Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
1. INTRODUCTION
Consider N point massesm,, m,, .... mN > 0 with positions ql, q2, .... qN,
qi E R2 in the Euclidean plane. These points form a central configuration,
if, for some positive constant 1, the following system of algebraic equations
is satisfied
u,q m,m,
i<j 1/4i-4jll
The central configurations are important in n-body problems because
they are bifurcation points for the topological classification of the co-planar
n-body problem (see Easton [6] and Smale [12]); they are the limiting
configurations for colliding particles and for completely parabolic orbits
(see D. Saari [9, 111); they are the only configurations which can be main-
tained all the time in the n-body system; and often they are the starting
points for finding some new classes of periodic solutions. Questions about
these conligurations were raised by Wintner [lo], Smale [12], Palmore
[S], and many others. Recently there has been an active interest in finding
and classifying the central configurations, e.g., the recent work of Meyer
and Schmidt [ 11, Hall [2], Moeckel [3], and others.
For n = 3, this problem is completely solved and it is a classical result:
There are exactly 3 collinear solutions and 2 equilateral triangular solu-
tions for any 3 positive masses. In case of n = 4, even for 4 equal masses,
the exact number of central configurations is unknown. On the other
hand, Arenstorf [S] obtained the number of central configurations with
one of four masses being very small. He started with one zero mass and
then analytically continue it into positive masses. A very interesting
phenomenon arises in this case, i.e., there are two sets of masses such that
the numbers of central configurations corresponding to these two sets are
different. In this paper, we are going to find the exact number of central
configuration for some open sets of N masses for any N. It turns out that
for any N 2 4, the same phenomenon arises; there are always two open sets
of masses such that the numbers of central configurations corresponding to
these masses are different. From this we conclude that for any Nb4, the
set of masses for which the degenerate central configuration exists has a
positive (N- 1 )-dimensional measure (Haussdorff). This was proved first
by Palmore [S], but it seems difficult to complete all the steps in his proof.
The method we are going to use is analytical continuation. We start with
two zero masses and find the corresponding central configurations, especially
the ones in which two zero masses are at the same point. Under certain
conditions, these central conligurations can be analytically continued into
a full n-body problem with all masses positive. We will see that the central
conliguration with two zero masses at the same point will bifurcate into
several distinctive central configurations of a full n-body problem. By
repeating this processes, for any choice of II, we can find the exact number
of central configurations for at least some open subset of N positive masses.
The central configurations are also called relative equilibria. However,
throughout this paper we only use the name central configuration and save
the relative equilibria for those special central configurations with one or
more zero masses.
Let q* = (q:, qf, .... 9.:) be a solution of Eq. (1), i.e., qr, q:, .... 4.: form
a central configuration; then it is easy to see that kq* = (kq:, kqr, .... kq,;)
and Aq* = (Aq:, Aq:, .... Aq,i$) is also a solution of (l), where k is a con-
stant and A E SO( 2, R j is a rotation matrix. We can define an equivalence
170 ZHIHONG XIA
mim.w+2
- 4;) (2)
+ IlqN+Z-qrl13 (qN+Z
for i = 1, 2, .... N, and
N
mNf Iq.%‘+ I = c n1im.w + I
!=, Ilq,+,-qil13(q~-q4.y+1)
mlv, ,nlJv,2
--qN+l) (3)
+ IIqNfI -9.v+zI13 (qY+z
(4)
CENTRAL CONFIGURATIONS WITH SMALL MASSES 171
Note that in Eqs. (3) and (4) there are factors nz,,,+, and mN+?,
respectively. When dealing with small massesit is convenient to factor out
nlN+ It nl,v+2. Assume this is done. Our next step is to consider the extreme
case where ~~~+,=rn,+~=O. For m,V+,=m,V+2=0, (2) reduces to (l),
and (3 ) and (4) become
.\ m,
q.v+1= c
t-l lIq,~+,-q:l13~q‘y+L-qr*)
m,
q.v+2= i * 3 (q,v+z-q?).
!=I IlqY+2-qi II
Note that the equations for q ,,,+ , , qNfz are identical; we may write them
in a different way as
v,\r(.Y)
=;=,i Ilq,*-*ul[
mi +P12.
l
Let I* be a solution of DV,,(x) =O; then it is called a relative equi-
librium of the restricted (n + l)-body problem, and it is nondegenerate if
the Jacobian of DVJx) has a rank 2; i.e., if ID’V,(s)l # 0 at x = .Y*. Since
D’V,(x) is symmetric, it has two real valued eigenvalues, say a, and a2,
and by selecting appropriate coordinates, we may write
and
Then,
for qN+,-&+I, qN+r-q~~+z very small, Eqs. (2), (3), and (4)
become
i= 1, 2, .... N (8)
(9)
CENTRAL CONFIGURATIONS WITH SMALL MASSES 173
We still have a singularity for (10) at Mu+, =O, rr~~+~=O, and r,=O; a
scaling of variables will remove this singularity. Let rz= (mv+, +mN+2),‘3r;;
(10) becomes
1
aI - W~l13 0
ri = 0, (12)
[ 0 a, - l/llrSII’
a,--a2 0
=3(a,-a,)a~#o;
0 30,
rk=
_ +(0 u-“~)
--r2 , r2 = +(a;’ 3, 0),
al--a2 0 I I 3a, 0
0 3a, ’ 0 a,--a, .
Both of them do not vanish provided that a, #a?, and hence if a, #a,
these four solutions can be continued into four different classes of central
configurations for the full (N+ 2)-body problem with two small masses.
174 ZHIHONG XIA
The following graph illustrates the relative positions of two small masses
in a (4 +2)-body problem.
* “22
* 1124
mI *
*
*+*
*
*-*
7 7
minimum saddle
m, *
Now we have a way for finding the central configurations of the (N+ 2)-
body problem with two small masses even when these two small masses
have the same limit position as their masses tend to zero. We point out
that the central configurations thus found exhaust all the possible ones
with two sufficient small masses and with the large masses m,, m,, .... mN
close to q1*, q2*, .... 4:.
In case (iii), when a, = a, < 0, more complicated bifurcation occurs, and
it depends on high order terms of V,v(x) at s*, it is possible that we have
more than four solutions that can be continued into a full (N+ 2)-body
problem instead of four solutions in the case of a, # a,, a, < 0, and a, < 0.
For example, let us put N equal masses at the vertices of a regular
N-polygon; then the origin is a relative equilibrium and is a local minimum
of V,,v(x), by symmetry, a, = a, < 0. One can show that at least 2N classes
of central configurations of the (N + 2)-problem, with two small masses at
the origin, can be obtained, again by symmetry.
As an application of above results, let us consider a 4-body problem with
two small masses. There are live relative equilibria for a restricted 3-body
problem; two of them are equilateral triangle solutions which are minima
of Vz(.u), and two eigenvalues of D’V(x) are different at these two points,
and the other three of them are collinear solutions which are saddle points
of VJx). All of the live relative equilibria are nondegenerate. Thus,
Proposition 1 gives 5 x 4= 20 classes of central conligurations for the
4-body problem with two small masses which are away from each other
and there are 2 x 4 + 3 x 2 = 14 classes of central conligurations with two
small masses close to each other; therefore, there are totally 34 classes of
central configurations for the 4-body problem with two small masses. This
same result was obtained by Arenstorf [8] by a different method.
CENTRAL CONFIGURATIONS WITH SMALL MASSES 175
with
D’V,v(.x)=
[0 a2
a’
1 ’
1+O(?v+
1).
176 ZHIHONG XIA
So two of the relative equilibria are saddle points, while the other two are
minimal of V,,,+ ,(x).
We conclude that, for any saddle point of V,,(x), then for each central
configuration continued from this relative equilibrium, with m,, , suf-
ficiently small, there are always two saddle points of V,+,(x) nearby
mN+ , , and correspondingly, for any minimum of V,%(X), there will always
be two minima and two saddle points nearby mN+ 1 for VI”+ ,(x).
Now, we come back to the 4-body problem we discussed at the end
of the last section. First consider the case with m, and m, large and
m3 sufficiently small. From the above result, there are together
2 x 4 + 3 x 2 + 5 x 4 = 34 relative equilibria. Among these 34 relative equi-
libria of the restricted (3 + I)-body problem, 16 belong to the equilateral
triangle formed by m,, m2, and m, with 2 x 3 = 6 local minima and
2 x 5 = 10 local saddle points. The other 18 belong the collinear central
configurations of m, , m,, and m3 with 3 x 2 local minima and 3 x 4 local
saddle points. By a very careful counting, we conclude that there are
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3 x 4 x 2 + 3 x 4 x 5 = 294 relative equilibria for the restricted (4 + 1)-body
problem.
We now turn our attention to the central configurations with three large
masses. First for any three positive masses, there are five classes of central
configurations; three of them are collinear and two of them are equilateral.
Arenstorf [S] and Palmore [4] showed that for any collinear central con-
figuration there are six relative equilibria for the restricted 4-body problem;
two of them are local minima and the other four of them are saddles. For
the equilateral central configuration, an interesting phenomenon happens.
Arenstorf [S] showed that there are 8 or 10 relative equilibria of the
restricted 4-body problem depending on the three positive masses, and for
most of the masses they are all nondegenerate. A simple topological
argument shows that in the nondegenerate case, among these 8 (or 10)
relative equilibria, 3 (or 4) of them must be local minima and the other 5
(or 6) must be saddle points. Since thee eigenvalues continuously depend
on the masses, it can be shown easily that for most of the masses, at the
minima relative equilibria, D2 V3(x) has distinct eigenvalues.
Therefore for the restricted 4-body problem, for some open set of three
large masses, there are totally 34 relative equilibria. Among them, 22 are
saddle points and the other 12 are local minima, and for some other open
set of three large masses, the total number of relative equilibria is 36.
Among them 24 are saddle points and 14 are local minima. From these
results about the restricted (3 + 1 =4)-body problem and the technique
we developed above, it is easy to see (again, by carefully counting the
numbers), for a 5-body problem, for some open set of masses (with two
large masses, one small mass, and two even smaller masses), there are 294
CENTRAL CONFIGURATIONS WITH SMALL MASSES 177
classes of central configurations, while for some other open set of masses
(with three large masses and two small masses), there are 374 classes of
central configurations, and all of them are nondegenerate.
Repeating the above procedure, for any N there are two open sets of
massesfor which we can find the exact number of central configurations
corresponding to these two sets of masses.For n from 4 through 10, the
numbers thus found are listed in Table I.
It would be nice if we could develop a general formula depending on
the number of bodies for the number of central configurations thus found.
It turns out that it is even hard to write a recursive formula for these
numbers. In fact, the numbers in Table I are produced by a small Fortran
program.
It is interesting to see how fast the number of the central configurations
increasesas n increases.Let t, be the total number of central configurations
we find for some n positive masses. Since local minima produce more
central configurations when continued into positive masses, t,, must be
greater than the number obtained by considering all local minima as saddle
points. In this way, we have a simple estimate:
TABLE I
n 4 5 6 1 8 9 10
THEOREM. Let .Z, be the set of masses (m,, m2, .... mN) E G2.yfor which
the degenerate central configuration exists; then .Z,v has a positive (N - l)-
dimensional Hausdorff measure.
This is a theorem of Palmore [S], but there is a step in his proof that
seems difficult to complete. Here we give a different proof.
Proof: We have shown that for any n 24, there exist m E 52’: and
m*E[Wy, such that m$Z‘, and rn* $Z,, and the numbers of central
configurations corresponding to m and rn* are different. Let r be any arc
in R’y joining m and m *, then l-r\ Z,\, # 4; otherwise, by analytical con-
tinuation of the central configuration along the arc, we will get the same
number for both m and m* but that is impossible. This proves the theorem.
5. SOME REMARKS
ACKNOWLEDGMENTS
Part of the writing of this paper was done at Northwestern University and the University
of Cincinnati. The author thanks D. Saari, K. Meyer, D. Schmidt, and G. Hall for reading the
original maniscript and making many useful suggestions. The author also thanks R. Moeckel
for pointing out an error in the original manuscript.
REFERENCES
Primed in Belgium