FUNDAhlENTAL THEOREM OF STRUCTUR>.L STABILITY
neighborhoods and e l e m e n t a r y quadrangles (S17.2, L e m m a 5). W e choose

and fix one of t h e s e partitions, which w e denote n .
We introduce the following notation:
Ui ( i = 1,2, . . . , p ) are the canonical neighborhoods of the s o u r c e s and the
sinks of the partition G;
.
V , ( j = 1, 2, .. , q) are the canonical neighborhoods of the saddle points 0,;
l$), @, I n ) , I:? ( i =1, 2, . . ., q) are a r c s without contact making up the
boundary of the canonical neighborhood of the saddle point Oi;
F g ) , F:?, Fi:), F i y ) (i = I, 2, ..
.,q) are the e l e m e n t a r y quadrangles of the
partition ll whose boundaries incorporate the a r c s l\y),l$y, l$y), respec-
tively. W e will u s e the following abbreviated notation f o r these quadrangles
and arcs: F1, F2, . .., FLq; I t , I,, . .., l A q ;
.
& ( k = 1, 2, .., r ) are all the o t h e r e l e m e n t a r y quadrangles of the partition
JJ;
(bk) ( k = 1, 2, . . ., r) are the a r c s without contact makingup the boundary
of quadrangle RR through which the paths of s y s t e m (A) e n t e r Rk (leave
Rk 1;
A i (i = 1, 2, . . .. s) a r e the v e r t i c e s of the e l e m e n t a r y quadrangles which
belong to the boundaries of the canonical neighborhoods of the s o u r c e s or
to the boundary arcs without contact through which the paths of ( A ) e n t e r
the region H ;
.
Bi ( i = I , 2, . ., s*) are the v e r t i c e s of the e l e m e n t a r y quadrangles which
belong to the boundaries of the canonical neighborhoods of the s i n k s or to
the boundary arcs withmt contact through which the paths of ( A ) 1 e a v e
the region H (all the c o r n e r points of the boundary of H are evidently
included-among the A i and B i ) .
Let (A) be a dyn_ami,: s y s t e m sufficiently c l o s e t o (A). We w i l l f i r s t
define the region i n which this s y s t e m is considered. Let C be s o m e
closed boundary c u r v e of g . Lf C is a cycle without
contact, i t w i l l be used a s the boundary c u r v e of a.
(
-
J ,/-\
G Suppose C is made up of arcs without contact
11, I2, ..., ln and arcs of paths 21, z,, ..
., z, of
s y s t e m (A), so that when the c u r v e C is t r a v e r s e d i n
%x,
* /
the positive direction the v a r i o u s arcs are encoun-
.
t e r e d in the o r d e r I t , zl, 1,. z2, . ., I,. z, ( F i g u r e 60),
t h e i r end points ( c o r n e r points) being respectively
Xi. YI, X2, Y l r . . .. X,, Y , . Let Ai, X,, . . ., l,, be s o m e
r, x3 5 fixed elongations of the a r c s I , , 12, .. ., I,. ; Through
* -
y a x the points Y,,Y z . ., Y n we p a s s the arcs ztr z2, ..., z,of
I
.-%.- / paths of s y s t e m (A) to their intersection with the

3: arcs k2, ha, ..., A,, a t the points%:, X,, ..., F,, XI,
respectively (in Figure 60 these arcs are m a r k e d by
FIGURE 60 dashed c u r y e s ) . T h e arcs without _contact TI Y I , - -
x, YL... .. X , Yn will be designated I , , I,, . .., In,
- -
respectively. Lf (A) is sufficiently c l o s e to (A), the
c u r v e c made up of the arcs &, E. &, &, ..., I,. z, is a simple closed c u r v e .
F o r e a c h boundary c u r v e Ciof H we substitute a c u r v e ci constructed in
t h i s way. The region limited by all the c u r v e s ziis 8.
L e t u s now c o n s t r u c t a partition I? of H I , analogous to the r e g u l a r
partition n of B , making e a c h element Ui, i j , F,, Rk of the partition 11 t o
* I f li is an arc without contact a i d I is the partofthis ~ r consisting

c entirely of interx points ofb. then
li is called an e I o n g a t i o n of the arc without ivntact 1 .
157
c o r r e s p o n d to a n element G I , v j , Py,,,-hh. In p a r t i c u l a r , we a s s u m e that U i

coincides with Ut ( i = 1, 2, . . ., p ) .
Each canonical neighborhood V j of a saddle point O j ( j = 1, 2, . . ., q) is made
to c o r r e s p o n d to a region v j , whose construction is described i n the s t a t e -
ment of L e m m a 6 ( s e e F i g u r e 59; the regions V j and vj i n L e m m a 6 a r e
designated y and $, respectively). F u r t h e r m o r e , we w i l l identify LE),
p,z j , J = 1.2, . . ., q, with Liy, &7), respectively. 752 @YJ) is an a r c without con-
tact entering the boundary of Tj. It is p a r t of the a r c i\$ (ic)) or of i t s
elongation (we a s s u m e that all the elongations have been chosen and fixed
beforehand).
Every e l e m e n t a r y quadrangle F , (m = 1, 2, . . .. 4) of the partition I 1 is
made up of a r c s of paths of s y s t e m (A) and is limited on one side by the
a r c l,, and on the o t h e r by an a r c without contact which belongs e i t h e r to a
l i m i t cycle r of the neighborhood of a s o u r c e o r a sink, o r to a l i m i t cycle
(or a n a r c without contact) of the region @. The quadrangle F , is made t,o
c o r r e s p o n d to the quadrangle 3, made up f r o m a r c s of paths of s y s t e m (A)
and limited on one side by the a r c I,, and on the other by a n a r c without
contact belonging to the s a m e l i m i t cycle (or the s a m e boundary cycle
without contact, or the s a m e boundary a r c without contact).
The c o r n e r points of the boundary of H a r e the c o r n e r points Y 1of the
boundary of H and the points k i d e s c r i b e d in the construction of
( F i g u r e 60). C o r n e r s e n i i p a t h s and corner-arcs of paths in a a r e deter-
mined by the region itself and the s y s t e m (A).
Consider an e l e m e n t a r y quadrangle Rk and the a r c aL (k = 1, 2, . . ., r ; s e e
above) entering i t s boundary. Let Ah and A f be the end points of the a r c a*.
L e t r(a)be the cycle without contact (a boundary cycle or a cycle belonging
to a s o u r c e ) o r the boundary a r c without contact incorporating a i r .
The point Ak) (or Aka) may be a c o r n e r point of the boundary of H , or a
point of a c o r n e r a r c or of a c o r n e r semipath of s y s t e m (A), or i t may be
a v e r t e x of one of the quadrangles F,. In each of these c a s e s , the c o r r e -
sponding point Ai1)(or zf) is naturally determined by the preceding con-
s t r u c t i o n . If Ak (or AA) is not a point of one of the above types, we may
take 2 2 ) to coincide with Ai ( &*to coincide with AhPJ, r e s p e c t i v e l y ) . The
points and 22 a r e thus well defined. The a r c ah is made to c o r r e s p o n d to
a n a r c without contact a h which is p a r t of the cycle (or the arc without
contact) r between the end points 22and 22. Finally, the e l e m e n t a r y
quadrangle Rh consisting of the arcs of paths of s y s t e m (A) and delimited
(on one s i d e ) by the a r c ak is made to c o r r e s p o n d to the elementary
quadrangle a),consisting of the a r c s of paths of s y s t e m (A) and delimited
by the a r c without contact a,.
.
The points g1 (i =2, 3, . ., s*) corresponding to the points B I , and the
a r c s without contact 8 k corresponding to the a r c s b k ( k = 1, 2, . . ., r ) a r e
determined in a n a t u r a l way using the previously constructed a r c s a h and
points Xi.
(A)
Let s y s t e m be sufficiently c l o s e to s y s t e m (A). Moreover, when
going around the cycle without contact I=)whichbelongs to a s o u r c e or to
a boundary of H (or when traveling along a boundary a r c without contact)
in a c e r t a i n direction, l e t the points At be encountered i n the o r d e r
Ai,, Ai,, ..., Ai,. F r o m L e m m a s 2 and 3 it follows that when going around the
- - . -
cycle F(a)in the s a m e direction, the points
Ail, Ai,,
lie on F(a) in the o r d e r
. ., Ai,. A s i m i l a r proposition is valid f o r the cycle r()(orthe
boundary a r c without contact) and for the points Bj and BI.
158
518. THE FUNDAMENTAL THEOREhl OF STRLrCTURAL STABILITY
It is r e a d i l y s e e n that if s y s t e m (A) is sufficiently c l o s e to (A), then:

a) '4' (they coi_ncide with Ut)are canonical neighborhoods of s o u r c e s or
s i n k s of s y s t e m (A); i f V iis a canonical neighborhood of a n equilibrium
s t a t e (a limit c y c l e ) of s y s t e m (A), d iis a canonical neighborhood of a n
equilibriu_m s t a t e ( r e s p e c t i v e l y , a l i m i t c y c l e ) of the s a m e stability of
s y s t e m (A).
b ) To-every _saddle point 01 of s y s t e m (A) c o r r e s p o n d s a sa_ddle point Gj of
.
s y s t e m ( A ) and t', ( j = 1. 2, . ., q ) is a canonical neighborhood of O j .
c ) T h e previously constructed r e g i o n s p, ( m - 1, 2. . .., 4q) and f i k (k =
= I , 2, .. ., r ) are elemenl.ar-y quadrangles of s y s t e m (A).
d ) T h e s e t of all points of the quadrangles F,,, and E k and the neighborhoods
ci and r?, coinciGes with the e n t i r e region 8.
e ) System (A) h a s no o t h e r equilibrium s t a t e s and closed paths in z,
except those which lie i n the canonical neighborhoods c; and f,.
T h i s proposition foll.ows d i r e c t l y f r o m the preceding s t a t e m e n t and f r o m
the fact that each e l e m e n t a r y quadrangle Fmand i& adjoins e i t h e r the
boundary of E or the boundary of one of the canonical neighborhoods oi o r r j .
f ) T h e canonical neighborhoods di_and fj and the e l e m e n t a r y quadrangles
F,,,and m. f o r m a r e g u l a r partition of H I , which w i l l be designa_ted fi.
g ) The canonical p a r t i t i o n s ll and H of the regions B and 8, r e s p e c t i v e l y ,
are isomorphic in the cbbvious sense."
W e w i l l now show that f o r any e > O and a sufficiently s m a l l 6 > 0 , w e have
(8,A ) 2 (3,2)
for a s y s t e m (3 which is & - c l o s e to (A).

Choose a fixed e > O . The corresponding 6 > 0 w i l l be s e l e c t e d i n s e v e r a l
steps.
1 ) W e choose a number b n > 0 s u c h that if s y s t e m (A)is an-close to
s y s t e m (A), R satisfies the above conditions (a) through (g).
2 ) For e v e r y canonical neighborhood Ui w e s e l e c t two n u m b e r s q i > 0 and
61 > 0 ( i = 1, 2, . . ., p ) i n a c c o r d a n c e with L e m m a s 5 and 7 (i.e., such that i f
s y s t e m (A)is Gi-close to s y s t e m (A) and 'piis a topological mapping of the
boundary of U,into itself, which is a n q;-translation, t h e r e e x i s t s a mapping
Ti of Ui into itself which i s a n e-translation, c o n s e r v e s paths, and coincides
with 'pi on the boundary of U t ) .
Let
du=mir:{6i, d2, .. ., aP}, qu=min (q,, q2, . .', qP).

3) Consider a n elenientary quadrangle F j and a n a r c without contact
I, (; = 1, 2, . . ., 4q) e n t e r i n g i t s boundary (note that l j e n t e r s the boundary of
a canonical neighborhood of a saddle point). Let c,and d j be the a r c s of
paths of s y s t e m (A) e n t e r i n g the boundary of the quadrangle F j . L e t
eF = m i n ( e , qu}.
By L e m m a 9, 54.2, t h e r e e x i s t 6,>0 and q j > O s u c h that i f s y s t e m (5)
is
aj-close to s y s t e m (A), and q j is a mapping of the a r c s l j , c j , dl onto the
* To be precise, if any two elen-ents El and E 2 0 f the pamtion are incidental. i.e., one of these elements
enters the boundary of its cOUnl.erpart, the corresponding elements El and of partition Rare also
incidentdl. By e I e m e n t s of t h e p a r t i t i o n II we mean the neighborhoods U i and V i , the elementary
quadrangles F, and R h , the arcs without contact and the arcs of paths entering their boundaries, and the
end points of these arcs. The previous construction established a natural one-to-one correspondence between
n.
the e k m e n t s of paKitiCms ll and
159
I Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS OF STRUCTURAL STABILITY
corresponding boundary a r c s 6, &, d; which is a l s o an qj-translation, t h e r e

e x i s t s an eF -translation T j o f the elementary quadrangle Fj onto which zj
coincides with cpj on l j , c j , d j and c o n s e r v e s paths.
Let
..., &},
I b=min(8i,8z7 qF=min{qt.q~,
4 ) Consider a n elementary quadrangle Rh and the a r c without contact ah

..., qdP}.
and a r c s of paths CJ and d j ( j =l, 2, ..., r ) entering i t s boundary. A s i n c a s e 3,

we take E R = e F = min {e, qv} and s e l e c t f o r each of the quadrangles Rh the two
n u m b e r s d j and q j (along the s a m e l i n e s a s before). Let
8~ = min {ai, 6 2 , .- ., b},q~ = min {si,qzl ... . qr}.
5) Consider the a r c without contact ah entering the boundary of the

quadrangle R h . Let AAi and be the end points of ah. Let chi and ch2 denote
the a r c s of paths of s y s t e m (A) passing through A h , and A , , , respectively,
and entering the boundary of the quadrangle Rh (k = 1, 2, . ., r ) . The a r c ah lies .
e i t h e r on a boundary cycle or a boundary a r c without contact, or on a cycle
without contact which belongs to a source.
If(z) is sufficiently c l o s e to (A), the elements of the r e g u l a r partition fi
corresponding to the quadrangle Rhr i t s s i d e s ah. c k i , c u , and the v e r t i c e s Ah,
and Ahz a r e the quadrangle a),, its s i d e s
.R. Gzr and the v e r t i c e s &,and 21,~.By
L e m m a s 7 and 8 , 54.2, w e can_select & > o
and a k > O such that if s y s t e m (A) is ak-close
to s y s t e m (A), and p ( A h i t & I ) < 6 k ,
p (&, &) <61. t h e r e e x i s t s a mapping (PA
defined on the a r c s ah, chi and CkZ which maps
these a r c s into &,&, &, respectively,
and the points Aki and Ah2 into &i and .&a,
respectively, and which i s a l s o an qR-trans-
lation ( V R is defined i n step 4 of the p r o c e -
dure; s e e Figure 61). Let
6 A =min Ki, 62, . , 6&

bA = min {ai, a2, . . ., b}.
6 ) Consider a canonical neighborhood
VJ of the saddle point 01 and the a r c 8 without
contact i ~ ) , @ ,ih), ig entering i t s boundary.
!fr Let aV>O be a number with the following
property: if s y s t e m (A) is aV-close to
s y s t e m (A), v j is a canonical neighborhood
IFIGURE 61
of the partition n
corresponding to the
neighborhood V j , and ev = min { e , q p ,V R } , then
a ) t h e r e e x i s t s a mapping T, of ~j onto V j
which c o n s e r v e s paths, is an ey-translation, and coincides with the
mapping cp described before the statement of L e m m a 6 on each of the a r c s
@),-ij$);z
* In other words, the mapping Tjtopologically maps each of the arcs ly,, 1% onto itself, leaving the end
points of these arcs fixed and moving the intersection points of the arcs with the separatrices of system (A)
into intersection points with the corresponding separatrices of system (x).
160
; l e . THE FUPiDAhlEXTAL THEOREhl OF STRUCTURAL STABILITY
b ) e v e r y a r c of a pa.th c (or d ) entering the boundary of the elementary

quadrangles p$(or F;;), pc, FiY) c a n be mapped onto the corresponding arc
--
c ( d ) by a n q*-translatior., where q* = min {qF,qR};
.,
c ) to e v e r y v e r t e x A, of the quadrangle F:?) or Fj:) ( j = 1, 2, . . q) t h e r e
c o r r e s p o n d s a v e r t e x 2. of the partition li such that
P (Am 2.1 <L4

(LA is selected in s t e p 6 of the procedure).
The existence of 6v follows f r o m L e m m a 6 (18.1),from L e m m a s 7 and 8,
$ 4 . 2 , and f r o m the r e m a r k to L e m m a 3, 59.2.
7) Consider a n e l e m e n t a r y quadrangle R k and the a r c without contact ( I k
.
(k = 1, 3 , . ., r ) entering i t s boundary. Let E , , E 2 , . . ., E,. be the end points
of the arcs a h , which are c o r n e r points or belong to c o r n e r arcs or c o r n e r
seniipaths.:: Let 8 ~ >0 be s o sniall that i f s y s t e m (A)is GE-close to s y s t e m
( A ) and 2, is a point corresponding to point E , ( s = 1,2, . . ., r*), then
p (ES, E,) < S A .
The existence of 8~ follows f r o m the method of construction of R and
from L e m m a 5, 54.1.
Let
6 = - m i n ( 8 n 3 .aU, AF3 8 R . By, BE}.
LVe will prove that if s y s t e m (6)is 6 -close to s y s t e m (A), then

e = -
( R , A ) E ( H , A). (4)
To prove t h i s propos,ition, w e a s s u m e s y s t e m (B)to be 8 - c l o s e to (A)

and construct a mapping T which r e a l i z e s the relation (4).
C o n s t r u c t i o n o f m a p p i n g T.
S t e p I. In e a c h saddle-point neighborhood ( j = 1, 2,rj .
. ., q ) w e con-
s t r u c t a mapping Tj satisfying the conditions described in s t e p 6 above,
and a s s i g n the symbol ? to T j .
S t e p 11. Ue now c c n s t r u c t a mapping T of the arcs of paths c and d
entering the boundaries of the quadrangles FY, F) onto the arcs and 2 s o c
that this mapping is a n q*-translation (see s t e p 6 above) and coincides with
the previous mapping T a t the points of the neighborhood vj.
S t e p 111. The mapping T completed on the boundary arcs I , e, and of
the quadrangles F j ( j = 1, 2, . . ., 4q) is now continued to Fj s o that i t is a n
e,-translation and c o n s e r v e s paths. T h i s is feasible because of condition 3.
S t e p Il. E v e r y point E , (s = 1, 2, . . ., r * ) is made to correspond to a point
E,, taking T ( E , ) = Es.
S t e p V . Steps I11 a:id K have defined the mapping T a t the end points
of e a c h a r c without contact a b entering the boundary of the e l e m e n t a r y
quadrangle R h . W e continue this mapping to a mapping T of the e n t i r e arc air
onto &, so that T is a n TR-translation. T h i s is feasible because of 5.
S t e p VI. Let E , be the end point of the a r c without contact a h e n t e r i n g
the boundary of the quadrangle R k (see s t e p 7), c, the arc of a path of
s y s t e m (A) passing through the point E, and entering the boundary of the
9
The end points of the arcs ah may also belong to the elementary quadrangles 4.
161
quadrangle Rk, the corresponding a r c of a path of s y s t e m (A) entering the

boundary of the quadrangle Rk. The mapping T defined a t the points E , by
s t e p N is now continued to a mapping T of c, onto Ls (s = 1, 2, . . ., r*) , so that
T i s a n qR-translation. T h i s is feasible because of 5.
S t e p VII. The mapping T h a s been defined by s t e p s I through VI on the
a r c without contact ak and on the a r c s of paths entering the boundary of each
.
e l e m e n t a r y quadrangle R k (k = 1, 2, . ., r ) ; this mapping-is a n qR-translation.
We continue i t to a mapping T of the quadrangle Ekonto Rk,which c o n s e r v e s
paths and is an ER-translation. T h i s is feasible because of 4.
S t e p VIII. The mapping T is defined by the previous s t e p s on the
boundary of each canonical neighborhood Ui (i = 1, 2, ..
., p ) and is an q v -
translation. We continue this mapping to a mapping T of the neighborhood D,
onto itself which c o n s e r v e s paths and is an e-translation. T h i s is feasible
because of 2.
The mapping T defined by s t e p s I through VI11 is evidently a n e - t r a n s l a t i o n
which maps into and c o n s e r v e s the paths. We thus have the r e l a t i o n
e -
( H , A ) E z (E?, x). (4)
This completes the proof of the fundamental theorem.
3. The fundamental t h e o r e m f o r a s p h e r e
The definition of a s t r u c t u r a l l y s t a b l e dynamic s y s t e m on a s p h e r e w a s

given i n Chapter 111 (56.2, Definition 12). It amounts to the following: a
dynamic s y s t e m ( A ) defined on a s p h e r e S is said to be struc_turally s t a b l e
i f for any e > 0 t h e r e e x i s t s 6 > 0 such that f o r any s y s t e m (A) 6 - c l o s e to (A),
(S, x):(S, A ) .
The n e c e s s a r y and sufficient conditions of s t r u c t u r a l stability of a
s y s t e m defined on a s p h e r e p r e c i s e l y coincide with the corresponding con-
ditions for a s y s t e m on a plane. We w i l l now prove the following t h e o r e m .
T h e o r e m 24. A dynamic system (A) defined on a sphere s is
structurally stable if and only if
I . Each of the equilibrium states of system (A) i s a simple node,
saddle point, m focus.
II. The closed paths of system (A) are simple limit cycles.
III. System (A)has no saddle-to-saddle separatrices.
P r o o f . To fix o u r ideas, l e t ( A ) be a n analytical s y s t e m on a s p h e r e ,
and we will prove s t r u c t u r a l stability r e l a t i v e t o t h e s p a c e Rh". S t r u c t u r a l
stability r e l a t i v e to the s p a c e Rh"(r > 1)i s proved in the s a m e way, and
the proof r e l a t i v e to the s p a c e R g ) is even s i m p l e r .
Su f f i c i e n c y. The sufficiency of conditions I through 111 f o r the
s t r u c t u r a l stability of s y s t e m (A) on a s p h e r e is proved along the s a m e lines
a s T h e o r e m 23, but on the whole the proof is somewhat s i m p l e r . The
simplification naturally d e r i v e s f r o m the fact that t h e r e a r e no boundary
a r c s of paths and boundary a r c s and c y c l e s without contact on a s p h e r e .
N e c e s sit y . We w i l l only prove the n e c e s s i t y of condition II, namely
that e v e r y closed path of a s t r u c t u r a l l y dynamic s y s t e m on a s p h e r e is a
162
$13. THE FLiXDAhlESTAL THEOREhl OF STRUCTURAL STABILITY
simple limit cycle. The absence of multiple equilibrium s t a t e s and saddle-

to-saddle s e p a r a t r i c e s for a s t r u c t u r a l l y stable s y s t e m is proved along the
s a m e lines, with s o m e simplification.
Without loss of generality, w e may consider the s y s t e m s on a s p h e r e S
in a three-dimensional s p a c e R3, defined by the equation
z2-L y 2 + z B = i . is)
As the closed covering of the s p h e r e (%see 5 3 . 2 ) we choose the covering
which c a n be r e g a r d e d as the s i m p l e s t in a c e r t a i n s e n s e : i t c o m p r i s e s
two r e g i o n s zl and 4 , where 3 , is the s e t of a l l the points of the s p h e r e S f o r
which z , , ( z < l , and c2is the s e t of all the points of the s p h e r e for which
-1<z,<z2. U e f u r t h e r a s s u m e that - - I < t , = = z O and z , < z 2 < 1 (Figure 62).
Let U,.J'i be the local coordinates in Gi (i = 1, 2); ci c o r r e s p o n d s to a region
kT in the plane ( u i , v i ) . which may be r e g a r d e d as a c i r c l e centered at the
origin.
FIGr'RE 82 FIGLIRE 6 '
The intersection of 6; and 8, is the r i n g E .

Consider a s t r u c t u r a l l y stable analytical s y s t e m (,A) defined by the s e t
of analytical equations
where i = 1, 2 , and Pi,Qi a r e functions defined in Gi (or equivalently, in Hi);

i n the intersection Bof these regions, s y s t e m (AI) is t r a n s f o r m e d into
s y s t e m (Az) i n virtue of the transformation equations ( a l s o analytical)
between the coordinates u,,u1 and u 2 , r2 (15.2).
T h e proof will be done by reductio a d absurdum. Indeed, suppose t h e r e
is a closed path L of s y s t e m (A) on the s p h e r e S which is not a s i m p l e limit
cycle (,i.e., a path with i t z e r o c h a r a c t e r i s t i c index). Again without l o s s of
'
generality, w e may take the path L to lie in c,,
outside the ring 8, and w e may
then t r e a t i t as a closed path of the s y s t e m
ClU
--
J; -
--P,(Ul. VI). %=Q1(u,, ut), (AI)
163
Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS OF STRUCTURAL STABILI IY I
defined i n the plane region at. Since the c h a r a c t e r i s t i c index of L is z e r o ,
the r e s u l t s of Chapter V show that e i t h e r
(1) the path L is a multiple l i m i t cycle of s y s t e m (A), o r
( 2 ) a l l the paths passing i n a sufficiently s m a l l neighborhood of L a r e
closed paths.
Let u s f i r s t consider c a s e (1). In v i r t u e of Remark 2 to Theorem 19
(15.2)> for any 6, > 0 i n t h i s c a s e t h e r e e x i s t s a s y s t e m (AT) bi-close to
(AI) which h a s a t l e a s t two s t r u c t u r a l l y stable l i m i t c y c l e s in any
a r b i t r a r i l y s m a l l neighborhood of L . W e will denote these limit c y c l e s by
L: and L:.
To simplify the presentation, we identify c, with H , , i.e., consider as
a c i r c l e of r a d i u s R with a c e n t e r a t the origin in the plane (ulrv i ) . Let the
ring B be made up of points of the c i r c l e c, for which the r a d i u s - v e c t o r p
s a t i s f i e s the inequality R , , < p , ( R . F u r t h e r l e t the cycle L lie inside the
c i r c l e O<p<<Rz where Hz< Rt ( F i g u r e 63), and l e t the s y s t e m (A;) have
the f o r m
Let u s construct a s y s t e m (AI) of c l a s s 1 which is sufficiently c l o s e to

(AI) and coincides with (AT) in the c i r c l e p<Rz and with the s y s t e m (AI) in
the r i n g z . W e w i l l u s e a function cp(p) with the following p r o p e r t i e s :
a ) rp ( p ) is a function of c l a s s 1 defined for a l l p, O,cp,< R ;
b) ' P ( P ) = ~ for 0 4 p g R z ,
(p(p)=O for R,,<p<R,
O,<cp(p)<.l for &<p<R1
( F i g u r e 64; t h e r e evidently e x i s t functions of any c l a s s r > l satisfying
condition b).
FIGURE 64
The right-hand s i d e s of s y s t e m (AI) - the functions Pl(ui, vi) and Q1(ul, vi) -
a r e defined by the equalities
+
pi = pi ( ~ -f pi)'P (P), 6,= Q$+ (QT-QJ 'P (P), (5)
where p = v m .
System (AI) with right-hand s i d e s defined by (5) is a s y s t e m of class 1
which coincides with s y s t e m (AI) i n the ring B and with the s y s t e m (A;) i n
the c i r c l e O,<p,<RI. Moreover, it is readily s e e n that if 6, is sufficiently
s m a l l , (AI) c a n be made a r b i t r a r i l y c l o s e to (AI).
Since (AI) and (Aa) coincide i n the ring B a s before, taken together they
constitute a c e r t a i n dynamic s y s t e m on the s p h e r e S, which w e denote by (A).
164
$18. THE FUNDAMENTAL THEOREhl OF STRUCTURAL STABILITY
In general, t h i s is a s y s t e m of class 1. We will show that it c a n be a p p r o x i -

mated with any d e g r e e of a c c u r a c y by a n analytical s y s t e m .
To this end, consider the vector field w(M) = w (sr, y, z ) (where M (I, y, z ) is
a point of the s p h e r e ) corresponding to the s y s t e m (A) i n the s p a c e R 3 .
The v e c t o r w ( M ) lies in a plane tangent to the s p h e r e a t the point MI
Let &(x, y, z), i = 1, 2. 3, be the coordinates of the vector W(I, y, 2). Consider
a cube E with i t s c e n t e r a t the origin and with f a c e s p a r a l l e l to the coordinate
planes, enclosing the sphere S . Consider a s p h e r i c a l l a y e r defined by
the inequalities
l-q<r<I+-q,
where r = V z a + y y $ + z e , and q is a positive number sufficiently s m a l l f o r the

layer to lie inside the cube E . We define a function Ff(I,y , z ) a t any point
(I. y, z ) of the l a y e r Z by the r e l a t i o n
(i = 1, 2, 3; T = Vz2 -/- yz + 2 ) . Since (A) is a dynamic s y s t e m of c l a s s 1,

it is readily s e e n that F i (z, y. z ) a r e functions of class 1 in the l a y e r f By .
U'hitney's theorem (see /ll/, V o l . I, Sec. 2601, the functions Fi (z,y,z) can be
extended o v e r the e n t i r e cube E without changing t h e i r c l a s s . Let the
functions Ft be components of the vector vi; t h i s approach yields a v e c t o r
field of c l a s s 1 defined :n the cube 2 which coincides with the field w on the
s p h e r e S. By the W e i e r s t r a s s theorem, the field w can be approximated
with any a c c u r a c y to r a n k 1 with a n analytical field w,. On the s p h e r e S , the
v e c t o r s w, are i n g e n e r a l not tangent to the s p h e r e . However, projecting
these v e c t o r s on the corresponding tangent planes to the s p h e r e , we obtain
a field or v e c t o r s tangent io the s p h e r e which define s o m e d y r a m i c
s y s t e m (A) on S. Clearly (A) is a n analytical s y s t e m which can be made a s
c l o s e as desired to (A) cind thus to the initial system. Now, since the
c y c l e s L: and L*,are s t r u c t u r a l l y stable, s y s t e m (A) h a s s t r u c t u r a l l y stable
c y c l e s L, and E , in the neighborhood of e a c h of these cycles.
We have thus established that if a n analytical s y s t e m (A) h a s a multiple
l i m i t cycle L on the s p h e r e S , t h e r e e x i s t s a n a r b i t r a r i l y c l o s e analytical
s y s t e m (A)which h a s a t l e a s t two closed paths i n any a r b i t r a r i l y s m a l l
neighborhood of the cycle L . T h i s , however, i m p l i e s that (A) is not
s t r u c t u r a l l y stable, as demonstrated i n the proof of T h e o r e m 2 0 ( S 1 5 . 3 ) .
It now r e m a i n s to c o n s i d e r c a s e (2), when all the paths passing sufficiently
c l o s e to L are closed. In t h i s c a s e , a s i n the proof of T h e o r e m 20, w e c a n
c o n s t r u c t a n a r b i t r a r i l y c l o s e s y s t e m of c l a s s 1 with L a s i t s s i m p l e limit
cycle, and then proceed to approximate to i t , a s before, with a n analytical
s y s t e m (A). System (A) will have a simple l i m i t cycle 2 a r b i t r a r i l y c l o s e
to the cycle L . If e is sufficiently s m a l l , the mapping which r e a l i z e s the
E -identity of the partition of the s p h e r e by the paths of (A) and (A) moves
a n isolated closed path E of (A) into a n o n - isolated closed path of (A) lying
n e a r L , which is impossible.
The problem is thus proved f o r the s t r u c t u r a l stability r e l a t i v e to the
s p a c e R!;'. In o t h e r c a s e s , the proof is analogous with obvious modifica-
tions.
165
Ch. VI. NECESSARY A N D SUFFICIENT CONDITIONS OF STRUCTURAL STABILITY
4. R e m a r k s and supplements
a) R e m a r k r e g a r d i n g s t r u c t u r a l l y s t a b l e s y s t e m s
inside a cycle without contact.
A s w e have already noted (6.1), s t r u c t u r a l l y stable s y s t e m s were
originally considered in a region limited by a cycle without contact 141, and
not i n any g e n e r a l region. The definition of a s t r u c t u r a l l y stable s y s t e m
can be significantly simplified in this c a s e (compared to Definition 10, 96.1).
Indeed, l e t (A) be a dynamic s y s t e m defined in G, and c* a closed sub-
region of G , limited by a cycle without contact r.
c*
F i r s t l e t u s a s s u m e that (A) is s t r u c t u r a l l y stable i n G*. Since is
evidently a region with a n o r m a l boundary, Theorem 2 3 applies and condi-
tions I through I11 a r e thus satisfied in 8*. Repeating the s a m e a r g u m e n t s
as in the proof of the sufficiency i n Theorem 23, we r e a d i l y see that the
following l e m m a holds t r u e .
L e m m a 8 . F o r any e > 0 , there exists 6 > 0 such that if system (A) is
&close to system (A), we have
(a,
A, (@*,A ) . (6)
Now suppose that L e m m a 8 is satisfied f o r a s y s t e m (A). Let f7 be a

o-extension of G* ( s e e L e m m a 1, 18.1), where u > 0 is sufficiently s m a l l .
CleaTly, 3 i n t h i s c a s e is a l s o a region limited by a cycle without contact,
and G+ c H .Then by the previous l e m m a and by L e m m a s 12 and 8 of 54.2
w e see (this conclusion is e a s y to d e m o n s t r a t e ) that f o r any e > 0
( H , A, :(R,A ) ,
provided that s y s t e m s (a) and (A) a r e sufficiently c l o s e , and t h i s i m p l i e s
that (A) is s t r u c t u r a l l y stable i n C*.
We have thus shown that i f (A) is s t r u c t u r a l l y stable i n G*, L e m m a 8
holds true, and vice v e r s a . Hence, the statement of Lemma 8 can be
adopted a s a definition of s t r u c t u r a l stability of a s y s t e m (A) in a region
limited by a cycle without contact.
A s i m i l a r r e m a r k applies to the case when the s y s t e m is considered i n
a region limited by s e v e r a l nonintersecting c y c l e s without contact. The
proof is e n t i r e l y analbgous.
In the g e n e r a l c a s e , when @ is a region with a n o r m a l boundary, the
statement of L e m m a 8 no longer provides a definition of s t r u c t u r a l
stability. Indeed, if ( 6 ) is s a t i s f g d , the boundary arcs of paths of c* should
be a r c s of paths of both (A) and (A). This, however, is not generally t r u e .
b) S t r u c t u r a l l y s t a b l e s y s t e m s o n c l o s e d s u r f a c e s .
The conditions of s t r u c t u r a l stability of dynamic s y s t e m s (of c l a s s 1) on
closed s u r f a c e s of non-zero kind, both oriented and non-oriented, w e r e
considered by M. Peixoto 171. These conditions amount to the following:
a dynamic s y s t e m (A) of class 1 defined on a s u r f a c e of the kind p > O is
s t r u c t u r a l l y stable i f and only i f
1) i t h a s a finite number of equilibrium s t a t e s , a l l of which
are s t r u c t u r a l l y stable;
2 ) i t h a s no saddle-to-saddle s e p a r a t r i c e s ;
166
$18. I H E FUSDAhlEh'TAL THEOREhl OF STRL'CTI'RiL STABILI ITY
3 ) i t h a s a finite number of closed paths, all of which a r e

simple limit cycles;
4) the D -limit (w-limit) s e t of each path is e i t h e r a n equilibrium
s t a t e or a l i m i t cycle.
The only new condition added f o r s t r u c t u r a l stability on a closed s u r f a c e
of a nonzero kind is thus condition 4. On a s p h e r e (or on plane), condition1
automatically follows f r o m conditions 1 through 3 by the PoincarC -
Bendixson t h e o r e m (QT, S4.6). On a s u r f a c e of a nonzero kind, however,
dynamic s y s t e m s may exist whose paths, say, a r e everywhere dense.
Condition 4 r u l e s out the possibility of the existence of these paths.
c ) S t r u c t u r a l l y s t a b l e s y s t e m s i n t h e s p a c e of
dynamic systems.
We have a l r e a d y noted (S6.1) that s y s t e m s which are s t r u c t u r a l l y stable
in a c e r t a i n region f o r m a n open s e t in the space of dynamic s y s t e m s . Ke
a r e now in a position to prove this statement. L e t be a bounded plane
region, the space of dynamic s y s t e m s of class 1 defined i n z , and G* a
region with a n o r m a l bcundary, c* cG.
T h e o y e rn 25. The d y n a m i c systems of class 1 defined in G which are
stable in G:r constitute an open set in R,.
stYuctliYa~~y
Proof .
Let (A) be a dynamic s y s t e m s t r u c t u r a l l y stable in c* which
py/
belongs to the space R,. We w i l l show that all the dynamic s y s t e m s which
a ' r e sufficiently c l o s e to ( A ) are a l s o s t r u c t u r a l l y
s a b l e . This will prove the theorem.
By T h e o r e m 1, 5123.1, and the r e m a r k to this
theorem, t h e r e e x i s t s a region If with a n o r m a l
boundary such that
B* c H c ~ T cG ,
/$ and s y s t e m (A) h a s no equilibrium s t a t e s and no closed

paths in p o t h e r than those which are located i n c*.
FiGCfHE 1-5 Let u be the distance f r o m E* to the boundary of FT.
Evidently, u>O (Figure 65).
Let e be a positive number, e < a / 2 . In o u r proof
of T h e o r e m 2 3 we have shown that i f 6 , > O i s sufficiently s m a l l and s y s t e m
( A ) is d,-close to s y s t e m (A), then
where 5 is s o m e region. W e will take 6,>0 so that i t s a t i s f i e s this condi-

tion.
F r o m (4)and the inequality E ( u / 2 , we have
@Cii
( s e e footnote on p. 67).
The region 3 was d e s c r i b e d i n detail in o u r procf to T h e o r e m 2 3 . In
particular, i t h a s been established th2t if s y s t e m (A) is d,-close to s y s t e m
(A), wher-e 6, is sufficiently small, (A) h a s no equilibrium s t a t e s and no closed
paths i n i? o t h e r than those which fall i n the canonical neighborhoods .Viand
.
I',
167
Finally, if (x) is B3-close to (A), where 5 is sufficiently s m a l l , the

equilibrium s t a t e s and the closed paths of (A) which a r e located in Ut and VJ
a r e s t r u c t u r a l l y stable (by R e m a r k 2 to Theorem 18, 58.2, Remark 2 to
T h e o r z m 18, 59.2, and r e m a r k to Theorem 18, -14). F r o m (4) it follows
that (A) h a s no saddle-to-saddle s e p a r a t r i c e s in 2.
We thus s e e f r o m the above that i f
8=min{dl, a2, &},

and (A) is 6 -close_ to (A), conditions I through I11 of Theorem 2 3 a r e s a t i s -
fied for system_(A)i n a region H with a n o r m a l boundary and, i n v i r t u e of
t h i s theorem, (A) is s t r u c t u r a l l y stable-in this region. But then, by ( 6 )
and L e m m a 1, 56.1, we conclude that (A) is s t r u c t u r a l l y stable inF*. T h i s
completes our proof.
An analogous t h e o r e m is evidently a l s o t r u e f o r a sphere.
T h e o r e m 25. The structurally stable dynamic systems on a sphere
f m m an open s e t in the space of dynamic systems.*
The validity of. Theorem 25 follows almost immediately f r o m Theorem 24.
Peixoto / 7 / h a s shown that Theorem 2 5 is applicable to dynamic s y s t e m s
of class 1 on any closed surface, whether oriented o r not.
W e will now show that the s t r u c t u r a l l y stable s y s t e m s f o r m a n e v e r y -
where dense s e t i n the s p a c e of dynamic s y s t e m s . W e will consider dynamic
s y s t e m s of class 1 defined i n some region 8. Let H be a simply connected
region limited by a simple closed c u r v e I, such that I? c G. F o r simplicity,
the proof will be given f o r the s p a c e of dynamic s y s t e m s defined i n f o r
which the c u r v e r is a cycle without contact. Let t h i s space be R * . Close-
n e s s i n t h i s space is defined as c l o s e n e s s to r a n k 1 inG.
T h e o r e m 26. Let
&
dt= P ( G Y), -$=Q(z, Y) (A 1
be a dynamic system in R*. F m any 6 > 0 , there exists a system (A) 8 -

close to (A) which is structurally stable in H.
P r o o f . Let 6 > 0 be fixed. W e may take 6 s o s m a l l that any s y s t e m
which is 6-close to (A) belongs to R*, i.e., r is a cycle without contact of
t h i s system.
By the W e i e r s t r a s s t h e o r e m on the approximation of continuous functions
with polynomials, t h e r e e x i s t s a s y s t e m
dz
-=Pi@,
df Y), $=Q(z, 6119 (A1)
d /5-close to (A), whose right-hand s i d e s a r e polynomials.

Let Pl and Qi be polynomials of d e g r e e m and n , respectively. In our
proof to T h e o r e m 10, 57.2, we have shown that t h e r e e x i s t i r r e d u c i b l e
polynomials a r b i t r a r i l y c l o s e to P, and Q l r respectively, which are m o r e o v e r
of the s a m e d e g r e e as P1 and Q l .
Let P a and Q2 be such polynomials, and l e t the s y s t e m
dt
-=
dt Pa (z,Y), S=92 ( ~ Y)
7 (A2
be a /5-close to (AI).
Structural stability relative to one of the spaces @, is naturally meant here.
168
518. THE FUVDAhlENTAL THEOREM OF STRC'CTCRAL ST.4BILITY
By the BCzout theorem ( / 1 2 / , Ch. 111, 3 . 1 ) , s y s t e m (Az) may only have

a finite number of equilibrium s t a t e s , which d o e s not exceed m . n . Let
0, (z,, y,), i = 1, 2, . . ., s, b e a l l the equilibrium s t a t e s of (Az) located i n
H (s,<m.n).
Suppose that s o m e of the equilibrium s t a t e s O;, e.g., Ol , is not s i m p l e ,
Le.,
For any choice of a and f3, the point Ol is a n equilibrium s t a t e of (A;),

and P ; and QZ a r e polynomials of d e g r e e s m and n , r e s p e c t i v e l y . We take
a a n d P sufficiently s m a l l , so that
(this is evidently always possible). System ( A t ) is then a r b i t r a r i l y c l o s e to

(Az), and O1 is a s i m p l e equilibrium s t a t e of (AS).
If the polynomials Pf and Q: are not i r r e d u c i b l e , we r e p l a c e them with
sufficiently c l o s e polynomials P , and Qz of the s a m e d e g r e e (i-e., m and n ,
r e s p e c t i v e l y ) which a r e i r r e d u c i b l e . W e then obtain the s y s t e m
which ?lso h a s a t m o s t ni.n equilibrJum s t a t e s .

If (Az) is sufficiently c l o s e to (A;), i t h a s a s i m p l e equilibrium s t a t e 6, i n
a sufficiently s m a l l neighborhood of 0,( s e e J2.2, R e m a r k 3 to T h e o r e m 6).
Suppose that this is indeed so. System ( A z ) is thus a r b i t r a r i l y c l o s e to (Az)
and h a s a finite number of equilibrium s t a t e s ( l e s s than n . n > , a t least one of
which, d l , is simple.
Suppose that one of the equilibrium s t a t e s of ( A z ) in H is multiple (we may
take i t as Oz). Then, just as we have p a s s e d f r o m (Az) to (Az), w e w i l l p a s s
f r o m ( A z ) to a n a r b i t r a r i l y c l o s e s y s t e m
where the right-hand s i d e s are i r r e d u c i b l e polynomials of d e g r e e s m and n

and-which h a s a s i m p l e equilibyium s_tate d2 i n the neighborhood of 4.
If (Az) is sufficiently c l o s e to (Az), (Az) a l s o h a s a s i m p l e equilibrium s t a t e
O1 in the neighborhood of 0,.
I69
Ch. VI. NECESSARY AND SUFFICIENT CONDITIONS OF S rKIIC rLlKHL S I AHILI I Y
Continuing along the s a m e lines, we obtain a f t e r a finite number of

I
s t e p s the s y s t e m
dz
-d=
l P , (2,Y), $= Q3 (2, Y), (A3
which is 6/5-close to (Az) and h a s a finite number of equilibrium s t a t e s in

H , a l l of which are s i m p l e .
If (As) h a s multiple foci o r saddle-to-saddle s e p a r a t r i c e s in H , they can
be eliminated by a suitable rotation of the v e c t o r field. Indeed, consider
the s y s t e m
1d.zt = P 3 - p Q 3 = P,(I,v), =Q3+ pPz=Q( (I,Y), (A4)
where p =+0. The vector field of this s y s t e m is obtained by turning the

vector field of (As) through the angle tan-' p.
The equilibrium s t a t e s of (A4) a r e all the equilibrium s t a t e s of (A3) and
I nothing but the equilibrium s t a t e s of (As) ( L e m m a 3, 8 3 . 2 ) . Since (As) only
h a s s i m p l e equilibrium s t a t e s i n H , (A4) will a l s o have only s i m p l e equilib-
I r i u m s t a t e s in H , provided p is sufficiently s m a l l . Suppose that (A3) h a s a
multiple focus (zo, yo). Without l o s s of generality, we may take zo = yo = 0 .
Let
I Since 0 ( 0 , O ) is a multiple focus, we have for the equilibrium s t a t e 0 of (As)
zl>0, a=a+d-0.
For the s a m e point considered as an equilibrium s t a t e of (A4), we have
u*=p(I)-c).
If 0 (0, 0) is a multiple focus of (A4), then b = e . Since a l s o d = - a , we

find A = -a2 - b 2 , < 0 , which c o n t r a d i c t s the condition A > 0. Thus, the point
0 (0, 0) cannot be a multiple focus of (A4). Clearly if p p 0 is sufficiently
s m a l l , we have A* > 0, and u* is s m a l l , Le., 0 (0, 0) is a s i m p l e focus of
(A4). Thus, we have incidentally established that a rotation of the vector
field of a dynamic system through a sufficiently small angle will reduce any
multiple focus to a simple focus.
We have shown i n Chapter N that a saddle-to-saddle s e p a r a t r i x of a
dynamic s y s t e m d i s a p p e a r s when the vector field of the s y s t e m is turned
through a sufficiently s m a l l angle ( i t "decomposes" into t w o s e p a r a t r i c e s ; s e e
the l e m m a i n 911.1 and the proof of T h e o r e m 16 i n 8 1 1 . 2 ) . Since system(A3)
may only have a finite number of equilibrium s t a t e s and s e p a r a t r i c e s , we
conclude f r o m the above that f o r a sufficiently s m a l l p f 0 the following
conditions a r e satisfied:
1) System (&) is a / 5 c l o s e to (As).
2 ) System (A4) h a s only a finite number of equilibrium s t a t e s in H, all
of which a r e s i m p l e and which do not include multiple foci (in o t h e r words,
H includes a finite number of equilibrium s t a t e s all of which a r e s t r u c t u r a l l y
stable ).
I I70
$18. THE FUE:DAhlNTAL THEOFFkl OF STRL'CTLIRAL STASII.IT1
3 ) System (A4) h a s no saddle-to-saddle s e p a r a t r i c e s in H .

Another condition fo:.lows f r o m conditions 1 through 3 combined with the
analyticity of (A4):
4) System (A4) may only have a finite number of closed paths in H .
T h e validity of condition 4 is proved along the s a m e lines as in
Theorem 2 1 (S16.1). T h e r e is only one distinctive feature in this proof
which is worth considering separately. Using the notation introduced in
the proof to T h e o r e m 2 1 , w e will. show that h', o r K" cannot be a closed
path (i.e., c a s e 5 i n the proof to T h e o r e m 2 1 is inapplicable to o u r con-
ditions). Indeed, l e t K,,, be a closed path. Then, by the definition of K , ,
this path i s a non-isolated closed path. The p r e s e n c e of a non-isolated
closed path in a n analytical systeni l e a d s to the existence of a cell which
is completely filled with closed paths (512.3). Let W' be such a cell. A s
we known ( s e e QT, 523,2), the boundary of the cell )I'should be made up of
tbvo z e r o - l i m i t continua, each of which is e i t h e r a ) a closed path, or
b ) a n equilibrium s t a t e classified a s a c e n t e r , or c ) a continuum of
saddle-to-saddle s e p a r a t r i c e s and equilibrium s t a t e s . However, in o u r
case, no such z e r o - l i m i t continua exist, since s y s t e m (A4) does not have
any c e n t e r s o r saddle-to-saddle s e p a r a t r i c e s in H , and a closed path of
a n analytical s y s t e m cannot be a z e r o - l i m i t continuum (any closed path of
a n analytical s y s t e m is e i t h e r a n isolated o r a n i n t e r i o r path in a c e l l ) . Thus,
hL cannot be a closed path. In all o t h e r r e s p e c t s , the proof of proposition 4
does not differ f r o m the proof of Theorem 2 1 .
Let the closed paths of (Ad) lying in H be L , , L n , . . ., ,Z, (they are a l l
limit cyclesj. If all these l i m i t c y c l e s are simple, (A4) is s t r u c t u r a l l y
stable (by T h e o r e m 23, 518.2) and o u r theorem is proved. Suppose now that
.
s o m e of the c y c l e s Li ( i = 1, 2. . .. r ) of (-44) a r e multiple. U'e may then
rotate the vector field znd consider the s y s t e m
In 512.3 w e defined the m u l t i p l i c i t y o f a l i m i t c y c l e and e s t a b -

lished that e v e r y limit cycle of an analytical s y s t e m h a s a definite multipli-
city. W e s h a l l now u s e s o m e r e s u l t s whose proof w i l l only be given l a t e r
on ( T h e o r e m s 60 and 61, 532.4). By these theorems, i f Li is a cycle of finite
niiltiplicity of a n analytical s y s t e m (Ah), t h e r e e x i s t e i > 0 and p: > 0 with
the following property: any s y s t e m (A,) for which I p I -== pr h a s a t most two
closed paths in U b i ( L i ) , znd t h e s e paths are simple limit cycles.
.
Let E = min { e , , E*, . . ., E ? } . LVe denote by iFi( i = 1, 2, . ., r ) a canonical
neighborhood of the c l o s e paths Li lying in U. ( L , ) and by y l and y ; the
~
c y c l e s without contact of (&) which f o r m the boundary of Vi. \ V e may take

a l l the V t to be nonintersecting neighborhoods in H . The set E\ 6Vi (i.e.,
the complement of the Lnion of the neighborhoods V iin g ) will be designated
F.
Let p*, 0 < y* < min {p:, p:, . . ., p:}, be so s m a l l that if 1 p 1 < p*, the
following conditions are satisfied:
a ) The s y s t e m (A,) is 6 1 5 c l o s e to (A4).
bj T h e s y s t e m (A,) has only a finite number of equilibrium s t a t e s i n I?,
which a r e a l l s t r u c t u r a l l y stable.
c ) T h e s y s t e m (A,) h a s no saddle-to-saddle s e p a r a t r i c e s in H .
171
d ) The cycles without contact y ; and y; (i = 1, 2, . . ., r) of (A4) a r e c y c l e s

without contact of (A,,).
e ) In e v e r y neighborhood V I , t h e r e a r e a t m o s t two closed paths
of (A,,), and these paths a r e s t r u c t u r a l l y stable l i m i t cycles.
The validity of conditions a through d f o r s m a l l p is self-evident.
Condition e follows f r o m the definition of the number e , the n u m b e r s pf , and
the neighborhoods V I .
W e will now show that i f p is sufficiently s m a l l i n absolute magnitude,
conditions a through e a r e supplemented by a n additional condition:
f ) Any closed path of (A,) lying in a is completely contained in one of
the neighborhoods V , .
To prove t h i s proposition, note that the s e t F is a closed region limited
by a finite number of c y c l e s without contact of (&), and that (A4) h a s only
a finite number of equilibrium s t a t e s i n F , a l l of which a r e s t r u c t u r a l l y
stable, and h a s no saddle-to-saddle s e p a r a t r i c e s and no closed paths in
t h i s region. Therefore (Ah) is s t r u c t u r a l l y stable i n F . But then any s y s t e m
sufficiently c l o s e to (A4) h a s the s a m e topological s t r u c t u r e in F (see
18.4a).
Hence it follows that f o r a sufficiently s m a l l p, (A,) cannot have closed
paths which a r e completely contained i n F. F u r t h e r m o r e , i f a closed path
of (A,,) h a s a point which belongs to one of the V I , the e n t i r e path belongs
to that V i , since otherwise the path would i n t e r s e c t one of the c y c l e s without
contact ri o r y ; , which is impossible. T h i s proves condition f .
We have thus shown that i f p is sufficiently s m a l l , the s y s t e m (A,)
s a t i s f i e s conditions b, c, d, e. Then t h i s s y s t e m i s s t r u c t u r a l l y stable in H by
the fundamental t h e o r e m of s t r u c t u r a l stability (Theorem 23, 5 18.2). Since
s y s t e m (A,,) is 8 -close to (A), the proof of Theorem 2 6 is complete.
Peixoto h a s shown that Theorem 2 6 is a l s o t r u e when R* is the space of
the dynamic s y s t e m s on any closed surface, whether oriented or not (see
IW.
It follows f r o m Theorem 2 6 that s t r u c t u r a l l y stable s y s t e m s f o r m an
everywhere dense s e t in the space R* of s y s t e m s of class 1.
In our proof of Theorem 26, we have actually established that any
s y s t e m i n R* h a s an a r b i t r a r i l y close a n a l y t i c a 1 s t r u c t u r a l l y stable
s y s t e m . Hence it follows that i n the space of a n a l y t i c a 1 s y s t e m s , the
s t r u c t u r a l l y stable s y s t e m s are a l s o everywhere dense.
T h e o r e m s 25 and 2 6 signify that s t r u c t u r a l l y stable s y s t e m s constitute,
so to say, the "bulk" of the space of dynamic s y s t e m s . Structurally
unstable s y s t e m s , on the other hand, constitute "partitions" partitioning
the space into regions, each filled with s t r u c t u r a l l y stable s y s t e m s of the
s a m e topological type.
d) R e m a r k r e g a r d i n g t h e c o n d i t i o n s of s t r u c t u r a l
s t a b i l i t y of a d y n a m i c s y s t e m r e l a t i v e t o t h e s p a c e s
R$) a n d Rg).
In Theorem 23, which f o r m u l a t e s the n e c e s s a r y and sufficient conditions
of s t r u c t u r a l stability of a dynamic s y s t e m , s t r u c t u r a l stability is naturally
understood in the s e n s e of Definition 10 ( S S . l ) , Le., r e l a t i v e to the space Ri
( s e e 96.3). L e t now (A) be a s y s t e m of class N > i o r a n analytical s y s t e m ,
which is considered a s a point of the space R * , where R* is one of the s p a c e s
R!$ ( r 4 . N ) o r RZI ( s e e 55.1). A s before, we consider a region c* with a
I72
$18. THE FUNDAMENTAL THE0GEb.L OF STRUCTURAL STABILITY
normal boundary, c* cG. Let (A) be s t r u c t u r a l l y stable in @ r e l a t i v e to

the s p a c e R*. Then:
(A) may only have a finite number of equilibrium s t a t e s , a l l of which
a r e simple (in v i r t u e of T h e o r e m s 10 and 11, 57.3);
(A) does not have equilibrium s t a t e s with pure imaginary c h a r a c t e r i s t i c
n u m b e r s (by v i r t u e of Remark 1 to m e o r e m 15, 510.4);
(A) does not have saddle-to-saddle s e p a r a t r i c e s (by v i r t u e of the r e m a r k
to T h e o r e m 16, 511.2);
(A) may only have closed paths which are simple l i m i t c y c l e s (by v i r t u e
of Theorem 20, 15.3).
A l l this m e a n s that conditions I through I11 of Theorem 23 a r e the
n e c e s s a r y conditions of s t r u c t u r a l stability of s y s t e m (A) i nz* relative
to the s p a c e R*.
Conversely, i f t h e s e conditions a r e satisfied, (A) is s t r u c t u r a l l y stable
r e l a t i v e to the space R i ,and t h e r e f o r e relative to the s p a c e R* ( s e e S6.3).
W e have thus established that conditions I through I11 of T h e o r e m 23 are
the n e c e s s a r y and sufficient conditions of s t r u c t u r a l stability of a s y s t e m
(A) in G* r e l a t i v e to any of the s p a c e s Rg),R; which contain (A) a s a point.
Hence, if w e a r e only dealing with the s p a c e s R$ and R;)* t h e r e is no need
to indicate explicitly that (A) is s t r u c t u r a l l y stable (or unstable) r e l a t i v e
to one of these s p a c e s .
133
C h a p t e r VII
C E L L S OF S T R U C T U R A L L Y S T A B L E SYSTEICIS. A N
ADDITION T O T H E T H E O R Y OF S T R U C T U R A L L Y
S T A B L E SYSTEMS
INTRODUCTION
The p r e s e n t chapter c o m p r i s e s t h r e e sections. The f i r s t , $19, d e a l s

with c e l l s of s t r u c t u r a l l y stable s y s t e m s . The concept of a c e l l o f a
d y n a ni i c s y s t e m w a s introduced in QT, Chapter VII, $16. If a l l the
singular elements, i . e . , boundary a r c s , equilibrium s t a t e s , limit cycles,
and orbitally unstable paths and semipaths, a r e removed f r o m the region
(or the s p h e r e ) where the dynamic s y s t e m is considered (assuming that
t h e r e is only a finite number of such elements), the remaining points of i?
f o r m an open s e t comprising a finite number of components. r h e s e c o m -
ponents are the cells of the dynamic s y s t e m . In e v e r y cell, the paths of
the s y s t e m show a n identical behavior in a c e r t a i n s e n s e . The general
p r o p e r t i e s of cells a r e investigated in QT, Chapter VII, and f o r general
dynamic s y s t e m s , t h e r e are infinitely many different types of c e l l s . How-
e v e r , dynamic s y s t e m s which are s t r u c t u r a l l y stable in c* and have a
n o r m a l boundary are c h a r a c t e r i z e d by relatively f e w cell types. An
investigation of these types is the subject of $19. It gives a complete
listing of all the different types of simply connected i n t e r i o r c e l l s of
s t r u c t u r a l l y stable s y s t e m s (i.e., cells which do not touch the boundary
of the region), and a l s o of all the different types of doubly connected c e l l s
(both i n t e r i o r and adjoining the boundary).
Some examples of s t r u c t u r a l l y stable s y s t e m s with their c e l l s are
considered in 520.
'The l a s t section, S21, h a s no relation to the subject of c e l l s . It p r o v e s
that we c a n do without the requirement of -identity of close s y s t e m s
(see J18.4,a ) in the definition of a s t r u c t u r a l l y stable s y s t e m within a
cycle without contact (or on a sphere), i.e., the following definition can be
advanced:
System (A) is said to be s t r u c t u r a l l y stable in a region enclosed by a
cycle without contact i f any sufficiently close s y s t e m h a s the s a m e
topological s t r u c t u r e like (A) in the relevant region.
The equivalence of this definition and the original Definition 1 0 (Q6.1),
incorporating the requirement of e-identity, is proved in 521, which thus
constitutes a n addition to the theory of s t r u c t u r a l l y stable s y s t e m s .
Note that the definition incorporating the condition of e-identity is
quite natural and is m o r e convenient f o r the derivation of the n e c e s s a r y
conditions of s t r u c t u r a l stability, since i t p e r m i t s r e s t r i c t i n g the a n a l y s i s
1 74
to local considerations (e.g., i n the neighborhood of a p a r t i c u l a r path),
r a t h e r than proceeding with global t r e a t m e n t . On the o t h e r hand, the
definition without e-identity i s intrinsically s i m p l e r and it d i r e c t l y shows
that s t r u c t u r a l l y stable s y s t e m s constitute a n open s e t i n the s p a c e of
dynamic s y s t e m s .
S l 9 . CELLS G F STRUCTURALLY STABLE

DYNALIIC S Y S rERIS
1. G e n e r a l considerations pertaining to cells of

dynamic s y s t e m s
The concept of a cel'. of a dynamic s y s t e m and the p r o p e r t i e s of c e l l s

a r e considered in detail in Q T , Chapter V I I . In this section, w e will
r e i t e r a t e withoui proof the e s s e n t i a l information about c e l l s that w i l l be
needed in o u r treatmen: of the c e l l s of s t r u c t u r a l l y s t a b l e s y s t e m s .
The a n a l y s i s c a n be c a r r i e d out e i t h e r on a s p h e r e or in a bounded
plane region E . In the case of a s p h e r e , we w i l l a s s u m e that the dynamic
s y s t e m ( A ) h a s a f i n i t e m n i b e r of s i n g u l a r p a t h s ( s e e QT, 816.9). If (A)
is defined in a plane region c, w e will c o n s i d e r the s y s t e m i n a subregion
G* with a n o r m a 1 b o 11n d a r y , a s s u m i n g that ( A ) h a s a finite number
of singular paths in c*. To fix ideas, let u s c o n s i d e r the c a s e of a s y s t e m
defined i n a plane region.
Let E be the s e t of a i l points which belong to the s i n g u l a r e l e m e n t s of the
systeni i n z * . ' Since W P a s s u m e a finite number of s i n g u l a r e l e m e n t s , E is
a closed s e t . I t s compiement, the s e t C*\E, is t h e r e f o r e open and c o n s i s t s
of disjoint components. These components are the cells of the dynamic
system (A).
The following g e n e r a l propositions a r e proved in QT, Chapter V I I , 4 16:
I. T h e n u m b e r of ct:lls is finite.
11. Any cell is e i t h e r s i m p l y connected or doubly connected.;:'.:
111. P a t h s belonging 1.0 a single c e l l a r e e i t h e r all whole paths, or all
positive {negative) semrpaths, or all arcs of paths.
If the c e l l c o n s i s t s of whole paths, i t s paths a r e e i t h e r a l l c l o s e d p a t h s , o r
a l l loops, or a l l nonclosed paths, i.e., a- and o - l i m i t continua without
common points.
IV. A l l t h e nonclosed whole paths i t hich belong t o t h e s a m e c e l l have the
s a m e a - l i m i t continuum and the s a m e o - l i m i t continuum.
V. If a c e l l consisting of nonclosed whole paths is doubly connected,
one of i t s boundary con.:inua is a n a - l i m i t continuum, and the o t h e r is a n
o-limit continuum of the c e l l .
' The set E comprises the points of all orbitally unstable paths and semipathr. the points of corner s e m i -
pdths ind corner arcs of paths, the points of bounJary arcs and cycles without contact and boundxy
arcs of paths, and all the equilibrium states. See 516.2.
'* We recall that a bounded region is said to be simply connected i i its boundary consists of one connected
set (the boundary continuum) ,md doubly connected if its boundxy consists of two n o n i t m w c t i n z
connected sets. In doubiy c o r n e c t e d regions, one of the two continua is an exterior boundary continuum.
and the other is an interior botmdarv continuum. The interior boundary continuum, in particular, may
comprise a single point.
I75
Ch. V11. CELLS OF STRUCTURALLY STABLE SYSTEMS. ADDITIONS TO THEORY
VI. If a point P is a boundary point of a c e l l consisting of whole paths,

all the points of a path L p through P a r e boundary points of the cell.
VII. The boundary of each cell c o n s i s t s of points which belong to
singular e l e m e n t s . If s o m e point P is p a r t of the boundary of a c e l l 2 and
belongs to a singular path L completely contained in G * , o r to a s i n g u l a r
semipath L ( )(orbitally unstable or c o r n e r ) , or to a c o r n e r a r c , the e n t i r e
path L (semipath LOor a r c 1 r e s p e c t i v e l y ) belongs to the boundary of the
c e l l 2.
If the point P is p a r t of the boundary of the c e l l 2 and belongs to a
boundary a r c I of a path and P is not a c o r n e r point of the boundary, the
e n t i r e a r c 1 a l s o belongs to the boundary of the c e l l 2.*
VIII. If a path o r a semipath L is p a r t of the boundary of the c e l l 2, all
the l i m i t points of L a l s o belong to the boundary of the c e l l 2.
We will u s e the t e r m a s i n g u 1a r a r c to designate a p a r t of a
boundary a r c without contact where all the points, except the end points,
belong to nonsinG1ar paths, and each end point is e i t h e r a c o r n e r point
o r belongs to a singular a r c or a singular semipath (a singular a r c may
coincide, in p a r t i c u l a r , with a boundary a r c without contact). A singular
a r c is called a s i n g u 1 a r a - o r o -a r c depending on whether the paths
of the s y s t e m e n t e r into G* o r leave G* through this a r c . Similarly, a
l i m i t cycle without contact with all i t s points belonging to nonsingular
paths is called a singular a - c y c l e (or o -cycle) depending on whether the
paths of the s y s t e m e n t e r into E* ( o r leave G * ) through this cycle.
IX. If a semipath (or a n a r c of a path) of c e l l 2 c r o s s e s a singular
w-arc ( a - a r c ) h , all the s e m i p a t h s ( a r c s of p a t h s ) of the cell 2 c r o s s the
s a m e a r c h a t i t s inside points and do not c r o s s any o t h e r o - a r c ( a - a r c )
except A.
If a semipath (or a n a r c of a path) of the c e l l 2 c r o s s e s a singular
w-cycle (a-cycle), all the s e m i p a t h s (or a r c s of p a t h s ) of the c e l l 2 c r o s s
this cycle.
X. A c e l l whose boundary contains a n o - a r c o r a n a - a r c is simply
connected.
XI. A cell whose boundary contains a singular o - or a-cycle is doubly
connected .
Our problem is to identify all the various types of c e l l s which are
allowed for s t r u c t u r a l l y s t a b l e s y s t e m s . Since s t r u c t u r a l l y s t a b l e s y s t e m s
have only a finite number of equilibrium s t a t e s and closed paths and have
no saddle -to-saddle s e p a r a t r i c e s (the equilibrium s t a t e s being only s i m p l e
nodes, saddle points, and foci), t h e r e a r e r e l a t i v e l y few different types of
c e l l s in s t r u c t u r a l l y stable s y s t e m s . For example, s t r u c t u r a l l y s t a b l e
s y s t e m s m a y not have c e l l s filled with closed paths or loops.
2. Doubly connected cells of s t r u c t u r a l l y

stable systems
We will now proceed with a detailed a n a l y s i s of the c e l l s of s t r u c t u r a l l y

s t a b l e s y s t e m s . In this subsection, w e shall identify all the different types
of doubly connected cells, which can be done v e r y easily.
* Ifthe point P belongs to the boundary of the cell 2 and is a corner point which belongs t o a boundary
arc of a path, this a r c will either belong entirely t o the boundary of the cell 2 or not belong altogether.
This c a n be verified by simple examples.
176
:I?. CELLS OF S ITRL'CTCMLLY STABLE SYSTEhlS
F i r s t consider c e l l s of s t r u c t u r a l l y stable s y s t e m s filled with whole

paths. F r o m propositi.on IV of S15.1 and f r o m the p r o p e r t i e s of
s t r u c t u r a l l y stable s y s t e m s w e obtain the following theorem which is
valid f o r both simply and doubly connected c e l l s .
T h e o r e m 27. If 2 is a cell of a structurally stable s y s t e m filled with
whole paths, all these paths f m t -+ i go to the same sink (Le., a stable
node, focus, m limit rycle), and for t -+ - thgy all go to the same s o w c e
(an unstable nmie, fociis, or limit cycle).
Theorem 27 evidently c a n be r e s t a t e d a s follows: any cell of a
s t r u c t u r a l l y stable s y s t e m filled with whole paths h a s one s o u r c e and one
sink.
Doubly connected cells filled with whole paths.
Theorem 2 7 and propositions IV and c' of S15.1 enable u s to establish
directly all the different types of doubly connected cells of a s t r u c t u r a l l y
stable s y s t e m which are filled with whole paths.
Indeed, the boundar,y of such a cell c o n s i s t s of a single sink and a single
s o u r c e , i.e., e i t h e r two limit cycle, or a l i m i t cycle and a n equilibrium
s t a t e , o r e l s e two equilibrium s t a t e s . The l a s t alternative, however, h a s to
to be rejected, since the boundary of a bounded cell cannot consist of two
points. The boundary of a doubly connected cell thus c o n s i s t s e i t h e r of two
limit cycle or a limit cycle and an equilibrium s t a t e . Evidently, in e i t h e r
case, the e x t e r i o r boundary continuum is n e c e s s a r i l y a l i m i t cycle,
enclosing all the paths of the c e l l . The i n t e r i o r boundary continuum may be
a limit cycle, a node, or a focus.
B e f o r e w e c a n proceed withour classification of the different types of
cells, w e should decide on a c r i t e r i o n for inclusion of two cells in one type.
Different c r i t e r i a are zvailable, and w e w i l l u s e the following definition.
D e f i n i t i o n 22. Cells Z , and Zzare said to be of the same type ( m to
belong to the same type) ** i f there exists an mientation-conserving, path-
conse,rving topological mapping T of 2, onto 3,ichich does not reverse the
direction of the paths in t . Otherrcise, ice shall say that the cells z,and z2
are of different types.
We are now in a position t o d e s c r i b e all the different types of doubly
connected c e l l s of a s t r u c t u r a l l y stable s y s t e m which are filled with whole
paths.
I. C e l l s b o u n d e d b y a l i m i t c y c l e a n d a n e q u i l i b r i u m
s t a te .
The boundary of e a c h cell in this category c o n s i s t s of a limit
cycle Lo and a n enclosed equilibrium s t a t e - a node or a focus. The
cycle L o may be e i t h e r stable o r unstable, and the i n c r e a s e in t may
correspond to motion in positive (counterclockwise) or negative s e n s e
along the cycle. Accordingly, t h e r e may be f o u r different cell types in
this category, which are shown in F i g u r e 66.
To structurally unstable systenu, this theorem in general is inapplicable. For example, in a cell
-
filled with loops, all the path$ go to the same equilibrium state for t -+ m and t -+ f m , i.e., the
source of this cell coincides with its sink.
* * The condition stated in this definition could be replaced by a requirement of the existence of a mapping
T with the relevant properties moving the cell Z, into ZI (Le., without considering the c e l l closures).
Alternatively. we could omit the requirement that T should be an orientation-consentine mapping or
conserve the direction of m o t i m in I along the paths. The choice of the particular criterion assigning
two cells to the same type is largely arbitrary. Sometimes, the actual choice depends on the problem
k i n g considered.
I77
It i s readily s e e n th2.t all the four c e l l s shown in F i g u r e 67 belong to
different types. On the o t h e r hand, e a c h c e l l in F i g u r e 68 is of the s a m e
type a s the corresponding cell i n F i g u r e 67." T h e r e i s
t h e r e f o r e a total of four types of c e l l s bounded by two
limit c y c l e s .
111. D o u b l y c o n n e c t e d c e l l s f i l l e d w i t h
s e m i p a t h s o r a r c s o f p a t h s . The boundaryof
w
a cell filled with s e m i p a t h s or a r c s of paths evidently
should c o m p r i s e l i m i t a r c s or c y c l e s without contact
{see S19.1). By proposition X in 19.1, the boundary of
a doubly connected c e l l cannot c o m p r i s e singular a r c s ,
and i: t h e r e f o r e i n c o r p o r a t e s at least one s i n g u l a r cycle
without contact. Hence it follows d i r e c t l y that the
boundary of a doubly connected c e l l filled with s e m i p a t h s
c o n s i s t s of a s i n g u l a r cycle without contact and a s o u r c e
or a sink (a l i m i t c y c l e , a node, or a focus). F r o m these considerations,
one c a n r e a d i l y find the different types of t h e s e c e l l s , and this a s s i g n m e n t
is left to the r e a d e r a s an e x e r c i s e . Doubly connected c e l l s filled with
a r c s of paths are bounded by two s i n g u l a r c y c l e s , one inside the o t h e r .
Only one type of such c e l l s is possible ( F i g u r e 69).
3 . I n t e r i o r cells of s t r u c t u r a l l y s t a b l e s y s t e m s .
Simply connected i n t e r i o r cells
A cell 2 in region G*:s s a i d to be i n t e r i o r i f zc

G*, i . e . , the boundary
of c e l l Z h a s no common points with the boundary of E*.
L e P I wi a 1 . Euery interior cell consists of whole paths.
P r o o f . If cell 2 c o n s i s t s of s e m i p a t h s , their end points evidently lie
on the boundary of the c e l l and t h e r e f o r e belong to c e r t a i n s i n g u l a r e l e m e n t s .
The only suitable singular e l e m e n t s a r e boundary a r c s or c y c l e s without
contact. The boundary of a cell consisting of s e m i p a t h s thus i n t e r s e c t s
the boundary of region GC, i . e . , this is not a n i n t e r i o r c e l l . It is s i m i l a r l y
proved that a n i n t e r i o r cell cannot c o n s i s t of a r c s of paths.
L e iii ni a 2. The botindary of an interior cell has no points which
belong to cmnev semipaths m c m n e r aycs ojpaths.
P r o o f . Let L b e a c o r n e r s e m i p a t h o r a n a r c of a path, P a point
of the semipath which belongs to the boundary of a n i n t e r i o r c e l l Z. By
proposition VII, 119.1, the end point MIof the semipath (or a r c of path)
L which is a point of the boundary of G* a l s o belongs to the boundary of
c e l l Z, i.e., 2 is notand interiorcell. This c o m p l e t e s the proof.
F r o m the definition of a n i n t e r i o r c e l l and f r o m L e m m a 1, supported
by proposition VI in Sl9.1, i t follows that the boundary of a n i n t e r i o r cell
of a s t r u c t u r a l l y s t a b l e s y s t e m c o n s i s t s of whole paths, which are l i m i t
c y c l e s , s e p a r a t r i c e s , or equilibrium s t a t e s . Doubly connected cells
were considered in the previous subsection, where w e s a w that their
boundaries contain n e i t h e r saddle points nor s e p a r a t r i c e s . W e w i l l now
prove that the r e v e r s e situation a p p l i e s to i n t e r i o r c e l l s .
The mapping T mavins cell a in Figure 67 into cell a in Figure 68, while conserving the orientation. the
paths and the sense of motion, clearly maps the exterior limit cycle of the cell into the interior limit
c i c l e , and vice v e n a . This, however, does not contradict Definition 22.
179
Ch. VII. CELLS OF STRUCTURALLY STABLE SYSTEMS. ADDITIONS T O THEORY
T h e o r e m 28. The boundary of a simply connected i n t e r i m cell of a

structurally stable system comprises a saddle point and at least two
separatrices.
P r o o f . Let Z be a n i n t e r i o r c e l l of a s t r u c t u r a l l y stable system whose
boundary contains neither saddle points n o r s e p a r a t r i c e s . By L e m m a 2,
the boundary of c e l l Z then c o n s i s t s of nodes, foci, and l i m i t cycles only,
i . e . , of sinks and s o u r c e s . Since these elements a r e disjoint in p a i r s and
a t l e a s t two of these elements - namely the a- and o - l i m i t coninua of the
paths of the c e l l Z - a r e included in the boundary of 2, Z cannot be a simply
connected cell, which contradicts the original assumption. The boundary
of Z thus must contain a saddle point o r a s e p a r a t r i x . Now, i f the boundary
of a cell filled with whole paths contains the s e p a r a t r i x of s o m e saddle
point, it a l s o contains the saddle point (519.1, VIII). On the other hand, if
the boundary of a c e l l contains a saddle point, it contains a t l e a s t two
s e p a r a t r i c e s of this saddle, which a r e continuation of each o t h e r . This
completes the proof.
Let u s now consider a r e g u l a r s y s t e m of canonical neighborhoods of a
s t r u c t u r a l l y stable s y s t e m (A) in e* ($17.1). Any a - s e p a r a t r i x ( o - s e p a r a -
triw) of a s t r u c t u r a l l y stable system entirely contained in G* invariably
c r o s s e s a cycle without contact which belongs to a sink (a s o u r c e ) .
In Chapter VI, 917.1, we introduced the concept of f r e e and non-free
cycles without contact, e l e m e n t a r y a - and o - a r c s , simple and cyclic
e l e m e n t a r y a r c s . L e t u s r e i t e r a t e s o m e of the propositions corresponding
to these concepts.
1) Any nonsingular path c r o s s e s e i t h e r p r e c i s e l y one a-cycle o r
p r e c i s e l y one a - a r c (simple or cyclic) a t a n i n t e r i o r point, or it c r o s s e s
e i t h e r one o-cycle o r one a - a r c .
2 ) Nonsingular paths c r o s s i n g an a - a r c ( o - a r c ) cannot c r o s s a f r e e
o-cycle (a-cycle), and a l l these paths c r o s s the s a m e o - a r c ( a - a r c ) . If
the a - a r c and the o - a r c a r e such that a nonsingular path c r o s s i n g one of
the a r c s inevitably c r o s s e s the o t h e r a r c , the two a r c s a r e said to be
conjugate. Similarly, f r e e a - and o-cycles which a r e c r o s s e d by the s a m e
paths a r e called conjugate f r e e c y c l e s .
3 ) Any a - s e p a r a t r i x ( o - s e p a r a t r i x ) of a saddle point of system (A)
p a s s e s e i t h e r through the common end point of tqvo simple o - a r c s ( a - a r c s )
o r through the end point of a cyclic o - a r c ( a - a r c ) .
4 ) The paths of the s a m e cell c r o s s a single p a i r of conjugate a- and
o - a r c s or a single pair of conjugate a - and o-cycles (QT, Chapter XI,
527, L e m m a 5).
W e will now prove that no two cyclic a r c s of a s t r u c t u r a l l y stable
s y s t e m a r e conjugate. Note that this proposition does not apply to
s t r u c t u r a l l y unstable s y s t e m s .
L e m m a 3. Of two conjugate elementury urcs of u structurally stable
system, at least one is a simple (not cyclic) ayc. ,
P r o o f . Let a be a cyclic a r c , M o i t s end point. To fix ideas, l e t a be
a cyclic a - a r c . The a r c a and the point Moconstitute a cycle without
contact C which e i t h e r belongs to a s o u r c e o r is a boundary cycle. Let Lo
be a path through M o . According to the definition of a cyclic a r c , Lo is
a singular path, i . e . , it is e i t h e r a s e p a r a t r i x , o r a c o r n e r semipath, or a
c o r n e r a r c of a path.
180
fl?. (;ELLS OF STRLCTUR.ALLY STABLE S Y S T E h l S
F i r s t a s s u m e that L o is a s e p a r a t r i x of s o m e saddle point 0. It i s

c l e a r l y a n w-separatrix, i . e . , it goes to O f o r t - t - UJ. Let L , and L? be
a - s e p a r a t r i c e s of the saddle O ( F i g u r e 70). Let b be the o - a r c conjugate
with a . All the paths passing through points of the a r c a , other than i t s
end point M<!, c r o s s the a r c b at i t s i n t e r i o r points as t i n c r e a s e s .
FIGURE 79 FIGURE 71
The s e p a r a t r i c e s L , and L , should p a s s through the end points of t.

Indeed, i f the s e p a r a t r i x L t d o e s not p a s s through a n end point of b , it
p a s s e s through the common end point of two a - a r c s 6, and br or through
the end point of a cyclic w-arc b J . But then paths p a s s i n g through the
points of the a r c a n e a r The point \I,, a l s o c r o s s the a r c s b,. bZ , or b3 a s t
i n c r e a s e s , i.e., they do not c r o s s the a r c b, c o n t r a r y to the assumption
that a and b are conjugatc- a r c s .
The s e p a r a t r i c e s L , and L ? thus must p a s s through t h e e n d p o i n t s of b.
Since these s e p a r a t r i c e s have no common points, the end points of b do not
coincide, i . e . , b is a s i m p l e a r c .
L e t now L o be a c o r n e r arc or a semipath. Let P be a c o r n e r point,
which is the commonenc point of the semipath {or a r c ) L o and the boundary
a r c without contact L ( F i g u r e 71). The paths through the points of the
a r c a sufficiently c l o s e t,3 the point M0,which l i e on the s a m e s i d e f r o m
L n as the a r c X , w i l l e v i d e n t l y c r o s s the a r c ;i. as t i n c r e a s e s . It is thus
readily s e e n that 1. and Q are conjugate a r c s . Since >. is a boundary a r c
without contact, i t is a s i m p l e o - a r c . This c o m p l e t e s the proof of the
lemma.
W e w i l l now consider the different types of simply connected i n t e r i o r
c e l l s of s t r u c t u r a l l y s t a b l e s y s t e m s .
Let Z be such a cell. By Theorem 28, the boundary of Z c o m p r i s e s
a saddle point and i t s s e p a r a t r i x . P a t h s of the c e l l 2 c l e a r l y do not
i n t e r s e c t the free cycle (- which belongs to a s o u r c e or a sink. Indeed,
i f the paths of Z c r o s s this cycle, all the points of the cycle belong to 2.
T h e r e f o r e , boundary points of the cell, and hence the boundary continua,
l i e both inside and outside the cycle L. These boundary continua should
not have any common points, i.e., the c e l l 2 is not simply connected,
c o n t r a r y to the original assumption.
181
182
Since -%I; and J1.i are two different points, b is a s i m p l e a r c . Let Ct, be a
cycle without contact c o m p r i s i n g the arc b. Cb belongs to the sink to which
the s e p a r a t r i c e s L; and L; go for t + 7 0 0 .
H e r e again w e shou;d consider two a l t e r n a t i v e s :
-461) The c y c l e s without contact C, and C b do not enclose one another.
X l 2 ) The cycle C,: l i e s inside the cycle Cb.
Consider case -411 f i r s t .
A l l ) T h e c y c l e s w i t h o u t c o n t a c t C, a n d C b d o n o t e n c l o s e
o n e a n o t h e r . Let y be a s i m p l e closed cur1.e consisting of the a r c s a
and b, the s e g m e n t s .%f2g2and M , O , of the s e p a r a t r i c e s L? and L , .
respectively, the s e g m e n t s 02M; and O J I ; of the s e p a r a t r i c e s L; and L ; , and
the equilibrium s t a t e s 17,and 02. Let A be the s i m p l y connected region
bounded by the c u r v e y T h e points of t h e c u r v e p which a r e the i n t e r i o r points
of the a r c s a and b belong to the c e l l 2, and a l l the o t h e r points of y a r e
boundary points of 2. All the paths of the cell cross the arc u and e m e r g e
froni the cycle C, with i n c r e a s i n g t . r h e s e paths should e i t h e r e n t e r the
region 3 or leave i t . M'e w i l l show that they n e c e s s a r i l y e n t e r this region.
FIGURE 74. a i S! unstable node, SIstable node: b) S unstable

node; c ) S unstable node: d) S s t j b l e node; e) S stable node.
In cases t.g, h , i . L , is an unstable cycle.
183
Ch. VII. CELLS OF STRUCTURALLY STABLE SYSTEMS. ADD1 IYONS TO IHEOKY
Indeed, let the paths of the c e l l 2 which c r o s s the a r c a leave the region
A as t increases (Figure 73). Then t h e r e a r e points of the cell 2 which lie
outside the c u r v e y , and t h e r e f o r e a l s o boundary points of Z with this
p r o p e r t y . Let E be the s e t of these boundary points (which l i e outside y ) .
The s e t E c o n s i s t s of the points of singular paths. It is readily s e e n that
the regions enclosed by the l i m i t c y c l e s C, and Cb lie inside the c u r v e .
T h e r e f o r e , the s e p a r a t r i c e s L t , L;, Lz, and L; have no points which lie
outside y , i.e., E is the s e t of points of singular paths, other than the
s e p a r a t r i c e s L,, L;, L z , L;and c l e a r l y o t h e r than the equilibrium states Oiand
0 2 . But a l l the points which a r e sufficiently c l o s e to the curve y f r o m the
outside belong to nonsingular paths. The s e t E is t h e r e f o r e a t a positive
distance outside the c u r v e y . This m e a n s that E does not i n t e r s e c t with
the boundary continuum which contains the s e p a r a t r i c e s L,, L z , L;, L; and
t h e i r limit points, i . e . , the c e l l 2 h a s a t l e a s t two boundary continua. This
evidently contradicts the assumption that Z is simply connected.
We have thus established that in the c a s e A Lall ~ the paths of the c e l l Z
c r o s s i n g the a r c a leave the cycle C, and e n t e r into the region A as t
i n c r e a s e s ( F i g u r e 72). The cell will be of one of the types shown in F i g u r e 74
according as the s o u r c e o r respectively the sink to which the paths of the
cell go is a n equilibrium s t a t e or a l i m i t cycle, and in the l a t t e r c a s e ,
according as the direction of motion along the l i m i t cycle is positive
(clockwise) o r negative. In c a s e A11 the boundary of the cell 2 evidently
c o n s i s t s of the s e p a r a t r i c e s L , , L;,L z . and L; and their a - and o-limit
without contact Cb c l e a r l y belongs to s o m e s t a b l e l i m i t cycle L o , and Cb lies

inside Lo ( F i g u r e 75). Considering, a s before, a simple closed c u r v e y and
the region A enclosed by this c u r v e , we r e a d i l y s e e that A is p a r t of the r i n g
region enclosed between the c y c l e s C . and C b . The paths of the cell Z c r o s s
the a r c a and with increasing t e n t e r into A . As t f u r t h e r i n c r e a s e s , these
paths c r o s s the a r c b, e n t e r into a canonical neighborhood of the cycle Lo
and go to L o . The s e p a r a t r i c e s L; and L i , with i n c r e a s i n g t , a l s o go to L o .
The l i m i t cycle Lo is t h e r e f o r e p a r t of the boundary of the cell 2 . A s before,
this boundary c o n s i s t s of the s e p a r a t r i c e s L , , L 2 , L ; , and L; and their a -
and w-limit points. The cell 2 belongs to one of the types depicted in
Figure 76. The exact type depends on the c h a r a c t e r of the s o u r c e to which
C, belongs and on the direction of motion along this s o u r c e (if i t is a c y c l e )
o r along the cycle L o .
FIGURE 75
184
a b C
e i
FIGURE 7 6
Let u s now c o n s i d e r case Az.
8
A ? ) r h e s a d d l e p o i n t 0, c o i n c i d e s w i t h 0 , ( F i g u r e s 7 7 , 7 5 , 8 0 ) .
In this c a s e , the s e p a r a t r i x L ; , which is an w-continuation of the s e p a r a t r i x
in the positive direction, is a l s o a n o-continuation of the
s e p a r a t r i x L , in the negative d i r e c t i o n and e n t e r s the
boundary of the c e l l 2 . L e t b be a n a r c without contact
conjugate with the a r c a , and C h a c y c l e without contact
comprising the a r c 6 . f i e s e p a r a t r i x L ; p a s s e s through
the end point -11; of the arc b. I? is r e a d i l y s e e n that p a t h s
c r o s s i n : ? the a r c a p a s s through the points of the c y c l e Cb
sufficiently n e a r ?he point -\I;, on e i t h e r s i d e of this point.
T h e r e f o r e b is a cyclic a r c .
Let ;m be a s i m p l e c l o s e d c u r v e consisting of the
c, segmen1.s J f I O l and .\I@,of t h e s e p a r a t r i c e s L , and L ? ,
respectively, of the a r c a , and the point 0,; l e t 1 be the
FIGL'KE '77 region Enclosed by the c u r v e y . T h e r e a r e two
possibilities:
A z l ) I ' h e p a t h s of t h e c e l l Z c r o s s t h e a r c a
a n d e n t e r i n t o _I a s t i n c r e a s e s ( F i g u r e 77). In t h i s case, the
s e p a r a t r i x Ll', and a l s o the c y c l e without contact C+, lie inside the c y c l e y,
and the sink to which CS 'Delongs l i e s inside Cb. The c y c l e s C, and Cb d o
not e n c l o s e one a n o t h e r . The boundary of the c e l l Z c o n s i s t s of the
s e p a r a t r i c e s L , . L ? . L ; and t h e i r a - and 6)-limit points. The cell Z belongs
to one of the types shown in F i g u r e 78. A s in the p r e v i o u s cases, the
exact type of the c e l l depends on the c h a r a c t e r of the s o u r c e s and s i n k s to
*xhich the p a t h s of the c e l l go and on the d i r e c t i o n of motion along the
s o u r c e (or the sink), aseuming that i t is a l i m i t c y c l e .
A22) T h e p a t h s o f t h e c e l l Z c r o s s t h e a r c a a n d l e a v e
1 a s t i n c r e a s e s ( F i g u r e s 75 and 80). In this case, e i t h e r the c y c l e s
(.,'? and Cb do not enclose one another ( F i g u r e 79) or cycle C, lies inside
('o ( F i g u r e 80).
Ch. VII. CELLS OF STRUCTURALLY STABLE SYSTEMS. AIIDITIONS T O 'THEORY
g h i
FIGURE 78
FIGURE 79 FIGURE 80
a s e Ail, we
However, reasoning along the s a m e l i n e s a s f o r c-~ -
- can..show
that the c a s e depicted in Figure 79 is inapplicable. 'Thus, cycle c. l i e s
. _ . ..
inside c b . Hence i t follows that the s o u r c e to which Cb belongs is a limit
186
c y c l e L o , and Cb lies inside L o . The boundary of the cell Z c o n s i s t s of the
s e p a r a t r i c e s L l , L 2 . L ; and t h e i r l i m i t points. The cell Z may belong to one
of t h e types shown in F i g u r e 81
a b
C d
e f
FIGURE 81
We have fully covered case A, when the. s o u r c e belonging to the c y c l e C,

l i e s inside C,.
B ) T h e s o u r c e w h i c h b e l o n g s t o c y c l e C, l i e s o u t s i d e C,.
An unstable l i m i t c y c l e L o , with C,J lying inside L o , is evidently a s o u r c e
of t h i s kind. A s i? case A, the s e p a r a t r i c e s L1 and L n through the end
points MI and , I f J of the arc Q may
B l ) go t o different sa.ddle points Ot and O,(Figure 82);
B2) go to the s a m e saddle point O1 ( F i g u r e 83).
FIGURE $ 2 FIG L'RE 8 3
187
Ch. VII. CELLS OF STRUCTURALLY STABLE SYSTEMS. ADDlTIONS TO THEORY
It is readily s e e n that c a s e B1 is analogous to c a s e Al2, differing f r o m

the l a t t e r only in the direction of a r r o w s along the paths, and c a s e E2 is
analogous to c a s e Az2.'" Correspondingly, t h e r e is no need to analyze
these c a s e s in p a r t i c u l a r detail. The types of c e l l s corresponding to
c a s e Bz a r e shown in F i g u r e 84. These c e l l s a r e analogous to those
shown in Figure 81.
a b
C d
e f
FIGURE 84
We have so far a s s u m e d that the a r c a is a simple a r c , and i t s conjugate

a r c b is e i t h e - simple or cyclic. The cases when b is a s i m p l e a r c and i t s
conjugate a r c a is cyclic a r e obtained f r o m A2 and B2 by r e v e r s i n g the
direction of the a r r o w s along the paths.
We can now count the total number of different - in the s e n s e of
Definition 22 - simply connected i n t e r i o r c e l l s which may e x i s t i n
s t r u c t u r a l l y stable s y s t e m s . We w i l l not do this, however. It should
only be s t r e s s e d that the above a n a l y s i s does not show that each of the
different cell types described actually e x i s t s f o r one of the s t r u c t u r a l l y
s t a b l e s y s t e m s . We have only established that s t r u c t u r a l l y s t a b l e s y s t e m s
* In case A**, the source lies inside the sink, and i n case B, the sink lies inside the source.
188
can have no other type; of simply connected i n t e r i o r c e l l s in addition to
those described. It c a n be proved without difficulty that a l l the c e l l types
described do exist, ho,,vever. This can be done, s a y , by constructing a
dynamic s y s t e m corresponding to each cell type in t e r n i s of its vector
field.
Me have established all the different types of doubly connected c e l l s
I,F19.2), and all the types of i n t e r i o r simply connected c e l l s which may
occur in s t r u c t u r a l l y stable s y s t e m s . It r e m a i n s to consider simply
connected cells which are not internal cells, i . e . , c e l l s whose boundaries
have points in common with the boundary of the region. W e w i l l not
analyze these cells, as t h i s c a n be done along the s a m e l i n e s as before.
F i g u r e 85 depicts the different types of simply connected c e l l s of a
s t r u c t u r a l l y stable s y s t e m which touch the boundary of c * f o r the c a s e
when this boundary c o n s i s t s of a single cycle without contact.
FIGURE 85
The concept of the r e g i o n o f s t a b i l i t y i n t h e l a r g e of a given

sink is of considerable importance in various applied problems. This
region is defined as the s e t of a l l the cells for which the given singular
element is the sink. Examples of r e g i o n s of stability in the large will be
found in the next section.
189
Ch. VII. CELLS OF STRUCTURALLY STABLE SYSTEMS. ADDITIONS TO THEORY
$20. EXAMPLES O F STRUCTURALLY STABLE

SYSTEMS
In t h i s section w e w i l l consider some examples of s t r u c t u r a l l y stable

systems.
E x a m p 1e 4 . Consider the system
z = ax +b y - x (39 +y2) = P (5, y),

Y + ~ Y - U +)'Y
(2' = Q (z,Y),
assuming that
and
D = a'-413. = (a-d)' + 4bc > O (U= a +d).
Since f o r s y s t e m (1)
then together with any solution x = cp ( t ) , y = $ ( t ) of ( l ) , x = - 'p ( t ) , y = - 9 ( t )

is a l s o a solution of (1). Geometrically t h i s signifies that t h e r e f 1 e c -
t i o n of a n y p a t h of s y s t e m ( 1 ) a t t h e o r i g i n i s a l s o a p a t h
or, alternatively, t h e p h a s e p o r t r a i t o f t h e d y n a m i c s y s t e m
(1) i s s y m m e t r i c a l a b o u t t h e o r i g i n .
Let u s consider the system (1) inside the c i r c l e
with the boundary
xr+ya=RS.
The path
of s y s t e m (1) is tangent to the c i r c l e ( 6 ) a t a point (x, y) when

. .
F r o m equations (1). we have for sufficiently l a r g e R
190
S 20. E IALIPLES OF STRL'CTLrRALLY ST.48LE SYSTEhlS
The l a s t relation sh3ws that f o r sufficiently l a r g e R, c i r c l e ( 6 ) is a

cycle without contact f o r paths of s y s t e m (1). Moreover, since
* .
l d
xx+yy=--(x*+f*),
2 dt
i t follows from (8) that, as f i n c r e a s e s , Z i 'y diminishes, i.e., a l l the

paths c r o s s i n g the c i r c l e ( 6 ) e n t e r into the cycle without contact ( 6 ) a s t
i n c r e a s e s . W e will take R to be so l a r g e that the l a s t condition is satisfied.
It follows f r o m ( 2 ) , as is readily s e e n , that the paths through the points
of the a x i s x = 0 c r o s s f r o m one s i d e of the a x i s to the o t h e r s i d e (with the
exception of the path which i s the equilibrium s t a t e 0 ( O m ) .
Let P (2,y) # 0. W e then change over f r o m s y s t e m ( 1 ) to a single
equation
Let u s d e t e r m i n e f o r what k the s t r a i g h t line

y=kX (10)
or p a r t of this line is an i n t e g r a l c u r v e of equation (9). F r o m ( 9 ) and (10)
w e have
=c + d k - - k r * ( i + k * )
u+bk -s* ( 1 +k * )
or
Hence
F r o m ( 2 ) and ( 4 ) it follows that t h e r e e x i s t p r e c i s e l y two values of k

satisfying equality (12). Let these two r o o t s be k, and k2, where k, c o r r e -
sponds to the plus i n (12) and kz to the minus.
Let u s now find the s t a t e s of equilibrium. One of these is the
point 0 (0, 0). T h e r e a r e obviously no o t h e r equilibrium s t a t e s on the
v e r t i c a l axis. I'he coordinates of any equilibrium s t a t e o t h e r than the
point 0 (0,0)therefore have the f o r m (xo,h0). Inserting these coordinates
in the equations P (x,y) == 0, Q (x, y) = 0, w e find that k s a t i s f i e s
equation ( l l ) , i.e., e v e r y equilibrium s t a t e of s y s t e m (l), o t h e r than the
point 0, l i e s on one of the s t r a i g h t l i n e s y = kiz and y = k g , where k, and
kz are defined by (12).
F r o m the equation P (xo, kxo) = 0 we now find that
so that
yo=kxo= i k m
191
The c b a r a c t e r i s t i c equation of the equilibrium state 0 (0, 0) is
A2 -n;l + A = 0. (15)
It follows f r o m ( 3 ) that 0 (0, 0) is a simple saddle point. Let A, and h2 be

the r o o t s of equation (15). D i r e c t calculations show that
a + bk, =Al, a + bk, = &, (16)
and, by (3), hi > 0 , hz ( 0 . F r o m ( 1 3 ) and ( 1 6 ) i t follows that the s t r a i g h t
line y = k2x h a s no equilibrium s t a t e s (except the point 0) and t h e r e f o r e the
r a y s y = k2x ( x > 0) and y = k g ( x -= 0)are
paths of the s y s t e m , i.e., s e p a r a t r i c e s of
the saddle point 0. The s t r a i g h t line
y = klxr on the o t h e r hand, contains p r e -
c i s e l y two equilibrium s t a t e s in addition to
the origin, 0,( x o , yo) and 0,(-xo, -yo), where
xoand yo a r e obtained f r o m (13) and (14) for
k = k,. The equilibrium s t a t e s 0, 01,and O2
partition the line y = k,x into four p a r t s ,
each of which is a path of the s y s t e m . The
s e g m e n t s 001and 002of the line y = kix are
s e p a r a t r i c e s of the saddle point 0.
F a i r l y s i m p l e calculations show that
A (XO. YO) = A (-- ~ o y -YO) = 2 L 1/B > 0.
FIGURE 86 T h e r e f o r e the points 0, and 0,a r e s i m p l e
nodes (these cannot be foci, since they are
l i m i t points of paths which a r e segments of the line y = kls).
System (1) h a s no closed paths. Indeed, a closed path should enclose
at l e a s t one of the equilibrium s t a t e s 0,Oi,0,and i t should t h e r e f o r e c r o s s
the line y = k,x. which is unfeasible, since this line is made up of a number
of paths of the s y s t e m . System (1) thus h a s no closed paths and no s a d d l e -
to-saddle s e p a r a t r i c e s ; it h a s t h r e e equilibrium s t a t e s - a s i m p l e saddle
and two simple nodes. T h e r e f o r e by Theorem 23, s y s t e m (1) is s t r u c t u r a l l y
stable in the c i r c l e (5). The configuration of the paths is shown in
F i g u r e 86. The direction of the a r r o w s along the paths is chosen so that
all the paths e n t e r into the c i r c l e (5) a s t i n c r e a s e s .
The c i r c l e (5) contains two c e l l s of s y s t e m ( l ) , 2, and 2,. Both c e l l s
a r e filled with s e m i p a t h s and are simply connected. The boundary of the
c e l l 2, (2,) is made up of the s e p a r a t r i c e s O A , , O A 2 , and 00, (00,),
equilibrium s t a t e s 0 and 0, (02),and a s i m p l e a r c without contact
A l B , A 2 ( A l B , A 2 ) . The conjugate a r c s of A , B , A , and A,B2A, a r e c l e a r l y cyclic.
The c e l l 2, (&)is the region of stability in the l a r g e of the s t a b l e node
01 ( 0 2 ) .
E x a m p 1 e 5. Consider the s y s t e m ( 1 ) as before
;=ux+~~-z(z'+~'), 6/=c~+dy--y(x'+y'),
a s s u m i n g that
u = a + d >0,
D = @ - 4 A >0.
192
A s in the p r e v i o u s example, c i r c l e ( 6 ) of a sufficiently l a r g e r a d i u s R is
a c y c l e without contact and all the p a t h s of o u r s y s t e m e n t e r into it as t
increases.
Consider the s y s t e 71 inside circle 15 J of sufficiently l a r g e r a d i u s . The
l i n e s y = k , and ~ y = k : ~ where
, k , and k 2 are the r o o t s of (11). are made up
of paths of our s y s t e m .
T h e c h a r a c t e r i s t i c roots 2., and 2.? of t h e equilibrium s t a t e 0 (0. 0)a r e
positive and different in v i r t u e of conditions (171, (181, (19j, so that 0 is a n
unstable node. F r o m (131, ( 1 4 j ,
( 1 6 ) and from the fact that both i.,
and i.? are positive we conclude that
e a c h of the s t r a i g h t l i n e s y = k , ~ ,
y : k,r contains, b e s i d e s the point
0, two o t h e r equilibrium states.
Let these equilibrium s t a t e s be
-4,. B: and A ? . B,, r e s p e c t i v e l y
\ F i g u r e 87J. Calculations show
that for the equilibrium s t a t e s -4,
and B, lying on the s t r a i g h t line
y = k,x, the d e t e r m i n a n t 3 =
= 2i., JTD> 0 , and for the equilib-
r i u m s t a t e s A: and LIZ, A =
- -%?
- 1 < 0 . Thus, d, and B,a r e
s i m p l e nodes, and A ? and B, a r e
s i m p l e s a d d l e points. The s y s t e m
h a s no closed p a t h s for the s a m e
FIGrrRE t.7
r e a s o n s a s before, and i t s p h a s e
p o r t r a i t is s y n i a e t r i c a l r e l a t i v e
to the origin 0. The configuration of the p a t h s i s shown in F i g u r e 87. The
d i r e c t i o n of the a r r o w s along the paths and the d i r e c t i o n of the s e p a r a t r i c e s
a r e determined so a s to e n s u r e that all the paths e n t e r into the boundary
c y c l e without contact as t i n c r e a s e s .
By Iheorem 2 3 , the s y s t e m is s t r u c t u r a l l y s t a b l e i n s i d e the circle (5).
T h i s circle c o n s i s t s O F four cells: Z,.Z 2 and the two c e l l s Z;, 2; in
s y n i m e t r i c a l position :elative to the o r i g i n . Zi is a n i n t e r i o r c e l l . It
c o n s i s t s of whole path.; extending f r o m the unstable node 0 to the s t a b l e
node :I,, and belongs t o the s a m e type as cell a in F i g u r e 74. Cell Z? c o n -
s i s t s of semipaths and belongs to the s a m e type a s c e l l Q in F i g u r e 8 5 . The
stability r e g i o n of the node -4, c o m p r i s e s the c e l l s Z, and Z ? , and that of
the node B, the cells Z ; and 2;.
R e m a r k . System (1) m a y be investigated by analogous methods for
any o t h e r combination of the coefficients a, 6 , c, d . See QT, J30,
Example 18.
E x a m p l e 6 . The s y s t e m
h a s two equilibrium s t a t e s , A (0, - 1 1) and B ( 0 , - i-JrT). H e r e , -4 is

a n unstable f o c u s and 13 is a saddle point. It is r e a d i l y verified that
193
Ch. VII. CELLS OF STRUCTURALLY STABLE SYSTEMS. ADDITIONS TO THEORY I
the c i r c l e
xa+y*---l=O (21)
is a path of the s y s t e m ( s e e QT, $30, Example 14). The focus A lies inside
this c i r c l e , and the saddle point B outside the c i r c l e .
W e will now show that the c i r c l e ( 2 1 ) is a s t r u c t u r a l l y stable limit cycle
of s y s t e m (20). To prove this, we have to evaluate the integral
is a solution of the s y s t e m corresponding to path (21), and 7 is the period

ofthefunctionscp and $ (z>O). By (34), 813
where Lo designates the c i r c l e (21). The l a s t i n t e g r a l is readily evaluated

by e x p r e s s i n g the c i r c l e Lo in p a r a m e t r i c f o r m x = cos t , g = sin t and i n -
s e r t i n g f o r P and Q t h e i r e x p r e s s i o n s f r o m
sin t + 2 cos t
,Infinity (20). As a r e s u l t , we f i n d J = '
2-cost dt=
4
=Tn: (3 -2v3) (0. Hence i t f c h o w s that
c i r c l e ( 2 1 ) is a stable s t r u c t u r a l l y stable
l i m i t cycle f o r o u r s y s t e m .
The topological s t r u c t u r e of s y s t e m (20)
in a plane was investigated in QT, 530,
Example 14.': It was established t h e r e that
s y s t e m ( 2 0 ) h a s no other closed paths,
except the c i r c l e (21), and h a s no saddle-to-
saddle s e p a r a t r i c e s . The configuration of
the paths of s y s t e m (20) is shown in
Figure 88.
FIGURE 88
Let c* be a region with a normal boundary
bounded by the a r c s without contact EF and
RS and the a r c s of paths F R and E S (Figure 88),
which encloses the equilibrium s t a t e s A and B of the s y s t e m and i t s limit
cycle (21). By Theorem 23, s y s t e m (20) is s t r u c t u r a l l y stable i n G*. c*
contains four c e l l s . 2, is a n i n t e r n a l doubly connected c e l l consisting of
whole paths which unwind f r o m the focus A and wind onto the l i m i t cycle L o .
The o t h e r t h r e e c e l l s a r e simply connected. The c e l l Zz c o n s i s t s of
semipaths and is bounded by a n a r c without contact E,F,, t h r e e s e p a r a t r i c e s
* This investigation analyzes the behavior of the paths of system (20) at infinity and considers the paths of an
auxiliary system
d+
-=2y+sa+ys-i,
dt
*=
dt
--2r,
+ -
which have a general integral (ze ya l)eY = C . The closed paths of the auxiliary system form a
topographic system of curves for system (20).
I94
f 21. DEFI?.ITLCh' OF STRIICWIUL STABILITY WITHOL'T E -1DEKTTITY
of the saddle point B, limit cycle L o , and saddle point B . The o t h e r

two c e l l s z3and 2; consist of arcs of paths. The conjugate of the arc
without contact E p , entering the boundary of the c e l l Z2 is c l e a r l y a
cyclic a r c without contact. The stability region of the cycle Lo c o m p r i s e s
the c e l l s Z1 and 2,.
5 2 1 . A DEFJBITION O F STRUCTURAL STABILITY

FOREGOING THE REQUIREILIENT OF &-IDENTITY
b e will consider in :his section dynamic s y s t e m s s t r u c t u r a l l y stable

in I T w h i c h a r e b o u n d e d b y a c y c l e w i t h o u t c o n t a c t F. We have
s e e n in Chapter V I (S18.4,a) that s t r u c t u r a l stability of s y s t e m (A) in such
a region c a n be defined as follows:
I. System ( A ) is s t r u c t u r a l l y stable i n region ts'bounded by a cycle
without czntact r if f o r any E > 0 t h e r e e x i s t s 6 > 0 such that f o r any
s y s t e m ( A ) 6-close t o (A) the following relation is satisfied:
- , e -
(ti-, A) = (it-, A ) .
Definition I not only r e q u i r e s that s y s t e m s sufficiently c l o s e to the
s t r u c t u r a l l y stable s y s t e m (A) have the s a m e topological s t r u c t u r e a s (A)
in E, but a l s o i m p o s e s a f u r t h e r condition, namely that the paths of any
such s y s t e m c a n be moved into the paths of s y s t e m (A) by a n a r b i t r a r i l y
s m a l l translation. All :hese f a c t o r s combine into a f a i r l y complex
definition of s t r u c t u r a l instability of a s y s t e m . And yet, according to the
most natural and straightforward definition, a s y s t e m is s t r u c t u r a l l y
unstable if i t s topologica: s t r u c t u r e can be a l t e r e d by infinitesimally s m a l l
i n c r e m e n t s . U e thus a r r i v e a t a b e t t e r definition of a s t r u c t u r a l l y stable
system:
11. System (A) is s t r u c t u r a l l y stable in. region bounded by a cycle
without contact r if t h e r e e x i s t s 6 > 0 such that any dynamic s y s t e m (A)
6-close to ( A ) h a s the s a m e topological s t r u c t u r e a s (A) i n F?.'
Definition I1 is s i m p l e r than Definition I, s i n c e i t i m p o s e s fewer
r e s t r i c t i o n s on s t r u c t u r i l l y stable s y s t e m s . It is clear that if s y s t e m ( A )
is s t r u c t u r a l l y stable in the s e n s e of I i t is a l s o s t r u c t u r a l l y stable in the
s e n s e of 11. T h e r e v e r s e is not immediately obvious: i t would s e e m that
there might be dynamic s y s t e m s which would be s t r u c t u r a l l y s t a b l e in the
s e n s e of I1 and s t r u c t u r d l y unstable in the s e n s e of I. In fact, however,
this is not so, and the two definitions are equivalent. T h i s fact w a s proved
by Peixoto in /14/, and the p r e s e n t section is devoted to Peixoto's proof.
To fix ideas, we w i l l consider s t r u c t u r a l stability i n relation to the s p a c e
R 2'. i.e., the s p a c e of analytical unctions with the distance defined in
t e r m s of the f i r s t derivatives only.
In case of s t r u c t u r a l stability r e l a t i v e t o the spaces RF),R'$, t h e
proof is completely analogous (or even s i m p l e r ) .
W e should specify in Definitions I and I1 in relation t o what space, RC) or R,;', the structural stability is
being considered. However, by virtue of remark d in 918.4, this is immaterial for Definition I. W e will
i c e in what foilows that it i s immaterial for Definition II. either.
195
Conditions I, 11, 111 of Theorem 2 3 (118.2) w i l l be designated CSS

(conditions of s t r u c t u r a l stability). These conditions postulate
1 ) existence of only a finite number of equilibrium s t a t e s of s y s t e m (A)
in region W , these equilibrium s t a t e s being simple nodes, saddle points,
or foci;
2 ) absence of saddle -to-saddle s e p a r a t r i c e s ;
3 ) absence of closed paths with c h a r a c t e r i s t i c index equal to z e r o .
To prove the equivalence of Definitions I and 11, we w i l l establish that
the CSS follow f r o m Definition 11. We will r e q u i r e a number of l e m m a s
in the p r o c e s s .
In L e m m a s 1 throught 8 it is assumed that s y s t e m (A) w i t h t h e
r i g h t - h a n d s i d e s P(x, y)and Q(x, y ) i s a n a n a l y t i c a l d y n a m i c
s y s t e m s t r u c t u r a l l y s t a b l e i n w i n t h e s e n s e of
D e f i n i t i o n 11, and 6 is the number mentioned in t h i s definition.
Analytical dynamic s y s t e m s 6-close to s y s t e m (A) will be called
f e a s i b 1e s y s t e m s . By Definition 11, feasible s y s t e m s have the s a m e
topologicai s t r u c t u r e in as the s y s t e m (A). W e w i l l a s s u m e that all the
paths of s y s t e m (A) which c r o s s the cycle without contact r e n t e r into W as
t increases.
L e m m a 1 . Structurally stable systems in the sense of Definition 11
form an open set in the space of all (analytical) dynamic systems.
The validity of L e m m a 1 follows directly f r o m Definition 11, since by
t h i s definition feasible s y s t e m s a r e s t r u c t u r a l l y stable.
L e m m a 2. System (A) has only a finite number of equilibrium states,
which a r e all simple.
P r o o f . By W e i e r s t r a s s s theorem, t h e r e e x i s t s a feasible s y s t e m (A)
whose right-hand s i d e s a r e i r r e d u c i b l e polynomials ( s e e proof ,Of
Theorem 10, 17.2). F r o m the Bgzout theorem i t follows that (A) only h a s
a finite number of equilibrium s t a t e s . But then (A) may a l s o have only a
finite number of equilibrium s t a t e s . The f i r s t proposition of the l e m m a
is thus proved.
Let u s now prove that a l l the equilibrium s t a t e s a r e s i m p l e . Suppose
that one of the equilibrium s t a t e s of (A) is multiple. Let this be the
point 0 (0, 0), i.e., P (0, 0) = Q (0, 0) = 0 and A (0, 0) = 0.
Let cp ( x , g) be a polynomial which is equal to 1 a t the point 0 (0, 0) and
vanishes f o r a l l the other equilibrium s t a t e s of (A). Consider the s y s t e m
where a and b a r e r e a l n u m b e r s s m a l l e r than 1 in absolute value, and e > 0 is

so s m a l l that s y s t e m (1) is f e a s i b l e . All the equilibrium s t a t e s of (A) a r e
c l e a r l y a t the s a m e time equilibrium s t a t e s of system (1). Since (1) is a
feasible s y s t e m , the number of i t s equilibrium s t a t e s in @ is the s a m e a s
that of (A). This means that the two s y s t e m s (A) and ( 1 ) have the s a m e
equilibrium s t a t e s in w.
Let C be a circle centered a t 0 (0, 0) which l i e s in W so that a l l the
equilibrium s t a t e s of s y s t e m (A), except the point 0, fall outside this
Similar propositions for structurally stable systems in the sense of Definition I were advanced in 918.4.c,
Theorems 25 and 26. Their proof, however, far from following directly from Definition I, conversely
requires a preliminary derivation of the conditions of structural stability.
196
c i r c l e . I 4 e choose E > 0 to be so s m a l l that f o r any two s y s t e m s of the
f o r m \ l ) , the v e c t o r s of the corresponding fields do not point in opposite
d i r e c t i o n s at any point of the c i r c l e C. Then by QT, S10.2, L e m m a 2,
the rotations of the v e c t o r fields of all t h e s e s y s t e m s on the c i r c l e C' a r e
equal to one a n o t h e r . Since C e n c l o s e s a single equilibrium s t a t e 0 of (A),
and hence of any s y s t e m ( I ) , the P o i n c a r e index of the equilibrium s t a t e o i s
constant for all the s y s t e m s (1).
On the o t h e r hand, i t follows f r o m equations (1) that
An a p p r o p r i a t e choice o f the n u m b e r s a and 6 will make 3 (0.(3) e i t h e r

positive or negative. Ey QT, S11.4, Theorem 30, in the f o r m e r c a s e the
P o i n c a r e index of the equilibrium s t a t e 0 is + 1, and in the l a t t e r c a s e i t is
-1, T h i s , however, c o n t r a d i c t s the previous r e s u l t . The proof is thus
complete.
L e )rz H I a 3. None qf the equilibritrw states of system ( A ) i s a centev.
Proof . is the divergence of the vector field ( P , Q ) , designated
f*
'7 Y
div ( P , 0). W e w i l l u s e this notation in the following.

Let U i ( a i . bi). i = 1, 2, j . ., n , be a l l the equilibrium states of s y s t e m (A).
Consider the l i n e a r functions
T,W ~ ) = m ( z - - n , ) t ~ ( y - - b ~ )
( j =1.2,. .., n), where the n u m b e r s m and p are chosen so that
div(qj, q j ) = m + p j ; O
and f o r j # k , q j ( a h . bk) # ( I . Let (1.y) = T , Q . . . q,,.
C l e a r l y , q(5, y ) = O a t any of the points O i . Now f r o m the equality
.. cp). d o e s not vanish at

and f r o m the previous conditions i t follows that div (cp,
any of the equilibrium states of s y s t e m (A).
d5.
dr =P ( X , y) $. E? (I,y) = P*, 2- = Q (5,y) 4- E'P (I,y) = Q*, A"' )
where ~ f - 0is s o s m a l l that s y s t e m (A*) is feasible. The set of the

equilibrium s t a t e s of (R') evidently coincides with the s e t of the equ libr iu m
s t a t e s of (A), i.e., i t corisists of the points U j . Since
div ( P * , Q*)= div ( P , Q)+ e div (9, (p),
f o r sufficiently s m a l l E , 'div ( P * , Q*)#O a t those of the points O j where

div ( P , Q) # 0 . At those of the points 0,where div ( P , Q) = 0, we have
div ( P ,Q*) = E div (cp, (p) =#=0.
197
Thus, div ( P * , Q*) # 0 for all 0,. i . e . , none of the equilibrium s t a t e s of

s y s t e m (A::-)is a c e n t e r . Therefore, s y s t e m (A) h a s no c e n t e r s e i t h e r ,
and the proof of the l e m m a is complete.
R e m a r k . The above proof r e m a i n s in f o r c e for s y s t e m s of c l a s s 1,
too. The fact that s t r u c t u r a l l y stable s y s t e m s of c l a s s 1 cannot have
center-foci follows directly f r o m the existence of feasible analytical
systems.
L e m m a 4. System (A)has no saddle-to-saddle separatrices.
P r o o f . Letq,O(q(n,besosmall t h a t f o r a n y h E 2 , where lis the
segment O<h,<1, the s y s t e m
$ = P cos (A?) -Q sin (nq),

(A A )
-$=Psin (ill) + Q cos (~zq)
is feasible. The vector field of the s y s t e m (A,) is c l e a r l y obtained f r o m the
v e c t o r f i e l d of (A) by rotation through the angle bq i n the
The equilibrium s t a t e s of (A,) and (A) coincide. Finally,
(A,) and (A) are equal a t e v e r y point, as is readily s e e n . Therefore the
saddle points of (A,) and (A) coincide.
Let these saddle points be 0,( i = I
= i , 2, . . ., m ) ; the s e p a r a t r i c e s of
each saddle point 0,will be denoted
Lit, L i z , Lis, Lib
On each s e p a r a t r i x L i f ( j = 1, 2, 3,4;
i = i, 2, -. ., m) we choose a point C t J
f which is sufficiently c l o s e to the c o r r e -
sponding equilibrium s t a t e 0,, s o that
no two s e g m e n t s OiCiJof the s e p a r a -
t r i c e s i n t e r s e c t . Let U i , be the
neighborhoods of these s e g m e n t s
satisfying the following conditions:
each neighborhood U i l contains only one
equilibrium s t a t e of s y s t e m (A),
namely Oi, and only one of the points C ,
FIGURE 89 namely C i f (Figure 89). If the points
Cit are sufficiently c l o s e to the
respective o,, such neighborhoods c l e a r l y e x i s t .
Through each point Cii, w e now p a s s a segment without contact I i J (e.g., a
segment of a n o r m a l to the path 1 , ~ )which is e n t i r e l y contained inUliand is
so s m a l l that the segments lil have no common points with one another and
each segment l i j h a s only one common point with a l l the segments Olei, of
the s e p a r a t r i c e s , namely the point C i J . E s y s t e m (A,) is sufficiently close
to (A), e.g., i f q is sufficiently s m a l l , we conclude f r o m S9.2, r e m a r k to
L e m m a 3, that to e v e r y segment OiCir of the s e p a r a t r i x L,, of the saddle
point oi c o r r e s p o n d s a segment OiC(,:)ofthe s e p a r a t r i x L$)of the saddle
point 0,of s y s t e m (A,) which is entirely contained i n U i , and is such that
C$ E Zij. Furthermore, as in the c a s e of s y s t e m (A), e v e r y segment ZiJ h a s
a single common point with all the segments 0,Cfj)of the s e p a r a t r i c e s of the
s y s t e m (A,). We will a s s u m e that q is s o s m a l l that t h i s condition is
satisfied for all (A,), I E I .
198
In view of the above, e v e r y neighborhood U i j thus neturally c o r r e s p o n d s
to a single s e p a r a t r i x of any s y s t e m (Ai), O.<b < i , namely the s e p a r a t r i x
L::! L V e will refer to tl-,is s e p a r a t r i x a s t h e s e p a r a t r i x c o r r e -
s p o n d i n g t o t h e n e i g h b o r h o o d Uij.
Let now s y s t e m ( A ) have a saddle-to-saddle s e p a r a t r i x y . To fix
ideas, l e t y extend f r o m saddle point 0,to O?. The case of a s e p a r a t r i x
extending f r o m s o m e saddle point to the s a m e saddle point, i.e., forming
a loop, is treated analogously.
Let Tt be a homomorphism of G- into itself, which moves the paths of ( A )
into the paths of (A,) (this homomorphism e x i s t s , s i n c e ( A ) is a s t r u c t u r a l l y
stable s y s t e m and ( A , ) is a feasible system; see Definition n). Evidently,
T , ( y ) i s a s e p a r a t r i x e x k n d i n g f r o m the saddle point T I (0,)to the saddle
point T k (/I2). Being a s e p a r a t r i x of two saddle points, i t c l e a r l y c o r r e -
sponds to two (different) neighborhoods U i i .
T h e r e are uncountably many s e p a r a t r i c e s T k (7) (since 1. E I ) , w h e r e a s
the p a i r s of neighborhoods U i i f o r m a countable set. Consequently, t h e r e
e x i s t s a n u n c o u n t a b 1 e s e t I* c I of the values of A and a l s o two
neighborhoods U i j , e.g., the neighborhoods ci33 and U16of the saddle points
O3 and Oh. such that i f 2. E I * , the s e p a r a t r i x T I ( y ) extends f r o m saddle point
O 3 to saddle point 0, and c o r r e s p o n d s t o the neighborhoods CIJ3 and U6,.
Since I* is a n uncountable s e t of points of the segment I = [O, 11, P c a n -
<
<
not consist entirely of isolated points. Hence t h e r e e x i s t s a t l e a s t one
E, E I * and a sequence of X i . i = 1, 2. 3, . . . *
such that rZi E I* and lim X i = A,. Without
t*.m
l o s s of generality, we m a y take l i to be
&- :WYJ&L4
a monotonically d e c r e a s i n g sequence.
The s e p a r a t r i x T b (y)> and l i k e w i s e
the s e p a r a t r i c e s T A i (y), extend f r o m
saddle point O3 to saddle point O6 and
correspond to the neighborhoods U J 3and
FIGURE 30 U44(Figure 90). Let 'c' be a neighborhood
of the s e p a r a t r i x T b (y). F r o m 89.2,
r e m a r k to L e m m a 3, and f r o m 8 4 . 2 , L e m m a 7 i t follows that f o r all
sufficiently l a r g e n, T I , ty) c V . But the vector field of (Aio) is obtained f r o m
the vector field of (A) by rotation through a positive angle. In o u r proof
of Theorem 16 (511.2) w e established that if t h e r e e x i s t s a saddle-to-
saddle s e p a r a t r i x of s y s t e m ( A ) and a sufficiently s m a l l neighborhood of the
s e p a r a t r i x , the s y s t e m generated f r o m ( A ) by a rotation of the vector field
through a sufficiently sniall angle cannot have a saddle -to-saddle s e p a r a t r i x
in that neighborhood.:'. We have thus reached a cor*tradiction, which
p r o v e s the l e m m a .
R e m a r k . The pr0c.f of the l a s t l e m m a actually shows that i f a dynamic
s y s t e m ( A ) with a finite number of simple saddle points h a s saddle-to-
saddle s e p a r a t r i c e s , the s y s t e m obtained by rotating the v e c t o r field of (A)
through a sufficiently s m a l l angle no longer h a s any such s e p a r a t r i c e s .
In our proof of Theorem 16 w e did not consider systems (Ai), but systems of the form
-
To move from ( A ) to ( A ) , w e t a v e to rotate the vector field through an angle a. tan a = b, and
v m .
stretch the field vectors by a factor
199
L e m m a 5 . Every closedpath L of system (A)i s isolated, Le., i t i s a

limit cycle.
P r o o f . Since (A) is an analytical s y s t e m , a closed path L e i t h e r i s a
l i m i t cycle or is enclosed in an open ring E entirely consisting of closed
paths ( s e e Chapter V, J12.3). W e w i l l show that no such ring can e x i s t .
Indeed, suppose that ring E does e x i s t . By QT, 523.2, the i n t e r i o r
boundary of this ring, which is a z e r o - l i m i t continuum, is e i t h e r a closed
path L o , o r c o n s i s t s of a finite number of saddle-to-saddle s e p a r a t r i c e s ,
or finally is a c e n t e r . The l a s t two a l t e r n a t i v e s a r e ruled out by
L e m m a s 3 and 4. The f i r s t alternative is unacceptable because all the
paths which extend outside Lo and sufficiently c l o s e to it a r e closed. But
all the paths inside Lo and sufficiently
close to it a r e a l s o closed, i . e . , Lo con-
s i s t s of the i n t e r i o r points of the r i n g E,
which contradicts the original assumption.
L e m m a 5 is thus proved.
R e m a r k . The proposition of the
l e m m a is c l e a r l y a l s o valid f o r s y s t e m s
of c l a s s 1.
L e m m a 6. System (A) may only have
a finite number of closed paths.
P r o o f of L e m m a 6 is entirely
analogous to the proof of Theorem 2 1
($16.1).
L e m m a 7. System (A) has no multiple
FIGURE 91 limit cycles.
P r o o f . By L e m m a 6 , s y s t e m (A) may
only have a finite number of l i m i t cycles. Let Li,Lz,. . ., L, be a l l the
c y c l e s of the s y s t e m . Suppose that one of t h e s e , L1 say, is a multiple
cycle.
Let U be a sufficiently s m a l l neighborhood of Ll which does not i n t e r s e c t
with the paths Lz,. . ., L, ( F i g u r e 91). Using the theorem of the creation of
a closed path f r o m a multiple limit cycle (515.2, Theorem 19) and applying
the s a m e construction as i n our proof of L e m m a 2, 515.2, w e obtain a
ti
s y s t e m (AI) of c l a s s 1 which is 7 - c l o s e to (A), coincides with (A) outside
the neighborhood U (L), and h a s i n this neighborhood a t l e a s t two closed
paths. Let L; and L; be two such paths (in U).
6
B y L e m m a 2, 515.2, t h e r e e x i s t s a s y s t e m (Az) of c l a s s 1 , g - c l o s e to
s y s t e m (AI), for which the c u r v e s L;, L;, L z , Ll, . . ., Lp a r e s t r u c t u r a l l y
stable l i m i t cycles.
Finally, l e t (A3) be a n analytical s y s t e m providing a n adequate
approximation to (Az). By Theorem 18 and the r e m a r k to t h i s theorem
(the t h e o r e m of the s t r u c t u r a l stability of a l i m i t cycle, 514), s y s t e m (&)
h a s one l i m i t cycle in the neighborhood of each of the c u r v e s L;, L;,
Lz, . . ., L,, i.e., i t h a s a t l e a s t p + 1 l i m i t c y c l e s .
6
If (A3) is a l s o 5 - c l o s e
to (Az), it is 6-close to (A), i.e., it i s a feasible s y s t e m . But then,
because of the s t r u c t u r a l stability of (A), (As) should have the s a m e number
of l i m i t cycles as (A) does, i.e., p . The assumption of a multiple cycle
200
4 "1. L>EFI?;I rlCtI OF STKCiCTC'RAL STAGILITY WlTtiOUT e - IDEXTITY
among the l i m i t c y c l e s of ( A ) thus h a s led to a contradiction. This c o m -

pletes the proof of the l e m m a .
L e t n ) n a 8. Systetu ( A ) has no multiple f o c i .
P r 0 0 f . Suppose that one of the equilibrium s t a t e s of (A), 0 , ( u t , b,) say,
is a multiple focus (we a s s u m e that i t is a s t a b 1e focus). At the point 0 , ,
the .Jacobian A > 0, and div ( P . Q )= 0 . Let (x', ) be the s y s t e m introduced
in the proof to Lemma 3 . U'e may take E = E' to be s o s m a l l and of such a
sign that for the corresponding s y s t e m (A- ) - w e c a n denote i t by A ( e ' ) -
the point 0, is a n u n s t a b l e focus. Simple standard reasoning (see,
e . g . , the proof of Theorem 14 about the c r e a t i o n of a closed path f r o m a
multiple focus, 10.3)shows that i n this c a s e , f o r s o m e E * between 0 and E',
the s y s t e m A' ( E " ) w i l l have a closed path contained e n t i r e l y in a neighborhood
of 0,. By L e m m a s 6 and 7, s y s t e m ( A ) h a s a finite number of closed
paths, which are simple l i m i t c y c l e s . Seeing that a transformation to a
sufficiently close s y s t e m produces only a slight translation of e v e r y simple
limit cycle,: we conclude immediately that the s y s t e m A ( E @ ) h a s a t least one
closed path m o r e than system (A} does. This, howeirer, contradicts the
condition that A:'(")is a feasible s y s t e m . The l e m m a is thus proved.
R e m a r k . The proof of this l e m m a is a l s o greatly simplified i f w e
consider s y s t e m s of c l a s s 1, and not analytical s y s t e m s . In this c a s e ,
ae c a n easily find a feasible s y s t e m whose right-hand s i d e s in the neighbor-
hood of a multiple focu:; 0, a r e l i n e a r p a r t s of P a n d Q. 0,is evidently a
c e n t e r of this feasible .system, which is again impossible.
T h e o r e tti 29. De,FitiitionsI and 11 of stnickwal stability of a dynattiic
sgsteiti in a regioti boiiilded b y a cycle witJtoiit contact aye equivalent.
Proof . Lf s y s t e m (-4) is s t r u c t u r a l l y stable in the s e n s e of Definition I,
i t i s s t r u c t u r a l l y stable in the s e n s e of Definition !I. This is obvious. Lf
s y s t e m (A)is structura.lly stable i n the s e n s e of Definition 11, L e m m a s 2
through 7 show that i t s a t i s f i e s the conditions of s t r u c t u r a l stability (CSS).
Finally, if s y s t e m (A) s a t i s f i e s the CSS, i t is s t r u c t u r a l l y stable in the
s e n s e of Definition I by T h e o r e m 23, $18.2. W e thus see that I-tII+CSS+I.
This m e a n s that Definit.ions I and I1 and CSS a r e all equivalent. Q. E . D.
Theorem 29 is a l s o valid for dynamic s y s t e m s on a s p h e r e . However,
in the case of a s p h e r e , we w i l l consider only s y s t e m s of class 1. Let the
ii. in Definitions I and I1 be identified with a s p h e r e S2. 'CVe thus obtain two
definitions of s t r u c t u r a l stability on a s p h e r e , where Definition I coincides
with Definition 12 (6.2), and Definition I1 is free f r o m the requirement
of t' -identity .
T h e o y e )ti 30. Dejinitiors I atld II of stnichwal stability of a dynamic
system of class 1 on a sphere S2 are eqiiivaleut.
P r o o f of Theorem 30 is conducted along the s a m e l i n e s a s the proof
of Theorem 2 9 . Our proofs of L e m m a s 1 through 8 r e m a i n in f o r c e f o r
dynamic s y s t e m s of c l a s s 1 on a sphere: the only difference i s that s o m e
additional arguments, itnalogous to those adopted in the proof of
t'heorem 24 (18.3), have to be used when dealing with a s p h e r e .
r h e o r e m 30 can be generalized without difficulty to s y s t e m s of class
r > 1. The case of analytical s y s t e m s on a s p h e r e is m o r e complicated,
and i t w i l l not be considered h e r e .
R e m a r k . Theorem 30 is valid for dynamic s y s t e m s of c l a s s 1 on any
closed s u r f a c e , whether oriented or unoriented, and not only on a s p h e r e
.-
I ( '.
T h i s fdlloas from the structural stability of a simple limit cycle (see 4 14, Theorem 18).
20 1
C h a p t e r VIIl
BIFURCATIONS O F DYNAMIC S Y S T E M S .
DECOMPOSITION OF A M U L T I P L E
EQUILIBRIUM S T A T E I N T O S T R U C T U R A L L Y
S T A B L E EQUILIBRIUM S T A T E S
INTRODUCTION
The f i r s t seven c h a p t e r s of the book dealt with the theory of s t r u c t u r a l l y

stable s y s t e m s . Chapter VI11 begins the second p a r t of the book, devoted
to c e r t a i n a s p e c t s of the so-called t h e o r y of b i f u r c a t i o n s of
dynamic s y s t e m s . The p r e s e n t chapter contains two sections, 522 and
$23. In 122, the main problems of the theory of bifurcations are f o r m u -
lated and a link is established between the f i r s t and the second p a r t of the
volume. In particular, the concepts of b i f u r c a t i o n s and d e g r e e of
s t r u c t u r a l i n s t a b i l i t y of a dynamic s y s t e m are defined. Since 522
is "narrative" (without any l e m m a s , t h e o r e m s , and proofs), w e will not
s u m m a r i z e i t s contents h e r e . It should be noted, however, that the theory
of bifurcations is concerned with the changes which occur in the
topological s t r u c t u r e of a dynamic s y s t e m i n a particular region when the
s y s t e m itself (i.e., i t s right-hand s i d e s ) is a l t e r e d , and the t e r m
b i f u r c a t i o n generally refers to these changes in topological s t r u c t u r e .
Bifurcations of a multiple isolated equilibrium state (i.e., bifurcations
of the dynamic s y s t e m in the neighborhood of such a n equilibrium s t a t e )
a r e the subject of 523. The discussion is confined to a n a l y t i c a l
s y s t e m s , and only the s i m p 1e s t multiple equilibrium state is con-
sidered, i . e . , such that the series expansions of the functions P and Q in
i t s neighborhood contain a t l e a s t one l i n e a r t e r m . F u r t h e r m o r e , the topic
of bifurcations of these equilibrium s t a t e s does not r e c e i v e a fully general
treatment in 523, as we only investigate t h e n u m b e r and t h e
c h a r a c t e r of t h e s t r u c t u r a l l y s t a b l e e q u i l i b r i u m s t a t e s
into which the multiple state decomposes on passing to c l o s e systems:'.
The topological s t r u c t u r e of these equilibrium s t a t e s is treated in detail
in QT, Chapter IX (521 and 122). It is established in QT that if for a n
equilibrium state 0 (0, 0)
(J = (0, 0)+Q; (0, 0) # 0,
the point 0 is a topological node, o r a topological saddle point, o r a
saddle-node. If, however, u = 0 , s i x different possibilities a r i s e : a
The problem of bifurcations of a multiple equilibrium state in its general form is formulated as follows:
establish the changes in the topological structure of a dynamic system in the neighborhood of an
equilibrium state on passing to close systems. A relatively narrow segment of this general problem is
considered in 123.
202
topological node, a topological saddle point, a focus o r a c e n t e r , a n
euuilibriutu s t a t e with an elliptical region, a degenerate equilibrium s t a t e ,
a saddle-node. In 523, a relationship is established between the
topological s t r u c t u r e of a multiple equilibrium s t a t e , on the one hand, and
the n u m b e r and c h a r a c t e r of the s t r u c t u r a l l y s t a b l e equilibrium states into
which the multiple s t a t ? decomposes when p a s s i n g to c l o s e s y s t e m s , on
the o t h e r . For example, i f u : P ; (0,0) - (I (0.
,; 0)F 0 and the multiple
ecluilibriuni s t a t e 0 is a topological saddle point, it may only decompose
into a n odd number of s t r u c t u r a l l y stable nodes and saddle points, the
number of s t r u c t u r a l l y stable nodes being of necessity one l e s s than the
number of s t r u c t u r a l l y stable saddle points. The o t h e r r e s u l t s of the
c h a p t e r a r e contained in T h e o r e m s 35, 37 through 39. Note, however,
that f o r o = 0 , the type of the multiple sinqular point is e n t i r e l y determined
by the number of s t r u c : u r a l l y stable equilibrium states into which it
decomposes and by the;r topological s t r u c t u r e . tf, on the o t h e r hand, 0 -= 0 ,
the difference (other than topological) between nodes and foci h a s to be
taken into consideration in s o m e c a s e s , and i n other cases it is altogether
impossible to e s t a b l i s h the c h a r a c t e r of the multiple equilibrium s t a t e
using the component s t r u c t u r a l l y s t a b l e equilibrium states.
A r e a d e r wishing to speed up h i s p r o g r e s s through the book may omit
the proof of T h e o r e m s 37- 35, and f a m i l i a r i z e himself with the s t a t e m e n t
of the t h e o r e m s only.
in conclusion note that $23.3 contains a t h e o r e m by PoincarC
( rheorem 36) wttich stztes that i f a dynamic s y s t e m h a s only s i m p l e
equilibrium s t a t e s and the isocline P ( 5 , y) = 0 ( o r Q (z, y) = 0 )h a s no singular
points (i.e., points a t which P ; = P!, = 0 ) , then saddle points a l t e r n a t e with
nodes and foci along th1.s isocline. T h i s t h e o r e m i s used in the proof of
r h e o r e m 3 8 , but it is a l s o of considerable independent i n t e r e s t .
S 2 2 . T H E D E G R E E OF S IRUC IURAL I N S rABILI IY AND

BIFURCATIOPZS OF DYKARIIC SYSTERIS
The previous c h a p t e r s dealt with s t r u c t u r a 11y s t a b 1 e dynamic

s y s t e m s on a s p h e r e or i n a plane region. r h e definition of a s t r u c t u r a l l y
stable s y s t e m in a plane region ff w a s formulated a s s u m i n g that tt7 is a n
a r b i t r a r y b o u n d e d region (S6.1, Definition IO). A c e r t a i n a d d i t i o n a l
r e s t r i c t i o n w a s imposed at a later stage on t i - , e . g . , in the derivation of
the n e c e s s a r y and sufficient conditions of s t r u c t u r a l stability, and It w a s
treated a s a region with a n o r m a l boundary (516.2, Definition 1 9 ) . This
r e s t r i c t i o n i s not fundamental, although it s i m p l i f i e s s o m e proofs.
r h e definition of s t r i c t u r a l stability in a n a r b i t r a r y plane region it
(,including a region with a n o r m a l boundary) h a s one distinct disadvantage:
together with W , we a r , i f o r c e d to c o n s i d e r o t h e r r e g i o n s c l o s e to it.. To
avoid the difficulties (again not of fundamental n a t u r e ) a s s o c i a t e d with this
approach and to achieve a m o r e p l a s t i c d e s c r i p t i o n of the concepts that
follow, we will a s s u m e i n this section that the boundary r of ti is a c y c l e
without c o n t a c t
:
* 7h< simplert 3nd most complet? picture is obtair,ed for + n a m l c systems on a sphere. where the dcfinition
of rtructural rtabillty is marked.y simpler. W e again wish to emphasize that the restriction imposed on the
relevant reeion is solely intendcd to simplif? the presentation. The concepts of bifurcition and degree of
structural instability can b e defined analoyously for dynamic systcms i n any bounded plane region.
203
ch.VIII. BIFURCATIONS OF DYNAMIC S Y S T E M S
rhus, let
dz dy
= P (2,Y), = Q (G Y) (A)
be a dynamic s y s t e m defined in and considered in a closed region

bounded by a cycle without contact J? (pc G). The exact s p a c e of dynamic
s y s t e m s in which our a n a l y s i s is c a r r i e d out should be indicated. Let this
be one of the s p a c e s R$) ( l < r < N ) or R?) ( r > 1 ) , defined f o r the region E,"
which we will denote by R*. System (A) c l e a r l y should belong to the space
R*. In the following, s t r u c t u r a l stability (or s t r u c t u r a l instability) of
s y s t e m (A) in w i l l be understood as s t r u c t u r a l stability (or s t r u c t u r a l
instability) r e l a t i v e to the s p a c e R*.
The s t r u c t u r a l instability of a dynamic s y s t e m may be considered using
the concept of e - i d e n t i t y o f p a r t i t i o n s (J6.1,Definition 10 and
J18.4, r e m a r k a). In Chapter VI1 ($21, Theorem 2 0 ) we have seen, how-
e v e r , that the e-identity can be dropped when dealing with a region bounded
by a cycle without contact. Indeed, s y s t e m (A) i s s t r u c t u r a l l y
s t a b l e i n ~ e g i o nw i f a l l s u f f i c i e n t l y c l o s e d y n a m i c
s y s t e m s (A) h a v e t h e s a m e t o p o l o g i c a l s t r u c t u r e a s (A)
in r. In other words, if (A) is a s t r u c t u r a l l y stable s y s t e m in w, all the
points of a c e r t a i n neighborhood Ub ((A) I R*)of ( A ) i n the space R* a r e
s y s t e m s with the s a m e topological s t r u c t u r e in a?.
Conver_sely, if (A) is s t r u c t u r a l l y unstable in W> t h e r e always e x i s t
s y s t e m s (A) a r b i t r a r i l y c l o s e to (A) whose topological s t r u c t u r e in W i s
different f r o m the topological s t r u c t u r e of (A).
In Chapter V I (918.4)we established that the s t r u c t u r a l l y stable
s y s t e m s f o r m a n open s e t in the space R* and that this s e t is everywhere
dense in R*. Structurally unstable s y s t e m s f o r m i n R+ "partitions"
s e p a r a t i n g between regions filled with s t r u c t u r a l l y s t a b l e s y s t e m s . Each
of t h e s e regions c o n s i s t s of dynamic s y s t e m s with the s a m e topological
s t r u c t u r e in V.
In QT w e investigated the different topological s t r u c t u r e s of a dynamic
s y s t e m in a p a r t i c u l a r region and the f a c t o r s determining this s t r u c t u r e .
Our p r e s e n t topic d e a l s with t h e c h a n g e s i n t h e t o p o l o g i c a l
s t r u c t u r e of a s y s t e m i n E w h e n t h e d y n a m i c s y s t e m
(i.e., the right-hand sides, the functions P a n d Q ) i s a l t e r e d . The
remaining c h a p t e r s of the book are concerned with the applications of this
topic to a number of important p a r t i c u l a r c a s e s .
It is naturally a s s u m e d that, despite the changes, the dynamic s y s t e m
always r e m a i n s in the s p a c e R*.
In applications, this question is s o m e t i m e s considered in a r e s t r i c t e d ,
p a r t i c u l a r f o r m . Indeed, given a c e r t a i n s e t E , E c R*, one c o n s i d e r s the
changes in the topological s t r u c t u r e in w as the dynamic s y s t e m r u n s
through the points of this s e t . The p a r t i c u l a r choice of the s e t E is
determined by the problem being c o n s i d e r e d . E is o f t e n c h o s e n a s a s m a l l
neighborhood of a given dynamic s y s t e m (Ao) or as s o m e line, s u r f a c e , or
h y p e r s u r f a c e i n the s p a c e R*. Dynamic s y s t e m s r e l a t e d to r e a l physical
p r o b l e m s generally contain one o r s e v e r a l p a r a m e t e r s , i.e., they have
See 55.1. is the space of dynamic systems of class N with a metric defined b y the distance ( i n G ) of
the functions P and Q and their derivatives to r-th order inclusive. Rr)is the space of analytical functions
with the same metric.
I
204
the f o r m
In these p r o b l e m s w e generally have to c o n s i d e r the changes i n the

topological s t r u c t u r e of the dynamic s y s t e m a s the p a r a m e t e r s Li v a r y i n
a c e r t a i n region. The :;et E i n this c a s e is c l e a r l y a n m-dimensional
hypersurface (or s o m e region of this h y p e r s u r f a c e ) . If the s y s t e m depends
on a single p a r a m e t e r , the corresponding s e t E is a 1 i n e in the s p a c e R*.
The question of the changes in the topological s t r u c t u r e of a partition
into paths with changes i n the dynamic s y s t e m is of g r e a t independent
theoretical i n t e r e s t . Two f u r t h e r f a c t o r s g r e a t l y enhance i t s i m p o r t a n c e .
The f i r s t is that the study of the changes in the topological s t r u c t u r e
associated with changes in a dynamic s y s t e m , i . e . , t h e t h e o r y o f
b i f u r c a t i o n s , provided the main tool in the investigation of p a r t i c u l a r
dynamic s y s t e m s . A s we have noted above, no g e n e r a l r e g u l a r methods
a r e available f o r this investigation, and without exaggeration w e can s t a t e
that a l m o s t all the avai:.able r e s u l t s in this direction have been obtained
using the theory of bifurcations. The theory of bifurcations t h e r e f o r e plays
a leading r o l e i n the study of p a r t i c u l a r s y s t e m s .
The second factor is a s s o c i a t e d with the importance of the theory of
bifurcations in applied p r o b l e m s , i n p a r t i c u l a r in physical and engineering
applications. Dynamic s y s t e m s corresponding to these p r o b l e m s always
contain a c e r t a i n n u m b e r of p a r a m e t e r s . The changes in topological
s t r u c t u r e following changes i n these p a r a m e t e r s a r e of the utmost i m p o r -
tance in the a n a l y s i s of the p r o p e r t i e s of physical s y s t e m s r e l a t e d to the
topological s t r u c t u r e of the corresponding dynamic s y s t e m (e.g., when
considering sustained oscillations i n a given physical s y s t e m j. The theory
of bifurcations in one f o r m o r another is t h e r e f o r e applied virtually to
e v e r y dynamic s y s t e m *correspondingto a physical p r o b l e m .
Lye will consider i n #hat follows the case of s m a 11 c h a n g e s of a
dynamic s y s t e m . This is the key to the study of "large" changes, and i t
a l s o h a s numerous important applications, e.g., in p r o b l e m s of stability
of physical s y s t e m s . O u r problem is thus formulated in the following
form: i n v e s t i g a t e t h e c h a n g e s i n t h e t o p o l o g i c a l s t r u c -
t u r e of t h e p a r t i t i o n o f r e g i o n w into paths following
small changes in the corresponding dynamic system.
Only s t r u c t u r a l l y u n s t a b l e s y s t e m s should be considered,
since i f ( A ) is s t r u c t u r a l l y s t a b l e in il',i t s topological s t r u c t u r e d o e s not
change as a r e s u l t of srnall changes in the s y s t e m . If, however, ( A ) is
s t r u c t u r a 11y u n s t $3.b 1e , dynamic s y s t e m s of different topological
s t r u c t u r e s always exist in any a r b i t r a r i l y s m a l l neighborhood of (A) (in
the s p a c e R*i.
We ' s a y i n this case that t h e p o i n t ( A ) o f t h e s p a c e R* i s a
b i f u r c a t i o n p o i n t o f a d y n a m i c s y s t e m . " B i f u r c a t i o n is
generally understood a s the c h a n g e i n t h e t o p o l o g i c a l s t r u c t u r e
of a dynamic s y s t e m o c c u r r i n g when i t p a s s e s through a bifurcation point.
Structurally unstable s y s t e m s , and only these s y s t e m s , a r e
bifurcation points in the s p a c e of dynamic s y s t e m s , and the problem
A mor? precise statement would be the following: !A) is a bifurcation point of tho topological structure
oid ,l)riamic system in F. W c will nevertheless use the more concise form given in the main text.
205
Ch.VIII. BIFURCATIONS OF DYNAMIC SYSTEMS
r e d u c e s to the following: i n v e s t i g a t e t h e c h a n g e s i n t h e
t o p o l o g i c a l s t r u c t u r e of a s t r u c t u r a l l y u n s t a b l e s y s t e m
on p a s s i n g to s u f f i c i e n t l y c l o s e s y s t e m s .
The next topic to consider is the classification of s t r u c t u r a l l y unstable
s y s t e m s . A s t r u c t u r a l l y unstable s y s t e m may be "more s t r u c t u r a l l y
unstable" or "less s t r u c t u r a l l y unstable.'' r h e s e i m p r e c i s e t e r m s c a n be
imbued with p r e c i s e mathematical meaning by introducing the concept
of t h e d e g r e e o f s t r u c t u r a l i n s t a b i l i t y of a s y s t e m . Originally
this concept was introduced in /S/. For the s a k e of simplicity, w e will-
only give a definition of the degree of s t r u c t u r a l instability in a region W
bounded by a cycle without contact. I'he corresponding definition f o r a
general bounded region will be given a t a l a t e r stage (S31, Definition 30).
W e shall a s s u m e that the relevant s y s t e m s are e i t h e r analytical in G o r
of c l a s s N , where N is a natural number whose magnitude, as w i l l be s e e n
f r o m the definition, depends on the degree of s t r u c t u r a l instability of the
system. is a subregion of G bounded by a cycle without contact.
Structurally stable s y s t e m s in w w i l l be called s y s t e m s o f z e r o
d e g r e e o f s t r u c t u r a l i n s t a b i l i t y . A s we know (18.4,a),
s y s t e m ( A ) is s t r u c t u r a l l y stable in it if h a s the following property: for
any e > 0, t h e r e e x i s t s 6 > 0 such that if (A) is I - c l o s e t o ( A ) , then
, e -
(W,A ) =(W, A ) .
W e will define the d e g r e e s of s t r u c t u r a l instability by induction.
D e f i n i t i o n 23. A dynamic system (A) of class N , 3 is said to be a
system of 1st degree of structural instability (or to have a degree of
structural instability 1 ) in @ i f it is not structurally stable in this region
and satisfies the following conditim for any e > 0, there exists 6 > 0 such
that for any structurally unstable system (A) 6-close to (A)to rank 3 we
have - - e -
(W, A ) (W, A ) .
System ( A ) of class N , 5 is said to be a system of 2nd degree of

structural instability in @ i f it is not a system of zero or first degree of
structural instability and the following condition is satisfied: f o r any e > 0,
there exists 6 > 0 such that any system (A)6-close to (A) to rank 5 i?
either a system of zero or first degree of structural instability in W or
satisfies the relation
- - e -
( W , A ) E (W, A ) -
System (A) of class N > 2 k f 1 is said to be a system of k-th degree of
structural instability in W g it is not a system of lower degree of
..
structural instability (i.e., zero, l s t , 2nd, , (k - 1) -th) and the following
condition i s satisfied: fmany e > 0, there exists 6 > 0 such that any
system (A) which is 6-close to rank 2k + 1 to (A) either has a degree of
w
structural instability of at most k - 1 in or satisfies the relation
A few r e m a r k s concerning d e g r e e s of s t r u c t u r a l instability. W e see

f r o m Definition 23 that f o r a s y s t e m to have a definite d e g r e e of s t r u c t u r a l
instability, i t must be a s y s t e m of a sufficiently high class. T h i s is not
a n unexpected conclusion: we have encountered a s i m i l a r situation in
206
Chapter I. Indeed, in defining the multiplicity of a root of a function, s a y
( 9 1 . 3 , Definition 21, w e could speak of a root of multiplicity r of a function
F @)only i f F (z) w a s a function of c l a s s . V > r . A somewhat puzzling point
in o u r definition of d e g r e e s of s t r u c t u r a l instability is the fact that only
s y s t e m s of c l a s s S > 2 k + 1 c a n be of a k-th d e g r e e of s t r u c t u r a l instability.
\\'e will not d i s c u s s h e r e the f a c t o r s r e s p o n s i b l e f o r this r e s t r i c t i o n . Note,
however, that t h i s f a c t is r e l a t e d to the p r o p e r t i e s of a multiple f o c u s d i s -
c u s s e d i n Chapter IX ($25.1, T h e o r e m 40).
A siqnificant shortconiing of Definition 2 3 is that i t d o e s not a s s i g n a
definite d e g r e e of struc:tural instability to e a c h and e v e r y dynamic s y s t e m .
For s y s t e m s of a finite class this is obvious. Indeed, let ?; = 3k L 1. k > I .
\[e c a n e a s i l y c o n s t r u c t a s y s t e m ( A Jof class -V, which is not a s y s t e m of
+
class .V + 1, with k 1 equilibrium s t a t e s , e a c h of niultiplicity 2, i n s o m e
r e g i o n E. It follows f r o m Definition 23 and f r o m the definition of the
multiplicity of a n equilibrium s t a t e (S7.3)that this ( A ) cannot have a d e g r e e
of s t r u c t u r a l instability less than o r equal to k. On the o t h e r hand,
Definition 2 3 s t a t e s that ( A ) cannot have a d e g r e e of s t r u c t u r a l instability
g r e a t e r than k. The s y s t e m ( A ) that w e have c o n s t r u c t e d thus d o e s not have
a definite d e g r e e of s t r u c t u r a l instability.
Let u s now c o n s i d e r the analytical case. It is r e a d i l y s e e n that
analytical s y s t e m s of any finite d e g r e e of s t r u c t u r a l instability e x i s t .
F u r t h e r m o r e , t h e r e e x i s t analytical s y s t e m s with o t h e r analytical s y s t e m s
of a l l finite d e g r e e s of instability contained in any of t h e i r neighborhoods.
r h e s e s y s t e m s a r e naturally a s s i g n e d a n infinite d e g r e e of s t r u c t u r a l
instability. However, it d o e s not follow f r o m Definition 2 3 that e a c h
analytical s y s t e m may have a definite finite o r infinite d e g r e e of s t r u c t u r a l
instability .*.
S y s t e m s of the 1 s t d e g r e e of s t r u c t u r a l instability a r e r e 1a t i v e l y
s t r u c t u r a l l y s t a b l e i n the set of all s t r u c t u r a l l y unstable s y s t e m s . ..'.
Similarly, s y s t e m s of k-th d e g r e e of s t r u c t u r a l instability a r e relatively
s t r u c t u r a l l y s t a b l e i n the s e t of all s t r u c t u r a l l y unstable s y s t e m s of d e g r e e
of s t r u c t u r a l instability > k.
b'e have s e e n before that dynamic s y s t e m s a s s o c i a t e d , w i t h physical
p r o b l e m s g e n e r a l l y contain one o r several p a r a m e t e r s . T h e r e f o r e ,
s y s t e m s dependent on p a r a m e t e r s a r e of p a r t i c u l a r i n t e r e s t . Let, for
simplicity, the right-hand s i d e s contain a single p a r a m e t e r , i . e . , H e a r e
considering a s y s t e m of the f o r m
do
-dt= FJ(1.Y, ).). $=Q((., Y, 1.). (A I )
The p a r a m e t e r 1. m a y v a r y o v e r a c e r t a i n s e t of r e a l n u m b e r s o r r u n
?hisough the e n t i r e real a x i s .
If w e a r e i n t e r e s t e d only in s y s t e m s which are obtained f o r v a r i o u s
v a l u e s of the paramete:- 1, w e should specifically c o n s i d e r the concept of
s t r u c t u r a l stability and d e g r e e s of s t r u c t u r a l instability i n r e l a t i o n to
s y s t e m s (A),). _The following definition of s t r u c t u r a l stability c a n be given
(for a r e g i o n W bounded by a c y c l e without contact): s y s t e m (A %) is
s t r u c t u r a l l y s t a b l e in T? r e l a t i v e to s y s t e m s (A ,,) i f t h e r e e x i s t s 6 > 0 such
.* This can he rhown without much trouble. See Gudkov / 8 L p.425.

In the following sense: if we consider the set of structurally unstable systems and closeness fo rank
!:items of the 1st degree of structural instdbility a r e stable relative to this set.
1,
207
Ch. VIII. BIFURCATIONS OF DYNAMIC SYSTEMS
that e v e r y dynamic s y s t e m ( A h )f o r which I h - ho I < 6 h a s the s a m e topological

s t r u c t u r e as (Ak,,) inz. The d e g r e e s of s t r u c t u r a l instability a r e s i m i l a r l y
defined.
If s y s t e m ( A A o is
) structurally stable in the usual s e n s e , i t is a l s o
s t r u c t u r a l l y stable relative to s y s t e m (Ah). The converse, naturally, is not
always true: a system may be s t r u c t u r a l l y stable relative to the
s y s t e m s ( A h ) , without being s t r u c t u r a l l y stable in the usual s e n s e .
The concept of bifurcation may also be redefined relative to s y s t e m s
(A W e thus a r r i v e a t the following definition of the b i f u r c a t i o n
v a l u e of t h e p a r a m e t e r :
D e f i n i t i o n 2 4 . The value ho of the parameter h i s called a
bifurcation value of the parameter if there exist values of the parameter E.
arbitrarily close to ho for which the topological structure of the dynamic
system in the relevant region is different f r o m the topological structure
of ( A ~ ~ )Values
. of the parameter which are not bifurcation values are
called ordinary values.
The situation is entirely analogous for a s y s t e m dependent on s e v e r a l
p a r a m e t e r s . r h u s , f o r instance, f o r a s y s t e m
dx
x = P ( x , Y, A, p), $ = Q ( G V , A, p), (Ah+)
depending on two p a r a m e t e r s h and p , w e c a n speak of s t r u c t u r a l stability

o r d e g r e e s of s t r u c t u r a l instability r e l a t i v e to the s y s t e m s ( A h , = ) . W e
can s i m i l a r l y speak of the b i f u r c a t i o n p a i r o f v a l u e s o f t h e
p a r a m e t e r s h , p and of the b i f u r c a t i o n p o i n t i n t h e p l a n e o f
the pa ram et er s . If the p a r t i c u l a r region in the plane of the p a r a -
m e t e r s contains no bifurcation points, all the dynamic s y s t e m s c o r r e -
sponding to t h i s region have the s a m e topological s t r u c t u r e . A change in
topological s t r u c t u r e may o c c u r (as a r e s u l t of a continuous change in the
p a r a m e t e r s ) only when the s y s t e m c r o s s e s through a bifurcation point.
It is readily s e e n that the investigation of all the bifurcations of a
dynamic s y s t e m in a given region r e d u c e s to a n inspection of the changes
which occur following s m a l l changes of the s y s t e m i n the neighborhood of
the elements determining the topological s t r u c t u r e . In other words, i t
suffices to investigate the changes i n the topological s t r u c t u r e in the
neighborhood of the equilibrium s t a t e s , closed paths, and l i m i t continua
in 1 7 . Some of the pertinent topics w i l l be treated fully o r partially in
this and next chapters.
In conclusion of this section, l e t u s consider two examples.
E x a m p 1 e 7 . Consider a s y s t e m
dx
= P ( x , -y, a) = x c o s a +y sin a- ( x cosa-y sin a) (5' + u*), (A ,I
-$=Q (5, y, a) = ssin a--y cos a- ( x s i n a + y cos a)(x* + -y*),
depending on a single p a r a m e t e r a. This s y s t e m is generated by a rotation

of the vector field of the s y s t e m .
d
d tz
-= x-x(x2+61?, ~d =
y -Y-u(ss+Y')
bo)
through the angle a . W e may therefore take a to v a r y f r o m 0 to 2n. W e .

shall consider the s y s t e m (A,) o v e r the e n t i r e plane ( x , y). Since(A,+,) is
208
obtained f r o m (A,! by r e v e r s i n g the direction of the field v e c t o r s a t e v e r y
point of the plane, w e need only consider the s y s t e m {A,) for O g a < x .
F r o m the relations
P(-t, -y.a);=--P(z, y,a), Q(-z, -y,a)=-Q(x, y,a)
it folloivs that the phase-plane representation of the dynamic s y s t e m ( A % )

for any (t is s y m m e t r i c z l about the origin (S20, Example 4 ) .
Me will f i r s t apply the Bendixson c r i t e r i o n . Since
- -
w(i,Y, Y, a) = - 4 ( t 2 + y t ) C O S a ,
a) + @(i,
dr UY
dP
w e see that f o r a # % , Cl,ta<n, t h e s u m x + $ d o e s not r e v e r s e i t s sign
a n y x h e r e in the plane. Therefore, by the Bendixson c r i t e r i o n (QT, 12.3),
s y s t e m ( A , ) with a#% h a s neither closed paths (in p a r t i c u l a r , limit
c y c l e s ) nor closed c u r v e s consisting of paths in the phase plane.
Let t = z ( t ) . y = y ( t ) b e paths of ( A a ) . Then, a s i t follows from the
s y s t e m equations,
+ y ( t )y ( t )=Ti
z ( t )z ( t )
* d
+y (t)'] =t z c o s a + 2zy sina--y2
[t( t ) 2 cos a - ( S f y2)* cos a.
r h e last relation shows that f o r Oga<-$, the infinity is absolutely unstable,

$ <a,tn
2nd f o r the inf-nity is absolutely stable (see 20, Example 4 , and
ais0 Q r, 5 1 3 . 1 1.
Me c a n now easily establish the configuration of the paths of s y s t e m ( A a )
for v a r i o u s values of the p a r a m e t e r a , O g a < x . W e w i l l consider
1 I 7
s e p a r a t e l y the four cases a = O , O<a<$, a = T , $<a<x.
1 ) a = 0. F r o m the equations of (Ao) w e readily see that the positive
and the negative coordinate s e m i a x e s are paths of the s y s t e m . Calculations
show that (&) h a s t h r e e equilibrium s t a t e s , the saddle point 0 (0,0) and two
stable dicritical nodes A (-i, 0) and B ( 1 , O ) . Hence i t follows that, in the
absence of closed c u r v e s consisting of path segments, the partition of the
phase plane into paths h a s the f o r m schematically shown in F i g u r e 9 2 .
FIGURE #9'i
209
Ch.VIII. BIFURCATIONS OF DYNAM[C SYSTEMS
2) 0 < a (5. System (A,) again h a s the s a m e three equilibrium s t a t e s

0 (0, 0 ) , A ( - 1 , 0), B ( 1 . 0 ) as in the previous case.': 0 (0, 0) is a simple saddle
point for any a. The points A and B a r e stable structurally stable foci.
There a r e no closed c u r v e s consiting of system paths, in particular, no
limit cycles and saddle -to-saddle s e p a r a t r i c e s . Therefore the a-
s e p a r a t r i c e s of the saddle point 0 r e a c h from infinity and the a - s e p a r a t r i c e s
wind onto the two foci A and B. To define the direction of winding of the
paths ( s p i r a l s ) around the foci, we have to establish the field of directions
on the x a x i s . For y = 0 , *d=l x sin a - 2 sin a = ( x - 2)sin a . Since f o r the
values of a being considered s i n a > 0, w e have
TdY> O for O<x< 1 and for - m<x< -1.
s<O for - l < x < O and f o r l<x< + 00.
a2
Moreover, on the x axis, ; i l = ( x - x 3 ) ~ ~ ~ a ,s o that $=tan a . The direction
of the s e p a r a t r i c e s of the saddle point 0 is determined f r o m the equation
sin a . k2 + 2 cos a . k-sin a = 0
( s e e QT, $9.2, c o r o l l a r y of Lemma 1). Its solutions are
k, = tan?, kz =tan( +++).

W e now readily see that the partition of the phase plane into paths should
have the f o r m shown schematically in Figure 93.
3 ) a = $ - . The s y s t e m takes the f o r m
A d i r e c t substitution shows that (Ax,2)h a s the common integral
(x2 + y2)Z- 2 (x%-)'y = c (11

( s e e QT, S1.13).
Since
(XS +y y - 2 (xZ--y2) 3 (XZ- l)2+ 292 (1 + S P ) +y4- 1,
w e conclude that C > - I . W e may therefore take C = a 4 - l , where a > O .

Equation (1) thus takes the f o r m
( ~ Z + y * ) ~ - 2 (x*-yz) =a4-I. (2)
The c u r v e s ( 2 ) constitute a family of Cassini ovals with the foci A (-1, 0) and
B (1, 0). Elementary a n a l y s i s shows that f o r a > n the c u r v e s ( 2 ) are
convex ovals, f o r 1 ( a ( flthese are "pinched" ovals, and for a = 2 the
c u r v e is a l e m n i s c a t e . For O < a < l the c u r v e s break into two s e p a r a t e ovals,
* When a vector field is rotated, the number and the position of the equilibrium states do not change. Only
their character may change (see Q T , 5 1 . 1 4 , remark preceding Example 7 ) .
543 8 210
P 29. DEGREE OF STRL'CTCR..\L ISST;\BILITY AKD BIFLHCAT[ONS
and finally f o r a = 0 they degenerate into two points d and B ( F i g u r e 9 4 ) .

Each of the o v a l s is a path of s y s t e m ( A n?), the lemniscate c o n s i s t s of
three paths (the saddle point 0 and the two s e p a r a t r i c e s forming loops),
and the points .+ and
l B a r e c e n t e r s . I'he direction along the paths is
readily defined by considerinq the sign of $ f o r y = 0. \Ve thus obtain the
configuration shown in Figure 94.
4) $ <a < x . This case is analyzed along the s a m e lines as case 2.
The configuration of pat.hs in the phase plane is schematically shown in
Figure 95.
FIGLIRE 94 FIGLIRE 35
For a = z , IC < a < -32x , 3 3

a = F1- C I1, T;I.- < a < 2% , the configuration of the
paths is the s a m e a s in F-igures 9 2 , 9 3 , 5 4 , and 95, respectively, but the
direction of motion along the paths is r e v e r s e d . F o r a = 2.7, w e r e t u r n
to the original s y s t e m (Ao).
Let be the i n t e r i o r of the circle
22 +y* = R*,
Lvhere the r a d i u s R i s so l a r g e that the l e m n i s c a t e
(5~fy"*-2(t*--ya)=O
is contained entirely insi3e W . F r o m the fundamental theorem of
s t r u c t u r a l stability (Theorem 23, 1 1 8 . 2 ) i t follows that f o r a between the
l i m i t s Oxcr < 2rr.a # 2 , a # %3 x , ( A z ) is s t r u c t u r a l l y stable in T,'' and
3
for a = 2
and a = yzc i t is s t r u c t u r a l l y unstable. The b i f u r c a t i o n
v a l u e s of t h e p a r a m e t e r are thus $ a n d $ n .
Theorem 2'3 was proved for a regicn with a normal boundary. whereas R need not be such a region for
( A m ) , However, a circle of a sufficiently large radius R , > R is a cycle without contact for ( A s ) , and
therefore inside this circle (A,) is structuralty stable by Theorem 23, and then it is also structurally
stable in w f 9 6 . 1 . Lemma 1).
21 1
I
I
If in the determination of the topological s t r u c t u r e w e c o n s i d e r not only
the configuration of the paths but a l s o the direction of motion along the
paths, bifurcation o c c u r s when the p a r a m e t e r p a s s e s through the value
-$, say, since the s t a b l e foci Aand Bchange into u n s t a b l e foci, I
i.e., the topological s t r u c t u r e changes. Note, hoflever, that if the d i r e c -
tions along the paths a r e ignored, the t r a n s i t i o n of the p a r a m e t e r through
the bifurcation value does not lead to a bifurcation in our example.
F i g u r e s 92 through 9 5 enable u s to t r a c e the changes in the configuration
of the paths of (Ao) during the rotation of the v e c t o r field. Originally, the
points A and B a r e nodes, and the s e p a r a t r i c e s of the saddle point O a r e the
coordinate s e m i a x e s . A s the v e c t o r field is r o t a t e d in the positive d i r e c -
tion, the nodes change into foci, the paths wind onto the foci i n the clock-
wise s e n s e , and the tangents to the s e p a r a t r i c e s a t the point 0 a l s o r o t a t e
in the positive direction with half the rotation velocity of the field v e c t o r s . *
For a = ,: all the paths a r e closed, except the foci (which a r e now
c e n t e r s ) and the saddle point 0 with i t s s e p a r a t r i c e s . The s e p a r a t r i c e s
join in p a i r s forming loops. A s the field is f u r t h e r r o t a t e d , the c e n t e r s
again become foci, but the paths unwind in this c a s e . r h e s e p a r a t r i c e s
of the saddle point 0 s e p a r a t e , the a - s e p a r a t r i c e s extending to infinity for
I+ f 00 and the w - s e p a r a t r i c e s going to the foci for t - c - 00. The change
of configuration with f u r t h e r rotation of the field is obvious.
Let u s consider s t i l l another example which i l l u s t r a t e s the v a r i a t i o n
of the topological s t r u c t u r e a s a r e s u l t of field rotation.
E x a m p l e 8. Consider the s y s t e m
2
dt
= --tsin a--y cos a + (sa+ya- 1)s (5 cos a- y sin a),
(E,)
3
df
=s c o s a - y sin a + (9+ g ~ - 1)s (=sin a + y casa).
I This s y s t e m is obtained by rotating the v e c t o r field of the s y s t e m

dx
-= (BO)
through the angle a . We may t h e r e f o r e take a as varying between -x and x.

However, (B,) and ( B a + xhave
) identical paths, which only differ in the
direction of motion, and w e may t h e r e f o r e consider only the s y s t e m (B,)
f o r --+<a<-% 2
System (Bo). and t h e r e f o r e (B,), h a s a single equilibrium s t a t e O(0,O).

The c h a r a c t e r i s t i c equation of this equilibrium s t a t e
I
* - I
0
] cos a +sin a cos'a-sin a-A =
h a s the roots
hi, =cosa-sina zk 1/ -sin %- 1.
* This follows from the relations k, = tafl 2, kz = tan

2
(++$) (p. 210).
f 22. DEGREE @F STRL'CTLTAL IKST.46ILI TY AND BIFCRC.4TIONS
T h e r e f o r e f o r a b e t w e e n t h e l i m i t s -$<a<-$-,
., the equilibrium s t a t e O ( 0 , O )
of (B,) is
a n unstable focus f o r -?<a<
2
--+ and f o r
't --~-==z<~;
4 4
a n unstable dicritical node f o r a = -5-
6'
a multiple focus or c e n t e r f o r a=:;
a s t a b l e focus f o r nlh<a<+.,
f i e paths s = s ( t ) , y=y(t)of (B,) are tangent to the c i r c l e
ga jy* = Ra
Nhen
F r o m the equations of (E.=),
. .
=;yy= -W+!P), (7)
ifa=?.
2
Finally note that, in polar coordinates, (B,) c a n be written in the form
9
dt -
p [(ps- cos a -sin a], (8 I
-= 2 (p* - i)asin a +cos a. (91
Equality ( 8 ) is c l e a r l y equivalent to ( 6 ) f o r a # and to (7) f o r a=;.

The above r e l a t i o n s enable u s to investigate the configuration of the
paths of (B,) f o r v a r i o u s values of the p a r a m e t e r a.
I) a = $. The equilibrium s t a t e 0 (0, 0) is a stable focus. It follows
f r o m ( 5 ) and ( 7 ) that all c i r c l e s ( 4 ) are c y c l e s without contact. The
s y s t e m therefore h a s no closed paths. Since the focus 0 is the only
equilibrium s t a t e , the s y s t e m h a s no l i m i t continua consisting of
continued paths. By ( 7 ) , the paths c r o s s i n g the cycle without contact (4)
e n t e r into the cycle as f i n c r e a s e s . Thus all the paths of the s y s t e m
wind onto the focus 0 (0, a s t i n c r e a s e s , and go to infinity as t
(1)
d e c r e a s e s ( F i g u r e 96). Equation ( 9 ) shows that a s t i n c r e a s e s , the motion

along the paths is counterclockwise.
2) 4 > a>:. In this case, the equation
(p' -2)* c a s a - s i n a =0 (10)
213
Ch.VIII. BIFURCATlONS OF DYNAMIC SYSTEMS
h a s a single r e a l root po = 1/i + m.By (8), the c i r c l e
p=po, i . e . , za+y2=p:, (11)
is a path of (B=). All the other c i r c l e s (4), in v i r t u e of (6), a r e c y c l e s

without contact; i f R > po (I? < p0)> the paths c r o s s i n g the c i r c l e ( 4 ) leave
the c i r c l e ( e n t e r into the c i r c l e ) a s t i n c r e a s e s . The c i r c l e (11) is t h e r e -
f o r e a n unstable l i m i t cycle of the s y s t e m . The point 0 (0. 0 ) i s a stable
s t r u c t u r a l l y s t a b l e focus. The s y s t e m h a s no o t h e r l i m i t continua, and all
the paths a r e t h e r e f o r e s p i r a l s unwinding f r o m the l i m i t cycle. It follows
f r o m (9) that the motion along these paths in the direction of i n c r e a s i n g t is
counterclockwise. The configuration of the paths is shown in F i g u r e 97.
The a n a l y s i s of the other c a s e s p r o c e e d s along the s a m e lines.
3 ) a = $. The s y s t e m h a s a single unstable l i m i t cycle, the c i r c l e
52 ty2 = 2. (12)
The equilibrium s t a t e 0 (0, 0) is a multiple ( s t r u c t u r a l l y unstable) s t a b l e
focus. The motion along the s p i r a l s with i n c r e a s i n g t is in the c o u n t e r -
clockwise s e n s e ( F i g u r e 98).
FIGURE 96. $. Stable focus.
FIGURE 97. $ - > a > ? ,

4
r>m FIGURE 98. a $ , r = 4.
Unstable cycle; stable focus. Unstable cycle; multiple
stable focus.
4 ) $->a>0. The s y s t e m h a s t w o l i m i t c y c l e s , a n unstable l i m i t cycle
za + yz= 1+1/GTG, (13)
214
which is a c i r c l e of r a d i u s g r e a t e r than 1, and a s t a b l e l i m i t cycle
$+lJZ=i- I/tan, (14)
s h i c h is a c i r c l e of r a d i u s s m a l l e r than 1. The equilibrium s t a t e O(0,O) is
a n unstable focus. T h o motion along the s p i r a l is i n the counter,clockwise
sense (Figure 59).
5 ) '1 = 0 . T h e s y s t e m h a s one s e m i s t a b l e , and t h e r e f o r e multiple ( s e e
$12.3), l i m i t cycle, the c i r c l e
22 +-y* = 1, (15)
and a n unstable focus 0 (0. 0). The motion along the paths is i n the c o u n t e r -
clockwise direction ( F i z u r e 100).
6) 0> a > - +. All ,circles ( 4 ) a r e c y c l e s without contact. The s y s t e m

therefore h a s no closed paths and no closed c u r v e s consistingof paths. All
the paths go to infinity a s t i n c r e a s e s , and for t+- - 00 they a i n d onto the
unstable focus 0 (0.0 ) .
\ \ e see f r o m ( 9 ) that
and
l'herefore inside the c i r c l e
the motion along the paths is counterclockwise, and outside this circle the
motion is clockwise (Fig-ire 101). T h e c i r c l e (16) is evidently orthogonal
to e a c h c r o s s i n g path.
215
Ch.VlII. BIEVRCATIONS OF DYNAMIC S Y S T E M S
FIGURE 101. 0 > a > - . No FIGURE 102. u = -$. NO cycles.

cycles. R >fi
unstable focus. R =fi
unstable dicritical node.
7) a = --. T h e configuration of the paths is the s a m e a s in the I

previous c a s e , but the point 0 (0, 0) i s an unstable dicritical node, and not a
focus ( F i g u r e 1 0 2 ) . C i r c l e (16) is replaced by c i r c l e (12).
8 ) -$>a>-$. The equilibrium s t a t e O ( 0 , 0) is a n unstable focus.
A l l c i r c l e s ( 4 ) a r e c y c l e s without contact. The derivative = (p2 - I)'sin a i
+ c o s a v a n i s h e s on the circles
and
pa=i- /-.
In the r i n g between t h e s e c i r c l e s , dB > 0, and e l s e w h e r e in the plane $ 0.
T h e r e f o r e the motion along the paths with i n c r e a s i n g t is in the c o u n t e r -
clockwise d i r e c t i o n inside the r i n g and in the clockwise direction outside
the c i r c l e ( 1 7 ) and inside the c i r c l e ( 1 8 ) ( F i g u r e 1 0 3 ) .
For a= -%, the s y s t e m h a s the s a m e paths a s f o r a=;, but the
d i r e c t i o n of motion is r e v e r s e d . A s a f u r t h e r diminishes f r o m - $ t o
- pn, we s u c c e s s i v e l y obtain the s a m e configurations a s in F i g u r e s 97
through 103, but with the d i r e c t i o n of the a r r o w s r e v e r s e d . Finally, for
3
a = - T n , w e r e t u r n to the o r i g i n a l s y s t e m (B,,z) ( F i g u r e 96).
The bifurcation values of the p a r a m e t e r a (for :>a> --%) a r e c l e a r l y
n , -,
y n 0, and
4
.;- Our a n a l y s i s and the figures c l e a r l y show how the paths
of the s y s t e m evolve as the p a r a m e t e r a d e c r e a s e s (i.e.> a s the field
v e c t o r s r o t a t e c1ockwise)and what bifurcations the s y s t e m e x p e r i e n c e s .
On passing through the bifurcation value of the p a r a m e t e r $, a n u n s t a b 1e
216 ,
2'1. DEGREE OF STRCCTIIRIL INST 4 6 t L I T Y AXiJ 6IFC'RC.ITIOKS
1i m i t c y c 1 e a p p e a r s -a circle of a l a r g e r a d i u s c e n t e r e d at O(this cycle,

as w e s a y , is c r e a t e (1 f r o m i n f i n i t y ). A s a d e c r e a s e s f u r t h e r , the
cycle c o n t r a c t s (the r a d i u s of the c i r c l e
monotonically diminishes ). F o r the bifurca -
tion value a = 5, the topological s t r u c t u r e
of the s y s t e m is the s a m e as before, but the
stable s t r u c t u r a l l y stable focus 0 changes
into a multiple ( s t r u c t u r a l l y unstable) focus.
When the s y s t e m p a s s e s through the bifurca-
tionvalue;, a s t a b l e l i m i t c y c l e is
c r e a t e d f r o m this multiple focus, and the
multiple s t a b l e focus changes into a n u n -
s t a b l e s t r u c t u r a l l y unstable focus, while the
existing unstable limit cycle continues con-
F I G ~ w E 10%. - $ > a > - $ , NO tracting. After that, the s t a b l e l i m i t cycle
, ).der. R , < y - 2 , R z < 1 unstabl? expands, and the unstable l i m i t cycle con-
io<.u 3 , t r a c t s , and f o r the bifurcation value a = 0
both c y c l e s m e r g e into a single multiple
s e m i s t a b 1e cycle. When the s y s t e m c r o s s e s the value a = 0 , this
multiple cycle d i s a p p e a r s , and f u r t h e r d e c r e a s e of a between the l i m i t s
0 >a > - 2
2
d o e s not produce any additional change in the topological
s t r u c t u r e . We see from F i g u r e s 1 0 1 - 103 how the direction of motion
along the paths is r e v e r s e d . Indeed, when the s y s t e m p a s s e s through
a = 6, a c i r c l e p 2 = l + p -'?
/-
a
a p p e a r s f r o m infinity on which t h i s change
occurs. T h i s circle c o n t r a c t s a s a diminishes. For a - 1?.
4'
the focus
0 (0, 0) changes into a d i c r i t i c a l node (of the s a m e stability). F u r t h e r
d e c r e a s e of a produces a second circle n e a r 0 on which the direction of
rotation is again r e v e r s e d , and the d i c r i t i c a l node changes back to a
focus. Both circles move one toward the o t h e r . For a = -$ they m e r g e
and the motion along the two circles is i n the s a m e direction, i.e., clock-
w i s e ( F i g u r e 9 6 with the direction of the a r r o w s r e v e r s e d ) .
Note that as a i n c r e a s e s , the changes of topological s t r u c t u r e o c c u r
in a r e v e r s e o r d e r . h F a r t i c u l a r , f o r s m a l l negative a, the s y s t e m h a s no
l i m i t cycles ( F i g u r e 101:'. When a r e a c h e s the value 0, a s e m i s t a b l e
(multiple) l i m i t cycle zz 4- y* = 1 a p p e a r s (Figure 100). In this case we s a y
that the l i m i t cycle is c r e a t e d f r o m p a t h c o n d e n s a t i o n (or p a t h
c 1u s t e r i n g ). F u r t h e r i n c r e a s e i n a produces two i i m i t c y c l e s . In this
case, w e s a y that a m u l t i p l e l i m i t c y c l e d e c o m p o s e s i n t o t w o
c y c l e s or a n a d d i t i o n a l c y c l e i s c r e a t e d f r o m a m u l t i p l e
c y c 1 e . In o u r example w e thus observed c r e a t i o n of c y c l e s f r o m
infinity, f r o m a multiple focus, f r o m a multiple l i m i t cycle, and f r o m
path condensation. In C t a p t e r s XI through XI11 w e s h a l l encounter additional
cases of l i m i t cycle c r e a t i o n .
217
Ch. VIII. RIFIIRCATIONS OF DYNAMIC SYSTLMS
$ 2 3 . DECOMPOSITION OF A MULTIPLE
EQUILIBRIUM STATE INTO STRUCTURALLY
STABLE EQUILIBRIUM STATES
1. The number of s t r u c t u r a l l y s t a b l e equilibrium

s t a t e s obtained f r o m a multiple equilibrium s t a t e
In the p r e s e n t section we w i l l only c o n s i d e r analytical dynamic s y s t e m s .

Some propositions regarding multiple equilibrium s t a t e s , however, w i l l
r e m a i n valid f o r s y s t e m s of class N a l s o . However, since the v a r i o u s
proofs given for the analytical c a s e e i t h e r r e m a i n without change f o r
s y s t e m s of c l a s s N or a r e actually simplified, we w i l l concentrate on
analytical s y s t e m s .
be a dynamic s y s t e m and 0 (0, 0) i t s equilibrium s t a t e . We a s s u m e that

0 (0, 0 ) i s a multiple equilibrium s t a t e of multiplicity r , where r is a
n a t u r a l number, r > 2 . By Definition 1 5 (57.3) and Definition 5 (2.1), an
equilibrium state 0 of s y s t e m ( A ) is of multiplicity r i f the following con-
ditions a r e satisfied:
(a) t h e r e e x i s t n u m b e r s E~ > 0 and 6, > 0 such that any s y s t e m (A) 60-
close to rank r to s y s t e m (A) h a s at most r equilibriu? s t a t e s in U r o ( 0 ) ;
(b) f o r any e < E, and 6 > 0 , t h e r e e x i s t s a s y s t e m (A) 6-close to r a n k r to
(A) which h a s a t l e a s t r equilibrium s t a t e s in U, (0).
F r o m the definition of r-multiplicity of the equilibrium state 0 and f r o m
the condition r>Z and Theorem 6 ( 5 2 . 2 ) i t follows that 0 (0, 0) is a n isolated
equilibrium state in our c a s e and that
In the following, eo > 0 and 6o > 0 are fixed n u m b e r s defined by condi-

tion (a). Moreover, we take e , to be so s m a l l that 0 (0, 0) is the only
equilibrium state of (A) in U0(0).If n e c e s s a r y , other conditions w i l l a l s o
be imposed on e,, and 60, provided they do not c l a s h with the b a s i c r e q u i r e -
ment of s m a l l n e s s of these n u m b e r s .
Analytical dynamic s y s t e m s 6,-close to r a n k r to s y s t e m (A) w i l l be
called f e a s i b 1 e s y s t e m s . Let C be the boundary of the n_eighborhood
U, (0). We take 6, to be s o s m a l l that no feasible s y s t e m (A) h a s
equilibrium s t a t e s on the c i r c l e C, and the v e c t o r s defined by s y s t e m s (A)
and (A) do not point in opposite d i r e c t i o n s a t any point of this c i r c l e . Then
w,C (c)= w A (c).

We will t r y to elucidat_e maximum information about the equilibrium s t a t e s
of a feasible s y s t e m (A) which lie in V.,(O), i f i t is known that they all a r e
s t r u c t u r a l l y stable. W e w i l l f i r s t p r e s e n t s o m e propositions which a r e .
21 8
4 ?. DECOLIPOS[TIOh OF A A!CLTIPL EQVtLtEiUL'hl STATE INTO S TRCCTCP..ALLY STASLE ST.ATES
applicable to any r-tuple equilibrium s t a t e ( L e m m a s 1 and 2 and

Theorems 31 and 32), and then r e s t r i c t the range of equilibrium s t a t e s by
imposing one f u r t h e r condition.
If the feasible s y s t e m (A) h a s p r e c i s e l y IC equilibrium s t a t e s
01,02,. . ., Ok in the neighborhood Ue,,(0),and they are a l l s t r u c t u r a l l y
stable, w e shall s a y that the multiple equilibrium s t a t e O ( o r the multiple
singular point 0 ) decomposes into str_ucturally stable equilibrium s t a t e s
..
O,, 02, ., O,+on passing to s y s t e m ( A L
L e m m a 1 . If a feasible systetn (A) has r equilibrium states in a
neighborhood u,,(0)of a,u r-tuple equilibrium state 0, all these equilibviutn
states are simple.
P r o o f . Let the feasible s y s t e m (A) have r equilibrium s t a t e s
.
O,, 02, . ., 0, in U,,(O),and a t least one of them, Ol say, is a multiple
equilibrium s t a t e . Let Ui be a neighborhood of O i , i = 1, 2, . . ., r , such that
U t c U,, (0)and no two Ui i n t e r s e c t . the point Oi, i = 2, 3, . ., r , is not a .
simple equilibrium s t a t e of s y s t e m (A), we c a n r e p l a c e t h i s s y s t e m in ,a
sufficiently s m a l l neghborhood Vt of the point Oi( V i c C r i ) by a s y s t e m (A;) a5
close a s d e s i r e d to ( A ) f o r which Oi is a s i m p l e equilibrium state.:' F u r -
t h e r m o r e , in a s2fficiently s m a l l neighkorhood ,'.I of the point O1,Vf c, UI.
we c a n replace ( A ) by another s y s t e m (Ai), as c l o s e as d e s i r e d to (A),
which h a s i n VI a t l e a s t two equilibrium s t a t e s 0;and 0;;this c a n be done
since by Theorem 6 ( S 2 . Z ) a multiple equilibrium s t a t e h a s a multiplicity
higher than 1. By the first footnote to this page, 0; and 0; may
be regard_ed as simple equilibrium s t a t e s . We c a n f u r t h e r construct a
s y s t e m (A') of class r , which would be as c l o s e to rank r as desire><to
s y s t e m (A) and would c o h c i d e outside the neighborhoods Vi with (A) and
inside each of the neighborhoods Fi with (Ai).**. Let now (A") be a dynamic
s y s t e m whose right -hand s i d e s are polynomials-providing a suffLciently
good approximation to th,? right-hand s i d e s of (A'). Evidently, (A") is then
a feasible s y s t e m with a t l e a s t r + l e q u i l i b r i u m s t a t e s in U,,(O). This
contradicts the assumption that 0 is a n r-tuple equilibrium s t a t e . The
proof of the l e m m a is complete.
L e wi >na 2. If 0 i s an r-tuple equilibviurn state of system (a),there
exist s y s t e m s a s close to rank r to s y s t e m ( A ) a s desired, which have r
structurally stable equilibrium states in tre, (0).
'The validity of L e m m a 2 follows directly f r o m the definition of
multiplicity of a n equilibrium s t a t e and f r o m the f i r s t footnote to this
Page.
T h e o r e m 31. If I A @ ) = l i s the Poincare index of an r-tuple
equilibrium state o of systetn (A), then
Z s r (mod2). (3)
P r o o f . Let C be the boundary of the neighborhood U.,(O). By
definition, the Poincard index l i s equal to the rotation W;t (C) of the v e c t o r field
Let the system have the f o r m * = o ( s - q )

dt
jb(y-yl)+ ..., ddyt = c ( + - - l i ) + d ( y - - y r ) + . . . i n the
n?tehborhood of the pointOi(ri, vi). If.=(," :l=O, adding sufficiently small increments to u and d
we obtain a close system withA =i; 0, 1.e.. a system whercOi i s a simple equilibrium state. Similarly, a
jtructurally unstable point c a n be convened into a nruccurally stable point by an arbitrarily small change
of the system.
* [he Fosstbtlity of constructing such a systcm of class r is established in Chapter V I (518.3) i n ow proof
to r'heorem 21.
219
Ch.VIII. BIFUXCATIONS OF DYNAMIC SYSTEMS
of system (A) along the curve C divided by 2 n ( Q T , 9 1 1 . 2 , Definition XI11

and 510.2, Definitio? XI). According to the preceding lemma, there e x i s t s
a feasible system (A) which h a s in U.,(O) p r e c i s e l y r equilibrium s t a t e s ,
a l l of which a r e s t r u c t u r a l l y stable. F u r t h e r m o r e , f r o m equation ( 2 )
wx (C) =w,(C)= 2nI.
On the o t h e r hand, by QT, 11.2, L e m m a 1,
I= (0,) +
+ I , (0,) .. . fIx (6). ( 41
But e v e r y equilibrium s t a t e Ot of (A) is s t r u c t u r a l l y stable, i . e . , i t s index
is e i t h e r + 1 o r -1, so that it follows f r o m the l a s t equality that
Izr(mod2).
The Droof is c o m d e t e .
all of which are structurally stable, then- I

k = r (mod 2). (5)
Proof . A s in the previous theorem, w e prove that
k=l(mod2). (6)
Relation (5) follows f r o m ( 3 ) and ( 6 ) . Q. E. D.
The previous l e m m a s and t h e o r e m s a r e valid, a s we have observed,
for any r-tuple equilibrium s t a t e . W e will now consider in some detail
a relatively narrow, but e x t r e m e l y important c l a s s of equilibrium s t a t e s .
T h r o u g h o u t t h e r e m a i n i n g p a r t of t h i s s e c t i o n , w e w i l l
a s s u m e t h a t t h e s e r i e s e x p a n s i o n s of t h e r i g h t - h a n d
s i d e s - t h e f u n c t i o n s p a n d Q - i n t h e n e i g h b o r h o o d of t h e
r e l e v a n t e q u i l i b r i u m s t a t e O ( 0 , 0) c o n t a i n a t l e a s t o n e
linear term, i.e.,
I pk (0, 0) I +I J'; (0, 0)I + I Q;(0, 0) I + IQ; (0,0) I + 0. (7 1
I To fix i d e a s , l e t
If Q; (0, 0) = 0, but P; (0, 0) does not vanish, say, nothing changes

significantly.
We w i l l f i r s t derive the n e c e s s a r y and sufficient condition for the
equilibrium s t a t e of the p a r t i c u l a r type to be of multiplicity r . Note that
f o r r = 1 and r = 2 , condition ( 7 ) is fulfilled automatically ( T h e o r e m s 6
and 7, 52.2 and 9 2 . 3 ) .
T h e o r e m 33. Let 0 (0, 0) be an equilibrium state of a system
of class r (in particular, analytical system) and let at least one of the f i r s t
derivatives of the functions Pand Q , say Q; (0, O), not vanish at the p d n t
0 (0, 0) . Let further I/ = 'p ( x ) be the solution of the equation
for y in some sufficieiitly small neighborhood of o,* ami
a (2) = P (2,'F ( 3 ) ) . (10)
Then the point 0 is an r-tuple equilibrium state of system ( A ) if and

only if the nriutber 0 is an r-tuple root of the function 0 (I).**
PI' o o f of Theorem 33 is analogous to the proof of the corresponding
proposition in Theorem 7 ( $ 2 . 3 ) . Using L e m m a 1 , $ 1 . 3 in o u r proof, 'we
take f o r P (t,y) the function
atz+&+ ... + a , - d - l + P ( z , y)
with appropriately chosen coefficients a i . Conditions ( a ) and ( b ) of

r h e o r e m 7 are represented by conditions ( 1 ) and ( 8 ) , respectively, and
are fulfilled in o u r c a s e by assumption.
We now proceed to the next theorem. Let 0 (0, 0 ) be a n r-tuple equili-
brium s t a t e of ( A ) f o r which condition ( 8 ) is fulfilled, and e, > Oand 6o > 0
the n u m b e r s introduced in the preceding. W e m o r e o v e r a s s u m e that e, is
sufficiently s m a l l s o that 91 # 0 in the neighborhood U,,(O), and the c u r v e
Q (z, y) = 0 may be defined in this neighborhood by a n explicit equation
Y = 'P (4,where El < z < E.2, Et < 0, 5 2 > 0.
T h e o r e m 34. For any whole number k satisfying the inequalities
= r (mod 2 ) , and fmany positive 6 < 6, and e <&,.there exists an
0 4 k +'r, k
analytical system (xi)Ic,hich is &close to rank r to system ( A ) and has in
c;,(0)precisely k , and at that structurally stable, equilibrium states, all
of which nimeouer lie in L:, (0).
Proof. Let, as before,
P (2,'P (I))= e (I). (10)
By r h e o r e m 3 3 , z = O i s a root of multiplicity rof function e@). There-
fore,
e(z)=Az'+ ... = A ~ ~ ( l f f ( t ) ) , (11)
where A Z O , and f ( z ) is a n analytical function, f ( 0 )= O .
Consider the system
g = P ( x , . Y ) = ~ ( Z , Y ) + P ( ~ ,Y), g = Q ( z , y ) = Q ( z , Y). (12)
where
P(Z, Y) = A (z--x1)(5-s2) . .. ( X - Q ) (ek

+a) (1 + f -e
(1)) (21, (13)
and k is a n integer, O < k < r , k = r ( m o d 2 ) , a > O , g , ( x i < E z . and all q a r e

different ( p ( z , y) is independent of y).
Since o=
Q, the c u r v e q(z, y)=Ocoincides with the c u r v e Q(z.y)=O, i.e.,
in the neighborhood U,,[O) i t may be e x p r e s s e d by a n explicit equation
l'he equation Q (z.y) = 0 has ,I single-valued solution y = cp (z)in a sufficiently small neighborhood of 0
in virtue of condition (8) and the theorem of implicit functions; here, 'p is a function of class r a n d
cp (0)= 0.
*I
If QI
(0, 0)= 0, but Pi (0,0). say, does not vanish, I3 (z)should be replaced by a function (y)
I* =
=Q (q* (y), y), where z = 'p* (y) is the solution of the equation P (I, y) = 0 for r.
22 1
y='p(x), E.,<x<E2.
c
QI.VIII.
rherefore
BIFURCATIONS OF DYNAMIC SYSTEMS
0 ( 5 )= p" (5.6 ( 5 ) )= p" (5, 'p ( 5 ) )= P (x, B) + P (G 'p (4)

=
I
= A (2- xi) (z-z~) .. a +
( z - x ~ ) (Z-k a)(1 +f (5)). ( 14 )
By assumption, O(0,O) is the only equilibrium s t a t e of (A) in V , ( O ) . It
thus follows f r o m (10) and (11) that l + f ( x ) + O f o r E.,<x(Ez. Moreover,
Z ' - ~ + ~ # O , since a > O , and r - k is even. But then we s e e f r o m (14)that
s y s t e m ( 1 2 ) h a s i n U,(O) p r e c i s e l y k equilibrium s t a t e s O i ( x i , tp(zi)).
i = l , 2, ..., k. By (11) and (13),
P ( X ~~ ) = A ( f + f(x)){xr~[(5-x*)(x-5~) @-4-zr1+
+a(x-x*) (x-xz) ... (2-5*)}.
If the numbers X i , i = 1 , 2, ..., k , and a a r e sufficiently s m a l l , the function
p ( z , y) in a finite interval is a r b i t r a r i l y close to z e r o with i t s derivatives,
and s y s t e m ( 1 2 ) is therefore a r b i t r a r i l y c l o s e to r a n k r to s y s t e m (A).
Moreover, f o r s m a l l xi, a l l the points 0,lie i n U , (0).
We will now show that the equilibrium s t a t e s Oi (21,cp ( x i ) ) of s y s t e m ( 1 2 )
a r e a l l simple. 'Indeed, f r o m the r e l a t i o n s P"(q ' p ( s ) ) = g ( x ) ,Q ( x , ' p ( x ) ) = O it
follows that , ,
Therefore
I
c
~(0,) =x ( x i , 'P (zi))= -e' (xt) QI(W, 'P ( X I ) ) (i = 1, 2 , ...,k).
But QI # 0 by a s s u p p t i o n , and ~ ( x does
( x l , rp (q)) J not vanish, since ZLis
a simple root of the function 9(x)(see (14)). Hence, h(O1)#0, i.e., the
..
points O1 ( i = i , 2, ., k) a r e simple equilibrium s t a t e s of s y s t e m (12).
If the points OI a r e s t r u c t u r a l l y stable equilibrium s t a t e s , the proof is
completed. If some of these are s t r u c t u r a l l y unstable equilibrium s t a t e s
( i . e . , multiple foci or c e n t e r s ) we can adopt the s a m e technique a s i n
L e m m a 1 and change over to a n a r b i t r a r i l y close analytical s y s t e m which
h a s in U,,(O) p r e c i s e l y k equilibrium s t a t e s , a l l of which a r e s t r u c t u r a l l y
stable and lie in U . ( O ) . This completes the proof of the theorem.
R e m a r k . Theorems 32 and 34 naturally complement one another
and completely determine the number of equilibrium s t a t e s of a
sufficiently c l o s e system which may lie i n U, (0)if they a r e a l l s t r u c t u r a l l y
stable. Theorem 32, however, h a s been proved f o r a general c a s e ,
w h e r e a s in m e o r e m 34 we a s s u m e d condition ( 8 ) (or (7)). It is therefore
interesting to t r y to establish whether or not Theorem 34 is applicable to
the g e n e r a l c a s e , i.e., when condition (7) is not satisfied.
2. The c h a r a c t e r of the s t r u c t u r a l l y s t a b l e equilibrium

s t a t e s obtained f r o m a multiple equilibrium state with a # 0
In the previous section w e determined the number of s t r u c t u r a l l y stable
singular points to which a n r-tuple equilibrium s t a t e 0 of a dynamic
* W e have previously encountered this formula in the proof of Theorem 7 (02.3).
222
13. DECOkIPOS[TION OF A hlULTIPLE EQLitLtBRIUhl STATE th CO JTRVCTVR.%LLY ST.4BLE iTATES
s y s t e m (A) may decompose on passing to c l o s e s y s t e m s . In this subsection

w e will t r y t o elicit s o m e information about the c h a r a c t e r of t h e s e
s t r u c t u r a l l y stable singular points. W e a s s u m e that the multiple equilibrium
s t a t e 0 is isolated and s a t i s f i e s the condition
IP ; (0, 0) I+ I P;(0, + IQ;(0,O) I C I Q;(0, 0) I # 0.

0) I (7)
The topological s t r u c t u r e of the dynamic s y s t e m in the neighborhood of

such a n equilibrium s t a t e h a s been studied in detail i n QT, 521 and S22.
W e will r e q u i r e the principal r e s u l t s f r o m QT, which a r e r e i t e r a t e d
below. W e distinguish Detween two cases, u = 0 and u # 0.
( a ) Let
u = P ; ( o , O)+Q;(O, O)#O. (15)
In this case, s y s t e m ( A ) is t r a n s f o r m e d by a non-singular linear

transformation t o the f o r m
where Pt and Qzare anal-ytical functions whose series expansions in the

neighborhood of the point O(0,O) consist of t e r m s of not lower than second
order.
Let
be the solution of the equation

Y+Qz(z,Y)=O (18)
in the neighborhood of O(0,O) and l e t the expansion of the function
where m > 2 , and
(the existence of these n u m b e r s m and A m follows f r o m the fact that the

equilibrium s t a t e is isolated).
The following proposrtion applies:
I (QT,521.2, Theorem 65)
1. If m is odd, and Am > 0, the equilibrium s t a t e 0 of s y s t e m (16) is a
topological node ( F i g u r e 104).
2 . If m is odd, and A m < O , 0 is a topological saddle point ( F i g u r e 105).
3 . If m i s e v e n , the equilibrium s t a t e O ( 0 , 0) is a saddle-node, i.e,, i t s
canonical neighborhood c o n s i s t s of a parabolic and two hyperbolic s e c t o r s
( F i g u r e s 106 and 107).
( b ) Let
o=P;(O, O ) + Q ; ( O , O ) = O . (22 1
223
Ch.VII1. RIWRCATIONS OF DYNAMIC SYSTEMS
FIGURE 104. Oddm, A,,, >0. FIGURE 105. Odd m, A,,, C O .
FIGURE 106. Even m, A,,, >0. FIGURE 107. Even m, Am <0 .
In t h i s c a s e , s y s t e m ( A ) c a n be t r a n s f o r m e d by a non-singular l i n e a r
transformation to the f o r m
dz
Y),
-;ii-=~+&.(~ $=QS@. Y). (23)
System (23) in i t s turn c a n be reduced to an even s i m p l e r form
3
dt =q* g=qz(E*q)
by the transformation
E=x, q=y+Pz(z, Y
)7
which is one-to-one in the neighborhood of O(0,O). Reverting to the

original notation, we may thus consider a s y s t e m of the f o r m
dx
-=
dt Yt $=Q2& Yh (24)
where the s e r i e s expansion of Qz(z,y) contains no l i n e a r t e r m s .
224
Since O ( 0 , 0) is by assumption a n isolated equilibrium s t a t e of
s y s t e m ( 2 4 ) , t h i s s y s t e m c a n be w r i t t e n in the f o r m
where h (z), g (z), f ( 3 , y) are analytical functions; h (0) = g (0) = 0; r > 2 ;

a f 0: b may be equal to z e r o ; i f b + 0 , we have n > 1.
in t h i s case, the following proposition a p p l i e s :
I1 (QT, S22.2, Theorems 66 and 6 7 )
1 . Lf r is odd, the equilibrium s t a t e O i s e i t h e r a topological saddle
point, o r a topological node, o r a focus (a c e n t e r ) , or finally a n equilibrium
s t a t e with a n elliptical region (whose neighborhood contains one hyperbolic
and one elliptical s e c t o r , F i g u r e 108 J.
FIGURE 108 FIGURE 109
2. J f r is even, 0 i s e i t h e r a d e g e n e r a t e e q u i l i b r i u m s t a t e (two hyper-

bolic s e c t o r s , F i g u r e 109), or a saddle-node ( F i g u r e 110).
FIGURE 110
Note that by Theorem 3 3 and the condition a # 0, the number r is the

multiplicity of the equilibrium s t a t e 0.
225
We c a n now consider the c h a r a c t e r of the s t r u c t u r a l l y stable equilibrium

s t a t e s obtained f r o m a multiple singular point. F i r s t w e c o n s i d e r the c a s e
when
CI = p ; (0, 0 ) +
9"(0, 0) # 0.
The solution of the problem is v e r y simple, since the c h a r a c t e r of the
component s t r u c t u r a l l y stable s t a t e s is e n t i r e l y determined by the Poincare'
index of the singular point 0. W e may a s s u m e without l o s s of generality
that the s y s t e m h a s the f o r m ( 1 6 ) .
Let H be the number of hyperbolic, and E the number of elliptical s e c t o r s of
a canonical neighborhood of the equilibrium s t a t e 0. According to
Bendixson's formula, the Poincard index of the point 0 is
E-I
I=-+12
( s e e QT, ..
- . Amendix, l o ) .
In our c a s e , as we s e e f r o m (16), u = 1 and A = 0 . Let (A)be a dynamic
s y s t e m 8,-close to (A) to the r e q u i r e d rank, and 01, Op, . . ., 01,the equili-
brium s t a t e s of this s y s t e m lying in U,,(O),all of which a r e s t r u c t u r a l l y
stable. If the n u m b e r s 6, and e, a r e sufficiently s m a l l , the number q = u(O1)
.
i s c l o s e to 1 f o r e a c h of the equilibrium s t a t e s O i , i = I , 2, . ., k , and
Ai = A (Oi) is close to z e r o . Then, i f A (01)> 0, O iis a s t r u c t u r a l l y s t a b l e
-=
node, and i f A i (Oi) 0, Oiis a s t r u c t u r a l l y s t a b l e saddle point. Thus, in
o u r c a s e , all the s t r u c t u r a l l y s t a b l e equilibrium s t a t e s into which a
multiple equilibrium s t a t e 0 decomposes a r e s t r u c t u r a l l y s t a b l e nodes and
saddle points.
By Theorem 3 3 and r e l a t i o n s ( 2 0 ) and (211, the multiplicity of the
equilibrium state 0 (0, 0) of s y s t e m ( 1 6 ) is m . Evaluating the Poincard index
I = I A(0)of the point 0 f r o m (26), we obtain the following r e s u l t s :
1 ) If 0 is a topological node, I = 1 ( E = H = 0).
2 ) If 0 is a topological saddle point, I = - 1 ( E = 0, H = 4 ) .
3 ) If 0 is a saddle-node, I = 0 ( E = 0, H = 2).*'
On the o t h e r hand, we know ( s e e the proqf of Theorem 31) that
IA (o)=I=I;T (01) +IT (02) f ... +IZ (01,)- (27)
Since the Poincar; indices of a s t r u c t u r a l l y s t a b l e node and a s t r u c t u r a l l y

s t a b l e saddle point are +1 and -1, respectively, equation ( 2 7 ) d i r e c t l y gives
the number of s t r u c t u r a l l y s t a b l e nodes and s t r u c t u r a l l y stable saddle points
among the points O,,Oz, . . ., Oh in each of the c a s e s 1, 2 , 3 above. The value
of the number k is completely fixed by T h e o r e m s 32 and 3 4 .
A l l the v a r i o u s r e s u l t s c a n now be s u m m a r i z e d in the following
theorem.
* Note that the Poincar6 index of the equilibrium state 0 (0,0 ) o f system (16)can be readily computed
without using Bendixson's formula (26). T o this end, we should count the number o f times the field vector
of the dynamic system crosses the direction of the y axis while moving along a circle of small radius
centered at 0. See Kranosel'skii et a l . /19/, 67.
226
I 03. DECOXIPOSITIOti OF A ,\IC'LTIPLE EQL'[LIBRIb,\l ST.4TE IKTO STRL'CTC'RALLY ST.4BLE STATES
be a dynamic systeni, 0 (0, 0) an equilibrium state of this system f a r which

A = 0 and o = P; (a, 0) + Q; (0, 0) f a , and let m > 2 be the multiplicity of the
equilibrium state 0. Titen the number k of struchwally stable equilibrium
states Of,02? . . .. Oh into which the tnultiple equilibrium state o decomposes
on passing to arbitrarily close systems* satisfies the conditions
O 4 k g m and k = m ( m o d 2 )
and may be eqtml to any number satisfying these conditions. Each of the
points Oi ( i = i , 2, . . .,k ) is either a structurally stable node or a structurally
stable saddle point. Moreover,
1) i f 0 is a topological node, k is odd and the number of structurally
stable nodes among the points O i , 0 2 , . . ., okis 1 more than the number of
structurally stable saddle points;
2) i f 0 is a topological saddle point, k i s odd and the number of
structurally stable nodes among the points oi, 02, . . ., oh is 1 less than the
number of structurally stable saddle points;
3) if 0 i s a saddle-node, k i s even ami the number of structurally
stable nodes among the ,boints Oi,02.. . ., Or i s equal to the number of
structurally stable saddle points.
R e m a r k . Theorem 35 fully d e t e r m i n e s the types of the s t r u c t u r a l l y
stable equilibrium state!; which a r e obtained f r o m a multiple equilibrium
s t a t e 0 (0, 0) in the case IJ = P; (0, 0) +
QI (0, 0) + 0 . It follows f r o m t h i s
theorem, in particular, that the topological s t r u c t u r e of the equilibrium
s t a t e 0 in this case is uniquely determined by i t s Poincarg index, which is
equal to the difference between the number of s t r u c t u r a l l y stable nodes
and the number of s t r u c t u r a l l y stable saddle points obtained f r o m the
multiple equilibrium s t a t e .
3. The c h a r a c t e r of the structurally s t a b l e equilibrium

s t a t e s obtained from a multiple equilibrium s t a t e with
o=o
In our proof of the basic t h e o r e m s of this subsection, w e will make u s e

of one proposition which belongs to Poincare' ( / 1 5 / , p . 43, Theorem V ) and
is a l s o of independent i n t e r e s t .
T h e o r e m 36 (Poincare t h e m e m ) . If a dynamic systein
has only simple equilibrium states and if the isocline P (3,y) = o(m
Q (2,y) = 0) has no singular points (Le., points at which both partial
derivatives P; and PI vaxish simultaneously), the equilibrium states with
A < 0, Le., saddle points, altemzate on this isocline with equilibrium
states with 6 > 0, Le., ntnies and foci.
P r o o f . Let O1 (q,y,) and O2(z2,yz) be two simple equilibrium s t a t e s of
( A ) and I a simple a r c of the curve P (2,y) = 0 between these points, which
contains no other equilibrium s t a t e s except the end points. We have to
show that A (0,)and A (02) are of different signs. Suppose that this is not
=-
so and l e t A (01)> 0 and A (0,) 0, say.
This number is defined before Lemma 1
227
Consider the v e c t o r s n ( M ) = n ( x , y) with the coordinates P k ( z , y), PI ( s , ~,)

where M (z,y) is a point of the c u r v e P (z,y) = 0. These v e c t o r s f o r m a
continuous field of n o r m a l s on the a r c I , and no vector of this field is z e r o
(i.e., the field h a s no s i n g u l a r i t i e s on the a r c I ) . Consider the function
Q (2. y) = Q ( M ) and i t s gradient grad Q ( M ) . Since 01 and O2 a r e equilibrium
s t a t e s of (A), we have Q (0,) = Q (0,)= 0 . F u r t h e r m o r e , the conditions
A, > 0 and A 2 > 0 c l e a r l y indicate that the vector n (Oi) makes a positive angle
with the vector gradQ(Oi), i = l , 2 ( F i g u r e 111).
FIGURE 111 I
To fix i d e a s , suppose that the tangent OITl a t the point Oi of the c u r v e I ,
corresponding to i t s direction f r o m Ol to 02,m a k e s a p o s i t i v e angle
with the n o r m a l n (0,). Then the tangent 0 2 T 2at the point O2 of the c u r v e 1 ,
corresponding to the d i r e c t i o n f r o m O2to O,, evidently m a k e s a n e g a t i v e
angle with the n o r m a l n (02).Computing the derivative of the function
Q (5, y) along the c u r v e I in the direction 0102 a t the point O1 (0,)and
10201)
r e m e m b e r i n g that this derivative is equal to the projection of the gradient
of Q (5, y) on the corresponding tangent, we find that this derivative is
negative a t the point0,and positive a t the point 02. This r e s u l t , combined
with the r e l a t i o n s
Q (01)= Q ( 0 2 ) = O
and with the definition of a derivative along a n arc shows that the function
Q ( x , y) is negative on the a r c I n e a r the point Ol and positive on this a r c
n e a r the point 0,. But then Q ( x , y)=O a t s o m e i n t e r i o r point of the a r c 1,
i . e . , this point is a n equilibrium s t a t e of (A), c o n t r a r y to the original
assumption. Q. E . D.
Let u s now consider the equilibrium s t a t e 0 (0, 0) a s s u m i n g that
+
I P; (0, 0) I I P; (0, 0) I tlQ;
(0, 0) I +I Q;(o, 0) I
f0 (7)
The problem is much m o r e difficult now than i n the previous c a s e , and

i t entails a number of s p e c i a l a l g e b r a i c propositions. We will t h e r e f o r e
l i s t a number of relevant r e s u l t s without proof. Detailed proof w i l l be
found in 1161.
The complications a r e a s s o c i a t e d with two f a c t o r s . First, if a singular
point 0 decomposes into s t r u c t u r a l l y s t a b l e equilibrium s t a t e s O1 ( i =
= 1, 2, . . ., k), A (0,)and u (Oi) m a y take any a r b i t r a r i l y s m a l l v a l u e s . In
p a r t i c u l a r , the difference u (0,) - 4 A (0,) may be e i t h e r positive o r negative.
228
r h i s m e a n s that the paints 0,may i n general include s t r u c t u r a l l y stable foc
and not only s t r u c t u r a l l y stable saddle points of nodes. Second, the
Poincare' index of the equilibrium s t a t e 0 no longer d e t e r m i n e s the topologi
cal s t r u c t u r e of this equilibrium s t a t e (as i t w a s in the previous c a s e ) .
Indeed, by proposition I1 of the previous subsection, if the multiplicity r of
O i s odd, this point may be
a ) a topological saddle point ( E = 0. H = 4 ) ;
b j a topological node ( E = 0, II = 0);
c ) a n equilibrium s t a t e with a n elliptical region ( E = 1, H = 1);
d ) a focus or a c e n t e r ( E = 0. H = 0).
If the miltiplicity r of 0 is even, the equilibrium s t a t e 0 niay be
e ) a degenerate equilibrium s t a t e ( H = 2 , E = 0 ) ;
f ) a saddle -node (II : 3, E = 0) .
By Bendixson's formula, the Poincard index of the point 0 h a s the
following values:
I - --I in case a;
I = + I in cases b, c , d;
I = u in cases e , f . _ii
The topological s t r u c t u r e of the equilibrium s t a t e is therefore completely

determined by the Poincarg index only i f i t is equal to -1, i.e., 0 is
a saddle point. The c h a r a c t e r i s t i c of the point 0 in t e r m s of the component
s t r u c t u r a l l y stable equilibrium s t a t e s f o r c a s e a is the s a m e a s i n
Theorem 35, provided we do not distinguish between s t r u c t u r a l l y stable
nodes and s t r u c t u r a l l y stable foci, i.e., between s t r u c t u r a l l y stable
equilibrium s t a t e s w i t h A > O . * :'' Using T h e o r e m s 32 and 34, w e obtain
the following r e s u l t .
T h e o r e m 37. A tnultiple equilibriurn state 0 ( 0 . 0 ) of system (A) fm
which condition (7)i s satisfied but u = P; (0, 0) Q&(0. 0) = 0 is a
topological saddle point if and only if the number of stmchtrally stable
nodes and foci into which i t decomposes i s 1 l e s s than the number of
structurally stable saddle point. The total number It of stmcturally
stable equilibrium states into which o decomposes may be atzy positive
odd number not exceeding r , where r i s the multiplicity of 0.
For the sake of com?leteness, the next s t e p in o u r a n a l y s i s should have
been a i m e d a t determining whether c e r t a i n relations e x i s t between the
number of s t r u c t u r a l l y stable nodes and the number of s t r u c t u r a l l y stable
foci or each number niay take any value between 0 3nd 9. W e w i l l not
consider this problem, however.
In cases b, c , d, a s i t follows f r o m ( 2 8 ) and f r o m the previous r e s u l t s ,
the number of nodes and foci among the s t r u c t u r a l l y stable equilibrium
s t a t e s O,, Oz, . . ., OR is I g r e a t e r than the number of saddle points, and in
cases e and f the number of nodes and foci is equal to the number of
saddle points .+ In t e r m s of decomposition into s t r u c t u r a l l y stable
As for system (16). the Poincars index of the equilibrium state 0 o i system (24) can be readily computed
without resorting to Bendixson's formula. See footnote on p.226.
* * Note t h a t f r o m a p u r e topological point of view, no such distinction is possible, since the node and the
focus have an identical topological structure.
* We recall that O,, 02, . . ., 0, are the structurally stable equilibrium states into which the muItiple
equilibrium state 0 decompose;.
229
equilibrium s t a t e s , we can thus distinguish between points b, c , d, on the

one hand, and points e , f , on the o t h e r . In s i m i l a r t e r m s , we can
distinguish between points of type e and points of type f , and a l s o fully
c h a r a c t e r i z e each of the types b, c , d. However, the resulting classification
is not pure topological, since we wish to differentiate between s t r u c t u r a l l y
stable nodes and s t r u c t u r a l l y stable foci obtained f r o m the multiple point 0.
Moreover, we a r e forced to consider in t h i s c a s e the decomposition into
the m a x i m u m number of s t r u c t u r a l l y stable equilibrium s t a t e s (equal to
the multiplicity of the original equilibrium s t a t e ) , and not into a n y
p o s s i b l e n u m b e r , as-before. To emphasize the l a s t factor, we w i l l
r e f e r to feasible s y s t e m (A) as a s p 1i t s y s t e m if on passing f r o m (A)
to (A) the singular point 0 of multiplicity r decomposes into r s t r u c t u r a l l y
stable equilibrium s t a t e s .
Let u s f i r s t formulate a r e s u l t relating to a quadruple equilibrium s t a t e
( I = 0, c a s e s e and f ) .
T h e o r e m 38. Let ( A )be a dynamic system, 0 (0, 0) a multiple
equilibrium state of the system of multiplicity r = 2m, m > 1 , for which
+
condition ( 7 ) is satisfied and 0 = P; (0, 0) Q; (0.0) = 0 (Le., a degenerate
equilibrium state m a saddle-node) . Then
1) If the equilibrium state 0 decomposes into k scrUcttmally stable
equilibrium states O i ,i = 1, 2, . . ., k,on passing to a feasible system, k is
even and the number of s h c t u r a l l y stable saddle paints among 0,is equal to
the number of s h c t u r a l l y stable nodes and foci.
2) If 0 i s a degenerate equilibrium state, there exist split systems
arbitrarily close to (A) for which Odecomposes only into structurally foci
and saddle points.
3) If 0 is a saddle-node, the structurally stable equilibrium states into
which 0 decomposes on passing to any suff.:ciently close split system
include at least one stnrcturally stable node.
Proof .
The validity of the f i r s t proposition follows f r o m I'heorem 32
and equation (28) and h a s i n fact been established before.
Let u s prove the second proposition. Without l o s s of generality, we
may a s s u m e ( s e e S3.2, L e m m a 2 ) that s y s t e m (A) h a s the f o r m (25):
dz
= Y = P (2.Y), $=aza"' 11 + h +
(31 bxn:^a,tl+g(41+Y'f (5, Y) = Q (z, Y),
where h ( z ) , g ( r ) , f ( x , y) a r e analytical functions, h ( O ) = g ( O ) = O . m > l ; a > O ;

n > l if b # O . Let O(0, 0) be a degenerate equilibrium s t a t e . In QT (522.2,
Theorem 6 7 ) i t is proved that i n t h i s c a s e
e i t h e r b = 0,
or b # 0 and n > m . (29)
Let, f i r s t , b+O. We take some positive number q>Oand a sequence

of a r b i t r a r y numbers xt, i = i . 2, .... n-I, such that
0 <Xi <2, < . . . Xn-i< '1. (30)
W e draw up a polynomial
b ( s )= b x ( z - - z 1 ) ( x - z 2 ) . . . (s-zn-J = b ~ ~ + b , s " - ~.+. +bn-15,
. (31)
whose r o o t s a r e the numbers 0, xl, r2, .... xn-!.
230
5 ?3. DECO!.IPOSLTION OF A AlL'LTIPLE EQUILIBRIUA! STATE INTO S TRVCTL'R.4LLY STABLE STATES
-
By (291, m - I .:n-l. Let y,, &. . . ., x , be any n u m b e r s satisfying the
inequalities
- - -
0 < s1<2,- :
2-2 <x* < .. . <x!n-t <z m - 1 < Z,-l< gm<q. (32)
W e now construct a polynomial

- -
a (T)=UZ(Z--IJ. . . ( r - z m . l)(z-z,) ... (r-rm)= ~ ~ 2 ~ t . . 2. +uzm-lz,
~-1- t z ~ - ~ (33)
whose roots are the numbers 0, si. rj. Clearly, if q is sufficiently s m a l l ,

the coefficients bi and aj may be made as s m a l l as d e s i r e d .
Consider a s y s t e m
We take q > O to be so s m a l l that the following conditions a r e satisfied:

(a) r l s e o ;
( b ) (A) is a feasible s y s t e m ; (34)
(c) if I z l < q , then i + h ( z ) > O .
All the equilibrium s t a t e s of s y s t e m (A) lie on the azis Ox. F r o m ( 3 3 )
(0)are
and ( 3 4 . c) it follows that a l l the equilibrium s t a t e s of (A) lying in U,,
- _
O(O,O), O i ( x i , O ) , i : = L 2 ..., m - f , and Oj(zj,O), j = l . 2 , . . ., m. (35)
System ( K ) thus h a s 2 m equilibrium states i n But Uq (0) c Ue,(0)

by ( 3 4 , a). Therefore ( A ) is a split s y s t e m , and by L e m m a 1 a l l the u, (0)
equilibrium s t a t e s ( 3 5 ) a r e simple. Let u s identify t h e i r c h a r a c t e r . To
this end, w e w i l l compute the values of A and (5 f o r each of these states.-
Direct computations show that if O* (z*, 0) is a n equilibrium s t a t e of (A),
then
and
u (0')= g
b (T', 0) b (z*)[f + g (z')].
L (37)
' is one of the points in ( 3 5 ) , w e have a ( x * ) = O and

If 0
A (O*) [ 1 + h (z')].
= -O' (i) (38)
Computing a' (z), w e i n s e r t f o r z* the a b s c i s s a s of the points in ( 3 5 ) ,

and using the inequalities ( 3 2 ) , ( 3 4 , c ) and the condition a > 0, w e obtain
f r o m the l a s t e x p r e s s i o n
A(0)>0; A(Oi)>O, i=i,2, ..., m - - l ;

(39)
A(Dj)<:O, j=l,2. ..., m.
23 I
I
F u r t h e r m o r e , since z=O and z 5 x i r i = 1 , 2, ..., m - I , a r e r o o t s of the

polynomial b ( Z ) , we s e e from ( 3 7 ) that
u(O)=O, o(zi)=0, i = l , 2, ..., m-I. (40)
Relations ( 3 9 ) a n d ( 4 0 ) s h o w t h a t the points a,, j = 1, 2, . . ., m , a r e

s t r u c t u r a l l y stable saddle points, and the points 0 a n d o t , i = 2 , 2, . . .,
. . ., m - 1 , a r e m u 1 t i p l e foci of (A). In accordance with the standard
points, whereas Oand O i a r e s t r u c & r a l l y stakle foci. C l e a r l y , (8)is a

split s y s t e m and, i f q is sufficiently s m a l l , (A) is a s close a s d e s i r e d to
(A). This proves the second proposition of the theorem f o r the c a s e b + 0.
z,,x 2 , . . ., z,,-~ can be taken as any positive numbers s m a l l e r than 7 .

We now proceed to the third proposition of the theorem. Let the
equilibrium s t a t e 0 (0, 0) of s y s t e m (A) be a saddle-node. By Qr, 522.2,
Theorem 67, we have in this c a s e
b#O and I g n c m . (41)
We w i l l divide the proof into two p a r t s .

a ) Consider a s y s t e m
dt
Z'Y,
%-=O(Z, y)=Q(.Z, Y ) f 4 ( Z , Y)== (A1

=a~'"' +h ( ~ ) l +bx"Y ti +g(41 + y s f (xt Y) + q ,I( u),
where b =# 0 and l < n < m . We will f i r s t prove that if (A) is sufficiently c l o s e

to r a n k 2 - 1 to s y s t e m ( A ) and h a s i n Up, (0) p r e c i s e l y 2 m equilibrium
s t a t e s , which a r e a l l s t r u c t u r a l l y stable, then a t l e a s t one of these
equilibrium s t a t e s is a node.
Evidently, c l o s e n e s s to rank 2 m - 1 of (A) and (A) indicates that the
function q (5,I/) is c l o s e to rank 2 m - 1 to z e r o .
Let Ot ( z i , 0), i = 1, 2 , . . ., 2m, be theequilibrium s t a t e s of (A) in U,, (0).
Consider the values of A, u, and IC = us-4Acorresponding to each of these
equilibrium s t a t e s . Clearly,
A ( x , 0)= -@%(z, O)=--q;(x, 0 ) - 2 m a ~ ~ ~ - ' [ 1 + h ( r ) ] - - ~ ~ h ' ( s ) , (42)
u (Z,0)= 6;(I,0 ) = q ; (Z, 0) + bz" I1 +gWI. (43)
It follows from these r e l a t i o n s that
IC (z)= u'- 4 A = b'xZn +'pi +'pz

(Z) (Z) = bax2"+CP (2). (44)
where
232
and
F r o m n < m i t follo.Ns that Z n < 2 m - - l . Since, m o r e o v e r , g ( O ) = O , the

function 'pl (z) may be written i n the f o r m
'p, (t)= z2n+Gj1 (3).
,where 'pl (s)is a n analytical function. Consequently, t h e r e e x i s t s a n u m b e r

e t satisfying the following conditions:
iaj O < e t
1
<eo;
( b / 6 2 s ~ " + ~ 1 ( z ) > 0f o r z - e i and f o r z = - E ~ ;
(47)
(h)!
ba
(c ) I ~ ; z n (2)
) 1< f o r -eI <J <e2.
K'e now choose a number ai. U < A , < d o ,
satisfying the following condition:
i f the function q ( z , y) is & - c l o s e to r a n k 2m--1 to z e r o , then
I
1 ) ?.(-,)>O, i*(E,):'0;
2 ) p<?nn, (Sn)! bz
, (2) <- for --E~<Z--IE~; (481
3 ) all the equilibrium s t a t e s Oi, i = I,?, . . ., 2m, lie in L T e , ( 0 ) .
Condition (48,l)c a n be satisfied i n v i r t u e of (44)and (47,b). Condition
(48,2)holds t r u e f o r a sufficiently s m a l l 6 , in virtue of (46)and the
inequality 2n +
1 .<2m-- 1. Finally, (48, 3) is automatically satisfied f o r a
sufficiently s m a l l ~ 5 ~ .
Let now (A) b e 6,-close to (A) to r a n k 2 m - 1, i.e., the function q ( J , y) is
&-close to z e r o to r a c k 2 m - 1. Consider the function
X (t)= b2X2" 4- 'p (t) (45)
on the segment [--si. ei].

F r o m the equality
'F (4= 'FI (4f 'F?(4 (50)
and f r o m (47,~)
and (48,2) it follows that on this s e g m e n t
I 'p'*"' I <(Zn)!ba.
(I) (51)
The function h (z)thorefore h a s a t m o s t 2 n r o o t s on the s e g m e n t [ - ~ ~ 1 . .

Let El, S t , . . ., E k be a l l the different roots of L (z) on this segment (some of
them may be multiple roots). C o n s i d e r the k + 1 i n t e r v a l s into which the
r o o t s Ejdivide the segment I- E , , E , ) . In e a c h of t h e s e i n t e r v a l s , the
function ?.(z) r e t a i n s a constant s i g n . By (48,1), h (2)> 0 i n the f i r s t and
the l a s t of these i n t e r v a l s . The i n t e r v a l s where I (I)< 0 w i l l be called
negative i n t e r v a l s and designated JI. J 2 , . . ., J l ( F i g u r e 1 1 2 ) . We .will now
prove that L < n . Indeed, both e n d s of ea'ch negative i n t e r v a l are r o o t s of
the function X (2). Lf the i n t e r v a l s J, and J,,, have a c o m m o n end point 5,
' If the function has N root: on some segment (counting their multiplicities), itc derivatire has at least
,J -1 roots. This iollor*.s irom the Eolle theorem and from the fact that e a c h multipie root o fa function
IS a rOat of the derivative of inultiplicity smaller than 1.
233
Ch.VIII. BIFURCATIONS OF DYNAMIC SYSTEMS
(e.g., the i n t e r v a l s J z and J 3 in F i g u r e 1 1 2 ) , &* is a root of even multiplicityof

h ( z ) , i.e., i t h a s a multiplicity of a t l e a s t 2. Therefore, the total number of
r o o t s of the function h (2)on I- e , , ell is a t l e a s t 21. Since, on the other hand,
the number of r o o t s is a t most 2n, we s e e that l ~ n .Then by (41)
l<m. (52)
All the equilibrium s t a t e s O,,i = 1, 2, . . ., 2m,lie i n U,,(O)(see (48)) and
a r e s t r u c t u r a l l y s t a b l e and t h e r e f o r e s i m p l e . If Oi ( s i . 0 ) l i e s in one of the
negative i n t e r v a l s , A (0,) = A, > 0 , since otherwise I (q)= hi = 03 - 4 A , > 0.
But then by Theorem 36 (PoincarB theorem), each negative i n t e r v a l J , con-
tains a t m o s t one point O i . This signifies, as we s e e f r o m inequality (52).
that the number of equilibrium s t a t e s Oi with I < 0, i.e ., the number of
foci, is less than m. On the o t h e r hand, the f i r s t proposition of the theorem
indicates that the total number of nodes and foci among the points 0,is
equal to the number of saddle points m. Hence, t h e r e is a t least oneAnode
among the points O,, which p r o v e s the third proposition f o r s y s t e m (A).
FIGURE 112
b ) Let u s now c o n s i d e r the g e n e r a l c a s e . Let
be a dynamic s y s t e m 6-close to r a n k 2 m to s y s t e m (A) which h a s in U,, (0)

p r e c i s e l y 2 m equilibrium s t a t e s , all of which a r e s t r u c t u r a l l y stable. We
have to prove that if 8 i s sufficiently s m a l l , t h e s e equilibrium s t a t e s
contain a t l e a s t one node. Let the equilibrium s t a t e s be Oi, 02, . . ., Ozm.
Consider the transformation
We a s s u m e that the region 3 in which all the s y s t e m s a r e t r e a t e d is convex

in y'* and that 6 < 1. Then, as is r e a d i l y s e e n , the mapping ( 5 3 ) is one-
to-one and r e g u l a r , moving into s o m e region in the plane (X,Y).
Indeed, if two different pcints (sl, y,) and ( x 2 , gz) are mapped by (53) onto the
+
s a m e point, then rl = s 2 , yl # g2 and y1 p (z,.11~)= ya + g Z ) . But then gz - y, =
p (ti,
= p (xi,yl)- p (5,. YZ) = (YI - V Z ) PI (51,y*), where 11,c y* < y2 (or y, > y* > yz).
T h e r e f o r e . / p; (2, y*) I = I , which c z n t r a d i c t s the condition 6 < 1.
Transformation (53) changes (A) into the s y s t e m
-_
dX --Y,
dt
%=Q(X, Y).
* Convexity in y indicates that if the end points of a segment parallel to the axis Oy l i e in G , the entire
segment lies in G.
234
f 23. DECOllPOSITION OF .4 :.!LLTIPLE EQUILIBRILhl ST.4TE INTO STRUCTLR.4LLY ST.4BLE ST.4TES
Let u s consider X and Y a s coordinates in the plane (t,y). i . e . , w e

r e p l a c e A, Y with 2, y, respectively. This gives a s y s t e m
dz (A )
;iT=y. $=O(t, y),
defined in E .
Transformation ( 5 3 ) is 6-close to rank 2m to the identity transformation
.V L t, Y = y ( s e e $3.2). Let E , be a closed region, such that G 13,3 Cco(0).
If 6 is sufficiently s m a l l , w e see by L e m m a 1 , 2.2 that s y s t e m (A) is
defined in F , and is a r b i t r a r i l y c l o s e to s y s t e m (A), and hence to
s y s t e m (A), t o r a n k 2 m, - 1 . To e v e r y equilibrium s t a t e O,,- i =
= 1, 3 , . . .. 2m,of systern ( A ) c o r r e s p o n d s a n equilibrium s t a t e 0, of
s y s t e m (A). Since ( 5 3 ) is a r e g u l a r mapping, 0, are s t r u c t u r a l l y s t a b l e
equilibrium s t a t e s by L e m m a 2, $6.1, and the corresponding p a i r s 0, and
0, are e i t h e r both nodes, or both foci, o r both saddle points. Finally, f o r
a sufficiently s m a l l 6, all the equilibrium s t a t e s 0, Of s y s t e m (A) lie in
Cre, (0) and t h e r e are 30 other equilibrium s t a t e s of (A) in this neighborhood.
But then s y s t e m (A) s a t i s f i e s all the conditions of proposition a (see
p. 2 3 2 ) , i . e., among that points O,, and therefore among O r , t h e r e is a t
l e a s t one s t r u c t u r a l l y stable node. This completes the proof of the
theorem.
In T h e o r e m s 37 and 38 w e considered cases when the P o i n c a r e index of
the original equilibrium s t a t e 0 (0. 0) is -1 o r 0. L e t u s now proceed to the
l a s t c a s e , when the Po-ncark index is f 1. The equilibrium s t a t e 0 (0, 0) of
s y s t e m (A), as w e have noted before (see (28)), is then one of the following:
b ) a topological node;
c ) a n equilibrium s t a t e with a n elliptical region;
d ) a focus or a c e n t e r (see p. 229).
In $23.2 w e s a w that the relevant s y s t e m c a n be reduced to the f o r m (25):
It is established i n QT, Chapter IX, 522.2, T h e o r e m s 6 6 and 67, that

c a s e s b, c , d obtain if a < 0 and r is a n odd number. W e m a y thus take
s y s t e m ( A ) in the f o r m
dt
-
d=t y, 2 = u32m*111 f h (41T b2Y 11 + g ( t j l f Y*/ (J, !A, (54)
where a < 0, m > 1, h , g , and f are analytical functions, and h (0) = f (0) = 0.
Then (see QT, Theorem 6 6 )
b ) 0 (0. 0) is a topological node if
b#0, n is even, and n < m (55)
o r if
b+O, n i s even, n = = m . and D = b 2 + 4 ( m + l ) a > O ;
c ) O ( 0 , 0) is a n equilibrium s t a t e with a n elliptical region i f
b+O, n is odd, n < m
235
or if
bfO, n is odd, n = m , and D = b 2 + 4 ( m + 1 ) a > 0 ;
d ) O ( 0 , 0) is a focus or a c e n t e r if
b=O,
or i f
b g O a n d n>m,
or if
The next theorem c h a r a c t e r i z e s - in t e r m s of decomposition into

s t r u c t u r a l l y s t a b l e equilibrium s t a t e s - the difference between c a s e s b
and c , on the one hand, and c a s e d, on the o t h e r . This classification,
however, is formulated f o r the c a s e when s y s t e m (A) is given in the
f o r m (54) and i f n = m , then D = ba + +
4 (m 1) a # 0 . We will show s o m e -
what l a t e r ( s e e r e m a r k to Theorem 3 9 ) that the l a s t condition is
fundamentally not r e s t r i c t i v e .
T h e o r e m 39. 1) If on passing from system (54) to a feasible system
the equilibrium state 0 (0,O) decomposes into k structurally stable equili-
brium states O i ,i = 1, 2, . . ., k , k i s odd and the number of structurally
stable saddle points among 0,is 1 less than the number of structurally
stable nodes and foci.
2) If 0 (0, 0 ) i s a topological node or an equilibrium state with an
elliptical region and if for n = m, D = ba + 4 (m f 1) a # 0 , the structurally
stable equilibrium states into which 0 decomposes on passing to a
sufficiently close split systerri include at least one structurally stable node.
3} If 0 (0, 0) is a multiplefocus or center, there exist split systems
arbitrarily close to (54) on passing to which 0 decomposes into structurally
stable foci and saddle points only.
P r o o f . The f i r s t proposition of the theorem follows d i r e c t l y f r o m the
fact that the Poincare' index is $. 1 in c a s e s b, c, d ( s e e (28)). The second
and the third proposition f o r m f n a r e proved p r e c i s e l y i n the s a m e way
a s the corresponding propositions of Theorem 3 8 . If, however, m = n , the
proof is m o r e complicated, and is omitted here.:.
R e m a r k . E n = m , and D = ba + 4 ( m + 1 ) a = 0 , t h e r e exist s y s t e m s
of the s a m e type, a r b i t r a r i l y c l o s e to (54), f o r which D > Oand D < 0. T h e r e -
fore, no c r i t e r i o n is applicable in this c a s e which would differentiate -
in t e r m s of decomposition into s t r u c t u r a l l y stable equilibrium s t a t e s -
between a node o r a n equilibrium s t a t e with a n elliptical region, on the
one hand, and a focus or c e n t e r , on the o t h e r .
For the sake of completeness, we r e q u i r e s t i l l another c r i t e r i o n , which
would differentiate - in t e r m s of decomposition into s t r u c t u r a l l y stable
equilibrium s t a t e s - between c a s e b and c a s e c, i.e., between a node and
an equilibrium s t a t e with a n elliptical region. One such c r i t e r i o n h a s been
* A number of special algebraic lemmas have to be used in this case. See /16/, Theorem 5.
236
5 '23. DECOhIPOSITION OF .A hlL'LTIPLE EQL'ILIBRIL'hl ST.ATE IKTO STRUCTUR.ALLY ST.4BLE STATES
established i n /16/ (p. 55, Theorem 6 ) . It m a k e s use, in p a r t i c u l a r , of the

c h a r a c t e r of the stability of the s t r u c t u r a l l y stable nodes and foci into
which the multiple equilibrium s t a t e 0 decomposes.
This c r i t e r i o n , howtwer, employs essentially nontopological concepts,
and i t is therefore inadequate f r o m the topological point of view adopted in
o u r t r e a t m e n t . The question of a satisfactory (topological) c r i t e r i o n of this
kind thus r e m a i n s open.
In conclusion of this section, note that o u r r e s u l t s indicate that the type
of a multiple equilibrium s t a t e 0 (0, 0) of s y s t e m ( A ) satisfying conditions ( 7 )
and ( 2 2 ) {i.e., when u = P ; + Q; = 0 ) is not determined in g e n e r a l by the
topological s t r u c t u r e and the number of s t r u c t c r a l l y stable points into which
the multiple point decomposes. To obtain a full c h a r a c t e r i s t i c of points of
types b, c , d, e, f (p. 229), w e have to consider the (non-topological) difference
between nodes and foci. If in (54), n = m and D = b2 f 4 (m + f ) u = 0,
cases b and c cannot be distinguished f r o m case d by considering the
s t r u c t u r a l l y stable points obtained f r o m the multiple point.
237
C h a p t e r IX
C R E A T I O N OF L I M I T C Y C L E S FROM A
M U L T I P L E FOCUS
IN TRODUCTION
In the previous chapter w e considered bifurcations of a multiple

equilibrium state, concentrating only on the number and the c h a r a c t e r of
the s t r u c t u r a l l y stable equilibrium s t a t e s into which a multiple state may
I
decompose. In Chapter IX we w i l l consider a simple equilibrium state, namely
a s t r u c t u r a l l y unstable focus (A # 0, pure imaginary c h a r a c t e r i s t i c n u m b e r s )
and determine the number of limit cycles which may be c r e a t e d in i t s
neighborhood on passing to close s y s t e m s . Once this problem is solved, w e
w i l l be able to d e s c r i b e the possible bifurcations of a dynamic s y s t e m in the
neighborhood of a s t r u c t u r a l l y unstable focus.
The chapter is divided into two sections. In 124, various auxiliary
propositions are considered. It studies i n m o r e detail the p r o p e r t i e s of the
s u c c e s s i o n f u n c t i o n , and a l s o defines the leading concepts of f o c a l
v a l u e s and m u l t i p l i c i t y o f a m u l t i p l e f o c u s . The succession
function w a s originally defined in l o : l e t 0 be a focus, 1 a r a y issuing
f r o m the focus, M o a point on the r a y , sufficiently close to 0, L a path
( s p i r a l ) through M o r M ithe next (after M 0 . in t e r m s of increasing t ) i n t e r -
section point of the path L and r a y 1 . Let O M o = po, O M , = pi ( F i g u r e 2 6 ,
p. 91). The function pi = f (po) is called the succession function on the r a y 1 .
The f o c a 1 v a 1u e s a r e defined a s the values of the derivative of the
function
d (Po) = f (Po)-Po
a t the point po = 0.
In J24 we prove that if n e x i s t s such that
d (0) = d (0) = ...= dn-1 (0)=0, d (0).f 0,
then n i s a n odd number ( L e m m a 5 ) . In this c a s e

k,!?l!
2
is called the m u 1t i p 1i c i t y of the focus 0. A s i m p l e focus has multipli-

city zero ( n = I), whereas a multiple focus either has no definite multipli-
city, o r its multiplicity is k > i . A number of f o r m u l a s a r e derived in $24.3
which c a n be applied to compute the focal values. All these formulas lean
238
$24. FOC.-\L VALL'ES
on the proof of the fundamental theorem of Chapter IX. E x p r e s s i o n s for the

focal values of a n analytical s y s t e m are d e r i v e d in 5 2 4 . 4 . T h e s e e x p r e s s i o n s
are useful i n t r e a t m e n t of p a r t i c u l a r s y s t e m s .
The fundamental theorem of this c h a p t e r is proved in 5 2 5 - the theorem
of the c r e a t i o n of l i m i t c y c l e s from a multiple focus (Theorem 4 0 ) . Its
formulation is v e r y s i m p l e : i f 0 (0,O) is a multiple focus of multiplicity
k ; > i of a dynamic s y s t e m (A), s y s t e m s (A) sufficiently c l o s e to ( A ) to rank
2k + 1 c a n have a t most k c l o s e d paths in a sufficiently _small neighborhood
of the focus. On the o t h e r hand, t h e r e e x i s t s y s t e m s (A) a s close a s d e s i r e d
to (A) \to rank 2k f 1 ) which have p r e c i s e l y k c l o s e d paths i n a n a r b i t r a r i l y
s m a l l neighborhood of :he focus. Thus a k-tuple focus may c r e a t e k, but
no m o r e than k, l i m i t c:ycles.
Theorem 4 1 e s t a b l i s h e s that for any s. 1 , < s < k , t h e r e e x i s t bifurcations i n
ivhich p r e c i s e l y s l i m i t c y c l e s a r e c r e a t e d f r o m a k-tuple focus.
T h e o r e m s 4 0 and 4 1 show that a dynamic s y s t e m may only have a finite
number of different bifurcations in the neighborhood of a focus of finite
multiplicity. In 5 2 5 . 2 a classification of these bifurcations is given.
At the end of the cha.pter ( 1 2 5 . 3 ) we c o n s i d e r one p a r t i c u l a r case often
encountered in applications, namely a s y s t e m dependent on a single p a r a -
m e t e r and i t s bifurcations in the neighborhood of a multiple focus of
multiplicity 1 when the p a r a m e t e r is v a r i e d . \Vhen the s y s t e m crosses the
bifurcation value of the p a r a m e t e r , the stability of the focus changes and
a l i m i t c y c l e is c r e a t e d , o r alternatively a n existing l i m i t cycle "contracts"
into the focus.
S24. FOCAL 1'ALUES
1. S o m e p r o p e r t i e s of the s u c c e s s i o n function
The contents of the F r e s e n t c h a p t e r is d i r e c t l y r e l a t e d to l o ,

S t r u c t u r a l I n s t a b i l i t y of a n E q u i l i b r i u m S t a t e w i t h P u r e
I m a g i n a r y C h a r a c t e r i s t i c R o o t s . W . e will t h e r e f o r e u s e the
notation and the concepts introduced i n 510, and s o m e of the p r e v i o u s
r e s u l t s . \ V e will c o n s i d e r a s y s t e m of class N which h a s the following
canonical f o r m in the neighborhood of a n equilibrium s t a t e 0 (0, 0) with
p u r e imaginary c h a r a c t e r i s t i c roots:
where p > O . This s y s t e m is a p a r t i c u l a r case of the s y s t e m
The functions 'p and 9 are d i s c u s s e d i n $10.1.

Changing o v e r to the p o l a r c o o r d i n a t e s p, 0, 'we f i r s t obtain the s y s t e m
239
Ch.IX. CREATION OF LIMIT CYCLES FROM A MULTIPLE F O C U S
and then the eauation
It is a s s u m e d h e r e that 0 (0, e) = Ofor all e, which e n s u r e s the continuity of

the function 0. Taking
P = f (0: 007 Po) (5)
to be the solution of equation ( 4 ) satisfying the initial condition
f (00; 00, P) = Po7 (6)
w e define the s u c c e s s i o n f u n c t i o n
P = feo (Po) = f (eo +2% 00, Po) (7)
on the r a y O = e o , and a l s o the function
de0 (Po) = feo (Po) -Po. (8)
For eo = 0, we designate these functions f (po) and d bo),respectively.

Thus,
In $10 w e considered the functions f e o ( p o ) anddeo(po)

f o r p o > O . In the
p r e s e n t section, we w i l l allow negative values of poalso. W e w i l l a s s u m e ,
however, that
I Po I < 87 (11)
where 6 is a sufficiently s m a l l positive number.

F i r s t note that equation ( 4 ) is not affected by a simultaneous substitution
of - p for p and of 0 f n f o r 8 . More precisely, if p = p (e) is a solution of
equation ( 4 ) and if p* = - p and e* = 0 -+- rc, then $$ = R (p*, e*). Indeed,
(the l a s t equality follows directly f r o m ( 3 ) and (4)). Thus if p = p (e) is a

solution of equation (4), then
L e m m a 1. The following equality ho2ds true:

deo (-p) = -deo+, (p) = -f (0, + 3n; 00+ fi, +
PI P.
240
$24. FOC.AL L.AL['ES
P roof . By definition
where r , = f ( O ; Go, -pO) is the solution of equation (4 J f o r the initial conditions

tt-ttyo. p = -Po.
Consider the solution of equation ( 1 2 ) f o r the initial conditions p* =
.-= - - t ~ - p ~ ,o * = 6 + z = O 0 + x . T h i s solution evidently h a s the f o r m
Thus for the p a r t i c u l a r initial conditions chosen
But the condition p*=po f o r 6 * = I o j x is equivalent to the condition p -po for

O = He. T h e r e f o r e , i n v i r t u e of the uniqueness t h e o r e m ,
By ( I d ) , ( l S j , (7), and (8), w e have
Substituting p f o r po, w e obtain (13). .This c o m p l e t e s the proof of the

lemma.
Geometrically, Lemtna 1 is self -evident. Indeed, l e t M o be a point with
polar coordinates (00, - ( t o ) , where p > 0 , and l e t L be a path through this
point which again c r o s s e s the r a y UJT, (as the polar angle 6 is i n c r e a s e d by
2;t) at a point .Ill with the p o l a r coordinates (0, + 2;t. - p , ) , pl>O ( F i g u r e 113).
By definition deo(- po) = .- p, - (- po) = ( i o - pl . On the o t h e r hand, Moand
.lll c a n be considered a s points with the p o l a r coordinates (0, -a, po) and
(0, f n, p i ) , respectively. But then deo+n (PO)= PI - P O = - deo (-- PO).
L e wt m a 2. If there exists r, > 0 such that fm-all p. 0 < ,o i r l . do, (p) >
> 0 (4,(p) (01, there also exists r2 > 0 such that for all p. 0 0, respectively). Therefare, for all p , 0 1 p l ~ =r
= min { r , , r 2 } ,
24 1
riiith the initial covuiitiows
P r 0 0 f . The validit,v of ( 2 0 ) follows f r o m QT, Appendix, 58.3,

r h e o r e m B". Relations (21) follow d i r e c t l y f r o m identity ( 6 ) . Q. E. D.
In Lvhat follows w e a s s u m e for simplicity that O p = 0 ( t h i s , evidently,
does not Lead to a loss i n g e n e r a l i t y j . F o r the unctions f o (po) = f (2x;0 . po)
and d o (,oo) = fo (po) - po we will u s e the r e s p e c t i v e notation f (po) and d (po) ( s e e
(9)and (10)). Kote that the function f e n t e r i n g ( 2 0 ) and (21) is a function of
t h r e e v a r i a b l e s , f (e; O,, po), whereas f (po) is a function of a single v a r i a b l e .
The r e l a t i o n between these two functions is f i r e d by (9).
Since f (0; eo. p,) h a s continuous p a r t i a l d e r i v a t i v e s with r e s p e c t to po to
order .V inclusive, the functions f (po) and d (p0) a r e continuously differentiable
.V t i m e s .
De f i IZ i t i o n 25. Tke ualue of the i-th derivatioe of the function d (pa)
at the point 0, i.e., dcij (u), is called the i - t h focal ualue of the focus 0 .
If ( 2 ) is a s y s t e m of c l a s s .V, the focal v a l u e s (0). 1 -:
ii -V> a p r i o r i
exist.
L e HI ))I a 5 . i f there exists k such that
d (0)= 0, Cr (0) = 0, . . . . d'k-" (0) = 0, d'" (0) # 0, (22)
k is an odd izuwibeu.
Proof. By ( 3 ) and (4),p E 0 is a solution of equation ( 4 ) . Therefore
f (0) = d ( 0 ) = 0. (23)
Applying M a c l a u r i n ' s formula to the function d (p,) and using r e l a t i o n s

( 2 2 ) and (23), we find
where 0 < q < 1. Therefore, i f k is even, d (p,) h a s the s a m e s i g n for all

sufficiently s m a l l po, both negative and positive ( i t s s i g n c o i n c i d e s with the
s i g n of the k-th focal value (0)). T h i s c o n t r a d i c t s L e m m a 2, however.
Thus k must be odd. Q. E. D.
D e f i n i t i o n 26. If conditions (22) are satisfied, andk = 2m i 1, m s 0 ,
ice shall say that the focus 0 (0, 0 ) is a focus of multiplicity m.
In Chapter W (10.2) i t is proved that the f i r s t focal value is
Lf m = 0, then k = 1, d (0) # 0 , a # 0. But then the f o c u s 0 (0, 0) h a s

complex, though not p u r e i m a g i n a r y , c h a r a c t e r i s t i c numbers, i.e., i t is
a s i m p l e focus. Conversely, i f m > 0 , then k > 3 , d' (0) = 0, a = 0, and the
c h a r a c t e r i s t i c n u m b e r s are p u r e i m a g i n a r y , i . e . , the f o c u s is multiple
($10.3, Definition 16). Thus for m > 1 we are dealing with a m u l t i p l e
f o c u s of m u 1t i p 1i c i t y m . Note that a multiple f o c u s d o e s not always
have a definite multiplicity. Indeed, if (1) is a s y s t e m of class -V, but not
.
of c l a s s N f 1, and i f d' (0) = d" (0)= . . =dW) (0) = 0, Definition 26 is inapplica -
ble.
243
ch.IX. CREATION OF LIMIT CYCLES FROM A MULTIPLE FOCUS
In what follows we w i l l d e a l with multiple foci only, since a s i m p l e

focus is s t r u c t u r a l l y s t a b l e and the topological s t r u c t u r e of the partition
into paths in a sufficiently s m a l l neighborhood of a simple focus does not
change on passing to c l o s e s y s t e m s .
When p > 0, the focus is s t a b l e if f o r a l l sufficiently s m a l l positive po,
d (pa) < 0, and u n s t a b 1e if d (po) > 0.':
Hence and f r o m (24) i t follows that if k = 2m +
1> 1 satisfies conditions
( 2 2 ) and p > 0, the focus 0 is
I
s t a b l e when d ( " )(0)<0, (26)
and
unstable when d'") (0) > 0. (27)
D e f i n i t i o n 27, The first (counted in the proper order) nonzero focal

value of a multiple focus is culled the LyaPunou value.
In other words, the Lyapunov value is the number d(*) (0) provided that
r e l a t i o n s ( 2 2 ) hold t r u e and k > 3 . If k = 2m f 1, the Lyapunov value w i l l a l s o
be called the m-th L y a p u n o v v a l u e ( m > l ) . * f 6
F r o m ( 2 6 ) and (27) i t follows that for p > 0, a multiple focus is s t a b l e
(unstable) i f i t s Lyapunovvalue is negative (positive).
3. Calculation of the focal v a l u e s of a multiple focus
Since we will be dealing with multiple foci only, we take in what follows
a = 0, i . e . , we will c o n s i d e r s y s t e m (1). It follows f r o m the r e s u l t s of
Chapter N (910.1, (9), and 9 1 0 . 2 , (23)) that in this c a s e
To compute the focal v a l u e s , we u s e L e m m a 4.
(the functions E , a r e defined by (20)).

F r o m (28), (29), and (30) i t follows that U i(e) = 1, Ri (e) E 0 and that
1
[ E l (e; 0,po)]pozo = (e) = Rk (e)4- . . ., k = 2, 3, . . ., N. (31)
I
The t r i p l e dots in (31) c o r r e s p o n d to a polynomial i n the functions
R~ (e), R~ (e), . . ., Rh+ (e)and the functions U z(e), . . ., u h - ' (e).
* See 5 10.1. For 0 < 0, d (po) > 0 in the stable case and d (po) < 0 in rhe unstable case.
** Logically, this term is not quite adequate, since every multiple focus has a single Lyapunov value. It is
convenient, however, in that it directly identifies the running number of the focal value which is the
Lyapunov value.
244
924. FOC;\L L'ALL'ES
Let now po = 0 i n ( 2 0 ) . We will s u c c e s s i v e l y i n t e g r a t e t h e s e equations as

f i r s t - o r d e r l i n e a r equations, using initial conditions ( 2 l ) , the r e l a t i o n s
j (0; 0, 0) = 0 and R (0, e) = 0

and the notation ( 2 9 ) , (30), ( 3 1 ) . The first of these equations h a s been
considered before ( S 1 0 . 2 ) . I t s solution is given by (28).
I
hrultiplying the k - t h equation in ( 2 0 ) (k = 2. 3. . . .. iV) by p and integrating,
we find
e
By definition, the foc:al v a l u e s are equal to the d e r i v a t i v e s of the function
d (Po) = f (2n; 0, Po) -Po
at the point po = 0. T h e r e f o r e , using ( 2 8 ) and ( 3 3 ) , we obtain f o r the focal

values
2%
d'(O)=O, d'kj(0)=k! N~(0)dd, k = 2 , 3, . . . . N. (34)
0
Let u s now e x p r e s s the focal values in t e r m s of the right-hand s i d e s P and

(2). To this end, w e f i r s t have to d e r i v e a n e x p r e s s i o n f o r
Q of s y s t e m
Rk (e). System (1) h a s the f o r m
where the functions 'p anti tp a r e continuously differentiable to o r d e r AVi n -

clusive, and t h e s e functions together with t h e i r f i r s t d e r i v a t i v e s vanish
at the point 0 (0, 0). As is known ( s e e Appendix, 2 ) , f o r any k, 2 < k <A', the
functions 'p and 9 c a n be x r i t t e n in the f o r m
'p (I,Y) =Pz(.& Y) fP3

O (I,y) = Qz 0, Y) + QJ
(z, Y)
(2,Y)
. . . +PR
- . . . f QA
(I.Y) +P*
( r ,y) i
(I,
Q* (I,
Y),
Y), I (35)
w h e r e Pi (I, y) and Qi (5,y) are homogeneous polynomials of d e g r e e i ( i =

..
= 2, 3. . . k), and
where Pz (3, y) and QZ (I, y) are continuous functions which vanish a t x = g = 0.
245
Ch.1X. CREATION OF LlMIT CYCLES FROM A MULTIPLE FOCUS
Consider the fuqction R ( p , 8). By definition (see (3) and (4))
(pcos8, psin8)cosB+$(pcos8, psin8)sinO

R (P, e) = pcp-b(pcose, p s i n o ) case- ~ @ c o s e . p s i n e )
. (37)
.
sin e
P P
By L e m m a 3, the function R(p, 8) h a s continuous p a r t i a l derivatives to o r d e r

N inclusive with r e s p e c t to p. Moreover, R(0, 8 ) = 0, [ f ] , - , = O . Theref o r e
we obtain the following relation, analogous to relations (35):
R (p, e) = Rt ( 8 ) p2-1- . . . -+ Rk ( 8 ) ph + R* (p, e) pk, (38)
where R i ( 8 ) , i = 2 , ..., k , are computed f r o m ( 3 0 ) (i.e., they are equal to
4- I5+]p=0
~ R ( P e)
)> and R * ( p , 8 ) is a continuous function of p and 8 which
vanishes a t p = 0.
Inserting the functions and $ f r o m (35) in (37), we obtain
k
2 pm [pn1 COS e i - ~ ,sin 01 + P* cos e + p s i n t)

R ( p , 0) = m=;
fJ+ p " ' - l [ Q ~ ~ ~ ~ e -P P , s iP n e ~ + ~ ~ ~ - ~
m=2
where
P, ==P , (cos 8, sin e), Qm= Qm (cos e, sin e),
P% = P& (p cos 8, p sin e), Qz = Q& (p cos 8, p sin 8).
(39)
W e write u, (cos 8, sin e) and u, (cos 8 , sin O), respectively, f o r the factor
a f t e r pm in the n u m e r a t o r and the factor a f t e r pm-' in the denominator.
Moreover, by (36),
+-
P* (p cos 0, p sin e) cos 0 Q" (p cos 8, p sin 0) sin e = phu* (p, cos 0, s i n e),
Q*(p cos 8, p sin 8 ) cos 8--P* (p cos 0, p sin 0) sin 8 = p*u* (p. cos e, sin e), (40)
where U* (p, cos 0 , sin e) and u* (p, cos 0 , sin e ) are continuous functions of p and
e which vanish a t p = 0.
F r o m ( 3 8 ) through ( 4 0 ) we find
Rz (e) pa+ . .. fHk (e) p"+ f?* ( p , 8')kp
$ p u m (cos e, s i n e ) + p k U * (p, cos e, sin 8 )

-
- m=?
. (41)
p+ 5
,r,=?
p m - l v , (cos B, sin O)+pk-lu* (p, cos e, s i n e)
Multiplying both s i d e s of (41)by the denominator of the right-hand side,

w e obtain a n equality which is valid f o r all (sufficiently s m a l l ) p, i . e . , a n
identity. All the coefficients in this identity are continuous functions of p,
and u* (0,cos 8 , sin 8) = R* (0, 8) = 0. Therefore, equating the coefficients
of the corresponding powers of p and using the usual a r g u m e n t s of
246
934. FOC.AL \'ALCES
continuity, w e obtain the following r e l a t i o n s :
I
U Z z= pR,,
~3 e fiR, f v ~ R ? .
l l k = p ~ ~ h - V : R ~ - 1 i . . . -kt'a-1R2.
I t now follows that
where
-
-R~nt-t iR .
~ m - z + .. - i - R m - ~ v ~
(44)
B
Evidently, iV,,4 is e x F r e s s i b l e in t e r m s of the functions u i ( c o s 8 . sine) and
with the indices i not exceeding n2-I.
J . , ( c o s ~s. i n e )
inserting f o r u , and u i in (43)and (44)t h e i r e x p r e s s i o n s in t e r m s of Pt and
f j , , we obtain
- P,, (cos@.sin01cosf3 +Qnl (cos 0, sin @ ) s i n 8J- wm

m-- (43)
P
(in - 2. 3. . . . , k), where W,,, is e x p r e s s e d i n t e r m s of p and in t e r m s of the
functions P,(cos 0. sin e), (ji (cos e, sin e) with indices i not exceeding ni- I.
In o u r c a s e of a multiple focus (i.e., for a = O ) , the f i r s t focal value is
z e r o . If .V>2, the second focal value d" ( 0 ) = 0 by L e m m a 5. The remaining
focal values d k ) (0) are computed f r o m (34)using e x p r e s s i o n s (45)f o r R , (0).
X c o m p a r i s o n of the v a r i o u s e x p r e s s i o n s derived above w i l l enable u s to
formulate a useful l e m m a .
( a ) F r o m (28), (291, (30), and (31) it follows that
HZ (e) = RZ (e).
( b ) F r o m (31) i t follcws that, fQr m = 3 , 4 , . .., S ,
H, (0) = R, (e)+ a,,

( R , (e), . . ., R,-, (e). U~ (e). .. ., u , - ~ (e)),
a h e r e 'b, is a polynomial i n the corresponding functions.

( c ) F r o m (33)and (b) it f o l l o w s that u z (e) is e x p r e s s i b l e i n t e r m s of
R? (e), u 3 (e) is e x p r e s s i b l e in t e r m s of R z (e) and R 3 (e), . . ., u, (0) is
e x p r e s s i b l e in t e r m s of Rs (e). R , (e), . . R , (8). ..
\ d l F r o m ( b ) and (c) it follows that
where @Ei is e x p r e s s i b l e in t e r m s of the functions Rz (e), . . ., Rm-*(e) using

a l g e b r a i c operations and integration.
247
Ch.IX. CREATION OF LIMIT CYCLES FROM A MULTIPLE FOCUS I
( e ) F r o m (45) and (d) it follows that
H, (e) = Pm(cos8, sin8)cos8+Q,(cos8,

B
sin8)sin8
+
. ., k), where @&* is expressible, using algebraic operations and
(m = 2, 3 , .
integration, in t e r m s of the number and the functions P, (cos 8, sin e),
Q,(COS 8, sin e) with indices i not exceeding m - 1.
Finally, f r o m ( 3 4 ) and (d) we have the following useful r e s u l t :
L e m m a 6. Consider a dynamic system of class N
dz
-=
& --PBy+Cp(& Y)=P(r*Y), $=Pz+Wx, y)=Q(s, Y) (A )
where the functions and $I are expressed by equations ( 3 5 ) and (361, with k
equal to one of the numbers 2 , 3, . . . ,N. Then the focaZ value dim) (0), m =
I = 2, 3, . . . ,k , may be computed f r o m the formula
d,m' (o) = m!
2n
P, (cos 8 , sin 8 ) cos 8 + Q , (cos 0, sin 0) sin 8
2%
Before proceeding with the fundamental theorem of this chapter, we will

apply the above r e s u l t s , and in p a r t i c u l a r equation (46), to compute the
Lyapunov value of the focus 0 (0,0) of a modified s y s t e m of one p a r t i c u l a r
form.
Let (A) be a dynamic s y s t e m of c l a s s N (see %bow), and s an integer
satisfying the inequalities I g s , 2s +
I 6 N . Let (A,) be the dynamic s y s t e m
where is a p a r a m e t e r . Let 2 (Po) be th_e analog of d (Po) f o r The point (x8). I

0 (0, 0) is evidently a multiple focus of (As) and d (0) = d (0) = 0.
L e m m a 7. If all the focal values of the focus 0 (0,O) of system ( A ) ,
+
up to (2s 1) -th inclusive, are zero, i.e., if
d' (0)= d (0) = . .. = d'as+l' (0) = 0, (47)
the focal values of the focus O ( 0 , 0) of system ( A , ) up to ( 2 s ) - t h inclusive

are also zero, i.e.,
I and the ( 2 ~ +1) -th focal

2 (0) = d" (0) = . . . = &*)
value is
(0) = 0, (48)
Zza+1)(0)=(2s+ I)[ $-.2n. (49)
P r o o f . We e x p r e s s the right-hand s i d e s of (6,)using f o r m u l a s

analogous to (35) i n the f o r m
$24. FOCAL VALIIES
Evidently,
From ( 5 1 ) and ( 4 6 ) vie conclude that f o r 2 < i < 2 s , E l ) (0) = d(')(O), i . e . ,

&c)(u) = 0 _by ( 4 7 ) . The first focal value is 2 (0) = 0 , since 0 is a multiple
focus of {Asj. Equalities ( 4 8 ) a r e thus proved.
To compute the ( ? s i ? ) -th focal value, we w i l l u s e equations (46), ( 5 1 ) ,
and ( 5 2 ) . They d i r e c t l y lead to the r e s u l t
Hence and using the equality d ' z 3 7 * ~ (=

0 )0 , we find that
3L>LlI (U)=('s?-f)!h.2;(.
B
T h i s completes the proof of the l e m m a .
4. The c a s e of a n analytical s y s t e m
In this subsection we will consider the computation of the focal values of

an analytical s y s t e m (A;l. We a s s u m e , a s before, that the multiple focus
coincides with the origin and that the s y s t e m h a s the canonical f o r m
dz
x=a--BY-kcp (f,Y), $=-BzLCty+*(z, Y), (A 1
where p > 0, and 'p and 1: a r e analytical functions.

The function R (p, 0) i n the right-hand side of equation ( 4 ) (see 524.1) is
c l e a r l y a n analytical function of 6 and p in t h i s c a s e in the s t r i p
--ca<e<+=, lp[<ro, (53)
where ro is s o m e positive number. T h e r e f o r e , i t can be series-expanded

in powers of p . Since by ( 3 ) and ( 4 ) , R (0, 6) =- 0, the s e r i e s expansion h a s
the form
Note that the function R (p. 0 ) and hence the functions Ri (e), i = 1, 2, . . .,
a r e periodic functions of 6 with period of 2 n . It follows f r o m the standard
p r o p e r t i e s of analytical functions that t h e r e e x i s t s ri > 0 such that the
s e r i e s ( 5 4 ) converges for a l l 8. O < B 4 2 n , and f o r all p . 1 p I d r , .
By QT, Appendix, 58.3, Theorem C, the solution
P = f (0; 0, Po) (55)
249
Ch.lX. CREATION OF LIMIT CYCLES FROM A tvIULTIPLE FOCUS I
of the equation
2 - R (P? e)
--
(4)
satisfying the initial condition
(56)
I
f (0;0, Po) 3 Po
is a n analytical function of the a r g u m e n t s . Let u s expand this solution in
p o w e r s of the "initial value" po. Since H (0, e) 0, we s e e that p E 0 is a
solution of equation (4), so that f (0; 0, 0) E 0. The expansion of f (e; 0, po) in
p o w e r s of po t h e r e f o r e h a s the f o r m
T h e r e c l e a r l y e x i s t s a number r z < r l such that the series (57) c o n v e r g e s

f o r all 0, 0 < : 8 g 2 n , a n d f o r a l l p, I p l s r z .
By (56) and (57),
Inserting f o r p and R in equation ( 4 ) t h e i r e x p r e s s i o n s f r o m (57) and (54)

and equating the coefficients of the corresponding p o w e r s of p o i n the r i g h t -
and the left-hand s i d e s , w e obtain the following r e c u r s i v e differential
equations f o r the coefficients u i (e) (i = 1, 2, 3, . . .):
U ; (0) =
U;
Ri (e)us (e),
u (e)=: R~(e)u2 (6) + f i 2 (e)U: (e),
(e)==R,(0) u 8 w + 2
. . . . . . . . . . .
4 (6) u1(e) u2 (e)-I-1 3 ~(e) u: (e),
, . . . . . ,
1 (59)
I
I
Relations ( 5 8 ) may be considered a s the initial conditions f o r the
functions (8) satisfying differential equations (59).* Using these initial
conditions and successively integrating equations (59) as l i n e a r differential
equations for the corresponding functions, we obtain
By definition, the s u c c e s s i o n function is f (po) = f (2n; 0, po) ( s e e S24.1, (9)).

Therefore, taking 0 = 2n i n (57), w e obtain a series e x p r e s s i n g the
s u c c e s s i o n function
P = f (Po) =u1 (2n) Po+u2(24 Pi+. ..

It is readily seen that equations (59) are obtained from equations (20) if we set pa = 0 and introduce an
appropriate notation.
250
5 "c. FOCAL L'ALC'ES
Let
ui (23) =ai, i = 1, 2, . . .
Then
P = i (Poj = alp, tn*p: - . .. (62)
F r o m the last e x p r e s s i o n and the equality d ( p , ) - f ( p o ) - p o , we obtain for

the focal v a l u e s
d'(G)=al-i=ul(2z)- 1, d A ) ( 0 ) = k ! a k = k ! ua(2rc) ( k = 3 , 3, ...). (63)
Let u s now d e r i v e e x p r e s s i o n s for the coefficients ai i n t e r m s of the

right-hand s i d e s of (A). The e x p r e s s i o n f o r the f i r s t focal value
. ' z
d' (I)) = e-'' 8- 1
h a s been d e r i v e d before (see J 2 4 . 2 , (25)). The s u c c e s s i v e focal v a l u e s are

of any i n t e r e s t only when 0 (0, I?) is a multiple focus, i.e., when CL = 0.
System ( A ) in this case h a s the f o r m
d i
5c = -k3Yf(F(z, !I), $=i3r+q(z, u).
Let
'p (2,Y) = P,(z,Y) iPJ(3, Y) 7- .. . . $((I, Y)"w,

Y) ==Q~(z, Y) .. ., (64 i
where P , (5. Y) and Q i(z, y : ~are homogeneous polynomials of i-th d e g r e e

( i = 2 , 3, . .-). By (3), (4:, and (64),
where
um(cos e, sin e) = P,* (cos 0, s i n e) cos iQm (cos Q , sin e) sin e
and (661
rm(COP 0 , sin e)= Qm (cos 8, s i n 0) cos O-P, (cos 8, s i n e) sin 0
( c o m p a r e with (39); the p r e s e n t t r e a t m e n t is in f a c t a repetition of the

t r e a t m e n t of the p r e v i o u s subsection f o r s y s t e m s of class -\).
In v i r t u e of the f i r s t r e l a t i o n in ( 2 8 ) , we have for CL = 0
P, (e) =
o R ( o H)
[A Of, 3 p=o -0. (67)
The series expansion o2 the function R ( p . e) is t h e r e f o r e given by
1: (P, 6 ) = R?( 8 )p2 +- RJ(e)p3 + .. . (68)
25 1
Ch.lX. CREATION OF LIMIT CYCLES FROM A MULTIPLE FOCUS
Equalities (65), (66), (68) yield, as in the previous subsection, the

relations
uz == BRz,
~3 + Rzvz,
= B R3
u4 + +R2~3,
BE4
............
H 3 ~ 2
Hence
Inserting these e x p r e s s i o n s in (60), setting 8 = 2 x , and using (67) and

(63), we obtain the focal values. For a multiple focus, the f i r s t focal value
is z e r o , d' (0) = 0 . The second focal value is a l s o z e r o , d" (0) = 2a2 = 0,
by L e m m a 5. Note that this fact a l s o e m e r g e s d i r e c t l y f r o m the r e l a t i o n
d" (0)= 2az= 2

7R, (e) d e .
The integrand, as is r e a d i l y seen, is a n odd periodic function of period 2 n .

The i n t e g r a l t h e r e f o r e vanishes.
For the third focal value of a multiple focus we have
d" ( 0 )= 31 aJ= 6
I 12R2 (e) u2(e) +- R~(e)]d e .
Using the e x p r e s s i o n s f o r R Z (e) and R 3 (e) in t e r m s of the polynomials
P z , Q 2 , P 3 . Q3 and writing these polynomials (of second and third d e g r e e ,
r e s p e c t i v e l y ) in the f o r m
we obtain a f t e r e l e m e n t a r y , but fairly lengthy, computations the following

e x p r e s s i o n s f o r a3:
[3(a30+bo3)-t (aiz-tbzdl-
-- G2 (71)
1 2 ~ ~ 2 0 ~ 2 0 - ~ 0 2 ~ 0 2 ~ - ~ 1 1 ~ ~ 0 2 $ ~ 2 0 ~ + ~ ~ 1 ~ ~ 0 2 T ~ 2 0 ~ 1 .
If a s # 0, d" (0) = 6a3is the Lyapunov value. From the r e s u l t s of 224.2

( s e e (24), (26), (27)) i t follows that i f f3 > 0 , 0 is a s t a b l e focus for all< 0
and a n unstable focus f o r a3> 0.
If a 3 . = 0, the c h a r a c t e r of the equilibrium s t a t e (with p u r e imaginary
c h a r a c t e r i s t i c n u m b e r s ) c a n be determined by considering as (if a3 = 0, then
252
6"4. FOCAL <;ALL'ES
by L e m m a 5, a, = 0 to'>); i f as = 0 , we have to consider u7, e t c . However,

the computational d i f f i , x l t i e s rapidly multiply a s the index increases.:
Expression (71) f o r a3 h a s been derived a s s u m i n g a s y s t e m of canonical
f o r m (1). W e w i l l a l s o give a n expression f o r a3 in t e r m s of the
coefficients of a s y s t e m e x p r e s s e d in the g e n e r a l f o r m
The c h a r a c t e r i s t i c n u m b e r s of the focus 0 are i pi, where
$=n=
+v-z=Ti. (74)
The expression f o r a3 in t e r m s of the coefficients of s y s t e m (72) h a s

been derived in / l a / . To d e r i v e this expression, the substitution
is applied to r e d u c e ( 7 2 ) to the canonical f o r m
*=-&+pZ
dt %? q)+F3(&q)+ * f . t
s=aE+q2(5+ q)+Q3(39 q)+ .. (75)
Expressing a3 f r o m (71) i n t e r m s of the coefficients of s y s t e m (75) and

r e v e r t i n g to the coefficients of the original s y s t e m (72), we find
aa= --
468
% +
{ [ a t( 4 1i attboz +aonbti)t a b (bi, f azobti + atibzo) +
+ c2 (aliao2+ 2aG2b02)-2ac (bX -a w d -2ab (a:-- bzoboz)-
- b2 ( 2 a d 2 0+ btlbZd + (bc-2a9 (bttbo2--aItapo)l-
-(a2f be) 13 (~bo3- bum) + 2a (a21 + biz) -(can-
i bbzdl} (76 )
(see /18/, p . 2 9 ) .
Equation (71) is a p a r t i c u l a r case of (76), and c a n be obtained f r o m the
l a t t e r f o r a = d = 0 , b = - p, c = i3.
Note that (76) actually gives a n e x p r e s s i o n - in t e r m s of the coefficients
of s y s t e m (75) - f o r the focal value of s y s t e m (75) and not the original
s y s t e m (72). This is i m m a t e r i a l , however, f o r the investigation of the
* For dynamic systems with quad.atic polynomials in the right-hand sides, i.e., systems of the form
P (+, Y) = ar-8y a& +qtw .o~Y? + +

Q(z7 Y)=BIioyib~+2+btily+bczyat
as,a, in terms of the coefficients ai, and b,, have been derived in / l B / .
the expressions for as,
253
ch.IX. <;REA'I'ION OF LIblIT CYCLES FROM A MULTIPLE FOCI!S
topological s t r u c t u r e , since s y s t e m s ( 7 2 ) and ( 7 5 ) a r e the i m a g e s of each

other by a non-singular l i n e a r transformation.':'
Note that the Lyapunov values were derived by Lyapunov from different
considerations.
R e m a r k c o n c e r n i n g t h e c a s e of a c e n t e r i n a n a n a l y t i -
c a l s y s t e m . A s we know (QT, 8.6), a n equilibrium s t a t e of a dynamic
s y s t e m of c l a s s N >/ 1 which h a s p u r e imaginary c h a r a c t e r i s t i c numbers
may be e i t h e r a m u l t i p l e f o c u s , or a c e n t e r , o r a c e n t e r - f o c u s .
For analytical s y s t e m s , the c a s e of a center-focus is ruled out. Indeed,
i f s y s t e m ( A ) is analytical, the corresponding function d (po) is a l s o
analytical. Therefore, f o r positive po which are sufficiently close to z e r o ,
d (po) r e t a i n s a constant sign and the point 0 (0, 0) is a multiple focus, or
alternatively d (po) zz 0 and 0 is a c e n t e r . In the f o r m e r c a s e , a t l e a s t one
of the focal values does not vanish. In the case of a c e n t e r , all the focal
values vanish, i.e.,
a,=l, az=a,= ...=0. (77)
Conditions (72) are c l e a r l y n e c e s s a r y and sufficient f o r point 0 (0, 0) to be
a c e n t e r . A c e n t e r is thus observed when a n infinite number of conditions
are satisfied.':'':
S25. CREATION O F LIMIT CYCLES FROM

A MULTIPLE FOCUS
1. The fundamental theorem
T h e o r e m 40 (theorem of the creation of limit cycles from a multiple

f o c u s ) . If 0 (0, 0) i s a multiple focus of multiplicity k (k>1) of a dynamic
s y s t e m (A) of class N > 2k f 1 m of analytical class, then_
1) there exist > 0 and 6o > 0 such that any system (A) go-close to rank
2k + 1 to system (A) has a t most k closed paths in u,,(0,;
2) f m any e < and 6 < 6or there exists a system (A) of class N oy
(respectively) of analytical class which as 6-close to rank 2k + 1 to ( A ) and
has k closed paths in U, (0).
P r o o f . 1 ) Let u s prove the f i r s t proposition of the theorem. Without
loss of generality, w e shall a s s u m e that s y s t e m (A) (of c l a s s N or analytical)
h a s the canonical f o r m
dd lr - - ~ y + ~ ( 3 , y ) 7
-- g=h&$((x,y)$ (A 1
where 'p and II, a r e functions which vanish together with their first d e r i v a -
tives a t the point 0 (0, 0).
Consider the function d (po) = f ( 2 x , 0, PO) - po corresponding to s y s t e m (A)
( s e e 924.1). L e t this function be defined f o r all po. I po I < r o , where ro is
s o m e positive number. A s we know (see 124.2, c o r o l l a r y f r o m L e m m a 3),
' The transformation E==, q= -aB=--!B.y is non-singular, since by (73) b # 0.
* * W e mean here conditions each o f which requires computation of one number. All these conditions are
of course equivalent to the single condition d (po) = 0, but the latter requires computation of a
function, and not of numbers.
254
d (p,) is a function of c l a s s N or (respectively) a n analytical function. By
assumption, the point 33 is a root of multiplicity 2k 4- 1 of the function d ( p o )
(see 524.2, Definition 2 6 ) . Consequently, t h e r e e x i s t numbers E , > Oand
uo > 0 such that any function a (p,) defined f o r p o , I P O I < r o , which iso,-close
to rank 2k +
1 to the function d (p,) h a s a t m o s t 2k 1 r o o t s in the interval +
(- Ea. eo).
Consider modified s y s t e m s (A), which are given in canonical f o r m and-,
f o r which the point 0 (0, 0) is a focus. Let 2 (p,) be the analog of d ( P O ) for-(A).
By Theorem 3 in Appendix, 1, t h e r e e x i s t s tio> 0 suclh that if s y s t e m (A)
is &,-close to rank 2k $- 1 to s y s t e m (A), the function d (po) is defined f o r
a l l p 9 , I po I -=
roI and fo- these po the functions d (p,) and d" (po) are u,-close to
each other to rank 2k $. 1. W e w i l l now prove that the n u m b e r s aG and E ,
satisfy the f i r s t proposition of the theorem. Suppose that this is not so,
i.e., suppose that t h e r e e x i s t s a modified s y s t e m (A) of canonical f o r m which
is Cio-close to r a n k 2% +
1to s y s t e m (A) and yet h a s m o r e than k closed paths
in U,, (0). Each of these paths c r o s s e s e v e r y r a y issuing f r o m 0 p r e c i s e l y
at one point (see QT, J8.4, L e m m a 1). L e t p , and p z be the a b s c i s s a s of the
intersection points of a. path with the r a y s 0 = 0 and 0 = x I respectively.
Then pt and p2 a r e respectively r o o t s of the functions a ( p ) = 2, (p) and & (p),
By ( 1 3 ) i n 524.1, 6 ( - p 2 ) = - a n @ 2 ) .Therefore
,-
i.e., d ( p l ) = O , & ( p Z ) = O .
a (- p2) = 0, i.e., - p z is a l s o a root of the function 3. Evidently pI and
- p2 a r e d i f f e r e n t n_umbr?rs, s m a l l e r than &,inabsolutevalue. Thus, to e v e r y
closed path of (A) lying in U,,(0)correspond t,"o r o o t s of the function 2 ( p )
f r o m the interval (- E , , E , ) . Hence, i f s y s t e m ( A ) h a s in Us, (0) m o r e than k
closed paths, the tunction B ( p ) h a s a t l e a s t 2k +
3 different r o o t s '.- in (- E O , E O ) .
This c l e a r l y contradict,s the choice of the n u m b e r s 6 , and E* W e hzve thus
proved the f i r s t proposition of the theorem f o r modified s y s t e m s (A) given
in canonical f o r m .
Let u s now consider a g e n e r a l modified s y s t e m (A) (not n e c e s s a r i l y
canonical). Suppose that the f i r s t proposition of the theorem is not satisfied.
Then, for any el > 0 and a1 > 0, t h e r e e x i s t s a s y s t e m (A$:) 6,-close to rank
2k + 1 to (A) which h a s m o r e than k closed paths in Ue1(0). If el and 61 are
sufficiently s m a l l , t h e r e e x i s t s a l i n e a r transformation, as c l o s e a s
d e s i r e d to the identity t_ransformation, which t r a n s f o r m s the s p t e m ( A )
to the canonical f o r m ( A ) . x ik For sufficiently s m a l l el and 6,, (A) is a l s o
6,-close to ( A ) and h a s m o r e than k closed paths in U,,(0), which are obtained
from the closed paths of (A:; ) in U e l ( 0 by ) the s a m e l i n e a r transformation.
This, however, contra'dicts the previous r e s u l t . The f i r s t proposition of
the theorem is thus completely proved.
2 ) W e can now proceed with the proof of the second proposition.
Consider a modified s y s t e m of a s p e c i a l f o r m
dz -
x = p ( x , y, xo, AI? .-a, hh-1)=
=P(Z, Y)+@fkf(ZliY')zi... +.kh-$(3"$.y')k-1Z,

dy -
=Q (3,Y, A,, . . . , ak-l)= (A 1
= Q ( hY) + 'by + h.1 ( z ' f !/'IY + . .. + h - I (z' -? Y')'-'Y.
* Since zero is also a r o o t of this function.
** This can be proved along the iame lines as Lemma 1, 19.1. In this lemma, a similar proposition is
proved for a saddle point. In the case of a focus, we should further make use of the fact that the
a
( b pL
'u-pi
) is transformed to the matrix ( 6a -6a) by a linear transformation with the matrix (-'' -'1).
255
Ch.lX. CREATION OF L I M I T CYCLES FROM A MClLTlPLE FOCUS
For sufficiently s m a l l Ai> (A) is evidently a r b i t r a r i l y c l o s e to (A) to any

( p o s s i b l e ) rank and belongs to the s a m e c l a s s a s (A). The point 0 is a n
equilibrium s t a t e of (A). Now, since the l i n e a r p a r t s of the functions P and
+
0 a r e respectively Lox - j3y and fix Ahoyt the equilibrium s t a t e 0 (0,O) is a
focus of s y s t e m (A) f o r h, # 0 and a focus, a c e n t e r , o r a c e n t e r - f o c u s f o r
ho = 0 . Let
'
(
P
O
? LO, i t , ... 7 Ak-1)
be the analog of d (p,) for s y s t e m (A). Clearly, 2 i,s a continuous function

of po and of the p a r a m e t e r s ho, Al, . . . , h k - l . Since (A) is obtained f r o m (A)
for A, = ;Il = . . . = hk-, = 0, w e have
Z(P0, 0, 0, . . ., 0) = d (Po). (1)
F o r any e > 0 and 8 > 0, t h e r e exist h* > 0 and r* > 0, r* < ro,* such that
for
lAtl<A*, i = O , 1, 2. ..., k-1, (2)
1 ) system (A) is
8-close to r a n k 2k 1 to (A); +
.
. ., Ak-l) is defined f o r all po, I po I < ro, and
2 ) the function a ( p o , A,, h l ,
e v e r y root satisfying the inequality I po I < r* c o r r e s p o n d s to a closed path
of (A) e n t i r e l y contained in U,, (0).
It is henceforth a s s u m e d that the n u m b e r s A i , i = 0, 1, 2, . ., k - 1 , s a t i s f y .
condition (2).
We w i l l show that f o r a: a p p r o p r i a t e choice of the p a r a m e t e r s
.
Lo, Al, L2, . ., Ak-, s y s t e m (A) h a s k closed paths in U. (0).
F r o m Maclaurin's f o r m u l a , f o r all sufficiently s m a l l po,
where h (po) is a continuous function and h (0) = 0 ( s e e proof of Theorem 5,

51.3).
By assumption, d(2k++') (0) # 0. To fix i d e a s , l e t d ( 2 k + l ) (0) > 0. Then f o r
all sufficiently s m a l l positive po, d (p,) > 0. We choose one of these numbers,
s m a l l e r than r*; designating i t rlr we have
O<ri<r*, d ( r l ) = 2 ' ( r l , O , 0, ..., o)>o. (4 )
Let now
... = r ~ ~ - ~ = oA, ~ - ~ + o ,
and consider the modified s y s t e m corresponding to these p a r a m e t e r s
* By definition, r, > 0 is a number such that t h e function d (po) is defined for I po I < ro.
256
$25. CRE4TION OF LIhlIT CYCLES FROhl A hLILLTIPLE FOCI'S
and the corresponding function d; (po) = a(p,,, 0, 0, .. ., 0,;?,?-I). By Lemma 7 of

the previous section,
(0) = . . . = d-(?k-2)
c
& (O)=d; I (0) = O , 6 ( 2 k - ' ) ( 0=) ( 2 k - l j ! k 2 ~ ,
B
(5)
and by hIaclaurin's formula, f o r all sufficiently s m a l l po,
where E, (pa) is a contir-uous function and Xl(0)= 0 .

T O fix ideas, let p:,O, and we choose h k - 1 s o that
The l a s t of these conditions is satisfied f o r any sufficiently s m a l l r?k-$ i n

virtue of ( 4 ) and the continuity of d-(p,, Lo,I.,, . .., X k - l ) .
F r o m p>O, X k - i s O and equation ( 6 ) i t follows that f o r all sufficiently
s m a l l positive po, dl (po)(0. W e choose one such value, s m a l l e r than r l , and
designate it r 2 . Thus,
0 < r2 < r I < r+ (8)

and
d", ( r l ) > 0, 21(rz) < 0. (91
F u r t h e r constructicm is completely analogous.9 Indeed, w e consider a
system
* A l s o see the proof of Lemma 1 . 9 1 . 3 .
257
By ( l l ) , &(po)>O f o r all sufficiently s m a l l positive po, and t h e r e e x i s t s
r3 such that
O<r3<rz
and
d", (ra) >0.
Continuing a s before, we eventually obtain a s y s t e m
d+ -
= P ( x , Y, 0, ai, . . ,, ak-l)=
= P (I,y) + h, (sa + ye) z+ . ..+a b l (2+ yZ)k-12,
x-= Q-
dy
(I,Y, 0,h,, . . ., hk-i) =
==Q (I,Y) + a, (I2+ye) Y + ... +L,(s* +
and
Continuing to the s y s t e m
we choose L, s o s m a l l that
and
i f k is odd,
(0
1 0 if k is even. I
By (25), S24.2, we have f o r the f i r s t focal value
2x "0
&(O)=e B -i.
Therefore, to satisfy (18), a negative I o should be taken if k is odd and a

positive one i f k is even. C l e a r l y , i f f o r this choice of Lo, the number
r k + l > 0 is sufficiently s m a l l , we have
(0 i f k i s odd,
'k(rk+d )>O if k is even.
258
$ 9 5 . CREATIOS OF LI:,IIT CYCLES FROhf A X'bLTIPLE FOCLS
I r e m o r e o v e r a s s u m e that ?-hi( < rk. All the A,, i = 0 , 1. 2, . . .. k - 1, that

w e have chosen a r e l e s s than_ >.*in absolute value. Therefore, by condition I
imposed on E.*, the s y s t e m (Ah) is 6 -close to r a n k 2k + i to s y s t e m (A).
F u r t h e r m o r e , in virtue of ( 1 7 ) and ( 1 9 ) and the continuity of d k , t h e r e is a t
l e a s t one root of the function Jk (pa) between e a c h p a i r rl and r 2 . r? and
r 3 , . . .. rA and rk+l. I'herefore. this function h a s at l e a s t k different positive
roots, each s m a l l e r than r*. By condition 2 f o r r * , e a c h of these r o o t s
c o r r e s p o n d s to a closed path entirely contained in C, (0). If E < and 6 60.
C', (0)c Uc0(0) and by the f i r s t proposition of the theorem, (xk) h a s a t most
k different closed paths in U,,(0). T h i s completes the proof of the t h e o r e m .
R e m a r k . In o u r proof of the first proposition of Theorem 40, eo and
60 should be chosen so s m a l l that any s y s t e m (A) 6,-close to (A) h a s a single
equilibrium s t a t e in U,,(0)) which i s m o r e o v e r a focus. Then w e can speak
of the existence of the s u c c e s s i o n function and the function d f o r the s y s t e m
( 3 ) . In what follows, the numbers E,, and 60 are always a s s u m e d to meet this
requirement. hIoreover, i f 6o is sufficiently s m a l l , the points move along
the paths of ( A ) in Ue, (0) in the s a m e direction with increasing t (clockwise).
Indeed, reducing (A) to a canonical f o r m by a transformation close to
identity. i.e ., by a n orientation-conserving transformation, w e obtain a
system
The number is c l o s e to fi and therefore h a s the s a m e sign. This e n s u r e s

identical directions of motion. U'e w i l l a s s u m e in the following that this
condition i s a l s o satisfied.
2. Bifurcations of a dynamic s y s t e m in the neighborhood

of a multiple focus
The following theorem strengthens the second proposition of Theorem 40.

Together with Theorem 40, i t plays a leading r o l e r e g a r d i n g the bifurcations
of a dynamic s y s t e m i n the neighborhood of a multiple focus.
T h e D r e m 4 1 . Let 0 Kt. 0) be a multiple focus of multiplicity k of a
dynamic system (A) of class .V 2% - - i m of analytical class, and let E, and
6 n be positive numbers defined by the f i r s t proposition of Theorem 40 and
the above remark following the themem. Then
1) f o r any P and 6 . 0:. E 4 el). 0 < 6 ..: 6,,, and f o r any S , 1 -IS ik , there exists
a system ( B ) of class -V ( m respectively, analytical) which i s 6-close to
rank xk - to system (A) and has in (0)
tI precisely s closed paths;
2) if system (E)i s 6,,-closeto rank 2k A 1 to system (A) and has k limit
cycles in i-eo (O), all these cycles, and liketoise the focus of system (B)
lyipg in L.eo (O),are structurally stable (simple).
Proof . Let u s prove the f i r s t proposition of the t h e o r e m . Let 1 c s< k
( f o r s = k, this proposition coincides with the second proposition of
Theorem 40): In $e proof of Theorem 40, we constructed a succession
of s y s t e m s (AI). (Az), . . ., (xk..,),(Ak).All these s y s t e m s evidently can
be a s s u m e d to be 6-close to r a n k 2k + 1 to s y s t e m (A).
Consider the s y s t e m (A;). According to o u r construction,
259
Ch.IX. CREATION OF LIMIT CYCLES FROM A MULTIPLE FOCUS I
and
-
d", (r,) > 0, & ( r z )< 0, . . ., d, (ra+l) I >U
(0
i f s is even,
if s is odd. (21)
( s e e proof of Theorem 40). By (21), (A,) h a s a t l e a s t s closed paths in U, (0).

Suppose that i t h a s s +1 closed paths L 1 , L z , . . ., Le+, in U , (0).
By
condition (201, the point 0 (0, 0) is a
m-ultiple focus of multiplicity k - s for
(A8). Let V and Wbe two neighborhoods
of 0, such that W c V c U, (O),and Y l i e s
inside all the closed paths L i (i=
= 1, 2, . . ., s + 1). These paths a r e
"concentric" ( F i g u r e 115). By
Theorem 40 t h e r e e x i s t s a s y s t e m (A:')'
of c l a s s N , a r b i t r a r i l y c l o s e to (A,) to
rank 2k + 1, which h a s k - s closed paths
L,,,, . . ., Lk+lin W. But then, using
the s a m e construction a s in the proof of
L e m m a 2, 515.2, we can c o n s t r u c t a
s y s t e m (A) of c l a s s N , a r k i t r a r i l y c l o s e
to rank 2k + 1 to s y s t e m (As),which
coincides with (A,) outside V and with
(A'$)inside W . The s y s t e m ( A ) evidently
FIGURE 115
can be r e g a r d e d a s 8,-close to rank
2k + 1 to (A), and i t h a s a t l e a s t k -f- 1
closed paths Li (i = 1, 2, . . k + 1) in U,, (0). This c l e a r l y c o n t r a d i c t s
Theorem 40. T h e r e f o r e , ik,) h a s p r e c i s e l y s l i m i t c y c l e s in U t(O),i.e., the
f i r s t proposition of the theorem is proved.':'
Let now (B) be &,-close to rank 2k + 1 to (A) and suppose that it h a s k
l i m i t c y c l e s in U,, (O), a t l e a s t one of which is s t r u c t u r a l l y unstable. Then
modifying the s y s t e m only in the neighborhood of this cycle (again using the
construction of L e m m a 2, %15.2), we obtain a s y s t e m which is c l o s e to (A)
and h a s in U,, (0)m o r e than k closed cycles, which is impossible. A s i m i l a r
contradiction is obtained if we a s s u m e that the focus of s y s t e m (B) i n U, (0)
is a multiple focus. The proof of the theorem is complete.
T h e o r e m s 4 0 and 41 lead to important conclusions r e g a r d i n g the possible
bifurcations of a dynamic system in the neighborhood of i t s multiple focus
of finite multiplicity. indeed, consider a k-tuple focus 0 (0, 0) of s y s t e m (A)
(k2,2). Let E,, and 6o be sufficiently s m a l l n u m b e r s (defined by Theorem 40
and the r e m a r k following the theorem), and V a neighborhood of 0 bounded by
a cycle without contact F, V c U,, (0). We choose 6L0 < 8 < 6os to be so
s m a l l that the following condition is s a t i s f i e d i i f (A) is 8-close to (A), the
c u r v e r r e m a i n s a cycle without contact f o r (A) and (A) h a s in V a single
equilibrium s t a t e a, which is a focus. By T h e o r e m s 40 and 41, s y s t e m (A)
6 -close to rank 2k + 1 to (A) may have in Vat m o s t k l i m i t cycles, and there
e x i s t s y s t e m s (A) which have in V p r e c i s e l y s l i m i t c y c l e s , where s is any
number, 1 G s d k . These l i m i t cycles a r e a r r a n g e d "concentrically" and
enclose the focus 0" inside them. Let L,,L z , . . ., La be these cycles, and
.
suppose that Li l i e s inside Li+, (i = 1, 2, . ., s - 1). The topological s t r u c t u r e
If (A) is analytical, (A;)is also analytical
260
$ 2 5 . C R E A \ f O N OF LI\IIT CYCLES FRO!,! .4 LILLTIPLE FOCCS
of s y s t e m (A} in J' i s completely determined by the stability c h a r a c t e r of the

focas 8, the number s cbf limit c y c l e s in I-, and their respective stabilities.
If the stability character. of the focus Ois known, i t suffices to know whether
each of the c y c l e s L , i s s e m i s t a b l e or not." In 12, w e introduced the
m u l t i p l i c i t y o f a l i m i t c y c l e *':: and established that a n even-
multiplicity cycle is semistable and a n odd-multiplicity cycle is e i t h e r stable
o r unstable (this follows f r o m J12.3, ( S ) , and S12.4, (11)). Hence, the
topological s t r u c t u r e of the dynamic systeni (A) in the neighborhood I. is
completely determined if w e know
( a ) the stability c h a r a c t e r of the f o c p 8 ;
j b j the nuniber s of liniit c y c l e s of (A) i n J';
\ c ) the parity of the multiplicity of each of these c y c l e s (whether odd or
even).
Since s < k , s y s t e m (6) c l e a r l y may have only a finite number of different
topological s t r u c t u r e s in I-. In o t h e r words, s y s t e m (A) may undergo only
a finite number of different bifurcations in a neighborhood of a focus of
finite multiplicity. +
Assigning + (-) to a stable (unstable) focus, and the n u m b e r s O ( 1 ) to a
cycle of even (odd) multiplicity, w e obtain the following classification,
corresponding to i t e m s \a), (b), ( c ) above:
f, 1, 1, 0, 1, 0, o...o, 1, 0 ,
where the nuniber of ones and z e r o s is s . Each s y s t e m (A) 6-close to

rank 2k Ito (A) is c h a r a c t e r i z e d by a definite s c h e m e of this kind. W e
will not consider the i n v e r s e question, namely whether or not e v e r y s c h e m e
of this kind fully c h a r a c t e r i z e s a dynamic s y s t e m a r b i t r a r i l y close to (A).
Note, however, that by I'heorems 40 and 41, t h e r e e x i s t s y s t e m s
a r b i t r a r i l y c l o s e to (A) whose s c h e m e s contain k ones.
3 . Bifurcations in the neighborhood of a multiple focus

of multiplicity 1
\ r e will consider a sFecial case which is often encountered in applications

Let
be a dynamic s y s t e m which depends on the p a r a m e t e r X. We will c o n s i d e r

the bifurcations of this s y s t e m associated with the variation of the p a r a m e t e r
1. in the neighborhood of a n equilibrium s t a t e 0 (0, O ) , when 0 is a multiple
focus of multiplicity 1. For simplicity, w e a s s u m e that the bifurcation
' If, say. the focus 8 is unstable, the cycle L, is either unstable or semistable (unstable from inside and
strble from outside). In the former case. L 2 i s either stable or semistable. and in the latter case i t is either
unstable or semistable, etc,
** In 912.3, the multiplicity of a limit cycle is defined for analytical systems. An analogous definition
for rystcms of class .V w i l l be foind in Chapter X (526.2. Definition 28).
t Wt Chould stress that in the case of a k-tuple focus, the closeness of systems is considered to rank 2 k+ 1 .
26 I
Ch.IX. CREATION OF LIMIT CYCLES FROM A MULTIPLE FOCUS
value of the p a r a m e t e r is h = 0. Let
Then
To establish the stability c h a r a c t e r of the multiple focus 0 (0, 0) of the

s y s t e m (Ao), w e w i l l proceed along the s a m e l i n e s a s in the end of S24,
when dealing with s y s t e m (72). We apply the transformation
which r e d u c e s (&) to the canonical f o r m
Since ( 2 6 ) is a non-singular transformation, 0 r e m a i n s a multiple focus of

multiplicity 1f o r s y s t e m ( 2 7 ) also, and i t s stability does not change e i t h e r .
The third focal value of the focus 0 of ( 2 7 ) therefore does not vanish, and
i s the L y a p u n o v or t h e f i r s t L y a p u n o v v a l u e of the focus O ( s e e
124.4, Definition 27). We used the symbol 6 u 3f o r this value in S24.4, (63). I
Equation (76) in $24 provides a n e x p r e s s i o n f o r u 3 in t e r m s of the coefficients
of the original system (Ao), and therefore transformation (26) need not be
actually applied in p r a c t i c e . W e have s e e n (S24.2, (26), (27)) that i f
u 3< 0, 0 is a stable focus (of s y s t e m (27), and therefore of (Ao)), and if
u g > 0 , it is a n unstable focus.
Let V be a sufficiently s m a l l neighborhood of the point 0 bounded by a
cycle without contact I? of s y s t e m (Ao) which contains no closed paths of (&)
o r equilibrium s t a t e s other than 0.
Let 6 , > 0 be s o s m a l l that any system (A,) for which I h I < 6,, h a s the
following proper t i e s :
( a ) the curve r is a cycle without contact f o r this system;
( b ) (A,) h a s no equilibrium s t a t e s , other than 0, in V ;
( c ) the point 0 is a focus of system (A,);
( d ) (A,) has a t most one closed path i n V.
The validity of the f i r s t t h r e e conditions for sufficiently s m a l l 6 , is s e l f -
evident. Condition (d) is satisfied because 0 is a focus of multiplicity 1 of
s y s t e m (&) and because c l o s e s y s t e m s have a t most one closed path in a
s m a l l neighborhood of this focus (by Theorem 40).
By condition (a), the paths of a l l s y s t e m s (A,) ( I I I < 6,,) simultaneously
c r o s s the c u r v e I? with increasing t , e i t h e r f r o m outside to inside o r from
inside to outside.
Suppose that u (A) r e v e r s e s i t s sign a s the s y s t e m p a s s e s through the
bifurcation value of the p a r a m e t e r h = 0, i.e., the focus 0 changes i t s
stability. This condition is clearly satisfied i f d (0) f 0.
262
$95. CREPIOb OF LIhlIT CYCLES FRU!,I h hlCLTIPLE FOCUS
Let u s now consider the different possible cases.

1 ) cz3 < 0. K-e a s s u m e that on passing through the bifurcation value of
the p a r a m e t e r h = 0, u (A) changes i t s sign f r o m minus to plus. JX G' (0)# 0,
this condition is satisfied
when h i n c r e a s e s , f o r IS' (0) > 0 ;
when >.d e c r e a s e s , f o r u' (0) < 0.
Since o3 < 0, the focus 0 is a stable focus of (Ao) f o r A = 0 . T h e r e f o r e a l l
the paths of (A,) e n t e r into the cycle without contact r a s t i n c r e a s e s . For
IS (i.) < O , Oisastablefocusof(A,). ByTheorems40and31, ( A , ) h a s a t m o s t
one limit cycle in I', and if this cycle e x i s t s , i t is a simple cycle, i.e.,
either stable or unstable. Clearly, f o r CJ (i <)
0 no such cycle e x i s t s . In-
deed, i f this cycle existed, i t would be stable f r o m outside and unstable f r o m
inside, i.e., i t could not he simple. We have thus established that i f a 3< Oand
(I (>.) < 0 , (A,) h a s no l i m i t cycles in V .
.;
Conversely, if u (k) 0, 0 is a n unstable focus of (Al). Then, reasoning
a s before, we conclude that t h e r e is a single limit cycle LAof (AA)inside V ,
and this is a simple sta'5le cycle. It is readily s e e n that if X is sufficiently
s m a l l , the cycle L A is a r b i t r a r i l y c l o s e to 0.
W e thus obtain the following r e s u l t s . If a3< 0 and (I' (0) > 0, s y s t e m (A,)
h a s no l i m i t c y c l e s in 1'for s m a l l negative X and h = 0, and 0 is a stable
focus. A s the s y s t e m c r o s s e s the bifurcation value of the p a r a m e t e r X (i.e.,
for h>O). the focus becomes unstable, and a stable l i m i t cycle develops
inside the neighborhood Y ( F i g u r e 116).
a b
FIGURE 116. ax< 0, 0' (0)> 0. a) X <O. stable
focus: h = 0. stable multiple focus; b) X > 0 , un-
stable tocus, stable cycle.
If T. is varied in the opposite direction, i.e.. we move f r o m positive to

negative 1, the stable l i m i t cycle which originally existed in V would
contract to the focus 0 and vanish f o r h = 0, and the focus w i l l change i t s
stability accordingly.
A s X is f u r t h e r d e c r e a s e d , the focus r e m a i n s stable and the topological
s t r u c t u r e of V d o e s not change.
For a3< 0, u' (0)< 0, the stable l i m i t cycle is c r e a t e d on passing f r o m
positive to negative h , a i d conversely i t d i s a p p e a r s when I i n c r e a s e s and
reaches zero.
2 ) a3 > 0. The investigation p r o c e e d s along the s a m e l i n e s as before.
If a 3 > 0 and u' (O)>O, the point 0 f o r s m a l l negative h is a stable focus of
(Ab)and the s y s t e m h a s one unstable l i m i t cycle in F'. A s b i n c r e a s e s , this
cycle c o n t r a c t s to the point 0, and a t the bifurcation point A = 0 i t
d i s a p p e a r s and the focus 0 becomes unstable. F u r t h e r i n c r e a s e of b l e a v e s
26 3
Ch.IX. CREATION OF LIMIT CYCLES FROM A MULTIPLE FOCUS
the focus 0 unstable and the topological s t r u c t u r e of the s y s t e m in 1' does

not change ( F i g u r e 117).
a b
FIGURE 117. U S 32 0 , u' (0)> 0. a) h -=
0,stable
focus, unstable cycle; b) h = 0,multiple unstable
focus: h > 0, unstable focus.
If a3 > 0 and u' (0) < 0, the unstable l i m i t cycle is c r e a t e d as h changes

f r o m negative to positive v a l u e s , and conversely d i s a p p e a r s when h d e -
c r e a s e s to z e r o .
The above r e s u l t s c a n be s u m m a r i z e d in the following table:
Unstable focus, Stable focus, Unstable focus,

no cycle no cycle stable cycle
Unstable focus, Stable focus, Stable focus,
stable cycle no cycle no cycle
Stable focus, Unstable focus, Unstable focus,
unstable cycle no cycle no cycle
Unstable focus Unstable focus, Stable focus,
no cycle no cycle unstable cycle
The above a n a l y s i s shows that the change in I b r i n g s about a change in

the stability of the focus if a l i m i t cycle is c r e a t e d f r o m the focus or d i s a p p e a r s
contracting into the focus. A s t a b l e focus c r e a t e s a stable cycle, and an
unstable focus, a n unstable cycle. Thus, a focus c r e a t e s a l i m i t cycle of
the s a m e stability, and the stability of the focus changes in the p r o c e s s .
Conversely, when the cycle d i s a p p e a r s (when i t is "absorbed" by the focus),
the focus a c q u i r e s the s a m e stability a s that of the cycle before "absorp-
tion." This s t a t e of things is not limited to the c a s e of a focus of
multiplicity 1: i t is observed whenever a focus c r e a t e s or a b s o r b s a cycle
of definite stability (i.e., not a s e m i s t a b l e cycle).
E x a m p 1e 9. Consider the s y s t e m
for s m a l l v a l u e s of the p a r a m e t e r h.*. This s y s t e m h a s two equilibrium

states, and we will consider the state 0 (0, 0) only.
* System (28) is of importance in the theory of sustained oscillations. It was investigated by Bautin /20/.
264
$25. CRE.4TION OF LIhlIT CYCLES FROLI .4 PIIL'LTIPLE FOCLIS
The c h a r a c t e r i s t i c equation k a - - h k + I = O h a s two roots k = ? + -

1
117.
A.2
Therefore, for s m a l l ?,, the point 0 (0, 0) is a focus, which is stable f o r

t < 0 and unstable for t > 0 .
For this s y s t e m , u ( 2 , ) = A , and therefore 0' (iZ) = 1. To compute a 3 , we
u s e equation ( 7 6 ) in S24.4, taking
u=O, b = 1 . ~ = - - l , d=O. b20--6, bi * -i3,

- b0 : ! = y , 1 . T = i 1 .
r h i s gives a 3 = n/4 @ (y -L6). F r o m the inequality u' (0) > 0 and the table
above w e conclude that if fi ( y +6) < 0. the focus 0 (0, 0) is stable f o r ?.GOand
t h e r e are no limit cycles in i t s neighborhood. A s we move to positive i . ,
the focus c r e a t e s a stable l i m i t cycle, and itself becomes unstable.
If @ ( y -+ 6) > 0, the focus 0 (0, 0) is stable f o r 1. < 0 and t h e r e is a n unstable
limit cycle in i t s neighborhood. A s ). i n c r e a s e s , the cycle c o n t r a c t s to a
point and d i s a p p e a r s f o r 2. = 0 . A t this instant, the focus 0 becomes unstable.
The topological s t r u c t u r e in the neighborhood of 0 for 2. > 0 is the s a m e as
for 1. = 0.
265
Chapter X
C R E A T I O N OF C L O S E D P A T H S F R O M A M U L T I P L E
L I M I T C Y C LE
INTRODUCTION
The p r e s e n t chapter is entirely analogous to Chapter IX, C r e a t i o n

of L i m i t C y c l e s f r o m a M u l t i p l e F o c u s b o t h i n r e g a r d to
methods used and the r e s u l t s obtained. It is directly related to Chapter V,
C 1 o s e d P a t h s , and may be r e g a r d e d as i t s continuation. A s in
Chapter V, the main emphasis is on the succession function f (n) on a n o r m a l
to the limit cycle Lo and the function d (n) = f (n)- n. In Chapter V we have
s e e n that i f d' (0) # 0, the l i m i t cycle Lo of s y s t e m (A) is structurally stable,
and if d' (0) = 0, the l i m i t cycle is unstable and t h e r e exist s y s t e m s
a r b i t r a r i l y close to (A) which have a t l e a s t two closed paths in any s m a l l
neighborhood of L o .
In the p r e s e n t chapter we consider not only the f i r s t derivative d' (0), but
a l s o the value of the higher derivatives of the functions d (n) a t the point 0,
and t h i s l e a d s to m o r e refined r e s u l t s regarding the creation of closed paths
from a multiple l i m i t cycle.
Chapter X is divided into two sections. The f i r s t , S26, although highly
significant f o r what follows, p r e s e n t s auxiliary background information. It
is mainly devoted to the derivation of e x p r e s s i o n s f o r the derivatives of the
succession functions in t e r m s of the right-hand s i d e s P and Q of the dynamic
s y s t e m . Moreover, the m u l t i p l i c i t y o f a l i m i t c y c l e i s d e f i n e d
in $ 2 6 (Definition 28, $26.2). A limit cycle Lo is said to have multiplicityr
(or to be a n r-tuple limit cycle) i f
d' (0) = d" (0) = ... dcr-l) (0)= 0, d(') (0) # 0.
The fundamental t h e o r e m s concerning the creation of closed paths f r o m
a multiple l i m i t cycle a r e presented in $ 2 7 ( T h e o r e m s 42 and 43).
Theorem 42 is analogous to Theorem 40 on the creation of closed paths f r o m
a multiple focus. It amounts to the following: i f Lo is a multiple limit cycle
of multiplicity k of a dynamic s y s t e m (A), s y s t e m s sufficiently c l o s e to ( A )
can have a t most k closed paths in a sufficiently s m a l l neighborhood of L o .
On the other hand, t h e r e exist s y s t e m s as c l o s e a s d e s i r e d to ( A ) with
p r e c i s e l y k closed paths in any s m a l l neighborhood of L o . In distinction
f r o m the c a s e of a k-tuple focus, when c l o s e n e s s to rank 2k + I is postulated,
h e r e we a r e dealing with c l o s e n e s s to rank k. Theorem 43 shows that if
Lo is a k -tuple l i m i t cycle, and s is a n integer, 1<s,<k, t h e r e exist s y s t e m s
a r b i t r a r i l y close to (A) which have precisely s limit c y c l e s in any s m a l l
266
506. I)ERIV.ATIVES CY T H E SLCCESSIOS FCNCTIOS
neighborhood of L o . It follows f r o m T h e o r e m s 4 2 and 43 that a dynamic

s y s t e m may undergo only a finite number of different bifurcations in the
neighborhood of a l i m i t cycle of finite multiplicity. These different
bifurcations c a n be readily c l a s s i f i e d .
5 2 6 . EXPRESSIOKS FOR T H E DERIVATIL-ES O F THE

SUCCESSIO?: FUKCTICN. hIULTIPLICITY O F A
L1hrI-r C Y C L E
1. Expressions for the d e r i v a t i v e s of s u c c e s s i o n functions
Let
be a dynamic s y s t e m of class i V g i o r a n analytical s y s t e m , Lo a closed

path of (A),
2 = F (0, Y = * (t) (1)
the motion corresponding to this path, T > O the period of the functions p and
$-
Consider the neighborhood Q of the path Lo d e s c r i b e d i n Chapter V (B13.1)
with the c u r v i l i n e a r c o o r d i n a t e s s and n , defined by the r e l a t i o n s
-
z=cp(s, n). y=3(s, n), (21
where
The functions (p and $ are considered i n the s t r i p
where 6 is a sufficiently s m a l l positive number.

The p r o p e r t i e s of mapping ( 2 ) and of functions ( 3 J are d e s c r i b e d i n
Chapter V, S13.1.
Changing o v e r to the v a r i a b l e s s and n in s y s t e m (A), using r e l a t i o n s (21,
w e obtain the s y s t e m
--- -_ -
ds
-= p(?. $).$,-Q(?t\P).CP, -dn
= Q($*$).6;-p(P,\P).lp; (5)
dt A h ni dt A h n)
where
361
ax. CREATION OF CLOSED PATHS FROM A MULTIPLE LIMIT'CYCLE
( s e e 513.2, ( 8 ) ) . A s in 913, we will reduce s y s t e m (5) to a single differential

equation
which is obtained when the second equation in (5) is divided through by the
f i r s t equation. This division is p e r m i s s i b l e by $13.2, (11). We w i l l
r e i t e r a t e h e r e the p r o p e r t i e s of the function R (s, n) and of equation (7).
R (s, n) is defined in the s t r i p (4), w h e r e i t is a continuous function,
periodic in s with a period of T . If (A) is a dynamic s y s t e m of c l a s s N
(analytical s y s t e m ) , R (s, n) is a l s o a function of c l a s s N (analytical function).
Moreover,
R ( s , 0 ) = 0. (8)
Therefore, n = 0 is a solution of equation (7). This solution evidently

c o r r e s p o n d s to the closed path Lo of (A).
Let
be a solution of ( 7 ) satisfying the initial condition
f (so; so, no) ==no.
i s a succession function on the a r c without contact s = so which is a n o r m a l

to the path L o . In what follows, we shall take f o r simplicity so = 0 (this
evidently does not r e s t r i c t the generality of o u r a n a l y s i s ) ; the c o r r e -
sponding s u c c e s s i o n function w i l l be designated f (no)and the n o r m a l s = 0 w i l l
be designated I . Thus,
f (no)= f (7; 0, no) (11)
is the s u c c e s s i o n function on the a r c without contact 1 .
Together with the s u c c e s s i o n function f (no), we consider the function
d (no)= f (no)--no, (12)
In Chapter V we computed the f i r s t derivative of this function a t no = 0 ,
s
I
[P;(rp(s). Ip(*))+Q~to(*). W))W

d' (0)= 20 -1. (13)
Our immediate a i m is to d e r i v e e x p r e s s i o n s for the higher d e r i v a t i v e s , i.e.,

d"(O), d"(O), ...,
or, equivalently, f o r the d e r i v a t i v e s of the s u c c e s s i o n function
f " ( 0 ) . f"(O), .e.
268
The derivation of t h e s e e x p r e s s i o n s is analogous to the d e r i v a t i o n of the
e x p r e s s i o n s for the focal v a l u e s in S24.3.
Since the right-hand s i d e R (s. n) of the differential equation ( 7 ) is a
function of c l a s s -V, the solution
n - f ( s ; 0. no)
of this equation h a s cortinuous p a r t i a l d e r i v a t i v e s with r e s p e c t to no to

o r d e r .I' inclusive.
As in 524.2, L e m m a 4, we c a n show that t h e s e p a r t i a l d e r i v a t i v e s ,
t r e a t e d as functions of F (i.e., for constant no), s a t i s f y the following s y s t e m
of differential equations: *
BY (lo), f (0; 0, no)= i t , . Therefore,
Equalities (15) provide the initial conditions f o r equations (14).

Proceeding a s in 924.3, we introduce the notation
( k = 2 , ..., 'V, the functions Ek are defined by equations (14)).

Since n = 0 is a solution of equation ( 7 ) , we have f (s; 0, 0) 0 ,
Therefore,
( k = 2, 3, . . ., N ; the missing t e r m s i n ( 2 0 ) correspond to a polynomial i n

the functions R2 (s), R 3 (s), . . ., Rk-l (s) and the functions ul (s), u2 (s), . ., uk-,(s)). .
The expressions for E f . E,, . . ., EN are not given i n explicit form; see 924.2, footnote to Lemma 4
269
ch.X. CREATION OF CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
Let n = 0 i n (14). W e will integrate them successively using initial

conditions (15) and the relations
f (s; 0, 0) = 0 and R (s, 0) 0
together with (16)-(20). The f i r s t of these equations h a s been considered
in Chapter V (513.3). Its solution is
( s e e 513.3, (25)).
1
Multiplying the k-th equation i n (14) by and integrating, w e find
F r o m ( l l ) , ( 2 1 ) , and (22) we obtain the following e x p r e s s i o n s f o r the

derivatives of the succession function a t the point no = 0:
Rl(*)d&
f(O)=eO , (23)
&k) (0) = f ( k ) (0)= k!ed R1()dr Hk (S) e (24)
Let u s now e x p r e s s the functions R k (s) and Hk (s) in t e r m s of the right-hand

s i d e s P (2, y) and Q ( x , y) of s y s t e m (A). The e x p r e s s i o n
w a s computed in Chapter V . Indeed, differentiating the function R (s, n) d e -

fined by ( 7 ) with r e s p e c t t o n ,
w e obtained a f t e r s o m e manipulations
w r i t e out in explicit f o r m only those t e r m s which contain k-th o r d e r

derivatives of P and Q . Differentiating ( 7 ) k t i m e s with r e s p e c t to n and
270
using the symbolic power notation,:'. we obtain, as is readily verified,
;vhere the t r i p l e d o t s correspond to t e r m s which contain no k -th o r d e r

d e r i v a t i v e s of P and Q ( t h e s e t e r m s confzili the functions P and Q and t h e i r
d e r i v a t i v e s to o r d e r ,<k - 1 I . L et n = 0 i n (261. By ( 3 ) ,
'p(s. O)=Cp(s), S(s, O)=*(s),
2 (s, 0 ) = q' (s), -3; (s. 0) = $' (s),
6, (s, n) = t'(4, tnl (s,n) = -Cp' (s).
hioreover, s i n c e 'p and 9 a r e solutions of s y s t e m (A), we have
9' (SI P ((P (SI 7 $ ( s ) ) $' ( s ) = Q (F (SI t t (SI)

r h e r e f o r e , for n = 0 , the n u m e r a t o r of the fraction before the b r a c e s in (26)
L-Linishes, and w e find
and the t r i p l e d o t s r e p r e s e n t t e r m s containing the functions P and Q and

t h e i r d e r i v a t i v e s to o r d e r k - 1.
By \16)-(22) and (28)--(27),
,where cbk (s) i s e x p r e s s e d - with the aid of a l g e b r a i c o p e r a t i o n s and i n t e -

gration - in t e r m s of the functions P and Q and t h e i r d e r i v a t i v e s to o r d e r
(12 - 1) and in t e r m s of the functions (s), 9 (s), q' (s), 9' (s).
ud is used as I n
If f ( x , y) is a function of two variables, the symbolic binomial power
(u xI'Jy)
a h h e v i a r e d nofarLon for the operator defined by the cquallty
271
Ch.X. CREATION OF CLOSED PATHS FROM A hlULTIPLE LIMIT CYCLE
We can now compute the derivatives of the succession function f (no) a t

I
the point no = 0 . F r o m (23) and (25) it follows (seeing that rp and rl,> and
therefore their derivatives, a r e periodic functions) that
s
T
[&-?(a), Ws))+Q&W, $(.))la*
f' (0) = e a (30)
(This expression was originally derived in Chapter V, 513.3, (30).)

F r o m (21), (23), (24), and ( 2 9 ) we have
d(k)(0) = f W (0) =
I
where k = 2. 3. . . .. N. I
r o l e in the proof of the fundamental theorem of t h i s section. , and
Wh (s) a r e e x p r e s s e d by (30), (25), and (28), respectively. We should again
emphasize that the second t e r m in the right-hand s i d e of (31) does not con-
tain derivatives of P and Q of higher than (k - i)-th o r d e r , and the values of
these derivatives a r e taken a t points of the c u r v e z=rp ( x ) , y = rl, (s).
2. Multiplicity of a l i m i t cycle
In Chapter V we defined the m u l t i p l i c i t y o f a l i m i t c y c l e f o r

the c a s e of analytical s y s t e m s ( s e e 812.3). We w i l l now define this concept
f o r s v s t e m s of c l a s s N. Let
be a dynamic s y s t e m of c l a s s N,i, Lo a closed path of this s y s t e m ,

x = rp ( t ) , y = rl, ( t ) the motion corresponding to this path, T > 0 the period of
the functions rp and 9 .
We r e c a l l that the path Lo is a limit cycle i f it is a n isolated closed path.
A limit cycle Lo is said to be m u l t i p l e ( s e e 513.3, Definition 18) i f
otherwise the limit cycle is s i m p 1e .

A s we have repeatedly mentioned, if (A) is a dynamic s y s t e m of c l a s s N ,
the function f (no)and hence the function d (no). for sufficiently s m a l l n o r have
derivatives to o r d e r N inclusive; in p a r t i c u l a r , the numbers d' (0). d" (0), . ..
. . ., d'"(0) e x i s t . We have seen in 812.4 that if a t l e a s t one of these numbers
does not vanish, the closed path Lo is isolated, i.e., it is a l i m i t cycle
(stable, unstable, or semistable).
D e f i n i t i o n 28. A closed path Lo of a dynamic system (A) of class N
is said to be a limit cycle of multiplicity r (man r-tuple limit cycle) if
d' (0) =d" (0) = ...=d"-" (0) = 0, d'r" (0) # 0 (33 1
( r is a natural number, r , < N ) .
272
926. Dl~RIV.\TIVES CY T H E SUCCESSlON FLXCTIOX
.This concept of mult..plicity c l e a r l y coincides f o r analytical s y s t e m s with

the definition introduced in S12.3. A s in 512.3, i t involves a c e r t a i n arc
without contact 1 , and w e have to prove that the definition is independent of
the p a r t i c u l a r choice of this a r c . * However, w e will not give this proof
h e r e , since i t will e m e r g e as a d i r e c t consequence of the fundamental
theorem of this section.
I t follows from Definition 1 8 (913.3) and Definition 2 8 that any s i m p l e
limit cycle is a limit cycle of m u 1 t i p 1 i c i t y 1 and vice v e r s a . With
r e g a r d to multiple cycles, the situation is as follows: if (A) is a n analytical
s y s t e m and L o is a mult:Lple limit cycle of (A), i t h a s a definite multiplicity r ,
where r may be any natural number g r e a t e r than 1.' If (A) is a s y s t e m of
,'
c l a s s :Y?.2 and L o is a multiple limit cycle of (A), i t generally h a s a definite

multiplicity r , 2 , < r , < S . In s o m e cases, however, i t may turn out that f o r
the cycle Lo
d' ( 0 )=d" ( 0 ) i ... = d'x-" (0) = d"" (0)= 0. (34)
If \ A ) is not a s y s t e m of class .Y -- 1, Definition 2 8 becomes meaningless

and the multiple limit cycle L o h a s no definite multiplicity.
Lf a closed path L o is a l i m i t cycle of multiplicity r , i.e., if relations ( 3 3 )
a r e satisfied, Maclaurin's formula gives
Hence, using the r e s u l t s of S12.2, we conclude that a l i m i t c y c l e o f

e v e n m u l t i p l i c i t y i s s e m i s t a b l e , a n d a l i m i t c y c l e of o d d
m u l t i p l i c i t y i s s t a b l e o r u n s t a b l e . For o d d r , ifd'"(O)<O, the
cycle is stable, and if d"' (0) > 0 , the cycle is unstable.
Let u s consider a spccial modified s y s t e m , f o r which a l e m m a analogous
to L e m m a 7 of 524.3 w i l l be proved.
Let (A) be a dynamic s y s t e m of class ,V. Lo a closed path of (A),
I = 'p ( t ) . y = 10 ( t ) the motion corresponding to this path ('p and $ a r e periodic
functions of period T > O ) . By L e m m a 1, S15.1 and R e m a r k 1 to that l e m m a ,
t h e r e e x i s t s a function 1' (t,y) of class .V L 1 in region G where s y s t e m (A)
i s defined, such that f o r all s , - 00 < s < L m , we have
i35)
and
(36)
Q e footnote on p. 108.
'I
If ( A I is a n analytical system, L p i s a closed path of ( A ) , and a l l the derivatives vanish, d ' k )(0)=
:= 0 tk = i . 2. . . .), we have (I ( n o ) 0 and a l l the paths clcse t o L o are closed paths. i.e., L o is no
longer a limit c y c l e . See 12.:3.
273
Ch.X. CREATION OF CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
On the other hand, differentiating identity (35) and i n s e r t i n g s = s o , we get

F; ((P ( S O ) , 9 ( S O ) ) 9' (SO) +F; ((P (SO). 9 ( S O ) ) 9'(SO) = 0
Consider the l a s t two r e l a t i o n s as l i n e a r homogeneous equations f o r
F; (cp (so), 9 (so)) and F; (cp (so), 9 (so)). It follows f r o m (36) that these equations
have a nonzero solution. But then the determinant of the s y s t e m vanishes,
[cp' + 19'(so)]' = 0 , i .e., cp' (so) = P (cp (so), 9 (so)) = 0 and 9' (so) = Q (cp (SO),
9 (so)) = 0. This is impossible, s i n c e the point (cp (so) 9 (sp))l i e s on a closed
path Lo and is t h e r e f o r e not a n equilibrium s t a t e . Relation (37) is thus
proved,
Together with system (A), consider a modified s y s t e m of p a r t i c u l a r
form
ds
-=
dt P" ( x , Y) = P (5,Y) + L F ~ F ; , $= Q(G Y) = Q (2,Y) + ~F*FI, (Xi,)
where h is a p a r a m e t e r , and m is a n a t u r a l number. Clearly, (A,) is a l s o
a s y s t e m of c l a s s N , and i f h is sufficiently s m a l l , (A,) is as c l o s e as
d e s i r e d to s y s t e m (A).
It follows f r o m (35) that a c _ l o s e d p a t h Lo o f s y s t e m ( A ) i s
a 1 s o a p a t h o f s y s t e m (A,). Indeed, since Lo is a path of (A), we
have
T' ( t )= P (T (t)t 9 ( t ) ) , $' ( t )= Q (9 ( t ) , 9 ( t ) ) .
F r o m these equalities and f r o m (35) it follows that

T' ( t ) = p"(T ( t ) ? $ ( t ) ) , 9' (4 = Q(cp (th 9 ( t ) ) 1
which i m p l i e s that Lo is a path of s y s t e m (xA).

Let
-
f = 7(no,a) (38)
be the s u c c e s s i o n function constructed f o r s y s t e m (6,)on the s a m e a r c
without contact I and f o r the s a m e choice of the p a r a m e t e r as the succesEion
function f o f o r s y s t e m (A) (for sufficiently s m a l l A , the function 7 is defined;
s e e 54, L e m m a s 1,2, and 11).
Clearly,
-f (no,0) = f (no).
(39)
Moreover, since Lo is a l s o a path of s y s t e m (A,), we have
c
f (0)=7(0, 1)=o. (40)
Together with the function f"(no),we a l s o introduce d"(no)= a ( n o , h ) , analogous
to the function d(no)of (A),
-d (no)= d d
(no,h) = f (no,h ) - n ~ = f (no)-no*

I
(41)
In topics r e l a t e d to the c r e a t i o n of l i m i t c y c l e s from- a closed path L o , the

s y s t e m (A,) plays p r e c i s e l y the s a m e r o l e as s y s t e m (A,) (024.3) in
connection with the c r e a t i o n of l i m i t c y c l e s f r o m a focus.
L e m m a 1 . Let fm a closed path Lo of a d y m m i c system (A) of class
N>1
d' (0) = d " ( 0 ) = . . . = d'" (0)=o,
(42)
5438 274
where 1er<S, and let m be an integer,
i<m::r, (43)
and 3. a nonzeYo veal nvtnbev. Then the closed path Lo is a tntiltiple littiit
cycle of mdtiplicity m .for the system Le., (xi,),
2 (0) =d""(O) =. . . =a(m-')(0) =o. (441
and
Also
a ( " ' ) ( o ) = ~ . m[ q! ' ( 0 ) 2 + ~ ' ( O ) 2 1 " - 1 ~
q IF; (q (s), * (4)he'((4S P T 9 ' Wlm%-

9' -F; ('0 (s). ( J ) ) B' (S)l"+'
X
'
(m-1) [p.&(s), t ( s ) ) f Q ( w ) , WS))~ dr
r e O ds. (46)
P r o o f . Since L o is a path of s y s t e m (Kk),
the d e r i v a t i v e s gi)(O),
i = 1, 2 , . . .. m, may be found using (30) and ( 3 1 ) (for the d e r i v a t i v e s d(i) (0)).
In t h e s e equations, P (2,! I ) and Q (I,y) and t h e i r d e r i v a t i v e s should be r e p l a c e d
with (3, y) and Q ( x , y) and t h e i r d e r i v a t i v e s ( t h e functions v, 9, q', rt;' c l e a r l y
r e m a i n a s before).
& e recall that i n computations using (30) and (31), the v a l u e s of the
functions P" and p and t h e i r d e r i v a t i v e s should be taken for x = (p (s), y = $ (s).
o,<S,<T.
F r o m the r e l a t i o n s
i t follows, as will be s e e n f r o m e l e m e n t a r y calculations, that at the points

(cp (s), tp (s)) the v a l u e s of tke functions and 0 and t h e i r p a r t i a l d e r i v a t i v e s
to o r d e r m - 1 inclusive a r e r e s p e c t i v e l y equal to the v a l u e s of the functions
P a n d Q and t h e i r d e r i v a t i v e s to o r d e r m - 2 , i.e.,
P(cpN79 ( s ) ) = P ((P (s), 4 (41,

-Q('F(49 (SI) = Q ('P (4,9 (4)t
( l = i , 2, ... ) m-i; O < i < Z ) . v=pc*) J

L e t u s now compute the m -th o r d e r p a r t i a l d e r i v a t i v e s of the function P"
and Q a t the points (cp (s),9 (s)). In v i r t u e of the r e l a t i o n F ((p (s), 9 (s)) 0, i t
275
Ch.X, CREATION OF CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
suffices to consider only those t e r m s which r e s u l t f r o m the differentiation

of the functions P,Q, and F" (all the o t h e r t e r m s produced by differentiation
contain the function F (2,y) a s a multiplicative factor and t h e r e f o r e vanish
f o r z = 'p (s), y = #I (s)). Simple manipulations t h u s l e a d t o t h e r e l a t i o n s
u=*(s) u=W)
+
Y'W.)
1
}
( i = O , 1, 2, ...,m). J I
L e t fii (s), pi (s), etc., be the functions of (x,)
which are analogous to the
functions R i ( s ) ,H i @ ) , etc., of (A). Relation (25), the analogous relation
for R, (s), and equations (48) lead to the equality
(s) = R* (s). (50)
Thus, using (231, we s e e that
P (0)= f' (01,
and t h e r e f o r e
-d' (0) =a' (0).
F u r t h e r m o r e , using (21), (28), (29), (31), the analogous r e l a t i o n s f o r (xA),
and also equations (48), (50), and (51), we conclude that
@k (s) = Wa (s) .(53)
for k 6 m - I ;
&k (s) = @h (s) (54 1
for k x m ;
fi, (S) = H k (s) (55)
for k g m - 1 , and finally,
d
"(k) ( 0) = d'k' (0) (56)
I f o r l i - 2 . 3, ..., m - I .
From (42), (43), (52), and (56) i t follows that
- (0)= . . . -
I W e have thus proved relation (44).
d' (0)= d - P-i)
(0) = 0.
L e t u s now compute a ( " ( 0 ) . To this end, we will u s e equation (31),

replacing f , W , R , e t c . , w i t h x w,
E , e t c . By ( 5 0 ) , (51), and (54),
s - s R~(.)dr
I 8
1 R,,,
Ri(a)d8
\
5 (k-i) T
(0) = f' (0) (s) e 0 ds+m!f' (0) @ k ( s ) e O ds. (57)
1
0 0
'The_ expression f o r 6f"(s) c a n be found using (28). Replacing P and Q with
p a n d Q and using (49), we obtain
227. CRE;ITIO\; OF L i h i i r CYCLES mox A MCLTIPLE Liwr CYCLE
From ( 3 1 ) and (58) w e obtain
By the conditions of the l e m m a , d i m f ( 0 ) - 0 . F u r t h e r m o r e , d ' ( 0 ) - f ' ( O ) - I -- 0

i . e . , i'(0j = 1 . Inserting these numerical values and expression (25) in the
l a s t relation, w e finally obtain
2"' (0) = ).. m ! [a' ((I)? $' (0)%]"'-1,.
w (9). $ t S ) ) $' W-FFy (V 9 W)p* (311q'-1
:. 1-IF;
[ff'(s)~-$'(s)qm
(S)?
\'
im-ib 5 jF;qs~s).*(*)I e Q L f ~ ! s ) ,+ i s l i ] d ~
,. e u ds.
i . e .., eaualitv
. (46)holds true.
Inequalit; (45) follows directly f r o m (46)and inequality ( 3 7 ) . This c o m -
pletes the proof of the l e m m a .
R e m a r k . M'e have a s s u m e d in the l e m m a that r>.)(an_d thus ,V>2) and,
imposing conditions (212:', considered the modified s y s i e m (A,) with m > 2 .
Cinder these conditions, the closed path Lo of s y s t e m (A,) is a multiple lim2t
cycle of multiplicity m . The case m = 1, i . e . , when the modified s y s t e m (A,)
h a s the f o r m
h a s been considered in connection with the proof of Theorem 19 ($15.2).

ife have s e e n t h e r e that if (A) is a dynamic s y s t e m of c l a s s S > 1 , Lo a
closed path of ( A ) for wkich d' (0) = O ( d " ( 0 )need not be equal to zeroL o r even
need not exist altogether), and i + 0 , L o is a simple limit cycle of (A),,)and
t h e corresponding derivative is e x p r e s s e d in the f o r m
$27. CREATION O F LIhlIT CYCLES FROM -4

h l U L T I P L E LIRIIT CYCLE
1. The fundamental theorem
In Chapter V we considered a multiple l i m i t cycle and showed that i t

may "create" closed paths (515.2, Theorem 19). In this section we w i l l
elucidate the number of paths that may be "created" in the neighborhood of
a multiple limit cycle on passing to sufficiently close s y s t e m s . W e will
in fact prove the following theorem, analogous to Theorem 40 of 825.1.
T h e o r e In 4 2 (theort?ni of the creation of littiit cycles f r o m a multiple
litnit c y c l e ) . If (A) is a ,dynaniic system of c l a s s .V > i or an atmlytical
system, and L o i s a multiple liniit cycle of mrcltzpllcitj k (2<k<.V), therz
277
(3.X. CREATION OF CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
1 ) there exist E , > 0 and 6, > 0 such that any s y s t e m (A) 6,-close to
rank k to (A) has at most k closed paths inUQ ( L o ) ;
2) f o r any positive e < e , and 6 < 6,, there exists a system (A) of class N
(of analytical class, respectively) which is 6 -close to rank k to (A) and
I
has k closed paths in U,, (Lo).*
P r 0 0 f . 1 ) Let u s prove the f i r s t proposition of the theorem. A s i n the
previous section, consider a n a r c without contact I , which is a normal to the
path Lo ( s e e 26.1), and the succession function f (no) of s y s t e m ( A ) on the
a r c 1 , together with the function d (no) = f (no) - n o . Let these functions be
I
defined for a l l no, I no I<.*, where n* is s o m e positive number. A s we have
noted in 526, d ( n o )is a function of c l a s s N.
Since Lo is a limit cycle of multiplicity k of s y s t e m (A), we s e e that
I d' (0) = d" (0) = . .. = d'"" (0) = 0, d(h)(0) # 0, (1)
i.e., the number 0 is a root of multiplicity k of the function d ( n o ) . r h e r e f o r e

( s e e Chapter I, S1.3) t h e r e exist positive n u m b e r s q < n * and u such that any
function 2 ( n o )defined f o r all n o , I no I 4 n * , and 0-close to d (no) to rank k may
have a t most k r o o t s on the segment L- q, 9).
FIGURE 118
A sufficiently s m a l l positive number is taken for E,,, so that a l l the points

of the normal 1 lying in U, (Lo)correspond to the values of the p a r a m e t e r no
l e s s than q in magnitude ( F i g u r e 118). 80 is-also taken s o s m a l l that the
following condition is satisfied: if s y s t e m (A) is 8,-close to rank k to (A),
the succession function f"(no), and hence the function a(,,), a r e defined f o r
(A) on the a r c 1 for a l l no, no I < n * , and f o r Ino 1 4 n * the function 2 (no) is u -
close to d ( n , ) to rank k . This 6 , e x i s t s by Theorem 3 , Appendix, 1. The
numbers e , and 6, chosen in this way evidently satisfy the f i r s t proposition
of the theorem. The first proposition is thus proved.
2 ) The second proposition will f i r s t be proved for a dynamic system (A)
of c l a s s N . Consider a modified system of a p a r t i c u l a r f o r m
* The closed paths of Theorem 42 are clearly isolated, i.e., they are limit cycles.
278
where i ; a r e p a r a m e t e r s , and F (I,y) is a function of c l a s s h + 1,defined in
7; and satist'.ving conditions ( 3 5 ) and ( ~ 3 6of ) the p r e v i o u s section j S 2 6 . 2 ) .
fiy \ 3 3 / , the function F (2.y) \ranishes at the points of tlie l i m i t c y c l e L o ,Of
s y s t e m ( X Jand t h e r e f o r e L o is a l s o a path of s y s t e m ( X i . Evidently, !A)
i s also a s y s t e m of c l a s s .V.
F o r sufficiently s m a l l A,, t-q) is a r b i t r a r i l y c l o s e to ( A ) to any ( p o s s i b l e )
rank, and by S4, Leninias 1,2, and 11, the function
. . .. A+!).
I
d ( ~ , , .i.!. >.?,
analogous to the function d (n,,)for ( A ) , is defined for on the a r c without (-x)

contact I for. a l l R,,, I no I < R * . .:< Clearly, 2 is a continuous function of r i Y and
o f the p a r a m e t e r s >.i. >.?. . . .. Since ( A ) is obtained f r o m (-5)for
i., - i.: E .. . . 2 ;.R.l = (?, .ve see that
-
d (rz,,. 0. U, . . ., V I 5 d ( R , , ) . (2i
f'or any E > 0 and 6 >0, there exist I* > 0 anit i &
s u c h~
that +
if
lAt[<k*, i = l , 2, .... k - - l , (3J
then
!a) systcni (x) is 6 -close to r a n k k to s y s t e m ( A ) ;
',b) the function d(n,,. i.!. i,:.,. ., I.*-,) is defined f o r a l l R ~ 1,no 1<n*, and any
root of this function satisfying the inequality I no 1 < k c o r r e s p o n d s to a
c l o s e d path of tLXJcompletely contained in 1 I E ( L o ) .
Let F and h be fixed positive n u m b e r s , e < e g . 6<:6,. Let f u r t h e r , i.* and
t i b e the n u m b e r s corresponding to e and 6 ( i . e . , s u c h that i f (3) is satisfied,
conditions t a ) and ( b ) hol-t t r u e ) . i V e Ixill show that by a n a p p r o p r i a t e choice
. . ., kS-,, s y s t e m (A) c a n be made 6-close to r a n k k
of the p a r a m e t e r s i I ,i.-,
tu { A ) and ,will have k closed paths in C', (Lo).
Lf'e :vi11 a s s u m e that the n u n i b e r s
).; henceforth s a t i s f y condition (3).
F r o m M a c l a u r i n ' s fo.rmula we conclude that, in v i r t u e of r e l a t i o n s ( 1 ),
t'ur a l l sufficiently s m a l l no,
where h (nnl is a continuous function and h (0) = 0 ( s e e proof of Theorem 5,

31.3).
By sssuniption, d ~ k ~ ( 0 ) = L OTo. fix i d e a s , let d(!+)(0)>0. Then for all
sufficiently s m a l l positive R ~ d(n,)
, > 0 . i V e choose one of these nunibei s ,
s m a l l e r than k, and denore it by n,. T h u s ,
o<n,.<n. d(n,)=d"(ni. 0, ... ) 0)>0. (5)
h o w suppose that
Ch.X. CREATION OF CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
I
and consider the n o d i f i e d system corresponding to these values of the
parameters,
dz
;i7- = P
- (5, y, 0, 0, .. s, 0, hk-1) = P (5, Y) +hk-lFk--'F;,
dy=
dt
-
Q ( 2 3 Y, 0, 0, .. . t 0, hk-1) = Q (5, Sr) +hi-iP-'FL, (A,)
and the corresponding function 2, (no)=z(no, 0, 0, . . ., 0, hk-1). By (1) and

L e m m a 1 of the previous section,
-
?..
d ; (0) = d; (0)= . .. = Z i k - 2 )
and & k L 1 ) ( 0 )is e x p r e s s e d by equality ( 4 6 ) of the previous section, which
(0) = 0,
(6) I
gives
-d l k - ' ) (0)=
where Cl is a nonzero constant (C, is the factor before h i n ( 4 6 ) f o r
Clhk-1,
I
m = k - 1. I t s explicit e x p r e s s i o n will not be needed h e r e , however).
A s before, Maclaurin's formula and r e l a t i o n s ( 6 ) give f o r a l l suf-
ficiently s m a l l no
+h", (no)nt-1,
I
dl (no) =
where ch, (no)is a continuous function and x,(0) = O .

clik-1
nt-' (7)
I
Let cl>o and choose ha-1 so that
1 hk-11 <A*, hk-i <0, d", (ni) = a(ni, 0, 0, . - 9 0, Lk-1) >0. (8)
The l a s t of these conditions is satisfied f o r any sufficiently s m a l l hk-1 in

virtue of equality (5) and the continuity of the function 2(no,icl, a2, .., h k - 1 ) . .
The inequalities Cl > 0, hk-1 < 0 and equality ( 7 ) show that f o r sufficiently
s m a l l positive no, & (no)< 0. Choose one o f t h e s e numbers, s m a l l e r than ni,
and denote i t by nz. Thus,
O< n2<n1 <G (91
and
-
di (ni) >0, gi (nz)<0- (10)
F u r t h e r construction p r o c e e d s along the s a m e lines a s above (compare
with proof of the second proposition of Theorem 40, 5 25.1). After the
(k- 1)-th step, * we end up with the s y s t e m
*=P(%,
dt y,
- hi, hz, .., hk-1) =
=P (5, y) +&FF; + hzF'F; + . - + bk-iFk-'F;,
(A)
dy=a(&Y, hi, hz,
dt hk-i) =
= Q (5, y) f k,F& f- h,ppu + . +hk-,F'-'FI,

* - - .
-
In our construction of the systems (Al), (A2), . ., (Ah+) and the numberr n,, na, ., n k - i , we always ..
use Lemma 1 of the previous section. However, in the last, (k- I)-th step, the lemma itself is replaced
by the remark to Lemma 1 and formula (59)of the last section is used instead of formula (46).
280
927. CRMTIOh OF LIhlIT CYCLES FROhI .4 hlULTPLE LIAlIT CYCLE
and the n u m b e r s n,, n2. . . ., n R a r e s u c h that 1 I <I.*,
and
F r o m inequalities ( 1 2 ) and the continuity of the function 2 (no)i t follows

that a t l e a s t one r o o t of the function 8 (no)falls between each p a i r of n u m b e r s
R, and n2, n2_and n 3 , . . ., i t k - , and nk. These r o o t s c o r r e s p o n d to closed paths
of s y s t e m (A). By ( 3 ) and (ll), these closed paths lie in C', ( L o ) . Ill_oreover,
the path Lo of ( A ) is itself a path of ( 3 ) . Thus, a t l e a s t k paths of (A) e x i s t
inside the neighborhcod U ,( L , ) . Since E < E, and 8 < &, the f i r s t proposition
of the theorem i n d i c a t e s that C; ( L a ) may contain a t most k paths of (A),
i.e., it contains preciseLy k paths. This p r o v e s the second proposition of
the theorem f o r s y s t e m s of class N .
3 ) Let u s now c o n s i d e r the analytical c a s e . Our proof is inapplicable
to this c a s e , since, in general, no analytical function F (I, y) e x i s t s in the
e n t i r e 6 which satisfies conditions ( 3 5 ) and ( 3 6 ) of the p r e v i o u s section.*
W'e will t h e r e f o r e proceed along the s a m e l i n e s as in the proof of
T h e o r e m 19 (S15.2).
Let (A) be an analytical s y s t e m , and L o a multiple l i m i t cycle of
multiplicity k of (A), E,, and 6o the n u m b e r s introduced in the f i r s t p r o p o s i -
tion of the t h e o r e m . Take any pzsitive n u m b e r s E < eo and 6 < 6 0 . As
before, we c o n s t r u c t a s y s t e m (A) of c l a s s S g k and n u m b e r s nl, n2, . . ., nk
such that the following conditions are satisfied:
(a) System (A) is 6 / 2 c l o s e to r a n k k to s y s t e m (A).
( b ) Relaticns (11) and (12) are satisfied, and h i s a number with the
following p r o p e r t i e s : any path E of s y s t e m (A) c r o s s i n g the a r c without
contact I a t point .Ql corresponding to a value no, I no I h, of the p a r a m e t e r
c_ro_sses the arc I with i n c r e a s i n g t at a n o t h e r point so that the arc
M i M 2 of the path is e n t i r e l y contained i n Ur,*( L o ) .
By equation (59) of the preceding section, z-(O)=#=0 , i.e., L o is a s i m p l e ,
and hence s t r u c t u r a l l y stable, l i m i t cycle of (A).
Let q be a number, 0 .= q <Tb , and ( A) a n analytical s y s t e m q - c l o s e
to rank k to s y s t e m (A) (e.g., a s y s t e m whose right-hand s i d e s ar_e
polynomials adequately approximating to the right-hand s i d e s of (A)).
C l e a r l y , if q is sufficiently s m a l l , the following conditions a r e satisfied:
(a' ) ?e function d* (no)corresponding to {A* ) is defined for all
no, I no I < n, and
d' (ni)>0, d* (nz)<0, .. ., d' (.A)

I > O i f k is odd,
< O if k is even.
(b;:<)Any path L* of s y s t e m (A':) c r o s s i n g the a r c without contact 2 a t
point .Tf; corresponding to the value n o , 1 no I <h, of the p a r a m e t e r c r o s s e s
W e have seen in Chapter V (915.21 that such a function may be connructed in the neighborhood of the
path L o This is not sufficient, h,wever, since ig should be defined in the entire d .
28 i
Ch.X. CKEATION OF CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
the a r c 1 again with increasing t a t point M:, s o that the a r c M:M: of the
path L* is entirely contained in U , (Lo).:.
(c*) T h e r e e x i s t s a closed path L: of (A:) which lies entirely in U , ( L o )
and c r o s s e s the a r c without contact 1 a t point M : corresponding to the
value n: of the p a r a m e t e r , where n: < n, (condition (e:) is satisfied for
sufficiently s m a l l 1 because Lo is a s t r u c t u r a l l y stable limit cycle of (A)).
F r o m (a.), (b), and (c.) i t c l e a r l y follows that (A*) is a n analytical
system, 6 -close to rank k to s y s t e m (A), which h a s a t l e a s t k closed paths
in U , (Lo),i.e., i t s a t i s f i e s the second proposition of the theorem. This
completes the proof of the theorem.
R e m a r k . If (A) is a system of c l a s s E , the condition - in the second
proposition of r h e o r e m 4 2 - that s y s t e m (A) is 6 -close to rank k to (A)
may be replaced by requirement of 6 -closeness of rank N . If (A) is a n
analytical system, we can find a n analytical s y s t e m (A) satisfying the
second proposition of the theorem and yet close to (A) to rank m , where m
is a n a r b i t r a r y natural number. The validity of this r e m a r k follows directly
from the proof of Theorem 4 2 .
2. Supplements
O n t h e c r e a t i o n o f l i m i t c y c l e s f r o m a f o c u s of f i n i t e
m u 1 t i p 1 i c i t y . Theorem 4 2 and the r e m a r k following the theorem enable
u s to strengthen the second proposition of Theorem 41. Indeed, the
following theorem obtains :
T h e o y e m 41 . Let 0 (0, 0 ) be a multiple focus of multiplicity k of a
dynamic system (A) of class N > 2k + 1 (or analytical), and let e, and 6o be
sufficiently small positive numbers (introduced in the f i r s t proposition of
Theorem 40 and the remark to that theorem). If system (B) i s 6,-close
to yank 2k f 1 to system ( A ) , the sum of the multiplicities of the focus
and the limit cycles of (B) lying in U,, (0)is at most k .
P r o o f . Let the s u m of the multiplicities of the focus and the limit
cycles of (B) lying in U,, (0)be k* > k. Then, using the construction of
L e m m a 2, 815.2, and T h e o r e m s 4 0 and 42, together with the r e m a r k to
Theorem 42, w e can modify ( B ) in the neighborhood of each of these l i m i t
c y c l e s and the focus, to obtain a s y s t e m (B.) which is a r b i t r a r i l y close
to rank 2k + 1 to s y s t e m (B) and h a s k* > k closed paths in U , (0). The
existence of this s y s t e m (B ), however, contradicts the condition that
0 (0, 0 ) i s a focus of multiplicity k . If only analytical s y s t e m s are considered,
the limit cycles of (B;:) lying in U,,(0)should f u r t h e r be made s t r u c t u r a l l y
stable (see $15.2, L e m m a 2 ) and (B;:) should be approximated with a n
analytical s y s t e m . The theorem is thus proved. It is clear that the second
proposition of Theorem 4 1 follows f r o m Theorem 41.
B i f u r c a t i o n s of a d y n a m i c s y s t e m i n t h e n e i g h b o r h o o d
of a l i m i t c y c l e o f f i n i t e m u l t i p l i c i t y . W e will f i r s t consider
a proposition which is analogous to Theorem 4 1 and strengthens the second
proposition of Theorem 42. This proposition, together with Theorem 42,
plays a fundamental role in the e n t i r e topic of bifurcations of dynamic
s y s t e m s in the neighborhood of a limit cycle of finite multiplicity.
* For a s m a l l q , condition (b*) is satisfied because of condition (b) and Lemma 11, 04.2.
282
$27. CRE.ATIO?: O F LI?,!If CYCLES FRO!.: .A ALC'LTIPLE LI.\!IT CYCLE
T h e o r e m 4 3 . Let (A) be a dyimrnic system of class S > 1 or an

analytical system, L o a uiultiple lirnit cycle of niitltiplicity k ( 2 g k 4 . Y ) of ( A ) ,
ard eo and h0 sufficiently small positir*eniirribers (introduced in the f i r s t
psoposition of Theorerii 42).* Theti
I) f o r a n y ~ a r ~ 6 . O . : E , < e ~ , O ( 6 ~andfovany
6~, s , I - s s < k , there
exists a system (B)of class s (or an analytical system) which is 6 -close
to rank k to ( A ) and has precisely s closed paths in L-e ( L J :
2) if (B)is 8,-close to rank k to (A), the sum of the mdtiplicities of all
the limit cycles o f ( B ) lying in ( L o ) is at wost k .
The proof of the first. proposition is analogous to the proof of the c o r r e -
sponding proposition of Theorem 41. The proof of the second proposition
is analogous to the proof of Theorem 41'. The
r e a d e r w i l l be able to r e c o n s t r u c t the detailed
rz proof without difficulty.
The investigation of the bifurcations of a
dynamic s y s t e m in the neighborhood of a limit
cycle of finite multiplicity is analogous to the
investigation performed a t the end of 525.2 f o r
a multiple focus. T h e o r e m s 4 2 and 43 play a
leading r o l e in the e n t i r e treatment. Let Lo be
a /c -tuple l i m i t cycle of s y s t e m (A) ( k ' > 3 ) , )'a
sufficiently s m a l l neighborhood of this cycle
bounded by c y c l e s without contact and r2.
6 a sufficiently s m a l l positive y m b e r . By
Theorems 1 2 and 43, s y s t e m (A)G-close to
rank 12 to (A) may have a t most k closed-paths
i n V . hIoreover, t h e r e exist s y s t e m s ( A )
F l G l i R E 119
which have p r e c i s e l y s closed paths in V , where
s is any number, l , < s , < k . These closed paths,
naturally, are l i m i t cyc12s and a r e a r r a n g e d "concentrically" ( F i g u r e 119).
As in the case of a multi?le focus (25.2), the topological s t r u c t u r e of ( A )
in I' is entirely determined by the number s of the limit c y c l e s lying in T7 and
t h e i r stability c h a r a c t e r i s t i c s . Let these c y c l e s be L 1 ,L 2 , . . ., L , , and w e
a s s u m e that Li l i e s inside L L - l ( i= i , 3 , . . .. s - 1 ) . Suppose that the behavior
of the paths of ( A ) in relation to the c p l e s without contact rl and is
known.' The paths of any s y s t e m ( A ) sufficiently c l o s e to (A) behave in
relation to rl and r2 just like the paths of (A). T h e r e f o r e the topological
s t r u c t u r e of in 1- is completely detezmined if we know:
( a ) the number s of l i m i t c y c l e s of (A) in J7 ;
(bj whether each of these c y c l e s is of even o r odd multiplicity.
Hence i t follows, a s f o r a multiple focus, that in the neighborhood of
a limit cycle of finite mu:tiplicity a dynamic s y s t e m ( A ) may only have a
finite number of different bifurcations. W e w i l l not t r y to d e s c r i b e these
bifurcations, since the situation is p r e c i s e l y the s a m e as for a focus (525.2).
E: x a m p 1e 10. Consider the s y s t e m
.
where k is a natural number.
* I t is further assumcd that a l l systcms ao-close t o ( A I have no equilibrium states i n U,.,( L O )
' 9
In other words, i t is known w h c t t r r thc pathi ~ r ~ s s thc
i n cycle
~ ri (i= 1, 2) enter into V d r leave Vas
t m-rraws.
283
0.X. CREATION O F CLOSED PATHS FROM A MULTIPLE LIMIT CYCLE
Direct computations show that this s y s t e m h a s a single equilibrium

s t a t e O(O,O), which is a n unstable focus f o r even k and a stable focus f o r
odd k. It is readily verified that the c i r c l e s
x' +'y = ca (14)
f o r C # 1 a r e c y c l e s without contact of s y s t e m [Bk)>
and the c i r c l e
x2 +'y =1 (15)
is a path of the s y s t e m . Hence i t follows that this c i r c l e is the only closed
path of the s y s t e m , i . e . , i t is a l i m i t cycle. W e w i l l u s e the symbol Lo to
designate the c i r c l e (15). All the paths of (Bk),except the focus 0 and the
l i m i t cycle L o , a r e s p i r a l s . Since
- - - i apa
x x + yy = 2 --=pa
df
(pS- 1)k,
the infinity is absolutely stable. These data uniquely d e t e r m i n e the

topological s t r u c t u r e of the dynamic s y s t e m ( B k ) . F o r a n even k , the l i m i t
cycle Lo is s e m i s t a b l e , and f o r a n odd k i t is unstable. The path configura-
tion is shown schematically in F i g u r e 1 2 0 (even k ) and in F i g u r e 121 (odd k).
FIGURE 120. For even k, FIGURE 121. For odd k ,

the focus is unstable and the the focus is stable and
limit cycle IS semistable. the limit cycle is un-
stable.
We will now show that the path Lo is a cycle of multiplicity k for (Bk).
This follows d i r e c t l y f r o m L e m m a 1, 5 2 6 . 2 . For the s t a r t i n g s y s t e m (A)
of the l e m m a we take the s y s t e m
dz
-=
and s e t
F (x, y) = ;c2 + ya- 1.
The n u m b e r s m and h are assigned the v a l u e s k and '/2. Then s y s t e m (A,)

introduced in L e m m a 1, S26.2, coincides with ( B k ) . The paths of
s y s t e m (16) a r e the equilibrium s t a t e 0 (0,O)(a c e n t e r ) and the concentric
c i r c l e s x = C c o s t , y = C s i n t , including the c i r c l e L o . Therefore, f o r
284
527. CRE.1TLOX OF LIhlIT CYCLES FROhl A hIULTIPLE LUIIT CYCLE
s y s t e m (16), the s u c c e s s i o n function is f (po) po. and d (po) =_ 0 . But then

conditions (42), S26.2, a r e satisfied for any r . The following conditions
are additionally satisfied on the path L o :
F (5, y) E 0, (Ti)' + (F")C# 0

(conditions ( 3 5 ) and (36), 26). I'hus all the conditions of L e m m a 1, 826.2
are satisfied, i.e., c i r z l e Lo is a l i m i t cycle of multiplicity k of (Bk).
The c a s e k = 2 w a s :onsidered in detail in Chapter VI11 ( S 2 2 , Example 8).
In that example we established the exact changes i n the topological s t r u c t u r e
of the s y s t e m as i t s field rotated and elucidated the fate of the l i m i t cycle L o .
285
C h a p t e r XI
CREATION OF LZMZT CYCLES FROM THE

LOOP OF A SADDLE-POINT SEPARATRIX
INTRODUCTION
Consider a dynamic s y s t e m (D) with a simple ( s t r u c t u r a l l y stable) saddle

point 0 ( x o . yo) and.a path Lo which goes to the saddle point 0 both for t + - 00
and for t + + 00. T h i s path is both a n a - s e p a r a t r i x and an o - s e p a r a t r i x of
the saddle point 0 , and we s a y that it f o r m s a loop. In Chapter IV
( T h e o r e m 16, 11.2) i t is shown that a s e p a r a t r i x forming a loop is a
s t r u c t u r a l l y unstable path and t h e r e exist modified s y s t e m s a r b i t r a r i l y
close t o (D) and such that the s e p a r a t r i x loop d i s a p p e a r s on moving to
these systems.
In the present chapter we consider the creation of closed paths from
a s e p a r a t r i x loop upon moving t o c l o s e s y s t e m s .
The f i r s t of the two sections, 28, p r e s e n t s auxiliary background
m a t e r i a l . The succession function on an a r c without contact c r o s s i n g
the s e p a r a t r i x loop is considered in S28.1, and s o m e p r o p e r t i e s of t h i s
function a r e established. A l l t h e p r i n c i p a l r e s u l t s o f t h i s
chapter are derived in what follows using this succes-
sion function . The behavior of the saddle point and i t s s e p a r a t r i c e s
on moving t o c l o s e s y s t e m s constitutes the subject of S28.2.
T h e principal r e s u l t s of the chapter are contained in $29.
T o fix ideas, suppose that two s e p a r a t r i c e s of the saddle point 0 which
do not belong t o Lo lie inside the loop formed by the s e p a r a t r i x Lo. W e
first prove ( T h e o r e m 44, 129.1) that i f the p a r a m e t e r
00 (zo. Yo) = P
z (% Yo) +Qk (% Yo)
is positive (negative), the loop Lo is unstable (stable) from inside, i . e . ,
all the paths passing through points i n t e r i o r to the loop which a r e sufficiently
c l o s e t o the loop g o t o t h i s loop f o r t + - 00 ( t + +00).
We f u r t h e r consider the creation of a closed path f r o m a s e p a r a t r i x loop

( T h e o r e m s 45 and 46, f 2 9 . 2 ) . It is established that i f the s e p a r a t r i x loop
is stable or unstable (in particular, if uo # 0), t h e r e exist modified s y s t e m s
a r b i t r a r i l y close t o the original s y s t e m such that the loop d i s a p p e a r s on
passing t o any of t h e s e close s y s t e m s , and yet a t l e a s t one closed path is
c r e a t e d in any a r b i t r a r i l y s m a l l neighborhood of the loop ( F i g u r e 122).
In S 29.3 it is proved that i f uo # 0, a s e p a r a t r i x loop w i l l c r e a t e a t
m o s t one closed path in a sufficiently s m a l l neighborhood of itself. For
the creation of the closed path it is n e c e s s a r y (but not sufficient) that the
286
$28. AUXILI.4RY hlATERIAL
s e p a r a t r i x loop disappear. If a closed path i s c r e a t e d on moving t o a c l o s e

s y s t e m , t h i s closed path is a limit cycle of the s a m e stability a s the stability
of the disappearing s e p a r a t r i x loop ( T h e o r e m s 4 7 and 48).
FIGURE 122
T h e c a s e IT,,= 0 is considered in 529.4. In t h i s c a s e , the uniqueness

theorem does not apply, i.e., t h e r e exist modified s y s t e m s a s c l o s e a s
w e d e s i r e t o ( D ) with at l e a s t two closed paths in any a r b i t r a r i l y s m a l l
neighborhood of the loop ( T h e o r e m 50).
In addition t o the above topics, 2 9 a l s o d i s c u s s e s the conditions when
the disappearance of a loop of necessity l e a d s , or conversely does not lead,
to the creation of a clo.;ed path in i t s neighborhood. T h e o r e m s 45 and 4 9
provide a comprehensi.re a n s w e r to t h i s question for the c a s e oo # 0 .
In Chapter XI we a r e dealing with analytical dynamic s y s t e m s only.
However, all the r e s u l t s r e m a i n valid for s y s t e m s of c l a s s n ( n > l ) , and
the proof is completely analogous t o that for analytical s y s t e m s .
A l s o note that the c l o s e n e s s of dynamic s y s t e m s in Chapter S I is
to b e understood, as always, in the s e n s e of c l o s e n e s s in s o m e fixed
closed region 8.
;I 2 8 . AUXILIARY MATERIAL
W e will present a number of l e m m a s which l a t e r on a r e actively used

in the proof of t h e principal proposition of t h i s c h a p t e r . Some of t h e s e
l e m m a s are contained in QT, but they are nevertheless reproduced h e r e ,
s o m e t i m e s without proof.
1. Correspondence function and s u c c e s s i o n function
be a dynamic s y s t e m , I , and l2 two arcs without contact of the s y s t e m which

have no common points, and
s=g,(u), Y=~,(u), a S ~ , < 6
281
Ch.XI. CREATION OF LIMIT CYCLES FHOM THE LOOP OF A SADDLE-POINT SEPARATRIX
and -
z=gz(i), y=h,(G), a<iig6
a r e the p a r a m e t r i c equations of the a r c s 1, and 1 2 , respectively. Let

g,, h , , g,, h , be functions of c l a s s 2 . A point on the a r c I , (I2) corresponding
t o the value u F) of the p a r a m e t e r w i l l be designated M (u) (or @ (z)).
Suppose that e v e r y path L of s y s t e m (D) which f o r t = to p a s s e s through
the point M (u) of the a r c 1, (a,<u,(b) will p a s s through the point @(u> of the
a r c l2 f o r s o m e other ?>to. Moreover, suppose that f o r to < t < f the path
L h a s no common points either with I , or with 1 2 . The p a r a m e t e r s i and u
a r e functions of u . We will designate them as x (u)and o (u) respectively:
-
?==X(U), u=o(u).
T h e functions x and o are defined for all u , a g u g b , and, as is shown in QT,

s 3 . 6 , R e m a r k 2 t o L e m m a 9, they a r e functions of c l a s s 1. The function
w ( u ) i s called the c o r r e s p o n d e n c e f u n c t i o n between a r c s I , and 12.
It is readily s e e n that o ( u ) is a monotonic function.
Let, as always,
2 = 'p ( t ; t o t Xoor Yo), Y = 9 (C t o , 20, Yo) (1)
b e the solution of ( D ) which p a s s e s through the point (zo, yo) for t = to. Let
'P ( t ; t o , gi (u), hi (u))= Q, (C u), 9 (ti to, gi (u),hi (u))= 'f' (1, u)- (2)
Then
@ ( t o , U)E gi (u), 'y (to, U ) hi (u). (3)
On the other hand, i n view of the above assumptions,
Q, u) = gz (0 (@I),
( x (4, 'y ( x (u). u ) = hz ( 0( u ) ) . (4)
Since 1, and I , a r e a r c s without contact, each of the determinants
r e t a i n s the s a m e sign f o r a l l u, a,(u<b. Without l o s s of generality, we

may a s s u m e that both t h e s e determinants are positive. Then o (u)is a
monotonically increasing function ( F i g u r e 123).
FIGURE 123
288
Also note that
@i ( t , u)= P (@ (t, u ) , Y ( t , u ) ) , Y; ( t , u ) = Q (av. u), Y ( t , 4). (5)
T h i s f o l l o w s f r o m ( 2 ) a n d f r o m t h e f a c t t h a t ( 1 ) is a s o l u t i o n of s y s t e m ( D ) .
L e in rn a 1 . The junctions
= cp ( t ; t o , 20, Yo), Y = 9 ( t ; t o , zo. Yo) (1)
satisfy the partial dijjerential eqtiations
2 - :To P (20. YO) a

i- Q b o . YO), = $$ P (zol YO) + S Q
(20,YO). (6)
P roof . From t h e d e f i n i t i o n of a n d y, if
E q u a t i o n s ( 6 ) follow from t h e l a s t r e l a t i o n s a n d f i
L e m m a 2. Let
where 'p = 'p ( t ; to, x0, yo), 9 =\I,( t ; to. xO, yo). Then
aJ
Proof . Let u s evaluate x.
289
Ch.XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPAKATRIX
nifferentiating the determinant,
aJ
-=
at
we obtain after simple manipulations
-=
at [Pi (9,9 +Q; ('p. 4 1 J. (12)
W e choose fixed to, x0, yo, i . e . , the l a s t relation is considered a s an ordinary

differential equation. Since
'p ( t o ; t o , r o , Yo) = Z O , Ip ( t o ; to, Zo9 Yo) = Yo,

we have
Therefore
Integration of equation ( 1 2 ) with the initial condition (13) gives ( 11).

Q. E . D.
Let A ( t , u ) denote the Jacobian
L e m m a 3.
where ~ ( tto,; g , (u),h, @))is defined by (10).

Proof. By ( 2 ) and (6),
where the values of the derivatives 2,

ax,
dp s,
ago azo avo
* are taken a t the
point ( t ; to, g, (u), hi( u ) ) . The determinant (15) is equal t o the product of the
determinants J ( t ; to, g, ( u ) , hl ( u ) ) and P(gl' ") Q(gi' '"1, which proves t h e l e m m a .
290
T h e a b o v e l e m m a s e n a b l e u s to d e r i v e a n e x p r e s s i o n for t h e d e r i v a t i v e
of t h e c o r r e s p o n d e n c e f u n c t i o n . T h i s e x p r e s s i o n w i l l p r o v e t o b e of c o n -
s i d e r a b l e i m p o r t a n c e i n w h a t f o l l o w s , a n d i v e w i l l now p r o c e e d tvith i t s
derivation.
We have a l r e a d v introduced the d e t e r m i n a n t s
As w e h a v e s e e n , t h e t w o d e t e r m i n a n t s c a n be t a k e n p o s i t i v e w i t h o u t loss
of g e n e r a l i t y . From (21, (4), a n d ( 5 ) w e h a v e
AI (3)=
A, (3) =
L e m ma 4 . The deriu,
expressible in the forin
T h e l a s t r e l a t i o n s c a n b e c o n s i d e r e d a s a l i n e a r s y s t e m i n ~ ' ( u ) ~, ' ( u ) .
I t s d e t e r m i n a n t i s e v i d e n t l y e q u a l to A z ( u ) , so t h a t w e a r e d e a l i n g w i t h a
C r a m e r s s y s t e m . Its s o l u t i o n is
T h e d e t e r m i n a n t i n t h e l a s t f r a c t i o n is A ( ( x ( u ) ,u ) . B y (14)
A (X ( 1 ~ ) .u ) = J ( X ( u ) ; to. gt (U),h~( u ) )Ai ().-
H e n c e , u s i n g (11) a n d ( I < > ) ,we o b t a i n (18). Q . E . D .

R e m a r k . The d e r i v a t i v e 0' (u)e v i d e n t l y d o e s not d e p e n d o n t h e c h o i c e
of m o t i o n ( 2 ) a l o n g t h e p a t h L p a s s i n g t h r o u g h t h e point ( g (u).h (u)) (i. e . , i t
is i n d e p e n d e n t of t h e i n i t i a l t i m e t o ) .
291
Ch.Xf. CREATION OF LlMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPARATRIX
Let u s now consider the succession function. Let the succession

function
b e defined in s o m e interval of the arc without contact 1 described by the

p a r a m e t r i c equations
z=g(u), Y=hW
(g and h are functions of c l a s s 2 ) .

Let M, be the point of I corresponding to the value uo of the p a r a m e t e r u ,
Lo the path through M o ,and
s=cP(t), Y=S(t)
the motion along t h i s path in which the point M o corresponds t o the t i m e Io

Let the succession function f (u)be defined at uo and let
According t o the definition of the succession function, t h i s m e a n s that

the path Lo c r o s s e s the a r c without contact 1 again f o r s o m e T > to at the
point Bo(Lo)and that for to < t < T i t h a s no common points with I ( F i g u r e 1 2 4 ) .
B y (ZI), M o and a r e two different points and the path Lo is therefore not
closed.
I.
I
FIGURE 124
rdinates of a,,.
(22)
292
so that
Ch.XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPARATRIX
I
-
v =f(v) = f* (0(u))=0 (f (u))= o (C).
Differentiation of ( 2 7 ) with r e s p e c t to v gives

- I
du= f* (v) =o (u)
du
- f (u)-=
du (Ti)
du
-r (u)
61
(4.
Equation ( 2 8 ) e s t a b l i s h e s a relation between the derivatives of the succes-
sion function on different a r c s without contact. T h i s expression will - be
used in what follows. Note that i f Lo is a closed path, then Go = uo, vo = vo
and by (28)
f* (vo) = f (4-
T h i s indicates that the value of the derivative of the succession function in
t h i s c a s e i s independent of the p a r t i c u l a r choice of the a r c without contact
and is a l s o evidently independent of the p a r t i c u l a r choice of the p a r a m e t e r
on the a r c without contact.
L e m m a 6. Let U = f (u)be the succession function on the arc without
contact 1 defined for all U, a,<u<b, and let fm all these values of the
parameter
1 (4< (f (4>i). (29)
Then there exists at most one closed path crossing the segment of 1 c m -
responding to the above v a b e s of U , and i f such a closed path does exist,
it will be a stable (cmrespondingly unstable) structurally stable limit
cycle.
P r o o f . Suppose that t h e r e a r e two points on the a r c I , MI (u,)and M Z(uz),
through which p a s s closed paths, where u, and uz belong to the segment [ a , b ] .
Then f (4 = U I , f (uz) = ~ 2 ,
f (UJ -f (4= ui-u2. I
F r o m Lagranges formula, f (u,)- f (uz) = f (u) (u, - u 2 ) , where is a value
of the p a r a m e t e r f r o m the i n t e r v a l ( u t , u 2 ) and hence f r o m the segment l a , b ] .
But then f (j) (u, - uz) = u, - U Z , i.e., f (i)= 1, at variance with (29). The
f i r s t proposition of the l e m m a is thus proved. The fact that a closed path
f o r whichf (u)< 1 (f( u ) > 1 ) i s a stable (unstable) s t r u c t u r a l l y stable limit
cycle was established in Chapter V (s 12.4, and also 9; 14, T h e o r e m 18).
W e w i l l give without proof two f u r t h e r l e m m a s which deal with (D) and
other close systems.
Let I , and 1, b e two a r c s without contact of s y s t e m (D) without common
points, and let a correspondence function = o (u)between the a r c s I, and l2
b e defined on t h e a r c I , for a l l values of the p a r a m e t e r u , a , 0 there exists 6 > 0 such that fm every (6)
6 -close to (D) , I , and I2 are arcs without contact and there exists a COT-
resbondence function between these arcs
-u = o- (u),
294
S 26. A IrX I LIAR Y \ I AT ERI .U
ichich is defined on the segtnent [ a . bl and is e-close to the function w (u)on

this segment.*
.I s i m i l a r p r o p o s i t i o n a p p l i e s to t h e s u c c e s s i o n f u n c t i o n . I n d e e d , l e t a
s u c c e s s i o n f u n c t i o n u = f (u) b e d e f i n e d for s y s t e m (D) o n t h e arc I for all
v a l u e s of t h e p a r a m e t e r u o n t h i s arc, a s u s b , s u c h t h a t t h e v a l u e s a = f (a)
a n d 3 = f ( b ) r e p r e s e n t i i t e r i o r p o i n t s of 1.
I e wi tn a 8 . F o r any E > 0 there exists 6 > 0 such that for every system
(616 - c l o s e t o ( D ) ,I is aflarcu'ithoutcontactandonthisarcexists thesucces-
sion firnction - c
't = f (u).
which is defined on the segment [a, bl and is e-close to the function f ( u ) on

this segment.
Lemma 7 f o l l o w s a l n i o s t d i r e c t l y from L e m m a 2, s4.1, a n d From
T h e o r e m 4, 5 1.1, i f w e r e m e m b e r t h a t t h e c o r r e s p o n d e n c e f u n c t i o n ( u ) ,
t o g e t h e r w i t h s o m e f u n c t i o n (u), s a t i s f y t h e e q u a t i o n s
w h i c h a r e a n a l o g o u s to e q u a t i o n s (4). L e m m a 8 m a y be c o n s i d e r e d a s a
p a r t i c u l a r case of Lemma 7 .
2 . S o m e p r o p e r t i e s of a s a d d l e point a n d
its s e p a r a t r i c e s
S i m p l e e q u i l i b r i u m s t a t e s c l a s s i f i e d as s a d d l e p o i n t s a r e c o n s i d e r e d i n
d e t a i l i n QT, C h a p t e r I V , 8 7 . 3 , a n d also i n C h a p t e r IV of t h e p r e s e n t book
( 5 9 ) . W e w i l l now g i v e those p r o p e r t i e s of s a d d l e p o i n t s a n d s e p a r a t r i c e s
w h i c h w i l l be n e e d e d i n what f o l l o w s .
L e t 0 (E,,, yo) b e a s a d d l e p o i n t of t h e d y n a m i c s y s t e m (D), w h i c h is a n
i n t e r i o r p o i n t of %.
L e in ni a 9 . (a) There exist E~ > 0 and a0 > 0 such that any s y s t e m (5)
S,,-close in C to ( D ) has single equilibrium state 6 in U e e ( 0 ) ,uihich is
moreover a saddle poini.
(b) F m euery E , 0 < E < E ~ ,there exists 6 . 0 < 6 == 6,, such that if (6)
is 8-close to ( D ) , it has a saddle point 0" lying in U , (0).
P r o o f . T h e v a l i d i t y of L e m m a 9 f o l l o w s d i r e c t l y from t h e d e f i n i t i o n
of a s i m p l e e q u i l i b r i u m s t a t e (see $ 7 . 3 , D e f i n i t i o n 15, a n d also 32.1,
n e f i n i t i o n 5) a n d from t h e f a c t t h a t A < O f o r a s a d d l e p o i n t .
R em ark . C o n s i d e r a dynamic s y s t e m whose right- hand s i d e s are
f u n c t i o n s of t h e p a r a m e t e r p, i . e., a s y s t e m
w h e r e (E,,,,) is i d e n t i f i e d w i t h the o r i g i n a l s y s t e m ( D ) .
T h e n t h e r e e x i s t e,, > 0 a n d a > 0 s u c h that if1 p-po I<a, (8,) h a s i n U, (0)
a s i n g l e e q u i l i b r i u m state, w h i c h i s a s a d d l e p o i n t , and its c o o r d i n a t e s
zo (p) a n d yo (p) are c o n t i n u o u s f u n c t i o n s of p; i n p a r t i c u l a r
lim zo(p) = z,,, lim yo (p) =yo.

P+BO @-&lo
W e recall that c l o s e n e s s i s to t e understood Always as closeness at l e x t to rank 1. See 53.1,

Definition 6 , and 51.1. Definition 1.
295
Ch-XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPARATRIX
The validity of the l a s t relations, i . e., the continuity of the functions

zo (p)and yo (p)a t the point po, follows directly f r o m L e m m a 9 . Continuity
f o r a l l the other values of p, close to po, follows f r o m the fact that each
of these p may be identified with po.
The next proposition is contained in the
+-=l
MO(UO) r e m a r k to L e m m a 3, 59.2, and i t is given h e r e
without proof. Let 0 be a saddle point of (D),
L+i t s a -s e p a r a t r i x , I a n arc without contact
c r o s s i n g the s e p a r a t r i x L+ a t a single point M o .
"I which does not coincide with either end point
of the a r c I , without c r o s s i n g the second
FIGURE 126
o - s e p a r a t r i x of the saddle point ( F i g u r e 1 2 6 ) .
Let, f u r t h e r m o r e , x = g(u),y = h (u) be the
p a r a m e t r i c equations of the a r c 1, the point
-
M , corresuondina t o the value u nof the u a r a m e t e r u.
-By L e m m a 9, any modified s y s t e m (%) sufficiently close to (D) h a s a
single saddle point 0" sufficiently c l o s e t o 0 in s o m e fixed neighborhood
ue, (0).
L e m m a 10. Forany E > O , there exists s > o s u c h that v ( 5 )i s 8 - C l O S e
to (D), then:
I
(a) One of the o-separatrices of the saddle point 0" (which we denote %+)
c r o s s e s the a r c 1 at a single point Mor c o r v e s p d i n g to the value Zo of the
parameter, such that R,E U. (M,,), and the second a-separatrix of the saddle
point 0" does not cross the a r c 1.
(b) If motion i s defined on the separatrices L* and E+ so that the points
M oand ~2~ correspond to the same time t = t o , then for any t > t o , the point
i@ ( t ) of the separatrix 3 cmresponding to the time t lies in an e-neighbm-
hood of the point M ( t ) of the separatrix L + c o r r e s p d i n g to the same time.
A s i m i l a r proposition is t r u e for the a - s e p a r a t r i c e s of the saddle point 0.
R e m a r k . A s in the r e m a r k to L e m m a 9, let (DP)be a s y s t e m whose
right-hand s i d e s a r e continuous functions of p and which coincides with
the original s y s t e m (D) for-p=po. Then by L e m m a 10, t h e r e e x i s t s a > O
such that i f I p - po I < a, (D,,) h a s a single saddle point 0 (p) in U , (0)one
z:
of whose o - s e p a r a t r i c e s a0
c r o s s e s the a r c I at a single point (p), and
the o t h e r s e p a r a t r i x h a s no common point with 1. Moreover, the value
Go(p) of the p a r a m e t e r u corresponding t o the point G0(p) is a continuous
function of p. A s i m i l a r proposition applies t o the case when the right-
hand s i d e s of the s y s t e m are continuous functions of s e v e r a l p a r a m e t e r s .
Still another useful proposition c a n be derived f r o m an analysis of
the behavior of paths in the neighborhood of a saddle point (QT, 3 7 . 3 ) .
T h i s proposition is formulated in the f o r m of a l e m m a , without proof.
Consider a c i r c l e C centered a t the saddle point 0; t h e r e a r e
no equilibrium s t a t e s of s y s t e m (D), except 0, either inside the a r e a
enclosed by the c i r c l e o r on the c i r c l e . Let L: and L; be o- and a-
s e p a r a t r i c e s of the saddle point 0. Suppose that each of t h e s e s e p a r a -
t r i c e s h a s points lying outside C , and let M i and M z be the l a s t common
points of t h e s e s e p a r a t r i c e s with C (so that the segments OMi and OMz
of the semipaths 15: and L; contain no points of C ; Figure 1 2 7 ) . The
s e g m e n t s O M , and O M z of the s e p a r a t r i c e s L: and L; divide t h e c i r c u l a r
a r e a enclosed by the c i r c l e C into two (curvilinear) s e c t o r s , one of
296
528. AUXILIARY LIATEWAL
which c o n t a i n s t h e o t h e r t w o s e p a r a -
t r i c e s of t h e s a d d l e point 0. L e t t h e
o t h e r s e c t o r b e d e s i g n a t e d K. L e t d
and B b e t w o points o n t h e s e g m e n t s
OM, a n d OM,of t h e s e p a r a t r i c e s L:
a n d L; w h i c h d o not c o i n c i d e w i t h .If1
and.MM., a n d L l , h2 a r c s without c o n t a c t
p a s s i n g t h r o u g h A a n d B which h a v e
no c o m m o n points.
L e n i tn a 11. There exist seg-
ments A.-ii and RBI of the arcs iaithotit
contact I., and I., which are entirely
contained (with the exception of the
FIGCRE 121 end points A and B ) in the sector K and
haue the follouiing property: every
path L of ( D ) which fm t = tI passes through a point M of the arc .-id1 other
than A , will cross f m t 2 ) ti the arc X 2 at some point .V other than R;
moreover,
(a) all the points of'the path comespanding to t , ti< t ( t 2 , lie iaside
the sector K ;
(b) a path passing through the point A i of the arc .-idlcrosses the arc
BB, at the point B,:
(c) the point -V goes to B when -If goes to .i;
(6, for any T > 0, there exists a point -If*of the arc Aril such that f o r any
path which f m t = ti crosses the segment -\f*.-i of the arc . l A I we have the
inequality t2 - tl > T.
R e t a i n i n g t h e n o t a t i ? n of t h e p r e v i o u s l e m m a , let u s c o n s i d e r a m o d i f i e d
s y s t e m (6).
F r o m t h e t h e o r e m s of t h e c o n t i n u o u s d e p e n d e n c e of t h e s o l u t i o n o n t h e
i n i t i a l c o n d i t i o n s a n d o n t h e r i g h t - h a n d s i d e and f r o m the p r e v i o u s l e m m a s ,
i t f o l l o w s t h a t t h e r e e x i s t e , > 0 a n d 8,>0 w i t h t h e f o l l o w i n g p r o p e r t i e s : i f
( A ) i s &,-close t o (D), :hen
1 ) U,,(O) c o n t a i n s o n e a n d o n l y o n e equili_brium s t a t e , t h e s a d d l e point 6:
2 ) hi a n d & are arcs w i t h o u t c o n t a c t of ( D ) ;
3) t h e r e e x i s t s e p a r a t r i c e s x:and z; of t h e s a d d l e point 6 of (6)w h i c h
cross t h e a r c s Ai a n d L! a t t h e p o i n t s a n d E , r e s p e c t i v e l y , t h e point
l y i n g o n hl i n t h e same d i r e c t i o n f r o m Al as t h e p o i n t A , and t h e point
lying on in t h e same d i r e c t i o n f r o m B1a s t h e point B ;
4 ) t h e p a t h of ( b ) p a s s i n g t h r o u g h A, c r o s s e s t h e a r c hz a t s o m e point E,
w h i c h lies on 4 i n t h e s a m e d i r e c t i o n from B a s t h e point Bl.
T h e f o l l o w i n g t w o l e m m a s a r e now s e l f - e v i d e n t .
L e m m a 12. Far any e > 0 ( E < E,) there exists 6 0 (6 <ao) such that =.
if ( E ) is 6-close to ( D ) , then
(a) ZEL'~(A), B E U ~ ( B ) BIG
, L', ( B J ;
(b) the path of ( 6 )which fm t = ti passes through some point .If of the
arc XA, icill cross the arc 12, at the point f o r t = & > t , ; -v
(c) the point R goes to E when ,M goes to x.
L e m m a 13. F o r ( I fixed T>O, let M* be a point of the arc 1, satisfying
condition (d) of L e m m a 11. Then there exists a>o such that if ( 6 )is
6 -close to ( D ), the point lies on the arc XI in the same direction f r o m
I ch.XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPARATRIX I
M b a s the point A , and fm every path of (fi) crossing
the a r c h, we have the inequality z 2 - t , > T .
I
In what follows, we w i l l often r e q u i r e the p a r a m e t e r
(J ( 2 0 9 YO) = P ; (zo? YO) +QI (503 YO), (30)

where xo, yo a r e the coordinates of the saddle point 0 of ( D ) . We w i l l show
that it is invariant under a transformation of coordinates.
We p e r f o r m in G a transformation of coordinates defined by the equalities
o r the equivalent equalities

E = f ( x , Y). q = g ( z , Y) (31)
I
x=(P(E, l)?Y = 9 ( E v l),
where f, g , (P, and 9 a r e functions of c l a s s 2 .

In the new coordinates, (D) takes the f o r m I
L e m m a 14. u (4,yo) = P; (x0,yo) +Q; (x0, yo) is invariant tinder a transfm-

mation of coordinates, i . e.,
u* ( E O ? lo) =Pi (Eo, rlo) + Q: (Eo, Yo) = @o, Yo). (34)
P ro of .
Differentiating the f i r s t equality in ( 3 3 ) with r e s p e c t t o E and
the second with r e s p e c t t o tl, adding them and i n s e r t i n g f o r 5 ar.d q t h e i r values
Eo and lo,respectively, we obtain, using the relation P ( x o , yo) = Q ( x o , YO) = 0 ,
u* (Eo. tlo) = {G(9.9)[ P i ((P? 9)(Pi +PI ((P7 N $1 +
+fI (CP,9)IQ; ((P, 9)( ~+iQI
((P, 9)%I +g; (CP,9)[ P ; ((P, 9)( ~ +6 P I (CP,
9)%I +
+g& ((P9 9)[Q; (cpl 9)(P6
Equality ( 3 4 ) follows d i r e c t l y f r o m the l a s t relation in virtue of the identities
2 = p (f (2,Y h g (5, Y = 9 (f (2, Y). g (2, Y))

which aive. when differentiated with resDect t o x and u.
298
S2a.CREATION O F LIMIT CYCLES F R O M THE
SEPARATRIS LOOP OF A SIhIPLE S-ADDLE P O I N T
1. Some p r o p e r t i e s of the s e p a r a t r i x loop
be a dynamic s y s t e m , 0 (x,, yo) a s i m p l e equilibrium s t a t e of t h i s s y s t e m ,

ivhich is a s a d d l e point. Suppose that one of the a - s e p a r a t r i c e s Loof t h e
saddle point 0 i s also a n w s e p a r a t r i s , i . e . , it f o r m s a whole path which
g o e s to 0 both f o r t - t -- 03 and for t + + 03. In t h i s c a s e we s a y that t h e
s e p a r a t r i s L o f o r m s a 1o o p Let C , denote
the s i m p l e closed c u r v e c o m p r i s i n g t h e path Lo
/ $ 7
"3 Lr
and t h e point 0. T h e c u r v e Cowill b e called a
l o o p o f t h e s e p a r a t r i s Lo o r s i m p l y a
loop.
E e s i d e s the path L o , the saddle point 0 also
h a s two o t h e r s e p a r a t r i c e s L ; and L; ( e i t h e r dif-
ferent o r coincident) which both lie e i t h e r inside
t h e c u r v e C o o r outside t h i s c u r v e . \Ve will
henceforth a s s u m e , without loss of generality,
FIGC'RE 128 that t h e s e p a r a t r i c e s L; and L; lies outside t h e
c u r v e C o ( F i g u r e 1 2 8 ) . T h e c a s e when they
l i e inside the c u r v e C o is e n t i r e l y analogous.
L e ni ni a 1 . A septzvatrix forming a loop may have ot most one cowiition
point with an arc without contact.
P r o o f , Suppose a s e p a r a t r i x Lo forming a loop h a s two common
points A and B with s o m e a r c without contact 1 ( F i g u r e 1 2 9 ) . Then the
saddle point 0 should evidently lie both inside the c l o s e d c u r v e formed
b y the s e g m e n t s AB of :he path L o and t h e a r c without contact 1 and outside
t h i s c u r v e , which i s cll2arly i m p o s s i b l e . T h i s p r o v e s t h e l e m m a .
L e t z = 'pa ( t ) , y = ( t ) b e a solution c o r r e s p o n d i n g t o t h e s e p a r a t r i x
I.,,. Moancl .Vi the points of the s e p a r a t r i x c o r r e s p o n d i n g t o the times 1,
and t , . We m a y t a k e to < ti. A r c s without contact 1, and I , , without
common points, a r e drawn through the points N o and MI ( F i g u r e 130).
FICL'RE 1 2 3 FIGLRE 130
* 6y Theorem 23 ($18.2). 3 separ,iaix forming d loop may exist only in ri ,tructurally unstable system.
299
L e t M O B obe a segment of the a r c 1, which is entirely enclosed by the curve

C,,, with the exception of i t s end point M , ; l e t M O A o be a segment of the arc
lowhich lies entirely outside the curve CO, with the exception of the point M , .
F u r t h e r m o r e , l e t M I B , and M I A Ibe the analogous segments of the a r c 1,
( F i g u r e 130).
L e m m a 2. F o r any e > 0, there exists 6 > 0 with the f o l l m i n g pro-
perty: every path L which f o r t = topasses through a point M , of the arc
M O B , which lies in Ub ( M , ) and does not coincide with M will again cross
the segment M O B oof the arc 1, at some point s f o r T > to, without leaving
the e-neighborhood of the loop C, during the time t,<t< T (h7may dzjcfer
f r o m M , as in Figure 130, or coincide with M ) .
The proof of L e m m a 2 follows directly f r o m L e m m a 11 of the previous
section, i f w e take into consideration the properties of paths c r o s s i n g two
arcs without contact (QT, 83.4, L e m m a 5).
R e mark . Let A > 0 be s o m e number. The number 6 introduced in
L e m m a 2 c a n b e made sufficiently small, s o that f o r s o m e ti, to < ti < T,
I ti - t, I < A, the path L will c r o s s the segment M I B i of the a r c Il.
In the next l e m m a , we consider paths which c r o s s the segment MoA,of
the a r c lo lying outside the loop C,.
L e m m a 3. There exist e, > 0 and 6 , > 0 with the following property:
i f the path L passes through the point M of the arc M O A , which lies in
ubo (Mo)anddoes not coincide with M , , the path L will leave the &,-neighbor-
hood of the path Co both with increasing and with decreasing t.
P r o o f . Consider the w- and a - s e p a r a t r i c e s L: and L; of the saddle
point 0 which lie outside the 'curve C,. Let N iand N zbe two points on t h e s e
s e p a r a t r i c e s , and IC, and IC2 a r c s without contact passing through t h e s e
respective points ( F i g u r e 130). By L e m m a 11 of the previous section,
t h e r e e x i s t s 6, > 0 with the following property: i f the path L p a s s e s through
the point M of the a r c M O A owhich lies i n u b o ( M 0 )and does not coincide with
M o , t h i s path will c r o s s the a r c A2 with i n c r e a s i n g t and the a r c ?.with
,
decreasing t . As e,> 0 we choose a number such that the neighborhood
U , (C,) does not i n t e r s e c t with the a r c s without contact IC, and A*. The
numbers, 6, and e, evidently satisfy the proposition of the l e m m a .
Let u s now r e t u r n t o t h e case of the arc M O B olying i n s i d e the c u r v e
C,. By L e m m a 2, e v e r y path which f o r t = to c r o s s e s t h i s arc at a point M
sufficiently c l o s e t o M , , will again c r o s s this a r c a t a point % ( t r s u c c e s s o r "
of M ) f o r T > to; when M g o e s to M , , ii? a l s o g o e s t o M , . Since the a r c s
M O B oand M I B , may be identified with the arcs B B , and A A I , respectively,
of L e m m a 11 of the previous section, t h i s l e m m a together with L e m m a 2
d i r e c t l y lead t o the following useful proposition.
L e m m a 4. When the point M an the arc M O B o(Figure 130) goes to the
point M , on a separatrix, the time T corresponding to the point I@ goes
to+ 53.
We will s a y that the path L goes t o the loop C , f o r t -e +

00 ( t -t - m) i f
i t s w - l i m i t (correspondingly, a - l i m i t ) s e t coincides with the loop C,.
T h e proof of the following l e m m a , which u s e s L e m m a 2, is self-
evident and is thus omitted.
L e m m a 5. If among the paths crossing the arc M O B oat points suf-
ficiently close to M , there are no closed paths, then either all these paths
go to a loop f o r t+ i- m o r they all go to a loop f o r t + - 5 3 .
300
$23. CRE.4l'lOS OF L I h ! I r CYCLES FROh! A SEP.4RATRIX LOOP
D ef i n i t i o n 29. A loop C , is said to be stable (unstable) if all the

paths crossing the arc ,MOB, at points sufficiently close to M o (and yet
different from -$Io) go to the loop c0for t + + co (correspondingly, f o r
t - + - m).
Pv L e m m a 2 , a s u c c e s s i o n f u n c t i o n i s d e f i n e d o n s o m e s e g m e n t M O B
of t h e a r c .MOBo ( t h e p o i i t M 0 e x c e p t e d ) .
Let
z = go (4, Y = ho ( u ) (1)
h e t h e p a r a m e t r i c e q u a - i o n s of t h e arc I , . A s in t h e p r e v i o u s s e c t i o n , w e
a s s u m e t h a t g, a n d hO a r e f u n c t i o n s of class 2 . :
S u p p o s e t h a t t h e p o i n t s .!Io, A,, Bo, B c o r r e s p o n d t o t h e v a l u e s u o , a,, bo, b
of t h e p a r a m e t e r , w h e r e a, < uo < h < bo.
Let
-
zf = f ( u ) (2)
b e t h e s u c c e s s i o n f u n c t i o n o n t h e a r c M O B . In v i r t u e of t h e a b o v e a s s u m p -
t i o n , f (u)is d e f i n e d f o r all u , uo < u < b ( w e s h o u l d s t r e s s t h a t w e are d e a l i n g
w i t h a s u c c e s s i o n f u n c t i o n i n t h e d i r e c t i o n of i n c r e a s i n g t , i . e . ,
e a c h " s u c c e e d i n g " p o i n t c o r r e s p o n d s t o a l a t e r t i m e t h a n t h e t i m e of t h e
"preceding" point).
In \ + h a t f o l l o w s , w e w i l l also c o n s i d e r t h e f u n c t i o n
d (u) = f (u) - u . (3)

By L e m m a 2,
lim d ( u ) = 0.
u-uo
If a c l o s e d p a t h L* p a s s e s t h r o u g h t h e point hf* ( u * ) of t h e a r c M O B , w e
h a v e d (u*) = 0 .
If t h e r e e x i s t s u t , u, ( u , , < b , s u c h t h a t for all u , u , ( u < u l , d ( u ) #O, i.e.,
if t h e r e a r e n o c l o s e d p a t h s c r o s s i n g the a r c M O B , a t t h e p o i n t s u , u , < Z G U , ,
then
t h e l o o p i s s t a b l e ifd(u)<O,
t h e l o o p i s u n s t a b l e ifd(u)>O (5)
(for uo<ugul).
In w h a t f o l l o w s w e w i l l c o n s i d e r t h e l i m i t of t h e d e r i v a t i v e of t h e
s u c c e s s i o n f u n c t i o n f' ( ( 1 ) for u -+ u,.
Let
2 = (P.H (% Y = *.\a ( t ) (6)
b e t h e s o l u t i o n c o r r e s p o n d i n g t o t h e p a t h L p a s s i n g t h r o u g h t h e point M (u)
of t h e a r c M O B . A s b e f ' 3 r e , w e a s s u m e t h a t t h e s o l u t i o n ( 6 ) p a s s e s t h r o u g h
t h e point X for t = to a n d t h r o u g h t h e s u c c e s s o r IG (i)of M for t = T > t o .
T h e c o o r d i n a t e s of t i e p o i n t s M 0 ,M , # a n d t h e s a d d l e point 0 are
( E o . qo), ( E , tl), (E, 6), a n d (z0,y o ) , r e s p e c t i v e l y . Let, a s b e f o r e ,
00= CJ (zov YO) K YO)
(20, +QC (207 YO). (7)
[I, in particular. I o is a segment of the normdl to the pdth L,at the point M a ,rhe functions go a n d h a c a n
be t&cn in the form
go (u:.= To ( t e ) - % ( t o ) U? ho (u) = $0 ( t o ) 4-Ti 0 0 ) u.
30 I
Ch.XI. CREATION OF LIMIT CYCLES FROM T H E LOOP OF A SADDLE-POINT SEPARATRlX
L e m m a 6 . Let the point M on the arc M O B go to the point M , , i . e . ,

U+ u 0 . Then
f (u) + + 03 for (J (zoo1 Yo) = p; (x07 Yo) +QI

b o 7 Yo) > 0, (8)
f(U)-+O f C W (J(zo,Y o ) < o . (9)
Proof. F i r s t note that l i m f ( u ) i s independent of the p a r t i c u l a r choice

u-UO
of the a r c without contact I o . This follows directly f r o m equation (28) of
the previous section. In virtue of t h i s equation, i f I,* is an a r c without con-
tact analogous to 1 , u and f* (0)are the p a r a m e t e r and the succession function
on t h i s a r c , s o that the s e p a r a t r i x Lo c r o s s e s I* a t the point vo, and i f w (u)
is the correspondence function between a r c s without contact 1 and I*, we have
I f u - + u o , t h e n v - + u o , ~ - u o , and since w ( u o ) # O i n v i r t u e ofS28.1, (18),

we have
lim f* (v) = lim f ( u ) .
u-co u-uo
F i r s t consider the c a s e when

I
(Jo = (J (xo,Yo) >0.
Let C be a c i r c l e centered at 0 with a sufficiently s m a l l radius s o that
I
a t any point ( x ,y) inside C we have
The points M o and Mi and the a r c s I , and 1% are chosen so that t h e s e a r c s

a s w e l l as the segments of the s e p a r a t r i x Lo corresponding to t,( to and
t > t i lie inside the circle C ( F i g u r e 131).
I
FIGURE 131
By L e m m a 5 of the previous section, we have

T
302
929. CRMTION OF L I h f f T CYCLES FROhl A SEPARATRIX LOOP
where
and
- _ _
When U - + U O , both d e t e r m i n a n t s A ( & q, u ) and A(E, q, u ) g o t o t h e s a m e non-
qo,u o ) , so that
z e r o l i m i t 4(E0,
lim -= 1.
u-uo A(:. q. u)
We should t h e r e f o r e find only t h e l i m i t
Clearly,
T :; T
5
10
1
1.
1
( J ( c p ~ ( t1l ,S ~ ( ( t ) ) d t = (J(T.M, $ . \ ~ ) d f f (J(T.M~$.w)dt,
1;
(13)
w h e r e t; is the value of t c o r r e s p o n d i n g to the i n t e r s e c t i o n point 32; of t h e

path L with the a:c without contact 2, ( F i g u r e 131 : we r e c a l l that t h e path
L p a s s e s through the point ,If f o r t = t o ) .
ti
When M + . t i o , we h a v e \ o(cp, \
t@'2.1)dt-+ a ( q o ( t ) ,q 0 ( t ) ) d t , w h e r e r = q o ( t ) ,
i. t,,
Lowhich p a s s e s
y = q 0 ( t ) is the solution c o r r e s p o n d i n g t o the s e p a r a t r i x
through t h e point Mo f o r t = t o . T h e f i r s t i n t e g r a l o n t h e right in (13) t h u s
g o e s t o a finite limit.
Let u s c o n s i d e r t h e second integral, i . e . ,
5
1;
(J (TM (0, $M ( t ) )dt.
If t h e point -kf is sufficiently c l o s e t o Mo,then M; is a r b i t r a r i l y c l o s e

t o Mi and t h e s e g m e n t .l.f;.U of t h e path L is e n t i r e l y enclosed inside the
c i r c l e C (see S28.2, L e m m a 11). But then, in virtue of inequality ( I O ) ,
t h e integrand in t h e l a s t i n t e g r a l is g r e a t e r than 9 for any t E [ t ; , TI, i . e.,
T
\ (J (TJa ( t ) ,$24 ( t ) )d t > 9 (T- u.
;;
For .ti-+ M o , t; -+ ti and T d + w (by L e m m a 4). T h e last i n t e g r a l ,
and hence ru(cpM,q X ) d t , go to + a f o r M--+M,,. T h i s r e s u l t and equalities

to
( 1 1) and (12) show that lim f ' ( u ) =

U-UO
+w . We h a v e t h u s proved (8). It c a n b e
303
Ch. XI. CREATION OF LIMIT CYCLES FROkl THE LOOP OF A SADDLE-POINT SEPARATRIX
s i m i l a r l y shown that i f uo< 0, then
lim
M+Mo
i u (9.M( t ) , +Ar ( t ) )d t = - oo, T. e. f (u)40.

R e mark . W e have assumed s o f a r that the s e p a r a t r i c e s L; and L; of
the saddle point 0 which do not coincide with Lo lie outside the loop C, and
that the segment M O Bof the a r c without contact 1, on which the succession
function f (u)is defined corresponds t o u > u o . If the s e p a r a t r i c e s L: and
L; lie i n s i d e the loop Co, the loop may only be an a- o r w-limit continuum
for the paths of the s y s t e m f r o m o u t s i d e , whereas in a l l other r e s p e c t s
the situation does not change. In particular, all the previous l e m m a s of
t h i s section r e m a i n valid. The assumption that the s e p a r a t r i c e s LT and L;
l i e outside the loop is thus of no consequence and h a s been introduced f o r
convenience only.
Let u s now see what happens i f the direction along the a r c 1, is r e v e r s e d
( e . g., by defining a new p a r a m e t e r u* = - u ) . First note that the succession
function f (u)on the a r c lo (or, m o r e precisely, on the segment of the a r c
adjoining the point M o , where i t is defined) is a monotonically i n c r e a s i n g
function and i t s derivative f (u)is positive r e g a r d l e s s of the particular
direction on the a r c l o which is chosen a s the positive direction. T h i s
follows immediately f r o m geometrical considerations (or, alternatively,
f r o m L e m m a 5 of the previous section and the r e m a r k t o this l e m m a ) .
Our L e m m a 6 thus r e m a i n s i n force f o r any choice of the p a r a m e t e r on
the a r c without contact l o .
Conversely, conditions ( 5 ) derived in the preceding a r e t r u e only when
the succession function f ( u ) is defined for u>u,. If, on the other hand,
it is defined f o r u s m a l l e r than u,(e.g., for uo > u > u , ) and t h e r e a r e no
closed paths c r o s s i n g the a r c I , at points u , uo < u,<u,, we c l e a r l y s e e that
t h e l o o p i s s t a b l e ford(u)>O,
t h e l o o p i s u n s t a b I e ford(u)<O (14)
=-
( f o r uo u > u i ) .
W e a r e now in a position to d e r i v e a sufficient condition of stability
(instability) of a s e p a r a t r i x loop.
T h e o r e m 4 4 . Let 0 ( x 0 , yo) be a saddle point of the dynumic system
xd r= P ( z , v), $-=QQz,Y),
(D)
and L, its separatrix which together with the saddle point 0 f o m s the
loop C , . Then, if u, = P; ( x o , yo) + Q; (x,, yo) > 0, the loop is unstable, and if
u0 < 0, the loop is stable.
P r o o f . Let u s f i r s t consider the c a s e when the succession function is
defined for u c l o s e to u,, but g r e a t e r than u,. Let u o > 0 . In virtue of (4),
lim d (u) = 0.
UUO
On the other hand, i n virtue of L e m m a 6, i f a,> 0, we have
lim f ( u ) =
U-tUO
+ 03.
304
T h e r e f o r e , t h e r e e x i s t s u t s u c h t h a t for all u , u a 1 a n d d' ( u ) = 1' ( u ) - 1 > 0. (15)
It f o l l q w s from ( 4 ) a n d ( 1 5 )t h a t for all u , u c 0, i . e., t h e l o o p

i s u n s t a b l e i n v i r t u e of ( 5 ) .
L e t now u o < 0 . T h e n , b y L e m m a 6
lim j ' ( u ) = 0 ,
u-u*
i . e . , t h e r e e x i s t s ul s u c h t h a t for all u , u , , < u < u I ,
f ' ( u ) < I and d ' ( u ) C 0 . (16)
E y ( 4 ) a n d ( 1 6 ) bye s e e t h a t for all u . u,, < u < u l . d ( u ) < 0, i . e., t h e l o o p

is s t a b l e i n v i r t u e of ( 5 ) .
If t h e s u c c e s s i o n f u n z t i o n is d e f i n e d or u u > u I ,
d ( u ) 0, ~ a n d t h e n b y ( 1 4 ) t h e 1OGp is u n s t a b l e . B y ( 4 ) a n d ( 1 6 ) w e see t h a t
d ( u ) i 0, i . e . , t h e l o o p i s s t a b l e b y (14). T h i s c o m p l e t e s t h e p r o o f of t h e
theorem.
R e m a r k 1. It f o l l o w s from T h e o r e m 44 t h a t i f t h e f o u r s e p a r a t r i c e s
o f t h e s a d d l e p o i n t 0 f o r m t w o loops ( w h i c h l i e o n e o u t s i d e t h e o t h e r o r
a
o n e i n s i d e t h e o t h e r , F i g u r e s 132 a n d 1331, t h e n for n,:,O b o t h l o o p s a r e
u n s t a b l e , a n d for oo < 0 b o t h l o o p s a r e s t a b l e .
/
-
F G C R E 13" FIGURE 133
R e m a r k 2 . T h e o r e m 44 is c o n c e r n e d w i t h t h e case uo = Pi (zo.go) +
A Q&(ro.yo)# 0 . We w i l l now s h o w t h a t if oo = 0, t h e r e m a y b e cases w h e n
a n a r b i t r a r i l y s m a l l n e i g h b o r h o o d of t h e l o o p c o n t a i n s c l o s e d p a t h s , as
w e l l as cases w h e n t h e l o o p is s t a b l e o r u n s t a b l e .
E x a m p 1e 11. C o n s i d e r t h e s y s t e m
-d -i 2 g = P , ( z , g), $=i123-39=Q1(*, Y),
dt (D1)
w h i c h is i n v e s t i g a t e d i n QT, 5 1.14, E x a m p l e L'III. It can b e c h e c k e d t h a t

this system has a general integral
9-6.9 + I J ~= C . (17)
305
T h e shape of the c u r v e s ( 1 7 ) is readily established f r o m t h e i r explicit

equation
considered together with the auxiliary c u r v e s
2 = 6sz-2 + C,
which can b e constructed without difficulty. C u r v e s ( 1 7 ) a r e shown i n
Figure 134. C = 0 corresponds t o the curve with a loop. For C > 0,
curve ( 1 7 ) c o m p r i s e s a single branch
located outside the loop. For O>C>-32,
the curve c o n s i s t s of two branches.
One of these b r a n c h e s lies to the left
of the y axis, and the other branch is
an oval enclosed inside the loop. For
C = - 3 2 , the c u r v e c o m p r i s e s a branch
I lying in the left half-plane and the
point O1 ( 4 , 0). Finally f o r C < - 3 2 ,
the c u r v e h a s a single branch in the
left half-plane. Each of the c u r v e s
-
( 1 7 ) is either a path of s y s t e m (D1)
(for C; 0 and C < -32). o r c o n s i s t s
I (I! I
of two (for O > C>-32) o r f o u r ( f o r
C = 0) paths. System (D1) h a s two
equilibrium states: 0 ( 0 , O ) and
Oi(4,O). The f i r s t of these equi-
FIGURE 134 librium s t a t e s is a saddle point, with
two of i t s s e p a r a t r i c e s forming the
loop L , , and the second is a c e n t e r .
H e r e uo (0, 0) = Pi= (0, 0)+Qiu (0, 0) = 0, i . e., we a r e dealing with the
c a s e u,, = 0, and any neighborhood of the loop contains closed paths.
Together with s y s t e m (Dl), let u s now consider the s y s t e m
dt =2 ~ - p( x 3 - 6.79 + y2) (i2.z- 328) = P , (3,y).

!:
-= 12r-3s2-t~(2?-6s*+y2)2y=Qz(s, y),
(D2)
where p is a s m a l l (in absolute magnitude) number.

Clearly,
It is directly verified that the four paths of the s y s t e m (D1) making the
curve
2?-6s* + ya= 0 (20)
a r e a l s o paths of s y s t e m (Dz). Moreover, (Dz) h a s the s a m e equilibrium

s t a t e s as (Dl), i. e., the points 0 (0, 0 ) and 01(4, 0). and the point 0 is a
saddle point of (D2), w h e r e a s 0,f o r s m a l l p # O is a s t r u c t u r a l l y stable I
5438 306
>- .
rn.j
CREATION OF LI!.IIT CYCLES FRO!.! -4 SEPAR.ITRLX L O O P
focus, which i s s t a b l e for p ? O and unstable f o r p<O. T h u s Lo i s a s e p a r a t r i s

o f (D2) forming a loop. T h e v e c t o r field of (D2) is obtained b y rotating t h e
v e c t o r field of (D1) through a n angle tan-' p j (3.y ) ( t h i s is found by d i r e c t
computation; see a l s o QT, 1.14, r e m a r k preceding Example V I I ) . But
than all closed paths of ( D l ) a r e c y c l e s without contact f o r (l32). Hence i t
c l e a r l y follows that inside the loop Lo (D2) h a s no closed paths, i . e . , all
the paths of (D2) lying inside t h e s e p a r a t r i x loop e i t h e r wind onto t h e loop
t h e focus 0%for k-+ - 00 and onto t h e s e p a r a t r i x loop f o r t -

f o r t -+ - co and onto the focus 0, for t + + -, o r , c o n v e r s e l y , wind onto
-,- 00. Since
for p > 0 (p < O), the focus 0,is s t a b l e (unstable), t h e s e p a r a t r i x loop L ,
is
s t a b l e for p C 0,
unstable for p > 0 .
Now a. = PbX (0, 0 ) C C)iy (0, 0) vanishes for any p. O u r e x a m p l e t h u s p r o v e s

that for o,, = 0 , any of the t h r e e a l t e r n a t i v e s mentioned i n Example 2 m a y
b e observed.
2. T h e o r e m s of the creation of a c l o s e d path f r o m

a s e p a r a t r i x loop
Assuming, as b e f o r e , that s y s t e m ( D ) h a s a s a d d l e point 0 whose

s e p a r a t r i x L o f o r m s a loop, w e will c o n s i d e r , alongside with ( D ) , a modified
svstem
dt -
x = P ( z , Y), %=ax,
- Y). (6)
Retaining the same notation a s i n t h e previous subsection, we w r i t e +IIo
and X ifor points of the s e p a r a t r i x L o c o r r e s p o n d i n g to t = to and t = t , , to < t , ,
Inand 1, for t h e a r c s without contact passing through t h e s e points, and u f o r
t h e p a r a m e t e r on the arc: I , ( F i g u r e 130). L e t L; h e t h e positive half of the
path Lo containing t h e pomt .%Io(and hence also t h e point MI; L; is a n a -
s e p a r a t r i s of 0, and its points c o r r e s p o n d to t>tz, w h e r e t: < t,), and L;
t h e negative half of t h e path Lo containing M oand Mi ( t h e points of L; cor-
respond to t < t;, w h e r e t; > t , ) .
E?y L e m m a s 9 and 10, % 2 8 , t h e r e e s i s t e, > O and S o > O s u c h that if t h e
modified s y s t e m (6)is S,,-close t o (D), then
( a ) U,, (0)contains a single equilibrium s t a t e of ( b ) , t h e s a d d l e point 6 ;
( b ) t h e r e e x i s t a n io-separatrix 2;; and a n a - s e p a r a t r i x EA- of t h e s a d d l e
point d c r o s s i n g t h e arc Io at t h e points -9,and .G;, respectively, which l i e
inside t h e a r c I , ;
4
( c , t h e o - and t h e a - s e p a r a t r i c e s c-
and c r o s s t h e a r c I , a t points
and .)Ti, respectively, which lie i n s i d e t h e a r c I , ( F i g u r e 135).
M o r e o v e r , b y L e m m a s 9 and 10, P 28, for any positive e < e,, t h e r e e x i s t s
6 < So s u c h that for e v e r y s y s t e m (8) which is S-close to ( D ) , t h e saddle
point 6 i s contained i n U,. (O),and e a c h point of t h e s e m i p a t h (I;-)lies i n
t h e -neighborhood of t h e point of t h e s e m i p a t h L; (L;) c o r r e s p o n d i n g to the
s a m e t i m e . T h e points 5 j 0 and -@; lie i n L', (M,), and t h e points .c, and .I?: in
U , w,).
307
FIGURE 135
Let Go and 2; be the values of the p a r a m e t e r u corresponding to the points

fi, and I& (a <Go < b , a < 2 < b ) . The following c a s e s a r e possible for
different modified s y s t e m s :
and M
- -
A a r e two different points, i . e . , u; + uo.
1) -
2 ) M a and 2;a r e the s a m e point, i . e., u; = u,.
In case 1, L e m m a 1 shows that (6)h a s no s e p a r a t r i c e s which form a
loop whose semipaths a r e E; and E;-. We will s a y in t h i s c a s e that on
moving f r o m (D) t o (E), the 1o o p i s b r o k e n . T h e r e evidently always
exist modified s y s t e m s as close a s d e s i r e d t o (D) and such that the loop
is broken when moving t o t h e s e s y s t e m s . An appropriate example is the
system
dz
x=P-pQ, $=Q+pP, (D9
where p # O is a sufficiently s m a l l (in absolute magnitude) number. By

the l e m m a of 11.1, i f p>O, we have u;
> u o , Lo< uo for s y s t e m (D*), con-
sidered a s (D), i.e.,
L
O <E;,
and if p < 0, then ;;<u,, uo>u0, i.e.,
In c a s e 2, when the points i@, and &: coincide, (6) h a s a s e p a r a t r i x Eo

which f o r m s a loop, z;
and E;- are i t s semipaths, and in t h i s c a s e the s e m i -
path E;- m a y be simply designated xi.
Let, a s before, the succession function be defined for (D) everywhere
along the arc M O B , with the exception of the point M o .
L e m m a 7 . There exists 6o ==. 0 such that i;f (a)
i s &,-close to (D),
2,i s an a r c without c a t a c t fm(6), and every path of (b) which for t = to
crosses the segment G O Bof the arc I , will cross fm some T > to the
segment KBoof this arc.
T h e proof of t h i s l e m m a follows almost immediately f r o m L e m m a s 7 and
-
9- 12, l28, and i t is therefore omitted. Note, however, that the cases
- c
u ; > u o , u; = uo, and E; <Lo should be treated separately ( F i g u r e s 135, 136,
and 137, respectively). T h e proof proceeds along the s a m e lines in a l l
the t h r e e c a s e s .
308
LIhlIT CYCLES FRO:,I A SEPMRATFUX LOOP
FIGLRE 136 FIGURE 137
L ern m a 8. There exist eo > (3 and a,, > 0 which satisfy the following
conditions: i f ( 6 )is&-close to(D),and;; < 5,(L; >E,), all the paths of (8)
crossing the segment .+i& of the arc I , w i l l leave the e,-neighbmhood of
the loop L o as t increases (decreases).
P r oof .T h e validitv of L e m m a 8 follows d i r e c t l y f r o m the p r o p e r t i e s
of the paths of s y s t e m (fi) which c r o s s the segment -@,.qiof the a r c without
contact & ( s e e F i g u r e s 135 and 137), if w e u s e L e m m a 3 of t h i s section and
a l s o L e m m a s 9 and 10 of 5 2 8 . 2 .
We c a n now prove one of the fundamental t h e o r e m s , t h e so-called
t h e o r e m of c r e a t i o n o f a c l o s e d p a t h f r o m a s e p a r a t r i x
Zoo p . It e s t a b l i s h e s sufficient conditions f o r the appearance of a closed
path when a s e p a r a t r i x loop d i s a p p e a r s ( b r e a k s ) .
Let 0 be a saddle point of a dynamic s y s t e m (D), L o i t s s e p a r a t r i x which
f o r m s a loop. H e r e , 1,. 2;. L;-,u,,&, A , . B o . a,, bo, b , etc., have the s a m e
meaning a s before; m o r e o v e r , a , -== bo and the point Bo lies i n s i d e the
loop ( F i g u r e 135 o r 1 3 7 ) .
T h e o r e m 45. Let the loop fmtned by the separatrix L o of the saddle
point 0 be stable (unstable). Then, f o r any F > 0 , there exists 8 > 0 such
that i f ( 8 )is 8-close to ( D ) and if 5,< Z ; (Z0 > L;), then in the e-neighbm-
hood of the loop there exists at least m e closed path E* of (8)
uhich crosses
the arc without contact L,, at the point .\f* (u*), where
- <u'<
ug (ii<u* < b).
b
A similar proposition applies if the separatrices of the saddle point o

other than Lo lie inside the loop.
.
P r o o f Consider the c a s e of a stable s e p a r a t r i x loop. Let e > 0 b e
given. On rhe segment .lloB of the a r c 1, we s e l e c t a point .Vi (n,)sufficiently
c l o s e t o .%fa, s o that the following conditions are satisfied:
( a ) the path LN of ( D ) passing through Nj g o e s to the loop f o r t -+ cm; +
( b ) the segment of the path LN between the point Ni and the "succeeding"
intersection point N2 (uz)of the path Ls with the a r c Lo, together with the
309
CheXI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPAKATIUX
segment NiN,of the a r c l o , f o r m a

simple closed c u r v e C N which i s
entirely contained in Ue,3 (Lo).T h e
region limited by the loop of the
s e p a r a t r i x Lo and the c u r v e C N
enclosed inside t h i s loop is a l s o
contained in ( F i g u r e 138).
Since the s e p a r a t r i x loop is
stable, i t is c l e a r that i f point N 1
is sufficiently close t o Mo,both
conditions are satisfied. It is
a l s o c l e a r that nz < n,.
Let u s now take 0 so small
that the following conditions a r e
satisfied :
( a ) 8i < a o , where is the
FIGURE 138 number defined by L e m m a 7;
( b ) i f (6)is 6,-close to (D),
the path ZN of ( 6 )which p a s s e s through the point Nl c r o s s e s again the
a r c l o at the point fi2 (&), s o that n, < ni, and the simple closed c u r v e
EN, analogous t o the c u r v e CN, is entirely enclosed inside the loop of
the s e p a r a t r i x L o , delimiting together with the loop a region which is
entirely contained in U./l ( L ) .
It is c l e a r that i f 8, is sufficiently s m a l l , both conditions a r e s a t i s -
fied. In particular, condition (b) is satisfied since f o r sufficiently s m a l l
the c u r v e c, m a y be obtained f r o m the c u r v e CN by an a r b i t r a r i l y s m a l l
translation. Let u s further choose 6, > 0 which satisfied the following
condition: i f (6)is8,-close t o (D), the simple closed c u r v e which con-
s i s t s of the segment G@, of the s e p a r a t r i x
separatrix g-, and the segment Eo#:
e, the segment GI@: of the
of the a r c l o , is entirely contained
in U J a (Lo) and encloses the c u r v e C N ; the region delimited by the c u r v e s
c and FN is a l s o contained in ULI3(Lo).
The existence of such follows f r o m L e m m a s 9 and 10 of 9 2 8 . 2 , and
a l s o f r o m the fact that, f o r sufficiently s m a l l &, the c u r v e m a y be
obtained f r o m the loop of the s e p a r a t r i x Lo by an a r b i t r a r i l y s m a l l
translation.
We will now show that the number 6 = min {6,, 62} s a t i s f i e s the proposi-
tion of the l e m m a . Let ( 5 )be 6-close to (D) and let Go< .,. Let, further,
K l (k,)be a point on the segment k o B of the a r c l o , sufficiently c l o s e to the
point Ma. Then the path zhthrough t h i s point c r o s s e s again the a r c lo at
a point K z (k,), such that the following conditions a r e satisfied:
( a ) 4 > k,:
(b) the closed c u r v e zk consisting of the segment K I K z of the path z h and
the segment K 2 K l of the path lo encloses the c u r v e FAand together with fN
delimits a region which is entirely contained in U. (Lo).
Let 7 ( u ) be the succession function on the a r c without contact lo c o r -
responding t o (6).By L e m m a 7, this function is defined on the segment
KIN,. A l s o
f ( k i )= & >ki
-
T ( n , ) = nz < n,.
310
f2Q. CRE.ATI@X OF LIhlIT CYCLES FROLI A SEPARATRIX L O O P
Hence i t follows that t h e r e e x i s t s a number u*. k i < u * < n t , such that

-
f (U) = u*. (22)
The point .%I* corresFonding to this-value of the p a r a m e t e r lies between

K ,and and the path L* of s y s t e m ( D ) passing through t h i s point is closed
(in virtue of ( 2 2 ) ) . hZor?over, L* l i e s inside the region delimited by the
s i m p l e closed c u r v e s CT, and cx, and i t is t h e r e f o r e contained in c, ( L o ) .
T h e t h e o r e m is thus proved f o r the c a s e of a loop which is s t a b l e f r o m
inside. An analogous proof c a n be given for the other c a s e s .
!\e have s e e n before that i f the s e p a r a t r i x Loof the saddle point 0 of
the s y s t e m
*=E(=,
dt y), $ y = Q ( r ,y) (D)
f o r m s a loop, the numbers Lo and L; exist f o r the s y s t e m

dt (D:)
-d- t= P - p Q , $=Q+pp,
where p i s sufficiently s m a l l in absolute magnitude, and
( s e e (20) and (21)). T h i s together with T h e o r e m 4 4 lead t o the following

theorem on the creation of a closed path f r o m a s e p a r a t r i x loop:
T h e o Y e m 4 6 . if the separatrix L o of system ( D ) f o r m s a stable or an
unstable loop, then f o r any e > 0 and 6 > o thes-e exists a modified system
( h )6 -close to ( D ) which has at least one closed path L: in the e-neighbor-
hood of the loop Lo.
R emark . It is readily s e e n that if 6 > O i s sufficiently s m a l l , the closed
path L:. whose existence is postulated by the theorem, c r o s s e s the arc lo
at a point M * which lies in C , ( M o ) . The point M * m o r e o v e r lies on the
segment J i o B of t h e a r c I , i f ii; < Lo, and on the segment ;@;B if Lo< E; (the
point .\I* may not lie between the points Ayoand JY;, i f t h e s e are two dif-
f e r e n t points, in virtue of L e m m a 8 ) .
We will s a y that the closed path L: of s y s t e m (6)introduced in T h e o r e m 46
i s c r e a t e d f r o m t h e l o o p f o r m e d b y t h e s e p a r a t r i x Loof
s.ystem ( D ) ( F i g u r e 122).
T h e o r e m s 44 and 4 6 show that i f
GO (10, YO) = P; (=a. YO) f Q I (20. YO) # 0 (23)
for t h e saddle point 0 (zo, yo) of s y s t e m ( D ) , t h e r e always exist modified

s y s t e m s , a r b i t r a r i l y c l o s e to (D), in which the loop formed by the
s e p a r a t r i x Lo c r e a t e s at l e a s t one closed path.
3. T h e uniqueness of the closed path c r e a t e d from

a s e p a r a t r i x loop
We will now show that if t h e s e p a r a t r i x Lo of the saddle point 0 (zo, yo)f o r m s a
loop and condition (23) is satisfied, the s e p a r a t r i x loop m a y c r e a t e at most one
closed path.
31 I
Ch.XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POINT SEPARATIUX
Let u s first prove one l e m m a .

L e m m a 9. F o r any C > o there exist E > oand 6 > 0 such that zy
s y s t e m (E) i s &close to system (D) and has a closed path L: contained
in the e-neighbmhood of the loop L o , the period T of the path L: is gveater
than C .
P r o o f . Alongside with the a r c without contact l o , we consider, a s
before, an a r c 1, ( F i g u r e 139). The s e p a r a t r i x Lo c r o s s e s the a r c s lo and
I , a t points M Qand M icorresponding
t o the t i m e to and t , > t o .
By the r e m a r k t o TheoEem 46, the
closed path L: of s y s t e m (D) intro-
duced in the l e m m a c r o s s e s the I
a r c lo a t a point M*. Suppose that
t h i s intersection point c o r r e s p o n d s I
to t = t o . Clearly, i f E > 0 and 6 > 0
a r e sufficiently s m a l l , L: a l s o i n t e r -
s e c t s the a r c l, for s o m e t: > tQa;
a point M: lying on the segment M t B ,
of the a r c I , , and t: is a r b i t r a r i l y
close to ti ( F i g u r e 139). F u r t h e r -
m o r e , f o r s o m e T* > t:, the path
L: again c r o s s e s the a r c 1, a t the
point M*.
FIGURE 139 Evidently,
T*= t o -+ 7,
where z is the period of the closed path L:.
If E > 0 and 6 > 0 a r e sufficiently s m a l l , the point M * is a r b i t r a r i l y
c l o s e t o the point M o (and t h e r e f o r e to n;i, and @;), and the point M: is
a r b i t r a r i l y close t o M,, i . e . , by L e m m a 13, s28.2, T* is a s l a r g e a s
d e s i r e d . This means, in i t s turn, that f o r sufficiently s m a l l e and 6 ,
w e have
T = T*-lo >c.
Q.E.D.
T h e o r e m 47. I f o o ( x o , ~ o ) # O , t h e r e e x i s t eo>Oand 6,>Osuchthat
any system ( 6 )6,-close to (D) may have at most one closed path in the
e,-neighbmhood of the loop L o . if such a path exists, i t i s a limit cycle
of the same stability as the loop of the original system, L e . , it i s stable
f o r uo ( x 0 , y o ) < 0 and unstable for uo ( x o , yo) > 0.
Proof . To fix ideas, let
00 = 8 0 (XO,Yo) > 0.
By T h e o r e m 44, the loop of s y s t e m (D) is unstable in t h i s c a s e .

Let ei > 0 be so s m a l l that at every point M ( x , y) of U,,(O)we have
Let f u r t h e r x = q0 ( t ) , y = va (4 be the solution corresponding to the

s e p a r a t r i x L o . T h e points M o and M , corresponding t o the t i m e t o and t l
and the a r c s without contact 1, and I , are chosen s o that t h e s e a r c s and
the semipaths O M Qand MIO of the s e p a r a t r i x Lo a r e entirely contained
312
$'7?. CRE.4TI31. OF LI!,fIT CYCLES FRO>,I .A SEP.IR;\TRIX LOOP
i n C:,, (0).L e t , a s b e f o r e , M O B , a n d MlB1 b e t h e s e g m e n t s of t h e a r c s

I , a n d 1, w h i c h a r e e n t i r e l y e n c l o s e d i n s i d e t h e l o o p Lo, e x c e p t for t h e i r
end p o i n t s J I o a n d S I 1 ( F i p r e 139).
Consider the integral
I,
J= I, o(cfo(t), *to(t))dt.
10
Let x >0 be such that
T o g e t h e r w i t h t h e s y s t e m ( D ) , w e w i l l now c o n s i d e r m o d i f i e d s y s t e n i s ( 6 ) .
L e t 6 , > 0 b e a n u m b e r s a t i s f y i n g the f o l l o w i n g c o n d i t i o n : if (6)is
6 , - c l o s e t o ( D ) , t h e n at e v e r y p o i n t .\I@, y) of r , , ( O ) we h a v e
cr (x. Y) = Px (T,Y) t Q; (T,y) >4.

F o r a n y C > 0, t h e r e e x i s t 6 ? > 0 a n d e ? > 0 s u c h t h a t i f (E) is 6 2 - c l o s e t o
(ni a n d h a s a c l o s e d p a t h
L; c o n t a i n e d in C-?(L0) w h i c h c o r r e s p o n d s t o a
solution
t = F* ( t ) , Y = ** ( 0 ( L:)
with p e r i o d T , and if this p a t h crosses the a r c lo a t point .TI* for t = t o a n d

t h e a r c I , a t point -%I: for t = t:. c) < t: - to < T, t h e n t h e f o l l o w i n g c o n d i t i o n s
are satisfied:
(a) T > C ;
b) 1s 6(4;*(t).
1:
to
$*(t))dll<:2x;
(cl t h e p o i n t s of t h e p a t h L,: c o r r e s p o n d i n g t o t t < t<to - T a r e e n t i r e l y

c o n t a i n e d i n C', (0).
For s u f f i c i e n t l y s m a l l 6? a n d e?, c o n d i t i o n ( a ) is s a t i s f i e d i n v i r t u e of
L e m m a 9, c o n d i t i o n (b) i n v i r t u e of t h e t h e o r e m o n t h e c o n t i n u o u s d e p e n d e n c e
of s o l u t i o n s on t h e r i g h t - h a n d s i d e s a n d i n v i r t u e of t h e c o n t i n u i t y of o (I,y).
C o n d i t i o n ( c ) is s a t i s f i e d b e c a u s e t h e d y n a m i c s y s t e m ( D ) is s t r u c t u r a l l y
stable i n a r e g u l a r s a d d l e - p o i n t r e g i o n ( t h i s s t r u c t u r a l s t a b i l i t y f o l l o w s ,
e . g . , f r o m L e m m a -1, 8 9 . 2 ) .
Now l e t 6, = min (&, 6!}. eo =.min {e,. e 2 } .
R'e w i l l s h o w that 6, a n d e,
d e f i n e d i n t h i s w a y m e e t t h e c o n d i t i o n s of t h e t h e o r e m . To t h i s e n d ,
l e t L; b e a c l o s e d p a t h of ( 6 )w h i c h is 6 , - c l o s e t o ( D ) . We a s s u m e t h a t
t h i s p a t h lies i n V , ( L , ) ar,d p r o c e e d t o e v a l u a t e t h e i n t e g r a l
lo TIC
J*= \ o ( ( ~ * ($ f* )( t.) ) d t .
;0
We h a v e
t: toLr
J*= G ( t p * ( t ) ,$ * ( f ) ) d t f j o ( & * ( t )$, * ( t ) ) d f .
:0 1:
313
Ch.XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF A SADDLE-POIN'I SEPARATRIX
By condition (b), the first integral on the right is g r e a t e r than -22. The
second integral is g r e a t e r than 9 (to+z--t:) by inequality ( 2 6 ) . Therefore
I
or, by condition ( a ) I
Since for s m a l l 6, t: is close to the constant number t l , we conclude

that for sufficiently l a r g e C (i. e., f o r sufficiently s m a l l tio and E ~ ) ,J*> 0 .
R u t thenby Theorem 1 7 ( 5 1 3 . 3 ) , the closed path L: of s y s t e m (6) is an
unstable limit cycle.
We have thus proved that i f E,, > 0 and bo > 0 are sufficiently small, all
the closed paths of a s y s t e m (6)6,-close to (D) which l i e in U % ( L o )a r e
unstable limit cycles. Suppose that m o r e than one such path e x i s t s . We
c a n then evidently find two closed paths L: and L: contained in Lreo(Lo),such
that one is enclosed by the other and the region delimited by the two paths
contains no other closed paths. T h i s is an obvious contradiction, however,
since both paths t: and L: a r e unstable limit c y c l e s (we naturally a s s u m e
that U e o ( L 0contains
) no other equilibrium s t a t e s of (B), except the saddle
point 6 ) . T h i s proves the theorem.
1
A s i m i l a r proof can be given f o r u0 (zo, yo) < 0.
R e ma r k . T h e o r e m 4 7 can be generalized. Indeed, let s y s t e m ( D )
have a closed contour y consisting of the s e p a r a t r i c e s of the saddle points
0, ( x ' , yz), i = 1, 2, . . ., n, n 2 2 , and the saddle points themselves. Reasoning
along the s a m e lines a s before, w e can show that i f u (z', y,) < 0 ( i =
= 1, 2, . . . , n), the contour y is stable, and if u (x', y,)> n, i t is unstable,
and that y may c r e a t e a single limit cycle, which is stable in the f o r m e r I
c a s e and unstable in the l a t t e r .
The following theorem follows almost immediately f r o m T h e o r e m 47.
T h e o r e m 4 8 . Let the separatrix Lo of saddle point 0 (xO,go) of system
(D) form a loop and let condition ( 2 3 ) be satisfied, i. e . , u0 (x0, go) + 0.
Then there exist E > Oand 6 > osuch that V system (6) i s &close to (D)
and in the &-neighborhoodof the loop Loit has a separatrix Loof a saddle
point 6 which f o r m s a loop, (6)can have no closed paths in the &-neighbor-
1 * r . . * . -
Prbof. F o r $ a n d 8 we m a y t a k e
where eo and h0 a r e numbers defined by T h e o r e m 46. Indeed, let (5) be

6-close to (D) and suppose that in the &-neighborhoodof the loop Lo (6)h a s
a s e p a r a t r i x Lo which f o r m s a loop and a closed pathz,. We a s s u m e that
E , i s not a simple limit cycle of (6). By Theorem 19 ( 5 15.2), t h e r e e x i s t s
a s y s t e m (D) 8-close to (D) which ha: a t l e a s t two closed paths e,
and t,in
t h e s-neighborhood of the path &. (D) is evidently &-close to (D), and the
paths t1and & lie in thee,-neighborhood of the loop Lo. Now, t h i s contra-
d i c t s T h e o r e m 47. Thus, E, may not be a multiple limit cycle of ( E ) .
Now let zl be a simple, i . e . , structurally stable, limit cycle. By
314
52'. CRE.'ITIOE; OF LLh'LT CYCLES FROhl A SEP.4RATMX L O O P
T h e o r e m 46, t h e r e e x i s t s a s y s t e m ( D ) a r b i t r a r i l y c l o s e t o ( E ) w h i c h h a s
a . c l o s e d p a t h io i n a n y a r b i t r a r i l y small n e i g h b o r h o o d of t h e l o o p &,. If
i s s u f f i c i e n t l y . c l o s e to ( E ) , t h e n i n v i r t u e of t h e s t r u c t u r a l s t a b i l i t y
ol t h e c y c l e E l , ( D ) may' h a v e a l i m i t cycle e,. which is a r b i t r a r i l y close t o
t h e c,ycle El, w h i l e t h e two c y c l e s e, a n d il d o not c o i n c i d e a n d a r e b o t h
c o n t a i n e d i n t h e e - n e i g h ? o r h o o d of t h e l o o p L o . B u t t h i s a g a i n c o n t r a d i c t s
T h e o r e m 45.
We w i l l p r o v e a n o t h e r p r o p o s i t i o n Lvhich, i n a s e n s e , s u p p l e m e n t s
T h e o r e m 15,
T h e o r e m 49. If the separatrix L , of the saddle point 9 ( x , , yo) of
systein ( 0 )forms a looi? and U, (I,, yo) > 0 (4), there exist e > oand 6 > 0
with the folloiringpropc~rty: any system ( 6 )which is 6-close to ( D i and
f o r which iTo < G; (cowespondingly, Lo> Lo)has no closed paths in r J e (Lo).
P r o n f . To fix i d e a s , l e t u s c o n s i d e r t h e case uo (to,yo) < 0 . In t h i s
c a s e , t h e l o o p f o r m e d b y t h e s e p a r a t r i x L,, i s s t a b l e (s29.1, T h e o r e m 44).
Let e , > 0 a n d 6, > 0 b e t h e n u m b e r s d e f i n e d b y T h e o r e m 4 7 .
I . C o n s i d e r a point .!I1 (u,)of t h e a r c l o , w h e r e u t > u o . If t h e point M I
is s u f f i c i e n t l y close t o .lIo, t h e p a t h L 1 p a s s i n g t h r o u g h t h i s p o i n t for 1 = to
w i l l c r o s s t h e a r c l o a g a i n a t point .Yl a s t i n c r e a s e s . Let Cl b e a s i m p l e
c l o s e d c u r v e f o r m e d b y t h e s e g m e n t JI,.Y, of t h e p a t h L1 a n d t h e s e g m e n t
.\.IJfIof t h e a r c I , . U'e c h o o s e u, i n s u c h a w a y t h a t t h e f o l l o w i n g c o n d i t i o n s
are satisfied:
( a ) t h e c u r v e C1, a n d t h e r i n g r e g i o n e n c l o s e d b e t w e e n t h e l o o p of t h e
s e p a r a t r i x L o a n d t h e c u r v e CI a r e c o n t a i n e d i n C.?,(L0);
!bJ d ( u , ) < o .
R n t h c o n d i t i o n s are c l e a r l y s a t i s f i e d i f u I is s u f f i c i e n t l y close to u,(the
s e c o n d c o n d i t i o n h o l d s t r u e i n v i r t u e of o u r a s s u m p t i o n t h a t t h e l o o p is
stable).
11. W e c h o o s e 61 > 0 s o small t h a t if s y s t e m (6)is 6 , - c l o s e t o ( D ) , t h e
following conditions are s a t i s f i e d :
i c ) d"(u,)<O;
( d ) t h e c u r v e El a n d t h e r e g i o n b e t w e e n t h i s c u r v e a n d t h e l o o p of
s e p a r a t r i x Lo a r e c o n t a i n e d i n ( L o ) ( t h e f u n c t i o n d" is a n a l o g o u s t o d ,
a n d C?, i s t h e c u r v e p a s s i n g t h r o u g h M iw h i c h i s a n a l o g o u s t o Cl; F i g u r e 140).
FLCCQE 140
111. We choose E? > 0 and 6,> 0 s o t h a t t h e y s a t i s f y t h e f o l l o w i n g c o n d i t i o n s ,
(e, E,<?:
315
(f) i f(a) is &-close to (D), the curve i?, and the region enclosed by
t h i s c u r v e do not i n t e r s e c t with U s 2 ( L o ) .
IV. We choose e3 > 0 and 63> 0 s o that the following conditions a r e
satisfied :
(g) e3<%;
(h) i f (0)is 6,-close t o (D), and xis a closed path of (b) contained in
U e 3( L o ) , the pathzc r o s s e s the a r c without contact lo a t point M ( L ) , where
<> La(&is the valce of the p a r a m e t e r corresponding t o the point Ma).
T h e existence of the numbers e3 and 8,follows f r o m the fact that i f (b)
is sufficiently c l o s e to (D) and moreover u< Loo, the closed path should
enclose the saddle point 6 and, with it, all of i t s s e p a r a t r i c e s . But then
i t cannot b e contained in a sufficiently s m a l l neighborhood of the loop.
We will prove that the n u m b e r s
e =min{EZ, Ea), b=min {+, d,, 62, 6,) (29)
satisfy the proposition of the t h e o r e m .

Suppose that this is not so. Then t h e r e e x i s t s a s y s t e m (b) 6-close to
( D ) f o r which
u,>E; (30)
and which h a s a closed path in U.(L,). Let this path c r o s s the a r c without
contact I , at point h?(ii).
F r o m ( 2 9 ) and conditions ( f ) and ( h ) it follows, a s is readily seen, that
Lo <L <u,. (31)

Consider the function d(u) of system (5).Since L is a closed path, we
have
d (L)= 0. (32)
Moreover, by condition (c)
h(u,) <0. (33)
Finally, i f u2 is sufficiently close to ioand ~ o < u 2 ( ~ we
, have
;l(u2)< 0 (34)
in virtue of inequality ( 3 0 ) .
It follows f r o m ( 3 2 ) , (33), and ( 3 4 ) that e i t h e r the function d (u)h a s at
l e a s t one m o r e root ;*, u p <*; < u,, b e s i d e s the root E, o r z(ii)
= 0 . If
the first alternative is t r u e , a closed path z* of (b) will p a s s through the
point a* (;*); in virtue of condition (d), t h i s path is contained in U, (La),
i . e., t h i s neighborhood contains a t l e a s t two closed paths. T h i s c l e a r l y
contradicts the choice of eo > 0 and 6o > 0 .
If the second alternative is true, the closed path z
is not a s i m p l e
liAmitcycle. But then, by T h e o r e m 19, 5 15.2, t h e r e e x i s t s a s y s t e m
(D) a r b i t r a r i l y c l o s e t o (E), and in p a r t i c u l a r &-close t o ( D ) , which h a s
a t l e a s t two closed paths in U, (La), and t h i s again contradicts the choice
of 8o and eo. Q. E. D.
316
929. CREATION OF LIhlIT CYCLES FRO41 .4 SEPARATMX LOOP
R em ar k . yo) b e a saddle point of the s y s t e m

Let 0 (q,,
whose s e p a r a t r i x Lo f o r m s a loop, and let
00 ( I o . Yo) = P; (zo, Yo) +QI (20. Yo) f 0.

X rotation of the vector field of ( D ) through the angle tan-' p produces
a s y s t e m of the f o r m
-&
=P-pQ, *=Q+pP.
dt dt
Depending on the s i g i of p, we obtain e i t h e r inequality ( 2 0 ) o r (211, i.e.,

- r
u , t u ; or uo> u;.
e -
It follows from t h e s e inequalities that when the vector field of ( D ) is rotated

through a sufficiently s m a l l angle, the s e p a r a t r i x loop is broken.
T h e o r e m s 45, 47, and 49 show that when the field is rotated in one of
the two possible directions, the broken s e p a r a t r i x loop is replaced by a
limit cycle of the s a m e stability, which a p p e a r s i n the neighborhood of
the loop: when the field is rotated in the opposite direction, no closed
paths a r e observed in a sufficiently s m a l l neighborhood of the broken
loop.
4. The case P ; (5.yo) + (ZO.YO)= 0
Let u s now c o n s i d e r :he c a s e when the saddle point 0 (IO,yo) of s y s t e m

(n) h a s a s e p a r a t r i x Lo forming a loop, but a. (xo, yo) = 0.
We will show that in t h i s case t h e r e exist dynamic s y s t e m s a r b i t r a r i l y
c l o s e t o ( D ) which have a t l e a s t two closed paths in any a r b i t r a r i l y s m a l l
neighborhood of t h e loop L o .
L e m m a 10. Let ( E ) be a dynamic s y s t e m , 0 (x0, yo) a saddle point
of the s y s t e m , L~ the se9aratrix of this saddle point forming a loop, and let
a. (20, yo) = P; (rotyo) +Q;(x0.yo) =0. (35)

=-
Then for any e o and 6 > 0 , there exists a system ( 6 )which satisfies
(a) ( 6 )i s a-close to ( D ) .
(b) 0 (z~, yo) is a saddle point of ( 6 ) ,and
GJ (5,Yo) = E (zo, Yo) + 0;(zo, Yo) >0 (m<0 ) .

(c) The saddle point ,9 of (b) has a separatrix z,, which f m r n s a loop and
i s entirely contained in the &-neighborhoodof the loop Lo of system (D).
Proof . Without l o s s of generality, w e take the saddle point 0 a t the
origin and a s s u m e that the s y s t e m is ,given in canonical f o r m . Then
xo = yo = 0 and, by ( 3 5 ) , the s y s t e m h a s the f o r m u > uo.
317
where v + 0, and the functions P z and Q 2 together with t h e i r f i r s t derivatives

vanish at the origin.
Let, as before, I , be an arc without contact through s o m e point Moof
the s e p a r a t r i x L o , and u a p a r a m e t e r on t h i s arc. The point h2, c o r r e s p o n d s
t o u = u,,, and the points of the a r c I , lying inside the loop correspond to
> Go.
Consider a modified s y s t e m of the f o r m
dr
x = V z + P z ( z 7 i / ) = P a ( z , u), $=- ( ~ - a ) y + Q z ( ~ , Y ) = Q ~ ( z y).
, (Da)
If a is sufficiently s m a l l in absolute magnitude, 0 ( 0 , 0) is a saddle point

of t h i s s y s t e m , and
P L (0, 0 ) + Qh (0, 0 ) =a. (36)
Together with ( D o ) , we consider another s y s t e m
dx
x=Pa-pQa = v x + P ~ (I,Y)- u r - (Y -a) Y+ Qz (z,Y)I =Paw ( ~ v),
9
(Dab)
$ = Qa + p p a = -( v - a ) Y +~z,I( Y) + P ~ ~ z + p z Y)I =Vow
(2, ( ~ Y)*
7
whose vector field is obtained by rotating the vector field of ( D o ) through

the angle tan-lp. F o r any sufficiently s m a l l (in absolute magnitude) p,
0 is a saddle point of ( D a w )and
K p x (0, 0) + Qbpy (0,

Let a > 0.
Let, a s before, L: and Lb b e t h e semipaths comprising the s e p a r a t r i x Lo
which contain the point M o . For any e > 0 and 6 > 0, t h e r e exist a,> 0 and
pa > 0 such that i f I a I < a, and I p I < po, ( D a w )is & c l o s e to (D) and has
two s e p a r a t r i c e s LtDp and L&,of the saddle point 0 in the e-neighborhoods
of the s e p a r a t r i c e s L: and L6, respectively. Let t h e s e s e p a r a t r i c e s c r o s s
the a r c lo at points M o ( a , p) and M ; (a,p), which correspond to the values
u,, (a,p) and u; (a,p)of the p a r a m e t e r u .
F i r s t let p = 0, i.e., consider the s y s t e m ( D a ) . Two c a s e s a r e possible
a priori:
1) T h e r e exist a r b i t r a r i l y s m a l l positive numbers a* such that
ug (a*,0) = u; (a*,
O), (37)
i . e., the s e p a r a t r i c e s Lf.0 and L&.o m e r g e into a single s e p a r a t r i x of ( D o * )
which f o r m s a loop. Clearly, f o r any e > 0 and 6 > 0 and f o r a sufficiently
s m a l l a*, (D,,) is & c l o s e t o ( D ) , and t h i s loop is contained in U,(L,), which
proves the l e m m a .
2 ) T h e r e e x i s t s f3> 0, such that f o r all a, 0 < a < p, e i t h e r
uo(a,0 ) > u; (a. 0). (38)

or
( i f t h e r e exist a r b i t r a r i l y s m a l l a f o r which inequality (38) is satisfied and

a l s o a r b i t r a r i l y s m a l l a f o r which (39) is satisfied. t h e r e a l s o exist
318
$29. CREATION OF LIkIIT CYCLES FRO!.' .4 SEP.%R.\TRIX L O P
at*bitrarily s m a l l a* f o r which ( 3 7 ) is satisfied, and we r e t u r n t o

c a s e 1).
To fix ideas, suppose that inequality { 3 8 ) is satisfied for a l l a,0 < a 0 and is sufficiently s m a l l , we have
uo (0. p) <u; ( 0 , p) (40)

(see (20)).
0 and E > 0 b e fixed and let a. and po b e the positive n u m b e r s
Let i?~>
introduced above, corresponding t o t h e s e 6 and E .
Let a, be a fixed positive number, a l < a , , a,<p. Then
F r o m the l a s t inequality and r e m a r k to L e m m a 10, s 2 8 . 2 , i t follows

that i f p* is sufficiently s m a l l in absolute magnitude, w e have
P*) > u; (at, p*).

uo (a1, (42)
Let 0 <p* <po. By inequality (40). f o r a s m a l l positive p*,
ua (0,P*) <u; (0,p*). (43)
F r o m inequalities ( 4 2 ) and ( 4 3 ) and continuity of the functions u, and u; it

follows that f o r s o m e a*. O(a*<a,,
uo(a*,p*) = u; (a*.p').
T h i s m e a n s that the s y s t e m
ddtl
-= VZ+-~Z(Z, Y)-p*I-(v-aa*)yr42(J, Y)I,
(Dr;t&A*)
2
-= -( v - a * ) y+Qa (I, Y) + P* I v z f P , Y)I
(1,
h a s a s e p a r a t r i x Eo which f o r m s a loop. Since then O<a*<aO, 0 <p* f of the l e m m a .
T h e o r e m 50. Let o(to.yo) be a saddle point of dynamic system (D),
and L , its separatyix f o m i i n g a loop. if oo ( x 0 , yo) = 0, then for any e oand =-
6 > 0 there exists a modified system ( 6 )which i s 6-close to ( D ) and which
has at least two closed paths in the e-neighbmhocd of the loop L o .
Proof . For simplicity, l e t I,
= yo = 0 . If any neighborhood of the
loop Lo contains closed paths, ( D ) itself may be chosen as (6). It thus
suffices t o consider the c a s e when s o m e neighborhood of the loop Locon-
t a i n s no closed paths, i.e., the loop Lo is e i t h e r stable o r unstable. Let
lo b e a n a r c without contact passing through the point M o of the loop Lo, u
a p a r a m e t e r defined on I , , ~0 the value of t h i s p a r a m e t e r corresponding
t o hio. W e a s s u m e , a s before, that the points of the a r c lo which lie
inside the loop correspond to u , u0 < u g b o , and the succession function
319
Ch-XI. CREATION OF LIMIT CYCLES FROM T H E LOOP OF A SADDLE-POIhT SEPARATRIX
-
u = f (u) of ( D ) on the a r c lo is defined for all u ,u,, < u Q 6, where b o.

Let e > 0, 6> 0 b e given.
I. Choose ut >uo sufficiently c l o s e t o uoso that the following conditions I
a r e satisfied:
(a) d(ud>O.
( b ) T h e path LI which f o r t = top a s s e s through the point M I (u,)of the
a r c without contact lo will again c r o s s the arc lo at point N , as t i n c r e a s e s ,
s o that t h e c u r v e C , consisting of the segment M , N , of t h e path L, and the
s e g m e n t NiMs of t h e a r c l o , together with the region enclosed between the
c u r v e C, and t h e loop of the s e p a r a t r i x L o , are contained in t h e e14-neighbor-
hood of the loop L,(Figure 141).
FIGURE 141.
11. Let q be the distance between the c u r v e C, and the s e p a r a t r i x loop Lo.
We choose 6 , > 0 so s m a l l that the following conditions a r e satisfied:
tl
(a)
(b) If (B) is 6,-close to (D), then
and c u r v e -
- t h e analog of C, is contained i n U,,,'(C)and a l s o in UE,' (0.
( c ) If ( 6 )is 6,-close t o (D) and fi, (Lo)is the intersection point of the
corresponding s e p a r a t r i x z, with t h e a r c without contact io, we have < ut
and_the s u c c e s s i o n function F(u) of (b) on the a r c lo is defined f o r all
ur ~ o < U < u f .
111. We choose p* and a* s o t h a t the following conditions are satisfied:
(a) T h e s y s t e m (D&,,,,) considered i n L e m m a 10 is 6-close to (D).
(b) T h e s e p a r a t r i x Eo of the saddle point 0 of (De,,,*) f o r m s a s t a b l e
loop, which is e n t i r e l y contained both in U t / , (Lo)and i n UnlC(Lo).
T h e existence of t h e n u m b e r s p* and a+ satisfying conditions (a) and (b)
h a s been established i n t h e proof t o L e m m a 10, In p a r t i c u l a r , f o r the
loop z, t o b e stable, we should have a*< 0 .
320
$2'). CREATION OF LIhlIT CYCLES FROhl A SEP.AR-1TfUX LOOP
On the a r c I,, choose a point M , ( u , ) sufficiently c l o s e t o A ~ , ( u 2 > & ) .

Let Z ( U ) = ~ - ( U ) - - Uw'qere
, F ( u ) is the succession function of (Doc,,+). Since
the loop Eo of (D,*,,,) i s stable by assumption, w e have
iT(ut) < 0. (45)

On the other hand, inequality (44) is satisfied. F r o m t h e s e relations it
follows that, for s o m e u*, u 2 < u * < u i r
d (u*)= 0,
i . e . , a closed path I?* o f (Do*,,:) p a s s e s through the point 11.1' (u*).

Let (D1.$.)be designated ( 0 ; : ) . F r o m I I ( a ) and I I I ( a ) it follows that (6:s)
is b'2-close to ( D ) . F u r t h e r m o r e , as we have seen, ( D ) h a s a s e p a r a t r i x
zowhich f o r m s a stable loop and a closed path E*. Both the loop Eo and the
closed path .E* a r e contained in U,2 ( L o )- the loop in virtue of I11 ( b ) and the
path i* in virtue of I ( b i , II(b), and I I I ( b ) .
If the closed path E* of (D';) is not a simple limit cycle of t h i s s y s t e m ,
Theorem 19, $15.2, indicates that t h e r e e x i s t s a s y s t e m (8)d/2-close
to (E,':)which h a s in U,. ,(E*)at least two closed paths Et and z,. But
then ( 6 )s a t i s f i e s the proposition of the t h e o r e m . If t*i s a simple, i . e.,
s t r u c t u r a l l y stakle, limit cycle of (E::), then T h e o r e m 4 6 shows that t h e r e
e x i s t s s y s t e m (DI a r b i t r a r i l y c l o s e to (E.;:)which h a s a closed path in
any a r b i t r a r i l y s m a l l neighborhood of the loop E , . When (6)is sufficiently
close to (E*), s y s t e m tb), in virtue of the s t r u c t u r a l stability of the cycle
E*, h a s a limit cycle which lies a r b i t r a r i l y close to 2*, and the paths
L, and L2a r e different and are both contained in U, ( L o ) . Thus, s y s t e m ( D )
s a t i s f i e s the propositicm of the t h e o r e m . T h i s completes the proof.
R emark . T h e o r e m 50 proves that i f u, (30,yo) = 0, the s e p a r a t r i x
loop of the saddle point 0 (so, yo) may c r e a t e a t l e a s t two closed paths.
The question of the l a r g e s t number of closed paths which may be c r e a t e d
from a s e p a r a t r i x loop in the c a s e uo (z,, yo) = 0 and the conditions d e t e r -
mining this number r e q u i r e s a much m o r e complex analysis. T h e problem
$was considered by E . A . Leontovich in h i s t h e s i s and the r e s u l t s a r e p r e -
sented i n !21/.
32 1
C h a p t e r XII
CREATION OF A LIMIT CYCLE FROM THE

LOOP OF A SADDLE-NODE SEPARATRIX.
SYSTEMS OF FIRST DEGREE OF STRUCTURAL
INSTABILITY AND THEIR BIFURCATIONS
INTRODUCTION
T h e f i r s t of the two sections in t h i s chapter, $30, d e a l s with the creation

of a limit cycle f r o m the loop of a saddle-node s e p a r a t r i x . Let (A) be a
dynamic s y s t e m , Mo ( I O , yo) an equilibrium s t a t e of t h i s s y s t e m , which is a
saddle-node of multiplicity 2 . A canonical neighborhood of a saddle-node
c o m p r i s e s a parabolic s e c t o r and two hyperbolic s e c t o r s , s e p a r a t e d f r o m
one another by t h r e e s e p a r a t r i c e s . To fix ideas, suppose that the saddle-
node Mo h a s one a- s e p a r a t r i x L; and two w - s e p a r a t r i c e s L; and 1;:.
Suppose that the a - s e p a r a t r i x L; goes to M o for t - t . + 00, i. e., i t f o r m s a
loop, w h e r e a s none of the s e p a r a t r i c e s L; and L: is a continuation of L;
(Figure 142).
Since M o is a double equilibrium s t a t e of (A), t h e r e exist a r b i t r a r i l y
c l o s e s y s t e m s which have no equilibrium s t a t e s in the neighborhood of Mo.
The main result of 3 O s t a t e s that i f t h e e q u i l i b r i u m s t a t e M o , a n d
consequently the separatrix loop, disappear following a
s u f f i c i e n t l y s m a l l c h a n g e i n s y s t e m (A), o n e a n d o n l y o n e
l i m i t c y c l e i s c r e a t e d i n t h e n e i g h b o r h o o d of t h e l o o p
( T h e o r e m s 51 and 52, see Figure 143).
The second section, 31, d e a l s with the s i m p l e s t s t r u c t u r a l l y unstable

s y s t e m s , namely s y s t e m s of the f i r s t d e g r e e of s t r u c t u r a l instability. A
s y s t e m of the f i r s t d e g r e e of s t r u c t u r a l instability inside a cycle without
322
5 :30. CRI:\TION OF L I l l I T CYCLE FROhl SEPAR.4TRIX LOOP
c n n r a c t w a s d e f i n e d i n C h a p t e r VI11 ( s 2 2 ) . In $31, a d e f i n i t i o n of t h e s e
s y s t e m s is g i v e n for a n y b o u n d e d r e g i o n , a n d t h e n e c e s s a r y a n d s u f f i c i e n t
c o n d i t i o n s a r e e s t a b l i s n e d f o r t h e s y s t e m t o be of t h e first d e g r e e of s t r u c -
t u r a l instability in that region ( T h e o r e m 67). T h e s e conditions are then
a p p l i e d t o i n v e s t i g a t e t i e b i f u r c a t i o n s of s y s t e m s of t h e f i r s t d e g r e e of
s t r u c t u r a l i n s t a b i l i t y . A11 t h e s e b i f u r c a t i o n s t u r n o u t t o b e p a r t i c u l a r cases
of t h e b i f u r c a t i o n s c o n s i d e r e d i n t h e p r e v i o u s c h a p t e r s a n d i n 5 30.
930. C R E A T I O N O F A L I M I T C Y C L E F R O M T H E
LOOP O F A S A D D L E - N O D E S E P A R A T R I X
1. The existence theorem
We c o n s i d e r a n a n a l y t i c a l d y n a m i c s y s t e m
Lvhich h a s a n e q u i l i b r i u m s t a t e yo) of m u l t i p l i c i t y 2 of t h e s a d d l e - n o d e
t y p e , w i t h o n e of t h e c h a r a c t e r i s t i c n u m b e r s d i f f e r e n t f r o m z e r o . Without
l n a s nf g e n e r a l i t y , w e m a y a s s u m e t h a t t h i s e q u i l i b r i u m s t a t e is at t h e
o r i g i n , i . e . , JO = y o = 0 . T h u s ,
and
Q;( 0 , 0 )# 0.
Is (0, 0)= P; (0, 0 ) i-
E q u i l i b r i u m s t a t e s of t h i s t y p e w e r e c o n s i d e r e d i n C h a p t e r 1.111 ( $ 2 3 ,
1 and 2 ) . X c a n o n i c a l n e i g h b o r h o o d of s u c h a s t a t e c o n s i s t s of o n e p a r a b o l i c
sector a n d t w o hypet+bol.ic s e c t o r s . S u p p o s e t h a t t h e p a t h s of t h e p a r a b o l i c
s e c t o r g o t o 0 for t -r f 00. T h e n t h e e q u i l i b r i u m state 0 h a s o n e a - s e p a r a t r i u
L; a n d t w o a - s e p a r a t r i c e s L ; a n d L ; .
S u p p o s e r h a t t h e p a t h L o , f r o m w h i c h t h e s e p a r a t r i x L; i s
c u t o u t , g o e s t o t h e e q u i l i b r i u m s t a t e 0 for t - + - ~ ,
a s w e 1 1 a s f o r t -+ -- m , i . e ,, i t f o r m s a loop. W e m o r e o v e r a s s u m e
t h a t n o n e of t h e s e p a r a t r i c e s L ; a n d L; forms p a r t of t h i s l o o p ( i . e ., t h e
s e p a r a t r i x L; d o e s not merge i n t o a single p a t h e i t h e r w i t h L : o r with L f ,
see F i g u r e 1 4 2 ) .
S i n c e 0 i s a n e q u i l i b r i u m state of m u l t i p l i c i t y 2, t h e r e e x i s t 6*>0 a n d
cd= 0 s u c h t h a t i f s y s t e m (6)is &,-close t o ( D ) t o r a n k 2 , i t h a s a t m o s t t w o
e q u i l i b r i u m s t a t e s i n U,,(O) (see D e f i n i t i o n 15, $ 7 . 3 a n d D e f i n i t i o n 5, $ 2 . 1 ) .
T h u s , t h r e e cases are p o s s i b l e a p r i o r i :
1) S y s t e m (6)h a s o n e e q u i l i b r i u m s t a t e d i n U,,(0).
2 ) S y s t e m (A) h a s t w o e q u i l i b r i u m s t a t e s G, a n d G2i n U., (0).
3 ) S y s t e m ( 6 )h a s n o e q u i l i b r i u m s t a t e s i n U,, (0).
A11 t h e t h r e e cases a r e a c t u a l l y o b s e r v e d in pr_actice: cases 2 a n d 3 i n
v i r t u e of T h e o r e m 3 4 , $ 2 3 . 1 ; case 1 o b t a i n s i f ( D ) is i d e n t i f i e d w i t h ( D )
itself, say.
323
a.XU. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARAIRIX
It iz readily s e e n that if (E) is sufficiently close t o (D), the equilibrium

s t a t e 0 in c a s e 1 is a l s o of multiplicity 2 and is a saddle-node (this follows
f r o m the r e m a r k to T h e o r e m 35, $23.2).
In c a s e 2 , one of the equilibrium s t a t e s and G2 into which the saddle-
node 0 decomposes is a s t r u c t u r a l l y stable node, and the other is a s t r u c -
t u r a l l y s t a b l e saddle-point ( $ 2 3 . 1 , Lemma 1 and 9?3.2, Theorem 35).
We will now consider in m o r e detail c a s e 3, i . e . , the c a s e when
the equilibrium state disappears on passing to a close
s y s t e m . W e will show that a closed path n e c e s s a r i l y f o r m s in the
neighborhood of the loop Lo in t h i s c a s e . In fact, the following theorem
can be stated:
T h e o r e m 51. F m a n y E > O t h e r e e x i s t s 6 > O s u c h t h a t i f ( 6 ) i s
&close to (D) and has no equilibrium states in U, (O),(6)has at least one
c2osed path contained in the neighborhood of the loop L o .
P r 0 0 f . Let E > 0 be given. Consider a canonical neighborhood V of
the equilibrium s t a t e 0 of s y s t e m (D) delimited by the following path
segments:
1) the segment K i K z of the a r c without contact 1 withendpoints R , and R 1
which m e e t s the s e p a r a t r i c e s L;,L:, and L; at the points N o , N i , and NI,
respectively ( F i g u r e 144);
2 ) the a r c without contact lo with end points Mi and M z which m e e t s the
s e p a r a t r i x L; at the point M o ;
3) the a r c s of paths K , M i and K 2 M z .
FIGURE 144
The concept of a c a n o n i c a l n e i g h b o r h o o d h a s been introduced in

QT (Chapter VIII, 19.2), where i t is a l s o shown that a canonical neighbor-
hood can always b e constructed f o r any a r b i t r a r i l y s m a l l neighborhood of
an equilibrium s t a t e . We may thus a s s u m e that the canonical neighborhood
V and the a r c without contact 1 lie in Uc12(0)and that the following condition
is satisfied: the paths Lot and Lo2which for t = to p a s s through the r e s p e c -
tive end points Mi and M z of the a r c lo c r o s s , with i n c r e a s i n g t , the arc
without contact 1 at points M ; and M ; , so that the quadrangle A limited by
324
$ 30. CREATION OF LLhlIT CYCLE FROhI SEP.4R.ATRIX L 3 0 P
t h e a r c w i t h o u t c o n t a c t I , , t h e s e g m e n t JItAI;of t h e a r c w i t h o u t c o n t a c t 1 ,
a n d t h e a r c s Ml+%I; of r h e p a t h s Lol a n d Lo? is a n e l e m e n t a r y
a n d JIlc-1I.~
q u a d r a n g l e c o n t a i n e d i n t h e e / 2 - n e i g h b o r h o o d of t h e l o o p Lo. S i n c e t h e
p a t h L o crosses t h e a r c s without c o n t a c t lo a n d 1 a t t h e i r i n t e r i o r p o i n t s
.\To a n d So, t h i s c o n d i t i o n is s a t i s f i e d w h e n e v e r the a r c lo is s u f f i c i e n t l y
small.
L e t 6: 0 b e so s m a l l t h a t i f (6) is 6 - c l o s e to (D), t h e n
( a ) t h e a r c s lo a n d 2 are a r c s w i t h o u t c o n t a c t for (6);
(b) the paths zol and of (b) w h i c h f o r t = fo p a s s t h r o u g h t h e p o i n t s 11.11
and r e s p e c t i v e l y , cross, w i t h i n c r e a s i n g t , t h e a r c t a t t h e p o i n t s JY;
a n d .%?;, and t h e r e s u l t i n g q u a d r a n g l e a ( t h e a n a l o g of A ) i s a n e l e m e n t a r y
q u a d r a n g l e of (fi) c o n t a i n e d i n U, ( & ) ( F i g u r e 145);
( c , as t d e c r e a s e s , t h e paths zol a n d Eo*c r o s s t h e a r c 1 a t p o i n t s g , a n d
Et,a n d t h e n e i g h b o r h o o d of t h e point 0, d e l i m i t e d by t h e a r c without
c o n t a c t lor t h e s e g m e n t Elk2 of t h e a r c I , and t h e a r c s of p a t h s WIMl a n d
i?J12> i s c o n t a i n e d i n U ,( L o ) .
FICLiRE 145
C o n d i t i o n s ( a ) , ( b ) , a n d ( c ) are s a t i s f i e d for a s u f f i c i e n t l y s m a l l 6 in
v i r t u e of L e m m a s 1 a n d 5 of 3 4 . 1 a n d L e m m a 7 of 3 4 . 2 .
W e w i l l s h o w t h a t t h e n u m b e r 6 s e l e c t e d in t h i s w a y f u l f i l l s t h e p r o -
p o s i t i o n of t h e t h e o r e m .
L e t (6)b e a s v s t e m 6 - c l o s e to (D) w h i c h h a s no e q u i l i b r i u m states i n
U, (0). L e t 6' b e t h e u n i o n of t h e s e t s and ( t h e s e s e t s are a s s u m e d t o
b e closed). Evidently, i s a c l o s e d n e i g h b o r h o o d of the l o o p L o , a n d i n
rv
v i r t u e of ( b ) a n d ( c ) c U, ( L o ) . W e d e f i n e a p a r a m e t e r u o n t h e a r c lo,
s o t h a t t h e e n d p o i n t s Mr a n d MI of t h e arc c o r r e s p o n d to t h e v a l u e s ul a n d
u: o f t h e p a r a m e t e r , u1 < L:~. C o n s i d e r a n y point hi fu) of t h e a r c lo (ul4 u g u - )
and t h e path of s y s t e m (8)t h r o u g h t h a t p o i n t . As t i n c r e a s e s , t h i s p a t h
crosses t h e a r c w i t h o u t c o n t a c t 1 a t some point $of t h e a r c I a n d e n t e r s
into t h e neighborhood v. Since t h e r e are n o e q u i l i b r i u m s t a t e s , and t h u s
n o l i m i t c o n t i n u a of (E), in P, t h e p a t h should leave v a s f f u r t h e r i n c r e a s e s .
325
Ch-XII. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX
It m a y evidently leave P only through the a r c without contact l o , i . e., it

again c r o s s e s the arc Lo at s o m e point M * ( u * ) . T h i s m e a n s that f o r all
u,u i f u , u I , f ( u Z ) < u 2 .

It follows f r o m the l a s t inequalities that t h e r e e x i s t s a number ii, u , < ~ c u 2 ,
such that f (u) =&.
The path L through the point fi(t)of the a r c Io is a closed path. C l e a r l y ,
e c iii c u, (Lo).
T h i s proves the t h e o r e m .
We shall saythat t h e c l o s e d p a t h i i s c r e a t e d f r o m t h e
l o o p o f t h e s e p a r a t r i x Lo o f t h e s a d d l e - n o d e O .
R e m a r k . T h e o r e m 5 1 and the above proof r e m a i n valid i f the point 0
is a n y e q u i l i b r i u m s t a t e o f m u l t i p l i c i t y 4 ( a n d not only 2 )
f o r w h i c h c o n d i t i o n s 1 a n d 2 a r e s a t i s f i e d (thisequilibrium
s t a t e is a l s o a saddle-node, see $23.2, a ) .
2. T h e uniqueness t h e o r e m
W e will prove the uniqueness of the closed path created f r o m the loop of
a saddle-node s e p a r a t r i x . As in the previous subsection, w e consider an
equilibrium s t a t e 0 (0, 0) of multiplicity 4 which satisfies conditions 1 and 2,
i . e . , a saddle-node, and the path Lo f r o m which the a - s e p a r a t r i x L; is cut
out f o r m s a loop, without merging with either of the o - s e p a r a t r i c e s L: and L:.
L e m m a . Let M o be a point of the path Lor and lo an arc without contact
through M o for which Mois not an end point. There exist E, > 0 and bo > 0
such that if(b) is 8,-close to (D), and L is a closed path of (6)contained in
u., (to),t mosses the arc I , at one and only one point.
P r o o f . It is readily s e e n that i f the proposition of the l e m m a is t r u e
f o r any sufficiently s m a l l a r c I , cut out f r o m s o m e fixed arc without contact
which c r o s s e s the path Lo, it is a l s o t r u e f o r any fixed a r c without contact
I,(see L e m m a 7, 8 4.2). By assumption, u (0.0) = P; (0,O) +
Q; (0,O) # O .
T o fix ideas, let
u(O,O)=uO>O.
Let e > 0 b e s o s m a l l that f o r any point ( x , y), (2, y) c U, (0), u (z, I/) > 0.
Consider the canonical neighborhood V and the e l e m e n t a r y quadrangle A of
(P) described in the previous subsection (in o u r proof of Theorem 51), and
a l s o the canonica1 neighborhood f and the quadrangle of t h e c l o s e s y s t e m
(fi), alongside with t h e i r union i?. Let V c U ,(0).A s Io we choose the a r c
without contact M i M z entering the boundary of V ( F i g u r e s 144 and 145).
Choose b o > 0 and^,> 0 so that i f (6)is bo-close t o ( D ) , the following
conditions a r e satisfied:
(a) I o i s an a r c without contact for (b);
(b) f o r any point ( x , Y)EW) ,
Z(z, Y) = R ( x , Y) +G (z,Y) >0;

326
5 ?O. CRKhTIOE; OF LlhliT CYCLE FROM SE?AR.\TRIX LOOP
(c) rcc',(0!(
Id) re0(Lo)cli'.
T h e s e c o n d i t i o n s a r e e v i d e n t l y s a t i s f i e d f o r s i i f f i c i e n t l y small e, a n d 60.
L e t E b e a c l o s e d pE.th of s y s t e m ( D ) w h i c h i s 6 0 - c I o s e t o ( D ) , a n d l e t
i c C'&* ( L o ) . T h e n , b y ( c ) , L c lT'.
If t h e p a t h is e n t i r e l y c o n t a i n e d i n r?, it is c o n t a i n e d i n U , (0). Ey
c o n d i t i o n ( b ) a n d F e n d i . s s a n ' s c r i t e r i o n (QT, 5 12.3, T h e o r e m 31, c o r o l l a r y ) ,
s v v t c r n (PI h a s no closed pat& w h i c h a r e e n t i r e l y c o n t a i n e d i n U, (0).
T h e r e f o r e , t h e c l o s e d [path L h a s p o i n t s w h i c h l i e i n t h e q u a d r a n g l e 3 .
T h e n , b y t h e p r o p e r t i e ; of e l e m e n t a r y q u a d r a n g l e s , t h e p a t h crosses
t h e a r c lo (at o n e a n d o n l y o n e point, b y c o n d i t i o n ( a ) ) . T h e a r c without
c o n t a c t In clearly c a n b e c h o s e n as small as d e s i r e d i n t h i s case.
T h p p r o o f of t h e 1 e r . m a is c o m p l e t e .
T h e o r e vi 52. Let 0 (0. 0 ) be a saddle-node of a dynamic system ( D )
for which cl0 = o (0. 0)+ 0 , and Lo a separutrix of this saddle-node foviriing
a loop.
Tnere exist ttco nuwbers e > Oand 6 > 0 such that i f ( f i i , is 6-close to ( D ) ,
it may have at most ont? closed path in U e ( L ~ ) . If this closed path exists, it
is a stable strrtctiirally stable limit cycle for 0, < 0 and an unstable struc-
turally stable limit cycle f o r (J>,0.
P r o o f . C o n s i d e r t h e case o o > O .
L e t E" > 0 b e so s m a l l t h a t f o r e v e r y point (I,y) E U q (0)w e h a v e
0 (I,y) = P ; (I.Y) iQ; (I.9) >4 . (3)
.,
L e t I , I o . i - , A , it', F, e c h e t h e a r c s w i t h o u t c o n t a c t , t h e c a n o n i c a l
n e i g h b o r h o o d s , t h e e l e m e n t a r y q u a d r a n g l e s , e t c ., c o n s i d e r e d i n t h e p r e -
v i o u s s u b s e c t i o n ( F i g u r t t s 114 a n d 1 4 5 ) . Let t h e c a n o n i c a l n e i g h b o r h o o d 1'
h e SO s m a l l t h a t
a n d l e t A b e t h e c o r r e s F o n d i n g e l e m e n t a r y q u a d r a n g l e of (D) ( F i g u r e 1 4 6 ) .
C h o o s e 6 , > 0 so s m a l l t h a t if (b) i s &-close to (D), t h e f o l l o w i n g c o n d i -
t i o n s are s a t i s f i e d :
( a ) The a r c s L a n d 1, ,ire a r c s without c o n t a c t for the p a t h s of s y s t e m (6).
( b ) T h e c a n o n i c a l n e i g h b o r h o o d F c o r r e s p o n d i n g t o (6)l i e s i n U 4 (LO)
t o g e t h e r with i t s c l o s u r e .
( c ) T h e p a t h s of (fi) p a s s i n g t h r o u g h t h e p o i n t s of t h e a r c I , cross t h e
a r c 1 w i t h t h e i n c r e a s e i n t , a n d t h e s e g m e n t s of_these p a t h s c o n f i n e d b e t w e e n
t h e arcs I, and 1 form an e l e m e n t a r y quadrangle A
( d ) For all (3, Y) E U,(O). z(z, Y) > 3 .
L e t iP be t h e u n i o n of t h e s e t s a n d 5 . W e c h o o s e e l , 0 < el < so
s m a l l t h a t for a n y set FF c o r r e s p o n d i n g t o a s y s t e m (B) 8,-close to ( D ) ,
we have
T h e e x i s t e n c e of t h e a p p r o p r i a t e 8 , a n d el i s s e l f - e v i d e n t .
L e t k h e a c l o s e d p a t h of s y s t e m (E) w h i c h i s 6 , - c l o s e to (D!: a n d l e t
c U,, ( L o ) . Then, b y (5), E c iv = FU 5 . We w i l l s h o w t h a t the p a t h
327
Ch.XII. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX
cannot be entirely contained in 8. Indeed, if L c v", then by ( 4 ) , c U , ( O ) .

T h i s is impossible, since relation (5) and B_endixson's c r i t e r i o n (QT, S 12.3,
T h e o r e m 31, corollary) show that s y s t e m (D) cannot have closed p%hs
which a r e entirely contained in U , (0). Thus, any closed path of (D) con- z
tained in U,,(Lo)h a s points which l i e in the quadr_angle Then, from the x.
p r o p e r t i e s of elementary quadrangles, the path L c r o s s e s the arcs_ 1 and lo,
and a l l t h e points of the path which lie outside h' a r e contained in V and
hence in U 5 ( 0 ) .
FIGURE 146
Let
x = Po ( t ) , Y = $0 (4 (6)
be a solution corresponding t o the path Lo of (D), Mo and No the intersection

points of Lo with the a r c s without contact lo and I , respectively, tt and t,,
t t < t 2 , the values of t corresponding t o points Mo and No for motion ( 6 )
( F i g u r e 146). Let, f u r t h e r ,
J= $'
11
(cpo ( t ) , $0 ( t ) ) +Q;(To (t), $0 (tNl d t . (7)
Let x be any number such that

IIl<X? (8)
and let C be any number such that

C > 8X
z .
(9)
Let So be a point of the path Lo corresponding to t = t , + C f o r motion ( 6 ) .
328
5 30. CRE4TIO?: OF LlhlLT CYCLE FRO\I SEPIR-\TRIX LOOP
Let 8 a n d E b e p o s i t i v e n u m b e r s s u c h t h a t 6 < 6 , , : < E , ; l e t (Fi) b e a

r l v n a m i c s y s t e m 6 - c l o . j e t o ( D ) , L a c l o s e d p a t h of ( D ) e n t i r e l y c o n t a i n e d
i n G e ( L a ) , .B a n d .q t h e i n t e r s e c t i o n p o i n t s of w i t h t h e a r c s lo a n d 1
( F i g u r e 116). Let
b e t h e m o t i o n along t h e path f o r w h i c h r h e point .Q c o r r e s p o n d s to i =ti.

Let a n d & b e t h e p c i n t s of t h e c l o s e d p a t h w h i c h c o r r e s p o n d for
m o t i o n ( I O ) to t = t , a n d t = f , + C , a n d l e t ? b e t h e p e r i o d of t h e s o l u t i o n
of ( 1 0 ) . If E is s u f f i c i e n t l y s m a l l , t h e point. .I2 w i l l be a s close a s
d e s i r e d t o Ma. H e n c e and from t h e t h e o r e m cf t h e c o n t i n u o u s d e p e n d e n c e
of t h e s o l u t i o n o n t h e i i i t i a l v a l u e s a n d t h e r i g h t - h a n d s i d e s i t f o l l o w s
t h a t if 6 a n d E a r e s u f f i c i e n t l v small, t h e f o l l o w i n g c o n d i t i o n s a r e s a t i s f i e d :
1) T h e point a n d all t h e p o i n t s of t h e p a t h 2 b e t w e e n & a n d so, are
Contained i n U&).
21 T h e a r c .\IS,) of t h e p a t h t c o r r e s p o n d i n g t o t h e v a l u e s of t , t , < t < t 2 - t C ,
i s s o close t o t h e a r c ill,&'a of t h e p a t h Lo c o r r e s p o n d i n g to t h e s a m e v a l u e s
o f 1 that
I
t t +c< ti A T (11)
( w e recall t h a t 7 is t h e p e r i o d of t h e p a t h E).
3)
tr
T h e l a s t e q u a l i t y is s a t i s f i e d for s u f f i c i e n t l y small 6 a n d E i n v i r t u e of ( 8 ) .
Let 8 and e b e c h o s e n so that conditions 1 t h r o u g h 3 are s a t i s f i e d . L e t
u s e s t i m a t e the number
'? -
By ( 1 2 ) , \ odt>-2~. Tt f o l l o w s from c o n d i t i o n 1 t h a t all t h e p o i n t s of t h e
i,
path 2 corresponding t o t2<t<t,+? a r e c o n t a i n e d i n CeO(O). T h e r e f o r e , b y
( 5 ) a n d (11).
H e n c e a n d f r o m i n e q u a l i t y ( 9 ) it f o l l o w s t h a t h ; 0, i . e., t h e c l o s e d p a t h
of (6) i s a n u n s t a b l e s t r u c t u r a l l y s t a b l e l i m i t c y c l e (913.3, Theorem 1 7 ,
214, T h e o r e m 18).
W e h a v e t h u s e s t a b l i s h e d t h a t a n y c l o s e d p a t h of s y s t e m (6)6 - c l o s e
to ( D ) w h i c h lies i n U ,(Lo) is an u n s t a b l e limit c y c l e . B u t t h e n , (6)c a n
h a v e a t m o s t o n e c l o s e d p a t h i n U, ( L , ) ( s e e t h e e n d of t h e p r o o f t o
T h e o r e m 47, 9 2 9 . 3 ) .
T h e case oo< O i s c o n s i d e r e d a l o n g t h e same l i n e s . T h e t h e o r e m
i s proved.
329
Ch.XI1. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX
931. DYNAMIC SYSTEMS O F THE FIRST DEGREE

O F STRUCTURAL INSTABILITY AND
THEIR FIFURCATIONS
1 . The definition of a s y s t e m of the first d e g r e e of

s t r u c t u r a l instability
In the previous c h a p t e r s and in s 30 we considered bifurcations of the

following types:
1 ) The decomposition of multiple equilibrium s t a t e s into s t r u c t u r a l l y
stable equilibrium s t a t e s .
2 ) The creation of l i m i t c y c l e s f r o m a multiple focus.
3) T h e creation of limit cycles f r o m a multiple limit cycle.
4) T h e creationof a l i m i t cycle f r o m t h e loop of a saddle-point s e p a r a t r i x .
5 ) The creation of a limit cycle f r o m the loop of a saddle-node s e p a r a t r i x .
In t h i s section w e will consider dynamic s y s t e m s of the f i r s t d e g r e e of
s t r u c t u r a l instability and elucidate the conditions satisfied by t h e s e s y s t e m s
and the bifurcations that they m a y undergo. Since s y s t e m s of the f i r s t
d e g r e e of s t r u c t u r a l instability a r e , in a s e n s e , the simplest s t r u c t u r a l l y
unstable s y s t e m s , their bifurcations may naturally be considered a s the
simplest bifurcations. Any one of the simplest bifurcations creating a
l i m i t cycle proves to be a bifurcation of one of the types 2 through 4 .
T h e definition of the d e g r e e s of s t r u c t u r a l instability of dynamic s y s t e m s ,
and in particular of the f i r s t degree of s t r u c t u r a l instability, will be found
in Chapter VI11 (s22, Definition 23). It is assumed in t h i s definition, how-
e v e r , that the s y s t e m i s considered in a region bounded by a cycle without
contact. W e therefore have t o define a s y s t e m of the f i r s t d e g r e e of s t r u c -
t u r a l instability in such a way that the definition will apply t o any bounded
closed region. The requirement of c l o s u r e is not essential: a s i m i l a r
definition c a n be stated f o r any bounded region.
Like the concept of a s t r u c t u r a l l y stable s y s t e m (see 96.3), the concept
of a s y s t e m of t h e f i r s t d e g r e e o f s t r u c t u r a l i n s t a b i l i t y
is associated with a c e r t a i n space R* of dynamic s y s t e m s . For s t r u c t u r a l l y
stable s y s t e m s , R*can be identified with any of the s p a c e s R: o r R f ) ,
k>r>l. W e have noted in 9 2 2 that the concept of a d y n a m i c s y s t e m
o f t h e f i r s t d e g r e e o f s t r u c t u r a l i n s t a b i l i t y is meaningful
only in relation to the s p a c e s R$)or R f ) , where r > 3 . T h e r e f o r e , in what
follows i t is invariably assumed, without any explicit indication, that we
a r e dealing with s t r u c t u r a l instability in relation t o one of the s p a c e s R:)
o r Rt, where r > 3. Let this space be Re.
A s in the c a s e of a structurally stable s y s t e m (96.1, Definition 10 o r
$6.3, Definition 13), in defining a s y s t e m of the f i r s t degree of s t r u c t u r a l
instability i n s o m e W , we have to consider, alongside with W , s o m e wider
region, Let G be the region u s e d to define the m e t r i c in the s p a c e R*.
Dynamic s y s t e m s a r e henceforth understood a s s y s t e m s which belong t o
the space R*, and c l o s e n e s s is interpreted a s c l o s e n e s s in R*. When
considering s o m e subregion of c, we will always a s s u m e that i t s c l o s u r e
is contained entirely in G I i . e ., i t s distance f r o m the boundary of G is finite.
Let
dx
-dr= P k u). %=QkY) (A)
be a dynamic s y s t e m , and W a closed subregion of G.
330
531. aisrtvs OF T H E FIRST DEGREE OF ~ T R L ~ C T L RI ~N LT ~ B I L I T I
De f i n i t i o n 30. .4dymzinic systein ( A )is called a system of the first

degree of structural instability in \t. if it is not striictirvally stable in It'
and if there exists an open region Jr,
K' c ff c B c G .
satisjying the folloicin,: condition: f o r any e > 1 1 , there exists 6 > 0 such that
for any systeni (A) 6-close to ( A ) , which is stnictirrally unstable in K, there
exists a subregion B f o r ichich
- - e
( H , .A) 3 ( H , -4).
T h e m e a n i n g of thiz s o m e w h a t c l u m s y d e f i n i t i o n c a n b e e l u c i d a t e d as
f o l l o w s : (.A) i s a s v s t e m of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t v i n f i . if
i t i s s t r u c t u r a l l y u n s t z b l e i n W , w h e r e a s a n y s u f f i c i e n t l v close s y s t e m (XI
i s e i t h e r s t r i t c t u r a l l y s t a b l e i n W , o r (41 and ( A , h a v e p a t h p a r t i t i o n s of
t h e s a m e ? o p o l o g i c a l s t r u c t u r e i n c e r t a i n n e i g h b o r h o o d s of Ii', a n d t h e
t r a n s f o r m a t i o n f r o m o n e p a r t i t i o n to a n o t h e r c a n b e i m p l e m e n t e d b y a n
at,bir r a r i l v small t r a n s l a t i o n .
W e n-ill now deriL-e t h e n e c e s s a r y c o n d i r i o n s s a t i s f i e d b y a n y s y s t e m
of chc f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y . T h e i r d e r i v a t i o n i s b a s i c a l l y
a n a l o g o u s t o t h e d e r i v a t i o n of t h e n e c e s s a r y c o n d i t i o n s of s t r u c t u r a l
s t a b i l i t y o f a s y s t e m : t h i s is a n a t u r a l a n d r e l a t i v e l y s i m p l e d e r i v a t i o n ,
but it r e q u i r e s a n e x a m i n a t i o n of n u m e r o u s a l t e r n a t i v e s .
T h r o u g h o u t t h e r e m a i n i n g p a r t of t h i s s e c t i o n , ( A ) i s a s y s t e m o f
t h e f i r a t d e g r e e o f s t r u c t u r a l i n s t a b i l i t y i n it- i n t h e
3 f ? n s e of P e f i n i t i o n 30.
2 . E q u i l i b r i u m states of systems of the first degree

of s t r u c t u r a l i n s t a b i l i t y
\$'e vcill f i r s t p r o v e a n u m b e r of t h e o r e m s w h i c h e s t a b l i s h t h e kind of

e q u i l i b r i u m s t a t e s t h a t s y s t e m s of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y
m a v have.
L e i n a 1. Let P fx. y) ana' Q ( x . y ) be functions of class .V, defined in a
closed bounded region t7, arui .IIo(x0.yo) any point in that region. F o r any
> ( 1 and ti i.Y (ti are nztiiral numbers), there exist polynoviials P ( x , y) and
B (I, y ) with the jolloiring properties:
a) and Q are 6-close to rank .V in 5 to the functions P and Q ,
respectively.
b) B ana' Q a.re irretlticible, i. e., (F,Q) = i .
c) The ualties of the polynomial P (I. y) (7 ( x . y))anrlall its deriuatives to
order n inclusive at thti point M, ( x o . yo) coincide with the corresponding ualries
of the function P (2. y) (0(x, y)) and its derivatives at the same point.
P r o o f . L e t x o = y, = 0; t h i s m a y b e a s s u m e d w i t h o u t loss of g e n e r a l i t y .
L e t 6; 0 b e g i v e n . F y t h e W e i e r s t r a s s t h e o r e m ( T h e o r e m 2 , 5 1.11, t h e r e
e x i s t p o l y n o m i a l s P* (z, y) and Q* ( x . y) w h i c h a r e & / 2 - c l o s e to r a n k N t o t h e
f u n c t i o n s P a n d Q , r e s p e c t i v e l y , a n d w h i c h s a t i s f y c o n d i t i o n ( c ) . If t h e
p o l y n o m i a l s P * a n d Q* a r e i r r e d u c i b l e , t h e y c a n b e a d o p t e d a s P a n d 0,
a n d t h e l e m m a is p r o v e d . S u p p o s e t h a t P* a n d Q* a r e r e d u c i b l e p o l y n o m i a l s .
33 I
Ch.X[I. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX I
Then they c a n be written in the f o r m
P*=Pi(%,~ ) . R ( IY),
, Q*=Qi(st Y).R(I, g),
where (Pi, Qi)= 1, and R is a polynomial of higher than z e r o d e g r e e .

If ( R , Q1) = 1, we may take P"(x, y ) = P * (I,y). Now suppose that R and Qi
a r e reducible. Let
Qi (5,Y) = 'pi (I,Y)*Tz(z,Y) .. . T&(G Y). $1 (2,Y) * $ z (2,Y) ... $I (I,Y) ( 1)
b e the factorization of Ql into irreducible f a c t o r s . It is assumed that the

following conditions a r e satisfied:
R ( x , y) is divisible by q j , j = l , 2, . .., I ; (2)
(the number k may b e z e r o , i . e., the polynomials 'pi need not o c c u r in

the factorization ( 1 ) ) . Let f u r t h e r a;c+py, a#O, p # O b e a polynomial of
the f i r s t d e g r e e which is irreducible with any of the polynomials 91 ( i
= 1,2, . . . ,L), and r is a natural number.
Suppose that
Ri(5, Y) = R (19 Y) + (a+

BY)' 'P~'PZ .. * 'P&,
and
p(G Y) = PiRi =P* (2,Y) 4-Pi (2,Y) (w+ (Pi% .. . (Pk.
F r o m ( l ) , (2), (3) it follows that ( R l , Q i ) = 1. Since ( P i , Ql) = 1, we have
(F,Ql) = 1. It is further evident that i f the numbers a and f~a r e sufficiently
small, and r is sufficiently large, the polynomials and P* a r e 6/2-close
t o rank N and coincide at the point ( 0 , 0 ) , together with t h e i r derivatives t o
o r d e r n , inclusive. We have thug constructed a polynomial P" which is
irreducible with Qi.
Now consider the polynomials and Q* = Q,R. If (P", R ) = 1, then
(P" Q * ) = 1 and we may take Q ( I ,y) = Q* (5,y ) . If, however, a and R a r e
reducible, using the s a m e construction a s f o r R1 (I,y), we c a n construct
a polynomial R z (x,y) which will be i r r e d u c i b l e with P" ( E , y) and such that
the polynomial
Q (5,Y) =Qi ( ~ Y) R2 (I,V)
3
is 6/2-close t o Q * ( I , y) t o rank N , and and Q* coincide a t the point (O,O),

together with their derivatives to o r d e r n , inclusive. The polynomials P
and constructed in t h i s way fulfill conditions ( a ) , (b), and (c) of the l e m m a .
T h e o r e m 53. A system of the f i r s t degree of structural instability in
W may have only a finite number of equilibrium states in this region, each
of which is furthermore isolated.
P r o o f . Since W is a closed region, i t suffices t o prove that e v e r y
equilibrium s t a t e of s y s t e m (A) in W is isolated. Suppose t h i s is not so,
and that M , (I". v n ) is an unisolated eauilibrium s t a t e i n W . Then. bv
332
531. SYSTEhIS OF THE FIRST DEGREE OF STRUCTURAL LKSTABLLITY
R e m a r k 3 to T h e o r e m 5 (52.2),
Since ( A ) is a s y s t e m of the f i r s t d e g r e e of s t r u c t u r a l instability in W ,

t h e r e e x i s t s a region R, T t ' c H , described in Definition 30. Let e > 0
be an a r b i t r a r y number. By Definition 30, t h e r e e x i s t s 6 > 0 such that
if s y s t e m(x) is 8 - c l o s e to ( A ) , then
e i t h e r ( a ) ( A ) is s t r u c t u r a l l y stable in W ,
or ( b ) ( 3 , is s t r u c t u r a l l y unstable, and
where is s o m e region.
Consider a dynamic s y s t e m
8 - c l o s e t o ( A ) , where the right-hand s i d e s a r e i r r e d u c i b l e polynomials

which coincide, together with t h e i r f i r s t derivatives, with the functions
P and Q and t h e i r f i r s t derivatives a t the point M0(TO,YO) (such a _system
e x i s t s by L e m m a 1 ) . AT, (zo, yo) is then an equilibrium s t a t e of ( A ) , and
Zi (ZO. YO) = A (z, Y) = 0. (6)
Since (A) is 6 - c l o s e to ( A ) , one of the conditions ( a ) or (b) should_apply to

t h i s s y s t e m . However, condition (a) is ruled out, s i n c e by (6), (A) h a s a
s t r u c t u r a l l y unstable equilibrium s t a t e M o in W . Condition ( b ) is a l s o
unacceptable, since (A) h a s in tt' a n infinite number of equilibrium s t a t e s ,
whereas (A) h a s a finite number of equilibrium s t a t e s in the plane (since
(F,Q', = 1) and relation (5) is t h e r e f o r e inapplicable. T h i s contradiction
proves the t h e o r e m .
.
L e wt ni a 2. Let O,, 02, . ., 0 , be all the equilibriicrn states of (A) in
W'. There exist c0 > 0 and 8o > o with the foilacing property: if (A) i s
6,-close to ( A ) and is st,ructurally unstable in W , the ~,-neighbmhoOdof
each point O i ,i = 1, 2, . . S , contains precisely one equilibrium state
.?
Ciof (A1an.d the equilibrium states Oi and haue the same topological
structure. The number eo > 0 may
be taken as small as desired.
P roo f . Let H be a region s a t i s -
fying t h e condition of Definition 30,
', H 3 W ( F i g u r e 1 4 7 ) . Without l o s s
of generality, w e may a s s u m e that
Hcontains no o t h e r equilibrium s t a t e s
1
I
of ( A ) , except O,,Ox,. . ., 0, ( t h i s is
\ so if H is a sufficiently s m a l l neighbor-
hood of the closed region Wj. e , > 0
is chosen a s a number satisfying the
following conditions:
(a) T h e neighborhoods U,, ioi), i =
= 1,2, .. . , s, a r e contained in H and no
FIGifRE 147 two of t h e s e neighborhoods i n t e r s e c t .
333
Ch.XII. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE 8EPAKATRIX
( b ) T h e distance of each neighborhood U q , ( O i )f r o m the boundary of H is

g r e a t e r than 2 eo ( F i g u r e 1 4 7 ) .
By Definition 30, t o t h i s E , corresponds a,,> 0 such that if (A) is &,-close
t o (A) and is s t r u c t u r a l l y unstable, we have
z)2 ( H ,
(8, A), (7)
whet-e f7 is s o m e region.
It follows f r o m ( 7 ) that contains precisely s equilibrium s t a t e s of (A),
-
which we designate 6,, 62, .. ., 0,.
Let f be the mapping of onto H implementing the relation (7), (i. e.,
f is an eo-translation moving d into H and conserving the paths, see 84.1,
Definitions 8 and 9). Let, further, f (ai) = O t , i = 1, 2, . . ,s. Then .
6i c u,, (Oi). (8)
By condition ( b ) , each of the neighborhoods U , , ( O j ) is contained i n H , i . e . ,
u,, (Oi) c J7 (9)
(see footnote to p. 67): It follows f r o m ( 8 ) and (9) that ai is the only

equilibrium s t a t e of ( A ) contained in U e o ( O i ) . Since f (5')= Ot and f is a
path- conserving topological mapping, the equilibrium s t a t e s Oi and Gi have
identical topological s t r u c t u r e s . The proof of the l e m m a is complete.
L e m m a 3 . Let 0 be an equilibrium state of (A), 0 c W. There
exists 6, > 0 such that if (A)i s structurally unstable, &,-close to (A),
and has an equilibrium state at the point 0 , the topological structure
of the equilibrium state 0 of (A) coincides with the topological struc-
ture of the equilibrium state 0 of (A).
Lemma 3 follows directly from Lemma 2 . 6, can be chosen a s the
6, of L e m m a 2 .
T h e o r e m 5 4 . A system of the f i r s t degree of structural instability
in IZ' has no equilibrium states with A = 0, u = 0 in this region.
P r o o f . Let (A) be a s y s t e m of the f i r s t degree of s t r u c t u r a l instability.
Suppose that i t h a s an equilibrium s t a t e 0 (0, 0 ) in W for which A = 0, u -= 0 .
Let 6,>0 b e the number defined by L e m m a 3 . Ey the W e i e r s t r a s s
theorem, there exists a system
I
ed t = P (z, y), %=-Q(z, y) (A)

6,/2-close to ( A ) , where P and 0" a r e polynomials, and the values of and 0
and t h e i r f i r s t derivatives at the point O ( 0 , O ) coincide with the values of
the functions P and Q and t h z i r f i r s t derivatives at the s a m e point. 0 ( 0 , O )
is an equilibrium s t a t e of ( A ) f o r which
-A ( 0 , O ) = A ( O , O)=O,
I
~ ( 0 O)=U(O,
, O)=O. (10)
At l e a s t one of the polynomials P , 0 h a s l i n e a r t e r m s (otherwise, we

replace with h y+-3 , where h is a sufficiently s m a l l number; the point
0 r e m a i n s an equilibrium state with A = u = 0 ) . Since (A) is a s y s t e m of
the f i r s t degree of s t r u c t u r a l instability, and (A) is a0/2-close t o ( A ) and
334
4 91. SYSTEL!:; CF THE FIRST DEGREE OF STRLfCTL'RAL INSTABILITY
h a s an equilibrium s t a t e at point 0 , (A) i s a s t r u c t u r a l l y unstable s y s t e m

with an isolated equilibrium s t a t e at 0 .
Let u s now apply the r e s u l t s of 123. In virtue of t h e s e r e s u l t s , (x)
may
be arrittcn in the form
where h , g, f a r e a n a l r t i c a l functions, h ( 0 ) - g ( 0 ) - 0 , r>2. a#O, and n > l

if 6 2 0 ( s e e 923.2, ( 2 5 ) ) .
where 2C 0 and i s so s m a l l in absolute magnitude that (A,) and (A) a r e a 0 / 2 -

close.
F i r s t c o n s i d e r t h e c a s e r > 2 . The point 0 i s then an equilibrium s t a t e
of (-XI) f o r wkich A 0 . By 523.2, 11, p. 226, i f r is odd (even), the point 0
for s v s t e m ( A ) (for s y s t e m (A,)) i s e i t h e r a saddle-point, or a node, o r a
focus, o r a c e n t e r , or an equilibrium s t a t e with an elliptical region, and
f o r s v s t e m ( 3 , )( f o r s,ystem-(.x)) it is e i t h e r a degenerate equilibrium state
o r a saddle-node. Thus, ( A ) and ( A , ) are s t r u c t u r a l l y unstable s y s t e m s
5,cIose to ( A ) , and the topological s t r u c t u r e of the equilibrium s t a t e 0 of
t.4) is different f r o m the topological s t r u c t u r e of the equilibrium s t a t e 0 of
(-XI). T h i s c o n t r a d i c t s L e m m a 3 .
Let now r = 2 . In this c a s e (XI) h a s the f o r m
-
dr
dt
- y. * dt
= >x2. ax2 + ...
11 Th(.r)l
The point O ( 0 , O ) is an equilibrium s t a t e of t h i s s v s t e m with A = - h > 0

and u 1 0 , i . e . , a multil>le focus o r a c e n t e r . T h i s equilibrium s t a t e
is s t r u c t u r a l l y unstable, and (Al,) i s t h e r e f o r e a s t r u c t u r a l l y unstable
s y s t e m . With r e g a r d to (x),t h e point 0 i s a degenerate equilibrium
s t a t e for r = 2 ( s e e QT, $ 2 2 . 2 , T h e o r e m 6 i ) . T h i s l e a d s to the s a m e
contradiction with L e m m a 3 a s for r > 2. T h i s completes the proof of
the t h e o r e m .
It follows f r o m T h e o r e m 5 4 that e v e r y equilibrium s t a t e of s y s t e m
( A ) in W is e i t h e r a s i m p l e equilibrium s t a t e or an isolated equilibrium
sTate with A = 0 and u 0 . +
-4s we know, a s i m p l e equilibrium s t a t e h a s a multiplicity of 1 (s7.3,
Definition 15, S2.2, T h e 3 r e m 6 ) . Let u s find the multiplicity of a
multiple equilibrium staye of s y s t e m ( A ) .
T h e o r e in 55. If (A) is a s y s t e m of the first degree of instability
in It', any multiple equilibriririi s t a t e in FV has a Pnultiplicity of 2.
P r o o f . Let O ( 0 , O ) b e a multiple equilibrium s t a t e of ( A ) , 0 E W . By
t h e previous t h e o r e m A ( 0 , O ) = 0, o ( 0 , O ) #O. Without l o s s of g e n e r a l i t y
M'B may take ( A ) in t h e f o r m
where
335
( s y s t e m (A) c a n be reduced t o t h i s f o r m by a non-singular l i n e a r t r a n s -

formation, s e e QT, $ 2 1 . 2 ) .
I According to the t h e o r e m of implicit functions, the equation
(13)
h a s a unique solution f o r y in the neighborhood of O ( 0 , 0). Let t h i s solution

be cp(r). Then
+ = 0,
P (4 Q (I,'p (4)
and
cp(O)=O.
It follows f r o m (IZ), (14), and (15) that

93' (0) = 0.
Consider the function
9 (4= P (I, cp (4).
I
Direct computations show that
9(0)=0, V(O)=O, $"(O)=p:,(O, 0).
F r o m Definition 15 ($ 7.3) and T h e o r e m 7 ($2.3) it follows that the

equilibrium s t a t e U ( 0 , 0) of s y s t e m (11) h a s a multiplicity of 2 i f and only
i f pkX(0, 0) # 0. We will now show that t h i s condition is satisfied. Suppose
that
P i x (0, 0) =o. (19)
Let e,, and a,, be the positive numbers introduced in L e m m a 2 . A s i n

o u r proof t o T h e o r e m 54, we u s e the W e i e r s t r a s s t h e o r e m and construct
a system
6,/2-cIose t o ( I l ) , where 5 and 5 a r e polynomials which together with

t h e i r derivatives t o second o r d e r inclusive coincide at the point O(0,O)
with the functions p and q and t h e i r derivatives. Then
pc(0, O)=&(O, O)=Z(O, O)=?(O, O)=q;(O,

-
0) =&(0, O)=O (20)
and
gx(o,0) =o. ( 2 1)
Since h" ( 0 , O ) = 0, (A) is s t r u c t u r a l l y unstable in W . Therefore, by

L e m m a 2, the equilibrium s t a t e U ( 0 . 0 ) of (A) is isolated.
Let G(z) and $(z) b e the analogs of the functions 93 a n d $ , i.e., t h e s e
a r e functions defined by the relations
336
6 :?l. YYSTE.\!S OF T H E FIRST DEGREE OF STRLrCTC'R.AL INSTIBILITY
and
c
$(I
= ); ( I , Cp(4). (231
S i n c e and 5 are a n a l y t i c a l functions ( p o l y n o m i a l s ) , @ and 5 are also

a n a l v t i c a l f u n c t i o n s . B:J ( 2 0 ) a n d (211,
(0) =T (0) = o (24)

and
*
,.
(0)= 5' (0)= i j " ( 0 ) = u.
ILIoreover, $ ( x ) c a n n o t b e i d e n t i c a l l y z e r o , I n d e e d , i f $(x) = 0, t h e n f o r a l l
s u f f i c i e n t l y s m a l l x , t h e point (5. k ( x ) ) i s a n e q u i l i b r i u m s t a t e of ( A ) , so
t h a t 0 is n o l o n g e r a n i s o l a t e d e q u i l i b r i u m s t a t e . T h u s , n e a r t h e point
x = 0, t h e f u n c t i o n $(z) is e x p r e s s i b l e in t h e l o r m
-' p ( x ) - ~ ~ + & * z " - ' i ..., (26)
, & h e r e a d t o a n d n\,3.
C o n s i d e r t h e dynamic: s y s t e m
\<;herepf.0 is so small t h a t the s y s t e m s (A,\ a n d (-3)are d 0 / 2 - c l o s e .

S i n c e 0 , 0 , 0 ) is a m u l t i p l e e q u i l i b r i u m s t a t e of (&), t h i s s y s t e m is
structurally unstable in W .
L e t u s t r y to find m m e e q u i l i b r i u m s t a t e s of (A,) n e a r t h e point 0 .
R y (221, y = + ( x ) i s t h e s o l u t i o n of t h e e q u a t i o n Q ( x , y) = 0 for y . I n s e r t i n g
(F (x) for y i n t h e e q u a t i o n p,,(x. y) = 0, w e o b t a i n a n e q u a t i o n f o r 2, i . e.,
F p ( x , $(.I) = 0 , o r e x p l i c i t l y
or
w h e r e h (p, x) = p -,-cuc+a,z'+ . . . .
S i n c e h (0, 0)= 0, h; (0, 0) = a + O., the t h e o r e m of i m p l i c i t f u n c t i o n s
s h o w s t h a t t h e e q u a t i o n ,h ( p , 3) = 0 h a s a s o l u t i o n x = x, n e a r t h e point
O!O, 0 ) w h i c h goes t o z e r o for p -+ 0. C l e a r l y , if p P O , t h e n I,, # O .
T h e point 0, (I,,. ' p ( x , ) ) is a n e q u i l i b r i u m s t a t e of t h e s y s t e m(x,) which
i s d i f f e r e n t f r o m 0 2nd lies a s close a s d e s i r e d to 0 , w h e n p is s u f f i c i e n t l y
small.
We h a v e t h u s e s t a b l i s h e d t h a t i f p:. ( 0 , O ) = 0, t h e r e e x i s t s a svstem (A,)
&-close t o ( A ) w h i c h is s t r u c t u r a l l y u n s t a b l e in W' a n d y e t h a s tWG e q u i l i b r i u m
s t a t e s i n a n y a r b i t r a r i l y s m a l l n e i g h b o r h o o d of 0 , n a m e l y 0 a n d 0,. T h i s
c o n t r a d i c t s L e m m a 2. 'Thus, p:x (0, 0) # 0, i . e . , t h e e q u i l i b r i u m s t a t e 0 h a s
a m u l t i p l i c i t y of 2 . Q. E . D.
T h e o r e r)t 56. I f ( A ) is a system of the fi.rst degree of structural
iristability in i t ' , any ntriltiple equilibrium state of this system in itr is
a saddle-nede.
337
I
P roof . Theorem 56 follows directly f r o m T h e o r e m s 54 and 55 because
of the following g e n e r a l proposition: a m u l t i p l e e q u i l i b r i u m s t a t e
of m u l t i p l i c i t y 2 f o r w h i c h u # O i s a s a d d l e - n o d e . If(A)
is an analytical s y s t e m , t h i s proposition is contained in I, $23.2 ( o r ,
equivalently, in QT, $ 2 1 . 2 , Theorem 65). For a non-analytical s y s t e m ,
the proof can be obtained by the s a m e method a s the proof of Theorem 65
in Q T .
T h e o r e m s 54, 55, and 56 show that a n y m u l t i p l e e q u i l i b r i u m
s t a t e of a s y s t e m of t h e f i r s t d e g r e e of s t r u c t u r a l
instability is a saddle-node with u # O and a multiplicity
o f 2.
Let u s now consider simple equilibrium s t a t e s , i . e., those with A # 0.
We have s e e n (Chapter IV) that a simple equilibrium s t a t e is stable, except
i f i t s c h a r a c t e r i s t i c numbers a r e pure imaginary. It is readily s e e n
that a s y s t e m of the f i r s t d e g r e e of s t r u c t u r a l instability may have a
s t r u c t u r a l l y stable equilibrium state of any type. W e w i l l t h e r e f o r e con-
c e n t r a t e on equilibrium s t a t e s with p u r e imaginary c h a r a c t e r i s t i c n u m b e r s .
Each of t h e s e equilibrium s t a t e s , a s w e know, is e i t h e r a multiple focus,
o r a center, or a center-focus ( a center-focus i s possible only f o r a non-
analytical s y s t e m , r e m a r k at the end of $24.4).
Without l o s s of generality, we a s s u m e that the equilibrium s t a t e is the
point O ( 0 , 0) and that (A) is given in the canonical f o r m
dx
x= --BY+cp(Z, Y), ~ = B 2 + W tY)? (28)
where @ > 0 , and cp and 21, a r e functions of c l a s s 3, which vanisp together

with t h e i r f i r s t o r d e r derivatives a t the point 0.
In Chapter I X we examined the succession function f (p) defined on a r a y
extending f r o m the point 0 , and the function d (p) = f (p) - p . Both t h e s e
functions are of the s a m e c l a s s a s cp a n d $ . The values of the derivatives
of d (p) at the point 0 a r e known a s the focal values. We have s e e n ( $ 2 4 . 2 ,
L e m m a 5 and (25)) that
d (0) = d' (0) = d" (0) = 0. (25)
If d" (0) # 0, O ( 0 , O ) is a multiple focus of multiplicity 1 (see 5 2 4 . 2 ,

Definition 2 6 ) . If d " ( 0 ) = 0, O i s e i t h e r a multiple focus of multiplicity
m > 1, or a multiple focus of indefinite multiplicity, o r a center, o r a
c e n t e r - focus.
T h e o r e m 5 7 . If a system of the f i r s t degree of structural instability
in W has equilibrium states with pure imaginary characteristic numbers
in this region, it is a multiple focus of multiplicity 1 , i . e . ,
this point.
Proof . Suppose that the theorem is not true, i . e., f o r the point
O ( 0 , O ) of s y s t e m (28) we have
d" (0)= 0. (30)
A s we know ($24.3, (35), (36)), s y s t e m (28) may be written in the f o r m

3
dx
=p ( 2 9 Y) = -BY + p , (2. Y) +pt (2,Y) + 3 53-allapz ( x ,
a=O
3 (31)
2 =Q (2,Y) = BS +Qz (2,Y) +0 s (2,Y) + 3 S~-~YOLQZ
a=O
(2,Y)
54.38 338
$ 3 1 . SYSTEhlS OF THE FIRST DEGREE OF S l R U C T C R A L IKSTABILITY
where P2 and Qz ( P a and Q 3 ) a r e homogeneous polynomials of d e g r e e 2 (of

d e g r e e 3), and P:(z, y) and Q: (z, Y) a r e continuous functions which vanish
for z = y = O .
Let do> 0 be the number defined by L e m m a 3 . Let further Fz and 0; be
polynomials which provide an adequate fit of the functions P; and QI:and
vanish at the point O ( 0 , 0 ) . Then the s y s t e m
3
dz
= -BY + P2 (G Y) fP3 (z,Y) 4-
G O
+3-a?/@ (I,Y) = P(z, ?A,
3 (A)
dY
z = P +Qz (17 Y) + 0 3 (5, Y) + 2 z3-aPQz(z, Y) =Q(zf
a=o
Y)
is 6,/2-close to ( A ) and the point 0 i s i t s equilibrium s t a t e with p u r e
imaginary c h a r a c t e r i s t i c numbers. Let J(p) be the analog of d ( p ) for
the s y s t e m (A). It follows from equality ( 3 0 ) and f r o m L e m m a 6, 9 2 4 . 3 that
2 ( 0 ) =o. (32)
Since (A) is an analytical s y s t e m , the l a s t equality shows that O ( 0 , O ) is

e i t h e r a c e n t e r or a multiple focus of multiplicity p > 1 for t h i s s y s t e m .
W e will now show that i t cannot be a cente_r.
Indeed, suppose that 0 is a c e n t e r of ( A ) . Consider the s y s t e m
dz
-_
dt -F(z, Y)iPP+Y9$/, %=W,Y ) + p ( ~ - F i ~ ) G (A,)
where p f O , with the corresponding function d;(p). For a sufficiently
s m a l l p , s y s t e m (A,) is 80-close to s y s t e m ( A ) (i.e., t o s y s t e m ( 2 8 ) ) .
By L e m m a 7 , 9 2 4 . 3 , we see that
2; (I)) = d"; (0)= 0, 2;,*(0)= i2x P # 0,
i.e., 0 is a multiple focus of s y s t e m (XI). Thus, (A) and (-TI)

are struc-
turally unstable s y s t e m s 8,-close to ( A ) , and the equilibrium s t a t e 0 is a
focus f o r (A,) and a c e n t i r for ( A ) . T h i s c l e a r l y contradicts L e m m a 3 .
W e have thus established that 0 is a multiple focus of ( A ) .
Then, by L e m m a 3, 0 is also a multiple focus of the original s y s t e m
(A). We will show that t h i s conclusion involves a contradiction.
Indeed, l e t 0 be a multiple focus of ( A ) . Then, if q > 0 i s sufficiently
s m a l l , all t h e paths of s y s t e m (A) contained in U , (0)(with the exception
of the point 0 ) are s p i r a l s which wind onto 0 for i + - 00 or f o r t + f 00.
Choose q > 0 which s a t i s f i e s t h i s condition.
Let e b e a positive nurnber, E < 2, 8, 0<8<&, a number corresponding
t o e in virtue of Definition 3 0 . Then, i f s o m e s y s t e m (8)is structurally
unstable in it' and is 8 - c b s e t o (A), we have
(H,A ) & ( a , , E),

where 8, is s o m e region, and isIthe region introduced i n Definition 3 0 .
We m a y take q > 0 s o s m a l l that U , ( O ) c R . Then, f r o m the above relation
A) 2 (7%
(U,(0), B), (33)
where is s o m e region.
339
Ch.XK. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX
F u r t h e r suppose that 0 is an equilibrium s t a t e of (6): W e will now

t r y t o extract s o m e information about the neighborhood V .
By ( 3 3 ) , all the paths of (e)
contained in 7 a r e s p i r a l s winding onto
the point 0 (the point 0 itself excepted). It follows f r o m the inequality
E -= $that U,,, (0)c 7 . Thus, i f (B) is a s t r u c t u r a l l y ucstable s y s t e m
Let the s y s t e m (A) constructed above b e 6/2-clos_e to ( A ) . Since

6<6,,, 0 i s a multiple focus of multiplicity m > 1 of ( A ) , a s we have s e e n .
Let (8)be the s y s t e m
where p # O is so s m a l l that (3,)is 6/2-close t o (A). In o u r proof of

T h e o r e m 40 we have s e e n that if q is taken sufficiently s m a l l and of an
appropriate sign, (xi,) will have at l e a s t one closed path in U,,,,@ ) ( s e e
$ 2 5 . 1 , proof of Theorem 40, i n p a r t i c u l a r equations (3)- (9); f o r our
purposes k = 2 ) . T h i s contradicts the previous proposition that a l l the
paths of (A,) in U,jZ (0)are s p i r a l s . Q. E. D.
T h e o r e m s 5 3 - 5 7 provide a complete solution to the problem of
equilibrium s t a t e s of a s y s t e m of the f i r s t d e g r e e of s t r u c t u r a l instability.
T h e y show, i n particular, that if ( A ) is a dynamic s y s t e m of the f i r s t
d e g r e e of s t r u c t u r a l instability in W , i t may have only a finite number
of equilibrium s t a t e s in t h i s region, and each of these equilibrium s t a t e s
is e i t h e r a s t r u c t u r a l l y stable equilibrium s t a t e , o r a multiple focus of
multiplicity 1, o r , finally, a saddle-node with u # O and a multiplicityof 2 .
3 . Closed paths of s y s t e m s of the first d e g r e e

of s t r u c t u r a l instability
Closed paths of dynamic s y s t e m s w e r e considered in Chapters V and X.

In the present subsection, w e will u s e the e a r l i e r notation and some of
the previous r e s u l t s .
F i r s t , w e will prove a number of l e m m a s relating t o the succession
I function on a n o r m a l t o a closed path.
be a dynamic s y s t e m of c l a s s N o r an analytical system, Lo a closed

path of t h i s s y s t e m ,
x=TJ(t), !4=9(4 (LO)
the motion c o r r e s p o n d i n g t o t h i s path, T > 0 the period of the functions cp and 9 .
In 13.1, we introduced curvilinear coordinates s, n in the neighborhood
of the path L o , defined by the relations
- -
5 ='p (s, n), Y = + (s, n ) ,
$31. SYSTEXIS OF Tb!E FRST DEGREE OF STRC'CTC'RIL [E;ST.AEILITY
w h e re -
~ ( s n, ) = = q(s)+-n.*' (s), $(s. n)=$(s)-nncp'(s). (34)
S y s t e m ( A ) is d e s c r i b e d i n t h e s e c o o r d i n a t e s by t h e d i f f e r e n t i a l e q u a t i o n
dn
-&
dS
R (s, n ) , (35)
w h e r e R ( s . n ) i s a f u n c t i o n of class S ( o r an analbytical f u n c t i o n ) * h i c h is
p e r i o d i c i n s with t h e period T and s a t i s f i e s t h e i n i t i a l condition
R ( s , 0 ) = 11.
Let. n = f (5, no) b e t h e s o l u t i o n of e q u a t i o n ( 3 5 ) s a t i s f y i n g the i n i t i a l

c o n d i t i o n f (0, n,))= n o .
Then
( n J = f (T. no)
i s a s u c c e s s i o n f u n c t i o n o n t h e a r c without c o n t a c t I d e f i n e d b y t h e e q u a t i o n s
I = :q ( 0 )+n$' (0). y = $ (O)-nq' ((I),
w h i c h is n o r m a l to t h e p a t h L, at t h e point s = 0 . Alongside x i t h t h e
s u c c e s s i o n function, we c o n s i d e r e d t h e function
d ( n o )= f ( n o ) - no.
f ( n o )a n d d ( n o )a r e d e f i n e d for all s u f f i c i e n t l y small ( i n a b s o l u t e m a g n i t u d e )

v a l u e s of no, s a y for I no 1 < n,f, and t h e y a r e f u n c t i o n s of t h e same c l a s s a s
t h e o r i g i n a l s y s t e m ( A ) (see 9 1 2 . 2 , e q u a t i o n ( 5 ) a n d t h e t e x t t h a t f o l l o w s ) .
L e t u s s t a t e t h r e e l e m m a s ( L e m m a s 4,5, 6 ) r e g a r d i n g t h e s u c c e s s i o n
f u n c t i o n s of close s y s t e m s . T h e s e l e m m a s follow d i r e c t l y from g e n e r a l
c o n s i d e r a t i o n s ( t h e c o n t i n u o u s d e p e n d e n c e of t h e s o l u t i o n s of d i f f e r e n t i a l
e q u a t i o n s o n t h e r i g h t - h a n d s i d e s a n d t h e d i f f e r e n t i a b i l i r y of s o l u t i o n s
with r e s p e c t t o p a r a m e t e r s and i n i t i a l v a l u e s ) , a n d w e swill o m i t t h e
r c s p e c t i v e p r o o f s . It is a s s u m e d t h a t t h e m o d i f i e d s y s t e m s
-
Td r= P ( x . Y). ===
dg O(Z. y)
b e l o n g to t h e s a m e class as s y s t e m (-A), and \ y e c o n s i d e r c l o s e n e s s t o

r a n k k , w h e r e 1 .:k<..V ( i f ( A 1 i s a n a n a l y t i c a l s y s t e m , k m a y b e a n y
natural number).
L e ~i m a 4 . For any e > 0 , there exists 6 > 0 rcilh the follotcirzg
property: i f (A) is 6-close to ( A )to rank k , thepi
(a, the nortrial I is an arc ivitfzoiit contact of (A) also, and
a siiccession
,-.
ficnction T(rt,) of systein (A) and t h i s also the function 2 (no)=
- f (nolavedefined on this arc fmall no. I no I < n,l ;
(b) the functions f ( n o ) and f ( n , ) and thus also the functions d ( n o )and
d ( n o )aye 6 -close to rank X. .
In \$.hat f o l l o w s , w e w i l l u s e m o d i f i e d d y n a m i c s y s t e m s of a s p e c i a l f o r m ,
namely
34 1
Ch. XU. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX
where pi a r e p a r a m e t e r s , and p t (I,y) and qt ( x , y) are functions of one c l a s s

with P and Q .
L e m m a 5. There exists p* > 0 such that if Ipl I <p* the normal 1 is an
arc without contact for &,,) and a succession function fc(no, pt, pa, . ., pn) of .
the same class as P and Q is defined m this arc fav all no, Inol<n:.
Moreover,
-f ( n o ,0, 0 , ..., O ) = f ( n o ) .
where P*, Q*, p : , q: a r e functions of the s a m e c l a s s a s P and Q. Let p* be

a number satisfying the conditions of L e m m a 5.
L e m m a 6 . For any e >0, there exists 6 >0 with the following property:
i f [ p , ( < p * and the functions P*, Q*, p t , qf are 6-close to rank k to the
functions P, Q , p i , qi , respectively, the normal I is an arc without contact
for the system (A;, ,) and a succession function f* (no,pi, p a , . ., pn) is defined .
on this arc for all no,I no1 <n:. The function f* is of the same class as P
and Q in all its variables and it is e -close to rank k to 7 .
The next l e m m a is m o r e r e s t r i c t e d . It d e a l s with the s y s t e m
dz
Z = ~ ( G Y ) + ~ P ( Z ,Y ) = F (Y,
~ p,), $ = Q ( G Y ) + P P ( ~ Y)=Q"@,
, Y). (A) I
Let 7 ( n o , p) be the succession function f o r t h i s s y s t e m on t h e n o r m a l 1 .
L e m m a 7. For the succession function ?(no, p) of system (A) on the
normal I , we have the equality
- 1
e
i
W'L+Q;)& 5 -6: (P;tQ;)&
f ;(O,O)= (#@(()))a+($' (0))s
0
tP+'(s)-F+-J'(s)l e ds, (36)
where the values of the functionsp;, Q;, p , and q are evaluated at the
+
Point (cp (s), (s)).
P r o o f . In Chapter V we rewrote the differential equation (35) c o r -
responding t o s y s t e m (A) i n the form
where (p and $ a r e defined by ( 3 4 ) (see s 13.2, ( 1 2 ) ) . The corresponding

equation for (A) is
where the functions P, Q, p , q a r e evaluated at the point (G(s, n), $(s, n)).
Let n = F ( s ; no, p) be the solution of equation (37) satisfying the initial
condition 7 (0, no, p) s no. Then identically
342
I ?I. SYSTEhI3 O F T t I E F R S T DEGREE OF S T R C C I ' I P ! INST-\HIL[I'Y
and t h e s u c c e s s i o n f u n c t i o n of ( 3 )on t h e n o r m a l 1 is d e f i n e d bv t h e r e l a t i o n
( t h e a s s i g n m e n t of rhe sixme svnibol 7 to TLVO d i f f e r e n t f u n c t i o n s s h o u l d not

l-arl to c o n f u s i o n , s i n c e o n e of t h e m is a fr3nction oi twn a t y J m e n t s , and
rhe n t h e r is a f u n c t i o n of' t h r e e a r g l l m e n t s i ,
r.pt
tJ (s)- 7;(s. 0 . 0). (401
r l i i f e r e n t i a t i n g ( 3 8 ) with r e s p e c t to p, c h a n g i n g t h e o r d e r or' clifferentiation

i n t h e l e f t - h a n d s i d e , s e t t i n g n,,=p= 0 a n d [using ( 4 0 \ , we o b t a i n
3-13
(:h. XII. CREATION OF LIMIT CYCLE PROM L O O P OF SAI)IILE-NOI>t bEI'ARATRIX
-f;
i . e . , if t h e v e c t o r f i e l d of t h e s y s t e m is a r o t a t e d f i e l d of ( A ) , t h e n
(0, 0 ) f 0 .
P r o o f . T h e v a l i d i t y of t h e l a s t i n e q u a l i t y f o l l o w s d i r e c t l y f r o m (36)
s i n c e i n t h e p r e s e n t case
satisfies the eqiiuiities

Z(0) = 0, 3(O)=e~r-f,
where
I -= s
T
0
[(F;((P(s), rl, (s)))* + (F1('P (s), $ (s)))'l ds,
i. e., f o o v p # ~ , L~ is (i sirrrple limit cycle 0;. the system (A;).

Proof .
'l'hc firs! p r o p o s i t i o n of t h e l e m m a i s c o n t a i n e d i n 926.2,
L e m m a 1, a n d t h e s e c o n d p r o p o s i t i o n w a s e s t a b l i s h e d i n c i d v n r a l l y i n
o u r p r o o f of T h e o r e m 19 ( 3 15.2 ( 1 8 ) ) . T h e n u m b e r I3 i n the l a s t q u a l i t y
i n ( 4 5 ) is a c o n s t a n t d e p e n d i n g on t h e f u n c t i o n s P , Q,F , rp, and 9
F e m a r k . If (-4) i s a s y s t e m of c l a s s N a n d for i t s c l o s e d p a t h La
d'(0) = d (0)= . . .= d'"'-l'(o) = 0,

w h e r e l < m < N , t h e n f o r all s u f f i c i e n t l y s m a l l p f 0 , t h e p a t h Lo is an
m - t u p l e l i m i t c y c l e of (A;). I n d e e d , i f d'""(0) = 0, oui' p r o p o s i t i o n f o l l o w s
f r o m L e m m a 8. If, h o w e v e r , d("(O)+ 0 , t h e n f o r * s u f f i c i e n t l y s m a l l p ,
t h e d e r i v a t i v e & ) ( O ) d o e s not v a n i s h e i t h e r , a n d t h e v a l u e s of z(O!,
&O), . . .( Zm-l)(o)a r e r e s p e c t i v e l y e q u a l to ~ T ( o )~, Y ( o ) ., . .. ~ - 1 ( )o ) , i . c . ,
they all v a n i s h ( s e e 5 2 6 . 2 , p r o o f of L e m m a 1 ) .
344
. f ::I. SYSTEhIS OF r t i E FiKST DEGQEF O F S T R C f C T I ' R 4 L [hSTABIL[TY
' T i i f , nr l e m m a i s o f f u n d a m e n t a l i m p o r t a n c e for what f o l l o w s .

\,i
L <: Lct (A) be n dynamic systen: qf class N 2 2, and m and K

/ : I t i ! t i :I.
..Li,/:L,wl, M ! ; I & C ) * . < , 2 < m & k < N . If (A) has a lit)iit cycle L o of rmiEtip[icity
m , Itiex ,for a ; i y e > 0 niid 6 > 0 , there exists ail arinlytical system (A)
6 . ciosc lo rc:r;k k to (A) which has n limit cycle of iniiltiplicity m in U,( L o ) .
P r o v L ' . T h c proof ?%illb e g i v e n f o r t h e case m = 3. T h e p r o o f for
m 79 nt' m ? 3 i s c o m p l e t e l v a n a l o g o u s . F u r t h e r m o r e , i n o.ur t r e a t m e n t
~ j sf v s r e m s of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y , we w i l l o n l y r e q u i r e t h e
c ;1ii m :. ,
7
T!ius, IC,: Lo be a l i n i - t c y c l e of m u l t i p l i c i t y 3 of t h e s y s t e m ( A ) . T h e n
d (11)= I ) . d' (0)= 0, d" (0)= 0 , d" ( 0 )# 0. (-17)
:\.c, fix rhv n u m b e r s E .' 0 and 6 > 0 a n d c h o o s e F ( r , y) to b e t h e f u n c t i o n

(b!' c l a s s iV A 1 d e f i n e d akLove( s a t i s f v i n g c o n d i t i o n s ( a ) a n d ( b ) ) . C o n s i d e r
: h e i?ioclificcl s v s r e n i
and : l i e + c o r r e s p o n d i n g . f u n c t i o n 6=a(n,, po, pl). L e t p,, an: p, b e so s m a l l

- h a t I .:+I i s h ? - c l o s e to i.ank k ?o ( A ) , a n d t h e f u n c t i o n d is d e f i n e d for
:.k! no, 1 no [ <is:, for :vhich .i(no) i s d e f i n e d . C l e a r l y ,
I
d (no, 0, 0 ) d (no). (48j
'.'t*ci:i: A i i and (481 i t 6% see t h a t t h e e q u a t i o n s
d(n,,. {to. pl) i

; 0,
-
dC,(n,, ito, pl) -0. %;(no, [IO. pl) = 0 (49)
iit7v+,; i s;it::u!raneous sc~!ution no=O, p o = O , pi = O . C o n s i d e r t h e J a c o b i a n of

r h i s s\-s:etii c ) f equations. \[.e h a v e
~ I 1-18'
Y and ~ ' 4 7~) it F o ! i c ~ u i ihar d;,(O,
Le: u s c o n i p u t e t i l e v i c m c n t s d L t , dLo, anci

- -
0. O ) = O , dz;(O, 0 , 0 ) = 0, aAi(0, 0, 0 ) f O .
at t h e p o i n t (0, 0, 0 ) . In t h e
c-t+i:ipuT;ttion ~f t i i t ? nui1ibi5rs (0, 0, (J) and dto,,, (0, I). 0 ) vce m a y e v i d e n t l y
r a k e f t ~ ) : : : rht: s t a r ! po = 0 , i . e . , w e m a v c o n s i d e r t h e f u n c t i o n d c o r -
r c s l ~ ~ ~ t i c i i nto. ; ? h e s v s t e t r -=PpplFF;.
dr
dt
$=Q+plFFI. Lemma 8 i s t h u s
applicable. By r h e f i r s t e q u a t i o n i n (46), a ( 0 , 0, p I ) = O . Therefore,
2u J 1 1 . ( I , ij) =Of Now, b; t h e s e c o n d e q u a t i o n i n (461, d:J0,0,pl)= e"!'- 1,
whtLrc. 1 2 0 . Therefore,
-d~o,,l(O,0 , O ) = i # O .
345
Ch.XI1. CREATION OF LIMIT CYCLE FROh.1 LOOP OF SADDLE-NODE SEPARA'I'RIX
F o r the computation of &o(O, 0, 0) we may a s s u m e f r o m the s t a r t pl = 0,

i . e., we may consider the function d corresponding to the s y s t e m
Then, by the c o r o l l a r y f r o m L e m m a 7, d; (0, 0,0) # 0 .

Thus, f o r no = po = p, = 0, none of the elements of the Jacobian
A (no,po, pl)along the second diagonal vanish, whereas all the elements
left of t h i s diagonal a r e z e r o , i. e., A (0, 0, 0) # 0. T h e r e f o r e , by the
t h e o r e m of implicit functions, no = po = pl = 0 is the only solution of s y s t e m
(49) in a sufficiently s m a l l neighborhood of the point (0, 0 . 0 ) .
Consider the analytical s y s t e m
dz
-=
dt
where P , Q are polynomials which a r e 6,-close t o rank k t o the functions

P and Q , respectively, and F is a polynomial 6,-close to rank k + 1 t o the
function F . Let 2 (no, po, pl)be the function corresponding t o t h i s s y s t e m ,
and
d^(no. PO, p1)= 0, k, (no, PO, PJ =o, 4;;(no, po, CL,)= o (50)
the equations corresponding t o s y s t e m ( 4 9 ) . By T h e o r e m 4 (the theorem

of a s m a l l increment of implicit functions, 1.2), s y s t e m (50) h a s a unique
solution in a sufficiently s m a l l neighborhood of the point (0, 0, 0) i f a is
sufficiently c l o s e to rank 2 t o a . On the other hand, by L e m m a 6, i f 6, is
sufficiently s m a l l , d is a s c l o s e as d e s i r e d t o rank k t o 2, where k > m = 3 .
T h e r e f o r e , we can choose 6 , > 0 s o s m a l l that the following conditions are
satisfied:
a
1)6,<T;
2 ) s y s t e m (50) h a s a unique solution (no, Go, bi) in a c e r t a i n neighborhood
of the point (0, 0 , O ) which is as c l o s e as d e s i r e d t o z e r o .
F u r t h e r m o r e , if 6,, io,
a r e sufficiently s m a l l , an additional condition
is satisfied:
3) 2;; ( i o , io, i l l # 0. (51)
T h i s follows f r o m ( 4 7 ) .
Suppose that all t h e s e conditions a r e satisfied. Then
$= P -boQ + ilPF;, g =Q + ji4 + Q P ; (a 1

is a n analytical s y s t e m 6 - c l o s e to (A) which by (50) and (51) h a s a t r i p l e
limit cycle io corresponding t o the value A. of the p a r a m e t e r no and contained
in (I, (L,). T h i s completes the proof of the l e m m a for the c a s e m = 3 .
F o r m = 2, the s y s t e m (A,) should be taken i n the f o r m
dz
= P-poQ, -$= Q + pop,
and in the g e n e r a l c a s e , we should take the s y s t e m
dz
~ i i - = P - ~ o Q + p j F F ; + p z P F ; f . * +P~-zF'"-~F;~
3= Q +pop + pIFFI + pzFzF; + ... +pm-zF'"-24,..
346
531. SYSTEhlS CF T H E FRSI' DEGREE OF STRCCTC'R..\L ISSTABILITY
In all o t h e r r e s p e c t s , the proof f o r m # 3 is the s a m e as f o r m = 3 .

T h e o r e r?i 58. If the dynamic system (A) is a system of the f i r s t
degree of structural insr'ability in It., it may only hace isolated closed
paths in this region.
P r o o f . Let Lo be a closed path of ( A ) in W . We w i l l consider, a s
before, the n o r m a l I t o the path Lo at one of i t s points and the function
d ( n o ) . Let this function b e defined f o r all no, I n o I < n : , and let the path
Lo correspond t o no = 0. H is the region introduced in the definition of
a s y s t e m of the f i r s t d e g r e e
of s t r u c t u r a l instability, H 3 IV.
H
Let Q , Q c H , be a neighbor-
hood of the path L o with the
following property: any path
of ( A ) passing through a point
of Q c r o s s e s the n o r m a l 1 both
for increasing and decreasing t
in the segment I no I < n : . A
sufficiently s m a l l canonical
neighborhood of Lo m a y be
chosen a s Q ( F i g u r e 1 4 8 ) .
Consider a neighborhood U
of the path L contained inside
Q at a positive distance f r o m
F[GL'RE 148 i t s boundary (in F i g u r e 148,
Q is the e n t i r e diagonally
hatched region and U is the
densely hatched region). Let e > 0 be so s m a l l than an E - translation
l e a v e s the neighborhood U inside Q .
Let f u r t h e r 6 > 0 be the number
corresponding t o e according :t the t e r m s of Definition 3 0 . Let 6 1 ,
0 < 6, < 6 , be so s m a l l that if (A) is 6,-close to ( A ) , each path of (A)
passing through points of the region Q c r o s s e s the a r c without contact
I in the segment I co I < n: both for i n c r e a s i n g and d e c r e a s i n g t , and t h e
function a (no)of ( A ) is defined f o r t h e corresponding v a l u e s of t h e
p a r a m e t e r no. By assurnption d (O)=O. Since ( A ) is a s y s t e m of
c l a s s 3, the numbers d' ( 0 ) ,d" (O),d"' (0) exist. If a t l e a s t one of t h e s e
n u m b e r s does not vanish, L o is a limit cycle ( a simple, a double, o r
a t r i p l e one), i.e., i t is an isolated closed path, and the t h e o r e m is
proved.
Now suppose that
(3 (0)= d' (0)= d" (0) = d'(0) = 0.
For p # 0 and a sufficiently s m a l l 6 , , (-4) is a j / 2 - c l o s e t o (A), and by

L e m m a 8, Lo is a limit c,ycle of multiplicity 3 of this. system.
By L e m m a 9, t h e r e e x i s t s an analytical s y s t e m ( A ) 6,/2-close to (A)
which h a s a limit cycle eoof multiplicity 3 i n any a r b i t r a r i l y s m a l l neighbor-
hood of the path L o . S y s t , ? F (A) is &,-close t o (A) and is s t r u c t u r a l l y un-
stable. T h e r e f o r e , ( H , A ) ( H , A ) , and consequently
(U,A ) :,u, A),

Ch.XII.
w e have U c P .
CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRIX
where ?i is s o m e region. Since 0 is obtained by e-translation f r o m U ,
Consider the function d (no)corresponding t o the s y s t e m (A). It is

I
I
defined f o r all no, I no I < n:, and is analytical. Let iobe the value of the
p a r a m e t e r no corresponding t o the cycle 2,. W e may take I io I < n: .
Since eo is a limit cycle of multiplicity 3, 2 (no) f 0 and t h e r e f o r e (z (no)9 0.
But then, because of analyticity, the function 2 (no)may have only a finite
number of r o o t s f o r I no I < n:, i . e., (A) m a y only have a finite number of
closed paths i n Q , and therefore in U . These paths are isolated, i.e.,
they are limit cycles. T h e mapping of fi onto U which implements the
relation (52) moves one of t h e s e cycles into the path L o . Therefore, Lo is
a l s o a limit cycle, i.e., an isolated closed path. T h i s completes the
proof of the theorem.
T h e o r e m 59. I f ( A ) is a system of the f i r s t degree of structural
instability in W , every closed path L~ of the system contained in W i s
either a simple ( i . e . , structurally stable) m a double (of multiplicity 2)
limit cycle.
Proof . Suppose that the t h e o r e m is not t r u e , i. e., (A) h a s a closed
path Lo in W which is neither a simple n o r a double limit cycle of the s y s t e m .
Let Lo correspond t o the value no = 0 c
the path. Let f (no)be the succession function on 1 , which is a p r i o r i defined
f o r a l l no, 1 no I < n:, and d (no) = f (no)- no.
Then
d (0)= d (0)= d(0)= 0. (53)
According t o the previous theorem, .Lo is a limit cycle of ( A ) . Therefore,

i f q > O is sufficiently s m a l l , Lo is the only closed path in U, (Lo),and a l l the
o t h e r paths passing through t h i s neighborhood wind onto L o . Let t h i s condi-
tion be satisfied. Moreover, let U, (Lo) c H, where H is the region intro-
duced i n Definition 3 0 .
Let V b e a neighborhood of the path Lo such that vc U, ( L o ) : let e > 0 be
s o s m a l l that i f is generated f r o m U, by an e-translation, then 3 =I V .
Let f u r t h e r 8 > 0 be the number corresponding t o t h i s e in viztue of
Definition 3 0 . Then f o r any s t r u c t u r a l l y unstable s y s t e m (A) 8 - c l o s e
t o (A) we have ( H , A ) 4 (fi, A, and t h e r e f o r e
where 6 is s o m e region containing V .

Since Lo is a limit cycle of (A), d (no)r e t a i n s a constant sign for all
sufficiently s m a l l no> 0. Suppose that d (no)> 0 and nr is a sufficiently
s m a l l number. Then d (nil))>0 .
where F ( I , y) is the function of c l a s s 4 repeatedly encountered in the

preceding. d (no)is the function corresponding to (A).
Let p # O be s o s m a l l that the following conditions a r e satisfied:
1 ) (A) is 13-close t o (A).
2 ) 2(n:,> 0 .
348
9 31. SYSTEXIE OF THE FIRST DEGREE OF STRLrCTUR4L INSTABILITY
B y ( 5 3 ) a n d Lemma :3 w e see t h a t t h e p a t h L o of (A) is a d o u b l e l i m i t

c y c l e of (A) a n d t h a t 2 ( 0 ) = >'(0) = 0, d" (0) = pfi, w h e r e fi
?
* 0 is a c o n s t a n t i n -
d e p e n d e n t ofy. If t h e s i g n of y is c h o s e n as m i n u s t h e s i g n of 0, a n a d d i - ,
tional condition i s satisfied:
3 ) k (0) = yfi < 0 .
By L I a c l a u r i n ' s f o r m u l a d (no).=d+ n2 +- o (ni), a n d f o r s u f f i c i e n t l y s m a l l
no, d (4< 0 .
L e t nk') > 0, < n;'' b e s u f f i c i e n t l y small. T h e n
4) a (nb")< 0 .
F r o m c o n d i t i o n s 2 arid 4 it f o l l o w s t h a t t h e r e e x i s t s nf), n:) < nap < n i l ) ,
s u c h t h a t 2 ( n p ) = 0 . T h e n u m b e r n:" c o r r e s p o n d s t o a c l o s e d p a t h of
s y s t e m ( A ) w h i c h d o e s riot c o i n c i d e with L o . If n:' a n d p a r e s u f f i c i e n t l y
s m a l l , w e h a v e E c i - . T h u s , for a S u f f i c i e n t l y s m a l l p of a n a p p r o p r i a t e
s i g n , (A) h a s a t l e a s t t w o c l o s e d p a t h s Lo an: L i n tr a n d h e n c e i n 0 .
S i n c e Lo i s a l i m i t c y c l e of m u l t i p l i c i t y 2, ( A ) is s t r u c t u r a l l y u n s t a b l e .
T h i s c o n t r a d i c t s r e l a t i o n (54), w h i c h i n d i c a t e s t h a t d c o n t a i n s o n l y one
c l o s e d p a t h of ( A ) . T h e t h e o r e m is p r o v e d .
R e m a r k . W e c o n s i d e r e d s y s t e m s of class 3, i. e . , s t r u c t u r a l
i n s t a b i l i t y of t h e f i r s t d e g r e e w a s t r e a t e d i n r e l a t i o n t o t h e s p a c e R : J p ( s e e
:5.1). T h i s p r o o f is i n a p p l i c a b l e t o t h e class of a n a l y t i c a l f u n c t i o n s , s i n c e
t h e a n a l y t i c a l f u n c t i o n F (z, y) w i t h t h e d e s i r e d p r o p e r t i e s i n g e n e r a l c a n
b e c o n s t r u c t e d o n l y i n a n e i g h b o r h o o d of t h e p a t h L o . a n d not i n t h e e n t i r e
r e g i o n 6. To p r o v e t h e t h e o r e m i n t h e a n a l y t i c a l case, w e m a y p r o c e e d
as follows: c o n s t r u c t , as b e f o r e , a s y s t e m (.?) of class 3 w h i c h has a
riGuble c y c l e Lo a n d a c l o s e d p a t h i n V . E y Lemma 2, S 1 5 . 2 , t h e r e
e x i s t s a n a r b i t r a r i l y clclse s y s t e m of class 3 , ( A l ) , w h i c h c o i n c i d e s w i t h
(A) e v e r y w h e r e e x c e p t E. s m a l l n e i g h b o r h o o d of t h e p a t h a n d w h i c h h a s
i n t h a t n e i g h b o r h o o d , arid t h e r e f o r e i n V also, a s t r u c t u r a l l y s t a b l e l i m i t
c y c l e E,. B y Lemma 9, o n t h e o t h e r h a n d , t h e r e e x i s t s a n a n a l y t i c a l
s v s t e m ( A 2 ) a r b i t r a r i l y close to (A,) w h i c h h a s a d o u b l e l i m i t c y c l e i n
a n e i g h b o r h o o d of t h e p a t h L o . If (A*) is s u f f i c i e n t l y close to (A1), i t h a s
a s t r u c t u r a l l y s t a b l e l i m i t c y c l e E2 i n t h e n e i g h b o r h o o d of t h e c y c l e z,.
T h u s , t h e a n a l y t i c a l s y s t e m (Az) h a s t w o c l o s e d p a t h s i n V , a n d w e a g a i n
end up with a c o n t r a d i c t i o n t o identity ( 5 4 ) .
4. A saddle-point separatrix forming a loop
A s a d d l e - t o - s a d d l e s e p a r a t r i x w a s c o n s i d e r e d i n d e t a i l i n C h a p t e r s IV
a n d XI. In C h a p t e r IV (!I 11) w e e s t a b l i s h e d t h a t a s t r u c t u r a l l y s t a b l e
s y s t e m c a n h a v e n o s u c h s e p a r a t r i c e s . In C h a p t e r XI w e c o n s i d e r e d a
s e p a r a t r i x f o r m i n g a locip a n d d e r i v e d some of i t s p r o p e r t i e s .
In t h i s s u b s e c t i o n w e w i l l d e a l w i t h a s a d d l e - t o - s a d d l e s e p a r a t r i x of
a s v s t e m of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y . The case of a
s e p a r a t r i x b e t w e e n t w o d i f f e r e n t s a d d l e p o i n t s is of n o i n t e r e s t for o u r
p u r p o s e s , since these s e p a r a t r i c e s are not c h a r a c t e r i z e d by a n y a d d i -
t i o n a l n e w p r o p e r t i e s i n systems of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y .
W e w i l l t h e r e f o r e c o n s i d e r t h e case w h e n a s a d d l e - p o i n t s e p a r a t r i x goes
to t h e same s a d d l e point for b o t h t + - 00 a n d t - + + 00, i . e., i t f o r m s a
l o o p . T h e r e s u l t s follow almost d i r e c t l y f r o m t h e f i n d i n g s of C h a p t e r X I .
349
Let (A) be a s y s t e m of the f i r s t degree of s t r u c t u r a l instability in

W , 0 (zo,yo) a saddle point of (A), Lo a s e p a r a t r i x of the saddle point 0 which
is contained in W and forms a loop t h e r e . A s in Chapter XI, we a s s u m e
that the other two s e p a r a t r i c e s of the saddle point 0 lie outside the loop
formed by the s e p a r a t r i x L o .
L e m m a 10. There exists eo > 0 such that the eo-neighbmhood of
the loop Lo does not contain any closed paths of (A).
P r o o f . Suppose that the l e m m a is not t r u e , i . e . , any neighborhood
of the loop Lo contains closed paths of ( A ) . Then by T h e o r e m 4 4 , 3 2 9
(Jo = R (50, Yo) +QI (50, Yo) = 0. (55)
Closed paths passing sufficiently c l o s e to Lo c l e a r l y may only lie

i n s i d e the loop.
Let H be the region introduced in Definition 30, H x W . Let f u r t h e r
q > 0 b e s o s m a l l that the neighborhood U , ( L o )C H does not contain any
equilibrium s t a t e s of ( A ) , except the point 0 , nor any closed paths which
lie inside the loop Lo ( F i g u r e 1 4 9 ) . Take s o m e e, O < e < + . Definition 30
a s s i g n s a c e r t a i n number 6 > 0 to t h i s e such that i f (8)
is 6 - c l o s e t o (A)
and is s t r u c t u r a l l y unstable, we have ( H , A ) (B,2). Then
e - -
(ut,(Lo), -4) (V, A ) , (56)
where F is s o m e region.
FIGURE 149
F r o m identity (56) and the inequality a<+ we have
We m o r e o v e r see f r o m (56) . , that contains o r e c i s e l v one saddle Doint 6

of s y s t e m (8) and precisely one loop Eo f o r m e d by the path of (A) c o r -
responding t o the path L o . Finally, by the s a m e relation (5_6), any
neighborhood of the loop z0 should contain closed paths of (A). This,
however, contradicts the r e s u l t s of Chapter XI. Indeed, by L e m m a 10,
1 2 9 . 4 , t h e r e e x i s t s a s y s t e m (A) &-close to (A) for which 0 is a saddle
point; t h i s saddle point h a s a s e p a r a t r i x zo
entirely contained in u~(Lo)
and m o r e o v e r
350
5 31. SYSTEMS CY THE FIRST DEGREE OF STRUCTC'RAL iNST.4BILITY
P y T h e o r e m 44, s29.1, on the other hand, the loop Lo i s unstable, i . e . ,

a sufficiently s m a l l neighborhood of the loop can contain no closed paths.
T h i s proves the 1emm.i.
Lemma 10 evident1:y implies that if ( A ) h a s i n W a s e p a r a t r i x which
f o r m s a loop, t h i s loop is e i t h e r stable o r unstable.
T h e o r e m 60. If a system of the f i r s t degvee of structitral instability
in W has a saddle-poi8r2t0 (z0, yo) whose separatrix forrns a loop contained
in rr', then oo (z0. yo) + 0.
P r o o f . Suppose that the theorem is not t r u e , i . e . , s y s t e m (A) h a s a
saddle point 0 (zo, yo) u hose s e p a r a t r i x Lo i s entirely contained in W. and
f o r m s a loop, and yet o0 (zo,YO) = O .
I n o u r proof to T h e o r e m 50 (529.4) we showed that if the loop Lo is
s t a b l e ( o r unstable), and oo (zo, yo) = 0 , then for any q: 0 and 6 > 0 t h e r e
e x i s t s a s y s t e m ( A ) 6 - c l o s e t o (A) which h a s in U,(Lo) a s e p a r a t r i x that
2
f o r m s a loop and at l e a s t one closed path.
(A) i s a s t r u c t u r a l l y unstable s y s t e m , and therefore f o r appropriate
e and 6, relation (56) should b e satisfied f o r t h i s s y s t e m . On the o t h e r
hand, this relation carnot be t r u e s i n c e f o r sufficiently s m a l l and e ,
U , ( L O )d o e s not contair. closed paths of ( A ) and V contains a t l e a s t one
closed path of ( A ) . The contradiction e s t a b l i s h e s the validity of the
theorem.
T h e o r e m 60 signifies that i f a s y s t e m of the f i r s t d e g r e e of s t r u c t u r a l
instability in W h a s a saddle point 0 (.to,yo) for which uo (zo,yo) = 0, no
s e p a r a t r i x e x i s t s f o r t : i i s saddle point which f o r m s a loop and is contained
entirely in W .
5. T h e s i m p l e s t dtructurally unstable paths
A s we know, singular paths of a dynamic s y s t e m , i . e . , equilibrium

s t a t e s , limit cycles, and s e p a r a t r i c e s , a r e the most important e l e m e n t s
f o r the analysis of the topological s t r u c t u r e of a dynamic s y s t e m on a
plane. By T h e o r e m 23, 9 18.2, s t r u c t u r a l l y stable s y s t e m s m a y only
have singular paths of the following types:
(a) s t r u c t u r a l l y stable equilibrium s t a t e s - nodes, saddle points, and
simple foci;
(b) s t r u c t u r a l l y s t a b l e limit cycles;
(c) saddle-point s e p a r a t r i c e s going to a simple node, a s i m p l e focus,
o r a s i m p l e limit cycle o r leaving t h e region of definition.
Any s t r u c t u r a l l y unstable s y s t e m , in p a r t i c u l a r a s y s t e m of the f i r s t
d e g r e e of s t r u c t u r a l instability, should have a t l e a s t one singular path
which d o e s not fit the a.bove classification. Consider the following addi-
tional types of paths:
1 . A multiple focus of multiplicity 1.
2. A saddle-node of multiplicity 2 with o = P ; + Q; # O .
3. A limit cycle of multiplicity 2 .
4 . A s e p a r a t r i x f r o m one saddle point t o another saddle point.
5. A s e p a r a t r i x of saddle point M (zo, yo) which f o r m s a loop when
0 b o , Yo) # 0.
35 1
Ch.Xl1. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARA'I'KIX
T h e p a t h s of t h e s e f i v e t y p e s w i l l b e c a l l e d t h e s i m p 1e s t s t r u c -
t u r a l l y u n s t a b l e p a t h s . T h e e x i s t e n c e of t h e s e p a t h s i n s y s t e m s
of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y d o e s not c o n t r a d i c t T h e o r e m s 53
t h r o u g h 60. @n t h e o t h e r h a n d , a d y n a m i c s y s t e m ( A ) of t h e f i r s t d e g r e e
of s t r u c t u r a l i n s t a b i l i t y i n W s h o u l d h a v e i n t h i s r e g i o n a t l e a s t o n e s i m p l e s t
s t r u c t u r a l l y u n s t a b l e p a t h ( o t h e r w i s e , it would b e s t r u c t u r a l l y s t a b l e i n W ) .
We w i l l s h o w i n t h i s s u b s e c t i o n t h a t a d y n a m i c s y s t e m of t h e f i r s t
d e g r e e of s t r u c t u r a l i n s t a b i l i t y i n W c a n n o t h a v e m o r e t h a n o n e s i m p l e s t
s t ~ ~ u c t u r a l ul ynstable path i n t h i s region.
F i r s t w e s h a l l p r o v e a n u m b e r of l e m m a s .
L e m m a 1 1 . Let Lo be a separatrix of the saddle point 0 of a dynamic
system (A) of class N which extends to another saddle point O1. There
flXl
exists a simple closed curve of class k ( k being a
given number<N +
1) which passes through the
points 0 and 01, encloses L ~ and , does not enclose
,.; 2: 4" \, 5; any other separatrix o r any equilibrium state of
system (A).
P r o o f . We s h a l l f i r s t s h o w t h a t a n a r c 1 of
a p a r a b o l a c a n b e p a s s e d t h r o u g h t h e point 0 ,which
h a s n o c o n t a c t s w i t h t h e p a t h s of s y s t e m (A) e x c e p t
at t h e point 0 ( F i g u r e 150). Without loss of g e n e r a l i t y ,
8p.' A p ,' w e m a y p l a c e t h e point 0 a t the o r i g i n a n d r e p r e s e n t
P"
t h e s y s t e m ( A ) i n t h e c a n o n i c a l form
dz
-=
dt X,z+Cp(z, Y), $=-2y+wz, Y). (57)
FIGLIKE 15U
w h e r e 'p and $ a r e f u n c t i o n s w h i c h v a n i s h at t h e
point O ( 0 , 0 ) t o g e t h e r w i t h t h e i r f i r s t o r d e r d e r i v a t i v e s
(see s 8 . 1 , ( l ) , ( 2 ) ) , a n d I l l z < O .
B y a s s u m p t i o n , all o u r s y s t e m s , s y s t e m (A) i n c l u d e d , a r e s y s t e m s of
c l a s s 3 . T h e f u n c t i o n s 'p and +,
b y T h e o r e m 5 of t h e Appendix (see
A p p e n d i x , s u b s e c t i o n 2), m a y t h e r e f o r e be w r i t t e n i n t h e f o r m
w h e r e P: a n d Qg a r e c o n t i n u o u s f u n c t i o n s , w h i c h v a n i s h at t h e point O ( 0 , 0 ) .
C o n s i d e r t h e p a r a b o l a y = C x 2 , w h e r e t h e v a l u e of the c o e f f i c i e n t C w i l l
b e c h o s e n a t a l a t e r stage. T h e c o n d i t i o n t h a t t h e p a r a b o l a I/ = Cxp is
t a n g e n t to a p a t h of s y s t e m (57) c a n b e w r i t t e n at t h e p o i n t of c o n t a c t i n
t h e form
2cx [hlx + 'p (I,!/)I - I1zY +9 (G Y)I = 0.
I n s e r t i n g for 'p a n d 0 t h e i r e x p r e s s i o n s f r o m (58) a n d s u b s t i t u t i n g Cx*

for y. w e o b t a i n
=o.
X Z I ( ~ ~ ~ - - ~ ~ ) C - - ~ ~ J + O ( ~ Z ~ ~ ) (59)
S i n c e hlh2<0, we h a v e 21, - hz P O . Therefore, for any C #A and

2al -i2
for
all I, I x I -== zo, w h e r e x,, is a s u f f i c i e n t l y small p o s i t i v e n u m b e r , t h e arc
352
4 .I. 5YSTEh S OF THE FIRST DEGREE OF STRUCTCRjL IhST4BILIT'i
o f t h e p a r a b o l a y = L'Z' is without c o n t a c t w i t h t h e p a t h s of s y s t e m ( 5 7 ) ,
e x c e p t a t t h e point 0 . if'e d e s i g n a t e t h i s a r c 1 .
T h e a r c 1 c l c a r l v c a n b e d r a \ v n so t h a t t h e s e p a r t r i x L o of t h e s a d d l e
point 0 lies o n o n e s i d e of t h e a r c 1 (on t h e c o n c a v e s i d e of t h e p a r a b o l a } ,
and t h e p t h e r t h r e e s e p a r a t r i c e s l i e on t h e o t h e r s i d e of t h e a r c . An
a n a l o q o u s a r c I t of t h e p a r a b o l a c a n b e d r a w n t h r o u g h t h e s a d d l e point 0,
iF'iqu r e 1 5 0 1 .
[.et 0 b e a n a - l i m i t point of t h e p a t h Lo, a n d 0,i t s a - l i m i t p o i n t . T h e
p n i n t s 0 and 0,r e s p e c t i v e l y d i v i d e e a c h of t h e a r c s 1 a n d 11 i n t o t w o p a r t s :
1' a n d l ' , i, a n d 1 ; . It i s r e a d i l y s e e n t h a t a n y p a t h c r o s s i n g t h e arc 1 s u f -
f i c i e n r l v close t o t h e [point 0 w i l l also c r o s s t h e a r c 1;. a n d a n y p a t h
c r f l s s i y t h e arc 1" u - i d c r o s s 1 ; . C h o o s e o n e arc of e a c h of t h e s e p a i r s .
i Y e o b t a i n a s i m p l e pieccLr-ise-smooth c l o s e d c u r v e m a d e up of t h e s e g -
m e n t s P ' P " and Q'Q" of t h e a r c s 1 a n d it and t h e a r c s P'Q' and P"Q'of p a ? h s .
T h i s closed c u r v e evidently e n c l o s e s t h e s e p a r a t r i s Lo.
O n t h e a r c OP' we '2hoose a point d w h i c h d o e s not c o i n c i d e e i t h e r w i t h
0 01' with P ' , and o n t h e a r c P'Q' of a p a t h w.e c h o o s e a p o i n t B . S i n c e 1 i s
at? a r c without c n n r a c c , we c a n a l w a y s j o i n t h e point d to t h e point B b y a n
arc ( w i t h o u t c o n t a c t ) , w h i c h at t h e point .1 a n d a t t h e point B h a s a p o i n t of
r w ~ f a c c.f r a n v d e s i r e d o r d e r k < N L 1 with t h e a r c 1 a n d r e s p e c t i v e l y w i t h
t h e a r c nt' t h e p a t h ( t h i s i s r e a d i l y p r o v e d u s i n g QT, 5 3 . 5 , L e m m a 8).
P r a u i n q t h r e e o t h e r analogous arcs ( F i g u r e 150) we obtain a s i m p l e
c l o s e d curve of c l a s s k w h i c h c o m p l e t e l y e n c l o s e s t h e s e p a r a t r i s Lo a n d
clfies not e n c l o s e o t h e r s e p a r a r r i c e s o r a n y e q u i l i b r i u m s t a t e s of t h e
s y s t e m . Q. E . D.
R Fm a r k . T h e n u m b e r k , i n g e n e r a l , s h o u l d b e GLV+ 1, s i n c e t h e
p a t h s of s y s t e m s of c l a s s :V a r e c u r v e s of class A'+ 1, and acelording t o
the. c o n s t r u c t i o n u s e d i n t h e p r o o f of t h e l e m m a , s e g m e n t s of t h e s e c u r v e s
are included i n t h e c l o s e d c u r v e .
L e t n tti a 12. Let L o be a separatrh of system (A) of class N which
originates atul ends in the saddle point 0 Cfmtning a loop L ) , and let there
be a neighborhood of L which does not contain any closed paths (so that
the loop is stable os unstable). Then there exist two sittiple closed curces
N + 1 , ane of which
C' and C" of class k , iohere k is any fixed integer, k -:
encloses the loop L a d the other is enclosed b y the loop L, such that the
annular region bcticem C' and C" contains no equilibrium states, no closed
citrues, and no separatrices (except L ) of systetn (A).
P r o o f . S u p p o s e t h a t t h e o t h e r t w o s e p a r a t r i c e s of t h e s a d d l e - p o i n t 0
( d i f f e r e n t from L o ) l i e o u t s i d e t h e l o o p f o r m e d b y t h e p a t h L .
!I'e c a n a I w a v s p a s s t h r o u g h t h e s a d d I e point 0 a s e g m e n t of a s t r a i g h t
l i n e which w i l l b e without c o n t a c t at all p o i n t s s u f f i c i e n t l y close t o 0 ( o t h e r
Than 0 i t s e l f l (see QT, 37.3, p . 1 5 5 j . L e t I be s u c h a s e g m e n t w h i c h c c n -
r a i n s 0 a n d i s withou- c o n t a c t w i t h t h e p a t h s of ( A ) a t all p o i n t s o t h e r t h a n
0 (Cigure 151).
L e t 1' a n d I" b e t h e t w o s e g m e n t s i n t o w h i c h 1 is d i v i d e d b y t h e s a d d l e
p(kinr 0 . It i s r e a d i l s s e e n t h a t a n y p a t h c r o s s i n g t h e s e g m e n t I' a t a
point P' sufficient1.y close t o 0 w i l l c r o s s 1" a t s o m e point P " .
\Ye t h u s o b t a i n a s i m p l e c l o s e d p i e c e w i s e - s m o o t h c u r v e e n c l o s i n g
t h e l o o p L , s u c h t h a t t h e r e g i o n b e t w e e n t h i s c u r v e a n d L c o n t a i n s no
e q u i l i b r i u m s t a t e s , no c l o s e d p a t h s , and no s e p a r a t r i c e s ( t h i s i s r e a d i l y
353
verified, since a r c s of paths c r o s s i n g the

segments I' and 1" of the a r c 1 p a s s through
a l l the points of t h i s annular region).
"Smoothing" t h i s piecewise- smooth curve
by the s a m e technique as in the previous
l e m m a , w e obtain one of the two simple
closed c u r v e s , C'say, postulated by
the l e m m a .
The second curve C" can be chosen a s
- - a cycle without contact, sufficiently close
t o the "loop" (Figure 15 l ) , T h e exis;tence
FIGURE 151
of t h i s cycle is established in QT, 124.3,
Lemma 2.
The s a m e proof can be used in the c a s e when the two other s e p a r a t r i c e s
of the saddle point 0 lie inside the loop formed by the s e p a r a t r i x L .
L e m m a 13. Let y be a simple closed curve of class k in some region c .
There exists a function z = Q, ( x , y) of class k - 1, defined in B , which vanishes
at the points of the curve y , takes 072 positive values inside y and negative
values outside y , and at the points of the curve y satisfies the relation
o;*+ 0;*
z 0.
P r o o f . Choose the arc length s reckoned f r o m s o m e fixed point of
the c u r v e a s the p a r a m e t e r of y . Then, a s is readily seen, y can be
written in p a r a m e t r i c form x = ' p (s), y = $ (s), where 'p and q a r e functions
of c l a s s k + 1, s v a r i e s f r o m 0 t o s o m e T , and 'p (T)= 'p ( O ) , q (T) =I@(0). Since
y is smooth, 'p' (s)and 0' (s) do not vanish simultaneously f o r any 8 , O<s< T.
In the neighborhood of y we define a curvilinear s y s t e m of coordinates
(s, n) by the equalities
I = 'p (s) + nq' (4, y =9 ($1-nq' (SI, (60)
where the right-hand s i d e s a r e evidently functions of class k . According t o

Chapter V (see 15.1, Lemma 1 and R e m a r k I t o the l e m m a ) , equations (60)
define n a s a single-valued function of the coordinates x and y in s o m e
neighborhood of the c u r v e y :
n = p ( s , y),
where F is a function of class k - 1 which vanishes on y , is positive on one
side of y and negative on the other side of y , and on y s a t i s f i e s the relation
F;'+F;+O. W e c a n choose the direction of the normals so that the function
F ( x , v) is positive inside the c u r v e y and negative outside this c u r v e . Con-
s i d e r any two closed c u r v e s y , and y 2 defined by the equations n = n, and
n = nL, respectively, i. e ., F (2, y) = n, and F ( x , g) = n2, where the numbers
n, and n2 are sufficiently s m a l l and 0 < n, < n2 (Figure 152).
Let f = f (n)be a function of class k- 1 defined for O < n < 00, which
s a t i s f i e s the following conditions:
1) f (n) 3 n f o r O < n < n , ;
2 ) f (n)E n2 f o r n > n 2 ;
3) n , < f (n)<n2 f o r ni < n < n2 (a specimen graph of the function f (n) is
shown in Figure 153; a s i m i l a r function was constructed in 4 15.2
354
s 31. SYSTELIS CF THE FlRST DEGREE OF STRUCTlrR.4L WSTABILITY
i n t h e p r o o f of L e m m a 2 ) . T h e ? u n c t i o n a (2, y) will b e defined i n s i d e t h e

c u r v e y b y t h e following conditions:
1) i n t h e r i n g b e t w e e n y a n d y ? ,
( r , y) = f (n)= f (F (f,y));
2) i n s i d e y ~ ,
Q, (3, y) = n*.
F[GL'RE 152 FIGL'RE 153
Q, (I, y) i s s i m i l a r l y d e f i n e d o u t s i d e t h e c u r v e y . T h e f u n c t i o n 0 c o n -
s t r u c t e d i n t h i s w a y e v i d e n t l y m e e t s a11 the c o n d i t i o n s of t h e lemma. Q . E. D .
L e m m a 14. Let
be a dynaiitic system of class ,v, L its separatrix extenclingfrotn saddle

point 0' to saddle point 0"( o ' a n d 0"may be the same point). F m any
6 > 0 and r s L V ( r is a natural number), there exists an analytical systein
(A) &-closeto rank r to (A) which has a saddle-to-saddle separatrix.
P r 0 0 f , T h r o u g h some point M of t h e s e p a r a t r i x L , w e d r a w a n a r c
w i t h o u t c o n t a c t I a n d d e f i n e a p a r a m e t e r u o n t h i s a r c , so t h a t t h e point 31
c o r r e s p o n d s t o u = 0 . To fix i d e a s , l e t t h e s e p a r a t r i x L m a k e a p o s i t i v e
a n g l e with t h e a r c I ( F i g c r e 151).
F y t h e f u n d a m e n t a l t h e o r e m of t h e s t r u c t u r a l s t a b i l i t y of d y n a m i c
s y s t e m s ( T h e o r e m 23, 3 18.2), t h e d y n a m i c s y s t e m (X)i s s t r u c t u r a l l y
s t a b l e in s u f f i c i e n t l y s m a l l n e i g h b o r -
h o o d s of t h e s e g m e n t s O'M a n d 0"M
+-+%
of t h e s e p a r a t r i x L. H e n c e i t fol-
R7zy p l o w s , a s is r e a d i l y s e e n , t h a t a
n u m b e r q > 0 e x i s t s with t h e f o l l o w i n g
p r o p e r t y : i f (A) is q - c l o s e to (A),
0' L' t h e n i n a s u f f i c i e n t l y small n e i g h b o r -
hood of t h e s a d d l e p o i n t 0' (0") there
e x i s t s a s i n g l e s a d d l e p o i n t @ (8') of
the system (x) and t h e s e p a r a t r i x
FIGURE 154 ?j (E') of the s a d d l e p o i n t 6' (6.") crosses
t h e a r c w i t h o u t c o n t a c t 1 at t h e p o i n t
- .%' (1G") c o r r e s p o n d i n g to t h e v a l u e
u' (L") of t h e p a r a m e t e r u , s o t h a t t h e s e g m e n t ail?* of t h e s e p a r a t r i x
2' Gw)i s c o n t a i n e d i n s i d e a n a r b i t r a r i l y small n e i g h b o r h o o d of t h e s e g m e n t
O ' M (0"M)of t h e s e p a r a t r i x L ( F i g u r e 1 5 4 ) .
355
Let 8 , be a positive number, 8 , <%, a 1 C6T .
dz
-=
dt P-hQ, s=Q+W.
Let Xo: 0 be so s m a l l that f o r a l l h , 1 h I 0, At < A,, hz < 0 , I h2 I < Lo. On passing f r o m ( A ) to (Ah,)
( i :1, 2 ) , the s e p a r a t r i x L branches into two s e p a r a t r i c e s L; and L; which
c r o s s the a r c I at the points N i and N:, respectively. Let these points
correspond t o the values u (A,) and uv (A,) of the p a r a m e t e r u. By the
l e m m a of 311.1, we have
u >0,
(L) u(2,) <0 ,
u <0
(b) u ( A Z ) >0,
(Figure 155).
Let P* and Q* be polynomials which provide a sufficiently c l o s e fit of
the functions P and Q s o that the following conditions be satisfied:
1) For a l l 1, lhl<A,,, the s y s t e m
is 6,-close t o rank r t o ( A ) .
2 ) The s e p a r a t r i c e s (h,) and c(Xt) of (Ah,) c r o s s the a r c without contact 2
at the points &(A,) and N ( h , ) corresponding t o the values ;(hi) and C(hl) of
the p a r a m e t e r which a r e so close t o the points N ; and N ; that
I
3 (A,) >o, u (A,) <0.

3) The separatrices_Z (A2) and E (h2)of s y s t e m (AA*)c r o s s the ar_c without
contac I at the points N (b) and (b)corresponding t o the values u (h,) and
the p a r a m e t e r which a r e so close t o the points N; and N ; that
u: (h2)< 0, 2 (5)> 0. (63)
62) and (63), we have
- (hi)--?
24 (h,) > 0,
2 ($) -2 (&) < 0.
FIGURE 155
356
f. 31. SYSTELIS OF TtiE FIRST DECREE OF STRLCTL'fLlL IZSIABILITY
S i n c e 3 and t? a r e c o n t i n u o u s f u n c t i o n s of t h e p a r a m e r e r h , t h e r e e x i s t s
a nuniberX, h 2 < ~ < h l , s u c h that
T h i s i m p l i e s t h a t t h e s e p a r a r r i s of t h e s y s t e m ( 3 1 ) e x t e n d s f r o m s a d d l e
point a' to s a d d l e point 6". S i n c e (-xi)i s 6 - c l o s e to rank r t o (-4) and is a n
a n a l v t i c a l s y s t e m , i t s a t i s f i e s all t h e p r o p o s i t i o n s of t h e l e m m a . Q. E . n.
T h e o r e m 61. If (A) is a dyiminic systerti of the f i r s t degree of
structural itstability in W , it may hace at most one simplest stnicttrrally
unstable path in this region.
P r c) o f . S u p p o s e t h a t (A) h a s t w o s i m p l e s t s t r u c t u r a l l y i l n s t a b l e p a t h s
i n i i ' . In o u r p r o o f , w e d o not d i s t i n g u i s h b e t w e e n p a t h s of t y p e 4 and t y p e 5
I see p. 351), i . e ., a s a d d l e - to- s a d d l e s e p a r a t r i s m a y b e a s e p a r a t r i x b e t w e e n
t w o d i f f e r e n t s a d d l e p o i n t s o r a s e p a r a t r i x f o r m i n g a l o o p for a s a d d l e point
with o Z 0 . We m o r e o v e r a s s u m e t h a t all t h e s y s t e m s a r e a n a l y t i c a l , i . e . ,
s-e a r e d e a l i n g w i t h s t r u c t u r a l i n s t a b i l i t v of t h e f i r s t d e g r e e u-ith r e s p e c t t o
t h e s p a c e Rb", w h e r e r > 3 . O u t - p r o o f r e m a i n s i n f o r c e , a s i s r e a d i l y s e e n ,
i n i h e n o n a n a l y t i c a l cast? as w e l l ( i . e . , for t h e s p a c e R$),3 g r g . V ) , and
s o m e a r g u m e n t s c a n a c t u a l l y b e s i m p l i f i e d i n t h e nonanalq.tica1 case.
L F ~ C~S c o n s i d e r s u c c e s s i v e l y all t h e p o s s i b l e cases of t w o s i m p l e s t
structurallv unstable paths in ti-.
1 ) S y s t e m ( A , h a s tu.,? s a d d l e - n o d e s i n 15..
Let o n e of t h e s e s a d d l e - n o d e s b e t h e p o i n t O ( O , O ) , and t h e o t h e r O , ( a , b ) ,
a n d l e t t h e s y s t e m ( A \ h.ave the form
dr
;iT.=p(z,y). 2:
-= Y - r 4 ( 2 ? Y), (A)
w h p w p ( 0 , O ) = p ; ( O , O ) = & ( O , O ) = q ( O . 0 ) = q ; ( 0 . 0)=q;(O. O ) = O ( s e e 531.1, (11)

and ( 1 2 ) ) .
S i n c e Ol(a, b) is a s a d d l e - n o d e , w e h a v e
C o n s i d e r t h e modified s v s t e m
dd xl
-= p @ . Y) +p I V - ~ lP ( 1 , Y),
(~~-b*)= =Y
; Q (t,Y) =Qtz, Y). (6)
u h e r e p3-0. For a n y p , t h e point O , ( a , b) is a m u l t i p l e e q u i l i b r i u m
s t a t e of (At, since x ( a , t ) = A ( a , b)= 0 . ( A ) is t h e r e f o r e s t r u c t u r a l l y
unstable.
Let y = q ( z ) b e a s o l u t i o n of t h e e q u a t i o n y + q ( z , y)= 0 i n t h e n e i g h b o r -
hoo-l of t h e point O ( 0 , 0 ) , a n d l e t $ (I) = P (2,q (3)) = p (I,9: (2)). S i n c e 0 is
a n e q u i l i b r i u m s t a t e of m u l t i p l i c i t y 2 for s y s t e m (A), t h e f u n c t i o n 9(2)h a s
the form
9 (1)=a&+ap3+. .., (65)
w h e r e a,+O ( T h e o r e m 3.3, s 2 3 . 1 ) .
357
Ch.XI1. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPARATRLX
We will now s e a r c h f o r the equilibrium s t a t e s of (8)

in the neighborhood
of 0 ( 0 , 0 ) . T o t h i s end, we have t o solve the simultaneous equations
p" (2,Y) = 0, Q"@. u) = 0

or, equivalently,
Y = cp ( 4 7 0 ( x , 'p (4)
= 0.
Since 'p(z)='p'(z)=O ( s e e (15), (16)), the series expansion of the function

cp (2)does not contain any t e r m s below second o r d e r . Therefore,
O(X, ' p ( z ) ) = p ( s , 'p(z))+p I(zS--a*)*+('p'(")-b*)'l =

=azz* + + ... +
aazs p [a4+ bd]-2pa'z' + - .. = p (a4 + b.')+a& + ... ,
where az J. 0, and the missing t e r m s a r e of third o r higher o r d e r in all
the v a r i a b l e s p and x jointly.
We t h u s have t o find the roots of the equation
p (a4+b4)+a&+. .. = O (66)
which a r e c l o s e t o zero; i n t h i s equation p # 0, at # 0. A detailed investiga-

tion of the roots of t h i s equation will b e given in the next chapter, in E 3 2 .
In particular, i t follows from L e m m a 2 of this section that if p i s sufficiently
s m a l l and i t s sign is minus the sign of a 2 , equation (66) has p r e c i s e l y two
real roots, which both go t o z e r o for p -+ 0 . T h i s signifies that f o r a
sufficiently s m a l l p of a n appropriate sign, s y s t e m (A) h a s at l e a s t two
equilibrium s t a t e s in any neighborhood of the point O ( 0 , O ) . Since (A) is
s t r u c t u r a l l y unstable, this contradicts L e m m a 2 of the present section.
2 ) System ( A ) h a s a multiple focus (of multiplicity 1) and a saddle-node
in W .
Without loss of generality, we may place the focus at the origin O ( 0 , O )
and write ( A ) in the f o r m
dz
-=
dt --Y+'p(x, y ) = P ( s , Y), $=z+*(x, Y)=Q@, Yh (A)
where 'p and *

are analytical functions whose series expansions s t a r t with
quadratic t e r m s . Let 0 , ( a , 6)cW be the saddle-node of ( A ) .
T h e modified s y s t e m is taken in the f o r m
dt
= P - pQ =p, df
dll = Q+ pP = @, (A)
which is obtained f r o m s y s t e m (A) by rotating i t s vector field through
the angle tan-' p.
Since 0, (a, b) is a saddle- point of ( A ) , we have P (a, b) = Q ( a , b) = A (a, b) = 0.
-
F r o m t h e s e relations and f r o m the obvious equality 8 ( x , g) = A (3, g) (i+p*) we
conclude that P ( a , b ) = g ( a , b ) = A ( a , b ) = O , i.e., the point 0, is a s t r u c t u r a l l y
unstable equilibrium s t a t e of (A).
(8) is thus a s t r u c t u r a l l y unstable s y s t e m . Reasoning as in the proof
t o T h e o r e m 5 7 ($31.2, ( 3 3 ) and (34)), we can show that i f p i s sufficiently
s m a l l , (A) c a n have no closed paths in s o m e neighborhood V of 0 . By
R e m a r k 3 t o T h e o r e m 14, 9 10.3, f o r a sufficiently s m a l l p of an appropriate
358
4 31. SYSTEVS OF TliE FIRST DEGREE OF STRL'CTL'RAL L\STAEILITY
sign, ( A ) has at l e a s t one closed path in any a r b i t r a r i l y s m a l l neighborhood

of 0 . We have established a contradiction, which proves the t h e o r e m f o r
this case.
3 ) System ( A ) h a s two multiple foci in W .
Let one of t h e foci be O ( 0 , 0 ) and the other O1 (a, 6 ) . W e write the s y s t e m
in the f o r m
T h e modified s y s t e m is
Since the point O1 ( a , b) i s a multiple focus of ( A ) , we have A ( a , b)> 0,

0 (a,6 ) = 0 . a
A d i r e c t check w i l l show that (a, 6 ) = A (u, b ) and ( a , 6 ) = u ( a , b)
T h e r e f o r e , 0,(a. 6) is a i equilibrium s t a t e of (A) with pure imaginary
c h a r a c t e r i s t i c numbers, i . e., ( A ) is s t r u c t u r a l l y unstable.
( A ) can be written in the f o r m
-=
ddtr -yY-p'a2+b*)z+ ... , t = z-p(a*ib*))y+
dd y ...,
where the missing t e r n s are of second or higher o r d e r . The point O ( 0 , O )
is t h e r e f o r e a s t r u c t u r a l l y s t a b l e focus of ( A ) , stable or unstable
depending on the sign of p. Then, a s in R e m a r k 3 t o T h e o r e m 14, we
can show that f o r a sufficiently s m a l l p of an appropriate sign, (A) h a s
a t least one closed path in any a r b i t r a r i l y s m a l l neighborhood of 0. W e
again reach a contradiction as in case 2 .
4) System ( A ) h a s a saddle-node and a limit cycle of multiplicity 2 in W .
Let O ( O . 0 ) be the saddle-node and L the limit cycle. T h e modified
s y s t e m is again taken i n the f o r m
which is obtained by a n appropriate rotation of the vector field of ( A ) .

( A ) is s t r u c t u r a l l y unstable, since 0 ( 0 , 0 ) i s i t s equilibrium s t a t e and A ( 0 , O ) = 0 .
We s h a l l again u s e oiie of the r e s u l t s of the next chapter, namely
T h e o r e m 71 (S32.4). By t h i s theorem, for a sufficiently s m a l l pof a n
appropriate sign, (A) h a s two closed paths in any neighborhood of the
limit cycle L of multiplicity 2. On t h e other hand, L i s a limit cycle
of ( A ) , and ( A ) is a s t r u c t u r a l l y unstable s y s t e m . T h e r e f o r e , reasoning
a s w e have done often before, w e can show that f o r a sufficiently s m a l l
p, ( 3 )may have only one closed path in a sufficiently s m a l l neighborhood
of L. W e have again established a contradiction.
5) System ( A ) h a s a saddle-node O ( 0 , O ) and a saddle-to-saddle
s e p a r a t r i x L in W .
A s in the previous c a s e , (A) is chosen as the modified s y s t e m .
The point O ( 0 , 0 ) is a multiple equilibrium s t a t e of (A), and t h e r e f o r e
(A) is s t r u c t u r a l l y unstable. T h e r e f o r e , f o r e v e r y E > 0, t h e r e e x i s t s p,,,
such that for I p I < p,, WE have the relation ( H , A ) ( E , 2)and hence the
359
relation (U (L),A ) ZE
e
(v, A ) , where U (t)is a neighborhood of the s e p a r a t r i x L .
F r o m the l a s t relation i t follows that contains a saddle-to-saddle
s e p a r a t r i x of (A). On the other hand, in o u r proof t o Theorem 16, 9 11.2,
w e have s e e n that i f the neighborhood U (L)of the path L and the numbers e
and p a r e sufficiently s m a l l , may riot contain any saddle-to- saddle
s e p a r a t r i c e s of ( A ) . We have established a contradiction.
6 ) System (A) h a s a multiple focus and a limit cycle of multiplicity 2
in W , and
7 ) System ( A ) h a s a multiple focus and a saddle-to-saddle s e p a r a t r i x
in W .
The proof which r u l e s out c a s e s 6 and 7 is the s a m e . Let O ( O . 0 ) be
the multiple focus. The modified s y s t e m is taken in the f o r m
The point O ( 0 , O ) is a multiple focus of (A) also, and (A) is t h e r e f o r e

s t r u c t u r a l l y unstable. F u r t h e r m o r e , everywhere (with the exception of
the point O), the vector field of (A) is obtained f r o m the vector field of
(A) by a rotation in t h e s a m e s e n s e (through an angle equal t o tan-'p (.'+y')).
T h e r e f o r e , c a s e 6 leads t o the s a m e contradiction a s case 4, using
T h e o r e m 71 and R e m a r k 2 t o T h e o r e m 72. In c a s e 7, the contradiction
is established as in c a s e 5, using R e m a r k 2 t o Theorem 16 (5 11.2).
8 ) System (A) h a s two cycles of multiplicity 2 in W .
Let L, and L z be the two cycles, U = U (Lz) an a r b i t r a r i l y s m a l l neighbor-
hood of & . Using the t h e o r e m of the creation of a closed path f r o m a
multiple limit cycle ($27.1, Theorem 42) and employing the s a m e con-
struction a s in the proof of L e m m a 2, s 15.2, we obtain a s y s t e m (AI)
of c l a s s N > r a s close as d e s i r e d to (A) t o rank r > 3 , which coincides
with (A) outside the neighborhood U and h a s two closed paths, which
a r e s t r u c t u r a l l y stable limit cycles, inside U . F u r t h e r m o r e , using
L e m m a 9 of t h i s section, we conclude that t h e r e exists an analytical
s y s t e m (A) a s c l o s e a s desired t o (AI) t o rank r , which h a s a limit cycle
of multiplicity 2 in any a r b i t r a r i l y s m a l l neighborhood of the cycle L,
and is th_erefore s t r u c t u r a l l y unstable. If ( A I ) and (A) a r e sufficiently
close, (A) willhave at l e a s t two closed paths in U (L2). The contradiction
is established a s in c a s e 4.
9) System ( A ) h a s a limit cycle Lt of multiplicity 2 and a saddle-to-
saddle s e p a r a t r i x Lz in W ( F i g u r e 156a).
Let the s e p a r a t r i x L z extend f r o m saddle point 0,t o saddle point 02.
We choose a sufficiently s m a l l neighborhood U = U (Lz) of the s e p a r a t r i x
LZ which contains no equilibrium s t a t e s except 0,and 02. Modifying the
s y s t e m ( A ) by a rotation of the vector field through a s m a l l constant
angle and employing the s a m e construction as for the proof of L e m m a 2,
15.2, we obtain a s y s t e m (A,) of c l a s s N , r a r b i t r a r i l y c l o s e t o rank
r > 3 t o (A) which
(a) coincides with (A) outside U ;
(b) h a s no equilibrium s t a t e s in U , except the saddle points 0,and 0,;
(c) h a s no saddle-to-saddle s e p a r a t r i c e s in U ( F i g u r e 156b).
360
E 31. SYSTEVS OF THE FIRST DEGREE OF STRL'CTL'RAL ISST.-\BILITY
FIGL'RE 156
F v Lemma 9 of t h e p r e s e n t s e c t i o n , t h e r e e x i s t s an a n a l y t i c a l s y s t e m
( A ) as close as d e s i r e d to ( A I ) t o r a n k r w h i c h h a s a limit c y c l e of m u l t i p l i c i t y
2 i n a n y a r b i t r a r i l y s r n a l i n e i g h b o r h o o d of t h e c y c l e L I , a n d is t h e r e f o r e
s x r u c t u r a l l y u n s t a b l e . B u t ( A l ) is s t r u c t u r a l l y s t a b l e i n U . Therefore,
if (x) is s u f f i c i e n t l y ci.ose t o ( A l ) , c o n d i t i o n s ( b j a n d ( c ) a r e s a t i s f i e d for
(.xi. We h a v e t h u s e s t a b l i s h e d t h e e x i s t e n c e of a s t r u c t u r a l l y u n s t a b l e
s y s t e m (AI a r b i t r a r i l y close t o ( A ) w i t h o u t a n y s a d d l e - t o - s a d d l e s e p a r a t r i c e s
i n U . T h i s l e a d s t o t h e same c o n t r a d i c t i o n a s i n case 5.
IO) S y s t e m ( A ) h a s t w o s a d d l e - t o - s a d d l e s e p a r a t r i c e s L1 a n d L2 i n W .
T w o d i f f e r e n t case:; are p o s s i b l e :
( a ) a t l e a s t o n e of t h e a e p a r a t r i c e s Ll a n d L Z e x t e n d s b e t w e e n t w o
different s a d d l e points;
( b } each s e p a r a t r i r f o r m s a l o o p for i t s s a d d l e p o i n t .
L e t us f i r s t c o n s i d e r case ( a ) . L e t Ll b e a s e p a r a t r i x e x t e n d i n g f r o m
s a d d l e point 0 t o a n o t h e r s a d d l e p o i n t 01, w h i c h d o e s n o t c o i n c i d e with 0 .
T h e s e p a r a t r i x LZ e i t h e r p a s s e s a t a f i n i t e d i s t a n c e f r o m L l o r at l e a s t
o n e of t h e t w o e q u i l i b r i u m s t a t e s of t h i s s e p a r a t r i x c o i n c i d e s w i t h 0 or O1.
S i n c e n o n e of t h e p o i n t s of the s e p a r a t r i x Li i s a l i m i t p o i n t for L 2 , t h e
s i m p l e c l o s e d c u r v e of class k ( w h e r e k is a p r i o r i known to b e e q u a l to 4)
w h o s e e x i s t e n c e is e s t a b l i s h e d i n L e m m a 11 m a y b e c h o s e n s o t h a t i t
e n c l o s e s Ll without e n c l o s i n g L z .
In case ( b ) , s e v e r a l s u b c a s e s s h o u l d b e c o n s i d e r e d . S p e c i f i c a l l y ,
t h e d i s t a n c e b e t w e e n t h e l o o p s f o r m e d b y t h e s e p a r a t r i c e s L, a n d L z m a y
b e e i t h e r p o s i t i v e or z e r o , o n e of t h e t w o l o o p s m a y e n c l o s e t h e o t h e r
l o o p o r t h e y m a y lie o n e o u t s i d e t h e o t h e r ( F i g u r e s 1 5 7 a n d 158).
FIGURE 157 FIGURE 15s
36 1
362
S '?I. S Y S T E L 5 C F THE FIRST DEGREE OF STRL'CTI RAL ISST.IBIL[TY
To fix i d e a s , s u p p o s e t h a t t h e s e p a r a t r i s Lz f o r m s a p o s i t i v e a n g l e
with t h e a r c without c o n t a c t 1 2 . T h e n , b y ( 6 7 ) a n d t h e lemma of B 11.1,
ci<c; if E, = 0, p > O ;
r+;>ro" i f h > O , p - 0 .
( T h e d a s h e d c u r v e s i n F i g u r e 150 are t h e s e p a r a t r i c e s L;. L;, L;, L; for X =O,p'O.)

r ; m d ri a r e f u n c t i o n s o f t h e p a r a m e t e r s .l a n d p: r; = T: O., p), r; = r; (),+ p),
a n d by r e m a r k t o Lemma 3, S9.2, t h e s e f u n c t i o n s are c o n t i n u o u s . W e
c h o o s e a s u f f i c i e n t l y s m a l l f i x e d A I : 0 . By ( 6 8 ) , 1; O.,.0 ) > r; (iI,
0), a n d
h v t h e c o n t i n u i t y of ri m d r i , w e h a v e f o r a s u f f i c i e n t l y s m a l l p > 0
F v ( 6 8 ) , u i ( 0 , b)<z(O, p) when F>O is s u f f i c i e n t l y small. But then, f o r a

s u f f i c i e n t l y s m a l l h: > 0 ,
1; (X2. ji, < r; 0.2, p). (701
F y ( 6 0 ) a n d ( 7 0 ) t h e r e e x i s t s x > O , suchthat^.^(^, that^.^(^, ;)=I.;@. E).

T h e s e p a r a t r i c e s L; ar.d Li of (-41;) t h e r e f o r e c o i n c i d e , f o r m i n g a s i n g l e
s a d d l e - t o - s a d d l e s e p a r a t r i s . T h e s y s t e m ( A ~ p is ) thus structurally
unstable. T h e n u m b e r s x and F c a n b e t a k e n as small a s w e d e s i r e , i n
p a r t i c u l a r , t h e y c a n be m a d e s u f f i c i e n t l y small for (Ax;) to have no
s a d d l e - to- s a d d l e s e p a l - a t r i c e s i n a s u f f i c i e n t l y s m a l l n e i g h b o r h o o d of
t h e s e p a r a t r i s LI .
F u r t h e r m o r e , i f ( A ) i s a s y s t e m of class .V2,3, Lemmas 11 a n d 1 2
i n d i c a t e t h a t y c a n b e c h o s e n as a c u r v e of class .V + 1 a n d L e m m a 13
s h o w s That UJ (2.'~) c a n ' D e c h o s e n a s a f u n c t i o n of class A'. T h e n (Ax;;)
i s also a s y s t e m of c l a s s :V.
We h a v e t h u s e s t a b l i s h e d t h a t i f ( A ) i s a s y s t e m of t h e f i r s t d e g r e e
of s t r u c t u r a l i n s t a b i l i t y of class :V>3 w i t h t w o s a d d l e - t o - s a d d l e s e p a r a t r i c e s ,
t h e r e e s i s t s a s t r u c t u r a l l y u n s t a b l e s y s t e m ( A X E )of t h e same class, as
c l o s e a s d e s i r e d t o (A). w h i c h h a s n o s a d d l e - t o - s a d d l e s e p a r a t r i c e s i n a
s u f f i c i e n t l y s m a l l neig'nborhood of t h e s e p a r a t r i s L , . T h i s , h o w e v e r ,
c o n t r a d i c t s t h e d e f i n i t i o n of a s y s t e m of t h e f i r s t d e g r e e of s t r u c t u r a l
i n s t a b i l i t y . T h e r e f o r e , case 10 i s r u l e d out f o r a s y s t e m of t h e first
d e g r e e of s t r u c t u r a l i n s t a b i l i t y i n r e l a t i o n to t h e s p a c e R$), 3 < r g N .
For s y s t e m s of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y i n r e l a t i o n
to t h e s p a c e Rg' ( t h e s p a c e of a n a l y t i c a l s y s t e m w i t h c l o s e n e s s to r a n k r ) ,
L e m m a 11 also m u s t b e a p p l i e d t o p r o v e i n a p p l i c a b i l i t y of case 10. L e t
(-4) b e a n anal-y-tical s y s t e m of t h e first d e g r e e of s t r u c t u r a l i n s t a b i l i t y
w h i c h h a s t w o s a d d l e - to- s a d d l e s e p a r a t r i c e s Li a n d Lz . Let further
r>, 3 b e a n a t u r a l n u m b e r , a n d 6 a n d E p o s i t i v e n u m b e r s , E b e i n g s u f -
f i c i e n t l y s m a l l . A s b e f o r e , w e c a n c o n s t r u c t a s y s t e m ( A x ) of class
A'>r w h i c h is 6 / 2 - c l o s e t o r a n k r t o ( A ) a n d w h i c h h a s a s a d d l e - t o -
s a d d l e s e p a r a t r i x i n U ,( L z ) a n d n o s u c h s e p a r a t r i x i n U, ( L , ) .
L e t 6,> 0 b e a n a r b i t r a r y n u m b e r (612. B y L e m m a 14, t h e r e
e x i s t s a n a n a l y t i c a l s y s t e m ( A ) b,-close to r a n k r to (Ax;;) _which h a s a
s a d d l e - t o - s a d d l e s e p a r a t r i s . If 61 is s u f f i c i e n t l y small, ( A ) e v i d e n t l y
h a s n o s a d d l e - t o - s a d d l e s e p a r a t r i c e s i n U, (LI) and w e a r r i v e , as b e f o r e ,
363
at a contradiction with the definition of a s y s t e m of the f i r s t degree of

s t r u c t u r a l instability. C a s e 10 is thus ruled out f o r s y s t e m s of the f i r s t
d e g r e e of instability in relation t o the space I?:). T h i s completes the
proof of T h e o r e m 61.
3
6 . T h e p r o p e r t i e s of the s e p a r a t r i c e s of a saddle-node
in s y s t e m s of the f i r s t d e g r e e of s t r u c t u r a l instability
Let ( A ) b e a s y s t e m of the first d e g r e e of s t r u c t u r a l instability in W

which h a s a saddle-node 0 (0,O)in t h i s region. Consider s o m e s e p a r a t r i x
L of t h i s saddle-node. T o fix ideas, let t h i s be an a - s e p a r a t r i x , i. e., a
s e p a r a t r i x which goes t o 0 f o r t+--03. By T h e o r e m 61, the saddle-node
0 is the only s t r u c t u r a l l y unstable path of (A) in W , and t h e r e f o r e a s t
i n c r e a s e s , one of the following c a s e s is a p r i o r i possible for the
s e p a r a t r i x L:
1) L will leave W .
2 ) L will go t o a s t r u c t u r a l l y s t a b l e node o r focus, o r t o a s t r u c t u r a l l y
stable limit cycle.
3) L will go t o the equilibrium state 0 without being i t s o s e p a r a t r i x -
(i. e., the positive semipath making up L is one of the i n t e r i o r semipaths
of the parabolic s e c t o r of the saddle-node 0 ) .
4 ) L will be the o - s e p a r a t r i x of the equilibrium s t a t e 0 .
5 ) L will go t o a s t r u c t u r a l l y stable saddle point.
W e will prove that cases 4 and 5 a r e unfeasible.
F i r s t we w i l l formulate, without proof, two l e m m a s that will be useful
l a t e r on.
L e m m a 15. Let the separatrix L of the saddle-node 0 of system (A)
cross an a r c without contact 1 at the i n t e r i m point M oof the a r c . F o r
every e > 0 , there exists 6 > 0 such that i f ( A ) i s 8-close to (A) and 0 i s
a saddle-node of (A), then
(a) there exists a single separatrix of the saddle-node 0 of (A) which
c r o s s e s the a r c 1 at the point a. contained in u, (M&
(bj if the points M o and 8,on the separatrices L and E correspond to
the same time t = tor the points of the segments MOO and L O O of the
separatrices L and corresponding to the same times t are distant
l e s s than e f r o m each other.
The proof of L e m m a 15 is analogous t o the proof of L e m m a 3, S9.2
and the r e m a r k following the l e m m a .
Let again L be a n a - s e p a r a t r i x of t h e saddle-node 0 of ( A ) , M o any
point of the s e p a r a t r i x , 1 an arc without contact through M o which h a s
no common points, except M o , with the s e p a r a t r i x L and with the other
s e p a r a t r i c e s of 0 . The a r c 1 is defined by
the p a r a m e t r i c equations
x=f(u), Y=g(u)
and the point M corresponds t o the value uo

of the p a r a m e t e r .
FIGURE 160 Moreover, let the positive direction on 1
correspond t o the increasing p a r a m e t e r and
l e t the paths of (A) make positive angles with the arc 1 ( F i g u r e 160).
364
5 31. SYSTEhlS OF THE F[RST DEGREE OF STRLrCTL'RAL INSTABILITY
We w i l l consider modified s y s t e m s (A) of the f o r m
which a r e obtained f r o m ( A ) by an appropriate rotation of the vector fie_ld.

L e wt m a 16. I f p + 0 is sufficiently small, 0 i s a - s a l e - p o i n t of (A,)
and the arc without cm:tact 1 has precisely one point M o in common with
one of the separatrices of the saddle-node 0 of (A,) and has no common
points with other separatrices of 0. If the point .IToon the arc 1 c m -
responds to the value ;io = u 0 (p) of the parameter U , then lo (p)-+ uOfor
p -+ 0. =-
If p > 0 ( p < O), then Lo(p) u o (Zo (p) < u,,). A similar proposition,
ivith suitably modified wording, applies tchen L is an w-separatrix of
the saddle-node 0 ,
T h e proof of L e m m a 16 is analogous t o the proof of the corresponding
l e m m a for the separatl-ix of a saddle point ( $ l l . l ) , and it is therefore
omitted.
T h e o r e m 62. If ( A ) is a system of the f i r s t degree of structural
instability in if', it camot have in this region a separatyix which goes
froni a saddle-node to a saddle point.
P r o o f . Suppose that the theorem is not t r u e , i.e., s y s t e m ( A ) h a s
in IL' a s e p a r a t r i x L of the saddle-node 0 which at the s a m e t i m e i s a
s e p a r a t r i x of the saddle point 0,
( F i g u r e 161). Let 1 be a n a r c without
contact passing through the point .Wo
$++
L
,l of the s e p a r a t r i x L . Consider a
s y s t e m (A,,). Since 0 i s a saddle-
f
node of (A,), (A,) is s t r u c t u r a l l y
unstable. L e m m a 16 and the c o r -
4 responding proposition f o r a saddle
point (9 11.1) then show that f o r a
FIGURE 161
sufficiently s m a l l p # 0 ( A d d o e s
not have in a sufficiently s m a l l neigh-
borhood of the path L a s e p a r a t r i x
which g o e s f r o m 0 t o 0,. T h e contradiction is now established by the
usual argument using t i e relation ( U ( L ) , A );(?,A,,). T h i s p r o v e s the
theorem.
T h e o y e ni 63. I f ( A ) is a dynamic system of the f i r s t degree of
structural instability in i f - , it cannot have in this r e g i a a path L which
is at the same time an a-separatrix and an a-separatrix of a saddle
point of ( A ) .
P r o o f of T h e o r e m 6 3 is analogous t o the proof of the previous
t h e o r e m . It is conducted by reductio ad absurdum, using a s y s t e m ( A J ,
L e m m a 16, and the fact that a path forming a loop may c r o s s a segment
without contact a t most in one point.
T h e o r e m s 62 and 63 show that c a s e s 4 and 5 listed at the beginning
of t h i s subsection are unfeasible.
Let u s now r e t u r n tcb closed paths of a s y s t e m of the f i r s t d e g r e e of
s t r u c t u r a l instability and d e r i v e a f u r t h e r property of t h e s e paths on the
b a s i s of the above theol-ems.
365
Ch.XlI. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPAKATRIX
T h e o r e m 6 4 . System (A) of the f i r s t degree of structural instability

in W may have only a finite number of closed paths in this region.*
P r o o f of T h e o r e m 64 is analogous to the proof of the corresponding
proposition for s t r u c t u r a l l y stable s y s t e m s (Theorem 21, 9 16.1). Suppose
that ( A ) h a s an infinity of closed paths in W . Consider a sequence Ll,
Lz,Ls, ... of these paths and choose a single point on each. Let M1,
M 2 , M B ,... b e the sequence of t h e s e points, M i E Li . Since W i s a compact
region, we may a s s u m e without l o s s of generality that the sequence M i is
convergent. Let t h i s sequence converge t o s o m e point M*. W e w i l l now
show that no such point may exist.
Two c a s e s a r e possible a priori:
1) M* is an equilibrium s t a t e .
2 ) M* is not an equilibrium state.
In c a s e 1, M * is neither a node nor a focus (whether simple o r multiple),
since a sufficiently s m a l l neighborhood of a node o r a focus may not contain
points of closed paths. Hence, M * i s either a saddle point o r a saddle-node.
But then an infinity of points Mi belong t o one of the hyperbolic s e c t o r s of
the equilibrium s t a t e 0 , and w e can show that t h e r e e x i s t s a sequence of
points which belong t o the closed paths Li and which have a condensation
point which is not a n equilibrium state (see 9; 16.1, proof of T h e o r e m 21,
c a s e 2 ) . Thus, c a s e 1 is reduced to c a s e 2 . Let u s consider the second
case.
Let L* be the path of ( A ) through M * . L* c l e a r l y cannot leave the region
W , nor' can it be a path which goes t o a node, a focus (whether. simple o r
multiple), o r a limit cycle (otherwise, closed paths Ll would p a s s a r b i t r a r i l y
c l o s e to the node, the focus, o r the limit cycle, which is impossible).
Similarly, L* m a y not be an i n t e r i o r path of the parabolic s e c t o r of a saddle-
node. Finally, by T h e o r e m s 6 2 and 63, L* may not be a s e p a r a t r i x of a
saddle-node, and by Theorem 58, L*may not be a closed path. W e are
thus l e f t with one l a s t possibility, namely that L* is a path which for
t+ - 00 goes t o the saddle point O1 and f o r t+ + 00 goes t o the saddle
point 0,.
Let u s f i r s t a s s u m e that the points 0,and 0 2 coincide, i. e., L* is a
saddle-point s e p a r a t r i x which f o r m s a loop.
Let (zO.yo) be the coordinates of the saddle point Oj. By T h e o r e m 60, in
t h i s c a s e u (zo, yo) = P ; (q, yo)+ Qb (q. yo) # 0. But then by Theorem 44,
s29.1, the loop L* is either stable o r unstable,
i . e., a sufficiently s m a l l neighborhood of the
+&f
loop may not contain points of closed paths.
We have thus established a contradiction.
Let now Ol and 0,be two different saddle
points, L* going t o Ol for t -+ - 00 and to 0,for
FIGURE 162
t+ + m ( F i g u r e 162). Consider that hyperbolic
s e c t o r of the saddle point 0,which contains an
infinity of points MI. Let L** be the o -continuation
of the s e p a r a t r i x L* on the side of t h i s hyperbolic s e c t o r . Clearly, L**
is analogous to the path L* in all r e s p e c t s , i. e., i t is a saddle-to-saddle
s e p a r a t r i x . But then (A) h a s two s t r u c t u r a l l y unstable paths L* and L**,
and t h i s contradicts Theorem 61. T h i s completes the proof of the
theorem.
' Theorem 64 is not a direct consequence of Theorem 58 (which states that the closed paths of a system of
the first degree of structural instability are isolated.
366
5 31. SYSTEhiS OF THE FIRST DEGREE OF STRL'CTLIRAL INSTABILITY
7. P r o p e r t i e s of s e p a r a t r i c e s of s a d d l e p o i n t s
of systems of the f i r s t d e g r e e of
s t r u c t u r a l instability
We h a v e a l r e a d y d e r : v e d some p r o p e r t i e s of s e p a r a t r i c e s of s a d d l e
p o i n t s of s y s t e m s of t h e first d e g r e e of s t r u c t u r a l i n s t a b i l i t y . I n d e e d ,
w e h a v e s e e n t h a t , first, t h e s e p a r a t r i x of a s a d d l e p o i n t 0 ( x o , yo) m a y
not form a l o o p if u (zo,y3)= 0 (Theorem 60) a n d , s e c o n d , a n w - s e p a r a t r i x
( a - s e p a r a t r i x ) of a s a d d l e p o i n t m a y n o t be a t t h e same time an a - s e p a r a t r i x
( o - s e p a r a t r i x ) of a s a d c l e - n o d e (Theorem 6 2 ) . In t h i s s u b s e c t i o n , w e w i l l
e s t a b l i s h t w o f u r t h e r p r o p e r t i e s of s a d d l e - p o i n t s e p a r a t r i c e s of s y s t e m s of
t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y .
T h e o r e m 65. A system of the f i r s t degree of structural instabi2ity
in IV may not have in this region two saddle-point separatrices tchich go
to a limit cycle of multiplicity 2, me f m t -+ - 00 and the other fm
f+ t 03.
P r 00f . S u p p o s e t h a t s y s t e m ( A ) h a s a l i m i t c y c l e L o of m u l t i p l i c i t y 2
i n W' a n d t w o s a d d l e - p o i n t s e p a r a t r i c e s L , a n d L I , of w h i c h Ll goes t o L o
f o r t -+ +- co a n d L2 goes t o t h e same l i m i t c y c l e for t -+-a. O n e of t h e s e
s e p a r a t r i c e s e v i d e n t l y lies o u t s i d e Lo a n d t h e o t h e r lies i n s i d e L o .
L e t e o > O b e s u c h t h a t U,,(Lo) d o e s not c o n t a i n a n y e q u i l i b r i u m s t a t e s of
(A) a n d a n y c l o s e d p a t h s , e x c e p t L o .
An a r c w i t h o u t c o n t a c t 1 i s p a s s e d t h r o u g h some point 31 of t h e c y c l e L o .
L e t s b e t h e p a r h m e t e r o n t h e arc 1 , c h o s e n s o t h a t t h e point 31 c o r r e s p o n d s
t o s = 0. Let f u r t h e r
=
; f (s)
be t h e s u c c e s s i o n f u n c t i o n o n t h e a r c 1 d e f i n e d f o r all s, I s I < q, w h e r e q
is some p o s i t i v e n u m b e r . S i n c e t h e s e p a r a t r i c e s L, a n d L z b y a s s u m p -
t i o n g o to L O ( f o r t - + +- 00 a n d t - + - 0 0 ,
r e s p e c t i v e l y ) , t h e arc 1 c o n t a i n s a n
Z
i n f i n i t e n u m b e r of p o i n t s w h i c h b e l o n g
%'
to L2 a n d a n i n f i n i t e n u m b e r of p o i n t s
w h i c h b e l o n g to L l . Let .Vi (s:) a n d
-11;(si) b e t w o s u c c e s s i v e ( i n t e r m s oft)
p o i n t s a t w h i c h t h e p a t h L 1 c r o s s e s the
a r c 1 , a n d -11;(si) a n d -ri; (s;)two s u c c e s -
s i v e ( i n t e r m s o f t ) p o i n t s at w h i c h L2
crosses 1 . W e a s s u m e t h a t M ; a n d M i
a r e so close t o 111 t h a t t h e coils Jl&W;
a n d IMeM; of t h e p a t h s Ll a n d L z r r e s p e c -
t i v e l y , a r e c o n t a i n e d e n t i r e l y i n U,,( L o )
andIs;I<tl,ls;I<?, i = 1,2.
To fix i d e a s , l e t t h e p a r a m e t e r s
o n t h e a r c I b e c h o s e n so t h a t SA < 0, si > 0
( F i g u r e 163). T h e n
FIGURE 163 s; < s; <0 < s: <s;.

Also note that s; = f (si), s; = f (s;) .
A l o n g s i d e w i t h ( A ) , c o n s i d e r the s y s t e m
367
and let po> 0 be s o s m a l l that for all p, lpl<po, the following conditions
a r e satisfied:
1) 1 is an a r c without contact of (A,), and f o r a l l s , / s l < q , the succession
function ;=f,,(s) of t h i s s y s t e m is defined on the a r c .
2 ) U,(Lo)contains no equilibrium s t a t e s of (A,) and - either f o r all
p > 0 or f o r a l l p < 0 - it contains no closed paths of (A,) (the l a t t e r proposi-
tion is t r u e by T h e o r e m 71, 32.4).
3) T h e r e exist s e p a r a t r i c e s L,, and L , of (A,) which c r o s s the a r c 1 at
the points M;,, Mi,, M;,, M i , corresponding t o the values si,, si,, s&,, s;, of
the p a r a m e t e r s , such that
sb,<si, <s;,< s;,
and
si, = f, (sb,), s;, = f,
(sip).
4 ) The coils Afb,MiF and Mz,M;, of the paths Li,and L2, a r e contained
in (Lo)( F i g u r e 164).
FIGURE 164
Consider successive iterations of the function f,, i. e., the functions
and let
( k = 2,3, ... 1.
In virtue of o u r assumptions, i f p = 0, then for any k , 2 ,
Let Ci, and CZ,,be the simple closed c u r v e s formed respectively by the
c o i l s Mb,Mi, and M;,rMi, of the paths L,, and L Z pand the segments Mi,Mb, and
M;,M;,of the a r c 1 .
It is readily s e e n that in o u r c a s e the curve C,, is enclosed by C,,, and
i f po is sufficiently s m a l l , the region r,, between these two c u r v e s is
368
c o n t a i n e d i n L ' e o ( 0 ) . We a s s u m e t h a t t h i s c o n d i t i o n i s i n d e e d s a t i s f i e d . T h e
l i m i t cycle L o i s clearly e n c l o s e d b e t w e e n t h e c u r v e s C I Oa n d ( ' 2 0 .
P v c o n d i t i o n 2 , for all ci of a c e r t a i n s i g n (it' O < I ~ 1 -=cp0), C;,(L,)contains
n o c l o s e d p a t h s of (-A,,j. L e t t h i s b e s o Sur p o s i t i v e p . C n n s i d e r a p a t h L I P ,
r:;hcr.e o < p < p 8 , . As t i n c r e a s e s , t h e p a t h Lrll e n t e r s i n t o t h e r e g i o n r,,
between the c u r v e s and C'?, ( r , , c l : e o ( f . o )c) r, o s s i n g t h r o u g h t h e point N i b .
S i n c e CEO ( L , , ) c o n t a i n s n e i t h e r e q u i l i b r i u m s t a t e s n o r c l o s e d p a t h s of (Ap),
t h e p a t h L , , w i l l l e a v e L.e,,(L.o), a n d t h u s t h e r e g i o n ru, as t i n c r e a s e s f u r t h e r .
T h i ; m a y o c c u r o n l y i f L.,, c r o s s e s t h e s e g m e n t JI,JIi,. o f :he a r c without
r m t a c r 1 . C o n s e q l i e n t l y , t h e r e e s i s r s a natiAral n u m b e r A\
fS, (sip) 3 s;p. (72)

On t h e o t h e r h a n d , b>, ( ' i l t , w e h a v e for p = 0
/S&(Sl,) <s;p. (73)
3 i n e e si,, s:,, , an.! f.~, a r e c o n t i n u o u s f u n c t i o n s of p (bj- r e m a r k to L e n i n i a 3 ,

2 9 , ? ' ! , i n e q u a l i t i e s (721 and ( 7 3 ) p r o v e t h e e x i s t e n c e o f a n u m b e r p*, U ( p* < p,
.iu<-iit h a t
/S,* (s1),*) = s;,
i . e., s u c h That
SS&* = siw*.
*l'he l a s t c q u a l i t v shoir-s t h a t t h e s e p a r a t r i s L I P * O S (-A,,*) c o i n c i d e s \Kith

t h e s e p a r a t r i s Le,,*, i . e . , (Ap.) h a s a s a d d l e - t o - s a d d l e s e p a r a t r i x . T h e
n u m b e r p* m a y b e t a k e n a s s m a l l a s d e s i r e d . We h a v e t h u s e s t a b l i s h e d
rhat i f (.A) h a s a l i m i t c y z l e of m u l t i p l i c i t y 2 i n 11. to w h i c h o n e s e p a r a t r i s
+-
g o e s for t + - ~3 a n d t h e o t h e r s e p a r a t r i s f o r t + 00, t h e r e e x i s t s a s y s t e m
A,,") a s c l o s e as d e s i r e d to ( A ) %whichh a s a s a d d l e - t o - s a d d l e s e p a r a t r i s
i n W . (-1,')i s a s t r u c t u r a l l y u n s t a b l e s y s t e m . T h e r e f o r e , s i n c e ( A ) i s a
s y s t e m of t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y , we sho:ild h a v e
(H, A)&@, A&*).
T h i s r e l a t i o n i s c l e a r l v u n f e a s i b l e , s i n c e (Afi.) h a s a s a d d l e - t o - s a d d l e
s e p a t - a t r i x and (-4) h a s n'3 s u c h s e p a r a t r i c e s i n t h e n e i g h b o r h o o d of W .
LVe h a v e t h u s e s t a b l i s h e d a c o n t r a d i c t i o n , w h i c h p r o v e s t h e t h e o r e t n .
T h e o v e m 66. A system of the f i r s t degree of strrictiiral instability
in If' cannot have in this region a saddle-point separatrix which goes (for
t 4 - 00 or f o r t -++- 00) lo a saddle-point separatrix forming a loop.
P r n n f . Again s u p p o s e t h a t t h e t h e o r e m i s not t r u e , i . e . , s y s t e m ( A )
n f t h e f i r s t d e g r e e of s t r u c t u r a l i n s t a b i l i t y h a s i n Lf? a s e p a r a t r i s Lo of t h e
s a d d l e - p o i n t 0 (q, yo) which forms a l o o p a n d a
s e p a r a t r i x L i of t h e s a d d l e point 0, (z,. yl) w h i c h
goes t o t h e l o o p L o for t + C 00 ( t h e s a d d l e p o i n t s
(3 and 0,a r e c l e a r l y d i f f e r e n t ) . By T h e o r e m 60,
[J (zo,yo) # 0 f o r t h e s a d d l e point 0 , a n d b y
'Theorem 44, s29.1,
(J ( Z O ? Yo) <0. (74)
To fix i d e a s , s u p p o s e t h a t t w o s e p a r a t r i c e s of
FIGURE 165 t h e s a d d l e p o i n t 0 w h i c h a r e d i f f e r e n t f r o m Lo l i e
369
outside the loop formed by the s e p a r a t r i x L o , so that the s e p a r a t r i x Ll goes

t o the loop Lo f r o m the inside ( F i g u r e 165).
Let eo > 0 be such that U,, (Lo)contains no equilibrium s t a t e s , except 0 ,
and no closed paths of ( A ) .
Through a point P of the path Lo draw an a r c without contact I , and let s
b e the p a r a m e t e r on the a r c 1 chosen so that the point P corresponds to
s = 0 and the points of the a r c 1 lying outside the loop correspond to
positive values of s ( F i g u r e 1 6 6 ) . Since by assumption the loop IS stable,
w e c a n always find a point A of the a r c 1 which lies inside the loop such
that a l l the paths c r o s s i n g the segment A P of the a r c 1 go t o the loop Lo
and t h e r e f o r e c r o s s the segment A P at infinitely many points. In particular,
the s e p a r a t r i x Ll of the saddle point Ol c r o s s e s the segment A P of the a r c 1
at infinitely many points.
FIGURE 166
Let the value of s corresponding t o the point A be a . The succession

function
;= f (s)
is thus defined on the segment APof the a r c I , i . e., for a < s < O .
Moreover, f o r e v e r y integer N, t h e r e is an N-th successive point for
e v e r y point of the a r c A P ,
~ L=
V fN (s).
Consider two successive points M o and M 1 among the intersection points of
the s e p a r a t r i x Liwith the a r c A P . Let so and s, be the values of the p a r a -
m e t e r s corresponding t 3 these points, and to and ti the corresponding t i m e s
on the path L,. C l e a r l y to t l , and so <si. Moreover, l e t the point M obe
s o close to the loop Lo that the segment M a l of the path Lo is contained in
U,, (Lo)( F i g u r e 1 6 6 ) .
Consider the modified s y s t e m
dx
-dt
= P (2, Y) - PQ (ZVA), -$= Q Y)
(I, + PP (Z, V) t (Ad
which is obtained from (A) by a rotation of the vector field. For all suf-
ficiently s m a l l p # 0, (A,,) h a s no s e p a r a t r i c e s forming loops in U., ( L o ) ,
5438 370
S 31. SYSTELIS OF THE FIRST DEGREE OF STRUCTUR.4L INSTABILITY
bur i t h a s t w o d i f f e r e n t s e p a r a t r i c e s L; ( p ) a n d Li (p) c r o s s i n g t h e a r c 1 i n
t h i s neighborhood.
L e t Li (p) b e a n a - s e p a r a t r i x a n d Li (p) a n a - s e p a r a t r i x of t h e s a d d l e
point 0 .
L e t P' (p) and P" ( p ) b e t h e " f i r s t " i n t e r s e c t i o n p o i n t s of t h e s e p a r a t r i c e s
L, (p) a n d Li (p), r e s p e c t i v e l y , with t h e a r c I , so t h a t t h e s e g m e n t OP' (p)of
t h e p a t h L; (p) a n d t h e s e g m e n t OP"(p)of t h e p a t h L i ( p ) c o n t a i n s n o p o i n t s of
t h e a r c 1 , e x c e p t P' ( p ) a n d P" (p). L e t s ' ( p ) a n d S* ( p ) be t h e v a l u e s of t h e
p a r a m e t e r s c o r r e s p o n d i n g t o t h e p o i n t s P' (p)a n d P" (p). C l e a r l y ,
lims' (p) = lim s* (p)= 0. (75)
k-0 P-tO
L e t po > O b e s u f f i c i e n l y s m a l l . T h e n , i f I p I < po, w e h a v e s' (p) > O ,

S' ( p )> a , a n d t h e s e g m e : ? t s OP' (p) a n d OP" (p) of t h e s e p a r a t r i c e s L; (p) a n d
1,; (p) l i e i n Uco(Lo). If, n i o r e o v e r , p > 0, a n d eo i s s u f f i c i e n t l y s m a l l , w e
h a v e bv t h e l e m m a o f S11.1 s" (p)> 0 a n d s' ( p ) < 0. Ey T h e o r e m 1 9 a n d t h e
r e m a r k t o t h e t h e o r e m ( 5 29.3), (.A& h a s n o c l o s e d p a t h s i n UE0( L o ) . We
-+vi11a s s u m e i n w h a t f o l l o w s t h a t t h e s e v a r i o u s c o n d i t i o n s a r e s a t i s f i e d .
F v L e m m a 7, $ 2 9 . 2 , a s u c c e s s i o n f u n c t i o n s- f (s. p) i s d e f i n e d f o r
all I p I CI is a n a p p r o p r i a t e l y c h o s e n n u m b e r , o n t h e
s e g m e n t .4P' ( p ) of t h e a r c w i t h o u t c o n t a c t I , i . e., for all s, Q.::s < S' (p).
S i n c e Cc0( L o )c o n t a i n s no c l o s e d p a t h s , w e r e a d i l y c o n c l u d e t h a t f (s, p) > s .
It i s m o r e o v e r o b v i o u s t h a t
lim f (s, p) = sD (p)
SS
' (PI
s
a n d t h a t for p = 0 t h e futiction = f (s, p) r e d u c e s t o t h e s u c c e s s i o n f u n c t i o n
of t h e o r i g i n a l s y s t e m ( f (s,0) f (s)) d e f i n e d o n t h e s e g m e n t d P of t h e a r c I .
S i n c e for t h e o r i g i n a l s y s t e m , e v e r y p o i n t of t h e s e g m e n t -4P of t h e a r c I
h a s a n iV-th s u c c e s s o r f o r e v e r y i n t e g e r .V, t h e n for a n y g i v e n .V e v e r y
f i x e d p o i n t of t h e s e g m e n t A P of t h e a r c I w i l l h a v e a n N - t h s u c c e s s o r
for t h e s y s t e m A,, p r o v i d e d p is s u f f i c i e n t l y small. W e w i l l d e s i g n a t e t h i s
successor by
-
s.\- = f.r (s, p).
C l e a r l y , f X ( s , p) i s a c o n i i n u o u s f u n c t i o n of p ( f o r t h o s e v a l u e s of p f o r w h i c h
it i s d e f i n e d ) .
R v L e m m a 3, 3 9 . 2 , a n d t h e r e m a r k to t h i s lemma, w e c o n c l u d e t h a t if
p # O is s u f f i c i e n t l y s m a l l , (-43has a s e p a r a t r i x 51 (p) of t h e s a d d l e p o i n t O1
w h i c h crosses t h e a r c w i t h o u t c o n t a c t 1 a t t h e p o i n t s i%lo ( p ) a n d M , (p), w h e r e
-%Io ( p )-+ :Ifo a n d MI ( p ) -+ MI for p --t 0 . Let so (p) a n d si (p)b e t h e v a l u e s of
t h e p a r a m e t e r s c o r r e s p o n d i n g t o M 0 ( p ) a n d Mi (p). B y t h e l e m m a of 11 .I,
i f p 0, w e h a v e so ( p ) > s,, a n d s1 (p) > sI.
W e m o r e o v e r a s s u m e t h a t po is so s m a l l t h a t if I p 1 < po, t h e coil of t h e
p a t h LI (p)b e t w e e n t h e po:.nts -%Io(p) a n d Mi (p) is e n t i r e l y c o n t a i n e d i n U,, ( L a ) .
B y t h e r e m a r k t o Lemma 3 , 9.2, so (p),sl (p), s' (p), a n d s" (p) a r e c o n -
t i n u o u s f u n c t i o n s of p. -41~0n o t e t h a t t h e e q u i l i b r i u m s t a t e s of (A) a n d (AJ
c o i n c i d e , a n d t h e r e f o r e (A,) h a s no e q u i l i b r i u m s t a t e s , e x c e p t t h e s a d d l e
point 0 , i n o', (Lo).
It f o l l o w s f r o m the a b o v e t h a t if eo > O a n d pa >O a r e s u f f i c i e n t l y s m a l l ,
a n d Oe p Q po, t h e f o l l o w i n g i n e q u a l i t i e s a r e s a t i s f i e d
so @'I< si (p)<s' (I*) 0, s" (PI > 0
371
Ch.XII. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLt-NODE SEPAKATRIX I
and U,, (Lo)contains no closed paths of (A,,) and no equilibrium s t a t e s , except
the saddle point 0 .
Since M i(p) is the s u c c e s s o r of M o (p), we have
$1 (p)= f (so (p), P).
Moreover, it is readily s e e n f r o m the above that f o r a given integer N and

f o r a l l sufficiently s m a l l p , t h e r e e x i s t s an N-th s u c c e s s o r of the point M o (p),
i . e ., the function
-
SN (p)= f N (SO (p), p)
e x i s t s , and i t is a continuous function of p (naturally f o r a l l p f o r which
it e x i s t s ) .
Let u s now consider the behavior of the s e p a r a t r i x L1 (p)p r e s c r i b e d by
the above assumptions.
Two c a s e s a r e possible:
1) t h e r e exist a r b i t r a r i l y s m a l l p, I p I < po, such that f o r s o m e N
fN(SO(p)v p)Es(p),
i . e . , t h e r e exist a r b i t r a r i l y s m a l l p such that L1 (p) coincides with the

s e p a r a t r i x L; (p) ( F i g u r e 167);
2 ) t h e r e e x i s t s pl > 0 , p i < p 0 , such that f o r all I p [<pi, the s e p a r a t r i x
Li (p) does not coincide with L; (p).
FIGURE 167
Let u s f i r s t establish the behavior of the s e p a r a t r i x in c a s e 2 .

We will consider
( a ) a simple closed c u r v e C consisting of the coil of the path Li(p) between
the points M o (p) and M i (p) and the segment of the a r c I between the points
M o (p) and Mi (14:
(b) a simple closed curve c consisting of the segment P ( p ) O of the
s e p a r a t r i x L; (p), the point 0 , the segment OP (p) of the s e p a r a t r i x L; (p),
and the segment of the a r c t bet*rreen the points P (p) and P (p).
372
9 31. SYSTELLS OF THE FIRST DEGREE OF STRCCTCRAL U'STABILITY
L e t r b e t h e r i n g e n c l o s e d b e t w e e n t h e c u r v e s C' a n d C" ( F i g u r e 168).

It is r e a d i l y s e e n t h a t I' i s e n t i r e l y c o n t a i n e d i n Lye, ( L o ) ( s i n c e e a c h point of
t h i s r e g i o n b e l o n g s to a coil of a p a t h of s y s t e m (A+) e n t i r e l y c o n t a i n e d i n
I-,,, ( L o ) ) a n d t h e r e f o r e c o n t a i n s n o c l o s e d p a t h s a n d no e q u i l i b r i u m s t a t e s ,
e x c e p t t h e s a d d l e point ( 3 . T h e s e p a r a t r i x L , ( p ) e v i d e n t l y e n t e r s i n t o r
t h r o u g h t h e point -Ifl(u) its t i n c r e a s e s . S i n c e b y a s s u m p t i o n it d o e s not
c o i n c i d e w i t h t h e s e p a r z t r i s L; ([I) of t h e s a d d l e p o i n t 0 , f u r t h e r i n c r e a s e
of t w i l l i n e v i t a b l y c a u s e i t t o l e a v e .'I It m a v l e a v e r o n l y b y c r o s s i n g
t h e s e g m e n t P' (p) P " ( p ) c-f t h e a r c 1 .
FIGURE 168
L e t Q ([I) b e t h e i n t e r s e c t i o n point of t h e s e p a r a t r i s L l (p) w i t h t h e s e g m e n t

P' (11) P" (p)of t h e a r c 1 . 'The p o i n t Q ( p ) is c l e a r l y t h e S - t h s u c c e s s o r of
.Ifo (p), w h e r e .V i s some n a t u r a l n u m b e r d e p e n d e n t o n p .
T h u s , if w e w r i t e so (11) f o r t h e c o o r d i n a t e of t h e point Q (p) o n t h e a r c I ,
t h e n for a n y p* > 0, p* s' W ) . (76)
C h o o s e a f i x e d p* a n d t h e c o r r e s p o n d i n g X*.
For p = 0, f o r e v e r y . V , a n d i n p a r t i c u l a r f o r N = N * , t h e r e e x i s t s a
n e g a t i v e n u m b e r / X I (so (0), 0). P u t so (p), s' (p), s" (p), fw (sa (p), p) are c o n t i n u o u s
f u n c t i o n s of p . T h e r e f o r e , if p*+, 0 < p * * < p*, is s u f f i c i e n t l y s m a l l , t h e
n u m b e r f.v* (so (p**), p * ) is close t o f,v. (so (p*), p * ) , a n d t h e n u m b e r s' (p**) is
close to zero, so that
f s . ($0 (I***). p*) < s' (P**). (77)
373
I
Ch.X[I. CREATION OF LIhlI I CYCLE FKOhi LOOP OF SADDLE-NOUE S E P A R A T R I X
F r o m ( 7 6 ) and ( 7 7 ) , using the continuity of a l l the relevant functions, w e

conclude that t h e r e e x i s t s a number p, p**<;<p*, such that
i , e ., the s e p a r a t r i x L ,(F) of the saddle point Oicoincides with the s e p a r a t r i x

L; (F) of the saddle point 0 . This, however, contradicts assumption 2 .
W e have thus established that i f a s y s t e m of the f i r s t d e g r e e of s t r u c t u r a l in-
stability (A) contains in W a s e p a r a t r i x L o of the saddle point 0 which f o r m s a
loop and a s e p a r a t r i x I,, of the saddle point 0,which g o e s t o t h i s loop, t h e r e
e x i s t s a s y s t e m (A,) a s close a s d e s i r e d to (A) which
( a ) has a s e p a r a t r i x extending f r o m saddle point 0 t o saddle point 0,;
(b) does not have s e p a r a t r i c e s forming a loop and contained in U,,(Lo),
where E,, i s a sufficiently s m a l l positive number.
F y ( a ) , (A,) is a structurally unstable s y s t e m , and by ( b ) , the relation
(17, A,) 2 (11, A ) is impossible for the s y s t e m s (A) and (A,). But then, by
Definition 30 (31.1), ( A ) may not be a s y s t e m of the f i r s t d e g r e e of
s t s u c t u r a l instability. This contradiction proves the t h e o r e m .
8 . The fundamental t h e o r e m (the n e c e s s a r y and sufficient

conditions f o r s y s t e m s of the first degree of
s t r u c t u r a l instability)
Cnllecting all the previous r e s u l t s , w e s e e that a dynamic s y s t e m (A)

o f the f i r s t d e g r e e of s t r u c t u r a l instability in a closed region W s a t i s f i e s
I . ( A ) has in W one and only one s i m p l e s t structurally unstable path,
i . e ., a path of one of the following types:
1) a multiple focus of multiplicity 1;
2 ) a saddle-node of multiplicity 2 with uo = P i + Qu # 0;
3) a limit cycle of niultiplicity 2;
4) a s e p a r a t r i x from one saddle point t o another;
5) a s e p a r a t r i x forming a loop f o r a saddle point with CY,,# 0.
TI. ( A ) does not have in W any s t r u c t u r a l l y unstable limit cycles,
saddle- point s e p a r a t r i c e s forming a loop, or equilibrium s t a t e s other
than those listed in I.
111. If ( A ) has a saddle-node in W , none of the s e p a r a t r i c e s of this
saddle-node may go t o a saddle point and no two s e p a r a t r i c e s of the saddle-
node a r e a continuation of each other.
I P . The s e p a r a t r i x of a saddle point of (A) contained in W may not go
for t -f - 00 or f o r t -+ + 00 t o a s e p a r a t r i x forming a loop. W may not
contain two saddle-point s e p a r a t r i c e s going t o the s a m e limit cycle of
multiplicity 2, one for t - c - 00 and the other for t+- + 00.
Conditions I through I V prove t o be not only n e c e s s a r y but a l s o s u f -
ficient for (A) t o be a s y s t e m of the f i r s t degree of s t r u c t u r a l instability
in W , i . e . , w e have the following theorem:
T h e oYe?n 67. For a dynamic system (A)to be a system of the f i r s t
degree of structural instability in a closed region W , i t i s necessary and
sufficient that the above conditions I through IV be satisfied.
314
5 ?l. SYSTEhlS OF THE FIRST DEGREE OF STRL.CTL'R.AL IiiST.4YILITY
P r o n f . T h e necessity of conditions I through 11- follows directly from

T h e o r e m s 5 4 - 57, 5 9 - 63, 65, and 66. The proof of sufficiency is omitted
h e r e . Note, however, that although the proof of sufficiency is fairly lengthy,
t h p underlving idea i s simple and it follows the proof of T h e o r e m 2 3 (the
n c c e s s a r v and sufficient conditions of s t r u c t u r a l stability of s y s t e m s , 9 1 8 . 2 ): ' :
9. Bifurcations of s y s t e m s of the first d e g r e e

of s t r u c t u r a l instability
The above properties enable u s to identify without any further difficulties

all the possible bifclrcat.ioiis of a dvnaniic s y s t e m ( A ) i n a region Pi7 where
(.At is of rhe first d e g r e e of s t r u c t u r a l instability. These bifurcations
evidentlv depend on the particular simplest s t r u c t u r a l l y unstable path that
thc s v s t e m h a s .
Let u s consider the different c a s e s .
1 1 ( A ) h a s a m u l t i p l e f o c u s o f m u l t i p l i c i t y 1 i n W . Only
one bifurcation is possi'2le in this c a s e , namely the creation of a limit cycle
from the multiple focus. T h i s bifurcation t r a n s f o r m s the multiple focus
into a s i m p l e ( s t r u c t u r a l l y stable) focus and changes i t s stability, while the
c r e a t e d cvcle is s t r u c t u r a l l y stable and i t s stability i s identical t o the
stabilitv of the original focus.
2 ) ( A ) h a s a l i m i t c y c l e o f m u l t i p l i c i t y 2 i n T i ' . Bilurca-
tions of two types a r e possible in this case: the disappearance of the limit
cycle and decompositior. of the limit cycle into two limit c y c l e s . In the
l a t t e r c a s e , the two neu cycles a r e s t r u c t u r a l l y stable, one being stable
and t h e other unstable.
3) ( A ) h a s a s e p a r a t r i s f r o m o n e s a d d l e p o i n t t o a n o t h e r
s a d d 1 e p o i n t i n W. . In t h i s c a s e , the saddle-to-saddle s e p a r a t r i x may
decompose into two s e p a r a t r i c e s which a r e not a continuation of one another.
This is the only possible bifurcation of such s y s t e m s .
4) IA) h a s a s a d d l e - p o i n t s e p a r a t r i x f o r m i n g a l o o p
i n it' . Fifurcations occur only i f the s e p a r a t r i s loop d i s a p p e a r s on moving
to c l o s e s y s t e m s . Two different bifurcations a r e possible: e i t h e r a
s t r u c t h r a l l y stable limit cycle i s c r e a t e d in the neighborhood of the dis-
appearing s e p a r a t r i s loop (of the s a m e stability a s the loop), o r the loop
d i s a p p e a r s without c r e a t i n g a limit cycle.
5 ) ( A ) h a s a s a d d l e - n o d e Jfo w i t h c r o + O i n W . T w o s u b c a s e s
should be considered here:
5a) T h e s a d d l e - n o d e -Ilo has no s e p a r a t r i x forming a
loop.
Two bifurcations a r e possible: disappearance of the equilibrium s t a t e
Jl,or i t s decomposition into two s t r u c t u r a l l y stable equilibrium s t a t e s -
a saddle point and a node. In e i t h e r c a s e , no limit c y c l e s a r e c r e a t e d .
5b) T h e s a d d l e - n o d e Noh a s a s e p a r a t r i x f o r m i n g a
lonp.
.Again birfu rcations of' two types a r e possible:
1) Decomposition of the equilibrium s t a t e M o into a s t r u c t u r a l l y stable
saddle point and a s t r u c t u r a l l y stable node. No limit cycle is c r e a t e d .
' Dynamic syrcems of the first degree q f srructurdl initabrlicyon torus wereconsidered by Aranson (see /38/).
3 75
Ch.XI1. CREATION OF LIMIT CYCLE FROM LOOP OF SADDLE-NODE SEPAPATRIX
2 ) Disappearance of the equilibrium s t a t e M o . The s e p a r a t r i x loop

naturally d i s a p p e a r s , and a limit cycle is c r e a t e d in its neighborhood.
The above bifurcations cover all the possible simplest bifurcations,
i . e., bifurcations which may occur in s y s t e m s of the f i r s t degree of
s t r u c t u r a l instability. This conclusion follows f r o m the r e s u l t s of the
s
previous c h a p t e r s and f r o m 30.
The above l i s t shows that all the simplest bifurcations involving
creation o r disappearance of a limit cycle a r e p a r t i c u l a r c a s e s of bifur-
cations considered in Chapters IX, X, XI and in 9 30 of the present
chapter. Note, however, that t h e s e c h a p t e r s , and Chapter VIII, a r e
by no means r e s t r i c t e d t o the analysis of the simplest bifurcations.
376
C h a p t e r XIII
L I J I I T C Y C L E S OF SOIlIE DYNAJIIC SYSTE,VS

DEPENDING ON A P A R A \ I E T E R
IXTRODUCTIO?S
T h e subject of t h i s chapter a r e analytical dynamic s y s t e m s depending

on a p a r a m e t e r , i.e., s y s t e m s of the f o r m
where H and 0 a r e analytical functions. Ke w i l l investigate the topic of

limit cycles c r e a t e d f r o m zi closed path Lo of the "original" s y s t e m (Ad)
on passing from p = 0 tci c l o s e values of the p a r a m e t e r p.
The c h a p t e r is divided into two sections, $ 3 2 and 3 3 . T h e f i r s t t h r e e
subsections of 5 32 are of auxiliary c h a r a c t e r . In S 3 2 . 1 , w e investigate
the succession function on an a r c without contact c r o s s i n g a closed path L o
and d e r i v e a number of I'ormulas f o r t h e coefficients in the series expansion
of this function. In $ 3 2 . 2 a m o r e detailed statement of the problem is given.
$ 3 2 . 3 is devoted t o one c l a s s i c a l problem of the theory of analytical func-
tions, and i t is t h u s of independent i n t e r e s t . It c o n s i d e r s the equation
F (u,.z ) = 0, where F i s a n analytical function satisfying the condition
F ( O , O ) = 0, and the so-called N e w t o n ' s p o l y g o n is applied t o
investigate the number and the behavior of the solutions of this equation
i n the neighborhood of the point ui = 0, J = 0 .
T h e main r e s u l t s of this section a r e presented i n 9 3 2 . 1 , in the f o r m
of T h e o r e m s 71 and 7 2 . In t h e s e t h e o r e m s , s y s t e m (A,) is taken in the
form
-d =z P - p ' Q , -$=Q+pP,
dt
whose vector field is obtained f r o m the vector field of (XO)by a rotation

through a constant angle. It is proved that if Lo is a limit cycle of even
multiplicity of (Ao), rotation of the vector field in one direction decomposes
t h i s cycle into two strucyurally stable cycles, and rotation in the opposite
direction m a k e s the cycle disappear ( T h e o r e m 7 1 ) . If, however, Lo is a
limit cycle of odd multiplicity of (Ao), any s y s t e m obtained by a s m a l l
rotation of the vector field of (Ao) in any direction h a s a single limit cycle
in a s m a l l neighborhood of L o , which i s moreover s t r u c t u r a l l y stable
(Theorem 72).
T h e creation of a limit cycle f r o m a closed path of a conservative
s y s t e m is considered in s 3 3 . T h e main definitions a r e given in I 3 3 . 1 .
A conservative s y s t e m is defined as a s y s t e m which h a s a n i n t e g r a 1
377
Ch.XIII. LIMIT CYCLES OF SYSTEMS DEPENDING ON A PARAMETER
i n v a r i a n t w i t h p o s i t i v e d e n s i t y in the relevant region (see

833.1, Definitions 31 and 32).
However, 33 d e a l s only with one p a r t i c u l a r c a s e of conservative
s y s t e m s , namely s y s t e m s defined in a doubly connected ("ring") region
where all the paths a r e closed paths enclosing one another. In $33.1
it is proved that such a s y s t e m is indeed conservative ( T h e o r e m 74).
Dvnamic s y s t e m s of the f o r m
I
a r e known as H a m i 1t o n i a n s y s t e m s , and they constitute a particular
c a s e of conservative s y s t e m s .
In s33.2, we consider s y s t e m s which a r e c l o s e t o the l i n e a r conservative
system
x= -y 9 y=x, (Bo)
i . e ., s y s t e m s of the form
;= -Y +PPI ( x , Y) + P P Z (GY) + ... (Bu)

i =x+ PPI (G Y) +p'242 (Y) f
I, . . .,
and we establish under what coriditions the path x 2 +
yg = p: of the original
s y s t e m (e,) c r e a t e s a single limit cycle on passing t o a sufficiently c l o s e
s y s t e m (B,) (Theorem 75).
In 833.3, s i m i l a r conditions are derived f o r the g e n e r a l c a s e of s y s t e m s
which a r e c l o s e t o conservative s y s t e m s (see T h e o r e m 77). T h e o r e m 78
(s 33.4) shows that t h e s e conditions a r e particularly simple f o r s y s t e m s
close t o Hamiltonian.
Note that the condition of analyticity of the dynamic s y s t e m s introduced
i n Chapter XI11 is not essential: s i m i l a r r e s u l t s can be derived for non-
analytical s y s t e m s a l s o .
$32. THE BEHAVIOR O F LIMIT CYCLES O F SOME

DYNAMIC SYSTEMS FOLLOWING SMALL
CHANGES IN THE PARAMETER
1. T h e succession function i n t h e neighborhood

of a closed path
Consider a dynamic s y s t e m depending on a p a r a m e t e r p
where ( x , y, p) and p ( x , y, p) are analytical functions of x , y and of the p a r a -

m e t e r p, defined for I , y f r o m s o m e region G in the plane ( I , I/)and f o r p
from s o m e interval containing the point po. C l e a r l y (A,,) m a y be treated
as a o n e - p a r a m e t r i c family of analytical dynamic s y s t e m s defined i n G.
Suppose that for p = po, (A,,) h a s a closed path L o . Without loss of
generality, we may take po = 0, i.e., Lo is a closed path of the s y s t e m
378
5 2 % EFFECT OF S l I A L L CH.4NGES LK P.IR-Ih!ETOI
w h e r e q ( t ) a n d + ( t ) are p e r i o d i c f u n c t i o n s of p e r i o d T w h i c h c o n s t i t u t e t h e
s o l u t i o n c o r r e s p o n d i n g t o t h e p a t h Lo.
As i n 1 3 , we d e f i n e c u r v i l i n e a r c o o r d i n a t e s s, n i n t h e n e i g h b o r h o o d
of t h i s p a t h ,
-
z = ( P ( s . n). y = i ( s , n), (2)
w.here t h e f u n c t i o n s 6 a n d 3 h a v e t h e following p r o p e r t i e s :
I ) @ and 5 a r e defined i n t h e s t r i p
- 00 < s < 00, -n* < It <R*, (3)

w h e r e n* i s some p o s i t i v e n u m b e r , a n d t h e y a r e a n a l y t i c a l f u n c t i o n s i n
this strip.
2) - (p a n d $ a r e p e r i o d i c f u n c t i o n s of s with p e r i o d T.
3 ) 'p (s, 0 ) = (p (s). .1.(s. 0) E (s) . (4)
4) T h e f u n c t i o n a l d e t e r m i n a n t
d o e s not v a n i s h e v e r b y w h e r e i n t h e s t r i p ( 3 ) , i . e . , it r e t a i n s a c o n s t a n t
sign in this strip.
T h e f u n c t i o n 'p(s. n) a n d q ( s + n) m a y b e t a k e n , i n p a r t i c u l a r , i n t h e s a m e
f o r m as i n 13
G = 'p (s) +R$' (si, s= 9 (s) -R'p' (s), (6)
a s for s u f f i c i e n t l v s m a l l n* > 0 t h e s e f u n c t i o n s e v i d e n t l y s a t i s f y all t h e c o n d i -

t i o n s a b o v e . We s h a l l see i n t h e following, h o w e v e r , t h a t in some cases
f u n c t i o n s of a d i f f e r e n t f o r m a r e m o r e c o n v e n i e n t f o r a n d 5 (see 3 3 3 . 3 ,
(561). F r o m r e l a t i o n (4) we evidently h a v e
cp; (s. 0 ) = 'p' (s). 5.;(s. 0)= I#' (s). (7)
L e t u s c h a n g e o v e r t o t h e v a r i a b l e s s, n i n all t h e s y s t e m s (A,,) w i t h
s u f f i c i e n t l v s m a l l p . Differentiating ( 2 ) with r e s p e c t t o t and using t h e
e q u a t i o n s of (-Ip), w e fiqd
S o l v i n g t h e s e e q u a t i o n s for $, $ and eliminating t , we obtain the

d ifferential equation
379
F o r n = 0, p = 0, the denominator in the right-hand s i d e of (R,) is evidently

equal t o
(4.(I,(SI. 0 ) 4% (s, 0) -C(Cp (4,(I,(4.0) cp;l (s, 0) =
= 7 (s)$, (s, 0) - (I,' (s) CP; ( 8 , 0) = A (s, 0)
B y condition 4, A (s,0) # 0 f o r all s, in particular, f o r all 0 B: s < T. Therefore,
i f n* > O and p * > 0 a r e sufficiently s m a l l , the denominator in the right-hand
side of equation (R,) does not vanish for a l l s, n, p satisfying the respective
inequalities
O<s<r, Inl<n*, lpl<p*.
.,
i . e because of the periodicity in s of the functions $ and for - 00 < s < + oc,
In I < n*, 1 pl <p* . Hence it follows that R (s, n, p) is a n analytical function f o r
--oo<s<+w, [ n ( < n * , lpl<p*. Therefore, i t can be expanded in a series
i n powers of n, p in the neighborhood of any point in t h i s region, and the
coefficients of this series w i l l be analytical functions of s. It is readily
s e e n that
R (s,0, 0) 3 0. (8)
T h e r e f o r e , the expansion of R ( s , n, p) in the neighborhood of the point (so, 0,O)

(where so is any fixed number) h a s the f o r m
R (s, n, = Aio (s) n +AM (8) P + 40(4n2-t 41 (S) nP + AJZ(S)Pa+ .. (9)
Since R (s, n, p) is periodic in s, i t is readily s e e n that the coefficients

Ai,(s) a r e a l s o periodic functions of period T.
It follows f r o m (8) that the function n = 0 solves the equation (Ro). In
general, however, it does not solve the equation (RL)for p + 0 .
We will consider the succession function on the a r c without contact I ,
defined by the equation s = 0 (for functions % and 3 of the f o r m (6), t h i s
a r c is the n o r m a l to the path (1) at the point s = 0 ) . The succession
function is constructed as in Chapter V ($13.3). Let
n = f ( s ; 0, no. p) (10)
be the solution of equation (R,) satisfying the initial condition
f (0; 0, no, p.)= no. (11)
According to g e n e r a l t h e o r e m s , t h i s solution is defined i n the region

-Tl<s<7+Tl, lnol<A, Ipl<i. (12)
where q > 0, and i < n * and b & p * a r e sufficiently s m a l l positive numbers:

in t h i s region, the solution is an analytical function of its arguments.
Since ne= 0 is a solution of (Ro),we have
f (s; 0, 0, 0) =0
and the expansion of the function f in powers of no and p h a s the f o r m
f (s; 0 , no,~ ~ = a l o ~ ~ + a o , ~ + a ~ o n ~ + a l , n. .d r + ~ ~ B (13)
+.
H e r e the coefficients aij=a,j(s) are analytical functions of I in the interval
- q < s <7 t q .
380
I32. EFFECT OF ShlALL CH.-\NGES IN P.4RALIETER
I n s e r t i n g ( 9 ) a n d ( 1 3 ) f o r R ( s , n, p) a n d n i n e q u a t i o n (R,) a n d e q u a t i n g t h e
c o r r e s p o n d i n g c o e f f i c i e n t s in t h e r i g h t - a n d t h e l e f t - h a n d s i d e s , w e o b t a i n
t h e f o l l o w i n g r e c u r s i v e d i f f e r e n t i a l e q u a t i o n s for t h e f u n c t i o n a,, (s):
. . . . . . . . . . . . .
F y ( 1 1 ) a n d ( 1 3 ) , f o r i := 1, j = 0, w e h a v e aij ( 0 ) = 1, a n d o t h e r w i s e ,
aij (0) = 0.
T h e equalities
a10 (U, = 1. ai j (U) =- u (15)
c n n s t i t u t e t h e i n i t i a l c o n d i t i o n s for e q u a t i o n s ( 1 4 ) . T h e f u n c t i o n s a i j (s)
t h e r e f o r e c a n b e found I l y s u c c e s s i v e l y s o l v i n g e q u a t i o n s (14) w i t h b o u n d a r y
c o n d i t i o n s ( 15).
Note t h a t t h e f i n a l e l p r e s s i o n s for t h e f u n c t i o n s ai,,(s), w h i c h a r e t h e
c o e f f i c i e n t s of t h e t e r m s free f r o m 11 in e x p a n s i o n (13), a r e t h e s a m e
e x p r e s s i o n s a s i n C h a p t e r S (s26.1). I n d e e d , t h e s e f u n c t i o n s m a y b e found
b y s e t t i n g p = 0 . B u t t h i s l e a d s u s t o s y s t e m (Ao) a n d e q u a t i o n (Ro) c o n -
s i d e r e d i n C h a p t e r S.
L e t t h e s u c c e s s i o n f u n c t i o n of (A,,) o n t h e a r c without c o n t a c t I b e f (no,p).
B y d e f i n i t i o n , it i s o b t a i n e d f r o m t h e f u n c t i o n (IO) f o r S = T , i . e . ,
n = f (no. p) = f (T;0, n d , P). (16)
T h i s a n d (13) s h o w t h a t t h e s e r i e s e x p a n s i o n of t h e s u c c e s s i o n f u n c t i o n h a s
the form
-
f (no,~ ) = " ~ O ( T ) ~ O T ~ O I ( . T ) ~ ~ ~ ~ O ( T ) ~ : I I. X I ~ ( T ) ~ (~1~7 ~) .
Setting
ai, (T)= U i j ,
we obtain
n = f (no, p) = 40no f uo1p +u*on: iu11nop f .. . (19)
St" w i l l now find e x p r e s s i o n s f o r t h e c o e f f i c i e n t s u i o a n d u p , a n d i n d i c a t e

i n g e n e r a l o u t l i n e the s t r u c t u r e of t h e c o e f f i c i e n t u l l t w h o s e e x p l i c i t
e x p r e s s i o n is h i g h l y c o m p l e x . T h e s e c o e f f i c i e n t s w i l l b e n e e d e d i n w h a t
follows.
We have
Ch.XII1. LIMIT CYCLES OF SYSTEMS DEPENDING ON A PARAMETER
Expanding t h e functions g ( s , n, p) and h ( s , n, p) i n powers of n and p and

seeing that g ( s , 0, 0) = 0 and h ( s , 0, 0)# 0, we obtain
g (s, n, p) = g10n +g01p +&on* + . . . 1
h (s, n, P)= hoo + hlon +ho1p + + . ..

h20n8 I
(21)
w h e r e gij==gij(s) and h i j = h i j ( s ) are analytical functions of S, and h o o ( s ) # O .
gion + go+ + gzon' + . . . E (Aion + Aoip + &on' + + . . .) (hoo+ big+ h o i ~+ . . .)

- 4 1 1 ~
Equating the coefficients on the right- and left-hand s i d e s of t h i s identity

and solving for A i r , we obtain
Let u s d e t e r m i n e the functions gij(s) and h i j ( s ) entering (22). To t h i s

$nd, we expand the functions is and 0 in powers of p, and the functions and
9 i n powers of n. T h i s g i v e s
H ( x , Y, p) = p (IY) +PPl
?(GI/) + P*P2 (5,Y) + . . ., (23)
B ( x , Y , c ~ ) = Q ( x , Y ) + c ~ Q ~ ( ~ , Y ) +...,
~*Q~(z,~)+
and
:ye c a n v e r i f v d i r e c t l y that
( c o m p a r e e q u a t i o n s (26:- (28). S 13.3).

IJsing (221, ( 2 9 ) , ( 3 0 ) . (31), w e find
($1= P; (v (4.1:(SI)
-40 +Q I (q (4,$ (s)) - d
In hoo(s) =
d
= P ; ( ( P . ~ ) + Q ; ( F .$ ) - X l n A ( s , O ) , (32)
Irp (J). $! Wrp (S)--P1 (9(SI. rp (s)) $ (4 .

4 1 (4 = QI 0)
(33)
T h e l a s t e q u a t i o n s , t o g e t h e r with e q u a t i o n s (14), i n i t i a l c o n d i t i o n s (15j, a n d

r e l a t i o n s ( 1 8 i , e n a b l e u s t o find t h e c o e f f i c i e n t s ul0 a n d uoi of e x p a n s i o n (IO).
T h e e x p r e s s i o n for t h e Zoefficient ul0 i s i d e n t i c a l t o t h e e x p r e s s i o n for t h e
d e r i v a t i v e of t h e s u c c e s s i o n f u n c t i o n o b t a i n e d i n C h a p t e r V (s 13.3, ( 3 0 ) ) ,
indeed:
%
lU;(cpW. @(91W?&(~)~ +@))Ids
ulo == alo( 7 ) = eo (34)
I n t e g r a t i n g t h e s e c o n d e q u a t i o n i n ( I ? ) w i t h i n i t i a l c o n d i t i o n s (15) a n d
u s i n g ( 3 2 ) a n d ( 3 3 ) , w e iind
J[P;(T(:L *ls\l--Q;(T<s~. *(:iiI?.

a,,(s) -e
Ch.XII1. LIMIT CYCLES OF SYSTEMS DEPENDING ON A PARAMETER I
Since A (z, 0) = A ( 0 , 0 ) , w e obtain
uo1=a01 (4=
T
(P;tVp, V)+U;te. W)da - 1( P ; 4 & s
= -e0 !e 11 (Pi (939cp' (4 ---Pi (cp? $1
0)
Let u s now proceed with the determination of u , , . T h e corresponding

expression is much m o r e complex than either (34) o r (36). Integrating
the fourth equation in ( 1 4 ) with the initial condition a l , ( 0 ) = 0, w e obtain
ti I
SAio(s)dr -jAlo(s)dr
a,, ( s ) = e 0 e +
(2A10a,oaol A,,alo) ds.
Inserting the expression f o r A,, f r o m ( 2 2 ) , we obtain
uio
u,,.~ 5
0
( g t ~ - 4 t h o - ~ i o h o ~+ 2A20a oi
hoo
ds. (39)
By (29), the expression f o r gll h a s the f o r m
g1, = (&Bl+ d " Y J cp' (4 - (P;,P, +P;,Yl) 9'(4-I-Pip; -piy;, (40)
where the functions p , and q1 and t h e i r derivatives a r e evaluated a t the

point ( c p (s), J, (s)).Using the obvious relations
p ; , (cp (s), $ (s)) $1 (s) =

d p i y ) , *('))-p =; ('p (s) 7 rl,(4)
cp' (4
and
9)
dx(9%49 9' = d9l(T>
- d r P ; " (cp? $) 9' (4
we r e w r i t e the expression f o r gli in the f o r m
384
S 32. EFFECT OF SJI.-\LL CHAKGES IN PARAXIETER
where
E x p r e s s i o n s ( 3 3 ) f o l ~ d O ,(30)
, f o r h o i , and ( 3 5 ) f o r a,,,
show that if
= Q l (cp
PI (cp (s). $ (4) ($1, t (-9)= 1 ' . (14)
i , e . , if t h e functions p 1 (I. y) and qI (2,y ) vanish on a closed path L o , we have

f,, = 0 . However, the i n t e g r a l
''.! [Pix ((P (s), 3 (s)) - q i V ('F ($1. 3' (s))JA- (-151
need not vanish in t h i s case. Indeed, let F ( I . y) b e an anal-ytical function

satisfving t h e following conditions:
(a) F (cp (s). $ (s)) = v.
(b) F; (cp (s). t (s))' P&('F (4. $ (s))* f U
( t h e proof of the existence of t h i s function in s o m e n e ig h b o r h o o d of
the path Lo is conducted p r e c i s e l y in t h e s a m e way a s t h e proof of L e m m a 1,
9 1 5 . 1 ) . Then, i f
PI (2.Y) - F (I.
Y) F; (I.Y). Y1 12. Y) - F (I* Y) F ; (2,!I), (46)
t h e integral in ( 4 5 ) doe:$ not vanish.
2. Statement of t h e problem
We a r e i n t e r e s t e d i r t h e number of l i m i t c y c l e s of (A,,) located in a

sufficiently s m a l l neigtborhood of the path Lo for sufficiently small p # 0 ,
i . e . , the number of limit c y c l e s c r e a t e d f r o m a closed path L o on passing
f r o m p = 0 t o c l o s e values of p.
We define the function
d (no, P) f (-no. P)--no. (47)
O u r problem is c l e a r l y equivalent to t h e d e t e r m i n a t i o n of t h e number of

sufficiently s m a l l real r o o t s of t h e function d ( r t o , p ) for sufficiently small
p F 0. *
T h e c r e a t i o n of limi.: c y c l e s f r o m a closed path of (A") on p a s s i n g t o
modified s v s t e m s w a s considered in C h a p t e r S. However, i n C h a p t e r X
w e dealt with all t h e possible modified s y s t e m s sufficiently close t o (Ao),
i . e . , the t r e a t m e n t w a s c a r r i e d out in a sufficiently s m a l l neighborhood
of t h e point (-4,) in t h e s p a c e of all dynamic s y s t e m s R. For t h e case
w h e n t h e c l o s e d p a t h L o i s a closed k-tuple l i m i t c y c l e of ( A o ) , we e s t a b -
lished that the maximum number of limit c y c l e s c r e a t e d f r o m Lo on
passing to o t h e r systerrls of t h i s neighborhood is k ( T h e o r e m 1 2 , s 2 7 . 1 ) .
In the p r e s e n t section we will c o n s i d e r t h e c r e a t i o n of l i m i t c y c l e s
f r o m a closed path Lo of (Ao) i n a m o r e r e s t r i c t e d s e n s e . The modified
s v s t e m s a r e confined t o the s y s t e m s (A,,) c o r r e s p o n d i n g t o sufficiently
385
s m a l l values of the p a r a m e t e r p , and we no longer consider any possible

s y s t e m sufficiently c l o s e t o (Ao). In geometrical t e r m s , t h i s means that
we a r e no longer dealing with the e n t i r e neighborhood of the point (Ao) in
the space R , but only with s o m e c u r v e & i n t h i s neighborhood which p a s s e s
through the point (Ao). We will consider the creation ( o r disappearance)
of limit cycles in the neighborhood of the path LO of (Ao) f o r motion along
the curve 1,in the space R.
We will d e r i v e the s u f f i c i e n t conditions f o r creation ( o r disappearance)
of limit c y c l e s in a number of simplest c a s e s . T h e s e conditions evidently
coincide with the sufficient conditions of appearance o r disappearance of
sufficiently s m a l l real r o o t s of the equation
on passing f r o m p = 0 t o sufficiently s m a l l finite p .

F i r s t note that i f the closed path Lo is a simple limit cycle, i . e., when
(see (34)), t h e r e e x i s t s n* > 0 and p*>O which satisfy the following condition:
if I p I < p*, equation (48) h a s one and only one solution no = no (p), such that
I no id I < I".*.
T h i s follows f r o m the t h e o r e m of s t r u c t u r a l stability of a simple limit
cycle (see 14, r e m a r k t o T h e o r e m 18). T h i s also follows directly f r o m
the t h e o r e m of implicit functions (s 1.2, Theorem 3 and r e m a r k t o
T h e o r e m 4 ) . Indeed, we s e e f r o m (47), (19), and (49) that i f L o i s a
simple limit cycle of (Ao), then
d (0, 0) = 0, dbo(0, 0 ) # 0,
so that the t h e o r e m of implicit functions is applicable.

However, i f u j o = 1, we have
d (0, 0 )= 0, dho (0, 0)= 0
and the conditions of the t h e o r e m of implicit functions a r e not met.

Suppose that in t h i s c a s e t h e r e e x i s t s k, 2 such that
d (0, 0 ) = dko(0, 0) = . . . = d$l')

0
(0, 0), d($ (0, 0)# 0.
(50)
It is readily s e e n that we then c a n find n*>O and p*>O with the following
property: f o r all p, I pI <p* the equation
d (no, P) = 0
h a s at most k r e a l roots which a r e s m a l l e r than n* in absolute magnitude.
Indeed, let n* and p* b e such that if InoI<n*, l p l < p * , we have
Then, if I p I < p* and the function d (no,p ) h a s k + 1 r o o t s in the interval

+
(- n*, n*), its first derivative with r e s p e c t t o no h a s a t l e a s t k r o o t s
386
5 2 2 . EFFECT OF ShI.lLL Ctl.WGES Lh PARA!.LETER
i n t h i s i n t e r v a l , t h e s e c o n d d e r i v a t i v e h a s a t l e a s t X. - 1 r o o t s , e t c ., a n d
t h e k - t h d e r i v a t i v e h a s a t l e a s t o n e root, which c o n t r a d i c t s ( 5 1 ) . T h u s ,
for s m a l l n o a n d p , e q u a t i o n (18)c a n n o t h a v e more t h a n k real roots. T h e
q u e s t i o n of e x i s t e n c e r*f t h e s e r o o t s , if a n y , r e q u i r e s s p e c i a l a n a l y s i s ,
h o w e v e r . T h e a n a l y s i s is c o n v e n i e n t l y c a r r i e d o u t w i t h t h e a i d of
Y e w t o n s d i a g r a m or p o l y g o n , which i s d e s c r i b e d i n t h e next
section.
3. N e w t o n s p o l y g o n and s o l u t i o n of t h e e q u a t i o n F (u,.z ) = 0
A s i g n i f i c a n t point i n o u r a n a l y s i s of t h e roots of a f u n c t i o n is t h e
a s s u m p t i o n that t h e v a - i a b l e s are c o m p l e x - v a l u e d .
k5e w i l l t h e r e f o r e c o n s i d e r a n a n a l y t i c a l f u n c t i o n F (w,z ) , w h e r e u7
and. z a r e c o m p l e x v a r i a b l e s . T h e r e s u l t s w i l l t h e n b e a p p l i e d to t h e
c a s e of real u- a n d 2 .
Let F j O , 0 ) 0 . The e x p a n s i o n of F (m,z ) in p o w e r s of u a n d J a r o u n d
t h e point ( 0 , O ) h a s t h e form
We i n t r o d u c e o n e fur-ther a s s u m p t i o n r e g a r d i n g t h e f u n c t i o n F (u-,z ) ,
n a m e l v t h a t a t l e a s t one of t h e c o e f f i c i e n t s uil) a n d a t l e a s t o n e of t h e
c o e f f i c i e n t s u O jd o not v a n i s h , i . e . , t h e e x p a n s i o n ( 5 2 ) c o n t a i n s a t l e a s t
o n e t e r m w i t h o u t z a n d a t l e a s t o n e t e r m without IC. T h i s a s s u m p t i o n
does not c o n s t i t u t e a f u n d a m e n t a l r e s t r i c t i o n of o u r p r o b l e m ( t h e p r o b l e m
(i -
of s o l u t i o n of t h e e q u a t i o n F (u,,z ) = 0 for u,). I n d e e d , f o r all u , =~ 0
1,2, . . . ), t h e f u n c t i o n F (a*.z ) m a y be w r i t t e n i n the form
F (w,3) = zF,( w , 3)
a n d r h e p r o b l e m r e d u c e s t o t h e a n a l y s i s of t h e e q u a t i o n F , (u-.z ) = 0 . The
s i t u a t i o n is s i m i l a r w h e n all u o , = 0 ( j > 0).
T h e f o l l o w i n g a n a l v s i s i s b a s e d on a n u m b e r of t h e o r e m s from t h e t h e o r y
of ana1,ytical f u n c t i o n s , w h i c h are p a r t l y g i v e n h e r e w i t h o u t p r o o f .
T h e o r e tn 68 (the theorem of implicit functions). Let F (u,.z ) be an
analytical junction in the neighborhood of (0. 0 ) and let
F (0. 0 )= 0, F&( 0 , 0 ) + 0.
There exist 6 > 0 am! E > 0 such that f o r every 3 , I z I < 6, the eqiuztim
f iu.. z ) 11has one and only one root U = f ( 5 ) satisfying the inequality
I f ( z ) I < E . The junction f ( z ) can be expanded in positive integral powers
of z , and the series tcili be convergent for 1 z 1 < 6 , i . e., it will be a
single- valued analytical function of z which vanishes at 3 = 0 .
T h e o r e m 68 is the t h e o r e m of i m p l i c i t f u n c t i o n s i n t h e case of c o m p l e x
v a r i a b l e s . I t s p r o o f c a n b e found i n / 2 2 / , C h . IV, p . 354. T h e p r o o f
u s e s a m a j o r a n t series. Theorem 68 is also a d i r e c t c o n s e q u e n c e of
t h e following t h e o r e m .
387
T h e o r e m 69 (Weierstrass ' s preparatory theorem). Let F (w, Z) be

an analytical function in the neighborhood of (0, 0) satisfying the conditions
I
Then in some neighborhood 1 W I c E, I z I < 6 of (0, O), F (w, Z) may be represented
in the form
F (w, Z) = [wk + A , (2) W h - l + ... +Ah (Z)]0 (W,Z). (54)
where @(w, z ) is an analytical function which does not vanish in this

neighborhood, and A , (z). A2 (z), . . ., Ah (z) are analytical functions for I z I e 6.
The proof of Theorem 69 can be found in 1 2 2 1 , Ch.IV, p. 352. It
follows f r o m Theorem 69 that in a sufficiently s m a l l neighborhood of
the point ( 0 , 0 ) , the equation
F (w, Z) = 0 (55)
is equivalent t o the equation
W k - t A , (Z)W"-'+ . . . +Ah-, (Z)W + d h ( Z ) = 0, (56)
whose left-hand side is a polynomial in w . W e i e r s t r a s s ' s p r e p a r a t o r y

t h e o r e m thus reduces the local investigation of the g e n e r a l c a s e of a n
implicit function w ( 2 ) defined by equation (55) to the c a s e of an implicit
function defined by an algebraic equation in w (which, in general, is not
an equation in 2).
Note that relations (53) and (54) and the condition @ (w, z ) + 0 lead
directly t o the equalities
At (0)= 0, A2 (0)= 0 , . . ., dk ( 0 )- 0. (57)
T h e o r e m 70. Let F ( w , z) be an analytical function in the neighborhood

of W , O ) which satisfies conditions (53). There exist E >0 and 6 >0 such
that f o r every z , 1 ~ 1 ~ 6the
, equation
F(w. z ) = O
has precisely k roots (either different m coinciding) wit w 2 , . . ., wh, which
a r e smaller than E in absolute magnitude. Moreover, i f Z + O , each of the
roots w i also goes to z e r o .
P r o o f . T h e validity of T h e o r e m 70 follows directly f r o m the previous
t h e o r e m and f r o m the theorem of the continuous dependence of the roots of
a polynomial on i t s coefficients ( s e e 1291, 73). Indeed, T h e o r e m 69 shows
that the roots of the equation F (w, z) = 0 coincide with the r o o t s of equation
(56). By (57), equation (56) h a s k r o o t s a t z = 0, which a r e a l l z e r o . T h e r e -
fore, the roots of equation (56) go t o z e r o f o r z - t 0.
R e m a r k . The roots wl, w 2 , . . ., wh of equation (55) depend on z.
Suppose that f o r s o m e z = zo all t h e s e r o o t s are different. It is readily
s e e n that in t h i s c a s e the roots w,, w2, . . ., wIcan be defined in the neighbor-
hood of the point zo so that they a r e analytical functions of z in thisneighbor-
hood (to e n s u r e all this, it suffices t o r e q u i r e that the r o o t s w t (z) v a r y
continuously with the variation of z).
388
s32. EFFECT OF SI\I;\LL CHANGES IN PARAhlETER
In what follows, we will only c o n s i d e r the case when t h e r o o t s wt, w2. . . ., ( c ~

a r e different for e v e r y z # 0 i n s o m e neighborhood U of t h e point 0 ( t h i s
will b e s o i f t h e discrirninant D ( z ) of equation (56), being a n analytical func-
tion of z , d o e s not vanish identically). Then equation (55) naturally d e t e r -
mines n e a r e v e r p p o i n t J +
0 of t h i s neighborhood k anal-ytical functions
which are t h e single-valued b r a n c h e s of the implicit function w . However,

i n t h e n e i g h b o r h o o d C ' o f t h e p o i n t 0 , t h e s e functions, i n g e n e r a l ,
a r e not single-valued analytical functions, Indeed, it c a n b e shown that if
the point z m o v e s along s o m e J o r d a n c u r v e around the point 0, the value of
t h e function wz (2) in g e n e r a l will have changed when we r e t u r n t o the i n i t i a l
point z . T h e function ( 5 8 ) (see /35/, Ch.XIII, 1) forms o n e o r s e v e r a l non-
intersecting s y s t e m s with t h e following property: if t h e point z t r a v e l s once
along t h e c u r v e r around the point 0, the functions of e a c h of t h e s e s y s t e m s
undergo a cyclic permu-ation. T h e s e s y s t e m s a r e known a s the c y c 1i c
s v s t e m s o f s o l u t i o n s of equation (55), and t h e point O i s t h e b r a n c h -
i n g p o i n t of t h e function u' (J). C l e a r l y , f o r z -+ 0, all t h e functions
U" (J) +o.
If the d i s c r i m i n a n t D (J) = O , the situation is m o r e complicated. This
c a s e , however, will not b e r e q u i r e d i n t h e following, and it is not con-
sidered here.
We now have to c o n s i d e r t h e form of t h e solutions u', ( z ) o f ( 5 5 ) i n the
neighborhood of 0.
If k = 1, T h e o r e m 68 s h o w s that t h e r e e x i s t s a unique solution, which
m a y b e written as a ser:es i n i n t e g r a l p o w e r s of z , i . e., as a series
rr=yI:+yL.ze+y3z3+ ....
which c o n v e r g e s f o r sufficiently s m a l l z .
For k > 1, e a c h of t h e k solutions which e x i s t by T h e o r e m 70 and the
r e m a r k t o t h e t h e o r e m can no l o n g e r b e r e p r e s e n t e d a s a series i n i n t e g r a l
p o w e r s of i. To e s t a b l i s h the n a t u r a l form of t h e solutions i n t h i s c a s e ,
let u s f i r s t c o n s i d e r s o m e s i m p l e e x a m p l e s . T a k e equation (55) i n t h e f o r m
UoI iUZOU" f ~ 0 3 ~=' I).
w h e r e none of the coefficients vanish. Then
where 7
-3% and f q . 9 are t o b e r e g a r d e d as o n e (fixed) value of e a c h
u20
of t h e s e r o o t s . For s m a l l z, the second root c a n be s e r i e s - e x p a n d e d i n

i n t e g r a l p o w e r s of z . T h e solution w t h e r e f o r e h a s t h e form
w=T*,za''~+5Lzz3t+ ...,
w h e r e a,# 0. T h e last series g i v e s both solutions f o r w, s i n c e zl'a is a
two-valued function of z .
If equation (55) h a s the form
uo1z + u& + Ug& = 0,
389
Ch.XII1. LIMIT CYCLES OF SYSTEMS DEPENDING ON A PARAME'IEK
i t s solutions can be expressed by the series
w = &'/a f $zz4/s + . . .,
where p i # 0, and t h i s notation gives all the t h r e e solutions of the equation,
since Z1/3 is a three-valued function of z.
In the above examples, the solutions of equation (55) a r e expressed by
series in fractional (rational) positive powers of z. This suggests that the
s a m e representation w i l l obtain in the g e n e r a l c a s e also. W e will therefore
s e e k solutions of (55) in the f o r m
W=yza+y~zal+y,z-+
where y # 0, and a and a , are positive rational numbers, a < a, < az < ...
..., (59)
I
Let u s f i r s t establish what power exponent a will e n s u r e convergence of
the power s e r i e s (59) to a solution of equation (55) f o r s m a l l z .
Let ukobe the f i r s t nonzero coefficient of the f o r m u i oin expansion (52),
and uo( the f i r s t nonzero coefficient of the f o r m uoI (these coefficients exist
by assumption).
If (59) i s a solution of equation (55), we have
F(yza+y,z"+ ..., Z ) E O , (60)
i . e . , inserting f o r w in the series ( 5 2 ) i t s expansion f r o m (59), we obtain

identically z e r o . T h i s enables u s t o determine the numbers a and Y . A s
t e r m s in s e r i e s (52) obtained after substitution f r o m (59). We can neverthe-

less isolate a finite number of t e r m s which definitely include the lowest-
order terms.
T h e s e a r e p r i m a r i l y t e r m s of the f o r m
uolz' and UkoykZak (61)
and a l s o the t e r m s
uij,y'zRi+ji,
where 1 < i < k - 1, and the index ji s a t i s f i e s the inequalities l < j i < 1- 1
and is m o r e o v e r the s m a l l e s t index j f o r which the coefficient ut, with
fixed 1 does not vanish.
It is readily s e e n that the t e r m s of lowest o r d e r in z must be contained
among the t e r m s (61) and ( 6 2 ) .
If identity (60) is satisfied, t e r m s of the lowest o r d e r in E mutually
cancel. Since none of the coefficients U ~ in J (61) and (62) is z e r o , we con-
clude that a t l e a s t two such t e r m s should b e of the s a m e o r d e r , which,
however, should not exceed the o r d e r of the remaining t e r m s . Thus,
the numbers
ka, ( k - i ) a + j k - i r (k--)a+jR-21 . - . $ a+iiT 1 (63)
include a t l e a s t two equal numbers which a r e not g r e a t e r than all the other
numbers. Suppose that j k = 0 and i o = [ , i . e., the numbers in (63) have the
f o r m iz+ j , , i = O , 1, 2, . . . , k. Then a should satisfy a t l e a s t one of the l i n e a r
equations
$2:. EFFECr OF SlIALL CH.WGES IS PARAh!ETER
,,vhere i , & i ? , a n d for a n v i = O , 1, 2. . . .. Is w e s h o u l d h a v e

ia + i ; 2. itaf j,, ( = i2.* ji?). (65)
T h e s e v a l u e s of a w i l l tte c a l l e d f e a s i b 1e . T h e r e m a y b e s e v e r a l
f e a s i b l e v a l u e s . To find all t h e f e a s i b l e v a l u e s , w e w i l l u s e a g e o m e t r i c a l
m e t h o d known a s t h e m e t h o d of N e w t o n ' s p o 1 y g o n .
Lye i n t r o d u c e r e c t a n # u l a r c o o r d i n a t e s i. j i n t h e p l a n e a n d a s s i g n t o
e v e r y t e r m in ( 6 1 1 and ( 6 2 ) a point -4; o n t h e p l a n e Lvith t h e c o o r d i n a t e s (i. ji)
( i - I). 1. 2. . . ., k). T h e t e r m s i n ( 6 1 ) e v i d e n t l y c o r r e s p o n d t o p o i n t s of t h e
f o r m ito(0, I ) a n d -4h(k. 0)w h i c h lie o n t h e c o o r d i n a t e axes: the o t h e r p o i n t s
A i fall i n t h e f i r s t q u a d r a n t , a n d t h e i r a b s c i s s a s a n d o r d i n a t e s d o n o t
e x c e e d k - 1 a n d 1 - 1, r e s p e c t i v e l y .
S u p p o s e t h a t for s o m e r e l a t i o n ( 6 4 ) is s a t i s f i e d , a n d il # i2. T h e n
11% -i i ,
i2--l1 '
i.e., a i s t h e s l o p e f a c t o r of t h e s t r a i g h t l i n e t h r o u g h t h e p o i n t s A , , ( i , . ji,) a n d
-L2tLiiJ.
T h e e q u a t i o n of t h i s . s t r a i g h t l i n e is
j-ji, = --a (i - i l ) o r j - j i , - a ( i - i I ) =O. (66)
C l e a r l v , f o r t h e p o i n t s fi. j ) w h i c h lie n o t 1o w e r t h a n t h e l i n e ( 6 6 ) , w e
h a v e j - j , , + a (i - i I ) 2.0 , i . e.,
i +-ai 3 it, + ai,, (67)
a n d for t h e p o i n t s b e 1o w t h i s l i n e , w e h a v e
.
C o n d i t i o n ( 6 5 ) s i g n i f i e s t h a t e a c h of t h e p o i n t s -4, (i, j , ) , i = 0 , 1, 2, . . , A-,
l i e s e i t h e r o n t h e l i n e (E6) o r a b o v e t h i s l i n e . H e n c e i t follorvs t h a t e v e r y
f e a s i b l e a i s i n a o n e - t o - o n e c o r r e s p o n d e n c e to a s t r a i g h t l i n e p a s s i n g a t
l e a s t t h r o u g h t w o p o i n t s A;, w h i c h h a s a n e g a t i v e s l o p e f a c t o r a n d is s o
l o c a t e d t h a t n o n e of t h - p o i n t s .A, lies b e l o w t h i s l i n e . W e w i l l r e f e r t o
t h e s e l i n e s a s f e a s i b 1e 1i n e s . a is e q u a l t o m i n u s t h e s l o p e f a c t o r
of t h e c o r r e s p o n d i n g l i n e .
T h u s , i n o r d e r t o find all t h e f e a s i b l e v a l u e s of a , w e s h o u l d find all
t h e f e a s i b l e l i n e s . T h i s c a n b e a c c o m p l i s h e d i n t h e following way. L e t
s b e a m o v i n g l i n e w h i c h i n i t i a l l y c o i n c i d e s w i t h t h e axis i. W e t u r n t h i s
l i n e c l o c k w i s e a r o u n d the point A h (k,0) u n t i l it p a s s e s t h r o u g h o n e of t h e
p o i n t s A i , t h e point A*, (hI,j k , ) say. T h e l i n e -A!,;ik, o b t a i n e d i n t h i s w a y w i l l
b e d e s i g n a t e d sl. T h i s is e v i d e n t l y a f e a s i b l e l i n e . T h e l i n e s1 m a y p a s s
t h r o u g h m o r e t h a n t w o p o i n t s A t . T h e n , l e t An, b e t h e l e f t m o s t of t h e s e
p o i n t s ( i . e . , t h e o n e with t h e l e a s t a b s c i s s a ) .
If .4kl d o e s not c o i n c i d e w i t h t h e point .-to (0, I ) , t h e m o v i n g l i n e w i l l b e
f u r t h e r r o t a t e d i n t h e c l o c k w i s e d i r e c t i o n a r o u n d t h e p o i n t AR, u n t i l i t
p a s s e s t h r o u g h s o m e point Aka (k,, j k t ) (if t h e r e a r e s e v e r a l s u c h p o i n t s ,
.Ah, is chosen a s the l e f t m o s t ) . The r e s u l t i n g c u r v e is d e s i g n a t e d SZ,
a n d t h e p r o c e s s is c o n t i n u e d u n t i l t h e m o v i n g l i n e p a s s e s t h r o u g h t h e
39 I
-
Ch.XU1. LIMIT CYCLES OF SYSTEMS DEPENDING ON A PAMMETER
point A , (0,1). T h e r e s u l t is a s t r a i g h t
line designated ,s ( F i g u r e 169). In p a r t i c -
u l a r c a s e s , m m a y b e equal t o 1.
T h e convex polygonal line AkAa,Ak,. ..
...
Ak,-iAo obtained in t h i s construction
i s known a s N e w t o n ' s p o l y g o n (or
N e w t o n ' s d i a g r a m ) . It is c l e a r
f r o m the construction that each s i d e of
Newton's polygon is a segment of a
feasible line. It is r e a d i l y seen, however,
that the c o n v e r s e is a l s o true: e v e r y
feasible line contains a segment which
is one of t h e s i d e s of Newton's polygon.
D A*no' We have thus established that all the
FIGURE 169 feasible values of a are equal t o minus
the slope f a c t o r s of the s i d e s of Newton's
polygon.
Now let the a i n expansion (59) b e one
of the feasible values of the index. We will d e t e r m i n e the corresponding
value of the coefficient y . T o t h i s end, we s h a l l make u s e of the condition
that the lowest t e r m s in t h e left-hand s i d e of t h e identity (60) should mutually
cancel. T o fix ideas, let a be t h e slope f a c t o r of the s i d e Ak,Attz of Newton's
polygon and let .411,A,2,.. ., A + k, > lI > l 2 > . . . > l a > k2, b e those among
the points A i which lie on t h i s side. The lowest t e r m s of the corresponding
e x p r e s s i o n will cancel out i f t h e coefficient of the lowest d e g r e e of 3; (equal
+
t o k,a j k l = Ita + j l l =. . ..= k2a +
j h z ) is z e r o , i. e., if
g (Y)=Ukij~IYkl+flijlIYtl~ . * fU.lijl,y"+Ukzj12Yk'=0. (69)
Since b y assumption y +
0, we a r e only i n t e r e s t e d i n the nonzero r o o t s
of equation (69). i . e., t h e r o o t s of t h e polynomial
h ( y )= U b l j h , ~ k l - k n + Ul,jli~ll-k'f - .fUk2jkZ. (70)
T h e number of t h e s e r o o t s (counting each root according t o its multiplicity)

is k, - k2. T h e r e f o r e , the number of values of the coefficient y corresponding
t o each s i d e of Newton's polygon is equal t o the number of units accommo-
dated by the projection of t h i s s i d e onto the a b s c i s s a a x i s . T h e values of
t h e coefficient m a y be real o r complex, and s o m e of t h e m m a y b e equal t o
one another. Since t h e length of t h e projection of the e n t i r e Newton's
polygon is k u n i t s , we conclude that t h e f i r s t c o e f f i c i e n t y o f t h e
s e r i e s (59) h a s k v a 1u e s (not all of which are n e c e s s a r i l y different).
W e have established t h e n e c e s s a r y conditions t o b e satisfied by a
and y when w = yza + ylzal + . . . is a solution of the equation F (w.z ) = 0 .
T h e s e conditions can be stated a s follows: a s h o u I d b e e q u a 1 t o t h e
a b s o l u t e v a l u e o f t h e s l o p e f a c t o r of a n y o f t h e s i d e s o f
N e w t o n ' s p o l y g o n , a n d v- f o r a f i x e d a- s h o u l d b e e q u a l
t o o n e o f t h e n o n z e r o r o o t s o f e q u a t i o n (69).
Let u s consider the case when y f 0 is a s i m p l e root of (69). W e
will show that the above conditions are a l s o s u f f i c i e n t f o r the equation
F ( w ,z ) = 0 t o have a solution of the f o r m (59) with t h e given a and y .
392
$3'2. EFFECT OF ShfALL CH.4NGES IN PAR.4IIETER
L e t a - .L w h e r e p a n d q are i r r e d u c i b l e n u m b e r s , b e a f e a s i b l e v a l u e
v'
of t h e i n d e s c o r r e s p o n d i n g , as b e f o r e , to t h e s i d e A-l,,,.4tt:of N e w t o n ' s polygon,
a n d y = = 0 o n e of t h e r o ' 3 t s of e q u a t i o n ( 6 9 ) . T h e e q u a l i t i e s
k,z+ j k , = ip.-r i l l = . .. 7 =k
l,a-..,ir. ~ j k+.
c l e a r l y s h o w t h a t e a c h of t h e n u m b e r s k , - k,. I , - k2. . . .. I , - k? is a multiple

k,-k2
of q , i . e ., h (y) is a p o l p o m i a l i n yq of d e g r e e 7.
Now l e t y F 0 b e a s i m p l e r o o t of t h e p o l y n o m i a l g (y) ( a n d h e n c e of h (y)).
Then
g(y)=V. g'(y) #I!. (71)
To i n v e s t i g a t e t h e e q u a t i o n F (w,)I = 0 for :.
i0, w e s u b s t i t u t e
u' = I':" = L ' p , ,r, (72)
w h e r e ;I 1 is o n e of t h e '7 p o s s i b l e v a l u e s of t h i s q - v a l u e d f u n c t i o n . Using
( 5 2 ) a n d ( 6 0 ) , we r e a d i l y see t h a t
hl? i'bl
F (u,3) = F ( l . 9 q , 3) = 2 [g(L.)-z*',lT(Y* ;*q1. (73)
w h e r e i. is a p o s i t i v e i n t e g e r , a n d q ( 2 . . zi"') is a p o w e r s e r i e s i n L' a n d :' 'z,

w h i c h is a p r i o r i known t o c o n v e r g e f o r all z a n d t.3' q = u of s u f f i c i e n t l y
s m a l l m a g n i t u d e a n d , t h e r e f o r e , c o n v e r g e s f o r all s u f f i c i e n t l y s m a l l I z [ ,
if 1: t a k e s i t s v a l u e s froin a bounded r e g i o n . T h e e q u a t i o n F ( r . 2 ) = 0 is
e q u i v a l e n t for z rf 0 t o t h e e q u a t i o n
g (L') -Z*"I (1'. ;' q = 0.

S u b st i t u t i n g
-
-1 7 - ' 5' (74)
Lve o b t a i n t h e e q u a t i o n
Ir) (t., +
); = g ( r ) ;*q (c, 5) = 0. (75)
T h e c o e f f i c i e n t s of t h e pori-er series r e p r e s e n t i n g t h e f u n c t i o n q- ( 1 . . 5) a n d
t h e r e f o r e t h e f u n c t i o n i t s e l f a r e i n d e p e n d e n t of t h e p a r t i c u l a r v a l u e of t h e
q-valued function q u s e d in t h e s u b s t i t u t i o n (72).
F v (71),
Cb(y,O)=O, (P;(y. O ) # O .
T h e r e f o r e , b y T h e o r e m 68 ( t h e t h e o r e m of i m p l i c i t f u n c t i o n s ) , e q u a t i o n ( 7 5 )
h a s a u n i q u e s o l u t i o n i n a s u f f i c i e n t l y small n e i g h b o r h o o d of t h e point 5 = 0,
w h i c h r e d u c e s t o y f o r := 0 . T h i s s o l u t i o n h a s t h e f o r m
c =y - y,; - yL.;z- . . , (76)

T h e c o r r e s p o n d i n g s o l u t i o n is
393
Ch.XIII. LIhlIT CYCLES OF SYSTEMS DEPENDING ON A PARAMETER
We have thus established that i f a = p / q is a feasible value of the index,

and y is a simple root of equation ( 6 9 ) corresponding t o t h i s a , the equation
F ( w , z) = 0 h a s a solution of the f o r m (59).
Let a,, a2, . . ., a,,, b e all the feasible values of the index a for the equation
F ( w , z ) = 0, and let ai = p i / q i , where pi and qj are irreducible n u m b e r s . Con-
.
s i d e r the c a s e when f o r any at (i = i , 2, . ., m), all the non-zero r o o t s of the
corresponding equation (69) a r e s i m p l e . Let the number of t h e s e simple
I ...+ r,
r,+rz+
A s we have just established, to e v e r y pair of numbers

=k.
ai, y t j ( i = 1,2, ...,m ;

I
j = 1, 2, . . ., si) corresponds to a solution
Suppose that f o r a fixed i , the s a m e value of the root is taken Y;=z~~

in (79) f o r all j , j = i , 2, .. ., r l . It follows f r o m (78) that t h e r e exist
p r e c i s e l y k solutions of the f o r m (79). We will show that i f z f O is suf-
ficiently s m a l l i n absolute magnitude, the values of the solution (79) are
all different. Indeed, i f t h i s is not so, t h e r e exist two solutions of the
f o r m (79), w, and w,, say, which take on identical values on a sequence
of points z i - + O . If t h e s e solutions correspond t o different a, they have
I different o r d e r s of s m a l l n e s s relative t o z; t h e r e f o r e l i m
0 o r m, and the equality w i( 2 1 ) = w z (21) b r e a k s down for l a r g e 1 . If the
W 2 (Id
is either
solutions w , and w2 correspond t o the s a m e a, they differ in the f i r s t
I coefficients y . T h e r e f o r e l i m m=+l
Wz(b1)
and the equality w 1 ( z I=) w z ( z i ) again
b r e a k s down. O u r proposition is thus proved. It follows that a l l the k
solutions (79) of the equation F ( w , z) = 0 which g o t o zero f o r z-+ 0 are
different. But then by T h e o r e m 70, the functions (79) exhaust all the
solutions of the equation F ( w , z ) = 0 which go to zero f o r z-+ 0.
We will now show how to find the cyclic s y s t e m s of solutions men-
tioned in the r e m a r k t o T h e o r e m 70. In the p r o c e s s we will d e r i v e a
different f o r m of solutions (79), which w i l l be useful a t a l a t e r stage.
A s before, we a s s u m e that f o r e v e r y a i , i = 1, 2, . . ., m , all the non-
z e r o roots of equation (69) are s i m p l e . It suffices t o consider one of
the feasible values at. W e will designate i t a, and l e t
a=plq, , (P, 9) =I.
Let
h (Y)= W l j h ; Y k l - - A a t U . ..
~ ~ j ~ , y ~ - ~ ~-f- - 6&hsj**
f - (70)
b e the polynomial h corresponding t o the given a.
We have s e e n before that h ( y ) is a polynomial of d e g r e e d = k k 2 in y f .
q
Let
r=yg (80)
and
394
5 3'7. EFFECT OF S h I A L L CHANGES Ih' P.W&:ETER
Let rl, r.. . . , r,

, be the roots of the polynomial H (r). Since
h(y)=ri(y4) (821
and all the r o o t s of the polynomial h ( y ) a r e simple, the n u m b e r s rl, I'z. ..., r,
a r e all different.
Let
~ Z I Si?.
I - . . .) Y i y (83)
h e all the q-th d e g r e e r o o t s of I'i. By (80) and (82j, t h e s e n u m b e r s a r e

roots of the polynomial h (y). Thus, e v e r y root Ti of the polynomial H i , T )
c o r r e s p o n d s to a sequence ( 8 3 ) of roots of the polynomial h ( y ) . W e know
from algebra that the numbers ( 8 3 ) can be expressed in the f o r m
YLI, Yil", Tile2, -..v Y~I"'-', (8-1)
w h e r e y i l is one of the q-th d e g r e e r o o t s of Ti (any of the q roots c a n b e

chosen, a s long a s i t is. kept fixed), and e is the fundamental q-th d e g r e e
root of 1.
Let io be one of the ITalues of {?;( a l s o quite a r b i t r a r , y , provided i t i s
fixed). A s we have s e e n above, the sequence (83) of the n u m b e r s yi, c o r -
responds to the sequence
V L , ,Ui?. . . .. U i q (85)
of solutions of the equation F(@, z ) - 0 , where
(L.. ',. -
- 5: (yijf yij,5, i-
y& . ..) ( j = I, 2, . ... q ) . (86)
If the c9 in the right-hand s i d e of ( 8 6 ) is replaced with s o m e o t h e r value

of the q-th d e g r e e root ,2f 3, we a l s o obtain a solution of the equation
F ( w , 3) = 0 . T h e r e f o r e , i f w e k e e p j fixed, setting for convenience j = 1,
and let the c0 in ( 8 6 ) run o v e r a 1 1 the values of the q-th d e g r e e root of z ,
i . e ., o v e r the n u m b e r s
L
CII,
C
=oE. :*E2, . . .. :(,Eq-'. (87)
w e obtain a s e t of Q solutions
- - >.. . . ,
U Z l . LC Vi*+ (88)
where
(2.,,=( 5gE')P[YiifYIlI;OE1-Yil~(50E') .'l....l ( I - 2 0 , 1, 2. ..., q-1) (89)

or
zZc[= 5: [yil"'p-t. y I I 1 ~(t+l);,
p -2 ... I (1 = 0 , i , ?+ . . ., q - 1). (90)
Since e is the fundamental q-th d e g r e e root of 1, and ( p . 9) = 1, the

n u m b e r s E'P ( I = 0. 1, 2, . . ., q - 1) constitute the s e t of a l l the r o o t s of
9-th d e g r e e of 1. T h e r e f o r e , the n u m b e r s y i l ~ ' p( 1 = 0,1. . . ., q - 1)
coincide with the n u m b e r s (8-1) and hence with the n u m b e r s (83). Since
e v e r y solution of the fovm ( 7 9 ) is determined by i t s f i r s t coefficient, it
thus follows that solutions (88) coincide with solutions ( 8 5 ) .
395
The s t r u c t u r e of the solutions (88), determined by (89), readily shows

that the functions (88) constitute a cyclic s y s t e m of solutions. Indeed, as
the point z moves around the origin tracing a simple closed c u r v e , the
n u m b e r s go&, which a r e the q-th d e g r e e roots of z, undergo a cyclic p e r -
mutation.
Let u s s u m m a r i z e o u r r e s u l t s . Let a = p i g be a feasible value of the
index (obtained by construction of Newton's polygon). To obtain the solutions
of the equation F ( w , z) = 0 corresponding t o t h i s a, we have t o find the r o o t s
rl, r2. . . ., I', of the equation H ( P ) = 0. Let y i i be one of the q-th d e g r e e r o o t s
of rr (i = 1. 2, . . ., d). The corresponding solution is
wit = Lf (Yii + YIIISO+ . . .)$ (86)
where Lo is any fixed value of 7;. Making torun o v e r all the values of t h i s
root, we obtain q solutions
W i t , Wiz, . .. t Wiqr
which constitute a cyclic s y s t e m . Since i = 1, 2 , . . . , d , we obtain a total

of d cyclic s y s t e m s of solutions, corresponding t o the feasible value a = p i q .
W e r e c a l l that d = k , - - k z , where k I - k z is the degree of the polynomial (70).
9
In o u r analysis of the creation of limit cycles f r o m a multiple limit
cycle, the function F (w, z) will be identified with d (no,p) (see (47)), i. e.,
with a real-valued function of real variables, and we will be concerned
with the existence o r absence of r e a l r o o t s of t h i s function.
The following l e m m a w i l l be of considerable importance in this
connection.
L e m m a 1 . If all the coefficients of the series expansion of the function
F ( w , z) are real, and y i s a real simple non-zero root of the equation g ( y ) = 0
( s e e (69)) or, equivalently, of the function h ( y ) , all the coefficients y,. yz. ... of
the cmvesponding solution (77)
w =YZP/'I+ Y,Z(P+" + yzztP+z)/r + . . .
of the equation F (w, Z) = o a r e real.
Proof . T h e numbers y and yt (i = 1, 2, ...
) are the coefficients of the
p o w e r - s e r i e s expansion of the function u ( s e e ( 7 6 ) ) satisfying equation (75).
'Since the series of the function F (w, z ) h a s real coefficients, the s e r i e s of
the function Q, (u, Z) a l s o h a s r e a l coefficients. When solving the equation
Q (27, z ) = 0 by the method of indeterminate coefficients, each successive
coefficient y i + l is a polynomial in the coefficients of the function Q and the
coefficients y , yl, y 2 , . . ., y i (see / l l / , Vol. 11, Sec. 450). But then all the
coefficients y t a r e real. Q. E. D.
4. The behavior of limit c y c l e s of s o m e dynamic s y s t e m s

following s m a l l changes in the p a r a m e t e r
In t h i s subsection we will apply the previous r e s u l t s in o r d e r to

establish what happens to a limit cycle of a dynamic s y s t e m depending
on a p a r a m e t e r following a s m a l l change i n the p a r a m e t e r . W e will
396
5 3 2 . EFFECT OF Shl.4LL CHANGES IN PARAhlETER
o n l y c o n s i d e r s y s t e m s 2f t w o p a r t i c u l a r t y p e s , o n e of w h i c h i n c l u d e s
s y s t e m s o b t a i n e d from some f i x e d s y s t e m b y a r o t a t i o n of t h e v e c t o r
f i e l d . T h e m e t h o d s u s e d , h o w e v e r , c a n b e a p p l i e d t o o t h e r t y p e s of
s v s t e m s depending on a p a r a m e t e r . A s throughout this c h a p t e r , we
will a g a i n c o n s i d e r o n l y a n a l y t i c a l d y n a m i c s y s t e m s .
W e w i l l f i r s t p r o v e ;I n u m b e r of l e m m a s , w h i c h a r e e x t e n s i v e l y u s e d
i n t h e m a i n p r o p o s i t i o n s . T h e s e lemmas follow d i r e c t l y f r o m t h e r e s u l t s
of t h e p r e v i o u s s u b s e c t i o n .
Consider the equaticn
F (u;3) = U +
~ ~ Z Uoi2
C + u?~u*1- . . . = 0 . (55)
L e t all t h e c o e f f i c i e n t s uil of t h i s e q u a t i o n b e real a n d l e t uqa b e t h e f i r s t

of t h e c o e f f i c i e n t s u i a w q i c h d o e s not v a n i s h . W h e n s p e a k i n g of s o l u t i o n s
or r o o t s of e q u a t i o n (55), w e w i l l i n v a r i a b l y m e a n roots of s u f f i c i e n t l y
s m a l l m a g n i t u d e which c o r r e s p o n d to s u f f i c i e n t l y small non- z e r o real
v a l u e s of 3.
L e v i m a 2. Let k be an euen number, and u 0 , # 0 . Then, if E<o
( $>0) the equation I (10, 3) = 0 has two different real roots far 3 >0 (: <O ) ,
which a r e simple roots, and has no real roots fm 3 < 0 ( 3 > 0).
Proof . L e t k = 21. l > 1. S i n c e uo, # 0, N e w t o n s polygon c o n s i s t s of
a s i n g l e s e g m e n t t h r o u g h t h e p o i n t s .Ali (21. 0) and .-io
(0,1) ( F i g u r e 170), i . e.,
t h e i n d e x n m a y t a k e o n o n e v a l u e only, = &. T h e v a l u e s of t h e c o e f f i c i e n t s are
o b t a i n e d from e q u a t i o n ( 6 9 ) , w h i c h t a k e s o n t h e f o r m
u2/,y*t + u,)i = 0. (91)
T h e r e f o r e , t h e e q u a t i o n F ( w , 2 ) = 0 h a s 21 s o l u t i o n s . A s w e h a v e s e e n
b e f o r e , all t h e s e s o l u t i o n s m a y b e w r i t t e n i n t h e form
&? = 21/21 ( y + y*z=+ y,z*/?+ . . .). (923
w h e r e t h e c o e f f i c i e n t s y a n d yi a r e t h e same, a n d z1 successively runs over

all t h e 2 1 - t h d e g r e e r o c t s of z.
If %< 0, e q u a t i o n (!31) h a s t w o real r o o t s . L e t o n e of these roots
UZIO
b e t h e y i n ( 9 2 ) . T h e n a l l t h e c o e f f i c i e n t s y i , i = 1, 2 , . . . , a r e also real
b v L e m m a 1.
If z > O , t h e r e e x i s t t w o real v a l u e s of t h e f u n c t i o n z * / z l . T h e s e v a l u e s
c o r r e s p o n d , b y ( 9 2 ) , t o real v a l u e s of w . T h e c o m p l e x v a l u e s of z ~ / cor- * ~
r e s p o n d for s u f f i c i e n t l y small z to v a l u e s
of w w h i c h a r e close to t h e c o m p l e x n u m b e r
+,i . e . , a r e also c o m p l e x . If z c 0 , all
L==-
zt
47/41) t h e v a l u e s of t h e f u n c t i o n ~ 1 2 1are c o m p l e x ,
a n d t h e f u n c t i o n w m a y not have real v a l u e s
for s u f f i c i e n t l y s m a l l z .
U &,kUI S i n c e all t h e r o o t s of e q u a t i o n ( 9 1 ) are
FIGURE 170 d i f f e r e n t , all t h e k = 21 r o o t s of t h e e q u a t i o n
F (w,z ) = 0 a r e a l s o different f o r sufficiently s m a l l z , and a r e thus simple.

W e have thus proved the l e m m a f o r z k <0 . The c a s e > 0 is reduced
t o the previous c a s e by substituting - z f o r z. T h i s completes the proof
of the l e m m a .
L e m m a 3 . If k i s odd and uO1+ 0, the equation F ( w , Z ) = 0 has precisely
one real root f o r both z > 0 and z < 0, and this root i s simple.
Pr oof . Let k = 21 + 1. Z > 1. Equation ( 6 9 ) h a s the f o r m
uzl+l,oYsl+l +uot = 0,
and the corresponding solution of the equation P(w,z) = 0 is
The validity of the l e m m a is established along the s a m e lines a s in the

previous proof.
C o r o 11 a r y . If k is odd, the equation F ( w , a) = 0 may have m o r e than
one r e a l solution only i f uol = 0.
L e m m a 4 . If k i s oddand
uo1= 0, Utt # 0, (93)
the equation F ( w , Z ) = 0 has either three real roots for z > 0 and one real
root f o r z < 0, o r three veal roots f o r z =-= 0 and one real root f o r 0. =-
All the roots of the equation, and in particular its real roots, are simple.
P r o o f . Let k = 21 +
1 , 1 > 1 and let uOmbe the f i r s t of the coefficients
uoI which does not vanish. Since uol = 0, w e have m > 2 .
It is readily s e e n that in t h i s p a r t i c u l a r c a s e Newton's polygon i s
made up of two segments, one through the points A. (0, m) and A , ( 1 , l )
and the other through the points A , (1, I )
and A21+,(21+1, 0) ( F i g u r e 1 7 1 ) . Con-
A sequently, a may have two values,
namely a,=m- 1 and
1
%=x.The equa-
tion for y corresponding to a i = m - l
is u , , y + uom= 0; t h i s equation h a s a
is an analytical function of z . By L e m m a 1,
all the yr ( i = 1,2, . . . ) are real and, t h e r e -
f o r e , f o r any r e a l a , the solution 10 is r e a l .
The equation f o r y corresponding t o a2=f is
UZl+l o y ~ ~ + 1u,,y
+ =0 (95)
(see ( 6 9 ) ) , and t h e r e a r e 2 2 solutions of the f o r m
398
E :23. EFFECT OF ShlALL CHANGES I X PARAhlETER
*
C o n s i d e r the c a s e %+r 0 < 0 . Equation (96) h a s a real r o o t and, as i.n
o u r proof of L e m m a 2 , we c a n show that t h e r e a r e two real solutions of
the f o r m (96) f o r z > O and no such solutions f o r z < 0 . T h u s , if *o< 0, t h e
equation F ( w , z) = 0 h a s t h r e e real solutions for z > 0 (one of the f o r m ( 9 4 )
and two of the f o r m ( 9 6 : ) , and only one real solution (of the f o r m ( 0 4 ) ) f o r
z4.0.
T h e case co>O i s reduced t o the p r e v i o u s case b y substituting - z
for z . In t h i s c a s e , t h e equation F ( w , z ) = 0 h a s one real solution for z > O ,
and t h r e e real solutions f o r z < 0 . Since all t h e r o o t s of equation (95) a r e
different, all t h e r o o t s of t h e equation F ( w . s) = 0 (and i n p a r t i c u l a r , i t s
real r o o t s ) are also different and hence s i m p l e . T h i s c o m p l e t e s the proof
of t h e l e m m a .
We c a n now proceed u i t h t h e main propositions of t h i s subsection. We
will f i r s t e s t a b l i s h what happens t o a l i m i t c y c l e of a dynamic s y s t e m
when t h e v e c t o r field is rotated. T h e o r e m s 71 and 72 provide the a n s w e r s
t o this question.
Let
b e the s t a r t i n g analvtical s y s t e m .
T h e s v s t e m (A,) dep3nding o n a p a r a m e t e r i s chosen in t h e f o r m
__
it = P (I,Y) - pQ (5. Y) = p(z. Y. P).
dz
'$=Q(z. Y) t p P ( 5 , y ) = b ( z 7 Y, P),
and the v e c t o r field of t h i s s y s t e m is obtained f r o m t h e v e c t o r field of ( A )

h v rotating through an angle equal t o tan-'p (see end of $ 3 ) .
We w i l l use t h e s a m e n o t a t i o n a s i n 5 3 2 . 2 . T h e n , by (231, rhe functions
pt and q' c o r r e s p o n d i n g t o (XJ have t h e f o r m
pt (t. Y) = - 0 (z, Y), ( 2 , Y) = P (2. y).

(97)
pt (2, y) = q i (z. y) = 0 f o r 1 9 2.
Suppose that s y s t e m ( A ) , or equivalently ( A o ) , h a s a l i m i t c y c l e L o .
T h e o r e rvi 71. If Lo is a limit cycle of even multiplicity of system
( A o ) , there exist E > 0 and p,,> 0 uith the follozcing property: either for
ecery p>O, ~ p I O, ( p I < p o ,
(A,) has no limit cycles in U e ( L o ) , whereas for p < O , Ipl<por ( A J has
precisely two limit cycles in Ue(Lo), uViich are structurally stable. The
number E may be chosen a s small a s desired.
P roof . L e t n = f (no,p) b e the s u c c e s s i o n function c o n s t r u c t e d f o r ( A J
on s o m e n o r m a l t o t h e l i m i t c y c l e Lo (see (16)), and d ( n o , p) = f (no,p ) - n 0
(see ( 4 7 ) ) .
A s ins32.1, we a s s u m e that t h e l i m i t c y c l e Lo of t h e s y s t e m (Ao) c o r -
r e s p o n d s t o no = 0. T h e series expansion of t h e function d (no,p) n e a r t h e
point (0, O ) , b y (19) and (47), h a s t h e f o r m
d (nth p) = ( 4 0 - 1) no iuo1p + Uzdtglf ... (98)
399
In particular,
d (no,0)= (ul0- 1) no +u m : + %n: + ... (99)
F y assumption, Lo is a limit cycle of even multiplicity of (AQ). T h i s

and Definition 28 ($26.2) show that ui0- 1 = 0 and that the f i r s t non-zero
coefficient in expansion ( 9 9 ) is a coefficient before an even power of no.
Let this coefficient be u z l ,o, where La 1. Thus,
d (no, P) = uzt, on;' + uoip + . . (100)
The coefficient u0, may be computed f r o m (36), 532.1. Let
X==(P(O? Y=11,(4,
where 'p and 11, a r e periodic functions of period T, b e the solution of (Ao)
corresponding t o the limit cycle Lo. Then
p ('p (4, l# (4) 3 9' (4, Q ((P (4, 11, (d) E 11,' (4.
By t h e s e relations and (97), we have
41 ('p (4, l# (4) E v' ( 4 7 PI (9(4, l# (SI) = -11,' (4.

Inserting the l a s t expressions in (36). we obtain
r *
s (P;+O;) ds r -$ (P;+QL)d.
UOI =
i
eo O
+V (s)*l d s .
tcp' ( s ) ~
Hence i t follows that uol # O . But then, in virtue of ( l o o ) , the conditions of

L e m m a 2 are satisfied f o r the function d (no,p) i f i t is considered a s F (w,2).
F y t h i s l e m m a , (Au), f o r sufficiently s m a l l p, h a s precisely two limit
c y c l e s in a sufficiently s m a l l neighborhood of the path L o . Since by the
s a m e L e m m a 2 the r e a l roots of the equation d (no,p) = 0 to which these
c y c l e s correspond a r e simple roots, the corresponding limit cycles a r e
s t r u c t u r a l l y stable ( s e e 13.3, (31), and a l s o 13.3, Definition 18, and
14, Theorem 18). Q . E . D .
The proposition of T h e o r e m 7 1 may be formulated in the following
graphic form:
When the v e c t m field of a dynamic system is rotated in one direction,
a limit cycle of even multiplicity decomposes into two structurally stable
cycles, and when the field i s rotated in the opposite direction, the limit
cycle disappears.
Let u s now consider the c a s e when the dynamic s y s t e m (A) h a s a limit
cycle Lo of odd multiplicity.
T h e o r e m 72. If Loi s a limit cycle of odd multiplicity of dynamic
system (Ao), there exist e > 0 , po > 0 such that jw all p # 0, I p I < po,
(A,) has a single limit cycle in U, (Lo), which i s moreover structurally
stable. The number e may be taken a s small as desired.
P roof . The first non-zero coefficient of n,k ( k = 1, 2, . ) in expansion ..
(98) is a coefficient before an o d d power of no, e . g . , the coefficient before
n:l+f, where L > O . The coefficient uol h a s the s a m e f o r m a s in the previous
400
53% EFFECT OF SLIALL CH.*GES [L P.4R.lilETER
t h e o r e m , i . e . , i t d o e s not v a n i s h . T h e p r o p o s i t i o n of T h e o r e m 7 2 t h e n
f o l l o w s d i r e c t l y from L,emma 3 . Q . E . D .
R e m a r k 1. If Lo is of m u l t i p l i c i t y 1. i . e., a s i m p l e l i m i t c y c l e of
( - A ) , t h e n it is a s t r u c t u r a l l y s t a b l e p a t h ( T h e o r e m 18, s 14). T h e r e f o r e ,
not o n l y t h e s y s t e m (A,) o b t a i n e d b y r o t a t i o n of t h e v e c t o r f i e l d , b u t a n v
s v s t e m s u f f i c i e n t l y close t o ( A ) w i l l h a v e p r e c i s e l y o n e l i m i t c y c l e i n a
S u f f i c i e n t l y s m a l l n e i g h b o r h o o d of L o . In case of h i g h e r m u l t i p l i c i t i e s of
t h e c y c l e L o , t h e r e a l w a y s e x i s t s y s t e m s a s close a s d e s i r e d to ( A ) , which
h a v e m o r e t h a n o n e l i m i t c y c l e n e a r L o . T h e o r e m 72 s h o w s t h a t i n a
s m a l l r o t a t i o n of t h e v e c t o r f i e l d , a m u l t i p l e c y c l e of odd m u l t i p l i c i t y
b e h a v e s l i k e a s i m p l e c y c l e , i . e . , t h e modified s y s t e m h a s o n e and o n l y
o n e l i m i t c y c l e i n a s u f f i c i e n t l y s m a l l n e i g h b o r h o o d of t h e o r i g i n a l c y c l e ,
which i s m o r e o v e r s t r u c t u r a l l y s t a b l e .
R e m a r k 2 . T h e o r e m s 7 1 a n d 72 a n d o u r p r o o f of t h e s e t h e o r e m s
r e m a i n v a l i d i f (-Ad is t a k e n i n t h e form
dr
-=
di J' (f,)'5 --Pf (2,Y) Q (I,Y). dy
7 = Q (2,Y) +pf (I,Y) P (2,Y).
w h e r e f (I.y) i s a f u n c t i o n w h i c h r e t a i n s a c o n s t a n t s i g n a t all p o i n t s of t h e
limit cycle Lo.
In c o n c l u s i o n of t h i s s e c t i o n , w e w i l l p r o v e a n o t h e r t h e o r e m , w h i c h i s
n o l o n g e r r e l a t e d to r o t a t i o n s of t h e v e c t o r f i e l d s .
L e t I,, b e a limit c y c l e of ( A ) , I = 'p ( t ) . y = 9 ( I ) t h e s o l u t i o n c o r r e s p o n d i n g
to t h e path L o . T > 0 t h e p e r i o d of t h i s s o l u t i o n . L e t f u r t h e r F (I,y) b e a n
ana1,ytical f u n c t i o n d e f i n e d i n t h e same r e g i o n a s s y s t e m ( A ) , w h i c h s a t i s f i e s
t h e following conditions:
(ai F ( c p ( s ) . ~ ( ( s ) ) = 0.
i (see e n d of 933.1).
l b ) I F ; ( v ( s ) . $ ( S ) ) ~ ~ + [ I ' ; ( ( F~( (Ss) ). ) I * . 0
dr -
-
dt
=P (I y) -i-ppl ( I ,y) 4-p*p2 (". y) I .. f = P (I,y, p),
(A,)
$=Q (2,y) t p41 (I,Y) p% ( 2 , Y) t ' . . = Q (I, Y7 p)
be a dynuiiiic s y s t e m , und let the fiinctions PI (3, Y) ana' q1 (I,y) have the forni
PI,I( Y) = -r (17 Y) F; (2,Y), QI (I.Y) = F ( 5 , y) F; (I,y). (101)
where F is a fimction satisfying conditions (a) ana' ( b ) . Then if Lo is a limit

cycle of mid intiltiplicity of system ( A ) , there exist po > 0 and E > 0 which
satisfy the follaoing c a d i t i o w for all p. p 1 < p o t hatting the same sign,
(A,) has precisely three limit cycles in L', ( L o ) , and these cycles are struc-
turally stable; fm all p.. I p I < pot of the opposite sign, (A,) has precisely
one limit cycle in C , ( L o ) , which is also stnrctuvally stable.
P roof. T h e e x p a n s i o n of d (no, p) i n t h e n e i g h b o r h o o d of ( 0 , O ) h a s t h e
f o r m (08):
d (no. P I = (uio - 1) no + UOIP f uiinp -+ UZOU: +
w h e r e u I o is c o m p u t e d f r o m (34). T h e c o e f f i c i e n t uol i s c o m p u t e d f r o m ( 3 6 ) .
Ey (101) a n d c o n d i t i o n ( a ) , it f o l l o w s t h a t U O I = 0 . The c o e f f i c i e n t u I 1is
c o m p u t e d f r o m ( 4 2 ) a n d , a s w e h a v e s e e n a t t h e e n d of 133.1, u l l = 0i= under
401
the conditions of o u r theorem. Thus, the conditions of L e m m a 4 a r e

satisfied f o r the function d ( n o , p) if it is considered as P (to,z ) . The proof
of the t h e o r e m follows directly f r o m t h i s l e m m a . Q. E . D .
S33. CREATION O F A LIMIT CYCLE

F R O M A CLOSED P A T H O F A
CONSERVATIVE SYSTEM
1. T h e integral invariant and conservative s y s t e m .

Statement of the problem. T h e method of
the small p a r a m e t e r
The concept of the i n t e g r a 1 i n v a r i a n t was introduced by Poincare

( s e e 1241, Sec. 235, p . 5 ) . We will give h e r e a definition of the integral
invariant f o r a dynamic s y s t e m of second o r d e r . In the g e n e r a l c a s e -
f o r a s y s t e m of n-th o r d e r - t h i s concept is defined analogously. A s in
the previous section, we a s s u m e that all the relevant s y s t e m s a r e analytical.
Let
b e a dynamic s y s t e m defined in G. For simplicity, we a s s u m e that G is

made up of whole paths of s y s t e m (A). Let D be a closed subregion of
G, M o (zo, yo) ED any point in t h i s subregion. Consider the path x = p ( t ; z,, VO),
y = ( t ; zO, yo) of s y s t e m (A) which f o r t = 0 p a s s e s through the point Mo.
Let the point with the coordinates p ( t ; zo, yo), $ ( t ; z, yo) be designated M ( t , M o ) .
The s e t of all points M ( t , M,) obtained when M Or u n s over a l l the points of
the subregion D f o r a fixed t , i . e., the set of point ( M ( t , M o ) ; M OE D ) , is
designated D t . D t is c l e a r l y homomorphic to D and is contained in G. Dt is
obtained f r o m D , so t o say, by a t r a n s l a t i o n o v e r a d i s t a n c e t
(intime) a l o n g t h e p a t h s .
Let p (z,y) be an analytical function in G which is not identically z e r o in
t h i s region.
D e f i n i t i o n 31. The integval
is called the integral invariant of a dynamic system (A) in every bounded

closed subregion D C G and f o r every t , we have
If (I) i s an integral invariant, the function p ( x , y) is called the density

of the integral invariant.
The integral invariant readily lends itself t o a hydrodynamic interpreta-
tion. Let (A) be regarded as a s y s t e m of equations describing the velocity
of steady- s t a t e motion of s o m e two-dimensional fluid filling the region G,
which contains neither s o u r c e s nor sinks. Let p (2,y) be the density of the
5438 402
5 33. CLOSED PATHS OF C0NSERL';ITIVE SYSTEAIS
fluid a t t h e point (2.y). 'The i n t e g r a l ( 1 ) t h e n e x p r e s s e s t h e mass of fluid

f i l l i n g t h e r e g i o n D , and e q u a l i t y ( 2 ) s i g n i f i e s t h a t t h e fluid m a s s is c o n -
s e r v e d a s t h e fluid p a r t i c l e s m o v e a l o n g t h e i r s t r e a m l i n e s i n a t i m e t to
f i l l t h e r e g i o n D t . For a n i n c o m p r e s s i b l e f l u i d , p (5,y) = c o n s t , a n d t h e
i n t e g r a l i n v a r i a n t i s t h e area of t h e r e g i o n D.
L e t u s d e r i v e t h e c o n d i t i o n t o b e s a t i s f i e d b y p ( 3 , y) for (1) t o b e a n
i n t e g r a l i n v a r i a n t . A s s u m i n g a fixed D f o r t h e t i m e b e i n g , w e w i l l u s e
t h e notation
J ( t )= 5 'i
(Dt)
p (E, Y) d x d y . (3)
Then 'i 5 p ( x , y ) d s d y = J ( O ) ,
(D)
a n d for e q u a l i t y ( 2 ) t 9 hold t r u e f o r a n y 1, it i s
n e c e s s a r v and sufficient that
J' ( t ) 3 0
( t h e d i f f e r e n t i a b i l i t y of J ( f )w i l l be p r o v e d b e l o w ) . To c o m p u t e J ' ( t ) , we
r e p r e s e n t J ( f )a s a n i n t e g r a l o v e r D. T h e s u b s t i t u t i o n of v a r i a b l e s
= 'p (1; Zo, Yo). Y = II,( t ; rot Yo) (4)
in ( 3 ) m a p s D onto Dt. U e thus obtain
where
A I f ; so, Yo) = D (T( t ; 50. YO). $(t: 20. YO))
D(zo+ YO^
is t h e J a c o b i a n of m a p p i n g (4). This J a c o b i a n h a s b e e n c o m p u t e d i n QT,
w h e r e i t i s s h o w n (QT, 83.5, L e m m a 6) t h a t
It f o l l o w s f r o m t h e l a s t e q u a l i t p t h a t A ( t ; zo, yo)+ 0, i . e . , (4) i s a regular

mapping, and
F v (5), J ( t ) =
(W
5 pAdxo,2yo. S i n c e t h e r e g i o n of i n t e g r a t i o n i n t h i s i n t e g r a l
i s i n d e p e n d e n t of t , a n d t h e i n t e g r a n d is d i f f e r e n t i a b l e , J ( f )i s d i f f e r e n t i a b l e
and
=
J' (I) 1 5 & [ p (9( t ;
( D)
ZO, yo), 9 ( t ;50, YO)) ( t ;q,yo)] dxo &/a. (8)
L e t u s c o m p u t e t h e int -grand i n ( 8 ) . U s i n g (7) and t h e obvious r e l a t i o n s
403
Ch.XII1. LIhlIT CYCLES OF SYSTEMS DEPENDING ON A PARAMETER I
we find
Since J ( t ) = O , we have J(0) = 0 . The l a s t equality should hold for

e v e r y region D , s o that the integrand in (10) must identically vanish,
i . e . , at any point (z, y) of G we should have
On the o t h e r hand, it follows f r o m (8) and (9) that if (11) is satisfied,

J ( t ) 3 0 f o r any D. W e have thus established that identity (11) is the
n e c e s s a r y and sufficient condition for (1) t o b e an i n t e g r a l invariant. I
R em a r k . Condition (11) can b e written in the f o r m v+ av = 0.
For an n-th o r d e r s y s t e m %=P,(zi, x2, .. ., x n ) , i 1, 2 , . . . n, the function
p (z,, x2, . . . , x,,) is the density of the integral invariant i f and only i f the
following identity is satisfied ( s e e /30/):
- + ? L O .L
,=i
L e t G, a s before, b e a region consisting of whole paths of ( A ) .

D ef i n i t i o n 32. System (A) is said to be conservative in G it has
an integral invariant with positive density in this region.
Consider a s y s t e m (A) which is conservative in G . Let 1$ p(z, y)dxdy,
(D)
where p(z, y) > 0, b e i t s integral invariant. The paths of ( A ) then coincide
with the paths of the s y s t e m
dx
z=P(X, Y)P@,y ) = f i , g = P ( z , Y)Q(z,y)=Q. (A)
By ( l l ) , we have i n G
-aQ
=_- aP
dy a+ *
T h e r e f o r e , i f G is a s i m p l y connected region, a single-valued function

H (z, y) exists in G which s a t i s f i e s the equalities
and (A) m a g be written in the f o r m

5 33. CLOSED P.iTHS OF COhSFRVATIVE SYSTEAIS
D e f i n i t i o n 33. A dynamic system of the f o r m
where H (J. y) i s a single- valued function defined in G is called a Harniltonian

system in G .
F r o m P e f i n i t i o n s 32 a n d 3 3 a n d c o n d i t i o n (11) it f o l l o w s t h a t a H a m i l t o n i a n
s y s t e m is a p a r t i c u l a r ~zaseof a c o n s e r v a t i v e s y s t e m . The d e n s i t y p of t h e
i n t e g r a l i n v a r i a n t of a Elamiltonian s y s t e m m a y b e c h o s e n a s t h e n u m b e r 1;
t h e i n t e g r a l i n v a r i a n t 0.) t h e s y s t e m i s t h e n r h e area of t h e r e g i o n . A
H a m i l t o n i a n s v s t e m h a s a g e n e r a l i n t e g r a l H (z, y) = C ( s e e QT, 5 1.13).
Ke h a v e t h u s e s t a b l i s h e d t h a t if ( A ) i s a c o n s e r v a t i v e s y s t e m i n a
s i m p 1 v c n n e c t e d r e g i o n G, and p (1.y) is t h e d e n s i t y of i t s i n t e g r a l
i n v a r i a n t , (-A) i s a H a m i l t o n i a n s y s t e m i n G. p (2,y) i s t h e i n t e g r a t i n g f a c t o r
of t h e d i f f e r e n t i a l e q u a t i o n Q (3.y) dx - P (z. y) dy = 0 c o r r e s p o n d i n g t o s y s t e m
( - A \ . It i s r e a d i l y s e e n ,'.hat if a n d 0 a r e a n a l y t i c a l f u n c t i o n s , II (3, y) i s
also a n a n a l y t i c a l function.
If G is not s i m p l y c o n n e c t e d , r e l a t i o n ( 1 2 ) i s n o t e n o u g h t o e s t a b l i s h t h e
e x i s t e n c e i n G of a s i n g l e - v a l u e d f u n c t i o n H (2, y) w h i c h s a t i s f i e s c o n d i t i o n s
(13). In t h i s case w e c a n o n l y s a v t h a t s u c h a f u n c t i o n e x i s t s i n a n y s i m p l y
c o n n e c t e d s u b r e g i o n of G.
In t h i s s e c t i o n w e w i l l o n l y c o n s i d e r t h e case of G w h i c h is an " a n n u l a r "
r e g i o n c o m p l e t e l y f i l l e d with c l o s e d p a t h s ' of s y s t e m (-4) e n c l o s i n g o n e
a ' i o t h e r . W e w i l l s h o w that i n t h i s case (-A) i s a c o n s e r v a t i v e s y s t e m i n G,
i . e ., it h a s a n i n t e g r a l i n v a r i a n t w i t h p o s i t i v e d e n s i t y p (J, y) i n t h i s r e g i o n ,
and that t h e s y s t e m
--'pp.
dr
dt dY=pQ
dt
i s a Hamiltonian s y s t e m
TILe o r e m 7 4 . Let ( A ) be an analytical s y s t e m , and G a closed annular
region coinpletely filled with concentric closed paths of system (A). There
exists an analytical function p (2,y) defined in G which satisfies the condition
i. e . , system is cmsevoative in G , and the system

-
ds =p P ,
dt
dy_pQ
dt
is Haniillonian.
P r o o f . C o n s i d e r a n anal-y-tical a r c without c o n t a c t 1 c o n t a i n e d i n C ,
w h i c h c o n n e c t s t h e p o i n t s of t w o c l o s e d b o u n d a r y p a t h s (see F i g u r e 172;
t h e e x i s t e n c e of t h i s a r c is p r o v e d i n QT,
$ 19.5, Lemma 6 ) . L e t 2 = f (s). y = g (s), a , < s , < b
b e t h e p a r a m e t r i c e q u a t i o n s of the arc I
( f and g are a n a l y t i c a l functions). A s we
know, e v e r y c l o s e d p a t h of G has p r e c i s e l y
o n e c o m m o n point w i t h 1 . L e t T (s)b e t h e
p e r i o d of t h e c l o s e d p a t h w h i c h crosses
t h e a r c 2 a t a point c o r r e s p o n d i n g t o t h e
v a l u e s of t h e p a r a m e t e r .
L e t x = (C q.yo). y = Q ( t ;zo. yo) b e t h e
FIGC'RE 172 s o l u t i o n of s y s t e m ( A ) s a t i s f y i n g the initial
405
I
conditions 'p (0; xo. yo) =z0, 9 (0; xo, yo) = y o . Consider the mapping T, defined
b y the equalities
z=a)((t, s), y = Y ( t , s), (15)
where
The functions 4, and Y a r e analytical, and T m a p s the region R of the

plane ( t , s) defined by the inequalities
a,<s,<b, O<t,<r(s), (It)
onto the region G . The mapping T i s r e g u l a r at all the i n t e r i o r points of H

I
( s e e QT, $3.5, Lemma 8), but i t is not one-to-one, since e v e r y point of
the a r c 1 corresponding to the value s ( a < s < b ) of the p a r a m e t e r is the image
of two points (0, s) and ( t (s), s ) belonging t o the boundary of the region under
t h i s mapping. This follows f r o m the obvious relations
Let
Q(0, s ) = Q ( T ( s ) , s), Y(0, s ) = Y ( r ( s ) , s). (17) I
t
-$ [ P ; ( Q ( t , r)'f'(!, s))+QL(m(t. *), Y ( f ,*)>I
r ( t , s)=e 0 at. (18)
The function r ( t , s) is defined in region R of the plane ( t , s). In virtue of
relations (15), i t may be considered a s a function of z, y defined in G .
Let the corresponding function be p (z, y).
We thus take
P (z7 Y) = r (1 (5, LA, s (z. 59). (19)
where t = t (I,y) and s = s (5, y) a r e the functions defined by (15).

W e will f i r s t show that the function p ( x , y) defined by ( 1 9 ) is single-valued
in G . If the point ( x , y ) does not belong to the a r c without contact 1 , i t c o r -
responds p r e c i s e l y t o one point (t, s) of the region R and therefore t o a
definite value of the function r . Let now M (I,y) be a point which belongs
t o the a r c 1 and corresponds to the value s of the p a r a m e t e r . In t h i s c a s e ,
a s we have s e e n before, the point M (z, y) is the image of two points of the
region R under the mapping, namely the points ( 0 , s) and ( T (s), s). T o
establish the single-valuedness of p ( x , y). i t s u f f i c e s to show that r (0, s) =
= r (T (s), s). But r (0, s) = 1 by (18). On the other hand,
'f'm (9 (t: f? g), 9 ( t ; f, g 1) + 0; ('p ( t ; f, g ) , $ ( t ; f, d ) l dt
is equal, a p a r t f r o m the factor T (s), t o the c h a r a c t e r i s t i c index of the closed

path LM passing through the point M ( s e e 13.3, Definition 1 7 ) . Since the
path LM belongs to the family of the concentric closed paths filling the region
G , i t is not a limit cycle and i t s c h a r a c t e r i s t i c index, together with the l a s t
intepral, vanishes ( s e e 5 13.3, (31) and 1 2 . 3 ) . But then, by (18), r (r (s), 8 ) -
-= 1 = r (0, s). W e thus established that the function p (z,y) is uniquely defined
by (19) in G.
406
S 23. (:LOSED P.4THS OF CONSERI.-\TIVE SYSTEhIS
We w i l l now s h o w t h a t t h i s f u n c t i o n f u l f i l l s t h e p r o p o s i t i o n of t h e t h e o r e m ,
i . e . , s a t i s f i e s i d e n t i t y ( 1 1 ) . By (18) a n d ( I S ) w e h a v e
(20)
5 I u l t i p l y i n g t h e e q u a l i t i e s i n ( 2 0 ) b y P ( I , y) a n d Q ( I , y), respectively, adding

t h e m up, a n d r e m e m b e r i n g t h a t
w e obtain
i . e . , relation (11).
We w i l l now s h o w t h a t t h e s y s t e m
-dd =
r
t pP=P, g=pQ=Q (A)
i s a H a m i l t o n i a n s y s t e m i n t h e r i n g G. S i n c e p > 0 , t h e p a t h s of ( A ) c o i n c i d e
with t h e p a t h s of ( A ) , i . e . , t h e y a r e c l o s e d p a t h s . F u r t h e r m o r e , i n v i r t u e
of ( X I \ ,
Q(0,P) U(P. V) = 0,
dr OY
I.t?.,
Let .I.~O(I,>. yo) b e a f i x e d , a n d N(s, y) a n a r b i t r a r y point i n G. L e t y b e a

smooth c u r v e c o n t a i n e d i n G w h i c h goes f r o m Moto JI. W e w i l l s h o w t h a t
the line integral
i s a s i n g l e - v a l u e d f u n c t i o n of I , y. T o t h i s e n d , it s u f f i c e s t o e s t a b l i s h
t h a t t h i s i n t e g r a l is i n d e p e n d e n t of t h e i n t e g r a t i o n p a t h y , i . e . , t h e i n t e g r a l
a l o n g a n y c l o s e d c u r v e C o n t a i n e d i n G i s z e r o . If t h e r e g i o n e n c l o s e d b y
t h e c u r v e C i s e n t i r e l y c o n t a i n e d i n G, w e h a v e [ odx- P*dy = 0 b y G r e e n s
():;
6
f o r m u l a a n d r e l a t i o n ( 2 1 ) . L e t now C b e t h e c l o s e d p a t h z = ( t ) , y = ( t ) of 4
s v s t e m ( A ) w h i c h p a s s e s t h r o u g h t h e p o i n t .\Io ( t h i s p a t h e n c l o s e s t h e i n n e r
b o u n r l a r v c u r v e of G , i.-e., i t is not h o m o t o p i c t o z e r o in G ) . L e t T b e t h e
p e r i o d of t h e f m c t i o n s g, and 9. T h e n
It follows f r o m the l a s t relation and f r o m ( 2 1 ) that the line i n t e g r a l vanishes

along a n y closed c u r v e C which belongs to G. But then the i n t e g r a l ( 2 2 )
i s independent of the integration pathy, i . e . , it i s a function of the point
M ( 5 , y). Let t h i s function be I f ( I , y), i. e ., we take
H (x,Y) = S6
CY)
( x , Y) d x - p ( x , Y) d ~ .
The function H (.T, y) defined i n t h i s way c l e a r l y s a t i s f i e s the relations

p = --, aH Q,?Xas , i . e . , ( A ) is a Hamiltonian s y s t e m . T h i s completes
the proof of the t h e o r e m .
In what follows, a c o n s e r v a t i v e s y s t e m w i l l be a n a n a l y t i c a l
system defined in a doubly connected region Gfor which
a l l t h e p a t h s c o n t a i n e d i n G a r e c l o s e d (they a r e evidently
concentric). These s y s t e m s evidently constitute a comparatively r e s t r i c t e d
c l a s s of conservative s y s t e m s in the s e n s e of Definition 3 2 . However, since
we w i l l only consider these s y s t e m s , w e w i l l use the general t e r m c o n -
s e r v a t i v e s y s t e m s in t h i s r e s t r i c t e d s e n s e f r o m now on. Similarly,
a H a m i 1t o n i a n s y s t e m in this section will be regarded a s a Hamiltonian
s y s t e m defined i n a doubly connected region which i s completely filled with
concentric closed paths.
%'e will a l s o consider s y s t e m s c l o s e t o a c o n s e r v a t i v e
s y s t e m i. e., s y s t e m s of t h e f o r m
I
ns
x = P ( + , Y ) + P P ( G Y, P), $=Q(x, y l + P q ( x , Y, P), (A,,)
where p is a s m a l l r e a l number, p and q a r e analytical functions of the

respective arguments, and system (Ao), i . e . ,
i s conservative. We will establish the s u f f i c i e n t c o n d i t i o n s to be

met by the functions p (2,y , p), q ( x , y, p) f o r a limit cycle of (A,,) to exist in
the neighborhood of one of the closed paths Lo of (Ao). In other words, using
o u r terminology, we w i l l establish the sufficient conditions f o r the creation
of a limit cycle L, of system (A,) f r o m the path LOof s y s t e m (Ao).
A s a p a r t i c u l a r c a s e of s y s t e m s c l o s e to a conservative s y s t e m , w e will
consider s y s t e m s c l o s e t o a H a m i l t o n i a n s y s t e m ,
where p and q a r e analytical i n the doubly connected region, and (Ho) i s a

Hamiltonian s y s t e m .
We will show that on passing f r o m the g e n e r a l c a s e of s y s t e m s c l o s e to
a conservative s y s t e m to the c a s e of s y s t e m s c l o s e t o a Hamiltonian s y s t e m ,
the sufficient conditions f o r the functions p and q a r e markedly simplified.
The situation i s particularly simple when the original conservative
s y s t e m is l i n e a r . In t h i s c a s e , we may take, without loss of generality,
408
5 33. CLOSED PATHS OF CONSER\'r\TI'I'E SYSTEIIS
( P o ) is a H a m i l t o n i a n s y s t e m a n d its p a t h s are t h e circles

xa f ya = c. (23)
A s y s t e m close to a l i n e a r c o n s e r v a t i v e s y s t e m h a s t h e f o r m
dr
x =- Y T P p ( x r y * p ) r ~==zP~(x.Y,P). (BP)
T h i s s y s t e m c a n b e i n v e s t i g a t e d e i t h e r a s a p a r t i c u l a r case of a s y s t e m
close t o a H a m i l t o n i a n s y s t e m , o r d i r e c t l y b y c h a n g i n g o v e r t o p o l a r
c o o r d i n a t e s . S i n c e t h e d i r e c t a p p r o a c h is v e r y s i m p l e i n t h i s case, a n d
a t t h e s a m e t i m e p r o v i d e s a b e t t e r i n s i g h t i n t o t h e g e n e r a l p r o p e r t i e s of
a s v s t e m close t o a c o n s e r v a t i v e s y s t e m , w e w i l l f i r s t i n v e s t i g a t e ( i n t h e
n e s t s u b s e c t i o n ) a s y s t e m of t h e f o r m (BJ b y c h a n g i n g o v e r t o p o l a r
coordinates.
T h e r e s u l t s of t h e p r e s e n t s e c t i o n m a k e i t p o s s i b l e to e s t a b l i s h i n
c e r t a i n cases t h e p r e s e n c e (or a b s e n c e ) of l i m i t c y c l e s i n n o n l i n e a r
d v n a m i c s y s t e m s . T h i s is a c c o m p l i s h e d b y t h e s o - c a l l e d P o i n c a r e'
m e t h o d or t h e m e t q o d o f a s m a l l p a r a m e t e r . A c c o r d i n g
i o t h i s method, a n o n l i n e a r d y n a m i c s y s t e m i s c o n s i d e r e d as a s y s t e m
close t o a c o n s e r v a t i v e s y s t e m . T h e e x i s t e n c e of l i m i t c y c l e s in t h e
n o n l i n e a r s y s t e m i s e s t a b l i s h e d b v f i n d i n g t h e c l o s e d p a t h s of t h e c o n -
s e r v a t i v e s v s t e m which c r e a t e t h e s e c y c l e s . T h e P o i n c a r g m e t h o d I S
n a t u r a l l v a p p l i c a b l e o n l y i f t h e n o n l i n e a r s y s t e m is i n d e e d close t u a
c o n s e r v a t i v e s y s t e m , 01- i f it c a n b e "fitted" w i t h a c o n s e r v a t i v e s y s t e m
b v some t e c h n i q u e .
2. Systems c l o s e to a l i n e a r conservative system
W e w i l l c o n s i d e r a s v s t e m of t h e f o r m
;--Y+PP(s.y.P), i = x + P q ( x . Y,P), ( B,)
:rhich is close to a l i n e a r c o n s e r v a t i v e s y s t e m (Bo) w h o s e p a t h s a r e t h e

c i r c l e s r"y*==C. T h e f u n c t i o n s p a n d q a r e a s s u m e d t o b e anal-ytical i n
t h e n e i g h b o r h o o d of t h e point (O,O, 0 ) . S y s t e m s of t h e f o r m (B,) a r e o f t e n
e n c o u n t e r e d i n a p p l i c a t i o n s . T h u s , for e x a m p l e , t h e e q u a t i o n
..
1+3= Pf(Z, 4,
..
u h i c h is
J + Z - 0,
close f o r small p t o t h e e q u a t i o n of t h e h a r m o n i c o s c i l l a t o r
g i v e s i n t h e p h a s e p l a n e ( 3 , y) ( s e t t i n g y =- ii
x=-y. y=Lr-pCLf(z. -g),
a n d t h i s i s e v i d e n t l y a n e q u a t i o n of t h e f o r m (B,,). T h e f u n c t i o n s p ( r , y, P) a n d
q ( r , y , p ) w i l l be a s s u m e d t o v a n i s h for r = y = 0, i . e . ,
P(o,o,Io=Q~o.o,~~=o. (24)
409
Ch.XIII. LIMIT CYCLES OF SYSTEMS DEPENDING Oh A PARAMETER
This assumption does not constitute a fundamental r e s t r i c t i o n . Indeed,

i
suppose that condition ( 2 4 ) is not t r u e . Consider the equations
---Y+P.P(Z, Y, p)=O, Z+W(Z, Y, p)=O. (25)

At the point x - y = p = O , t h i s s y s t e m s a t i s f i e s the conditions of the theorem
of implicit functions, and i t is therefore solvable for x and y. Let
s=f(I47 v=g(C1)
b e the solution of ( 2 5 ) n e a r the point (0, 0, 0 ) . Then
-g (P) + PP (f (PI? g (P),co = 0, f (P) +Pq (f (P). g (P)* co = 0,

where f ( p ) and g(p) a r e analytical functions, and
f (0) = g (0) =o.
Applying t o (B,) the substitution of variables
2- x + f (PO, Y = y + g (PI,
Ry ( 2 6 ) , p*(O, 0, p)=q*(O, 0, p)=O, i. e., condition ( 2 4 ) is satisfied for

equations ( 2 8 ) .
We will thus consider the s y s t e m (B,) assuming ( 2 4 ) . Let
P(Z,Y, P ) = P ( z , Y)+PPZ(Z, Y, P)* q ( ~ vY* P ) = ~ ( z ,Y)+PQZ(S, Y, P)* (29)
BY ( 2 4 ) ,
p (0. 0) = q (0, 0) = 0, pz (0, 0, P) = qz (0, 0, P) = 0. (30)
Using (29), we write t h e original s y s t e m (B,) in the f o r m

dX
x= ---Y+PP(G Y, P)= - Y f P P ( Z , Y)+P*PZ(G Y, P),
d (31)
J=z-t-lq(sc,
dr Y. P)=Z+Pq(z, Y)+P*qz(r, y, P).
Changing o v e r to polar coordinates x = p c o s O , y=psinO, we readily obtain
$==ppcosep(pcose, psine, p)+sineq(pcose, p s i n e , p)],
By (24), the right-hand s i d e of the last equation f o r IpI<p* (where p* is a

positive number) h a s the f o r m
1 + PF (P, 8, P),
where F ( p , 8, p) is an analytical function of i t s arguments (which need not
vanish at p = 0 ) . For s m a l l pr the l a s t expression does not vanish, and
410
933. CLOSED PA Ti;S OF COSSERV.4TIVE SYSTEhIS
'.VP mav t h e r e f o r e c h a n g e o v e r t o a s i n g l e e q u a t i o n
d!)
=P R (p, e. p). (33)
Lvhcre
R (P. e. ,
PI -
c o a 8 p ( p c o . 0 . p q i n 0 , p)-8inOq(pcoiR. ( > s i n @p)
I + P F ;P, e. p)
.
(34)
R ( p , 0, p) i s a n a n a l y t i c 2 1 f u n c t i o n w i t h a p e r i o d of 2 x i n 8 w h i c h v a n i s h e s a t
p :0 . T h e r e f o r e , p - 0 f o r e v e r y p is a s o l u t i o n of e q u a t i o n (33). S i n c e
t h i s s o l u t i o n i s i n d e p e r d e n t of 0 , it is d e f i n e d for all 0. B u t t h e n b y
T h e o r e m 1 of Appendix 1, t h e s o l u t i o n of e q u a t i o n ( 3 3 ) is a p r i o r i d e f i n e d
for all H . c ! , < ~ < ~ x , for all s u f f i c i e n t l v s m a l l p, for ipI<p*, say. L e t t h e
s o l u t i o n c o r r e s p o n d i n g t o t h e i n i t i a l c o n d i t i o n s Bo a n d po b e
P = f (0: eo, pa. PI,

where f i s a n a n a l y t i c a l f u n c t i o n of its a r g u m e n t s . Since f o r p = 0
equation (14) t a k e s t h e f o r m %= 0, w e h a v e f ( e ; Bo, po. 0) po. Therefore,
;(0; eo. pn. p) = p0 T- p i , (0: eo, p0, p ) .

Let eo = 0 , a n d t h e s o l u t i o n f (0; 0. po. p) w i l l b e w r i t t e n i n t h e form
P =Po -k P'Y (0; Po. p). (35)
Here ''(e; po, p ) i s a n anE.l.ytica1 f u n c t i o n i n t h e r e g i o n
o .<e G 2 3 , IPl<P*. lPl<P*.

T h e c o n d i t i o n of c o n t a c t b e t w e e n a p a t h of s y s t e m (E,,) a n d t h e r a y
0 7 c o n s t a t t h e point (2,y) h a s t h e form
. .
zy- yr = 9 + y' + p (42- py) = 0.
B e c a u s e of t h e p a r t i c u l i r c h o i c e of p* and p*, t h i s c o n d i t i o n is not s a t i s f i e d
in t h e r e l e v a n t region, i.e., t h e r a y s 8 = c o n s t are without c o n t a c t with t h e
p a t h s of t h e s s s t e m . It f o l l o w s f r o m f 3 5 j t h a t t h e s u c c e s s i o n f u n c t i o n o n t h e
r a y tl 0 h a s t h e form
p = P o i p'Y p x ; Po, p ) . (36)
T h e closed p a t h s of (BF; c o r r e s p o n d to t h o s e of p o # O for- bvhich
P'Y ( 2 % Po, p ) = 0,
i . e . , f o r w h i c h for p f 0
Y (2%; Po, p) = 0. (37)

S i n c e t h e s u c c e s s i o n f u n c t i o n i s a n a n a l y t i c a l f u n c t i o n of p, Y may be
i v r i t t e n i n t h e form
Y (2% p,. p) = Y e r r ; Po. 0)sp'u; (2n;po, 0) + ... (38)
41 1
QI.XII1. L I M I T C Y C L E S OF S Y S T E M S D E P E N D I N G O N A P A R A M E T E R
The continuity of Y shows that i f the equation Y (2n; po, p) = 0 h a s a root

po(p) which goes t o s o m e p1 for p 4 0 ( I p I / < p * ) , then
Y (2s; pi, 0)=o. (39)
Suppose that f o r s o m e pi, )plI<p*, the l a s t condition is satisfied and that,

moreover,
=,pl,+ 0.
apo
The t h e o r e m of implicit functions then shows that t h e r e exist p*> 0 and

6 > 0 which satisfy the following condition: if 1p1 <p*, the equation
h a s a unique solution po = po (p) such that I p J (p) - p i \ -== 6 and po (0) = pl. T h i s
evidently implies that for a sufficiently s m a l l p + O , (B,) h a s a single limit
cycle in a s m a l l neighborhood of the c i r c l e 2 +
ya = p:, and t h i s limit cycle
I
c o n t r a c t s to the particular c i r c l e for p - t . 0 . W e can naturally s a y that t h i s I
limit cycle of (B,) is "created" f r o m the path sa y* = p: of the original +
linear system.
xa+
Note that the equation Y (2n; po, 0)= 0, alongside with pi, may have other
solutions satisfying condition (40). If pz is one of these solutions, the path
y2 = p i of the original conservative s y s t e m a l s o c r e a t e s a limit cycle.
Let u s d e r i v e expressions f o r Y(2n; PO, 0) and Jy(2:j!'o) in t e r m s of the
I
functions p ( x , y, p) and ~ ( xy,, p) entering the right-hand s i d e s of ( B & . To
t h i s end, we expand the right-hand side of (33) in powers of p. The equation
thus t a k e s the f o r m
3 = PR,(P, e) +mz(P, e) + . .. (42)
Solution (35) of t h i s equation may be written in the f o r m
p = p o + p w e ; pa, o ) + p w , ( e ; poto)+ ... (43)
Inserting the l a s t expression in (42) and equating the right- and the left-hand
s i d e s , w e obtain
d'4
-= R~(pot e). (44)
F r o m ( 2 9 ) , (33), (34), and (42) i t follows that

R , (po, ~ ) = c o s O p ( p 0 c o s ~posinO)
, +sinB~(poccs8,posin8). (45)
On the other hand, since f (0; 0, po, p) po, w e have Ur(0; po, p) E 0 . In -
particular,
Y (0; pa, 0) = 0. (46)
Integrating (44) with the initial condition (46), using (45), and setting 8 = 2x,
we obtain the following expression f o r Y (2n; pa, 0):
Y (2%;po, 0)= s
2Z
n
[cos Op (po cos 0, posin e) +sin 6q (pocos 8, posin e)]d e . (47)
412
5 33. CLOSED PATHS OF COKSERC'.\TIVE SYSTEh!S
To obtain d y ~ ~ s ~ p " ~ "we

) , differentiate t h e l a s t equality with r e s p e c t t o po,
JPU
w h e r e p i , p i . 4.; q;a r e the values of the respective functions a t t h e point

Ipucos8, pa sin e). Simple manipulations then give
d ~ ( 2 PO,
~ ;0)=
JPO 0
@; q;) de +
f [( - p ; sin e e,
p ; cos e) sin
+- (q; s i n 0 - q; cos e) cos el de. (48)
The relations
d p (p,,cos 8, PO sin 0) -
dB
- ( -p ; sin 0 f p ; cos 0) p,,,
dq(p(cos8, posin8)-
d0
- (--;sin 0 &cos e)
po
show that the second i n t e g r a l on the right i n (48) is equal t o
Integration by p a r t s , r e m e m b e r i n g that p ( p o c o s e , p o s i n 8 ) and q ( p o c o s B , p o s i n 8 )

a r e periodic functions of 0 with a period of 2.2, e s t a b l i s h e s that t h e l a s t
i n t e g r a l is equal t o
-1 (p,,tos e, pa s i n e) cos e +- q (pa cost), p,, sin e) s i n el de,

Po
and by (41)it is t h u s equal to

--I Y (2.T; Po. 0).
PO
Therefore
?n
Po, 0) -
J Y (21;
"u
- [ [ p ; (pUcos 0, pUsin e) -+
0
i
+q;(pocosC), posine)]de--Y(2n; po, 0). (49)
PO
By ( 4 7 ) and (49), t h e conditions
d'p (2.7: pi. 0)
Y :2x; p,, 0) = 0 and a?o +O (50)
a r e equivalent t o t h e cocditions
?n
S ~p (p, cos 0 , p, sin e) cos 0 + q (PI cos 0, pI sin e) sin el de =0,
0
2%
S [ p ; (pi cos 0, p, sin e) +
0
q; (PI cos 0, pi sin de + 0.
T h e above r e s u l t s t h u s lead t o t h e following t h e o r e m .
41 3
Ch.XII1. LIMIT CYCLES OF SYbIEMS DEPENDING ON A PARAMETER
T h e o Y e m 75. If fm some pl, 1 p1 I < p*, conditions (50), OY equivalently

(51), are satisfied, the transformation f r o m system (Bo)to a sufficiently
close system (B,) creates one and only one limit cycle of (B,) f r o m the path
p = p1 of the original linear system ( R ~ in ) a sufficiently small neighborhood
of this path.
R e m a r k . If f o r p,, Y (217; pl, 0) = 0 and a y y 2 x : p o ) = 0, i . e . , only the
f i r s t condition in (50) is satisfied, it is no longer c e r t a i n that a limit cycle
is created f r o m the path p = p i .
T h e o r e m 75 is a local t h e o r e m in the s e n s e that it d e a l s with the creation
of a limit cycle in the neighborhood of one path of (Bo). W e will prove now
another theorem, which also r e l a t e s to s y s t e m s close t o l i n e a r conservative
s y s t e m s but is nevertheless a global t h e o r e m . Let a and b be s o m e positive
numbers, a < b.
T h e o r e m 7 6 . If the equation Y (2n; po, 0 ) = 0 has precisely s solutions
po = p i , i = 1, 2 , . . ., S , in the segment la, bl, each of these solutions satisfying
the conditions a 0 such that i f 0 < I p I p*,-=
the equation Y (2n; PO, p) = 0 h a s precisely one root in each of the intervals
( p i - 6 , pi + .
6 ) , i = 1, 2, . . , s. Let these intervals be designated f,. W e
may a s s u m e that no two of these Z i i n t e r s e c t and that they a r e all contained
inside ( a , b ) : these requirements a r e satisfied i f 8 is sufficiently s m a l l .
Let r be the s e t [ a , bl\fiIi
i=i
(i. e., the complement of the union of the s e t s 2,
to the segment [ a , b ] ) .
The proof is conducted by reductio ad absurdum. Suppose that the
t h e o r e m is not t r u e . Then t h e r e e x i s t s a sequence pi, i = 1,2,3, . . . ,
such that pi # 0, I pi 1 < p*, Iim p i = 0 , and (E,$) has m o r e than s closed
paths in the ring a,<p,<b, i . e . , the equation 4 (2n; PO. pi) = 0 h a s m o r e
than s r o o t s satisfying the condition a , < p , g b . Then t h e r e e x i s t s a t l e a s t
one root of t h i s equation which belongs to the s e t I?. W e designate t h i s
root p f ) .
Thus, p$i) E r, Y (231;p c ) , p i ) = 0 .
By passing t o a n appropriate subsequence, i f necessary, we can always
e n s u r e convergence of the numerical sequence pa. Let the sequence con-
v e r g e and l e t l i m pf) = p:. Evidently, p: E and Y (2n; p f ) , O)=O. But then p:
ha,
is a root of the equation Y (2n;p0,0) = 0 which does not coincide with any of
the r o o t s p i , i = 1, 2, ..
. , s, in contradiction t o the conditions of the theorem.
T h i s contradiction proves
3 . The g e n e r a l case of a s y s t e m c l o s e to a
conservative s y s t e m .. .
Let (Ao) be a conservative s y s t e m in s o m e doubly connected region G .
W e w i l l consider a c l o s e s y s t e m of the f o r m
414
& h e r e p (z,y, p), q (z,y , 11) a r e a n a l y t i c a l f u n c t i o n s i n G .
In 5 13.1, i n o r d e r tc d e f i n e t h e s u c c e s s i o n f u n c t i o n , we i n t r o d u c e d
c u r v i l i n e a r c o o r d i n a t e s i n t h e n e i g h b o r h o o d of t h e c l o s e d p a t h Lo i n which
Lo w a s d e s c r i b e d b y t h e e q u a t i o n n = 0 .
\C'e w i l l now s h o w t h a t if (X,j) i s c o n s e r v a t i v e i n G , a n d Lo i s s o m e
c l o s e d p a t h of ( A * ) , L o c G , we c a n i n t r o d u c e c u r v i l i n e a r c o o r d i n a t e s
s. n i n s o m e n e i g h b o r h o 2 d of t h e p a t h L o s o t h a t t h e p a t h s of (A") w i l l b e
described by the equati3n n = const.
T h i s i s obvious f r o m g e o m e t r i c a l c o n s i d e r a t i o n s . X f o r m a l proof
c a n h e f o r m u l a t e d as f c l l o w s . L e t x = Q, ( t ) , y = $ ( t ) b e t h e e q u a t i o n s of
t h e p a t h L o , and T ? ~ Cpc.riod of t h e f u n c t i o n s g, and q. b r e f i r s t d e f i n e
i n t h e n e i g h b o r h o o d of Lo c u r v i l i n e a r c o o r d i n a t e s s a n d m u s i n g t h e
+-quations
z-=q( s ) + m . $ ( s ) , y=$-(s)-m.q(s) (52)
(see $13.1, ( 6 1 ) . In t h e s e c o o r d i n a t e s , (A,) is d e s c r i b e d b y t h e d i f f e r e n t i a l
equation
dm
-=
ds R'P; m, 1.0. (531
L e t t h e s o l u t i o n of t h i s e q u a t i o n s a t i s f y i n g t h e i n i t i a l c o n d i t i o n m = m o for
s = Ohe
m = f * ( s ; mot PI (5-1)
(see ?+32.1,(RW),(10)).
S i n c e for p = 0 all t h e p a t h s of (A,) i n t h e n e i g h b o r h o o d of Lo a r e c l o s e d ,
f* (s; mo. 0) i s a p e r i o d i c f u n c t i o n of s with a p e r i o d of T f o r all mo, 1 mo I < m*,
w h e r e m* i s a s u f f i c i e n t l y s m a l l p o s i t i v e n u m b e r .
Lye now d e f i n e t h e c o o r d i n a t e s s, n , t a k i n g
s=s, m = f * ( s ; n, 0). (55)
S i n c e Lo i s a c l o s e d p a t h , w e h a v e /*(s; 0, 0) 0. By $ 32,
( s e e 533.1, (a), ( 1 3 ) , ( I - I ) , ( 1 5 ) ) . T h u s , f o r e v e r y s, f * ( s , 0, O)=O. an#

0,0)
df* 0.
(8.
T h e e q u a t i o n f * ( s , n, O ) = m is t h u s u n i q u e l v s o l v a b l e f o r n i n t h e n e i g h b o r h o o d
of t h e point n 0 w h e n m is s u f f i c i e n t l y small i n a b s o l u t e v a l u e , i . e . ,
e q u a t i o n s ( 5 5 ) c o n s t i t u t e a o n e - t o - o n e t r a n s f o r m a t i o n of c o o r d i n a t e s . T h e
r e l a t i o n s h i p b e t w e e n t h e c u r v i l i n e a r c o o r d i n a t e s s, n and t h e C a r t e s i a n co-
o r d i n a t e s x , y , i n v i r t u e of ( 5 2 ) a n d ( 5 5 ) , i s e x p r e s s e d by t h e e q u a l i t i e s
z = g , ( s ) - k f * ( ~ , n, O)+(s)=.(F(s, n), ~=$(s)-f*(~, n, O)i(s)=y(s, n)- (56)

It f o l l o w s f r o m the d e f i n i t i o n of f* (s,mo, 0) t h a t i n t h e c o o r d i n a t e s s, n t h e c u r v e s
n - c o n s t c o i n c i d e w i t h t h e c l o s e d p a t h s o f (Ao). E q u a t i o n s ( 5 6 ) f o r a f i x e d n t h e r e -
f o r e c o n s t i t u t e p a r a m e t r i c e q u a t i o n s of t h e s e p a t h s . In p a r t i c u l a r , n = 0 c o r -
r e s p o n d s to t h e p a t h LO of t h e original s y s t e m (&).
T h e p a r a m e t e r s c o i n c i d e s w i t h t h e t i m e t a l o n g L o . For o t h e r c l o s e d
p a t h s o f (Ao), h o w e v e r , t h e p a r a m e t e r s i n g e n e r a l d o e s not c o i n c i d e w i t h
415
Ch.XLII. LIMIT CYCLES OF SYSTEMS DEPENDING O N A PARAMETER
time, and the periods of motion along t h e s e paths i n g e n e r a l a r e different

and a r e not equal t o z. It is readily verified, however, that in the particular
c a s e of a l i n e a r s y s t e m (Ao), s always coincides with t and the period of
motion is constant for all t h e closed paths of (Ao). T h e right-hand s i d e s
of (56) have the s a m e period T in the variable s f o r all n.
It can b e directly verified that the functions c(s, n) and $(s, n) defined
by (56) p o s s e s s the p r o p e r t i e s 1 - 4 listed at the beginning of $32 ( s e e 932.1,
( 2 ) - ( 5 ) ) . W e w i l l t h e r e f o r e u s e the r e s u l t s of t h i s section, retaining the
s a m e notation.
In the coordinates s, n , (A,) is expressed by the differential equation
(57)
where R ( s , n, p) is an analytical function in the region

-co<s<+m, Inl<n*, IpI<p*
( n* and p* a r e sufficiently s m a l l positive numbers; see 9 32.1, (R,)).

The coordinates s, n a r e chosen s o that f o r p = 0, the functions n = const
a r e solutions of the equation (57). T h e r e f o r e R (s, n, 0) = 0. But then all
the t e r m s in the expansion of R (s, n, p) in powers of n, p in the neighborhood
of the point n = 0, p = 0 contain the f a c t o r p , i.e., this expansion h a s the
form
+ +
R ( S , n, = -401 ( s ) P -411 (4 pn Aoz (SI pa . . . +
(58)
A s in $32.1, we write
= f (s; 0, no, p) (59)
for the solution of equation (57) corresponding t o the initial condition n=no
f o r s = 0 . The succession function of (A,,) on the a r c without contact I
described by the equation s = 0 we designate f (no,p). Clearly,
f (no, p) = I(7; 0, no, PI. (60)

Let
d (no, p) = f (no, P) -no. (61)
F r o m the particular choice of o u r coordinates, f (s; 0, no, 0) = n o . T h e r e f o r e

d (no,0) = f (no,0)-no = O , and the series expansion of d (no,p) n e a r the point
(0, 0 ) h a s the f o r m
d (not p) = UOIP + ~ t i n o p+ UOZP' + . . ., (62)

s o that
i P),
d ( n o , ~ ) = p * d(no, (63)
where
di (no7 p) = UOI + Ull +nouozP + * (64)
A s we have noted before, the function d (no,p) and therefore dl (no, p) a r e

a p r i o r i defined f o r all sufficiently s m a l l no, p.
no = 0 c o r r e s p o n d s t o the closed path Lo of (Ao). Let d,(O, 0) = 0 . Then,
i f I no I and p a r e sufficiently s m a l l and p #O, we have d (no,p) f 0 . T h i s implies
416
5 33. ZLOSED PATHS OF CONSERC';\T'lt'E SYSTELLS
that t h e s y s t e m (A,) sufficiently close t o (Ao) h a s no closed paths i n the neighbor-

hood of t h e path L o of (&). T h u s t h e equality
d, ( 0 , 0) = o
i s the n e c e s s a r y cor,dition f o r t h e s y s t e m (A,) sufficiently close to (Ao)

t o have closed paths i n a sufficiently s m a l l neighborhood of LO.
T h e question of the number of closed p a t h s is equivalent t o t h e question
of the number of t h e r o o t s of t h e equation dl (no,p ) = 0 which a r e sufficiently
c l o s e to z e r o when p # C i s sufficiently small. By t h e t h e o r e m of implicit
functions, a s u f f i c i e 11t condition f o r the existence of only o n e s u c h
root i s
dl (0, 0 ) = 0 , din,(0. 0 )# 0. (65)
T h e o r e m 77. Let Lo be a closed path of systeni (A,,) corresponding to

no = v, and let the following conditions be satisfied:
d ; ( O , 0 )= V , d j n o ( O , 0)2.0. (66)
Then there exist e > 0 aiai 6 > 0 such that

(a) f o r any p , I p I < 6 , (A,) has one and only one closed path L , in r e( L o )
which contracts to L o f o r p + 0;
(b) this path is a strticturally stable limit cycle, which is stable when
p . 4 , (0,0 ) < 0 and unstable when p.dj,, (0, 0 ) > 0 .
Proof. By (63),
d;(O, O)=dl(O, 0). dj,,(O, O)=dino(O, 0). (67)
T h e last r e l a t i o n s show that conditions (66) are equivalent to conditions (65)

and proposition ( a ) follows d i r e c t l y f r o m t h e t h e o r e m of implicit functions.
Let u s now p r o v e proposition (b). L e t the closed path L , c o r r e s p o n d t o
no = h ( p ) ( p # 0, I p < 8 ) . , T h i s value of the p a r a m e t e r s a t i s f i e s t h e equation
dl (no. p ) = 0, i . e . , dl (h (p), p ) = 0 . h ( p ) is a n analytical function, and s i n c e
d t ( 0 , 0 ) = 0, we have h ( 0 ) = 0, i.e., the expansion of h ( p ) i n p o w e r s of p h a s
the form
h ( p ) = a i p f a,pa f . .. (68)
To prove s t r u c t u r a l stability of the c y c l e L p and to e s t a b l i s h i t s stability

c h a r a c t e r i s t i c s , w e have t o compute dk,(no. p) f o r n o = h ( p ) . F r o m conditions
(60), using (64), (65), and (67), we obtain
ug* ==0, 0 )# 0.
ui, = di, (0, 0)= djno(O. (69)
Differentiation of (62) w t h r e s p e c t t o no g i v e s
d,, (no, p ) = f ~ l + 3 ~~2 i n o p

T f .. f ,
whence
c',, ( h (p), p ) = p'u,, f ut,pa f 2uz,h ( p ) p T . . .
o r , by (68) and (69),
db,O(h(P). P)==CC.dC",(O, O ) + o ( p ) .
41 7
The l a s t equality shows that for sufficiently s m a l l p # 0, the derivative

d;, (h ( p ) , p) does not vanish and i t s sign is equal t o the sign of p.d;, (0,O).
T h i s signifies, in virtue of the r e s u l t s of Chapter V (see 12.3), that
proposition ( b ) is satisfied. The proof of the theorem is complete.
4. S y s t e m s c l o s e to a Hamiltonian s y s t e m
If the original s y s t e m is Hamiltonian, the conditions of the creation of

a limit cycle f r o m a closed path of the s y s t e m takes on a particularly
simple f o r m . We will derive these conditions using T h e o r e m 7 7 and the
relations of 32.
Let the s y s t e m close to a Hamiltonian s y s t e m be given in the form
The equations of the closed paths of the original s y s t e m (Bo) have the
f o r m H (z, y) = C. (B,) is considered i n a doubly connected region G filled
with the paths H (z, y) = C , where Ci < C <C2. Let Lo be one of t h e s e paths,
z = 'p ( t ) , y = J, ( t ) the motion corresponding t o this path, T the period of the
functions cp and $. In the neighborhood of L o , we introduce curvilinear
coordinates s, n described at the beginning of 933.3 (see (56)), in which
the paths of the dynamic s y s t e m (Bo) a r e described by the equations
n = const. Let the path Lo be described by the equation n = io. The ex-
pansion of d (no,p ) around the point n = Go, p = 0 then h a s the form
+
d (no, P ) =u o i ~ it (no-Lo) p +u O 2 p+ . ..
(see (59)- (62)). In t h i s case, the numbers d; (O,O),d;"@ (0,O) of T h e o r e m 77
evidently should be replaced with d; (no,
0) dLno (no,0).
Since d; (no,0) = u o i , equation
- ( 3 6 ) , 932.1, can be used t o compute d; (no,(I
replacing A (0, 0) with A (0, no). Since f o r a Hamiltonian s y s t e m
or, changing over t o a line integral,
(72)
where
(73)
Let Go be the region enclosed inside the c u r v e Lo. If (B,,)is defined

everywhere i n Go, using Green's function we can change o v e r f r o m the
418
33. CLOSED P A T H S OF COhSERL ;ITICE SYSTELIS
line integral t o a double integral. T h e c o n d i t i o n d;(Ko. 0) = 0 i s t h e r e f o r e

e q u i v a l e n t to t h e c o n d i t i o n
sG;
( l P i s ( Z , Y)-!-QiY(2, Y)ldZdY=o. (74)
If (B,) i s not d e f i n e d e v e r y w h e r e i n Go, t h e c o n d i t i o n d ; (no,0) = 0 is

e q u i v a l e n t to t h e c o n d i : i o n
J (no) = q, d x - p , dy = 0. (75)
(Lo)
Let u s now c o n s i d e r t h e n u m b e r d;F-O(&,0). S i n c e t h e n u m b e r Eoi s i n

n o w a y d i s t i n g u i s h e d from t h e o t h e r n u m b e r s no, d;,o(no, 0) c a n b e o b t a i n e d
b y d i f f e r e n t i a t i n g ( 7 2 ) w i t h r e s p e c t t o Go. T h u s ,
-
.
Let u s c o m p u t e 0 To fix i d e a s , w e a s s u m e t h a t a s t i n c r e a s e s , t h e
dG,>
c l o s e d p a t h Lo is t r a c e d i n t h e p o s i t i v e d i r e c t i o n . We h a v e
L e t LOh b e t h e c l o s e d p a t h of (Po) w h i c h c o r r e s p o n d s to t h e v a l u e G o t h of
t h e p a r a m e t e r , Gh t h e r e g i o n b o u n d e d b y t h e p a t h s Lo a n d Lob. T h e s i g n of h
is c h o s e n s o t h a t t h e p a t h LO i s e n c l o s e d i n s i d e L O t , ( F i g u r e 173). T h e n
J ( ~ o f h ) - J ( is ~ ) t h e l i n e i n t e g r a l (73) t a k e n a l o n g t h e b o u n d a r y of Gh, a n d
by Greens formula
J(nofh)-J(G~)= - 11 Ipi,(z, y)-!-qiU(s, y ) l d z d y . (78)

h
C h a n g i n g over to t h e c u r v i l i n e a r c o o r d i n a t e s s, n i n t h e l a s t i n t e g r a l
a c c o r d i n g t o t h e r e l a t i o n s Z = ? ( S , n), y = $ ( s , n), w e o b t a i n
J(;o-t-h)-J(%,)= -1 j [pix(;, $ ) + q i v ( G , $)]lA(s, n ) l d s d n , (79)

h
w h e r e A ( s , n) is the J a c o b i a n -.D ( s . n ) This J a c o b i a n , by a s s u m p t i o n , r e t a i n s

a c o n s t a n t s i g n n e a r t h e p a t h L o ; l e t A ( s , n)> 0 .
419
Ch.XIII. LIMIT CYCLES 0% SYSTEMS DEPENDING ON A PARAMETER
Because of the particular choice of the curvilinear coordinates s, n, the

expansion of ( 5 7 ) in powers of p and n - z o contains no t e r m s with p (see (58))
In particular, A i o ( s ) = O . In o u r c a s e , however, by $32.1, (32),
N + QI
d
Ai0 (4= (cp9 (cp, 9)-xIn (A (S, GO))?
d -
and P ; + Q ; = O . Therefore zln(A(s, no))=O, i . e . , A(s, Lo)f o r e v e r y i o i s
independent of s. and f o r e v e r y n we have
A ( s , n) = A (0, n). (80)
By assumption, the path Lo is t r a c e d in the positive direction a s t

i n c r e a s e s , and
I
T h e r e f o r e , the vector v(cph(s, KO), tpk(s, Go)) points inside the c u r v e Lo
( F i g u r e 1 7 3 ) . On the other hand, the direction of t h i s vector c o r r e s p o n d s
to increasing no on the curve z = G ( s , no), g=q(s, no), where s is constant.
T h e r e f o r e , since Lo is enclosed inside LO*,we have h < 0 . Changing o v e r
f r o m a double integral in ( 7 9 ) t o two successive integrations and seeing
that h < 0 and A ( s , n ) = A ( O , n)> 0, we obtain
where G = G ( s , n). $=$(s, n). Hence,
The above considerations and T h e o r e m 77 evidently lead to the following

theorem, f i r s t formulated by Pontryagin (see / 3 1 / ) .
T h e o y e m 78. Let Lo be a closed Dath of the Hamiltonian system
x=cP(t), Y=tpU)
the motion corresponding to this path, 'c the peyiod of the functions 'p and 9 ,
Go the region enclosed inside the path Lor and
420
$ ?:3. CLOSED P A l H S OF COSSERC',\TII'E SYSTEhlS
a sj*stevi close to a Hauiiltonian systetii (p is a small paratnetel-1. Then, if
there exist e == 0 and 6 >: 0 such that

(a) fbr ecery p, I p 1 < 6, (R,) has one and only one closed path L , in u, (L,,),
and this L , contracts to L o for p+O;
(b) this path is a structurally stable limit cycle, which is stable fbr
p l < 0 and iinstable far 111 > 0.
R e ni a r k 1. If (B,) i s defined only in the neighborhood of the path L o ,
and not e v e r - w h e r e in Go, condition ( 7 4 ) should be replaced with the equality
f 141 (T (4,IC- ( s ) ) 9 ' (S)-P1 (T (4,$ (4)

$'(s)l d s = 0
(see ( 7 5 ) ) .
R e m a r k 2. It is readily s e e n that T h e o r e m 7 5 (see S 3 3 . 2 ) , relating t o
ds
s v s t e m s which a r e c l o s e to the linear conservative s y s t e m - dt
= --y, 9at
= t,
is a p a r t i c u l a r c a s e of T h e o r e m 78. Indeed, the equation of a path of t h i s

l i n e a r s y s t e m h a s the f o r m x = pi cos t , y = pt sin t . Inserting t h e s e functions
for q and $ in (71) and ( 7 5 ) , w e obtain relations (51) of s 3 3 . 2 .
In conclusion of t h i s section, we wish t o comment on the c r e a t i o n of l i m i t
c y c l e s f r o m a focus o r a c e n t e r . We have repeatedly noted the deep anaIogy
between the investigation of a dynamic s y s t e m in the neighborhood of a
closed path and the investigation in the neighborhood of a focus o r a c e n t e r .
T h i s analogy a l s o extends to the creation of limit c y c l e s .
Suppose that the original dynamic s y s t e m (Ao) h a s at the origin an equi-
librium s t a t e with pure imaginary c h a r a c t e r i s t i c numbers ( i . e., a multiple
focus o r c e n t e r ) . Consider the modified s y s t e m
-_
dr
dt P + ppt T pap, f . .. , dt = Qf p41 ip'42 4*' (A,)
Changing o v e r to polar coordinates p , 0 and replacing the s y s t e m with a

single equation, w e obtain a differential equation analogous t o ( 5 7 ) in $ 3 3 . 3 :
where the coefficients R i , a r e periodic functions of 0 with a period of 2 x ,

and the series in the right-hand s i d e of the l a s t equation converges f o r
all sufficiently s m a l l p and p (see $24.1, ( 5 4 ) ) .
s
A s in 33.3, we s e e k i i solution f ( e ; 0, po, p) of (R,) which s a t i s f i e s the
condition j ( 0 ; 0, pa, p) ZE po in the f o r m of a series
P= ~loPT~Otll+~?0Pf+~IIP~+. f ..
where the coefficients u i j a r e functions of 0 satisfying r e c u r s i v e formulas
analogous t o equations ( 14) 3 2 . 1 ) . (s
Ch.XIII. LlhlLr CYCLES OF SYSTEMS DEPENDING ON A PAKAMEIEK
Considering the succession function p = f (2n; 0, po, p) and the function

d (po, p) = f (2x; 0, po, p) - p o , we can use Newton's d i a g r a m to derive the
sufficient conditions of the creation of limit cycles f r o m a multiple focus
o r c e n t e r , as i n S32.4.
Since the succession function constructed in the neigkborhood of a
focus o r a c e n t e r h a s a number of specific p r o p e r t i e s (see 324.1,
L e m m a s 1, 2, 51, the specificity is a l s o extended to the function d ( p 0 , k).
W e will not consider t h i s aspect of the m a t t e r , however.
Note that the function UC (0, po, p) constructed in 9 3 3 . 2 in connection with
a s y s t e m c l o s e to the l i n e a r dynamic s y s t e m (31) is defined f o r all suf-
ficiently s m a l l po and p. Therefore, the s a m e function can be applied
t o investigate the creation of limit cycles f r o m equilibrium s t a t e s of the
type of a c e n t e r , a s well a s from closed paths of the original conservative
system.
422
C h a p t e r XI\'
THE APPLICATION O F THE THEORY O F
BIFLJR CA TIOiVS T O T H E I i W E S T I G A TION OF
PAR TICC'LAR DY,VA,IIIC SYSTE.IIS
IN TROUUCTIOX
In this c h a p t e r , we c o n s i d e r s o m e examples of dynamic s y s t e m s c o n -

taining p a r a m e t e r s . Almost all s y s t e m s of this kind arose in connection
with p a r t i c u l a r physical or engineering topics. The main problem which
is encountered in applications is to e s t a b l i s h the p a r t i t i o n of the s p a c e
of p a r a m e t e r s into r e g i o n s corresponding to identical qualitative s t r u c t u r e s
of the dynamic s y s t e m . b
L i hen the p a r a m e t e r s a s s u m e v a l u e s f r o m e a c h of t h e s e regions, the
systeni r e t a i n s the s a m e qualitative s t r u c t u r e and, in g e n e r a l , r e m a i n s
s t r u c t u r a l l y s t a b l e (or, in any c a s e , "relatively s t r u c t u r a l l y stable,"
i . e . , s t r u c t u r a l l y s t a b l e r e l a t i v e to the s p a c e of the dynamic s y s t e m s
spanned by the r a n g e of variation of the p a r a m e t e r s ) . The points in the
p a r a m e t e r s p a c e which lie o n the boundary of two r e g i o n s correspond to
s t r u c t u r a l l y unstable s y s t e m s , which - with the exception of isolated
points - are s y s t e m s of the first d e g r e e of s t r u c t u r a l instability.
The a i m of the p r e s e n t c h a p t e r is to f a m i l i a r i z e the r e a d e r with c e r t a i n
techniques of the theory of bifurcations which enable u s to obtain information
r e g a r d i n g the qualitative s t r u c t u r e of dynamic s y s t e m s and to analyze the
changes i n qualitative s t r u c t u r e of s y s t e m s following a change i n p a r a m e t e r s .
In the application of t h e s e techniques, i t is v e r y i m p o r t a n t to be a b l e to
e s t a b l i s h the qualitative s t r u c t u r e of the dynamic s y s t e m a t l e a s t for s o m e
p a r t i c u l a r v a l u e s of the p a r a m e t e r s . The examples d i s c u s s e d in t h i s
c h a p t e r are t h e r e f o r e salved by the application of c e r t a i n p a r t i c u l a r
techniques developed in Q T (the isocline configuration, D u l a c ' s c r i t e r i o n ,
the topographic s y s t e m , e t c . ) .
In broad outline, the qualitative techniques based on the t h e o r y of
bifurcations c a n be d e s c r i b e d a s follows:
1 ) If for s o m e values of the p a r a m e t e r s the s y s t e m h a s a n equilibrium
s t a t e with 1 > 0 (i.e., a uode or a focus), we f i r s t have to e s t a b l i s h the
e x i s t e n c e \ o r the a b s e n c e ) of numerical v a l u e s of the p a r a m e t e r s for which
the equilibrium s t a t e changes i t s stability, i . e . , to find the v a l u e s of the
p a r a m e t e r s for %*hichthe s y s t e m h a s a n equilibrium s t a t e with p u r e
imaginary c h a r a c t e r i s t i c roots. The p r o c e d u r e d e s c r i b e d in S 2 5 . 3 e n a b l e s
us to detect in this case \if the equilibrium state is not a c e n t e r ) the
c r e a t i o n of a l i m i t cycle and thus to identify the r a n g e of p a r a m e t e r v a l u e s
f o r which the dynamic s y s t e m a p r i o r i h a s a l i m i t c y c l e .
Ch.XIV. APPLICATIONS OF THE THEORY OF BIFURCATIONS
2 ) If the equilibrium s t a t e s cannot be determined by e l e m e n t a r y

techniques, we s e a r c h f o r p a r a m e t e r v a l u e s a t which the equilibrium
s t a t e s have maximum multiplicity.
Investigation of the possible c h a r a c t e r of the equilibrium s t a t e s f o r
p a r a m e t e r values c l o s e to the values which correspond to maximum
multiplicity enables u s to e s t a b l i s h in c e r t a i n c a s e s all the a l t e r n a t i v e s
r e g a r d i n g the number and the type of the equilibrium s t a t e s .
3 ) If the qualitative s t r u c t u r e h a s been established by s o m e technique
in two different points of the p a r a m e t e r space, the transition f r o m one
of these points to another w i l l enable u s to e s t a b l i s h , say, the existence
of a saddle-to-saddle s e p a r a t r i x and hence the possibility of c r e a t i o n of
limit cycles.
The l a r g e s t difficulties i n qualitative investigation of dynamic s y s t e m s
a r e encountered when we attempt to prove the a b s e n c e or the p r e s e n c e of
l i m i t c y c l e s c r e a t e d f r o m a "condensation of paths." A complete a n a l y s i s
i s t h e r e f o r e often impossible.
S34. EXAMPLES
E x a m p l e 1 2 (the c r e a t i o n of a l i m i t cycle f r o m a multiple focus).

-$= 6xy + p ( 4 - b - 2 ~ -2' +2zy +3 ~ 3 ,

(1)
$ =4 - h - 2 y - k ' f 2xy + 3y'- 6p.7~4,
where p>O is sufficiently s m a l l .

The v e c t o r field of this s y s t e m is obtained f r o m the vector field of the
system
dx
-
dt = 6xy, -$-=4-&-2q- 225+ b y + 3y' (2)
by rotation through an angle tan-' p ( s e e S 3 . 2 ) . It is obvious that the

ordinate a x i s z = 0 is a path of s y s t e m ( 2 ) . System ( 2 ) h a s two equilibrium
s t a t e s , the points A (1, 0) and B (- 2, 0). They a r e both simple equilibrium
s t a t e s (A # 0). The c h a r a c t e r i s t i c equation of the equilibrium s t a t e A (1, 0)
is Aa + 36 = 0, i.e., A (1, 0) is a n equilibrium s t a t e with p u r e imaginary
c h a r a c t e r i s t i c n u m b e r s . Since ( 2 ) is a n analytical s y s t e m , A is e i t h e r a
multiple focus o r a c e n t e r . The c h a r a c t e r i s t i c equation of the equilibrium
state B (- 2, 0) is A* + + 61 72 = 0. Its r o o t s a r e = - 3 f 3 i p , i.e.,
B (- 2, 0) is a simple stable focus of s y s t e m ( 2 ) .
System (1) h a s i t s equilibrium s t a t e s at the s a m e points A (i, 0) and
B (-2, 0 ) a s s y s t e m (2) ( s e e footnoteonp. 210). Clearly, if p is sufficiently
s m a l l , B (-2, 0) is a s i m p l e s t a b l e focus of s y s t e m (1) a l s o .
To investigate the equilibrium s t a t e A (1, 0), we w i l l u s e the r e s u l t s of
Chapter IX. Moving the o r i g i n to the point A and r e v e r t i n g to the previous
notation, we obtain f r o m (1)
dt
-=
dt -.6~+66y--~*+(6+21r)~y+311~,
(3)
= --6~-6py-- 29 + (2-6p) q/+3y',
424
4 34. EXlh!PLES
and f o r p = O this s y s t e m r e d u c e s to
dX
-=
dl 6y+&y, $=; -6t-2.2?+2xy+3y2. (4 i
For s y s t e m ( 3 ) , u (p) = P; f Q; = - 12p, u'(0) = - 12 < 0. The value of a3
for ( 4 ) is computed using equation (76), S24.4; i t is found to be &, i.e.,

a 3 > 0 .: The table i n S 2 5 . 3 thus shows that the point .-(1,
l 0) is a n unstable
multiple focus of systeni ( 2 ) and - f o r a s m a l l positive p - a stable focus
of s y s t e m (1) containing a n unstable limit cycle in i t s neighborhood.
Let u s now investigate the behavior of the paths of s y s t e m (1) a t infinity,
following the s c h e m e described in QT, 5 1 3 . 2 . Applying the transformation
r-:--.y1 ,v ~ ~ w e e s t a b l i s h that the "ends" of the z a x i s a r e not equilibrium
1
s t a t e s of the s y s t e m . Applying the transformation z = f , y = to s y s t e m (I)
and multiplying the righr-hand s i d e s of the resulting s y s t e m by z , w e obtain
$= 3p- 2pz +- ( 3 + 2p) L' f 4pza 4 (2-2p) LVZ +

4 ( 4 p - 2 ) P - - ~ Z = ~ + 2 c a ~.+2c3 z2 P (L9, J), (5)
:;
-= -3- 3.(6p-2) uz + 2z* +- 2c*z + 2uza- 4;3 = 0 (c, 3).
Since 0 (c, z ) contains .a f a c t o r z , the a x i s z = 0 c o n s i s t s of paths of

s y s t e m ( 5 ) . To find the equilibrium s t a t e s lying on this a x i s , w e have to
consider the equation
P(L*,0) = 3 ~ i ( 3 + 2 p ) ~ , i ( 4 p - - ) o * + 2 ~ ~ = , 0 . (6)
The derivative 3; (L,, 0) = 3 +

2p f (8p - 4 ) u +
6ct h a s no real r o o t s f o r
s m a l l j . ~ . Therefore if p is s m a l l and p > 0, equation ( 6 ) h a s a single real
root q, and i t is readily s e e n that r,, < 0 and lim ro = 0. System (5) thus
P+Q
h a s a single equilibrium s t a t e on the a x i s := 0, B (c0, 0). Consider the
determinant
.=I3 $. 2~ t ( 8 p - 4 ) vo i 6 ~ [ 3- 3 +- ( 6 ~ - 2 )PO + 2141.

LVhen p is sufficiently s m a l l , ro is s m a l l and A (q,, 0) < 0, i.e., the
equilibrium s t a t e D (co, 0) is a saddle point of s y s t e m (5). T w o of i t s
s e p a r a t r i c e s coincide with the s e m i a x e s z = 0 adjoining the point D ( r o ,0).
The direction of motion along these s e p a r a t r i c e s is defined by the equation
dr;dt = P (c, 0). W e see f r o m t h i s equation that the two s e p a r a t r i c e s lying on
the a x i s z = 0 are a - s e p a r a t r i c e s . Therefore, the o - s e p a r a t r i c e s of the
saddle point D lie on the two s i d e s of the*axis z = 0, and the path configura-
tion of s y s t e m (5) n e a r the saddle point D c a n be schematically r e p r e s e n t e d
by the d i a g r a m shown in Figure 174.
The r u l e s f o r establishing the behavior a t infinity (QT, 1 3 . 2 ) now show
that the paths of s y s t e m (1) n e a r the equator are a r r a n g e d as shown in
Figure 175.
The indices Q;, and in particular as,a r e introduced in Chapter IX (E24.4, (61)). In this chapter i t is shown
that if 0 is 3 multiple focus and a.z 0 , t h e sign of a3 determines the stability of the focus.
425
Ch. X?V. APPLICATIONS OF THE THEORY OF BIFURCATIONb I
dx
FIGURE 174
BY (I), f o r x = o ,
4
-
A
1
dt =p(4-2y+3ya))0,
~ = 2 p ( x + 2 ) ( 1 - x ) , and$-=2(x+2) (1-4.
FIGURE 175
dv
i
and -;ii-=4-2y+-3yp>0. For y=O,
T h e r e f o r e the paths of s y s t e m ( 1 )
c r o s s the coordinate a x e s in the d i r e c t i o n s shown in F i g u r e 175.
We have thus established that s y s t e m (1) f o r a s m a l l p > 0 h a s two s t a b l e
foci A (1, 0) and B (- 2, 0) and a n unstable l i m i t cycle in the neighborhood of
the focus A . Using this fact, and taking into account the path configuration
a t infinity and the d i r e c t i o n of motion along the paths a t the points of i n t e r -
section with the coordinate a x e s , we conclude
that in the half -plane x < 0 s y s t e m (1) h a s a l i m i t
cycle C , u n s t a b l e f r o m t h e o u t s i d e to
which the s e p a r a t r i x of the saddle point D g o e s
f o r t -+-co, and in the half-plane x > 0 the
//
s y s t e m h a s a l i m i t cycle Cz s t a b l e f r o m t h e
o u t s i d e to which the s e p a r a t r i x of the saddle
point D' goes for t-++co. Evidently, any closed
path of s y s t e m (1 ) a o t h e r than Cl and C2, is
e n t i r e l y contained e i t h e r in the half -plane x < 0,
and then i t is enclosed inside C1 and e n c l o s e s
the point B , o r in the half -plane x > 0, and then
i t is enclosed inside C z and e n c l o s e s the focus A .
FIGUKE 176 It is r e a d i l y s e e n that s y s t e m (1) may only have
a finite number of closed paths. Indeed,
suppose that this is not so. Then t h e r e e x i s t s
an infinite s e t of closed paths {r}a r r a n g e d concentrically inside a cycle,
C , s a y . We c a n s e l e c t an infinite sequence of paths rl, rZ, r3,... with the
following p r o p e r t y : e v e r y s u c c e s s i v e path e n c l o s e s i t s p r e d e c e s s o r and
e v e r y path of the s e t {J?) is enclosed a t l e a s t inside one of the paths rr( s e e
QT, S l 6 . 9 , L e m m a 1 6 ) . L e t K b e the t o p o l o g i c a l l i m i t of the sequence
rL( s e e QT, Appendix, S l . 1 0 ) . It is r e a d i l y s e e n that e i t h e r K is a closed
path r*, such that closed paths l i e in any sufficiently s m a l l neighborhood
inside r* and no such paths a r e observed outside I'*, or K h a s the s t r u c t u r e
of a 0-limit continuum ( s e e QT, 523.2, Theorem 70), i.e., i t c o n s i s t s of
426
F 34. EXAh!PLES
equilibrium states a l t e r n a t i n g with s e p a r a t r i c e s which go to t h e s e

equilibrium s t a t e s . However, s i n c e (1) is a n analytical s y s t e m , i t cannot
have a c l o s e d path r* of this kind. The topological l i m i t K cannot have the
s t r u c t u r e of the null-limit continuum e i t h e r , because all the equilibrium
states of s y s t e m (1) are foci.
i V e haire thus e s t a b l i s h e d that s y s t e m (1) only h a s a finite number of
closed paths, and t h e r e is a t l e a s t one l i m i t c y c l e in the half -plane 1<0 and
at l e a s t two l i m i t c y c l e s in the half-plane t > O . T h e exact number of the
l i m i t c y c l e s of s y s t e m ( I ) cannot be e s t a b l i s h e d . It is clear, however, that
the number of l i m i t c y c l e s (counting e a c h a c c o r d i n g to its multiplicity) in
the half -plane I < 0 is odd, and the number of l i m i t c y c l e s i n the half -plane
I > o i s even. The configuration of the paths of s y s t e m (1) in a circle, which
is used a s a model of the Euclidean plane, is shown in Figure 176 (up to a n
even number of l i m i t c y - l e s ) .
E x a m p 1 e 13 (the c r e a t i o n of a l i m i t c y c l e f r o m a multiple f o c u s and a
s e p a r a t r i x loop, see / 2 0 / ) .
z=y. y'= --r+py+zy+z2+-y*. (7)
I t h a s two equilibrium s t a t e s , 0 (0.0) and d ( f , 0). -4 is a saddle point for

any p. The c h a r a c t e r i s t i c n u m b e r s of the e q u i l i b r i u m s t a t e 0 (0, 0) a r e
= 2 &
i.,.? {e - 1;
4 i t is t h e r e f o r e
1 ) a s t a b l e node for p.-< - 2;
2 J a s t a b l e focus for - 2 < p < 0 ;
3 ) a n unstable focus for 0 e p < 2 ;
4) a n unstable node for -,-2 d p .
For p = 0, 0 is a n eqiiilibriuni s t a t e with p u r e imaginary c h a r a c t e r i s t i c
n u m b e r s . To investigate i t s c h a r a c t e r , we will u s e , a s in the p r e v i o u s
example, the r e s u l t s of 1 2 5 . 3 . IVe have (I (p) = p> (I' (11) = 1 > 0. a3 is
computed f r o m (76), 1 2 4 . 4 , and is found to be equal to $ > O . Therefore,
for s m a l l p > 0, and a l s o for p = 0, the point 0 is a n unstable f o c u s of
s y s t e m (71, without any l i m i t c y c l e s in i t s sufficiently s m a l l neighborhood,
and for s m a l l p -== 0 the point 0 is a s t a b l e f o c u s with a n unstable c y c l e i n its
neighborhood (see table in 1 2 5 . 3 ) .
To investigate the situation a t infinity, w e f i r s t apply the P o i n c a r g
transformation I = 4 , y = +. hIultiplying the r i g h t -hand s i d e s of the re -
sulting s y s t e m by I, we obtain
-
di.
dt
= -u f z - pL'3 - y= - ut, - us, dz
-
dt
= -z- vz - p,* f L.za - usz. (8)
For 3 = 0, s y s t e m ( 8 ) h a s a single e q u i l i b r i u m s t a t e (0.0), which is a

s t a b l e node. The second equation i n ( 8 ) shows that the a x i s z = 0 is made up
1
of paths of s y s t e m (8). The PoincarC t r a n s f o r m a t i o n I = T, y= g i v e s the
system
du d:
-= 1+u-zspu;-3u2+u=. -= -u,?.
dt dt
Ch.XIV. APPLICATIONS OF T H E THEORY OF BIFURCATIONS
For z = 0, the l a s t s y s t e m h a s no equilibrium s t a t e s . Hence i t follows

that, in accordance with the r u l e s f o r investigation a t infinity (QT, 13.2),
the paths of s y s t e m ( 7 ) a r e a r r a n g e d a t infinity as shown in F i g u r e 1 7 7 .
FIGURE 111
3 FIGURE 178
To proceed with f u r t h e r investigation of s y s t e m (7), we w i l l u s e the

auxiliary s y s t e m
Its g e n e r a l integral, a s is readily seen, is
. . . .
in F i g u r e 178. One of the paths i s the loop i originating and terminating
a t the saddle point A , which is described by the equation
All the paths of system (9) enclosed inside the loop Lo a r e closed.
Consider the c o n t a c t c u r v e of s y s t e m s ( 7 ) and ( 9 ) ( s e e QT, 512.5).
I t s equation is
or
u2 ( p i x +l/) = 0. (12)
It c o n s i s t s of the points of contact of the paths of s y s t e m s ( 7 ) and (9)and

the equilibrium s t a t e s of these s y s t e m s . The contact c u r v e decomposes
into s t r a i g h t lines x + y + p = 0 and y = 0, but the line y = 0 constitutes a
428
5.34. EX i l P L E S
false contact, i.e., a path of s y s t e m ( 7 ) contacting a path of s y s t e m ( 9 ) a t 2.

point with y = 0 will inevitably cross this path f r o m one s i d e to another {see
QT, 512.5).
Since the s u m of the indices of the equilibrium states enclosed inside a
closed path is 1, any c l 3 s e d path of s y s t e m ( 7 ) e n c l o s e s the equilibrium
state 0 without enclosing A . This path t h e r e f o r e n e c e s s a r i l y c r o s s e s the
closed paths of s y s t e m , 5 ) .
Consider the i n t e r s e z t i o n of the c u r v e ( I l ) , p a r t of which is r e p r e s e n t e d
by the loop i, with the contact line z C y f p = 0. Inserting y = - z - p in
(1 1 ), we obtain the equation
Since f (z) = xf - 22 -. p h a s the r o o t s xi.?= 1 iz 1 m, f (z) > 0 for

kt <-1 and equation (13) only h a s one real r o o t . For p : - 1 equation ( 1 3 )
h a s a t r i p l e root J = 1, !..e., the contact line x -r y 4-p = 0 c r o s s e s the
c u r v e k 1 1 ) at the point d (1, 0). Thus, for 11 = - 1 the loop i h a s one point
in common with the contact line (the point -4). But then, a s w e s e e from the
position of the line x + - p = 0, the loop ?. d o e s not cross the contact line
~-y-p=Uforp<-l. Hence t h e c o n c l u s i o n t h a t f o r p < - I system(7)
h a s n o c l o s e d p a t h s . Indeed, l e t L be a c l o s e d path of s y s t e m ( I ) .
-4swe have s e e n before, L crosses the closed paths of s y s t e m ( 5 ) and
e n c l o s e s s o m e of t h e s e paths. T h i s m e a n s that t h e r e e x i s t s a closed path
of s y s t e m ( 9 ) which h a s a ( t r u e ) point of contact
K i t h the path L , namely the l e a s t path 2,
having a point of contact with L . L e t 31 be t h i s
point of contact. Jf lies inside the loop i on the
contact c u r v e (12), i.e., on the s t r a i g h t line
y = 0. This is a contradiction, however, s i n c e
all the points of this line correspond to f a l s e
contacts.
Having thus established that s y s t e m ( 7 ) h a s
no c l o s e d p a t h s for p S - 1, w e c a n find i t s
topological s t r u c t u r e . To fix i d e a s let p = - 1
(for p < - 1, the s t r u c t u r e is the s a m e , but f u r
p S - 2 the point 0 is a s t a b l e node, and not a
FIGCRE 17. p = --i < p*. f o c u s ) . The o - s e p a r a t r i c e s L 1 and L: of the
saddle point .4 (1, U) should extend froni the
unstable focus to infinity, s i n c e t h e r e a r e no
o t h e r a - l i m i t points in the plane. This, however, automatically fixes the
behavior of the a - s e p a r z t r i c e s L 3 and L,: one of them, L 3 say, winds onto the
focus 0 f o r t-+ + m, and the o t h e r , L , , g o e s to the s t a b l e node a t infinity.
Allowing for the d i r e c t i o n of the paths a t the i n t e r s e c t i o n points with the
coordinate a x e s , we obtain the path configuration shown s c h e m a t i c a l l y in
F i g u r e 179.
L e t u s now c o n s i d e r t.he path configuration of s y s t e m ( 7 ) f o r p = 3. In
this case, 0 (0. 0) is a n unstable node. The a b s c i s s a of the i n t e r s e c t i o n
point of the c u r v e (11) with the contact l i n e z y f 3 = 0 is d e t e r m i n e d ,
a f t e r eliminating y, f r o n i the equation f ( 5 ) = 33 - 3z2- 9z - 13 = 0. A
s t a n d a r d investigation of the function y = f (3) r e a d i l y shows that i t h a s a
single real root zo, and J:O > 3 (the c u r v e of f (z)is shown in F i g u r e 180).
429
-
Hence it follows that the contact line z +

y +- 3 = 0 does not c r o s s the loop
and, r e a s o n i n g as before, w e conclude that f o r p = 3 s y s t e m ( 7 ) h a s no
closed paths e i t h e r . The behavior of the s e p a r a t r i c e s of the saddle point
A ( 1 , 0) of s y s t e m ( 7 ) is thus fixed automatically ( a s in the c a s e p < - 1).
Indeed, the a - s e p a r a t r i c e s L3 and LI of the saddle point A may only extend
to the s t a b l e node a t infinity, and the closed c u r v e formed by these
s e p a r a t r i c e s should enclose the point 0. Consequently, one of the w -
s e p a r a t r i c e s of the saddle point A , Ll say, should go to the node 0 for
t - t - 00 and the other s e p a r a t r i x , L z , should go to the unstable node a t
infinity.
"t
Allowing f o r the direction of the paths a t the points of the coordinate

a x e s and at the node 0 and f o r the d i r e c t i o n of the s e p a r a t r i c e s a t the saddle
point A , we conclude that the path configuration of s y s t e m ( 7 ) f o r p = 3 can
be schematically illustrated by the d i a g r a m in Figure 181. It i s readily
s e e n that for p > 3 s y s t e m ( 7 ) s t i l l h a s no closed paths and i t s topological
s t r u c t u r e is t h e r e f o r e the s a m e a s f o r p = 3.
Let u s investigate the behavior of the s e p a r a t r i c e s Ll and L, a s the
p a r a m e t e r p v a r i e s f r o m -1 to 3. We should f i r s t note that on passing f r o m
a value pi of the p a r a m e t e r to s o m e value p z > p i , all the field v e c t o r s a r e
turned in the s a m e direction, specifically counterclockwise. 'This follows
f r o m the fact that by equations ( 7 )
Let the "first" i n t e r s e c t i o n point of the s e p a r a t r i x Li( L 3 )of s y s t e m ( 7 )

with the negative half-axis x be M i(p); the a b s c i s s a of this i n t e r s e c t i o n
point is z1 (p)( M , (p) and x 3 (p), r e s p e c t i v e l y ) . The negative half - a x i s z h a s
no contact points with the paths of the s y s t e m f o r any p. We may t h e r e f o r e
apply the r e s u l t s of $11.1. If we take into consideration the d i r e c t i o n s in
which the s e p a r a t r i c e s Ll and L , c r o s s the a x i s x , this l e m m a shows that
f o r i n c r e a s i n g p, z3(p)d e c r e a s e s and xi (p) i n c r e a s e s , i.e., the points M i(p)
and M 3 (p) move in opposite d i r e c t i o n s along the a x i s x . In accordance
with the previous r e s u l t s , xi (- 1)(2, (- I), and zl(3) > z, (3) ( s e e F i g u r e s 1 7 9
and 181). F u r t h e r m o r e , xi (p) and x 3 ( p ) are continuous functions (see
r e m a r k to L e m m a 3, 59.2). T h e r e f o r e , t h e r e e x i s t s one and only one value
330
5 34. EX;\! IPLES
of the p a r a m e t e r , p*, in the interval (-1, 3 ) such that zl (p*) = z3(p*). T h i s

signifies that the s e p a r a t r i c e s Ll and L3 corresponding to this p* coincide,
i . e . , f o r p = it* the systszm h a s a s e p a r a t r i x which f o r m s a loop. Let this
loop s e p a r a t r i x be L * . The value of the function o (2, y) = P; (2. y) + Q; (3. y ) a t
the saddle pointl(1,U)for p = p* is p* T 1. Since p* > - 1, we have p* - 1 > 0.
Therefore (see Theorem 14, 525.1), the loop L * i s unstable, i . e . , a l l the
sufficiently c l o s e paths enclosed inside this loop a r e s p i r a l s unwinding
f r o m the loop. For p = p*. the topological s t r u c t u r e of s y s t e m ( 7 ) o u t s i d e
t h e l o o p L* is deterniined unambiguously. The topological s t r u c t u r e of
the s y s t e m i n s i d e t h e 1o o p I-* depends on the number of limit c y c l e s
and the stability charac.:eristics of the equilibrium s t a t e 0, which a r e not
known. If p* ( 0 , then 0 is a stable focus and s y s t e m ( 7 ) h a s no closed paths
inside the loop L*, or e l s e i t h a s a n even number of such paths (counting
according to multiplicit:.esj. Lf, however, p* - 0 , 0 is a n unstable focus o r
node and s y s t e m ( 7 ) h a s a n odd number of closed paths inside the loop L * .
b-igure 1 8 2 shows the topological s t r u c t u r e of s y s t e m ( 7 ) f o r p = p*under the
assumption that this s y s t e m h a s no closed paths.
Applying the r e s u l t s 3 f Chapter XI> w e can t r a c e the behavior of the

s e p a r a t r i x loop a s the p a r a m e t e r v a r i e s n e a r i t s value p*. It follows from
the r e m a r k to Theorem 49 ( 5 2 9 . 3 ) that as the p a r a m e t e r p i n c r e a s e s
above it*, the s e p a r a t r i x loop L* d i s a p p e a r s , and a single unstable limit
cycle is c r e a t e d in i t s n5ighborhood (see F i g u r e 183, drawn under the
assumption that the s y s t e m h a s no o t h e r limit c y c l e s ) . A s the p a r a m e t e r p
d e c r e a s e s below p*, the s e p a r a t r i x d i s a p p e a r s without giving rise to
limit c y c l e s in i t s neighDorhood.
To follow the changes in the topological s t r u c t u r e of the s y s t e m as the
p a r a m e t e r v a r i e s , w e r e q u i r e the exact number of the l i m i t c y c l e s of the
s y s t e m f o r e v e r y p. The number of l i m i t c y c l e s may change only when
the s y s t e m p a s s e s through a bifurcation value of the p a r a m e t e r , and in such
a s y s t e m limit c y c l e s aye c r e a t e d e i t h e r from a multiple focus, o r f r o m a
loop of a saddle-point s e p a r a t r i x , o r from a c o n d e n s a t i o n o f p a t h s
( s e e $22, Example 8), or finally f r o m a multiple l i m i t cycle (if any). W e
will n e v e r t h e l e s s proceed with a tentative a n a l y s i s a s s u m i n g t h a t p* < 0
a n d t h a t s y s t e m (7) f o r e v e r y p h a s a t m o s t o n e c l o s e d
43 I
ch.XIV. APPLICATIONS OF THE THEORY OF BIFURCATIONS
I
path, which is m o r e o v e r s t r u c t u r a l l y s t a b l e . Asafirst
step, let u s consider what happens to a limit cycle of the s y s t e m when the
vector field is r o t a t e d .
R e m a r k r e g a r d i n g t h e v a r i a t i o n of t h e l i m i t c y c l e
f o l l o w i n g r o t a t i o n of t h e v e c t o r f i e l d
Let I
dx dy
~ = P ( XY,? P), x=Q(">
Y, PI (A,)
be a dynamic s y s t e m which f o r p = p o h a s a l i m i t cycle Lo. Suppose that Lo

is a s t a b l e l i m i t cycle with motion in the positive direction f o r i n c r e a s i n g t
( F i g u r e 184) and suppose that the field v e c t o r s r o t a t e in the positive s e n s e
a s p increases.
Through s o m e point M o of the cycle Lo w e d r a w an a r c without contact I ,
choose a point MI on this a r c , which l i e s inside Lo and is sufficiently c l o s e
to M O( F i g u r e 184), and consider the path Lithrough
the point M i . Li w i l l c r o s s the a r c 1 a t another
point M z , which lies between M iand M o .
Consider the s y s t e m (Aw), where p is sufficiently
c l o s e to pa, p > po. The path ZI of (A,) passing
through the point M ic r o s s e s the a r c without contact 1
a t a point k z ,which is sufficiently c l o s e to MZ. r h e
l i m i t cycle Lo of (ApJ is a cycle without contact f o r
(A,), and, a s f i n c r e a s e s , e v e r y path of (A,,o)
c r o s s i n g the cycle Lo w i l l e n t e r into the region F
bounded by the c u r v e Lo and the closed c u r v e c" con-
sisting of the coil M$Z of the path %$ and the segment I
FIGURE 184 M i @ , of the a r c 1 (in Figure 184, ? is diagonally
hatched). Since L" cannot leave 7 azd t h e r e a r e no
equilibrium s t a t e s in this region, I? contains one and
(by T h e o r e m 72, 132.4) only one closed path L"o of s y s t e m (Aw). We thus
conclude that a s the vector field of s y s t e m (AFo)is rotated in the positive
direction, the l i m i t cycle Lo c o n t r a c t s . Clearly, when the field is
rotated in the negative direction, the l i m i t cycle Lo e x p a n d s . A s i m i l a r
r e s u l t is obtained f o r a n unstable l i m i t cycle. Applying the s a m e reasoning
to a s e m i s t a b l e cycle (of even multiplicity) and using Theorem 71 (%32.4),
we conclude that as the vector field is r o t a t e d in one direction, the cycle of
even multiplicity d i s a p p e a r s , .and a s the field is rotated in the opposite I
direction, the cycle decomposes into two c y c l e s (a s t a b l e and an unstable
one), of which one c o n t r a c t s and the other expands.
After this digression, we can r e t u r n to s y s t e m (7). Thus, l e t p* -= 0 and
l e t s y s t e m ( 7 ) have a t m o s t one closed path, which is m o r e o v e r s t r u c t u r a l l y
stable. F o r p < - 1, the s y s t e m h a s the topological s t r u c t u r e shown in
F i g u r e 179; in p a r t i c u l a r , i t h a s no closed paths. For p ranging between
-1 and p*, s y s t e m ( 7 ) h a s no closed paths either. Indeed, a s p i n c r e a s e s
f r o m -1 to p* < 0, a closed path c a n be c r e a t e d only f r o m a condensation
of paths, but then i t is not a s t r u c t u r a l l y s t a b l e path. Thus, as p i n c r e a s e s
f r o m -CO to p*, the topological s t r u c t u r e of the s y s t e m r e m a i n s unchanged
432
S 34. EX-AhfPLES
( F i g u r e 179), the point; .Ifl and .If, of the s e p a r a t r i c e s L , and L , moving one
toward the other along the I a x i s .
F o r p = p*, the s e p a r a t r i c e s L, and L 3 m e r g e forming a n unstable loop.
Since the equilibrium s t a t e 0 is stable in this case, s y s t e m ( 7 ) may have no
limit c y c l e s under o u r assumptions, and i t s topological s t r u c t u r e is
correspondingly shown in Figure 182.
A s p i n c r e a s e s above p = p*, the s e p a r a t r i x loop c r e a t e s a n unstable
limit cycle. The s y s t e m may not have any o t h e r cycles, and i t s topological
s t r u c t u r e a s s u m e s the form shown in Figure 1 8 3 . A s p increase f r o m y * to
0, this cycle c o n t r a c t s and f o r p = 0 i t "collapses:' into the equilibrium
s t a t e 0 , which changes i t s stability a t this instant. Since f o r p = 3 the
s y s t e m h a s no closed p,&ths, i t may not have any closed paths f o r p -= 0
e i t h e r . Indeed, as I( d e c r e a s e s f r o m 3 to 0, a closed path may be c r e a t e d
only f r o m a condensaticm of paths, but then i t is not s t r u c t u r a l l y s t a b l e .
Thus, the topological s r r u c t u r e of the s y s t e m corresponding to p :;,O is shown
in Figure 181.
A s p d e c r e a s e s f r o m 03, the s y s t e m a t f i r s t h a s no closed paths, and i t
is only when the system crosses the value p = 0 that a n unstable l i m i t cycle
is c r e a t e d f r o m the focus 0, which then expands and f o r p = p* t r a n s f o r m s
into a s e p a r a t r i x loop which d i s a p p e a r s as the s y s t e m p a s s e s through the
value p = p*.
The r e a d e r is advised to w o r k out for himself a s i m i l a r ("tentative")
analysis assuming that p* > 0 or p* = 0 and that the s y s t e m a t any time h a s
the l e a s t possible number of closed paths.
E x a m p 1e 14 (the ci-eation of a l i m i t cycle f r o m the loop of a saddle-
node s e p a r a t r i x , see /34/).
dd t" = y ( ~ f p ) f 2 i y ' - l = P ( ~ , Y, p), ~ = - ~ ( z t p ) = Q ( Y,

~ p).
, (S,)
It is readily s e e n that the circle
ZZ-++-i=O
(which w e designate W ) is made up of paths of {S,).

The equilibrium s t a t e s a r e determined f r o m the equations P ( z , y. p)=O,
Q(.r. y. p) = O . T h e r e are two equilibrium s t a t e s on the y axis, A,,(O, a,,) and
8,( ( I , b,,), where
Up= -$+ / v > O , b,,= -$.-

I y q 7 c o .
The other equilibrium states a r e obtained f r o m the equations z + p = 0 and
x2 - ya - i = 0. For I p I > 1, these equations are contradictory. For
2, < 1 p 1 < 1, they d e t e r m i n e two additional equilibrium s t a t e s (other than
.-
A,, and B , , ) , C, (-p, c,,) and D, ( - p , - ~ , ) , where c,, = 1- 1 - p a . T'hus, f o r
0 < I p I < 1, the s y s t e m h a s four equilibrium states. For I p I = 1, the
equilibrium s t a t e s C,, and D, m e r g e into one, and the s y s t e m h a s t h r e e
equilibrium s t a t e s A t , BI,, and C, ( C , coincides with D l ) . Finally, f o r p = 0 ,
the s y s t e m h a s two equilibrium s t a t e s A , and B o (which may be r e g a r d e d as
coinciding with Co and Do*respectively). C, and D , lie on the circle W'.
Since a , + b , = - 1, then f o r p f 0, one of the points A , , B , lies inside Wand
the other outside the c i r c l e . The configuration of the equilibrium s t a t e s of
433
Ch.XIV. APPLICATIONS OF THE THEORY OF BIFURCATIONS I
(S,) f o r the v a r i o u s values of the p a r a m e t e r p is shown i n F i g u r e 185 (the
a r r o w s in this figure m a r k the direction of motion of the equilibrium s t a t e s
as p i n c r e a s e s ) .
The c h a r a c t e r of the equilibrium s t a t e s is determined by the usual
techniques. F o r p # 0, I p I # 1, the points A , and B , a r e nodes or foci
when p > 0 and saddle points when p<O, w h e r e a s the points B , and C, a r e
nodes o r foci when p < 0 and saddle points when p > 0 ( F i g u r e 185).
The equilibrium state C, of (S,) is multiple f o r p = & I , having A = 0,
+-
u = P; Q; # 0. To fix i d e a s , l e t p = - 1. We w i l l proceed by the method
outlinedinQT, $21.2, using Theorem 65 of this section. System (S,) for
p = - 1 h a s the f o r m
dz
-
dt
= y (z- 1) f2e + ya-I,
*=
dt
-z(z-I).
The coordinates of the equilibrium state Cl are ( 1 , O ) . Moving the origin

to the point C+ we effect a transformation of coordinates z= -2y, y = lG+y,
2
-
t = t
~, and r e v e r t i n g to the old notation x , y, t we obtain
Applying Theorem 65 f r o m QT, $21.2 to this s y s t e m , we find that the

point C-( is a saddle-node of (S-1) and that the path configuration n e a r this
point c a n be schematically shown a s in Figure 186 (this r e s u l t could be f o r e -
s e e n beforehand, by regarding the point C-l as the outcome of the merging of
the node C, and the saddle point D , of (S,) f o r p -+ - 1). The situation is
e n t i r e l y analogous with r e g a r d to the point Ci of (SI).
5438 434
S 34. EXAhlPLES
.ao and Bo are a l s o saddle-nodes (of s y s t e m (so)).

W e d o not prove this
proposition, however, since the only relevant fact f o r what follows is that
the Poincare index of each of these equilibrium s t a t e s is 0. The l a t t e r point
c a n be established i f w e note that on passing to a c l o s e s y s t e m (SA, two
equilibrium s t a t e s A , and C , (B,and D,, respectively) a p p e a r in the neigh-
borhood of the points A O(,orB o ) , the s u m of their indices being z e r o .
+
-As w e have s e e n before the c i r c l e kV (z* yz - 1 = 0) c o n s i s t s of paths of
(S,). If this c u r v e d o e s not contain equilibrium s t a t e s , i.e., if 1 p I > i ,
Il'is a closed path of s y s t e m (S,): W e will now show that (S,) h a s no closed
paths o t h e r than Wfor any value of p.
F i r s t consider the c a s e p = 0. (S,,) then h a s two equilibrium s t a t e s A , and
BO, each with PoincarC index 0. Since the s u m of the indices of the equili-
brium s t a t e s lying inside a closed path is 1 (QT, $11.2, Theorem 28,
Corollary l ) , (So)h a s no closed paths.
Let now p # 0. Consider a n auxiliary s y s t e m
-
(the two s y s t e n l s (S,) and (S,) can be considered a s p a r t i c u l a r cases of
the s y s t e m
i t is directly verified that (S,) h a s a g e n e r a l integral
t h e family of c u r v e s (15) for p > 0 is shown @ F i g u r e 187a and for I( 0 in -=

Figure 187b. The contact c u r v e of (S,) and (S,) h a s the equation
z2(z2- y2 - i ) = 0. It c a r be decomposed into a s t r a i g h t line z = 0 and the
c i r c l e Lt'. The contact points on the line z = 0 are all false (see previous
example, and a l s o QT, 12.5), and the sign of the expression fl (zz y' - 1) +
is r e v e r s e d only when we c r o s s the circle W.
--c
X
a b
FIGURE 187. a) p > 11; b) p < 0.
435
Ch.XLV. APPLICATIONS OF THE THEORY OF BIFURCATIONS
Suppose that (S,) h a s a closed path L . Since the s u m of the indices of the
equilibriuni s t a t e s of (S,) enclosed inside L i s 1, at least one node of (S,)
lies inside L and a t l e a s t one saddle point l i e s outside L. This mean?,
however, that the path L inevitably c r o s s e s closed paths of s y s t e m (S,) and
consequently h a s a ( t r u e ) contact ( s e e previous e x a m p l e ) with one of these
closed paths, Since the ( t r u e ) contact m a y only l i e on the c i r c l e W , which
is a path of (S,) for e v e r y p +
0, the closed path L of (S,) should be tangent
to the c i r c l e W . This is feasible only if L coincides with W(since W is a
path of (S,) or c o n s i s t s of paths of this s y s t e m ) .
We have thus established that (S,) h a s no closed paths for 1 p 1 4 1 and h a s
p r e c i s e l y one closed path - the c i r c l e W - for 1 p > 1. For p = - 1, (S,)
h a s a saddle-node C-, whose w-separatrix L+ f o r m s a loop which, together
-
with the point C+ constitute the c i r c l e W . For t + 00, the s e p a r a t r i x L+
goes to the saddle-node C-,> being one of the i n t e r i o r paths of the node
s e c t o r . At the point C-, (1, 0),
I
P ; ( x , Y, -I)+Q1(z, Y, -1)=2>0. (16)
Let u s now consider the bifurcations which take place n e a r the c i r c l e

-
W a s p v a r i e s around -1. For p > 1, the s y s t e m h a s no closed paths,
and on the c i r c l e W t h e r e is a saddle point and a node n e a r the point (1,O).
A s p d e c r e a s e s , these points draw c l o s e r , and they m e r g e into a saddle-
node f o r p = - i; one of the s e p a r a t r i c e s of the saddle-node f o r m s a loop.
A s p d e c r e a s e s f u r t h e r , the saddle-node CA d i s a p p e a r s and, by T h e o r e m s 51
and 52 (S30) and r e l a t i o n (16), the s y s t e m should have p r e c i s e l y one, and a t
that stable, l i m i t cycle in the neighborhood of the loop L + . This cycle is the
c i r c l e W . In this example, the l i m i t cycle is c r e a t e d f r o m a s e p a r a t r i x
loop in the following m a n n e r : the equilibrium s t a t e C-, is t r a n s f o r m e d into
a r e g u l a r point which f o r m s , together with the s e p a r a t r i x loop, the l i m i t
cycle W . The situation is e n t i r e l y analogous f o r the v a r i a t i o n of the
p a r a m e t e r p around p = 1
To elucidate the topological s t r u c t u r e of the dynamic s y s t e m (S,) on the
e n t i r e plane, w e should consider in m o r e detail the c h a r a c t e r of the
equilibrium s t a t e s and the path configuration a t infinity. This i s left to
the r e a d e r a s a n e x e r c i s e . The bifurcation v a l u e s of the p a r a m e t e r a r e
evidently p .= 0, 1, -1.
E x a m p 1e 15 (the c r e a t i o n of a l i m i t cycle f r o m the loop of a saddle -
point s e p a r a t r i x and a multiple focus).
d r= _
-
(Jt U--B)+v(l
it=-
& e (5 +US! (1 +
Physical considerations ( s e e / 2 6 / ) r e s t r i c t the v a r i a b l e x to the r a n g e
l+B.z>O, (17)
and the c o n s t a n t s fi and e a r e constrained by the inequalities
o<p<;, .O<e<oo.
436
5 34. EXAMPLES
In t h e half-plane defined by (17) (i.e., the half-plane to the right of the

1
line 2 = - -), (A) h a s two equilibrium s t a t e s O1 (-4, 1) and 0 2 y ~ ) , where
(q.
8
=? =7
1 Y2=-++/F$.
f+/-Fg
B
After s o m e manipulation:;, w e find that Q i ( - i , 1) is a node or a focus, which
i s stable f o r E>L
Z(1-8)
anti unstable for e<&). F o r ,"e 8 O iis a n
equilibrium s t a t e with p u r e imaginary c h a r a c t e r i s t i c n u m b e r s .
t'he equilibrium s t a t e O z is a saddle point for all the relevant values of
t h e p a r a m e t e r s . Let u s e s t a b l i s h s o m e p r o p e r t i e s of i t s s e p a r a t r i c e s .
The i s o c l i n e s of v e r t i c a l and horizontal inclinations are d e s c r i b e d by
the r e s p e c t i v e equations
and
The f i r s t is the equation $of a hyperbola and the second that of a parabola..
I'hese equations partition the half-plane 1 +&z>O into r e g i o n s where x and
r e t a i n a constant sign. Let t h e s e r e g i o n s be designated I, 11,111, I\'
( F i g u r e 188). The s i g n s of 3 and y, r e s p e c t i v e l y , a r e indicated in
p a r e n t h e s e s in the figure.
FIGURE 188.
The equation c h a r a c t e r i z i n g the d i r e c t i o n s of the s e p a r a t r i c e s f o r the

saddle point Oa h a s the f o r m
43 I
I
a.XIV. APPLICATIONS OF THE THEORY OF BIFURCATIONS
Since e > 0, 1 + pxz > 0, yz > 0, and 0 < p <f , the two r o o t s of (21) a r e
negative. Hence i t follows that the s e g m e n t s of the s e p a r a t r i c e s located
n e a r the point O2 lie in r e g i o n s I1 and IV ( F i g u r e 188).
Seeing that two s e p a r a t r i c e s go to the saddle point f o r t + 00 and +
the other two for t + - 00, and examining the variatior. of x and y in I
regions I1 and IV, we readily find that the configuration of the s e p a r a t r i c e s
n e a r the saddle point O2 can be r e p r e s e n t e d a s in F i g u r e 188.
Consider the point of the s z p a r a t r i x L, lying n e a r O2. Let this point
c o r r e s p o n d to the time to. Since this point l i e s in region IV, i t w i l l move
to the left and up as t i n c r e a s e s ( x d e c r e a s e s , I/ i n c r e a s e s ) . But then the
s e p a r a t r i x L, w i l l not c r o s s the parabola (20), which is an isocline of the
horizontal inclinations, and i t should evidently c r o s s the s t r a i g h t line
1 $. f3s = 0. We s i m i l a r l y conclude that the s e p a r a t r i x L, goes to infinity as
t increases.
The behavior of the s e p a r a t r i c e s Lz and Ls cannot be determined u n -
ambiguously, and a detailed a n a l y s i s is r e q u i r e d . To this end, consider
an auxiliary system
dt =-e ( x + y e ) .
dx
-=
dt y, 3 (E)
Dividing the second equation through by the f i r s t equation and substituting

ya = z, we obtain a l i n e a r equation, which c a n be integrated to give a g e n e r a l
i n t e g r a l of s y s t e m (B),
F (5, y) = e2ex ( - &+z+g9) = c. (22)
(B) h a s a single equilibrium state, the o r i g i n 0 (0, 0) (it c o r r e s p o n d s to

C=- &- in (22)). Investigation of the c u r v e s (22) by the usual e l e m e n t a r y
1
methods shows that they can be depicted as in F i g u r e 189. C = --
28
corre-
1
sponds to the equilibrium s t a t e 0 (0, 0). The v a l u e s - z < C <0 correspond
to closed paths of s y s t e m (B): as C i n c r e a s e s t h e s e closed paths expand.
C = 0 c o r r e s p o n d s to the p a r a b o l a -5
1
+ x + 61% = 0, which is obtained from
the p a r a b o l a y + x* = 0 as a r e s u l t 1
of a shift to the right by 5 (in F i g u r e 189,
this displaced parabola is shown by dashed curve; i t is not a path of
s y s t e m (B)).P a t h s which a r e not closed c o r r e s p o n d to C > 0.
L e t u s d e r i v e the equation of the contact c u r v e of s y s t e m s (A) and (B).
Taking the derivative of the function F ( x , p)
-
d-
F(z' ') -F'fi+F'&=F;P(x, y ) + G Q ( x , y)
dt dt dt
( s e e QT, 53.13) and using (A), we obtain
g
-= -2 e e Z C " (5 + y*) (1 - B). (23)
Hence i t follows that the contact c u r v e z=O is the parabola r + g * = O , which

is a n isocline of the horizontal inclinations both f o r s y s t e m (A) and f o r
s y s t e m (B).
438
5 34. EX .Ah!PLES
FIGL'RE 189
Since p < + , and e > O , *dtK O in r e g i o n s where x + y z > O , i.e., in r e g i o n s I1

and 111 in F i g u r e 188. This evidently i m p l i e s that the paths of s y s t e m (A) c r o s s
the closed paths of s y s t e m ( B ) i n t h e d i r e c t i o n f r o m outside to inside as
t i n c r e a s e s . In regions I and E , on the o t h e r hand, where r + y 2 < 0 ,
c>O and the paths of s j s t e m (A) c r o s s the closed paths of (B)f r o m inside
to outside as t i n c r e a s e s .
The a x i s z is a n isocline of the v e r t i c a l inclinations of (B). Each closed
path of (B) c r o s s e s the positive h a l f - a x i s I a t a single point in the downward
direction ( F i g u r e 189).
FIGURE 13C. The dashed curve shows the posslble be-

havior of the separatrut LI.
N o w c o n s i d e r the a - s e p a r a t r i x L2 of s y s t e m (A) which e m e r g e s f r o m

saddle point Oz into I1 ( 5 > 0,y < 0, see F i g u r e 190). Since O2 lies to the left
439
ch.XIV. APPLICATIONS OF THE THEORY OF BIFURCATIONS
of the parabola - 1
+ s + g* = 0, the path of s y s t e m (B) p a s s i n g through
this point is a closed path ( F i g u r e 1 9 0 ) . We designate this closed path z,.
zo,
At the point Oz, the s e p a r a t r i x Lz e n t e r s into s i n c e a t this point the
inclination of La is negative and the inclination of gois horizontal
( F i g u r e 190; Oz l i e s on the isocline of horizontal inclinations of s y s t e m (E)).
While in region where x + yp > 0, the s e p a r a t r i x L2 cannot c r o s s the c u r v e
Lo going f r o m inside to outside, since in this region i t s behavior is such
that i t c r o s s e s all the closed paths of (B) f r o m outside to inside. Moreover,
a s t i n c r e a s e s , the s e p a r a t r i x Lz cannot move f r o m region I1 into region I
c r o s s i n g the a r c OiOz of the parabola y f xp = 0, since in all the i n t e r i o r
points of this a r c the paths of s y s t e m (A) a r e d i r e c t e d f r o m left to right.
W e should t h e r e f o r e consider the following a l t e r n a t i v e s for the
s e p a r a t r ix Lz:
1) without leaving region 11, LZ goes to the equilibrium state 0,f o r
t-c + W;
2 ) remaining inside the c u r v e C , , Lz c r o s s e s the isocline of the v e r t i c a l

inclinations (the hyperbola (19)) and then c r o s s e s the parabola (20) to e n t e r
region IV.
A s t i n c r e a s e s f u r t h e r , the following t h r e e possibilities should be
considered f o r the s e p a r a t r i x Lz (they a r e consistent with the fact that in
regions I and J Y the paths of (A) c r o s s the closed paths of (B) f r o m inside
to outside):
2 ' ) L2 goes to the equilibrium s t a t e O,, if it is stable, o r to a s t a b l e
l i m i t cycle surrounding this equilibrium state, i f such a cycle exists;
2") Lz goes to the saddle point 0,;in this c a s e , i t evidently m e r g e s
with the s e p a r a t r i x L3 forming a loop;
+
2"') L z c r o s s e s the s t r a i g h t line 1 flx = 0 (Lzcannot go to infinity in the
lower half-plane since in region IV y > 0).
The behavior of the s e p a r a t r i x LI, as is readily s e e n , is determined by
the behavior of L2.
L e t u s now c o n s i d e r the e x i s t e n c e or otherwise of closed paths in
s y s t e m (A).
E v e r y closed path of s y s t e m (A) should enclose the point O,, while
leaving the point O2 on the outside, and i t m a y not have any points in
common with o t h e r paths, in p a r t i c u l a r , not with the s e p a r a t r i c e s L2 and
La. Hence i t follows that i f the s e p a r a t r i x Lz goes to the equilibrium state 0,,
(A) h a s no closed paths. Now suppose that the s e p a r a t r i x LZ c r o s s e s the
hyperbola (19), and l e t zobe the a b s c i s s a of the i n t e r s e c t i o n point M,
( F i g u r e 190). Every closed path of (A) of n e c e s s i t y c r o s s e s the segment
O,Oz of the parabola (20). Making u s e of this fact and taking into c o n s i d e r a -
tion the direction of the field of s y s t e m (A) in r e g i o n s I, 11, and IV and on
the i s o c l i n e s (19) and (20), we c a n r e a d i l y show that if (A) h a s closed paths,
they may only l i e inside the region between the v e r t i c a l l i n e s x = xz and
3 5 50 .
zo
L e t M* ( x * , y*) be the i n t e r s e c t i o n point of the path of the a u x i l i a r y
s y s t e m (E) with the positive half-axis z. Since zo < r* (this is c l e a r f r o m a n
examination of the behavior of Lz), we conclude f r o m the above that i f
s y s t e m (A) h a s c l o s e d p a t h s , t h e y m a y o n l y l i e b e t w e e n
the vertical lines x=xz and x=x*.
440
5 34. EXAhlPLES
\ V e will now apply I h l a c ' s c r i t e r i o n (QT, 512.3). Dulac's function is

taken in the f o r m
e ? ~ ~
Q(2. Y ) = v .
Computations show that
Since p -=+ ~ and 1 + k # O , the l a s t expression vanishes only a t the points

of the straight line
x=- p-2e
2Ep *
By Dulac's c r i t e r i o n , (A) may not have closed paths in the region where
d d
z ( Q p ) - k > (OQ)
~ 1 0 . Therefore, i f the above s t r a i g h t line d o e s not lie
between the v e r t i c a l s r=q and x = x * , i.e., if e i t h e r
or
s y s t e m ( A ) h a s no closssd paths.
The f i r s t of the two conditions in (24) c a n be written i u the f o r m
It is satisfied f o r any p O<p<+, if e > l .

Let_us now consider the second condition in (24). The equation of the
c u r v e Lo is given by (22), where C is determined f r o m the condition that
zo p a s s e s through the point 02(x2,y2). The equation of E , is thus
$ex x - ~ Z - - 1 ) =e?exz(x2+yi-x). 1
( 4
Setting y - 0 , w e obtain a n equation f o r x * :

e2ez*(x*-L) = e ? e x a (x2+y:-z) I
2e
or after e l e m e n t a r y manipulations,
i - 2=* ~ p2e(c?--x*)
I1 - 2e (x2 f yi)]. (25)
Since e v e r y c u r v e ( 2 2 ) c r o s s e s the positive half -axis x a t one point only

( F i g u r e 189), equation (.25) h a s a single positive solution f o r x*. Let
441
Ch. XIV. APPLICATIONS OF T H E THEORY OF BIFURCA'I'LONS
?e.x*=z. Equation (25) then can be written in the f o r m

1 -z -e-'f (E) = 0,
where f ( e ) = e " e x 2 [ I - 2 ~ ( ~ Z + (52

y ~ )and
l y, a r e independent of E).
For E = O , equation (26) h a s the f o r m
1 -z -e-' = 0.
It h a s a unique solution z = 0, and the derivative of its left-hand side a t
the pointz = Odoes not vanish. Therefore, by Theorem 3 and the r e m a r k
to Theorem 4 (the theorem of s m a l l i n c r e m e n t s of implicit functions, $1.21,
for sufficiently s m a l l e equation (26) h a s a unique solution c l o s e to z e r o ,
which goes to z e r o for E -+ 0. It is readily verified that f o r e > 0 this solution
is positive. Andsinceequation(25) h a s a u n i q u e p o s i t i v e s o l u t i o n
for I*, w e conclude that 2ex* -+ 0 f o r E -+ 0.
The second inequality in (24) can be written in the f o r m
p -- 2E > 2EX* .p.
This and the condition lim 2es* = 0 show that f o r any f3, the second
e-0
condition in ( 2 4 ) is satisfied a s soon as E > 0 is sufficiently s m a l l .
We have thus established that f o r e v e r y fixed p, O<f3<+. system (A)
h a s no closed paths i f E is sufficiently s m a l l o r sufficiently l a r g e . Let u s
now t r y to elucidate the topological s t r u c t u r e of (A) in these cases.
Let
A s w e have s e e n before, f o r R > 0, the equilibrium s t a t e 0,(-1, 1) is a n

unstable focus or node, and f o r R < 0 , i t is a stable focus o r node.
t
W e will now show that f o r e v e r y $, O < f 3 < ~ , there e x i s t s a c e r t a i n E >0
for which the s e p a r a t r i c e s Lz and Ls m e r g e into a loop.
Indeed, if E = > 0 is sufficiently s m a l l , s y s t e m (A) h a s no closed paths
and, moreover: R > 0. But then Ol is an unstable node o r focus; the
s e p a r a t r i x L Z d o e s not go to O1 f o r t -+ +
00 and i t either m e r g e s with the
s e p a r a t r i x L , o r l e a v e s the region IV by c r o s s i n g the segment ST of the line
x = - - Bt . In the f o r m e r case, our proposition is proved, and in the l a t t e r
case, the s e p a r a t r i c e s are a r r a n g e d as shown schematically i n Figure 151.
s Q X
FIGURE 191. e =e,. FIGURE 192. e = ez.
442
S 34. EXhllPLES
b e can s i m i l a r l y show that if E = e2 > 0 is sufficiently l a r g e (in p a r t i c u l a r ,

so l a r g e that R < 01, the s e p a r a t r i c e s L z and L 3 e i t h e r m e r g e forming a loop
o r are a r r a n g e d a s shown in F i g u r e 192.
Now suppose that the s e p a r a t r i c e s L 2 and L3 do not m e r g e e i t h e r for
E = F, or for e = E? (i.e., the configuration of the s e p a r a t r i c e s i s according
to F i g u r e s 151 and 192). \ Y e w i l l show that in this c a s e they m e r g e f o r
s o m e e = eo, el < e, < 2 First note that the s t r a i g h t line x = - h a s no
points of contact with the paths of the s y s t e m ( A ) (this is evident d i r e c t l y
f r o m the equations of (A)). AIoreover, s i n c e the con_tact c u r v e of s y s t e m s
( A ) and ( B ) is parabola (ZO), the a r c 02Sof the path Lo of ( B ) extending
between i t s i n t e r s e c t i o n points Ozand -% with the parabola (20) ( F i g u r e 1 5 2 )
h a s no c o n t a c t s with the paths of s y s t e m (A) e i t h e r . T h e r e f o r e , by the
basic p r o p e r t i e s of s e p a r a t r i c e s (99.2, L e m m a 3 and the r e m a r k to this
l e m m a ) , i f for s o m e value of the p a r a m e t e r E the s e p a r a t r i x L a c r o s s e s
t
the line x = - - ( F i g u r e 191) o r the s e p a r a t i r x L3 c r o s s e s the a r c 02-V of the
B
curve ( F i g u r e 1521, the behavior of these s e p a r a t r i c e s will not change f o r
a l l c l o s e values of the p a r a m e t e r s . Now suppose that the s e p a r a t r i c e s L2
and L 3 do not m e r g e f o r anys, e, < e < e?. Let E v a r y f r o m E, to e ~ . Then, a s
i s r e a d i l y s e e n , t h e r e e x i s t s s o m e first Lralue of the p a r a m e t e r E = e* for
which the s e p a r a t r i x Lz 110 longer c r o s s e s the line 3 = --1B ( w h e r e a s for all
e. E,<E < e*, it d o e s c r o s s this line). Since by assumption the s e p a r a t r i c e s
Lt and L 3 do not f o r m a loop, f o r e = e* the s e p a r a t r i x Lz goes f o r t .+ + 00
e i t h e r to OI-or to a l i m i t c y c l e e n c i r c l i n g 0,. In e i t h e r c a s e , L3 c r o s s e s the
a r c 02.V of L o . But then, f o r all c l o s e values of E, and in p a r t i c u l a r for
E < e*, L 3 c r o s s e s the a r c O Z N and consequently, L z may not c r o s s the line
2=-- B1 . \$e have established a contradiction, which p r o v e s that f o r s o m e
EO, 1 < EO < e2,
the saddle point 0 2 of (A) h a s a s e p a r a t r i x forming a loop.
Choose a fixed fi and follow the changes in the topological s t r u c t u r e of (A)
in the neighborhood of the equilibrium s t a t e 0,(-1, 1)a s E is v a r i e d .
-
Let e = B
-
2(1- BJ
F o r E = e , O,(-i, i) is a n equilibrium state with p u r e
imaginary c h a r a c t e r i s t i c n u m b e r s . Since at the point 0, (-1, f ) ,
o = p - 2e (1-p), we conclude that for e=; ando=O, g<O. a3 is computed
f r o m equation (76), 524.4, which gives
Therefore, f o r e = e, O1 1s a n unstable focus, and when e i n c r e a s e s above E,

the focus 0,is t r a n s f o r m e d into a s t a b l e focus and a single unstable l i m i t
cycle is c r e a t e d in i t s neighborhood ( s e e table in 925.3).
Thus, a s e v a r i e s f r o m a sufficiently s m a l l e, to a sufficiently l a r g e e2,
the following bifurcations definitely o c c u r :
1 ) a t l e a s t for one e = e,. the s e p a r a t r i c e s Lz and La of the saddle point
0 2 m e r g e forming a loop;
443
2 ) p r e c i s e l y once, f o r e = e, Oi is t r a n s f o r m e d into a multiple focus,

and a s e is f u r t h e r i n c r e a s e d , a n unstable limit cycle is c r e a t e d f r o m this
focus.
In addition to these bifurcations, w e may a s s u m e a p r i o r i the existence
of bifurcations of s t i l l another type in this s y s t e m :
3 ) appearance ( o r d i s a p p e a r a n c e ) of a l i m i t cycle of multiplicity 4, and
in p a r t i c u l a r of multiplicity 2, f r o m a "condensation of paths."
Very considerable difficulties a r e encountered when one t r i e s to e s t a b l i s h
the p r e s e n c e (or the a b s e n c e ) of bifurcations of this l a s t type, and nothing
c e r t a i n c a n be s a i d in o u r example r e g a r d i n g this point.
Let u s consider one f u r t h e r fact r e l a t i n g to s e p a r a t r i x loops. As we
know, the loop formed by a saddle-point s e p a r a t r i x is s t a b l e o r unstable,
according a s the value of P ; + Q, at the saddle point is negative or positive
( T h e o r e m 44, S29.1). In o u r example, f o r e = eo>
But
Let
eo G - b
(1-b)- -e (30)
(EOis the value of e f o r which the s e p a r a t r i c e s Lz and L , m e r g e forming a

-
loop; e is the value for which Osis a multiple focus). For O<p.=-+> it i s
readily s e e n that yz > 1. This and (29). (30) show that
We have ct . established that if the s e p a r a t r i c e s f o r m a loop for eo<;, the

loop i s unstable.
L e t u s now analyze the changes in the topological s t r u c t u r e of (A) a s E
i n c r e a s e s f r o m 'el to e 2 a s s u m i n g that two f u r t h e r conditions are satisfied:
(a) as e v a r i e s f r o m el to e2, no bifurcations of type 3 o c c u r ;
(b) t h e r e e x i s t s a single value eoof the p a r a m e t e r e at which the
s e p a r a t r i c e s Lz and L3 m e r g e forming a loop.
We should s t r e s s that the applicability of t h e s e conditions [or of e i t h e r of
them) is not known i n advance, and we introduce them a s a n assumption in
o r d e r to proceed with our a n a l y s i s .
For E = E ~ , the s y s t e m h a s no closed paths and the s e p a r a t r i c e s a r e
a r r a n g e d as shown i n F i g u r e 191.
A s e i n c r e a s e s , the topological s t r u c t u r e of the s y s t e m a t f i r s t does not
change. It may change only when the p a r a m e t e r c r o s s e s its bifurcation
444
9 34. EXAhlPLES
e
v a l u e . The bifurcation values of the p a r a m e t e r in o u r c a s e a r e and 0 . 5 5 ' ~
will show that E ( e o . Indeed, i f ;> e , , Oii s an unstable node or focus f o r
E = E and the s y s t e m h a s a s e p a r a t r i x forming a loop, which loop, a s w e
have s e e n before, i s d s o unstable. By condition (a), no closed paths may
develop in the s y s t e m . T h e r e f o r e , the paths enclosed inside the loop should
go e i t h e r to the loop or to the point O1 f o r f --t t o o , which is impossible s i n c e
both the loop and the point O l a r e unstable.
U'e have thus established that e< e o . For E = the point O1 is a multiple
unstable focus, and for all e between e l and E the s y s t e m h a s a constant
topological s t r u c t u r e (that shown on F i g u r e 191).
As E is f u r t h e r i n c r e a s e d , O1 is t r a n s f o r m e d into a s t a b l e focus, and a
s i m p l e unstable l i m i t cycle L o is c r e a t e d in its neighborhood (S25.3). The
s y s t e m a c q u i r e s the topological s t r u c t u r e schematically shown in F i g u r e 193.
?Is E v a r i e s between the l i m i t s < E < E ~ , this topological s t r u c t u r e is
retained.
-
FIGYRE 193. E -=z e <eo. FIGL'RE 1"4. e eo.
For e = e,, the s e p a r a t r i c e s of the saddle point O 2f o r m a loop. U'e will

show that a t this instant the l i m i t cycle L o d i s a p p e a r s (is "swallowed" by
the s e p a r a t r i x loop), and the loop itself is unstable, i . e . , the s y s t e m h a s the
s t r u c t u r e shown in Figure 194.
Indeed, suppose that the unstable l i m i t cycle Lo e x i s t s f o r e = E O . W e
know that f o r E = e- the s y s t e m h a s no closed p a t h s . Therefore the
s t r u c t u r a l l y s t a b l e l i m i t cycle L o should d i s a p p e a r with the i n c r e a s e in e.
This may o c c u r only by merging with s o m e l i m i t cycle c r e a t e d f r o m the
s e p a r a t r i x loop, i.e., as a r e s u l t of a bifurcation of type 3 , which i s r u l e d
out by coGdition (a). Thus, the l i m i t c y c l e m u s t d i s a p p e a r f o r E = E,, and
s i n c e the focus in this c a s e is stable, the s e p a r a t r i x loop is unstable
( F i g u r e 194).
A s e is f u r t h e r i n c r e a s e d , the s e p a r a t r i x loop b r e a k s up. It does not
produce closed paths, since eventually t h e s e p a t h s w i l l have to d i s a p p e a r ,
which is r u l e d out by condition (a). T h e r e f o r e , f o r E > eo, the topological
s t r u c t u r e of the s y s t e m is the s a m e a s f o r E = E ? ( F i g u r e 192).
We thus s e e that with the aid of conditions ( a ) and (b) w e succeeded in
deriving unambiguously the topological s t r u c t u r e of the s y s t e m f o r all e>O.
Note that if conditions ( a ) and (b) (or e i t h e r of t h e m ) are not satisfied,
the number of a p r i o r i possible topological s t r u c t u r e s of c o u r s e markedly
i n c r e a s e s . Thus, if w e allow bifurcations of type 3 - c r e a t i o n of a l i m i t
445
cycle of multiplicity 2 (from a condensation of p a t h s ) o r disappearance of a

l i m i t cycle - the topological s t r u c t u r e may undergo the following changes:
1) A l i m i t cycle of multiplicity 2 is c r e a t e d .
2) The l i m i t cycle s p l i t s into two: a n unstable outer cycle L' and a stable
i n n e r cycle L".
3 ) The o u t e r cycle L' is "swallowed" into the s e p a r a t r i x loop, and the
s e p a r a t r i x b r e a k s up.
4 ) The focus changes i t s stability, and an unstable l i m i t cycle L " i s
created.
5) The c y c l e s L" and L'" m e r g e into a cycle of multiplicity 2, which then
disappears.
T h e r e is nothing to prevent the s e p a r a t r i x loop and the type 3 bifurcation
from occurring several times!
E x a m p l e 16 (the c r e a t i o n of a l i m i t cycle f r o m a multiple focus,
see /37/).
==- [ i + ( x + y ) y + 6 ] = P ( x .
dz
y), $-= -y+(s+y)s=Q(x, y) (31)
f o r positive values of the p a r a m e t e r s y and 6.
The coordinates of the equilibrium s t a t e s satisfy the equations
x t ( x t Y) +6 = 0, Y- b+r)t= 0. (32)
Eliminating y between t h e s e equations, we obtain a single equation f o r

the a b s c i s s a s of the equilibrium s t a t e s :
F ( x , y, 6) = i + 2 y Z + + ( y ~ + 1 ) x + 6 = 0 . (33)
Since this is a cubic equation, ( 3 1 ) h a s a t l e a s t one equilibrium s t a t e

and a t m o s t t h r e e equilibrium s t a t e s f o r all the r e l e v a n t values of the
p a r a m e t e r s y and 6 .
To d e t e r m i n e the exact number of equilibrium s t a t e s , we w i l l follow
the g e n e r a l i n s t r u c t i o n s outlined in the Introduction to this c h a p t e r .
Specifically, we w i l l look f o r those values of the p a r a m e t e r s f o r which the
equilibrium s t a t e s have maximum multiplicity. A s we shall s e e , l o c a l
a n a l y s i s n e a r t h e s e values of the p a r a m e t e r s w i l l solve the problem of the
existence of regions with different number of equilibrium s t a t e s .
The ordinate y of the equilibrium s t a t e s is obtained f r o m the equation
+
y = I (x y), i.e., i t is a single-valued function of the a b s c i s s a . Therefore,
by Definition 15, 57.3, the multiplicity of the equilibrium s t a t e (I,v), where
x is a root of equation ( 3 3 ) , is equal to the multiplicity of the corresponding
root, i.e., the maximum multiplicity of the equilibrium s t a t e of our s y s t e m
may not exceed 3 . L e t u s t r y to e s t a b l i s h the values of the p a r a m e t e r s y and
6 when equation ( 3 3 ) h a s a t r i p l e root. As we know, a triple root should
simultaneously satisfy the t h r e e equations
F (x,y, 6) = 39 + 2 Y i + (Ya+ 1)x + 6 = 0,

Fk (5,y, 6) = 3x2+ 4y1 + (1 +ys) = 0, (34)
F;a(x, y, 6 ) = 6 1 + 4 ~ = 0 .
4
F r o m the l a s t equation, we have x = - - i - y .
446
5 34. EXAhlPLES
Inserting this r e s u l t in the second equation and r e m e m b e r i n g that w e a r e

i n t e r e s t e d only in posi?ive v a l u e s of the p a r a m e t e r s , we find y=+l'%.
9 '1'5 8
Hence x = - 3
1 and
, 1'8>0.
\$e have thus e s t a b l i s h e d the e x i s t e n c e of a single p a i r of positive

8 -
p a r a m e t e r s y and 6 , nzmely y o = k q , 6 0 = g J ' 3 , for which s y s t e m (31) h a s an
equilibrium s t a t e of multiplicity 3. L e t this equilibrium state be .Ifo(zo,yo),
21%
where q = -3. yo= -T. For t h e s e v a l u e s of the p a r a m e t e r s , the s y s t e m
evidently does not have o t h e r equilibrium s t a t e s .
Since Mo(z0,yo) is a multiple equilibrium s t a t e , A ( q , yo) = O (this a l s o c a n be
found by a d i r e c t compitation). F u r t h e r m o r e ,
4
a(% Yo)==P;(zo, Yo)-rQ;(Xo. Yo)= -2-y,= -s<p.
Investigating the equilibrium s t a t e M o by the methods d e s c r i b e d i n QT
(Chapter LX, $21.2, T h e o r e m 65), we conclude that -Ifoi s a s t a b l e topological
node. By Theorem 35, $23.2, t h e r e e x i s t s m a l l i n c r e m e n t s for which the
multiple equilibrium s t a t e M0decomposes into t h r e e s t r u c t u r a l l y s t a b l e
equilibrium s t a t e s (two nodes and one saddle point) and a l s o i n c r e m e n t s
which r e p l a c e .\Io with a single ( s t a b l e ) s t r u c t u r a l l y s t a b l e node. However,
the r e s u l t s of $23 cannot be applied to our c a s e , s i n c e w e are not dealing
with j u s t any increment. to the right-hand s i d e s of the corresponding s y s t e m ,
but only with i n c r e m e n t s which r e s u l t f r o m changes in the p a r t i c u l a r
p a r a m e t e r s y and 6 of t.he s y s t e m .
To investigate the possible e x i s t e n c e of t h r e e equilibrium s t a t e s , let u s
c o n s i d e r the function F (z, y , 6) ( s e e ( 3 3 ) ) in the neighborhood of the point
xo, yo. 60. L e t x = xo t 5, y = yo f h , 6 = 60 k. -
Seeing that q ,yo, 6,, s a t i s f y equations (34), w e readily find that
F@, y. 6)=F(zo++5* y o + h , 6O+k)==

=3: +2 h p + + 2yoh +he) E x h + 2ygoh +
+ k = E3 +AS*
+2,,h2 + BE + C ,
where A = 2 h , e t c .
Let k = - 2xth - 2 y g , h - x h e , i.e., C = 0 . The equation F (I, y . 6 ) = 0 then
t a k e s the f o r m +5 (Ea i A5 +
B ) = 0. One of i t s r o o t s is 5 , = 0, And the
c h a r a c t e r of the o t h e r r o o t s depends on the value of the d i s c r i m i n a n t
.-12 - = - 16x& - Sy& := - 562
3
xh. Therefore, i f h -=0 , and k is s e l e c t e d
a s above, the equation F ( x , y , 6 ) = 0 h a s t h r e e different r e a l r o o t s , one of
which is xOr and if h > 0 , the equation h a s a single r e a l r o o t zo and two
complex r o o t s . C l e a r l y , i f h is sufficiently s m a l l in absolute value, k is
a l s o sufficiently s m a l l and the p a r a m e t e r s y = yo h , 6 = 6, + k a r e positive.
W e have thus e s t a b l i s h e d that f o r c e r t a i n positive v a l u e s of the p a r a m e t e r s
y and 6 , s y s t e m (31) h a s t h r e e equilibrium s t a t e s , w h e r e a s f o r o t h e r v a l u e s
of the p a r a m e t e r s i t only h a s one equilibrium state.':
* Our prooi is confined to the neighborhood of a triple equilibrium state. In this case, the sought results also
can be obtained from d direct .rLalysis of the cubic equation F (z,y. 6 ) = 0.
Ch.xIV. APPLICATIONS OF T H E THEORY OF BIFURCATIONS
On c r o s s i n g o v e r f r o m one region of p a r a m e t e r values to the other

region, we evidently p a s s through the bifurcation values of the p a r a m e t e r s
for which the s y s t e m h a s a multiple (double or t r i p l e ) equilibrium s t a t e .
Since t h e r e e x i s t s only one p a i r of values y o , 8 , for which the s y s t e m h a s a
triple equilibrium state, w h e r e a s the values of the p a r a m e t e r s for which
the s y s t e m h a s one s t r u c t u r a l l y s t a b l e o r t h r e e s t r u c t u r a l l y s t a b l e
equilibrium s t a t e s f i l l whole regions in the p a r a m e t e r plane, we should
inevitably p a s s through the p a r a m e t e r v a l u e s corresponding to the double
equilibrium s t a t e on c r o s s i n g o v e r f r o m one of t h e s e regions (the region
with t h r e e equilibrium states, s a y ) into the o t h e r (that with one equilibrium
s t a t e ) . The equation of the c u r v e in the p a r a m e t e r plane corresponding to
s y s t e m s with multiple equilibrium s t a t e s c a n be found without difficulty.
To this end, we equate to z e r o the d i s c r i m i n a n t of the cubic equation (33).
After s o m e manipulations, w e obtain the equation
9 ( 7 , 6 ) = 1276--2y (y* + 9)]2-4.(ye-3)5 = 0, (35)
which is equivalent to two equations
6 ~ --vya-a-.h (7)
2v (vZ+ 9)i - 2 w - 3 ) (36)
27
and
6=2~(~~+%-2(P-3) 7/y'--3
27 =f(7). (37)
We a r e c l e a r l y i n t e r e s t e d only in positive v a l u e s of the p a r a m e t e r y which

a r e g r e a t e r than or equal to I/s. 8 '-
F o r y = y o = v g , h ( y o ) = f ( y o ) =gv3=6,,
i.e., w e obtain the point M o (yo, 6,) corresponding to a s y s t e m with a triple
equilibrium s t a t e . The c u r v e (35) consisting of two b r a n c h e s ( 3 6 ) and (37)
meeting a t a common point M o is shown in F i g u r e 195.
FIGURE 195
r h e c u r v e 3 5 partitions the f i r s t quadrant of the p a r a m e t e r plane into

two r e g i o n s I and I1 ( F i g u r e 195). It is r e a d i l y s e e n that the dynamic
448
4 34. EXAh!PLES
s y s t e m s corresponding to the points of region I (cp ( y , 6) > 0) have one

(structurally s t a b l e ) equilibrium s t a t e , w h e r e a s s y s t e m s corresponding to
the points of region I1 ('p ( y . 6) < 0) have three, a l s o s t r u c t u r a l l y stable,
equilibrium s t a t e s . Systems corresponding to the points of the c u r v e (35)
o t h e r than .IIo have one s t r u c t u r a l l y stable and oile double equilibrium s t a t e .
Let u s now consider the c h a r a c t e r of the equilibrium s t a t e s of s y s t e m ( 3 1 ) .
The f i r s t question is the possible existence of equilibrium s t a t e s with pure
imaginary c h a r a c t e r i s t i c n u m b e r s (a c e n t e r o r a multiple focus f r o m which
a limit cycle may be c r e a t e d ) . It should be established whether such
equilibrium s t a t e s e x i s t o r not, and i f they do, then f o r what values of the
p a r a m e t e r s . F o r such a n equilibrium s t a t e ,
and u = P; Q; = 0. + For a n equilibrium s t a t e (2. y) of s y s t e m ( 3 1 ) , a s is

readily seen,
A == 3x2 - 4xy y? f1, (38)
u z - (2% f Y X f 2 ) . (39)
Let u s f i r s t establish whether a n equilibrium s t a t e c a n e x i s t a t all f o r

which A = 0 and 0 = U simultaneously,"' The a b s c i s s a of a n equilibrium
s t a t e of s y s t e m (31) f o r which u = 0 simultaneously s a t i s f i e s equation (33)
and the equation
Eliminating y between these equations, we obtain
(y, 6) = 2y4-3y6 + 62 +a = 0. (41)
In the plane of the p a r a m e t e r s y , 6 equation (41)is a c u r v e (a hyperbola),

and only dynamic s y s t e m s corresponding to the points of t h i s hyperbola may
have equilibrium s t a t e s x i t h u = 0.
Let u s now d e t e r m i n e the values of the p a r a m e t e r s f o r which the dynamic
s y s t e m may have a n equilibrium s t a t e (3, y) with 1 = u = 0.
F r o m the equations
A = 312 +4xy +y'+ 1 = 0, u = - (12+yz + 2) = 0

w e readily find x =-'. Inserting this r e s u l t in the second equation, we
find
5
y=- 3 . , Y
Hence x = ---and,
2
1 3
by(33), &=A.
1 3
T h e r e is thus a single point A (vi, 6,) i n the p a r a m e t e r plane, with
Yl = -, 61 = 8
5
which c o r r e s p o n d s to a s y s t e m having a n equilibrium
15 '
~~~
s t a t e with A = 0, u = 0. Since the condition A = 0 is equivalent to the

existence of a multiple equilibrium s t a t e of a dynamic s y s t e m , i.e., the
existence of a multiple root of equation (33), the point A (yl, 6,) lies on the
We are rhus considering. as befo-e, equilibrium states of maximum multiplicity.
449
Ch.XIV. APPLICATIONS OF THE THEORY OF SIFURCATIONS
c u r v e cp ( y , 6) = 0 (see (35)) and is the only common point of this c u r v e and

I
the c u r v e (1 ( y , 6) = 0. It is r e a d i l y s e e n that the c u r v e s cp ( y , 6) = 0 and
9 (7, 6) = 0 touch a t the point A . Moreover, i t can be shown that the branch
+ ( y , 6) = 0 of the hyperbola located in the f i r s t quadrant is e n t i r e l y contained
in region I1 ( F i g u r e 195). Therefore, the stability of an equilibrium s t a t e I
with A # 0 c a n change only f o r those values of the p a r a m e t e r s when
s y s t e m ( 3 1 ) h a s t h r e e equilibrium s t a t e s .
Let u s t r y to identify the equilibrium s t a t e s with u = 0. Since
u = P ; + Q; = - (v + 2), a n equilibrium s t a t e (2, I/)with u = 0 h a s g = - 2.
The a b s c i s s a is x = 6 - 27 in v i r t u e of the f i r s t equation i n (31). Moving
the o r i g i n in the plane ( x , y) to coincide with this equilibrium state
(6 - 2 y , -2), i.e., using the substitution of v a r i a b l e s
x=X+6-2y, y=Y-2,
we obtain a s y s t e m
dx
df
= x +( y - .6)Y- .XY, g = (26- 3 y ) x- Y+X2+2y2+62-- 3y6 + 2.
In virtue of the conditionu=O, relation ( 4 1 ) is satisfied, and the l a s t
s y s t e m t a k e s the f o r m
dt -x+
25 ( y -6 ) Y - X Y , $= (26-3y) x- Y +X'. (42)
The equilibrium s t a t e (with u = 0 ) now h a s the coordinates (0,0). Its

c h a r a c t e r i s t i c equation is
Depending on the sign of the e x p r e s s i o n i n b r a c k e t s , the c h a r a c t e r i s t i c

r o o t s a r e e i t h e r complex conjugate n u m b e r s o r real n u m b e r s of
opposite sign.
It is r e a d i l y s e e n that the c u r v e s
I
Q ( y , 6 ) = 2ya-3y6+ 6*+2 = O
and
(y -6 ) (26- 3y) + i =0
have a single common point in the f i r s t quadrant, namely the point
A ( 55 p),
.8 and that this is t h e i r point of i n t e r s e c t i o n (and not a t r u e point
of contact). The e x p r e s s i o n (y - 6) (26 -- 3y) -I- 1 t h e r e f o r e r e t a i n s the s a m e
sign e v e r y w h e r e along the branch A B of the c u r v e (41)(except the point A )
and a n opposite s i g n everywhere along the branch AC of this c u r v e .
Consider the points (3, 5) and (9/2, 5) of c u r v e ( 4 1 ) ; we r e a d i l y f i n d t h a t t h e
e x p r e s s i o n ( y - 6 ) ( 2 6 - 3 y ) f 1 is negative on the branch A B and positive on
the branch A C . T h e r e f o r e , by (43), i f the dynamic s y s t e m ( 3 1 ) c o r r e s p o n d s
to a point on the branch A B , i t s equilibrium s t a t e with u = 0 is a multiple
focus o r c e n t e r (we shall see l a t e r on that this is a multiple focus of
multiplicity 1). If the s y s t e m c o r r e s p o n d s to a point on the branch A C , i t s
equilibrium s t a t e with u = 0 is a saddle point.
5438 450
9 ?4. EX.-\\IPLES
The branch A B of the hyperbola ( 4 1 ) p a r t i t i o n s the region SI into two

s u b r e g i o n s 11" and IIb ( F i g u r e 195).
Let u s c o n s i d e r the c h a r a c t e r of the equilibrium s t a t e s of s y s t e m s
corresponding to the points of each of t h e s e r e g i o n s . First note that for
any values of the p a r a m e t e r s y , 6 in (31), infinity is absolutely unstable.
Indeed, along a path of :;ystem ( 3 1 )
It i s readily s e e n tha: f o r 2 + y2 > 6* the last e x p r e s s i o n is negative.

T h e r e f o r e all paths of the s y s t e m e n t e r into a c i r c l e of a sufficiently l a r g e
r a d i u s c e n t e r e d a t the o r i g i n as t i n c r e a s e s .
Hence it follows, in p a r t i c u l a r , that the s u m of the P o i n c a r e indices
of all the equilibrium s t a t e s i s f 1.
S y s t e m s corresponding to points of region I, a s w e have s e e n before,
have one equilibrium state, which is m o r e o v e r s t r u c t u r a l l y stable. Since
i t s P o i n c a r e index is 1, i t is e i t h e r a node o r a focus. It is r e a d i l y s e e n -
the infinity being absolutely unstable - that this node (or f o c u s ) i s stable.
In region 11, the s y s t e m h a s t h r e e s i m p l e equilibrium s t a t e s whose
Poincard indices add up to 1. T h e r e f o r e one of t h e s e s t a t e s is a saddle
point, and e a c h of the o t h e r two i s e i t h e r a node o r a focus. I t c a n be shown
( w e omit the proof h e r e ) that the nodes o r the foci of a s y s t e m c o r r e s p o n d i n g
to a point inside region IIb are stable. On the o t h e r hand, the points of the
branch AB of c u r v e (41)r e p r e s e n t s y s t e m s with t h r e e equilibrium s t a t e s ,
one of which i s a focus Hith 0 = 0 .
LVhen w e c r o s s o v e r f r o m IIb into IT" a c r o s s the branch A B , o r e v e r s e s
i t s sign and one of the foci consequently b e c o m e s unstable. The points of
region II" t h e r e f o r e r e p r 3 s e n t s y s t e m s which have one s t a b l e and one
unstable focus (or node) .and one saddle point.
A s we have established before, the points of the c u r v e (35) o t h e r than .V0
c o r r e s p o n d to s y s t e m s which have one s i m p l e and one double equilibrium
s t a t e . On c r o s s i n g o v e r into region I, the double equilibrium s t a t e v a n i s h e s ,
and on c r o s s i n g o v e r into region I1 i t d e c o m p o s e s into two s i m p l e equilibrium
states.
The point M Oc o r r e s p o n d s to a s y s t e m with a triple equilibrium s t a t e .
T h e points of the branch A B of c u r v e (41) r e p r e s e n t s y s t e m s with one
multiple focus, and the points of the branch d C r e p r e s e n t s y s t e m s which
have a saddle point with (r = 0. Finally, the point A c o r r e s p o n d s to a double
equilibrium s t a t e with 0 == 0.
L e t u s now elucidate the position of the equilibrium s t a t e s in the phase
plane.
To this end, w e c o n s i d e r the i s o c l i n e s of horizontal and v e r t i c a l inclina-
tions. The isocline of the horizontal inclinations Q (2,y) = 0 is the p a r a b o l a
+
Y = I (5 Y). (44)
and the isocline of the v e r t i c a l inclinations P ( s , Y) - 0 is
y ( z f y) +z+6 =O. (45)
For 6 = 0 , this isoclin'? d e c o m p o s e s into two s t r a i g h t l i n e s
If y + 6, (45) is a hyperbola and, f o r e v e r y 6, the l i n e s (46) are i t s

jymptotes.
Figure 1 9 6 shows the isocline Q (z,y) = 0 and the family of the isoclines
( x , y) = 0 for a fixed y = y o >
4
X=-%
+ I-
b
FIGURE 196. a) 6 > yo; b) d E yo; c) 8 < yo.
On the isocline of the horizontal inclinations - the parabola y = ( x y) 5 -+

nodes and foci a l t e r n a t e with saddle points (by the Poincare' t h e o r e m , 823.3,
Theorem 3 6 ) . Since out of the possible t h r e e equilibrium s t a t e s , two are
nodes o r foci and one is a saddle point, the two e x t r e m e equilibrium s t a t e s
on the parabola a r e nodes or foci, a n d t h e m i d d l e s t a t e is a saddle point.
Let u s now e s t a b l i s h which of the two equilibrium s t a t e s that are not a
saddle point changes i t s stability on c r o s s i n g o v e r f r o m IIb into II', i . e . ,
which of the s t a t e s h a s a = 0. Since a = - y - 2, I/ f 2 cannot r e t a i n a
constant sign for this equilibrium state. Consider the equilibrium s t a t e
0, with the l e a s t a b s c i s s a . It is r e a d i l y s e e n that its ordinate vi is always
g r e a t e r than -1, since 0, is the i n t e r s e c t i o n point of the parabola with the
branch of the hyperbola extending above the asymptote y = - 1 ( F i g u r e 196).
Thus, y1 + 2 =-
0, i.e., only the focus with the l a r g e s t a b s c i s s a may change
i t s stability.
* If y < 3, cp ( y , d) > 0 (see (35)) and system (35)has one equilibrium state for a l l 6.
452
5 34. EX.1hlPLES
Let u s now c o n s i d e r the e x i s t e n c e of l i m i t c y c l e s f o r s y s t e m (31).

W e w i l l f i r s t show that on passing f r o m branch AB of c u r v e (41) into
region I', the multiple focus will c r e a t e p r e c i s e l y one limit cycle, which
w i l l be unstable.
Using the s a m e substitution of v a r i a b l e s as before, i.e., displacing the
o r i g i n to coincide with :he multiple focus, w e obtain
dY
E = x +- (y- 6 ) Y -XY, dt = (26- 3y) x- Y +x2. (42)
dt
To d e t e r m i n e the c h a r a c t e r of the equilibrium state, l e t u s find the value

of a3 (see S24.4, (76)). Computations show that the s i g n of a3coincides with
the s i g n of the r a t i o
36-4y
26-3y '
(47)
We thus have to d e t e r m i n e the sign of this e x p r e s s i o n at the points of the

c u r v e A B . This c a n be done as follows. Consider a n a u x i l i a r y s t r a i g h t line
56-8y = 0, (48)
which joins the o r i g i n with the point -4( 5 4 0 . 8 v5). It is readily s e e n that
the branch d B of c u r v e (41) and the h a l f - l i n e s 36 - 47 = 0 and 26 - 3 y = 0
located in the first quadrant l i e on the two s i d e s of the line (48).
E x p r e s s i o n (47) thus h a s the s a m e s i g n on the branch A B and on the s t r a i g h t
line (48). On the line (48) this e x p r e s s i o n is positive. Hence, as> 0 .
Then, a c c o r d i n g to the table at the end of 25.3, the multiple focus is
unstable and its multiplicity is 1; on c r o s s i n g o v e r into region IIb, it b e -
c o m e s s t a b l e and c r e a t e s a single unstable l i m i t cycle.
Let u s now i s o l a t e c e r t a i n r e g i o n s in the p a r a m e t e r plane which c o r r e -
spond to s y s t e m s without closed paths (in p a r t i c u l a r , without limit c y c l e s ) .
Consider the s t r a i g h t line x + y = O . W e s e e f r o m the f i r s t equation in ( 3 1 )
that f o r 6 = y this line is an i n t e g r a l c u r v e . If 6 # y a i t is a line without
contact. Indeed, for x = - y a n d d # y , ddt+ --
d-y#O, i.e., the paths do not
touch this line.
Take Dulac's function i n the f o r m
This gives
T h i s e x p r e s s i o n r e v e r s e 5 its s i g n on the s t r a i g h t line
z=6--2y. (51)
Hence, by Dulac's c r i t e r i o n , i t follows that e v e r y closed path of s y s t e m (31)

c r o s s e s the line (51).
For 6 = y the line (51) coincides with the line x +
y = 0, which is a n
i n t e g r a l c u r v e . Therefore, in this case the s y s t e m h a s no closed paths.
453
F o r 6 ( y , the line (51) p a s s e s to the left of the line x + y = 0, and all

the equilibrium s t a t e s of the s y s t e m lie to the right of this line (this c a n be
checked if we r e m e m b e r that for 6 y , the isocline Q ( x , y) = 0 (the p a r a b o l a )
may only i n t e r s e c t the right branch of the isocline P ( x , I/)= 0 (the hyperbola),
i . e . , the branch which extends to the right of the line x + y = 0; s e e
F i g u r e 196c). E v e r y closed path of the s y s t e m should c r o s s the s t r a i g h t
line (51) i n v i r t u e of Dulac's c r i t e r i o n . On the o t h e r hand, e v e r y closed
path should enclose a t l e a s t one equilibrium s t a t e . But then the closed
path should i n t e r s e c t the line without contact x + y = 0 at l e a s t at two points,
which is impossible.
We have thus established that f o r 6,(y the s y s t e m h a s no closed paths.
We w i l l now show that s y s t e m s corresponding to the points of region I
(i.e., s y s t e m s with one equilibrium s t a t e ) do not have closed paths e i t h e r .
F o r 6 < y , this h a s been proved above. Let 6 > y . In this c a s e , the
equilibrium s t a t e of the s y s t e m is the i n t e r s e c t i o n point of the parabola y =
= x ( x + y ) with the branch of the hyperbola P ( x , y) = 0 which extends to the
left of the line x + y = 0 ( s e e F i g u r e 196a), while the line (51) p a s s e s to the
right of the line x + y = 0. T h e r e f o r e , if t h e r e e x i s t s a closed path, i t
should c r o s s the i n t e g r a l c u r v e x+ y = 0, which is impossible.
Let u s now consider the s y s t e m s corresponding to the points of the line
( 3 5 ) , cp ( y , 6) = 0, i.e., the s y s t e m s with multiple equilibrium s t a t e s . The
line (35) c o n s i s t s of two b r a n c h e s (36) and ( 3 7 ) ( F i g u r e 195). It i s readily
s e e n that 6 < y on the branch (37) (and in p a r t i c u l a r a t the point M o ) .
T h e r e f o r e s y s t e m s corresponding to the points of this branch have no closed
p a t h s . Systems corresponding to the points of the branch (36) which l i e in
the region 6 > y have one s t r u c t u r a l l y stable state - a stable node or focus -
and one double equilibrium s t a t e with z e r o P o i n c a r e index. It is readily
s e e n ( s e e F i g u r e 196a) that the line without contact x + y = 0 p a s s e s between
the focus (or the node) and the line (51) in this c a s e . T h e r e f o r e , by
Dulac's c r i t e r i o n , no closed path enclosing a focus may exist. But then
the closed path should e n c l o s e the double equilibrium state, and this is
i m p o s s i b l e because i t s Poincare' index is z e r o .
Multiple equilibrium s t a t e s corresponding to the points of c u r v e ( 3 5)
can b e investigated using the r e s u l t s of QT, J 2 1 and J 2 2 . The point A
( F i g u r e 195) is found to r e p r e s e n t a s y s t e m with a degenerate equilibrium
state, and point Ma a s y s t e m with a stable topological node of multiplicity 3
( w e have indicated this b e f o r e ) . All the o t h e r points of the c u r v e cp (3, y) = 0
r e p r e s e n t s y s t e m s f o r which the double equilibrium s t a t e is a saddle-node.
It is readily s e e n that a s y s t e m with a saddle-node cannot have paths
forming a loop which goes to the saddle-node both f o r t -c - cw and t - c + 00.
T h i s is proved in the s a m e way a s t h e a b s e n c e of closed paths.
Now c o n s i d e r some point So on the segment A D of c u r v e (35) ( F i g u r e 195).
The dynamic s y s t e m corresponding to this point h a s no closed paths and no
paths f o r m i n g a loop, as w e have s e e n j u s t now. Moreover, this s y s t e m
h a s no multiple foci (a # 0, s i n c e So does not coincide with the point A ) .
But then w e c a n show that dynamic s y s t e m s corresponding to points
sufficiently c l o s e to So do not have closed paths e i t h e r ( t h e r e is "nowhere"
t h e s e paths c a n be "created" f r o m ; the r i g o r o u s proof of this proposition is
left to the r e a d e r ) . In p a r t i c u l a r , s y s t e m s corresponding to the points
sufficiently c l o s e to So in r e g i o n 11, i.e., s y s t e m s with t h r e e equilibrium
s t a t e s , have no l i m i t c y c l e s and no closed contours consisting of paths.
454
5 34. EX.4hlPLES
Consider the point S, which is sufficiently c l o s e to So in-IIa and point SZ

which l i e s in IIb below the s t r a i g h t line y = 6 . Let 3, and S, be the d,vnamic
s y s t e m s corresponding to these points. Since n e i t h e r s y s t e m h a s closed
paths and closed contours consisting of paths, their topological s t r u c t u r e
is determined unambiguously. Let u s prove this fact.
I. System $. It h a s one s t a b l e node or focus, one unstable node or
focus, and a saddle point. Since the infinity is unstable, the two a -
s e p a r a t r i c e s of the saddle point go to a s t a b l e focus.' The unstable focus
(or node) c l e a r l y lies inside the closed c u r v e f o r m e d by the a - s e p a r a t r i c e s ,
one of the o - s e p a r a t r i c e u of the saddle point goes to this unstable focus
for t + - 00, and the o t h e r goes to infinity. The corresponding topological
s t r u c t u r e i s shown in Figure 197.
FIGLIRE 197 FtGCRE 1'?8
SI. System s",. It h a s two s t a b l e n o d e s o r foci. 'The two o - s e p a r a t r i c e s

of the saddle point go to infinity f o r t + - 00 and s e p a r a t e the a - s e p a r a t r i c e s ,
which go to the foci for t + + 00. IXe
r e s u l t i n g topological s t r u c t u r e is shown
in F i g u r e 198.
Now c o n s i d e r the points of region I1
which are not c l o s e to So, but f o r which
we n e v e r t h e l e s s have 6 > y. To
e s t a b l i s h the possible configurations of
the s e p a r a t r i c e s of the s y s t e m s c o r r e -
sponding to these points, c o n s i d e r the
i s o c l i n e s of the horizontal and v e r t i c a l
inclinations, i.e., c u r v e s (32). These
i s o c l i n e s .partition the plane into r e g i o n s
in which x and y r e t a i n a constant sign
( F i g u r e 199).
FIGL'RE 1'23 Since 6 > y , one of the equilibrium
s t a t e s (0,on F i g u r e 199) lies to the left
of the line z = - y o , and the o t h e r two
equilibrium s t a t e s , O2 and O,, lie to the right of this line ( F i g u r e 196a). The
equilibrium s t a t e O2 is a saddle point, and O3is a s t a b l e focus or node. It is
' They form a closed line c o n w t i n g of paths, but these paths a r e not a continuation of one another. When
w t said above that the system had no closed contours consisting of paths, we meant contours of the type
ol limit continua.
455
Ch.XIV, APPLICATIONS OF THE THEORY OF BIFURCATIONS
r e a d i l y s e e n that the four s e p a r a t r i c e s of the saddle point 0%e n t e r - as t

i n c r e a s e s or d e c r e a s e s - into four of the previously identified r e g i o n s
adjoining the point 02 (this c a n be shown by the s a m e technique as that
used to c l a s s i f y the saddle point: s e e QT, 57.3). The following p r o p e r t i e s
can a l s o be established without difficulty:
The s e p a r a t r i x located n e a r the point O2 in the region where 3: < 0, < 0 a;
is a n o - s e p a r a t r i x . We designate i t L:. A s t d e c r e a s e s , it e i t h e r c r o s s e s
the isocline of horizontal inclinations and l e a v e s the region where
< 0, 3; < 0, o r i t does not leave this region, going to infinity instead
(i.e., i t c r o s s e s the cycle without contact of the s y s t e m ) .
The s e p a r a t r i x L; issuing f r o m the saddle point Ozinto the region where
> 0, y 0 c r o s s e s the isocline y = x ( x + y ) with i n c r e a s i n g t and e n t e r s into
the region where x > 0, I/ > 0 ( F i g t r e 199). Indeed, if the s e p a r a t r i x L; d o e s
not leave the region where x > 0, y < 0 a c r o s s the parabola, i t should go to
the equilibrium s t a t e Oi. This is impossible, however, since the saddle
point O2 is located below the point 01, and in the relevant region g -= 0, i.e.,
y d e c r e a s e s along L;.
W e a r e dealing with the c a s e when 6 > y , i.e., when the s t r a i g h t line (51),
x = 6 - Zy, p a s s e s to the right of the line without contact x = - y . In this
c a s e , a closed path, if i t e x i s t s , may not enclose the equilibrium state Os,
and i t only may enclose the focus (or the node) Ol.
It follows f r o m the above that t h r e e configurations of s e p a r a t r i c e s a r e a
p r i o r i possible f o r the s y s t e m s being considered (i.e., s y s t e m s with t h r e e
equilibrium s t a t e s and 6 > y ) :
1) The s e p a r a t r i x L: goes to infinity a c r o s s a cycle without contact, and
the s e p a r a t r i x L; e i t h e r goes to a s t a b l e node or focus Oi( a s for 6 < y ), o r
goes to a s t a b l e or s e m i s t a b l e l i m i t cycle e n c i r c l i n g the equilibrium
s t a t e 0%( F i g u r e 198).
2 ) The s e p a r a t r i x L; goes to a s t a b l e node o r focus 0,, forming a closed
c u r v e together with the s e p a r a t r i x L;, which e n c l o s e s the s e p a r a t r i x L:.
The s e p a r a t r i x L: g o e s to a n unstable node o r focus O$o r to a n unstable (or
s e m i s t a b l e ) cycle encircling the equilibrium s t a t e Oi( F i g u r e 197).
3 ) The s e p a r a t r i x L; m e r g e s with the s e p a r a t r i x L: to f o r m a loop
( F i g u r e 204).
Arguing a s before, we c a n show that i f configuration 1 (configuration 2 )
is observed f o r c e r t a i n v a l u e s of the p a r a m e t e r s , the s a m e configuration
is r e t a i n e d f o r all c l o s e v a l u e s of the p a r a m e t e r s .
A s we have s e e n before, however, both configuration 1 and configuration 2
a r e possible f o r different v a l u e s of the p a r a m e t e r s .
Thus, r e a s o n i n g p r e c i s e l y as the previous example, w e w i l l e s t a b l i s h
the existence of c e r t a i n values of the p a r a m e t e r s f o r which the s e p a r a t r i x
extends between saddle points. On e v e r y continuous line joining a point of
the type Sf with a point of the type Sz t h e r e evidently e x i s t s a point S, which
r e p r e s e n t s a s y s t e m with a saddle-to-sad
that region I1 contains a t l e a s t one continuous line r which p a s s e s between
the s t r a i g h t line 8 = y and the branch A B of c u r v e (41), whose points
c o r r e s p o n d to dynamic s y s t e m s with a saddle -to-saddle s e p a r a t r i x .
We c a n now investigate the topological s t r u c t u r e assuming, as before,
that no closed paths f o r m f r o m path condensations and that all the points
corresponding to s y s t e m s with a s e p a r a t r i x loop f o r m a non-closed line r
(extending between the branch A B of c u r v e (41)and the line 6 = y
( F i g u r e 200)).
456
8
D
I p
r/,
E
branch A D of the c u r v e q = 0 (other than the
point A ) have a stable node or focus and a
saddle-node with a n unstable node region.
Their topological s t r u c t u r e is shown in
F i g u r e 202.
On passing f r o m branch A D to region 11,
w e obtain s y s t e m s with one s t a b l e and one
unstable node o r focus, whose topological
s t r u c t u r e is shown in F i g u r e 197.
The topological s t r u c t u r e of the s y s t e m s
on the line A D is analogous to the s t r u c t u r e
FICL'RE 201 FIGURE 202
On r8 the l i m i t cycle v a n i s h e s , a f t e r "being swallowed up" by the loop

of the s e p a r a t r i x which o r i g i n a t e s and t e r m i n a t e s in the saddle point O?.
The topological s t r u c t u r e of the s y s t e m is shown in
F i g u r e 204.
@
In region 3 3 ( F i g u r e 2 0 0 ) , the s e p a r a t r i x loop
b r e a k s up and the s y s t e m a c q u i r e s the topological
s t r u c t u r e shown i n F i g u r e 198. The point M,,corre-
sponds to a dynamic s y s t e m with a triple equilibrium
state. It is a s t a b l e node, and the topological s t r u c -
t u r e of t h e s y s t e m is the s a m e as in F i g u r e 201.
Finally, the dynamic s y s t e m corresponding to the
point A h a s a s t a b l e focus and a d e g e n e r a t e
eq.iilibrium s t a t e . Its topological s t r u c t u r e is
FIGURE 203. s h o w n in F i g u r e 205.
457
FIGURE 204. FIGURE 205.
W e should emphasize that the preceding a n a l y s i s was c a r r i e d out using

f a i r l y a r b i t r a r y simplifying assumptions, whose validity i s by no m e a n s
certain.
E x a m p 1e 1 7 (the c r e a t i o n of a limit cycle f r o m a closed path of a
conservative s y s t e m ) .
Consider a system
which a r i s e s , in p a r t i c u l a r , in connection with a tube g e n e r a t o r operating

in the "soft mode" ( s e e / 6 / , p. 7 0 3 ) . Only positive values of the p a r a m e t e r s
a, p, y a r e meaningful, and we will therefore take in our a n a l y s i s u ==. 0,
B>O,Y>O.
System (52) may be considered f o r s m a l l p a s being c l o s e to the
Hamiltonian s y s t e m I
where H (5,
1
y) = T (x' + ye). The paths of s y s t e m (53) a r e c i r c l e s centered I
a t the origin.
Since (53) is a l i n e a r conservative s y s t e m , we may investigate system
( 5 2 ) using Theorem 75 ( 3 3 . 2 ) , o r the m o r e g e n e r a l Theorem 78. We w i l l
u s e Theorem 78 h e r e , a s it a l s o provides a n indication of the stability of
the c r e a t e d l i m i t cycle.
Let
x = pi cos t , y = pi sin t , (54)
where pl>O. be a closed path of s y s t e m (53), and Go the c i r c l e enclosed

within t h i s path. Let u s evaluate the i n t e g r a l
( s e e (B,) i n the statement of Theorem 78). We have
458
5 34. EXAhlPLES
T h e r e f o r e , condition (74) of Theorem 78 (S33.4) is satisfied if and only if

P , = i V / ,5.r
7
(we a r e evidently concerned only with positive p , ) .

Let u s compute the value of 1 f o r this pi (33.4, (75)). W e have
.ox
I = 1 ~ p i x ( p l c o s sp,sins)+-qi,(plcoss,
0
, p,sins)lds=
2.T
= 1 (a Bpi cos - yp: cos?
'7 s 2x2-
s) ds =
I ypi2n 1
2
0
- 2za - T1 y .2rr. *Y
= 2xa ( 0 .
Applying Theorem 78, we conclude that the closed path ( c i r c l e )
of the l i n e a r conservative s y s t e m (53) c r e a t e s - on -7.ssing to s y s t e m ( 5 2 ) -

a s t r u c t u r a l l y s t a b l e l i m i t cycle which is s t a b l e lur p > 0 and unstable f o r
P<O.
Ex a m p 1 e 18 (a s y s t e m c l o s e to a l i n e a r conservative s y s t e m ) .
This is a s y s t e m of type (B,) (33.2), and h e r e

p ( x , y, p) .=0, p ( x . y, p) = 'ly +By* ivu3 + 6y'- Eyj.
R e r e f o r e , by (47), 33.2,
2s
po, 0 ) =
$ (21; (ape sin e +- BpZ sin2e yp3 sins 6 .+
0
I$ (2n: p,,. 0) depends only on the p a r a m e t e r s a , y, and E.

In 3 3 . 2 w e considered the c r e a t i o n of l i m i t c y c l e s f r o m the paths of the
s y s t e m x = - y, y = x corresponding to the nonzero r o o t s of the equation
$ (2n; PO,0) = 0. The a r g u m e n t of that section, however, r e m a i n s valid for
a z e r o r o o t a l s o (provided s u c h a root e x i s t s ) . p o = 0 c o r r e s p o n d s to the
equilibrium s t a t e 0 (0. 0) of the s y s t e m x = - y. y = 2.
Thus, s e t t i n g po = 0, we are in fact dealing with the c r e a t i o n of a l i m i t
cycle f r o m a n equilibriurn s t a t e 0 (I). 0 ) of a l i n e a r c o n s e r v a t i v e s y s t e m .
For s m a l l p, the point 0 (0, 0) is a focus of s y s t e m (56), which is s t a b l e f o r
pa < 0 and unstable for va > 0 . T h e r e f o r e , the periodic solution c r e a t e d
f r o m the equilibrium state 0 (0, 0) when w e change o v e r f r o m the l i n e a r
s y s t e m to s y s t e m (56) with a s m a l l p is not a closed path ( i . e . > not a l i m i t
cycle). but r a t h e r the equilibrium s t a t e 0 (0, 0) itself.
Let u s investigate the nonzero r o o t s of the equation
rcI ( 2 G Pot 0)= nP0 ( a + 73 yP:-@) 5

=o, (58)
taking y > 0, e > 0, and assuming that a may take on both positive and
negative values (this c o r r e s p o n d s to the physical conditions of the actual
problem associated with this system; s e e f 6 f a Chapter IX, lo).
W e should consider only the positive r o o t s of (58). A s w e have s e e n
above, f o r a # 0, the equilibrium s t a t e 0 (0, 0) of the original s y s t e m may
not c r e a t e a limit cycle.
Setting in (58)
pa=r, a3 y = a , T5 e = b
and dividing through by np,, we obtain

a +a r - bra = 0, (59)
the r o o t s of this equation being a* F.

If a < 0 and aa+4ba < 0, equation (59) h a s no r e a l r o o t s .
+
If, on the other hand, a < 0 and aa 4ba> 0, the two r o o t s of equation (59)
a r e positive. T h e r e f o r e , equation (58) h a s two positive r o o t s which c o r r e -
spond to two l i m i t c y c l e s c r e a t e d f r o m the closed paths of the original
l i n e a r s y s t e m . It is readily s e e n that the conditions of Theorem 76 (S33.2)
a r e satisfied and, by this theorem, the s y s t e m h a s no other closed paths in
a sufficiently s m a l l neighborhood of 0. If p > 0, the equilibrium s t a t e
0 (0,0) is a stable focus for a < 0. Therefore, one of the l i m i t c y c l e s
c r e a t e d (namely, the inner c y c l e ) is unstable, and the o u t e r cycle is stable.
I
C d
FIGURE 206
460
S 34. EXAhfPLES
Between the (negative) a f o r a* $- 4ba <0 and those f o r which a* + 4ba > 0,
w e have one value a = a* = - $.c 0 f o r which a= +- 4ba = 0 . This value
c o r r e s p o n d s to a single positive root p: of equation ( 3 7 ) . Calculations show
that $A (2%; p:,O)= 0; therefore the theory of 5 3 3 . 2 is inapplicable to t h i s c a s e
( s e e r e m a r k to Theorem 75, 5 3 3 . 2 ) . It is r e a d i l y seen, however, that the
closed path of s y s t e m (56) corresponding to t h i s p i is a double l i m i t cycle
which s e p a r a t e s into t w o l i m i t c y c l e s - a stable and a n unstable one - a s a
increases.
A s a is f u r t h e r i n c r e a s e d and p a s s e s through z e r o , the unstable limit
cycle c o n t r a c t s to the equilibrium s t a t e , which becomes stable for a > O .
F i g u r e 2 0 6 , a - d, shows the path configurations of s y s t e m ( 5 6 ) c o r r e -
sponding to the c a s e s a < a * , a = a*,a* < a < 0, a > O . It is a s s u m e d that
p > 0 is sufficiently s m z l l .
Remark . It is readily s e e n that the problem of the c r e a t i o n of a l i m i t
cycle f r o m a n equilibricm s t a t e of the type of a c e n t e r cannot be solved
with the aid of the sufficient conditions of 5 3 3 . 2 . Indeed, these conditions
only prove the existence of a n equilibrium s t a t e p = 0 f o r s y s t e m (B&).
Investigation of the nonzero r o o t s of the equation 9 ( 2 ~ p, ; p) = 0, which
go to z e r o f o r p -+ 0, may naturally shed s o m e light on the problem of
c r e a t i o n of a l i m i t cycle f r o m a c e n t e r - t y p e equilibrium s t a t e of s y s t e m
(Bo). Note that the solution of the equation 9 (2n; p, p) = 0 corresponding
to this limit cycle should be sought a s a n expansion in fractional, and not
integral, powers of p, and, in the s i m p l e s t c a s e , a s a n expansion in powers of
6.
461
APPENDIX
1. T h e o r e m s of the continuous dependence of the

solutions of a s y s t e m of differential equations on the
right -hand s i d e s and of the differentiability of solutions
Let
-= r k ( t , xi, "'2, . . . , xn), k = l , 2, . . ., n, (A)
be a s y s t e m of differential equations defined in region G of the (n 1)- +

dimensional s p a c e . Let P I be continuous functions which have in G con-
tinuous p a r t i a l d e r i v a t i v e s of f i r s t o r d e r with r e s p e c t to the v a r i a b l e s
51, 1 2 , . . ., Xn .
Together with (A), we c o n s i d e r a modified s y s t e m
dzk - (t,
Pk 51, .. . t Zn) = Pk (tit 51, ... I sn) + p k It19 51,52, * . ., Xn), (A)
-=
dt
where P k and hence the " i n c r e m e n t s " a r e functions satisfying the s a m e

conditions a s the function Pk (k = 1, 2, . . .. n). If we are dealing with a
s y s t e m dependent on a p a r a m e t e r ,
the transition f r o m the value po of the p a r a m e t e r to another value p c o r r e -

sponds to a transformation f r o m s y s t e m (APO) to a modified s y s t e m (A,,),
and the i n c r e m e n t s a r e the functions
h = p k (t. 5 1 9 x2, * . Y xnv p ) - P k ( t , 51, 521 * .. v Xn, Po).
The situation is completely analogous f o r a s y s t e m dependent on s e v e r a l

parameters.
T h e o r e m 1 (theorem of the continuous dependence of solutions on the
right-hand sides of the equations). Let
Xk=(Pk(t;tO,XioIXzo, ..., X n o ) = ( P k ( t ) , k = l , 2, ..e) n, (1)
be the solution of s y s t e m (A) satisfying the initial conditions
Pk %%Oi
l, ... I G O ) =ckg (21
which i s defined f m all t from the interval (7, T ) , where r c t o < ~ . Let,
462
further
-
:rA = 'PA ( t ; t o ? TlOt XZO, .. t %;to) = (PA ( t ) ( 3I
be the solution of syste?m (A) satisfying the same initial conditions. Then,
f o r any+ I and ~ 2 T
. < T ~ < ~ , < T ~ < T , and any e > O , there exists a>Osatis-fying
the follozoing condition: if I PA ( t ;t l ,x 2 , . .., I,,) I < 6, k = 1, 2, . .., n , solution ( 3 )
i s defined for all 1 , T, G:t 4T ~ ,and f o r ecery t from this interval we have
1 (Fk ( t ; ZIO. I?,, * - 7 InO)-(Fk ( t ;t o , 1 1 0 . I?o* ... Xno) I< E . i4)
P r o o f . W e will only consider t E It,, ~ ~ 1The . s a m e reasoning can be

applied to t E IT^, tole Choose s o m e fixed E > 0.
Let GI be a closed bounded region completely contained i n G, which is
convex in a l l the coordinates I , (i = 1. 2, . . .. n ) and contains the p a r t of the
integral c u r v e ( 1 ) corresponding to t E [ t o , ~ ~ 1 . : ' Let po be the distance of
this p a r t of the integral c u r v e (1) to the boundary of GI (evidently, p , > 0).
Since El is a bounded closed region, the following inequalities are satisfied
at ei'ery point J I ( t ; I + ,r2, . . .. zn)of this region:
Ip i x , ( t i 11. 12. .. ., ~ n I <

) B. (5)
where B >0 is a constant.

Let
Ek
-
= '?k ( t ; to* IO, x?oo, . . . * XnO) (ti to, I?o, . . . ,X n o ) . (6)
The functions E k are defined f o r all t f o r which the functions <k and (Fk are
a p r i o r i defined. They are therefore a p r i o r i defined for all t sufficiently
c l o s e to to. For t = t o , Ek = 0. Clearly,
-PA ( t ; '?I, (F?. ... ? (Fn) T Pk

- - . ., -(Fa) =
( t ; T I ?'P?. ,
= 3 P h x , (ti q-k l - qk?.

n
- . . . * f-l k n ) E t 7PA
- -
( t ;' P I . (Fz. .. . t (Fn). (7)
i=1
where the point ( t : &, GI:, . . . , &,) lies on the segment joining the points
. c -
.
( t ;91, $ 2 , . . . Tn) and ( t l ; q l , Q. . . . (Fn).
Consider a n auxiliary differential equation
where c = u ( t ) is a function of t , and 6>0 is a constant. The solution of this

equation satisfying the #condition v ( t o )= O is
-
6 [ e n B ( f - - t o ) - 11. (9i
B
Convexity in all the coordinates ri indicates that q,

together with any pair of its points M ( t , z I . x ? , . . ., z,,)
.
and M' ( t , si, . ., r k ) , also contains the entire segment MM'.
363
APPENDIX
6 [enB(r*-to) - 11
0 <v ( t )< (10)
and v' ( t )> 0.
Let e * = min {e, $}. For b>O we choose any number satisfying the relation
6 [enB(ra--lo)-ll ce*.
(11)
Then, by (lo), for all t , to<t<zz,

v ( t )<e*. (12)
W e w i l l now show that the 6 chosen in this way s a t i s f i e s the proposition of
the theorem.
Let
IPk (ti 5 1 3 221 .. .Y zn) I <6 . (13)
Suppose that the point ( t ,Gt,

-
&, . . .. q,,) and t h e r e f o r e the e n t i r e segment
joining i t with the point ( t , q i ,V Z , . . ., Vn) a r e contained in Gi. By (7), (5),
and ( 1 3 ) we then have (since ( t , &, . . . , ;)"Rn) is a point of the relevant segment) '
I % l < B i I iS=*i l + b . (14)
This inequality is a p r i o r i satisfied for all t sufficiently c l o s e to to.
For t=to. Ei=O. Therefore, f o r t sufficiently c l o s e to to, 1$l<b, and hence
1% I<na.
On the other hand, f o r a l l tEIto. TZJ,
v' ( t ) = nBu + 6n >/ n6.

It follows f r o m the l a s t inequalities that for a l l t sufficiently close to
to ( t 2 to),
?*(T~ o r it is defined for a l l t E [to,t z ] , but f o r some t h this segment a t

l e a s t one of the inequalities
I&k ( t ) I <e, (16)
464
APPEXDIX
and hence a t l e a s t one of the inequalities

IEA ( t )I <e* (17)
is broken.
In the f o r m e r case, t h e r e e x i s t s t, f o < f < T * < T 2 , f o r which the
corresponding point of the integral c u r v e z1=Gi ( t ) lies outside Gl (this
follows f r o m QT, Appendix, S8.1, Theorem A). Then i t s distance
f r o m the point ( f , cp,(t), t p z . . . ., cp,(t)) in G, should be greater than po, i.e.,
p E 1 ( t ) * + .. . + E n ( t ) * > p o . This is possible, however, only i f for s o m e k ,
[Ek(t)l>%.>e*. n u s , :.n e i t h e r case, f o r s o m e t*, to<f*<T,, and a t l e a s t
for one k, 1Q k g n , w e have the inequality I z k (t*)1 >e*, and hence, by (12),
IEk ( t * ) 1 > v ( t ) - (18)
On the o t h e r hand, by (15), f o r a l l t sufficiently c l o s e to to and for all
i = 1 , 2, . . . , n, w e have
IEt ( t )I < 27 (t). 119)
It follows f r o m (18) and ( 1 9 ) that the interval of t values where the two
solutions cpt(t) and G t ( t ) a r e defined contains a point tl, t0<t,,<.r2, which
s a t i s f i e s the following conditions :
(a) f o r all t , t o < t < ti, and f o r all i = 1 , 2, . . ., n,
IEi ( t J IQ (ti), (21)

and a t l e a s t f o r one of these i, i = k say,
1 Ek (ti) I =U (ti). (22)
F r o m (a) and (b) and inequality (12), i t follows that f o r t E [ t o ,ti]
VE1(t) + (t)S + . . + En (t) c P i = c

* /qg< Pa-
i.e., the p a r t of the integral c u r v e rt=Gi(t) corresponding to t [ t o ,t ! ]lies in

ct. Then, by (7) and ( 8 ) and in v i r t u e of the assumption I p h I < 6 , k = l , 2, . . .. n,
w e have
i.e., I E k ( t l ) I < u ( t l ) r which contradicts (22). This contradiction p r o v e s the

theorem.
T h e o r e m 2 (thecn-em of the continuous dependence of solutions on the
right-hand sides and initial values). Let
XA=TA(t, tO,xl0? x20. - - * I %O)=FA(t) (p 1
465
APPENDIX
be the solution of system (A) corresponding to the initial values tor

xi0, xZ0, .. ., xno, which is defined in the interval ( T ~T),
, and
the solution of system (A)corresponding to the initial values To. Zlo, &,. . . ., x,,~.
-
Then, f o r any T~ and z2, r0 < q < t <72<T , and any E > 0, there exist 6 >0 and
q > 0 which satisfy the following conditions: if
(Pk(t.xl,XZ,...,xn)l<6, k=i, 2, . . , n ,
and
lG-tol<q u I.z*o-xItoI<?l, i = 1 , 2, . . ., n,
the solution (G) is defined f o r t , T , < ~ = S T ~ , and for every t f r o m this interval
I$k(t)-(Pk(t)I<E ( k = i , 2, . . ,n).
P r o o f of Theorem 2 is readily obtained if we change o v e r in (A) to the

v a r i a b l e s T and z k defined by the equalities
t ='c - t o +-10, xk = Zk-PhO fxko,
and apply Theorem 1, using the compactness of the segment ITl, r z l and the
fact that the solution(q) is defined on a segment [-cl-u, r 2 f u ] , where u is some
positive number.
C o r o 11 a r y 1. Let $2 be a compact s e t in the (n+2) -dimensional space
t , t o , XIO. PZO, . . .. G O , with the functions cpk(t, to; xlO,xz0, . . .. xno) defined e v e r y -
where in this space. For e v e r y E > O , t h e r e e x i s t 6 > 0 and q > O satisfying
the following conditions: if ( p k ( t ;xi, x2, . . ., Z ~ ) ~ < ~ ~ ~ ~ ' - - ~ " ~ I<t oT- &Jl < q ,
I
-
~ x k o - x h o ~ < qa_nd_the point ( t ' , to; xi,. . ., &)Ea, the functions (P,+ a r e defined a t
the point (t", to; %IO, . . . , Xno) and
c c
I (Pk (t", t o ;
-
110, ... 7
-
28~0) -(PA (t'. to; . . ., Xno) I <E
@ = = I , 2, ..., n).
Corollary 1 follows f r o m the previous theorem. It is proved again by
reductio ad absurdum, and the compactness of Q is used.
C o r o l l a r y 2 . For a s y s t e m
whose right-hand s i d e s a r e functions of c l a s s 1 in 1 , x,, xz, . . ., x,,, and

continuous functions of the p a r a m e t e r p. the solutions
a r e continuous functions in a l l the v a r i a b l e s f o r a l l those values of the

v a r i a b l e s for which they a r e defined.
Corollary 2 follows d i r e c t l y f r o m Theorem 2.
466
.APPENDIX
puow let the functions Pk and entering the right-hand s i d e s of (A) and
(A) be functions of c l a s s r > 1. In this case, the solutions ( F k ( t , to;rlo,q 0 .. . ..
. . . . % o ) , f o r all the values of the v a r i a b l e s f o r which they are defined, have
continuous (in all the v z r i a b l e s ) derivatives: ( a ) to o r d e r r -+ 1 with r e s p e c t
to t ; (b) to o r d e r r with r e s p e c t to all the v a r i a b l e s r i o ; ( c ) to o r d e r r Z +
with r e s p e c t to all the v a r i a b l e s , provided i t contains a t l e a s t one differentia-
tion with r e s p e c t to t ( s e e QT, Appendix, 8.3, Theorem B). An analogous
proposition applies to Cp,%.
T h e o r e m 3. Let &hesolution ( q )of system {A) be defined for all
t c ( T ~ ,T ) , and let ro < T ] < f < T~ e T . Then, for ecery e > 0 , there exist
6 > o and r) > o satisfyingthefollowingconditions: if (A) is 6-close to rank r
to \A) and if I io- to I < 1, I i i 0- x i 0 I < q (i = i , 2, . . ., n) the solution (Cp) i s
defined for all t , rl.<t<rt., and for these values of the variables
- ,).- _ . . ., -
i
dgr(t.
ot
r10, 1~0)-
oFI* Gj5 . , . &b

a v ~ ( t to,
rjti
, rl~.
d t i o cis::,
. .., r n o )
. .. ark 1 <E,
(23)
iti,+i,-t ...- ii,=msrfl, i,+z?+ ...+i n < r .
P r o o f . The p a r t i a l derivatives of the functions fFk s a t i s f y the following
s y s t e m of differential equations
with the initial conditions
(see QT, Appendix, 8 . 3 ) . The s i p a t i o n is anaiogous with r e s p e c t to the

p a r t i a l derivatives of the functions (Pk. The validity of Theorem 3 directly
f o l l o w s f r o m Theorem 2 , when the l a t t e r is applied to s y s t e m (24) with
initial conditions (25).
Consider the case when the right-hand s i d e s of the s y s t e m are functions
of s o m e p a r a m e t e r p. i.e., s y s t e m s of the f o r m
Jft
-- - P k ( t , 21, Z:, . . ., Znr p) ( k = 1, 2, . . ., n). (A,)
The functions P R are defined f o r all p, pi < p < p 2 , e v e r y w h e r e in s o m e

region G of the ( n T l)-dim?nsional s p a c e t , z,, x 2 , . . .. zn. These functions
are a s s u m e d to be continuous in all t h e i r a r g u m e n t s and to have continuous
p a r t i a l derivatives with r e s p e c t to a . I?, . . ., a;, and p to o r d e r r inclusive.
Let
x R L v k ( t t !o.Zio. . . . , X n o , p ) . k = Z , 2 , ..., 11, (26)
be a solution of (Au).
167
APPENDIX
Applying this formula to the function ~ ( t ) = f ( t z t,y ) and taking t = 1, we find

k--l
The l a s t t e r m in this equality is designated Q * ( I , y). Clearly,
The coefficient before i - " y a in the last e x p r e s s i o n is a continuous function

of I and y. W e designate i t g, (I,y). Let
and
P: (0,0)= 0.
Inserting ( 4 ) and (5) in (3), we obtain equation (l), where
Po (I,Y) = f (0,01,
k
pk (I,y) 2 &?a (0,0)
*=O
and
k
p* (I,.I) = y
*=O
lg, (I,Y) -gu ( 0 , 0)l i - a y o .
The theorem is proved. A s i m i l a r proposition c l e a r l y applies to functions

of n v a r i a b l e s .
3. The l e m m a about the normals of a s i m p l e smooth

closed curve
Let Lo be a s i m p l e closed c u r v e defined by the equations

z=(F(s), y=*(s),
where g, and * a r e periodic functions of period T. We a s s u m e that cp and $
a r e functions of the second c l a s s , and that g,' (s) and 9' (s) do not vanish
simultaneously for any s.
169
APPENDIX
Through each point M (9 (s), 9 (s)) of the path Lo we draw a normal to the
path a t that point and l a y off s e g m e n t s of the length 6j/rp (s)l + *
(s) on e i t h e r
side of the point M .
L e m m a 1 . If 6 > 0 is sufficiently small, no two segments of navmals
drawn through dgferent points of the curve Lo intersect.
P r o o f . Since Vrp (s)l + q (s)l is bounded f r o m above, i t
the l e m m a under the assumption that s e g m e n t s of length 6 a r e laid
off all the n o r m a l s . Moreover, i t suffices to c o n s i d e r the c a s e
when these s e g m e n t s a r e laid off on one side of the c u r v e Lo(either
all inside the c u r v e or all outside the c u r v e ) . We a s s u m e that both
t h e s e conditions a r e satisfied. Let the proposition of the l e m m a be false.
Then t h e r e e x i s t s a sequence of n u m b e r s 6, -+ 0 (6, > 0) and a sequence of
p a i r s of points P n , Qnof the c u r v e Lo such that the s e g m e n t s of n o r m a l s of
length 6, drawn through the points P, and Qn i n t e r s e c t . Since the c u r v e L, is
compact, we may take P, -+ P and Q, -+ Q. where P and Q a r e points of Lo.
Let u s f i r s t consider the c a s e when P and Q a r e two different points.
Let M , be the i n t e r s e c t i o n point of the n o r m a l s e g m e n t s through the points
P, and Q,. Then p ( P , , M,),<6,, p(Qn, M,),<6,, and by the triangle inequality
P (PT Q) <P ( P , Pn) + P (Pn,Mn)+P ( M n , Qn) -I-P (Qnt Q )Q I
4P (P,Pn) +26, fP (Qn, Q)
( F i g u r e 207). Since for a sufficiently l a r g e n, the right-hand s i d e of the
inequality is as s m a l l as d e s i r e d , w h e r e a s the left-hand side is constant,
this inequality cannot be satisfied. P and Q thus may not be two different
points, and we have to consider only the c a s e when they coincide.
We change over to a new r e c t a n g u l a r coordinate s y s t e m , placing the

origin 0 at the point P (which coincides with Q ) , and d i r e c t i n g the a b s c i s s a
a x i s along the tangent to Lo a t 0 ( F i g u r e 208). We r e t a i n the s a m e notation
a s before, i.e., the new coordinates a r e designated x and y and the
+
*
p a r a m e t r i c equations of the c u r v e Lo are s t i l l x = rp (s), y = (s). Clearly,
rp and 1p a r e functions of the second c l a s s and rp ( s ) ~ rp (s)* f 0 f o r any s.
L e t the point 0 of c u r v e Lo c o r r e s p o n d to the value so of the p a r a m e t e r s.
Near the point 0, the equation of the c u r v e Lo is y = f (2). Then
=a,
f (2)
v f (0) =o,
(8)
* (so) =o, rp (so) p 0.
470
Therefore, n e a r the point x = 0, the function f (I)h a s a continuous second
derivative f (I).
By assumption, the :lorma1 s e g m e n t s of length S, > 0 drawn through the
points P,,and 0,i n t e r s e c t f o r e v e r y n ( h e r e 6, -c 0, P, 0 , f& 0). Let
-f
(X,,, and l e t P, (a,, 6,) and Qn (an,b,). Then

.If,
- Yn) be a n intersection point,
a n f a n , and for n -+ 0, *e have a,, -.
0, z,,3 0. b, -+ 0, Kn -+ 0.
The equations of the n o r m a l s to the c u r v e Lo a t the points P,,and Qn are
respectively written in the f o r m
-
X-an = - f (an) (y - &), I-an = - f (&) (y -&).
Inserting f o r x and y in these equations the coordinates X,, and Y, of the
point .Cfn and subtracting t e r m by t e r m , w e obtain a f t e r s i m p l e manipulations
-
YnI/(~n)--f(a,,)~=an--anL(~n--b,) /(Gn)i-bn [/(Zn)-/r(~r.)j. (1)
F r o m the Lagrange formula
/ (G)- / (an) = (En) ( g n -an) (2 1

and
- -
bn- 6n = f -t (an) e t ( I n ) ( Z n - a n ) *
(an) (3 1
.xhere En and q R lie between a, and g,,and therefore go to z e r o f o r n -+ a.
Inserting ( 2 ) and ( 3 ) in ( L ) and dividing through by (&-an), w e obtain
I/( E n ) = 1 -7 / ( I n ) f (an) T bn/ (En)
and taking the limit
On the o t h e r hand,
Since b, -+ 0 and 6, + 0, we have k*,, - 0. Therefore
lim IF,/(E,,) = 0,
n-0
which contradicts ( 4 ) . This p r o v e s the l e m m a .
4. Proof of the differentiability of the function R ( p . 0)

w i t h r e s p e c t to p
&e will f i r s t prove a n auxiliary proposition. Let the function /(x, y) be

defined in the circle x2 - g 2 < R2 where i t i s continuously differentiable with
r e s p e c t to 5 and y to s o m e o r d e r N > 1 , and f ( 0 , 0)=0.
Consider the function j * ( p , 0) defined as follows:
f 8 ( p , O ) = p fi (pcos8, p s i n 0 ) f o r p#O.
(1)
f* (0, 8) = f ; (0, 0 ) COS 0 +- 1; (0,0)sin 0.
47 I
APPENDIX ~
Since by assumption f (I,y) is a function of c l a s s N , then f o r p # O and for

a l l 8, f * ( p , 8) h a s continuous p a r t i a l d e r i v a t i v e s with r e s p e c t to p to o r d e r N .
The next l e m m a e s t a b l i s h e s the existence and the continuity of the
d e r i v a t i v e s of the function j * ( p , 0) with r e s p e c t to p f o r p = O .
L e m m a 2. If f ( x , y) is a function of class N , 1 in the circle 2 + y14R1,
then
(a) for every 8, p (p, 0) is continuously dqferentiable in p to order N - i;
(b) for p +. 0, the function p - goes to z e r o u n g w m l y in e.
aP"
P r o o f . If f ( x , y) is a homogeneous polynomial of d e g r e e n 2 1, the l e m m a
can be d i r e c:tlv verified (it suffices to check the Droaosition for a s i n e l e v
t e r m of the f o r m It is m o r e o v e r c l e a r that i f the l e m m a is t r u e for

two functions, i t is a l s o t r u e for their sum. It is t h e r e f o r e sufficient to
prove the l e m m a f o r the function
N k
. .
k=i q=O
All the derivative of this function to o r d e r N inclusive vanish a t the point

(0,O). Thus, in our proof of the l e m m a , w e may a s s u m e without loss of
generality that all the d e r i v a t i v e s of the function f ( x , y) to o r d e r N inclusive
vanish a t the point (0, 0 ) . If this is so, we have f r o m T a y l o r ' s formula
p=o
where 2 and 7 a r e n u m b e r s lying between 0 and x and 0 and y > respectively,

which in g e n e r a l depend on the point (2,9). Inserting for x and y the p o l a r
coordinates pcose and p s i n e , we find f ( x , y ) = p N F F , T ) , where
== ~ g f $ - ~ , ~
N
1
F ( E , q) ( E , q) COSN-P e sinp 8. (3)
p=o
Similarly, using the Taylor s e r i e s of the functionf$)-rv-r,r,I < g < N-i, O,(r,<q,
under the s a m e assumption and r e m e m b e r i n g that the p a r t i a l d e r i v a t i v e s
of the function f ( x , y) to o r d e r N inclusive vanish a t (0, O), w e obtain
p=o
or
where
and zq, and (p, l i e between 0 and x and 0 and y, r e s p e c t i v e l y . F ( c , q) and

F,, ( E , q) a r e evidently continuous functions of E, q which vanish at E = q =O.
472
Differentiating the function
f (I,y) f (p C O S 0, p sin 0)
q t i m e s with r e s p e c t tc, p (i<qP--<S-l), we obtain
au
dp4 f ( x , y) = 2 C{f($rur ( r , y) cos+' 0 sinre. (6)
r=O
Inserting f o r f$!-?#, ($7, y) its e x p r e s s i o n f r o m (4)and dividing through by

pN-9. we obtain
r=o
- -
If p 40, w e have x -+ 0. y + 0, .,I. 3 0, Fqr -+ 0, and t h e r e f o r e F,, (I~,,
yur)-+ 0.
The e x p r e s s i o n
1 IjQf (G Y) -pq-N W!+, Y)
7 op'l OP', (7)
(1 <q<X--l) thus a l s o goes to z e r o f o r p-+ 0 (uniformly in e). It f u r t h e r

follows f r o m ( 2 ) that the e x p r e s s i o n
1
-
Pv-
f (I,Y) = p-"f (5, Y) (8)
a l s o goes to z e r o for p -4
0, uniformly i n 8.
Let u s now prove proposition ( a ) of the l e m m a . \Ve f i r s t take p + O and
apply the Leibnitz r u l e to obtain e x p r e s s i o n s for d e r i v a t i v e s of o r d e r k of
the function
1
f*(p, e ) = p f ( p c o s O , p s i n 8 ) ( s e e (I)),
where l , < k , < N - l .
The Leibnitz r u l e gives
k
- ( 2 C ~ ( - ~ ) k - q ( k - ~ ) ! p apq
q - . ~ ~ ) p ~ - k - i(9)
.
q=o
Since (8) and ( 9 ) for p -+ 0 go to z e r o uniformly in 8, w e c l e a r l y have f o r

P 4 0 ,
uniformly in 8. M o r e o v e r , by (1) and (2), f o r p-0,
uniformly i n 0.
f ' be) - 0 (11)
By (10) and (ll), proposition ( a ) of the l e m m a w i l l be proved if w e show

that f* (0, 6) = 0 and that the function f* (p, 0) f o r p = 0 (i.e., a t any point (0, e))
473
APPENDIX
is differentiable with r e s p e c t to p to o r d e r N - 1 and the corresponding

d e r i v a t i v e s vanish. Clearly ,
f * (0, e) =o
according to the original definition of the function f* (p, e) (since f; and f; by
assumption vanish a t the point (0,O)). Taking N > 1, we w i l l prove the
existence of the f i r s t p a r t i a l derivative with r e s p e c t to p of the function
f* (p, 8 ) a t the point ( 0 , O ) . E x p r e s s i o n (9) is c l e a r l y inapplicable, since i t
holds t r u e f o r p # 0 only. We should thus compute the p a r t i a l derivative
directly, as the l i m i t of the e x p r e s s i o n
The f a c t o r in p a r e n t h e s e s is (8), which f o r p -+ 0 goes to z e r o uniformly in 0.

The l i m i t of (12), i.e., the derivative of the function f* (p, 6) a t the point (0, 6).
thus e x i s t s and is equal to z e r o . By (lo), we thus conclude that the f i r s t
derivative with r e s p e c t to p of the function f* (p, 6) is continuous a t p = 0.
Now suppose that we have established the existence and the continuity I
of the p a r t i a l d e r i v a t i v e s with r e s p e c t to p to o r d e r p , < N - 2 for the
function f* (p, e) a t the point (0, 6). Then, by (lo), all t h e s e d e r i v a t i v e s (to
o r d e r p inclusive) vanish.
Let u s find the ( p f l)-th derivative with r e s p e c t to p of the function f (p, 0)
at the point (0, . . 0). by the d i r e c t techniuue, i.e., as the l i m i t of the e x p r e s s i o n
Using equality (9) (for k - p , < N - 2 ) and the p r o p e r t i e s of ( 7 ) and (8), we c a n

d i r e c t l y show that the l a s t e x p r e s s i o n goes to z e r o f o r p + O . This indicates
that the derivative Bp+fBpP+
e) a t the point (0, e) e x i s t s and is equal to z e r o .
Then by ( l o ) , this derivative is continuous a t the point (0, e). We have thus
proved proposition (a) of the l e m m a .
Let u s now prove proposition (b). F o r p p 0 , f(+, y ) = f ( p c o s 6 , p s i n 6 ) =
= f * (o. 6) o. Using the Leibnitz rule.
F o r o -+ 0, the left-hand s i d e cfgoes to z e r o uniformly in 6 (this follows

BPN
f r o m (6), which is a l s o valid f o r q = N , and f r o m our assumption concerning
the function f (z~ g)). The second t e r m on the r i g h t in (14)differs f r o m (10)
by a constant f a c t o r only, and i t t h e r e f o r e a l s o goes to z e r o uniformly in e
aNf*
for p -+ 0. But then the f i r s t t e r m on the r i g h t p - necessarily goes to z e r o
aPN
uniformly i n 8. Q. E . D.
Let u s now proceed with the m a i n proposition of this subsection. Consider
the s y s t e m
d+ du
x=--PBy+cp(z, Sr)? -;ii-=&+w+cp(x, Y)? (15)
414
where $ # O . and the functions rp and 9 together with their f i r s t derivatives
vanish a t the point (0,C). Let
F ( p , 8)=cr~+cp(pcos6, p s i n 8 ) c m O f g ( p c o s 8 , psin8)sine. (16)
f o r p + O and
(I,(P, e) = 9 (p cos 8, p sin 6) cos e - P (0 cos 6, P sin f4 sin e
P
e) = 0,
(I,(0,
P
} (17)
F ( P - 6)
R(pv e) = p+(0(p, e) (18)
L e m wt a 3 . If (15 ) is a system of class IV > i and p* >o i s sufficientljr

small, the function R (p. e) has continuous partial derivatives with respect
to p to w d e r S inclusive at every point (p, e ) of the region --oo <e< T-oo,
O,<lpl<p* (Lemma 3, 024).
Proof. J3e condition O(0, q E O shows that f o r p=O the functions
rp(pcos8. pctin6) and *(pcostl. p s i n 6 )
in (17) are defined and are equal to z e r o .
P 0
It is thus c l e a r that these functions a r e generated f r o m the functions ~ ( 2y),

and $(z, y)> respectively, l i k e the functions f * ( p , e) f r o m f ( 2 , y) (see ( I ) ) .
Therefore, L e m m a 2 is applicable to the functions
-
CF ((I cos 6. p sin 6)
and
P
and, using equations (15)- (18) and the inequality $ +- (I, (p, e) # 0 (which is
satisfied f o r e v e r y point (p, e) of the relevant region f o r a sufficiently s m a l l
p*), w e conclude that the function R (p, e) a l s o h a s continuous p a r t i a l
d e r i v a t i v e s with r e s p e c t to p to o r d e r .V - 1 inclusive a t e v e r y point
(p, 0) ( 0 4 I p I < p* for any 0 ) and to o r d e r X inclusive a t e v e r y point where p # V.
Lye thus only have to establish the existence of a continuous .V-th derivative with
r e s p e c t to p a t points where p = 0. We r e p r e s e n t R (p. e ) a s a product of two
functions
av- 1
and find -
aph-f using the Leibnitz r u l e (this is p e r m i s s i b l e , since each of the
f a c t o r s h a s p a r t i a l deri-gatives with r e s p e c t to p to o r d e r K- I inclusive at
e v e r y point (p, e)). Simple manipulations show that
R
&&--i ~ ( p e)
, ~ ( p e). aV-la
- ~ - - _ _ .
+N-I (p+@).V (p+(0ja apV-l (19)
where H ( p , e) is a polynomial in the functions
F, d.p, ..., __- ,

dF
@-
F
$-i
a,
do
dP
- 7
d-2(0
..., -
8a-2 .
The e x p r e s s i o n is c l e a r l y continuously differentiable in p a t e v e r y
(B+ (0)
point (p, e) (O,<lp((p* for any 6 ) . To prove the l e m m a , we thus have to

APPENDIX
establish existence and continuity of the derivative with r e s p e c t to p of the

expression
a t the point (0, 0). We will f i r s t prove the existence of this derivative for
p=O. For p=O, F ( p , e)-O, and expression ( 2 0 ) thus vanishes. Therefore,
to definite l i m i t s f o r p + 0, which a r e respectively equal to the values of

these derivatives a t the point (0, 0). F u r t h e r m o r e , l i m o @ , 0 ) = 0 , and
P-4
lim-=a. Hence i t follows that the limit on the right in ( 2 1 ) e x i s t s and
p-0 P
is equal to
i.e.,
W e have thus established the existence of the p a r t i a l derivative of ( 2 0 ) with

r e s p e c t to p f o r p=O.
We will now prove the continuity of this derivative a t the point (0, 0 ) . To
this end, l e t u s find
The last t e r m can be written in the f o r m
476
APPEXDIX
1
F o r p + O , w e have = -P+ a a , -+L
@Ab,)* pa '
and the e x p r e s s i o n in
b r a c k e t s goes to z e r o by proposition ( b ) of L e m m a 2 . M o r e o v e r ,
lim F ( p , e) = 0, and
p-.')
By ( 2 4 ) w e then have
Relations ( 2 2 ) and (25) show that the p a r t i a l d e r i v a t i v e with r e s p e c t to p of

the e x p r e s s i o n (20) and t h e r e f o r e of the function ---
e x$
aV-l
i s t s and i s con-
tinuous a t the point (0,). This c o m p l e t e s the proof of the l e m m a .
5. A r e m a r k concerning the definition of a

s t r u c t u r a l l y s t a b l e dynamic s y s t e m
In o u r definition of a s t r u c t u r a l l y s t a b l e dynamic s y s t e m i n If'

(Definition 10, J 6 . l ) , w e had to introduce a l a r g e r region H enclosing the
relevant region W . W e w i l l now explain why this w a s unavoidable. Consider
a dynamic s y s t e m ( A ) defined in s o m e region G , which h a s a n unstable
triple limit cycle Lo in this region (Definition 28, 526.2). Let W be the
closed region bounded by the cycle L o , and suppose that N7contains a
s t r u c t u r a l l y s t a b l e focus 0 and that all the paths of s y s t e m (A) in W, except
the focus 0 and the cycle L o , a r e s p i r a l s which go to 0 for t -+ -k 00 and to
Lo f o r t -+ - 00 (an example of s u c h a s y s t e m is provided by s y s t e m (B,) with
k = 3 in Example 10, 827.2, see F i g u r e 121). T h e r e e x i s t a r b i t r a r i l y s m a l l
modifications of s y s t e m (A) which c a u s e the cycle Loto decompose into t h r e e
c y c l e s ( T h e o r e m 4 2 , 27.1), and s y s t e m (A) t h e r e f o r e should be r e g a r d e d
as s t r u c t u r a l l y unztable in W'. On the o t h e r hand, f o r any E > 0, t h e r e e x i s t s
6 > 0 such that i f ( A ) is 6 -close to (A) to rank 3, w e have
where FF is s o m e region. This propositionfollows f r o m the fact that any (A)

sufficiently c l o s e to (A) h a s one, two, or t h r e e closed paths n e a r Lo
( T h e o r e m s 42 and 43, 827). The i n n e r m o s t of these paths is a l i m i t cycle
which is unctable f r o m :he inside, and the region bounded by this cycle may
be used as W. Relation (1) shows that if we w e r e to c o n s i d e r W itself, r a t h e r
than the l a r g e r region h', in Definition 10, 56.1, s y s t e m (A) with a triple
limit cycle would come out as a s t r u c t u r a l l y unstable s y s t e m .
477
BIBLIOGRAPHY
1. A n d r o n o v , A.A., E . A . L e o n t o v i c h , I . I . G o r d o n , and
A. G . M a i e r . Kachestvennaya teoriya dinamicheskikh s i s t e m
vtorogo poryadka (Qualitative Theory of Dynamic S y s t e m s of
Second O r d e r ) . - Moskva, Nauka. 1966.
2 . B a u t i n , N . N. 0 prodol'nykh dvizheniyakh samoleta, blizkikh k
fugoidnym dvizheniyam (Longitudinal A i r c r a f t Motions Close to
Fugoid Motions). - Uchenye Zapiski Gor'kovskogo Universiteta,
13. 1947.
3 . B e l l y u s t i n a , L . N . K dinamike simmetrichnogo poleta s a m o l e t a
(The Dynamics of Symmetric A i r c r a f t Motion). - Izvestiya
AN SSSR, tech. sci., No. 11. 1956.
4. A n d r o n o v , A . A . and L . S . P o n t r y a g i n . Grubye s i s t e m y
(Structurally Stable Systems). - Doklady AN SSSR, 14, T
5. B a g g i s , G. F. Dynamical S y s t e m s with Stable S t r u c t u r e s . - In:
Contributions t o t h e T h e o r y of Nonlinear Oscillations, Vol. II,
pp. 3 7 -59. P r i n c e t o n Univ. Press. 1952.
6. A n d r o n o v , A.A., A.A. V i t t , and S . E . K h a i k i n . Teoriya kolebanii
(The Theory of Oscillations). - Fizmatgiz. 19 59.
7 . P e i x o t o , M . M. S t r u c t u r a l Stability on 2 -Dimensional Manifolds. -
ploskikh algebraicheskikh krivykh ( T h e C o n c e i t s of S t r u c t u r a l

Stability and the D e g r e e s of S t r u c t u r a l Instability for Plane
Algebraic C u r v e s ) . - Matematicheskii Sbornik, 67( log), No. 4.
1965.
9. A n d r o n o v , A.A. and E . A . L e o n t o v i c h . K t e o r i i izmeneniya
kachestvennoi s t r u k t u r y razbieniya ploskosti na t r a e k t o r i i (A
Contribution to the Theory of Variation of the Qualitative S t r u c t u r e
of the Partition of a Plane by P a t h s ) . - Doklady AN SSSR, 21,
No.9. 1938.
10. A n d r o n o v , A.A. and E.A. L e o n t o v i c h . Dinamicheskie s i s t e m y
1-i stepeni negrubosti n a ploskosti (Dynamic Systems of the F i r s t
Degree of S t r u c t u r a l Instability on a Plane). - Matematicheskii
Sbornik, 68('110), 3. 1965.
11. F i k h t e n g o 1 t s, G . M . K u r s differentsial'nogo i integral'nogo
ischisleniya (A C o u r s e i n Differential and Integral Calculus). -
Moskva, Nauka, Vol. 1, 1966; Vol. 2, 1959; Vol. 3. 1963.
1 2 . W a 1k e r , R. Algebraic Curves. - New York, Dover. 1962.
478
14. P e i x o t o , M . hI. On Structural Stability.- -4nn. of Math., 69:155-
222. 1959.
15. P o i n c a r e , H . Thbses prdsente'es a la Facult6 d e s s c i e n c e s d e
Paris: Sur les propriCt6.s d e s fonctions definies p a r les
Cquations aux differences p a r t i e l l e s . P a r i s , Gauthier - V i l l a r s .
1875.
16. G u b a r I, N . A. Kharakteristika slozhnykh osobykh tochek s i s t e m y
dvukh differentsial'nykh uravnenii p r i pomoshchi grubykh osobykh
tochek blizkikh s i s t e m (Classification of the hiultiple Singular
Points of a SysLem of Two Differential Equations Using Structurally
Stable Singular Points of Close Systems). - Matematicheskii
Sbornik, 40\82), 1. 1556.
17. S t e p a n o Y , V . V . K u r s differentsial'nykh uravnenii (A Course in
Differential Equations). - hioskva, Fizmatgiz. 1955.
18. B a u t i n , N . N . Povedenie dinamicheskikh s i s t e m vblizi g r a n i t s
oblasti ustoichivosti (The Behavior of Dyanmic Systems N e a r the
Boundaries of the Stability Region). - hIoskva, Gostekhizdat.
1545.
19. K r a s n o s e l ' s k i i , 5 I . A . e t a l . Vekfornye polya na ploskosti
(Vector F i e l d s on a Plane).- Moskva, Fizmatgiz. 1963.
2 0 . B a u t i n , K .N . G t odnom differentsial'noni uravnenii, imeyushchem
oredel'nyi t s i k l (Concerning a Differential Equation with a Limit
Cycle). - Zhurrial rekhnicheskoi Fiziki, IX, N o . 7:601- 611. 1535.
2 1 . L e o n t o v i c h , E. A. C rozhdenii predel'nykh tsiklov ot s e p a r a t r i s y
( f i e Creation cf Limit Cycles f r o m a Separatrix). - Doklady
A h SSSR, X S V I I I , No.4. 1551.
2 2 . 31 a r k u s h e v i c h , A. I. Teoriya analiticheskikh funktsii (Theory of
Analytical Functions ). - bIoskva-Leningrad, Gostekhizdat. 1950.
23. L a V a l l d e P o u s s i n , C h a r l e s J., de. Cours d'analyse
infinitksimale. 8th edition revised and enlarged with the collaDora -
tion of F. Simonart. Louvain, L i b r a i r i e Universitaire, Paris,
Gau thier -Villars. 1938.
2 4 . P o i n c a r k , H . L c s mCthodes nouvelles d e la mdcanique cCleste,
t. 111. - P a r i s . 1899.
25, A n d r o n o v , A . A . and A.G. L y u b i n . Primenenie teorii Puankare o
"tochkakh bifurkatsii" i "smene ustoichivosti" k p r o s t e i s h i m
avtokolebatel'nym s i s t e m a m (Application of the PoincarC Theory
of "Bifurcation Points" and "Stability Changes" to the Simplest
Xutooscillatory Systems). - Zhurnal Eksperimental'noi i
I'eoreticheskoi Fiziki, 5, N o s . 3-4. 1935.
2 6 . A r o n o v i c h , G.V. and L . X . B e l l y u s t i n a . Ob ustoichivosti
kolebami gorizcnta v uravnitel'noi bashne (Concerning the Stability
of Horizon Oscillations in a Leveling T u r r e t ) . - Inzhenernyi
Sbornik AN SSSR, XIIT. 1552.
27. B a u t i n , N . E. K t e o r i i sinkhronizatsii (Synchronization Theory). -
Zhurnal Tekhnicheskoi Fiziki, 9 . 1939.
28. B e 1 l y u s t i n a , L . .?I. Ob odnom uravnenii iz t e o r i i elektricheskikh
mashin (Concerning a C e r t a i n Equation f r o m the Theory of
E l e c t r i c hIotors). - A. A. Andrcnov hIemoria1 Collection, AN SSSR,
pp. 173- 186. 1955.
379
BIBLIOGRAPHY
29. Su s h k e v i c h , A. K. Osnovy vysshei algebry (Elements of Higher

Algebra). - Mciskva-Leningrad, Gostekhizdat. 1941.
30. N e m y t s k i i , V . V . and V . V . S t e p a n o v . - Kachestvennaya teoriya
differentsial'nykh uravnenii (Qualitative Theory of Differential
Equations). - Moskva-Leningrad, Gostekhizdat. 1949.
31. P o n t r j a g i n , L. S. Uber Autoschwingungssysteme, die den
Hamiltonschen nahe liegen. - Phys. Zeitschrift d e r Sowjetunion,
Band 6 , Heft 1 - 2 . 1934.
32. M a r k u s , L.- The Behavior of the Solutions of a Differential System
N e a r a Periodic Solution.- Ann. Math., 73, No. 2 . 1960.
33. M a r k u s , L . Structurally Stable Differential Systems.- Ann. Math.,
73, No.1. 1961.
-
34. T s i n g Y u a n h s u n .Ob algebraicheskikh predel'nykh tsiklakh
vtorogo poryadka dlya differentsial'nogo uravneniya
I (Second-Order Algebraic Limit Cycles f o r the Differential Equation
-Acta Math. Sinica, 8, No. 1. 23. 1958.

35. P i c a r d , E. Trait4 d'analyse. t . 2, Paris, 1905.
36. S m a 1 e , S. On Dynamical Systems. -International Symposium on
Ordinary Differential Equations. - L a Univers. Nac. di Mexico.
1961.
37. B o l ' s h a k o v , V.M., E . S . Z e l ' d i n , R.M. M i n t s , and
.
N.A. F u f a e v K dinamike s i s t e m y ostsillyator - r o t a t o r ( T h e
Dynamics of a n Oscillator-Rotator System). - Izvestiya Vysshikh
Uchebnykh Zavedenii, Radiophysics, VIII, N o . 2 . 1965.
38. A r a n s o n , S. Kh. Sistemy pervoi stepeni negrubosti na t o r e (Systems
of the F i r s t Degree of Structural Instability on a Torus).-
Doklady AN SSSR, 164, No. 5. 1965.
480
t
SUBJECT INDEX
Attraction e l e m e n t s ( s i r k s ) 1 4 1 Conservative dynamic s y s t e m 401,
408
Bifurcation point 205 Continuation of boundary a r c 140
value of the p a r a m e t e r 208 of c o r n e r a r c 1 4 0
Bifurcations of a dynamic s y s t e m Covering of a s p h e r e
202,205 closed 52-54
in the neighbor.iood of a open 52
multiple limit cycle 282 Creation of a closed path f r o m a
in t h e neighborhood of a multiple limit cycle 127
multiple focus 259 of a l i m i t cycle f r o m condensation
in t h e neighborhood of a of paths 217
multiple focus of f r o m infinity 217
multiplicity 1 261 f r o m a loop of a saddle-node
of s y s t e m s of f i r s t d e g r e e of s e p a r a t r i x 332, 324, 326
s t r u c t u r a l instability 375 f r o m a loop of a saddle-point
Branching point of an analytical s e p a r a t r i x 309, 311, 319
function 389 f r o m a multiple focus 254
limit cycle 277
Canonical f o r m of dynamic s y s t e m Curvilinear coordinates in the
( n e a r an equilibrium s t a t e ) 68- 70 neighborhood of a closed path 110
Canonical neighborhood Cycle without contact 145
limit cycle 1 4 2 Cyclic s y s t e m s of solutions 389
node o r focus 1 4 2
saddle point 143 Decomposition of a multiple equili-
Canonical neighborhoods, r e g u l a r brium s t a t e into s t r u c t u r a l l y s t a b l e
s y s t e m 144 equilibrium s t a t e s 218
C e l l s of dynamic s y s t e m s 141,175 Degree of s t r u c t u r a l instability of
of one type 177 dynamic s y s t e m s 206,207
of s t r u c t u r a l l y s t a b l e s y s t e m s Density of integral invariant 402
doubly connected 1 7 7
simply connected 177,179 Elementary a - and a - a r c s 145
C h a r a c t e r i s t i c index of a closed Equilibrium s t a t e , multiple 65
path 1 1 8 decomposition into
E - C l o s e n e s s of dynamic s y s t e m s structurally stable
23.24 equilibrium s t a t e s 218
of functions t o rank r 2 of multiplicity r 65
of regions 27 simple 66
of r e g u l a r transformation t o e-Extension of a region 151
identity t r a n s f o r m a t i o n 27.28
to rank r 2 4 Focal value, f i r s t 9 2
Conjugate e l e m e n t a r y a r c s 146 calculation 244
Focal values 243, 244 Rotation of vector field of a dynamic
s y s t e m 29
Hamiltonian dynamic s y s t e m 405
Saddle-to-saddle s e p a r a t r i x 97
e-Identity of partitions into paths Semistable limit cycle 107
in a plane region 31 Separatrix loop of a saddle-node,
on a s p h e r e 58 creation of a l i m i t cycle 322,
Integral invariant 402 324,326
density 402 of a saddle point, creation
of a limit cycle 309,311,319
Limit cycle (also s e e Creation of a stable 301,304
limit cycle) saddle-to-saddle 97
multiple 103 Simple intersection point of two
of multiplicity k 108 c u r v e s 15
s e m i s t a b l e 107 root of a function 8
simple 103,109 Singular elements 1 4 1
stable 1 0 7 paths. s e m i p a t h , a r c s of paths
unstable 1 0 7 140, 141
Sinks 141
Lyapunov value 244 Sources 141
Stable limit cycle 107
Method of small p a r a m e t e r s e p a r a t r i x loop 301,304
(Poincare' method) 409 Structural instability, d e g r e e of
Metric in the s p a c e of dynamic 206,207
s y s t e m s in a plane region 51-52 Structurally stable dynamic s y s t e m in
on a s p h e r e 52 - 54 a plane region 55
Multiple equilibrium s t a t e , decom- on a s p h e r e 58
position into s t r u c t u r a l l y stable intersection point of two
s t a t e s 218 c u r v e s 15
Multiplicity of a common point of path 62
two c u r v e s 15 root of a function 8
of a limit cycle 108,272 Structural stability of dynamic s y s t e m s
of a multiple focus 243 relative t o a given s p a c e 59
of an equilibrium s t a t e 6 5 Structurally unstable path 6 2
of a root of a function 7, 1 0 closed path 133
relative t o a given c l a s s of s e p a r a t r i x loop 301,304
functions 13, 14 Succession function on an a r c without
contact n e a r a closed path 104, 378
on a n o r m a l t o a closed path
116
on a" r a y f r o m a focus 91 - 92,
Poincare' method 401 240
Symmetry of phase p o r t r a i t 190
Region of stability in the l a r g e of a Systems c l o s e t o conservative 408,409
s i n k 189 t o Hamiltonian 408,418
with n o r m a l boundary 139 of f i r s t d e g r e e of s t r u c t u r a l
Regular partition of a region 146 instability 206,331
e-Translation of a s e t 31
Repulsion elements ( s o u r c e s ) 141
Unstable limit cycle 107
482

Andronov Bifurcations

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Andronov Bifurcations

Caricato da

Copyright:

Formati disponibili

NASA

1. I. GORDON, and A.G. MAIER

Published for t h e National Aeronautics and Space Administratign , t '

Izdatel stvo Nauka

Translated from Russian

Israel ?rogram for Scientific Translations

Printed in Jerusalem by Katcz Pre8a

The present monograph is a d i r e c t continuation of o u r Q u a1 i t a t i v e

Preface ........................................ iii

Chapter I. LIULTIPLICITY O F ROOTS O F FUNCTIOKS AKD

$1. Xlultiplicity of a root of a function . . . . . . . . . . . . . . . . . . . 1

s2. T h e multiplicity of a common point of t w o c u r v e s ......... 14

Chapter 11. DYNXhIIC SYSTEAIS CLOSE TO A GIVEN SYSTEILI

54. Intersection of paths of c l o s e s y s t e m s with a r c s and c y c l e s

Chapter 111. THE S P A C E O F DYKAMIC SYSTEMS AKD STRUC-

55. T h e s p a c e of dynamic s y s t e m s . ..................... 51

7. Structurally stable and s t r u c t u r a l l y unstable paths.

Chapter IV. EQUILIBRIUM STATES O F STRUCTURALLY STABLE

5 8. S t r u c t u r a l stability of a node and a s i m p l e focus. . . . . . . . .

Si 9. S t r u c t u r a l stability of a saddle point .. .. . . . . . . .... . . 78

10. S t r u c t u r a l instability of an equilibrium s t a t e with pure

s 11. A saddle-to- saddle s e p a r a t r i x . . . . . . . . . . . .. . . . . . . . . 97

Chapter V. CLOSED PATHS I N STRUCTURALLY STABLE

Introduction ..................................... 103

$12. A closed path and its neighborhood. Succession function . .. 104

S 13. Curvilinear coordinates in the neighborhood of a closed path.

$14. Proof of s t r u c t u r a l stability of a s i m p l e l i m i t cycle . ... . . 118

C h a p t e r VI. NECESSARY AND SUFFICIENT CONDITIQXS O F

16. Singular paths ant3 s e m i p a t h s of dynamic s y s t e m s . . . . . . . 137

s 17. -4 r e g u l a r s y s t e m of neighborhoods and t h e partition of F into

18. The fundamental t h e o r e m of s t r u c t u r a l stability of a dynamic

C h a p t e r VII. C E L L S OF STRUCTURALLY STABLE SYSTEMS.

19. C e l l s of s t r u c t u r a l l y s t a b l e dynamic s y s t e m s . . . . . . . . . . 175

20. Examples of s t r u c t u r a l l y stable s y s t e m s . . . . . .. . . . . . . 190

21. A definition of s t r u c t u r a l stability foregoing t h e r e q u i r e m e n t

C h a p t e r VIII. BIFURCATIONS O F DYNAMIC SYSTEMS. DECObT-

23. Decomposition of a multiple equilibrium s t a t e into s t r u c -

Chapter M. CREATION OF LIMIT CYCLES FROM A MULTIPLE

124. F o c a l v a l u e s . . .............................. 239

25. Creation of l i m i t cycles f r o m a multiple focus ......... 2 54

Chapter X.. CREATION OF CLOSED PATHS FROM A MULTIPLE

Introduction .................................... 266

26. Expressions for the derivatives of the succession function.

S 27. Creation of l i m i t cycles f r o m a multiple limit cycle ..... 277

Chapter XI. CREATION OF LIMIT CYCLES FROM THE LOOP OF

Introduction .................................... 286

28. Auxiliary m a t e r i a l . ........................... 287

loop (311). 4. The case P i (x.. u d Q& (a,u.) -

Chapter XII. CREATION OF A LIMIT CYCLE FRON THE LOOP

Introduction .................................... 322

s 30. Creation of a limit cycle f r o m the loop of a saddle-node

s 31. Dynamic s y s t e m s of the f i r s t d e g r e e of s t r u c t u r a l instability

Chapter XIII. LIMIT CYCLES OF SOME DYNAMIC SYSTEMS

0 32. The behavior of limit c y c l e s of some dynamic s y s t e m s

s 33. Creation of a limit cycle f r o m a closed path of a conservative

34. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424

Bibliography. .. .... .. . . . .. . .. . . .... . . . ....... .. .. 478

The s u b j e c t - m a t t e r of the qualitative theory of dynamic s y s t e m s is

Let a s y s t e m of t h i s kind be defined in s o m e region G. What happens

In this chapter He consider s o m e fairly elementary concepts relating to

The conditions f o r a common point with r = 2 are m o r e complicated

i . e . , only the functions a s such a r e 8-close, but not t h e i r derivatives, the

can be made a r b i t r a r i l y c l o s e to z e r o to rank ni, but never to rank g r e a t e r

io.ime 1 = 1 , 2 , . . ., r , all 3, are non-negatire zohole numbers, and a i f a2-