The Special Theory of Relativity. Anadijiban Das

Universitext
Editorial Board
(North America):
S. Axler
F.W. Gehring
P.R. Halmos
Springer Science+Business Media, LLC

Universitext
Editors (North America): S. Axler, F.W. Gehring, and P.R. Halmos
AksoylKhamsi: Nonstandard Methods in Fixed Point Theory

Aupetit: A Primer on Spectral Theory
BoossIBleecker: Topology and Analysis
Borkar: Probability Theory; An Advanced Course
CarlesonlGamelin: Complex Dynamics
Cecil: Lie Sphere Geometry: With Applications to Submanifolds
Chae: Lebesgue Integration (2nd ed.)
Charlap: Bieberbach Groups and Flat Manifolds
Chern: Complex Manifolds Without Potential Theory
Cohn: A Classical Invitation to Algebraic Numbers and Class Fields
Curtis: Abstract Linear Algebra
Curtis: Matrix Groups
DiBenedetto: Degenerate Parabolic Equations
Dimca: Singularities and Topology of Hypersurfaces
Edwards: A Formal Background to Mathematics I alb
Edwards: A Formal Background to Mathematics II alb
Foulds: Graph Theory Applications
Gardiner: A First Course in Group Theory
Gardingffambour: Algebra for Computer Science
Goldblatt: Orthogonality and Spacetime Geometry
Hahn: Quadratic Algebras, Clifford Algebras, and Arithmetic Witt Groups
Holmgren: A First Course in Discrete Dynamical Systems
HoweITan: Non-Abelian Harmonic Analysis: Applications of SL(2, R)
Howes: Modem Analysis and Topology
HumiIMiIler: Second Course in Ordinary Differential Equations
HurwitzIKritikos: Lectures on Number Theory
Jennings: Modem Geometry with Applications
JonesIMorrisIPearson: Abstract Algebra and Famous Impossibilities
KannanIKrueger: Advanced Real Analysis
KellylMatthews: The Non-Euclidean Hyperbolic Plane
Kostrikin: Introduction to Algebra
LueckingIRubeI: Complex Analysis: A Functional Analysis Approach
MacLaneIMoerdijk: Sheaves in Geometry and Logic
Marcus: Number Fields
McCarthy: Introduction to Arithmetical Functions
Meyer: Essential Mathematics for Applied Fields
MinesIRichmanIRuitenburg: A Course in Constructive Algebra
Moise: Introductory Problems Course in Analysis and Topology
Morris: Introduction to Game Theory
PorterlWoods: Extensions and Absolutes of Hausdorff Spaces
RamsaylRichtmyer: Introduction to Hyperbolic Geometry
Reisel: Elementary Theory of Metric Spaces
Rickart: Natural Function Algebras
Rotman: Galois Theory
RubeVColliander: Entire and Meromorphic Functions
(continued after index)

Anadijiban Oas
The Special Theory

of Relativity
A Mathematical Exposition
With 27 Illustrations
, Springer
Anadijiban Das
Department of Mathematics and Statistics
Simon Fraser University
Burnaby, V5A IS6 British Columbia
Canada
Editorial Board
(North America):
S. Axler F.W . Gehring
Department of Mathematics Department of Mathematics
Michigan State University University of Michigan
East Lansing, MI 48824 Ann Arbor, MI 48109
USA
P.R. Halmos
Department of Mathematics
Santa Clara University
Santa Clara, CA 95053
USA
Library of Congress Cataloging-in-Publication Data
Das. Anadijiban.
The specialtheory of relativity: a mathematical exposition I
Anadijiban Das, author.
p. cm.
Includes bibliographical references and index.
ISBN 978-0-387-94042-7 ISBN 978-1-4612-0893-8 (eBook)
DOI 10.1007/978-1-4612-0893-8
1. Special relativity (Physics)--Mathematics. 2. Mathematical
physics. L Title.
QCI73.65.D38 1993
530.I'I--dc20 93 - 10256
Printed on acid-free paper.
1993 by Springer Science+Business Media New York

Originally published by Springer-Verlag Berlin Heidelberg New York in 1993
AII rights reserved. This work may not be translated or copied in whole or in part without the
written permission ofthe publisher Springer Science+Business Media, LLC, except for brief
excerpts in connection with reviews or scholarly analysis. Use in connection with any form
of information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter
developed is forbidden.
The use of general descriptive names, trade names, trademarks, etc., in this publication, even if
the former are not especially identified, is not to be taken as a sign that such names, as understood
by the Trade Marks and Merchandise Marks Act, may accordingly be used freely by
anyone.
Production managed by Jim Harbison; manufacturing supervised by Vincent Scelta.
Typeset by Asco Trade Typesetting Ltd, Hong Kong.
9 8 7 6 5 4 3 2(Corrected second printing, 1996)
ISBN 978-0-387-94042-7
SPIN 10528157
Dedicated to Sri Gadadhar Chattopadhyaya
Preface
The material in this book is presented in a logical sequenc:e rather than a

historical sequence. Thus, we feel obligated to sketch briefly the history of
the special theory of relativity. The brilliant experiments of Michelson and
Morley in 1887 demonstrated the astonishing fact that the speed of light is
independent of the state of relative linear motion of the source of light and
the observer of the light. This fact necessitates the modification of the usual
Galilean transformation (between two relatively moving observers), which
tacitly assumes that time and space are absolute.
Fitzgerald in 1889 and Lorentz in 1892 altered the Galilean transforma-
tion by introducing a length contraction in the direction of relative motion.
This contraction explained the Michelson-Morley experiment, but it was
viewed by both Fitzgerald and Lorentz as a mathematical trick only and not
indicative of the nature of reality. In 1898 Larmor introduced a similar time
dilation in an attempt to find the transformations which leave Maxwell's
equations invariant. Lorentz also introduced the time dilation independently
sometime before 1904. Poincare in 1905 also discovered the Lorentz transfor-
mation and asserted that it was the fundamental invariance group of nature.
Einstein in 1905 discovered the Lorentz transformation from physical con-
siderations. Einstein, alone among these mathematical physicists, recognized
the philosophical implications of the Lorentz transformation in that it re-
jected the commonly held notion that space and time were both absolute. He
postulated the equivalence of all inertial frames of reference (moving with
constant velocities relative to each other) with regard to the formulation of
natural laws. Furthermore, he recognized and postulated that the speed of
light is the maximum speed of propagation of any physical action. Therefore,
the speed of light must be invariant for all inertial observers. Thus the
Michelson-Moreley experiment was reconciled with theory. Minkowski, a
mathematician, combined both physical postulates of Einstein into one
mathematical axiom. This axiom is that "all natural laws must be expressible
as tensor field equations on a (flat) absolute space-time manifold." Thus, in
that there is no preferred inertial frame for the formulation of natural laws, a
universal democracy is postulated to exist among all inertiaI observers. This
vii
viii Preface
axiom is called the Principle of Special Relativity. Many experiments in-

volving atoms and subatomic particles have verified the essential validity of
this principle.
In the first chapter we introduce axiomatically the four-dimensional
Minkowski vector space. This vector space is endowed with a nondegenerate
inner product which is not positive definite. Therefore, the concepts of the
norm (or length) of a four-vector and of the angle between two four-vectors
have to be abandoned. A Lorentz mapping is introduced as an inner product
preserving linear mapping of Minkowski vector space into itself.
In Chapter 2 we introduce the flat Minkowski space-time manifold with a
proper axiomatic structure. It is proved that the transformation from one
Minkowski chart to another must be given by a Poincare transformation (or
an inhomogeneous Lorentz transformation). The conceptual difference be-
tween a Lorentz transformation of coordinate charts and a Lorentz mapping
of the tangent (Minkowski) vector space is clearly displayed. Minkowski
tensor fields on the flat space-time are also defined.
In the third chapter, by applications of a particular Lorentz transforma-
tion (the "boost"), length contraction, time retardation, and the composition
of velocities are explained. The group structure of the set of all Lorentz
transformations is demonstrated, and real representations of the Lorentz
group are presented. The proper orthochronous subgroup is defined and
discussed also.
The fourth chapter defines the spinor space (a two-dimensional com-
plex vector space) and the properties of spinors. Bispinor space (a four-
dimensional complex vector space) is also introduced. It is shown that a
unimodular mapping of spinor space can induce a proper, orthochronous
Lorentz mapping on Minkowski vector space. Furthermore, a unimodular
mapping of bispinor space is shown to induce a general Lorentz mapping of
Minkowski vector space.
In Chapter 5 prerelativistic mechanics is briefly reviewed. In the setting of
prerelativistic mechanics in space and time, [3 x IR:, the momentum conju-
gate to the time variable turns out to be the negative of energy! After this, the
relativistic mechanics is investigated. The Lagrangian is assumed to be a
positive homogeneous function of degree one in the velocity variables (which
makes the generalized Hamiltonian identically zero!). Examples from electro-
magnetic theory and the linearized gravitational theory of Einstein are
worked out.
In Chapter 6 the relativistic (classical) field theory is developed. Noether's
theorem (essential for the differential conservation laws) is rigorously proved.
As examples of special fields, the Klein-Gordon scalar field, the electromag-
netic tensor field, nonabelian gauge fields, and the Dirac bispinor field are
presented. However, at the present level of treatment, gauge fields are not
derived as connections in a fibre bundle over the base (Minkowski) manifold.
In each chapter, examples and exercises of various degrees of difficulty are
provided.
Preface ix
Chapter 7 deals with a research topic, namely, classical fields in the eight-
dimensional extended (or covariant) phase space. Historically, Born and
Yukawa advocated the extended phase space on the basis of the principle of
reciprocity (covariance under the canonical transformation p = -q, q = pl.
In recent years, Caianello and others have considered the principle of maxi-
mal proper acceleration arising out of the extended phase space geometry. We
ourselves have done some research on classical fields in the eight-dimensional
phase space. We can obtain, in a certain sense, a unified meson field and a
unification of fermionic fields. These fields, however, contain infinitely many
modes or particles.
We have changed the usual notation for the Lorentz metric Y/ij in favor of
dij (since Y/ijkl is used for the pseudotensor) and y == (1 - V 2 )-1/2 in favor of
r
fJ == (1 - v 2 l/2 (since y is used to denote a curve).
This book has grown out of lectures delivered at Jadavpur University
(Calcutta), University College of Dublin, Carnegie-Mellon University, and
mostly at Simon Fraser University (Canada). The material is intended mainly
for students at the fourth and the fifth year university level. We have taken
special care to steer a middle course between abstruse mathematics and theo-
retical physics, so that this book can be used for courses in special relativity
in both mathematics and physics departments. Furthermore, the material
presented here is a suitable prerequisite for further study in either general
relativity or relativistic particle theory.
In conclusion, I would like to acknowledge gratefully several people for
various reasons. I was fortunate to learn the subject of special relativity from
the late Professor S. N. Bose F.R.S. (of Bose-Einstein statistics) in Calcutta
University. I also had the privilege for three years of being a research asso-
ciate of the late Professor J. L. Synge F.R.S. at the Dublin Institute for
Advanced Studies. Their influence, direct or indirect, is evident in the presen-
tation of the material (although the errors in the book are solely due to me!).
In preparation of the manuscript, I have been helped very much by Dr. Ted
Biech, who typed the manuscript and suggested various improvements. Mrs.
J. Fabricius typed the difficult Chapter 7. Mrs. E. Carefoot drew the diagrams.
Dr. Shounak Das has suggested some literary improvements. I also owe
thanks to many of my students for stimulating discussions during lectures.
I thank Dr. S. Kloster for the careful proof reading.
Finally, I thank my wife Mrs. Purabi Das for constant encouragement.
Contents
Preface vii
Chapter 1. Four-Dimensional Vector Spaces and Linear Mappings 1

1.1. Minkowski Vector Space V4 1
1.2. Lorentz Mappings of V4 8
1.3. The Minkowski Tensors 13
Chapter 2. Flat Minkowski Space-Time Manifold M4 and

Tensor Fields 20
2.1. A Four-Dimensional Differentiable Manifold 20
2.2. Minkowski Space-Time M4 and the Separation
Function 25
2.3. Flat Submanifolds of Minkowski Space-Time M4 35
2.4. Minkowski Tensor Fields on M4 41
Chapter 3. The Lorentz Transformation 48

3.1. Applications of the Lorentz Transformation 48
3.2. The Lorentz Group ..'t'4 55
3.3. Real Representations of the Lorentz Group ..'t'4 59
3.4. The Lie Group ..'t'4++ 63
Chapter 4. Pauli Matrices, Spinors, Dirac Matrices,

and Dirac Bispinors 72
4.1. Pauli Matrices, Rotations, and Lorentz
Transformations 72
4.2. Spinors and Spinor-Tensors 79
4.3. Dirac Matrices and Dirac Bispinors 85
Chapter 5. The Special Relativistic Mechanics 89

5.1. The Prerelativistic Particle Mechanics 89
5.2. Prerelativistic Particle Mechanics in Space and
Time 1E3 x IR 95
xi
xii Contents
5.3. The Relativistic Equation of Motion of a Particle 100

5.4. The Relativistic Lagrangian and Hamiltonian
Mechanics of a Particle 108
Chapter 6. The Special Relativistic Classical Field Theory 120

6.1. Variational Formalism for Relativistic Classical
Fields 120
6.2. The Klein-Gordon Scalar Field 133
6.3. The Electromagnetic Tensor Field 140
6.4. Nonabelian Gauge Fields 147
6.5. The Dirac Bispinor Field 151
6.6. Interaction of the Dirac Field with Gauge Fields 160
Chapter 7. The Extended (or Covariant) Phase Space and

Classical Fields 168
7. 1. Classical Fields 168
7.2. The Generalized Klein-Gordon Equation 175
7.3. Spin-Ih Fields in the Extended Phase Space 190
Answers and Hints to Selected Exercises 202

Index of Symbols 204
Subject Index 207
1
Four-Dimensional Vector Spaces and
Linear Mappings
1.1. Minkowski Vector Space V4

The three-dimensional vectors in Newtonian physics are generalized into
four-dimensional vectors in the theory of relativity. This four-dimensional
vector space is called the Minkowski vector space and is denoted by V4 . This
vector space is over the real field III The mathematical axioms for addition
and scalar multiplication of Minkowski vectors are as follows:
AI. + bE V4 for all a, bE V4 .
a
A2. a + b = b + a for all a, b E V4 .
A3. (a + b) + c = a + (b + c) for all a, b, c E V4 .
A4. There is 0 E V4 such that a + 0 = a for all a E V4 .
AS. For all a E V4 there is -a E V4 so that (-a) + a = O.
(1.1.1)
M1. aa E V4 for all a E IR, for all a E V4 .
M2. a(fJa) = (afJ)a for all a, fJ E IR, for all a E V4 .
M3. la = a for all a E V4 .
M4. + b) = aa + ab for all a E IR, for all a, b E V4 .
a(a
MS. (a + fJ)a = aa + fJa for all a, fJ E IR, for all a E V4 .
We shall also assume the existence of an inner product for V4 satisfying the
following axioms:
11. abElRforalla,bEV4 .
12. a b = b a for all a, b E V4 .
( 1.1.2)
I3. ().a + Jib) c = A(a c) + Ji(b c) for all A, 1-( E IR,
for all a, b, c E V4 .
14. a x = 0 for all x E V4 if and only if a = O.
2 I. Four-Dimensional Vector Spaces
The axiom 14 is called the axiom of nondegeneracy. It is a weaker axiom than

15. a' a ;:0: 0 for all a E V4 , and a' a = 0 if and only if a = O. (1.1.3)
For a positive definite inner product axiom 15 replaces 14. In addition to these
axioms we impose the axiom of dimensionality on Minkowski vector space:
Dl. dim V4 = 4.
Let {e 1 ,e Z ,e 3 ,e 4 } be a basis set for V4 The metric tensor components rela-
tive to this basis are defined by
gij == e i ' ej for i, j E {1,2, 3,4}. (1.1.4)
From axiom 12 it follows that gji = gij for all i, j E {l, 2, 3, 4}. The four-
dimensional unit matrix is denoted by I == [biJ. The eigenvalues of the
matrix [gij] are the roots of the characteristic equation
(1.1.5)
Since the matrix gij is symmetric, the roots of (1.1.5) are all real. By the axiom
of nondegeneracy 14 it follows that all the eigenvalues of gij are nonzero.
The signs of the eigenvalues of gij are determined by the axiom of Lorentz
signature:
Sl. ).1 > 0, ).2 > 0, ).3> 0').4 < O.

The vector space obeying the sixteen axioms AI-A5, MI-M5, 11-14, DI,
and SI is called Minkowski vector space and is denoted by V4 .
In V4 , the two vectors a, b are defined to be Minkowski orthogonal (or
M -orthogonal) provided
ab = o. (1.1.6)
Theorem (1.1.1): There exists an M-orthonormal basis {e 1 ,e 2 ,e 3 ,e 4 } for V4
such that
(1.1.7)
I
where
I
010
o
D = [dij] == 0 0 1 oOJ .
000 ~1
The proof is rather involved and is omitted. The metric dij in (1.1.7) is called
the Lorentz metric. The signature of dij is defined to be the trace of [dij]. We
shall use a choice of [diJ so that the signature is equal to 2. Note that some
authors use the signature ~ 2.
Now we shall explain the Einstein summation convention. In a mathemati-
cal expression, wherever two repeated Roman indices are present, the sum
over the repeated index is implied. For example, we write

4 4
UkVk == I UkVk = I UIV I = UIV
"
k=l 1=1
4 4 4 4
gijuiV i == I L gijuiV i = L L gklUkv ' = gklUkV I.
i=l i=l k=l 1=1
The summation indices are called dummy indices, since they can be replaced
by other indices over the same range. In the summation convention, never use
dummy indices that repeat more than twice. This is necessary in order to
avoid wrong answers; for example,
4 4 4
UkVkUkV k == I UkVkUkVk # L L UkVkUIV I = UkVkUIVI = (U Vk)2. k
k=l k=l 1=1
Let {e 1,e 2,e3,e4} be an M-orthonormal basis (or tetrad) for V4. For any
vector U E V4 , there exists a linear combination
U = "iu e = u e,.
4
L..
1=1
i i (1.1.8)
The unique numbers or scalars u i are called the Minkowski components of the
vector u relative to the basis {e 1, e 2, e3, e4 }.
Theorem (1.1.2): In terms of the Minkowski components, the inner product

between vectors u, v is given by
(1.1.9)
By the axioms in (1.1.2) and (1.1.8) we have

u v = (uie.) (vie.)
lJ = d .. uV I} .

Note that from (1.1.9)

U U = dijuiu i = (U 1)2 + (U 2)2 + (U 3)2 _ (U 4 )2. (1.1.10)
The above expression is not positive definite. Thus the concept of the length
(or norm) of a vector in V4 is abandoned. Furthermore, if we define cos(u, v) ==
(u v)/ J(u u)(v v), then we are led to contradictions. For example, if we
choose Vo = e 1 + [en - 1)/nJe4 for n E 71+ and let u = el> then cos(e 1 , vn ) ==
n/J(2n - 1). Therefore, 1 :s; cos(u, vn ) and lim n _ oo cos(u, "n} --+ 00, which is
absurd. That is why the concept of an angle between two vectors u, v E V4
is abandoned as well. However, for a spatial vector subspace V3 == {v E V4 :
v4 = O}, the usual concept of the length and angle can be restored.
Since the expression (1.1.10) for u u is indefinite, we can define three kinds
of vectors in V4 :
4 I. Four-Dimensional Vector Spaces
(i) a vector U E V4 that satisfies U U > 0 is called a spacelike vector;

(ii) a vector U E V4 for which U U < 0 is called a timelike vector;
(iii) a vector U E V4 for which U U = 0 is called a null vector.
Example: Let {e l ,e 2 ,e 3 ,e 4 } be an M-orthonormal basis for V4 By (1.1.9),

e 1 e 1 = d ll = 1. Thus e 1 is a spacelike vector. Similarly e 2 , e 3 are spacelike
vectors. But e 4 e 4 = d44 = - 1, so e 4 is a timelike vector. Set U = e 1 + e4 and
observe that u . U = 0, so we see that U is a null vector. 0
The separation numher is a generalization of the concept of length and is

denoted by <J(u). It is defined by
<J(u) == ~ ~ o. (1.1.11)
Thus for either timelike or spacelike vectors U E V4 we have <J(u) > O. But for
a null vector n we have <J(n) = O. For an example choose U = (e l - 2e 4 )/2.
Then <J(u) = Ji=374j = .)3/2.
A vector e in V4 is called a unit vector if <J(e) = 1. Subsequently we shall use
only M-orthonormal bases for V4 unless mentioned otherwise. A spatial vec-
tor subspace relative to an M-orthonormal basis is defined as
(1.1.12)
Small Greek indices will take values in the set {l, 2, 3}, and small Roman
indices will take values in the set {1, 2, 3, 4}. The appropriate summation
convention will apply to each type of index.
Theorem (1.1.3) (Schwarz Inequality): For any two vectors u, v in V3 the follow-
ing inequality holds:
lu'v'l :::;; Ju'u'vfJv fJ . (1.1.13)
Equality holds if and only if u<X = AV' for some A E IR.
Proof: Suppose that u' == o. Then (1.1.13) holds trivially. Now suppose
u<Xu' > o. Then for any )0 E IR we have
(AU + V)(AU + v') = A2 UU + v'v' + 2AUV ~ O.
Setting the value A = -(u'v)/(uYu Y ), we obtain
[ -(U'V)2 + (uu)(vPvP)]/(uYu Y) ~ O.
From above (1.1.13) follows. The case of equality mentioned in the theorem
is left as an exercise.
The M-orthogonality between two vectors in V4 is not always intuitively

natural. We shall derive a few theorems on that topic now.
Theorem (1.1.4): No two time like vectors in V4 can be M-orthogonal.

Proof: Let D, v be two timelike vectors. Thus we have

D D = dijuiu j = u~u~ - (u 4f < 0,
V'V = dijviv i = VaVa - (V4)2 < 0.
Combining these two inequalities we have
Ju~uavPvP < IU4 V4 1.
By the Schwarz inequality (1.1.13) we obtain

lu~v~1 < IU4 V 4 1. (1.1.14)
Suppose, contrary to the conclusion of the theorem, that D' v = 0. Then

luaval = IU4 V 4 1. This last equality contradicts (1.1.14).
Corollary (1.1.1): For two timelike vectors D, v such that u4 > 0, v4 > 0, we
have D'V < O.
Theorem (U.S) (Synge): Let i, t be two timelike, future-pointing unit vectors

(past-pointing may replace future-pointing). Then - 00 0,
(J(t) = 1, iPiP- (i4 f = -1, 4> O.
We want to solve the above equations and inequalities. For that purpose
consider two three-dimensional unit spatial vectors:
a~aa = 1, dada = 1.
In spherical polar coordinates we can write
a l = sin Ocos tP, dl = sinOcos~,
a 2 = sin 0 sin tP, d 2 = sin 0sin~,
a3 = cosO, d3 = cosO,
where 0 :s; 0 :s; n, 0 :s; 0 :s; n, -n :s; tP < n, and - n :s; ~ < n. Thus we have
cos I{I = aad a = cos 0 cos 0 + sin 0 sin 0cos(t;6 - ~), where I{I is the angle be-
tween the two vectors aP , d P and 0 :s; I{I :s; n. We can always express t and i in
the form
t~ = (sinh x)aa, fa = (sinh i)d a,
t4 = cosh X > 0, f4 = coshi > 0,
6 1. Four-Dimensional Vector Spaces
where X, XE 1Ft Therefore,

-tt = coshxcoshi.[cos 2 (IjI/2) + sin 2 (IjII2)]
- sinh Xsinh X[cos 2 (IjI/2) - sin 2 (IjII2)]
= cosh(X - X)cos 2 (IjII2) + cosh(X + x)sin 2 (IjII2) > O.
Making another transformation x = sin(IjII2) so that x E [0,1], we get y ==
-t i = [cosh(X + X) - cosh(X - X)]x 2 + cosh(X - X) == f(x). Thus we can
graph the function f(x) over x E [0,1]. For the case cosh(X + X) -
cosh(X - X) > 0, this graph is a portion of a parabola with f'(x) ;;:::: 0 and
cosh(X - X) s y s cosh(X + X), For the case cosh(X + X) - cosh(X - X) < 0,
the graph is a portion of a parabola with f'(x) s 0 and cosh(X + X) s y s
cosh(X - X). In the case cosh(X + X) - cosh(X - X) = 0, the graph is a portion
of a straight line with f'(x) == 0 and y = cosh(X + X) = cosh(X - X). In all
three cases min {cosh(X + X),cosh(X - X)} s y s max {cosh(X + X), cosh (X - X)}.
Since the function cosh(x) has the lower bound 1 and no upper bound, we see
that 1 s y < 00, so - 00 < t i s - 1.
Corollary (1.1.2): Let t, t be two time like, future-pointing unit vectors such that
t i = - 1. Then t = 1.
Proof: From the proof of Theorem (1.1.5), it is clear that, for the case
cosh(X + X) > cosh(X - X), the minimum value y = 1 is at x = 0 and the min-
x,
imum value is cosh(X - X) = 1. Thus IjI = 0 and X = so = t/J = ~. Thus e e,
t a = ta, t 4 = 4; hence, t = 1. In the second case cosh(X + X) < cosh(X - X),
and the minimum value is cosh(X + X) = 1 at x = 1. Therefore, IjI = 1l and
e
X = - X and so = 1l - e, ~ = t/J + 1l. Thus t" = t", t 4 = 4, so t = 1. In the
case cosh(X + X) = cosh(X - X), the minimum value is 1, which is attained
by al! x E [0,1]. Then COs?(x + X) = cosh(x - X) = 1; hence, X = X = So o.
t" = t" = 0, t 4 = t 4 , so t = t.
Theorem (1.1.6): A timelike vector cannot be M-orthogonal to a nonzero null

vector.
Proof: Suppose that t is a timelike vector and n is a nonzero null vector.

In terms of their Minkowski components tat a < (t 4 f and nan a = (n 4 f with
It41> 0, In41 > O. Combining these expressions and using the Schwarz in-
equality (1.1.13) we have
(1.1.15)
Suppose that t n = 0, or (t ana )2 = (t 4n4 )2. This contradicts (1.1.15).
Now we shall prove a very counterintuitive theorem.
Theorem (1.1.7): Two nonzero null vectors are M-orthogonal if and only if they
are scalar multiples of each other.
Proof: (i) Assume that two null vectors m, n are such that m = An for some
A E lit Then m n = A(n n) = 0.
(ii) Suppose that two nonzero null vectors m, n are M-orthogonal. Then
mm = m"m" - (m4)2 = 0,
00 = n"n" - (n4)2 = 0, ( 1.1.16)
mn = m"n" - m4n4 = 0,
From the above expressions we obtain
(m"n,,)2 = (m 4n4)2 = m"m"nPn fJ , Im"n"l = jm"m"nfJn P. (1.1.17)
The above equation is the case of equality in the Schwartz inequality (1.1.13).
Therefore, m" = An" for some scalar A"# 0. Since n4"# 0, we have m4 =
m"n"/n4 = An"n"/n4 = An4. Thus m = An.
It is hard to plot the Minkowski vectors, since the concepts of length of a

vector and angle between two vectors do not exist. However, the parallelo-
gram law of vector addition still holds. It is worthwhile to draw Minkowski
vectors to gain some geometrical insight. We have to plot these vectors on a

piece of paper, which is part ofa Euclidean plane. Let us plot M-orthonormal
vectors e l , e 4 such that e l . e 4 = and e l . e l = -e4 e 4 = l.1t is quite natural
to plot these two vectors as i and j of the usual two-dimensional Cartesian
basis vectors; see Figure 1. As we have drawn, the Euclidean lengths Ilelll =
IIe 4 11 = 1 and e l , e 4 are Euclidean orthogonal. However, the vectors e l + e4 ,
-e l + e4 have Euclidean lengths Ile l + e 4 11 = II-e l + e 4 11 = and e l +.J2,
e4, -e l + e4 are Euclidean orthogonal. But a(e l + e4) = 0, a( -e l + e4) =
0, and (e l + e4 ) (-e l + e4 ) = - 2 "# 0. So we have to use caution in order to
interpret any plot of Minkowski vectors.
FIGURE 1. Minkowski vectors e l ,

e4, el + e4, -el + e4
EXERCISES 1.1
1. Let {e l ,e 2 ,e 3,e 4} be a Minkowski basis for V4. Let another basis be
{a, b, c, d} where
b = e2 - e3 + .j2e4,
d = e4
(i) Determine which of these are spacelike, timelike, or null vectors.

(ii) Obtain the separation numbers O'(a), O'(b), O'(e), O'(d).
(iii) Determine whether or not {a, b, e, d} is a basis for V4.
2. Determine which of the following subsets of V4 is a vector subspace.
(i) The union of the subset of all spacelike vectors and {OJ.
(ii) The union of the subset of all timelike vectors and {OJ.
(iii) The subset of all null vectors.
3. Prove that the nondegeneracy axiom of the inner product in (1.1.2)
implies that det[gij] #- O.
4. Prove that for any two vectors x, Y in V3 that
Ixayal = Jxaxayflyfl
holds if and only if x = AY for some AE IR.
5. Let s, t be two timelike future-pointing vectors in V4. Prove that
O'(s)O'(t) :s; Is' tl. (This is called the Reversed Schwarz Inequality for timelike
future-pointing vectors.) Does it hold for other cases, i.e., past-pointing or
mixed orientation?
1.2. Lorentz Mappings of V4

A linear mapping L: V4 -+ V4 is defined to be such that
L(Aa + Jib) = AL(a) + JiL(b) (1.2.1)
for all A, Ji E IR and all a, b E V4.
Suppose that {e 1 , e2' e3' e4} is a basis for V4 that is not necessarily M-
orthonormal. Since L(ed E V4, we must have
Cj == L(ej ) = lije" (1.2.2)
for some suitable scalars lij' i, j E {I, 2, 3, 4}, and the summation convention
applies. The 4 x 4 matrix L == [Iij]' where i stands for the row index and j for
the column index, is called the matrix representation of the mapping L rela-
tive to the bases ofV4, {e 1 ,e 2,e 3,e4}, and {cl,eZ,c3,e4}'
Theorem (1.2.1): Let e, == L(eJ The set of vectors {C 1 ,C 2,C 3,C4} is also a basis
for V4 if and only if det[l ij ] #- O.
The proof is left to the reader.

A linear mapping L with det[lij ] #- 0 is called invertible.
Example: Let us consider a linear mapping L such that

e 1 = L(ed == sin(n/4)e 1 + cos(n/4)e 2 = 1'1 ei ,
e2 = L(e z) == - cos(n/4)e 1 + sin(n/4)e2 = Ii 2ei,
e3 = L(e 3 ) == e 3 = li 3ei ,
e4 = L(e 4 ) == e4 = li4ei
Therefore, the 4 x 4 matrix representation is
-cos(nj4)
sin(nj4) o
o 0]
0
o 1 0 '
o o 1
with det[[ij ] = 1. The linear mapping L is invertible. The basis {e l , ez, e3 , e4 }
is M-orthonormal whenever {e l , e z, e 3 , e4 } is M-orthonormal. 0
Now we shall define the Kronecker delta:
b == {I
i. for i = j, (1.2.3)
} 0 for i i= j.
These numbers are the entries of the 4 x 4 identity matrix J relative to the
standard basis {e l ,e Z ,e 3 ,e 4 }. Similarly bOP are the entries of the 3 x 3 iden-
tity matrix with respect to the standard basis {i,j, k}. We shall work out some
examples involving the Kronecker delta.
Example: Consider the sum

blju j = bl l u l + b l 2 U 2 + b l 3 U 3 + J14U4
= lui + Ou 2 + Ou 3 + Ou 4 = u l .
Similarly
(1.2.4)
(1.2.5)
For an invertible linear mapping L, we denote the inverse mapping by A,

so that the corresponding matrices satisfy
A = L-l, AL = LA = J, aijl\ = [ija jk == Jik (1.2.6)
Example:
r ,i. (./4) cos (nj4) 0
~]
[I',] ~ -CO~('/4) sin(nj4) 0
0
0 0
n
r,i.(./4) -cos(nj4) 0
[a',J ~ CO'r4)
sin(nj4) 0
D
0 1
0 0
Now we shall discuss the transformation properties of vector components

under a change of basis sets.
Theorem (1.2.2): Suppose that {e 1 ,e2,e3,e4}, {e 1 ,e2,e3,e4} are two bases for
V4 such that ej = [ijei . Then the corresponding components of a Minkowski
vector u undergo the transformation
(1.2.7)
Proof: It is easy to see u = ukek = ujej = uj(l\ek) = Wjui)ek , so (Uk - Zkjuj)e k =

O. The linear independence of the basis vectors ek implies Uk = Zkju j, so
aiku k = aik[kjuj = /jijU j = u i.
Suppose that a linear mapping L: V4 --+ V4 satisfies L(a)' L(b) = a' b for
every pair a, b. Such a mapping is called a Lorentz mapping. A Lorentz
mapping preserves inner products of pairs of vectors in V4. The correspond-
ing matrix representation L == [Zij]' relative to an M-orthonormal basis, is
called a Lorentz matrix.
Theorem (1.2.3): A Lorentz matrix L must satisfy the matrix equation

LTDL = D, (1.2.8)
where L T denotes the transpose for Land D is the Lorentz metric defined in
(1.1. 7).
Proof: From the definition of a Lorentz mapping it follows that

L(e i) . L(ej ) = e i . ej = dij , (1.2.9)
where {e 1 ,e h e 3,e4 } is an M-orthonormal basis. Since L(e;) = Z\e k , Equa-
tion (1.2.9) yields Wiek)' (lmjem) = dij , so
Zk;lmj(e k' em) = zkizmjdkm = l\dkmZ mj = dij. (1.2.10)
Thus LTDL = D.
Corollary (1.2.1): The determinant of a Lorentz matrix L must satisfy

detL = det[lij] = 1. (1.2.11)
Proof: Taking the determinant of both sides of (1.2.8) we have

detCdetDdetL = detD. (1.2.12)
Recalling det L T = det Land det D = -1 we have from (1.2.12) that
(detL)2 = 1. (1.2.13)
Thus equation (1.2.11) follows.
A Lorentz mapping L with det L = 1 is called a proper Lorentz mapping. A

Lorentz mapping with det L = -1 is called an improper Lorentz mapping. A
Lorentz mapping L such that /44 > 0 when represented in an M-orthonormal
basis is called an orthochronous Lorentz mapping. Now we shall give some
elementary examples of Lorentz mappings. In all cases we use an M-
orthonormal basis for V4 .
Examples: (i) The identity mapping given by I(v) = v for all v E V4 . The cor-
responding matrix [c5 iJ shows that I is a proper, orthochronous Lorentz
mapping.
(ii) The space reflection P is given by the mapping P(e a ) = -ea and
P(e 4 ) = e4 . The matrix of P is
o
-1
(1.2.14)
o
o
P is an improper orthochronous Lorentz mapping.
(iii) The time reversal T is given by T(e a ) = ea and T(e4 Ji = -e 4 . The cor-
responding matrix is
0]
l 0 0
T
.
f0
= [t 'j ] = 0 0 1
1 0 o
o . (1.2.15)
000 -1
T is an improper Lorentz mapping that is not orthochronous.

(iv) A plane rotation is given by the mapping L such that
L(e 1 ) = cos8e 1 + sin8e 2 = e1 ,
L(e 2 ) = -sin8e 1 + cos8e 2 = e2 , (1.2.16)
L(e 3 ) = e 3 = e3 , where -:rr < 8 < n.

The matrix of L is
-sin8 0 0]
cosll 0 0
o 1 0 .
o 0 1
The Lorentz mapping L is proper and orthochronous. The plane rotation is

shown in Figure 2.
(v) The boost mapping is a very important Lorentz mapping and is defined
FIGURE 2. Plane rotation by the

angle 8.
FIGURE 3. Boost mapping associated

with a moving observer.
by
L(e 1 ) = f3(e 1 + ve 4 ) = C1 ,
L(e 2 ) = e2 = C2 ,
(1.2.17)
L(e 3 ) = e 3 = C3 ,
L(e4) = f3(ve 1 + e4) = C4'
The matrix of the boost L is
13
. 0 o1 00 Vf31
0
L = [I'j] = [ 0 o 1 0 .
vf3 o 0 13
The boost L is proper and orthochronous. This mapping is physically asso-
ciated with a moving observer (see Chapter III). The transformation (1.2.17)
is shown in Figure 3.
In Fig. 3, tan r/J == v, Ir/JI < n/4. The null vector D == (e 1 + c4)/..j2 is symmet-
rically situated among the basis vectors e 1, e4, and ~1' ~4. As the parameter v
tends to 1, both the vectors @1' @4 tend to collapse into D. The range of v is
restricted to Ivl < 1 in (1.2.17) to avoid this situation. 0
EXERCISES 1.2
1. Show explicitly that the mapping L in equation (1.2.17) has a matrix that
satisfies LTDL = D, detL = 1, and 144 > o.
2. Let L be a Lorentz mapping. Prove that L -1 and L T are also Lorentz
l
mappings by examination of their matrices.
3. Consider the mapping L whose matrix is
:]
-1 -1
. -1 0 -1
L = [/'j] = -10
-1 0 1 .
-1 -1 -1 2
(i) Show that L is an improper Lorentz mapping.
(ii) Show that the Lorentz mapping takes the lattice vectors nie i into the
lattice vectors fjj@j where both ni, lli E 71..
4. Prove that, for the entry 144 in the matrix L of a Lorentz mapping L, the
inequality 1 :s; 1/441 holds.
1.3. The Minkowski Tensors

In case a Lorentz mapping L is given bY@j = L(ej) = [kjek' the components of
a Minkowski vector u (a contravariant vector) change via the transformation
rules
(1.3.1 )
(1.3.2)
A Minkowski covector w (a covariant vector) is defined to be a linear map-
ping from V4 into IR. The components of ware defined by Wi == w(e i ).
Theorem (1.3.1): The components Wi of a co vector W, under a Lorentz mapping

@j = L(e), transform as
Wj = [kjWk (1.3.3)
Proof: Wj = w(@j) = wWjed = [kjw(e k ) = IkjWk.
Corollary (1.3.1): Let u i, Wi be components of a vector u and a covector w

respectively. The sum UiWi remains invariant under a Lorentz mapping.
The number UiW i , which is invariant under Lorentz mappings, is called a

Lorentz (or Minkowski) scalar. We shall often simply call them scalars.
Consider a mapping T: V4 x V4 ---+ R Let T be a bilinear mapping so that
we have
T(Aa + ,ub, c) = ).T(a, c) + ,uT(b, c),
T(a, Ab + ,uc) = ).T(a, b) + ,uT(a, c),
.u and all vectors a, b, c. We call T a second-order co-
for all scalars A,
variant Minkowski tensor and define its components relative to the basis
{e 1 ,e 2 ,e 3 ,e4 } as Tij == T(ei,eJ. The components Tij undergo the following
transformation under a change of basis:
iij = T(ei, ej ) = T(lmiem , lkjed = [m;lkj T(e m , ed
(1.3.4a)
The components Til of a second-order contravariant Minkowski tensor are
assumed to undergo the following transformation:
(1.3.4b)
Furthermore, the components Tij of a second-order mixed Minkowski tensor
are assumed to transform as
(1.3.4c)
An antisymmetric second-order covariant tensor IS defined as a tensor (X
whose components satisfy

(1.3.5)
A symmetric second-order covariant tensor is defined as a tensor (J whose
components satisfy
(1.3.6)
The number of linearly independent components of an anti symmetric
Minkowski tensor (Xij is six. Since (Xi) = - (Xji' we have the following relation-
ships on the components of (Xi/
(X22 =- (X22 = 0,
(X44 = - (X44 = 0, (1.3.7)
(Xij = -(Xji for i "* j.
Thus, there are six independent nondiagonal components of (Xi)' Similarly, for
a symmetric tensor (Ju = (Jji, there are four independent diagonal and six
independent nondiagonal components. Thus there are ten independent com-
ponents of a second-order symmetric Minkowski tensor.
Example: Suppose that Fij is an anti symmetric second-order Minkowski ten-

sor such that Fl2 = F23 = F3l = Fl4 = O. Under the Lorentz mapping given
in (1.2.17), the transformed components are
Fl4 = lmll\Fmk = 1lll44Fl4 + 1\ll4F4l = 0,
F24 = l m21\Fmk = 12 21\F24 + 1221l4F2l = f3F24 ,
~
F34 = I m 31 k 4Fmk = I 3 314 4F34 = f3F34
F23 = lm21k3Fmk = 12 213 3F23 = 0,
~
F3l = I m 31k lFmk = I 3 314 lF34 = vf3F34,
Fl2 = lmllk2Fmk = 1\122 F42 = -Vf3F24 ,
where f3 is defined as in (1.2.17). This example is relevant in the electromag-
netic theory. D
We can generalize the transformation properties (1.3.4a), (1.3.4b), (1.3.4c)

to higher-order tensors. The 4'+' components of an (r + s)-order (or rank)
(r indices contravariant, s indices covariant) Minkowski tensor is assumed
to transform as:
(1.3.8)
Note that, for multi-index tensors (r + s ~ 3), we shall write the contravari-
ant indices above the covariant indices rather than offsetting them as we do
for second-order tensors.
Example: Suppose that the components of a (r + s)-order Minkowski tensor

1:J: :::J: have the following values:
'''j: = 0 otherWlse.
1:!:::! = -1, 1:ili,"'i .
We want to compute fl:::l under the Lorentz mapping (1.2.17), so
= ( _1)'+1 (f3v)'+. 0
In case some Minkowski tensors are given, there are ways to produce new
tensors out of them. The following three theorems deal with the construction
of these new tensors.
Theorem (1.3.2): Given two (r + s)-order Minkowski tensors with components

O'J::::;:, 1:J::::;: the linear combination
AO'J::::;: + J,l1:j::::J: yields the components of another (r + s)-order Minkowski
tensor.
Proof: Scalars transform as A= A, {i = /t. Thus by (1.3.8) we have
Comparing the above with (1.3.8), the theorem is proved.
Theorem (1.3.3): Let the components of two Minkowski tensors be given by

a].i, ......]i and r!':.:~ . The tensor (or outer) product of the two tensors is given by
r
1 s 1 q .
the components aJ.' ....."j;r!', ... .~q and these form the components of an (r + p) +
(s + q)-order M inkowski tensor.
Proof: Using (1.3.8) we see that
so after a regrouping of terms the result follows.
Suppose that the components of a (r + s)-order Minkowski tensor are

given by rj,l . . j:. A single contraction of the tensor in terms of its components
is defined by ri,.ik-,dk+lir.
II "'Jh-ICJh+l "'}s
Similarly other single contractions can be de-
fined. Multiple contractions may also be defined.
Theorem (1.3.4): A single contraction rj: ...... j~=:~~~:~:::f. of the components of an

(r + s)-order Minkowski tensor yields the components of a [(r - 1) + (s - 1)]-
order M inkowski tensor.
The proof is left as an easy exercise for the reader.
Example: The Kronecker delta b ij , defined in equation (1.2.3), are the com-
ponents of a mixed second-order Minkowski tensor (the identity tensor). The
single contraction is given by the scalar bCc = b 1 1 + b 2 2 + b\ + b\ = 4. A
double contraction of the tensor product of the identity tensor with itself is
given by the scalar b\b\ = bCc = 4. The other double contraction is bCcb\ =
16. D
Recall the metric tensor components dij in (1.1.7). The contravariant metric
tensor components d ij = d ji are defined to be the entries of the inverse of the
metric such that
(1.3.9)
In an inner-product vector space, such as V4 , a contravariant tensor has
a corresponding covariant tensor and vice-versa. This correspondence is
brought about by the metric tensor in the following rules for lowering and
raising indices:
(1.3.10)
rk/ = dljr kj = dkjrj/; (1.3.11)
r/ = dljr jk = dkj'lj. 0
There exist some special numerical Minkowski tensors whose components

do not change values under Lorentz mappings. We shall list these tensors now.
(i) The (r + s)-order zero tensor has for its components nJ::: :): == o. These
values do not change under any Lorentz mapping.
(ii) The components of the identity tensor are b i j . The transformed compo-
nents are Jj = a~lrb! = a~lr = bj and thus remain unchanged.
(iii) The metric tensor components undergo the transformation dij =
[aJbjd ab = dij by equation (1.2.10). Similarly dij, and d ij == b ij n~main unchanged.
(iv) The tensor products of the above tensors produce other numerical
tensors of higher-order with similar properties. For example, the [(r + I) +
(s + I)]-order numerical tensor with components d a' ''2 ... dar-lardblb2 ...
dbs-1b s bk IIt. bk,h. can be mentioned.
These components do not change under any Lorentz mapping. There ex-
ists another important numerical symbol. It is the totally antisymmetric Levi-
Civita permutation symbol defined by
for (m n r s) an even permutation of (1 234),
for (m n r s) an odd permutation of (l 2 3 4), (1.3.12)
otherwise.
There are 256 components of this permutation symbol of which only twenty-
four have nonzero values. Some components are explicitly shown below
For the transformation properties of the permutation symbol Sm"", we first

notice that
Therefore, if we assign the transformation rule

emnrs = [det[liJr1la mlbnlerldsBabed
= ( l)lamlbnfrldsBabed' (1.3.l3a)
the 8mnrs are components of a numerical relative tensor. We shall assume
(1.3.l3a) and define the totally anti symmetric Levi-Civita pseudotensor and
its transformation properties by
(1.3.13b)
~mnrs = sgn[det[lij]]lamlbnlerlds"labed'
(Although in a Minkowski coordinate system there is no difference between
Babed and "lobed' these do differ in a curvilinear coordinate system of M4 in
Chapter 2.) In case of a proper Lorentz mapping ~mnrs = "labed' whereas in case
of an improper Lorentz mapping ~mn" = - "labed' Now, for the raising of the
indices of a pseudotensor, we follow the usual rules to get
"lobed = - "labed'
Example: Suppose that 0i = -Fij , which implies that Fji = _Fii. We define
the components of the Hodge dual pseudo tensor as
F*ij =(l/2)"lijkIFkl. (1.3.15)
We have by (1.3.12)
F*12 =(1/2)"l1234F34 + (1/2)"llZ43F43 = F 34 ,
(1.3.16)
F*Z4 = F 31 ,
Raising and lowering indices in (1.3.16) we get
F34 = _F*12, F14 = _F*23, F24 = _F*31,
FZ3 = _F*l4, F31 = _F*24, F12 = _F*34.
These equations can be neatly summarized into
Fij = -(l/2)"lijkIF*kl. (1.3.17)
This example is relevant to the electromagnetic field theory. D
EXERCISES 1.3
I. Consider the lowering of the index
Prove explicitly that the above components transform as those of a

[(r - I) + I]-order tensor.
2. (i) Consider the components 'ij of an arbitrary second-order covariant

tensor. Show that there exists a unique decomposition 'ij = Sij + a ij such that
sij = Sji, aij =- aji'
(ii) Consider a symmetric second-order contravariant tensor and an anti-
symmetric second-order covariant tensor with components sij and aij respec-
tively. Obtain the numerical scalar value of the double contraction Sij aij .
3. Consider the numerical tensor defined by the components d adbc == dabddc -
dacddb' Prove that:
(i) ddahc = - d adbc = - ddacb = d bcda ,
(ii) ddabc + ddbca + ddcab = 0,
(iii) dad be = ( - 1/2)1]admn1]bemn'
4. Consider an antisymmetric second-order covariant Minkowski tensor
with components aij' Let a mixed tensor be defined by the components (}ij ==
aikajk - (l/4)<5ijamnamn.
(i) Prove that (}ii == O.
(ii) Show that (}ij = (1/2)(a ik a jk + a*ika*jd, where a*ij == (1/2)1]ijkl akl .
References
1. P. R. Halmos, Finite dimensional vector spaces, Van Nostrand, Princeton, NJ, 1958.
[pp. 1, 12, 19J
2. W. Noll, Notes on Tensor Analysis prepared by C. C. Wang, The Johns Hopkins
University, Mathematics Department, 1963.
3. J. L. Synge, Relativity: The special theory, North-Holland, Amsterdam, 1964.
[pp. 6, 17J
2
Flat Minkowski Space-Time
Manifold M4 and Tensor Fields
2.1. A Four-Dimensional Differentiable Manifold

The space-time of events in special relativity is assumed to be a flat differ-
entiable manifold M 4 . Therefore, we shall go briefly through the definitions of
a four-dimensional manifold.
First, it is assumed that M4 is a topologized set, endowed with a Hausdorff
topology. That is, the question of whether a subset of M4 is open or not can
be answered. Furthermore, for two distinct points p, q E M4 there exist two
disjoint open subsets containing p and q respectively.
A chart (X, U) or local coordinate system is a pair consisting of an open set
U c;; M4 together with a continuous, one-to-one mapping X: U --+ D C;; [R4,
where D is an open subset of 1R4 in the usual topology. For apE U we have
x == (xl,x 2,X\X4) = X(p) E D. The coordinates (xi,x 2,x 3,x4) are the coordi-
nates of p in the chart (X, U) (see Figure 4).
Each of the coordinates Xi can be obtained by the projection mappings
ni: D --+ IR, defined by nk(x) = nk(xl, x 2 , x 3 , x 4) == Xk.
Consider two charts (X, U) and (X, 0), such that the point p is in the non-
empty intersection UnO.
From Figure 5 we obtain
x = (X 0 X- 1 )(x),
x = (X 0 ;e-l)(X).
By projection of these points we get
Xk = [n k 0 X0 X-1](x) == Xk(x) = Xk(Xl,X2,X3,X4),
(2.1.1)
Xk = [nk 0 X 0 ;e-l](X) == Xk(x) = Xk(Xl,X2,X3,X4),
where x E Ds and x E Ds'

The above equations define a general coordinate transformation between
two local coordinate systems. We shall assume that the eight functions xk, Xk
are continuously thrice differentiable or Xk E C(j3(Ds) and X k E C(j3(Ds). We can
also describe this fact by saying that the charts (X, U) and (X, 0) are C(j3_
20
.
)
IR
IR
FIGURE 4. A chart (X, U) and projection mappings.
"
X
Us
X
" X- 1
Xo
" -1
XO X
1R4
IR
t--+-IR
~k
FIGURE 5. Two charts in M4 and coordinate transformations.
related. From the assumption of differentiability, we can prove that if the

Jacobian of the transformation (2.1.1) satisfies
0(x , x2 "x3 x
1 4)
== det [OXi(X)]
--.- #0 0 (2.1.2)
o(xl, x 2 , x 3 , x 4 ) ox)
for all XED., then the mapping is one-to-one.

A subatlas of class '6'3 is a collection of charts (Xh, Uh), such that the union
of all Uh covers M4 and the charts are mutually '6'3-related. A maximal collec-
tion of '6'3-related charts for M4 is an atlas of class '6'3. Finally, a four-
dimensional differentiable manifold of class '6'3 is the set M4 together with an
atlas of class '6'3.
Example: Suppose that an inverse coordinate transformation in a domain of

22 2. Flat Minkowski Space-Time Manifold M4
M4 is given by
Xl = Xl sin x2 cos x3 ,
x2 = Xl sin x2 sin x3 ,
(2.1.3)
where 5s == {x: 0 < xt, 0 < x2 < IT, -IT < x3 < IT, X4 E IR}. The Jacobian of
this transformation is
a( X I ,x 2 ,x 3 ,x 4) Ai 2 "2
aCI A2 A3 A4)
X,X,x,X
= (x ) SIllX > O. D
Now we shall define a curve in M 4 . Let [tl' t 2] be a closed interval in IR. A

parametrized curve y is a mapping from [tl' t 2 ] into M 4 . Note that the curve
y is viewed as the mapping of [t I' t 2] to M4 and not the set of points in M 4,
which is in the range ofy. Let PI = y(t1), pz = y(t 2) as in Figure 6.
Let this curve be inside the coordinate neighbourhood U of M 4 . The coor-
dinates associated with this curve are
(2.1.4)
where t E [t[> t 2 ]. The functions ?Ii are assumed to be continuous and
piecewise twice differentiable. The image of this curve is the set r ==
[X a y] ([t I ,t2]) c 1R4. Furthermore, for a nondegenerate curve y,
ift [d:li(t)J
4
----at > 0, (2.1.5)
U M4
t
IR t1 t2
1R4
IR
FIGURE 6. A curve y in M 4 .
at each point of differentiability. The components of the tangent vector tp

(relative to the coordinate basis) to the curve y at p = y(t) are defined to be
df![i(t)/dt. The tangent vector tp of y at p E M4 can be visualized as a directed
line segment emanating from p tangential to y. There is an intrinsic way of
defining a tangent vector tp as the directional derivative
tp = tY(I) == d~:(t) O~i I p=y(r)

. (2.1.6)
Let us consider all possible differentiable curves passing through p. The set
of all possible tangent vectors, emanating from p, span a vector space. This
vector space is called the tangent vector space T"M 4 . It is clear that this vector
space is four-dimensional. For the differentiable manifold M 4 , it is assumed
that TpM4 is isomorphic to the Minkowski vector space V4 for each
pE V C M 4 .
Let the components of the tangent vector tp of a differentiable curve y be
df![i(t)/dt in a coordinate system. Let
Xk = gk(X) = gk(xl,x 2,x 3,X4),
Xk = Xk(X) = Xk(xl,x 2,x3,X 4)
indicate a general coordinate transformation. The curve y can be represented
in the hatted coordinate system by
(2.1.7)
The components of the tangent vector tp in the hatted coordinate system are
given by di'i(t)/dt. Using the chain rule, the transformation between the two
sets of components are
di"i(t) = [oii(~)JI df![j(t),
dt ax} x=1"(I) dt
(2.1.8)
df![j(t) = [oX~~x)JI di"i(t).
dt ax ~=:i(t) dt
The above equations imply that the transformation between the two sets
of components of a tangent vector field u(p) must be
_ Oii(x) j( )
uAi( XA) --!'l-.-u
ux}
x,
(2.1.9)
Example: Let a parametrized, differentiable curve y be given by

Xl = f![l(t) ;: t 2 ,
x 2 = f![2(t) == t,
x3 = ,q[3(t) == e"
X4 = ,q[4(t) == 1,
where t [0,2] and we have used a coordinate chart (X, U). For this curve

E
2:t=1 [dXi(t)jdtY = 4t 2+ 1 + e21 > satisfying (2.1.5). Let another chart
(i, 0) be defined by the transformation
Xl = XI(X) == Xl + x 2 ,
x2 = X2(X) == Xl _ x 2 ,
x3 = X3(X) == 2x 3 + X4,
X4 = X 4(x) == x 3 - x4j2,
o(xl, x 2 , x 3 , x 4 )
o( x,x,x,x
I 2 3 4) = 4 -=f. 0.
Therefore, the hatted components of the tangent vector, by equation (2.1.8),

are given by
d:i"l(t) = (l)d,q[l(t) + (1)~~2(t) = 2t + 1
dt dt dt '
d:i"2(t) = (1) d,q[l(t) + (-1) d,q[2(t) = 2t _ 1,

dt dt dt
d:i"3(t) = (2) d,q[3(t) (1) d,q[4(t) = 2 I
~ ~ + ~ e,
d:i"4(t) = (1) d,q[3(t) + (-lj2) d,q[4(t) = e l 0

dt dt dt
EXERCISES 2.1
1. Consider a coordinate transformation given by
Xl = J(X I )2 + (X 2)2;
Arctan(X 2j XI ), Xl> 0,
x2 = arc(x l ,x 2) == { (nj2)sgn(x 2 ), Xl
= and x 2 -=f. 0,
Arctan(x 2 jx l ) + n sgn(x 2), Xl < and x 2 -=f. 0;
x3 = x 3 ;
X4 = X4,
where - nj2 < Arctan(x 2 jx l ) < nj2 and the chart (X, U) = (X, M4)
(i) Obtain 15 == ;(0) S 1R4.
(ii) This transformation is ~r-related. Obtain the maximal r-value.
(iii) Obtain the domain Do S 1R4 such that the Jacobian
O(XI
,x "x
2 3 x4 )
o(xl,x 2 ,x 3 ,X4 ) > .
2.2. Minkowski Space-Time M4 and the Separation Function 25
2. A family of differentiable curves in M4 (using the coordinate chart

satisfy the differential equations
(X,M 4 ))
[d~:(t)J + [~i(t)]2 = 1
for all t E R Obtain this family of curves explicitly.
3. Consider a coordinate transformation
X4 = In 1~41,
where Ds == {~: 0 < ~1, 0 < ~2 < n, -n < ~3 < n, ~4 "# O}. The components
of a tangent vector field u(p) are given as
al(~) = (1/~1)COSX2cos2~3,
a2 (x) = (I/Xl) cos x2 sin 2~3,
a3(~) = (1/~1)sinx2,
a 4(x) = I/X4.
Compute ui(x) at x = (1,0,0,0).
2.2. Minkowski Space-Time M4 and the Separation

Function
A four-dimensional differentiable manifold M4 is flat (or Minkowski) pro-
vided it admits a global Minkowski coordinate chart (X, M4). For such a chart
the coordinate basis vectors (in the tangent space J;,M4) are assumed to be
M -orthonormal:
Jl
l 0 0
[ei(p) ej(p)] = [dij] = r

010
0 0 1 (2.2.1)
000
for every p E M 4 . [In case the tangent space has an inner product, the mani-
fold is called Riemannian or pseudo-Riemann (which need not be flat).]
The space-time universe of special relativity is assumed to be the flat
Minkowski manifold M 4 . A point p E M4 represents an idealized event.
An event p has four coordinates with respect to a chart (X, U), X(p) =
(X 1 ,X 2 ,X 3 ,X4 ). If (X, U) is a Minkowski coordinate system, then (X 1 ,X 2 ,X 3 )
represents the spatial Cartesian coordinates and X4 represents the time of the
event. The history of a point particle consists of a continuous locus of events

in space-time. This history can be represented by a continuous curve y in
M 4 . Such a curve y is also called the world-line of the particle. Usually a
world-line is a twice differentiable curve. The image r c [R4 of the world-line
y will be plotted (see Fig. 6) instead of the image of y in M 4 Subsequently we
shall restrict ourselves to the use of Minkowski coordinate systems on M4
unless explicitly noted otherwise.
The world-line of a particle is locally classified by the value of
In the case of a massive particle it is assumed that
d .. dpi(t) df!"i(t) 0
'J dt dt < .
Thus, the tangent vector to the world-line of a massive particle is timelike.

For a massless photon (electromagnetic field quanta) or other massless parti-
cles such as neutrinos it is assumed that
d .. dPi(t) df!"i(t) _ 0
"dt dt - .
Thus, the world-line of a massless particle has a null tangent vector at each
point. For the case ofthe hypothetical tachyon the tangent vector is assumed
to be spacelike. For a free particle (not acted upon by any force) the world-
line is determined by the differential equations
(2.2.2)
The assumptions made in this paragraph are compatible with the experimen-
tal results of physics. (Chapter 5 explores the details of motion curves in
space-time.)
The general solutions of (2.2.2) are given by
(2.2.3)
where x~, vk are constants of integration. For known free particles d;jViV i :::;: O.
The null-cone (light-cone), which has its vertex at Xo E [R4, corresponds to
the set of points
(2.2.4)
where Xi are the Minkowski coordinates (see Fig. 7). The circles shown in
Figure 7 represent (suppressed) spherical surfaces.
The events in space-time can be classified according to their corresponding
positions relative to Nxo ' The events that correspond to the points in the
FIGURE 7. The null-cone N xo
subset
{x: dij(x i - xb)(x i - x6) ~ 0, X6 < X4}
are the future events relative to the event corresponding to xo'
The events corresponding to the points in the subset
{x: dij(x i - xh)(x i - x6) ~ 0, X4 < X6}
are the past events relative to xo. The events in the subset
{xo} u {x: dij(x i - xh)(x i - x6) > O}
are called the present events relative to Xo. This concept of the present events
generalizes the Newtonian concept of present events (which are called simul-
taneous events in relativity). This classification of events into future, past, and
present subsets is independent of any observer.
The events on Nxo ' for which X4 "# X6, satisfy
[(Xl - X(\)2 + (x 2 - X~)2 + (x 3 - X6)2J1/2/lx4 - x61 = 1. (2.2.5)
The physical meaning of the above expression is that the speed of light is 1 in
our units. Thus the points on Nxo represent the history of a spherical wave
front of light converging to and then emanating from the event Xo.
The space-time separation along a differentiable (nondegenerate, rectifi-
able) world-line y: [a, b] -'> M4 is defined by the functional S,
s = S(y) == f ba[ty(t)]dt == fb Idii~

a
df![i(t) df![i(t)ll!2
a ~ dt. (2.2.6)
The above equation is the generalization of the concept of arc-length of a

curve. In the case where the world-line y is continuous but piecewise differ-
entiable, the equation (2.2.6) can be generalized to
s = S(y) == f'[-I
a IJ
df![i(t) df![i(t) 11/2
d .. - - - -
dt dt
dt + ...
(2.2.7)
where a < t 1 < ... < t. < b are the n jump discontinuities of df![i(t)/dt.
Example: Let a parameterized differentiable curve I' be given by

Xl = f![1 (t) == t\
X2 = f![2(t) == 0,
X3 = .?[3(t) == 0,
X4 = ,r4(t) == J3t 3 , t E [0, 1].
The corresponding separation is
s= L 19t4+0+0-27t411/2dt=.j2. D
A straight timelike world-line satisfying (2.2.2) can represent an idealized

point observer who is not subjected to any net external force. Using the
equation Xk = x~ + vkt, the separation along a straight world-line is
s= I IdiiviVil1/2 dt = Idij(x i - xb)(x j - X6W/2. (2.2.8)
An idealized point observer is assumed to follow a continuous, piecewise

twice differentiable, timelike world-line that is not necessarily straight. He
may carry a standard clock, which is an idealized point clock that runs per-
fectly. The elapsed time between two events of the history of the observer as
measured by the standard clock is called the proper time between the events.
The proper time elapsed between two events, y(td and y(t 2 ), is assumed to be
the separation
_ _f.'21-
s - S(y) -
df![i(t) d?j(t) 11/2
dij-d- --d- dt. (2.2.9)
I, t t
Example: Let us confront the problem of the twin paradox with help of the
preceding assumptions. Consider an identical twin at the same place on the
earth. One of the twins leaves in a rocket at t = t1 with a constant velocity v
along the xl-axis. After a while he reverses his velocity at t = t2 and returns
to the earth at t = t3.
The paradox is whether or not the returning twin is at that instant younger
than the twin that stayed at home. The piecewise straight timelike world-lines
FIGURE 8. The space-time diagram for the

twin paradox.
y, y' for the twins (see Figure 8) are given by
x '1 -
-
{cc ++ 2vt
v(t - tJ,
+ td,
t1~t~t2'
o < Ivl < 1; (2.2.10)
z - v(t t2<t~t3'
It is easily seen that t3 = 2t2 - t 1. The proper time measured by the twin
brothers are given by (2.2.6), (2.2.7) so that
(2.2.11)
The above equation clearly shows that s' < s and the returning twin will be
younger! The asymmetry of the aging processes is revealed :in the asymmetry
of the world-lines of the twin brothers. 0
Now we shall obtain spacelike separations from timelike separations.
Theorem (2.2.1): Let points a, b, q define a triangle in [R4 (see Figure 9). Let p
be an arbitrary intermediate point on the straight line joininy a to b. Denoting
the straight lines from a to p; b to p; P to q; q to a; and b to q by 1 2 , 1 1 , I, 1 4 ,
and 13 respectively and assuming that Y1' Y2 are timelike, Y3' Y4 are null, Y is
spacelike, one must have
[S(y)J2 = [S(ydJ [S(yz)]' (2.2.12)
FIGURE 9. Measurement of a spacelike

separation.
Proof: From the assumptions and equation (2.2.8), it follows that
di)qi - ai)(qi - a i ) = dij(qi - bi)(qi - b i ) = O.

We can write
qi _ ai = (qi _ pi) _ (ai _ pi),
qi _ bi = (qi _ pi) + (pi _ bi),
pi _ bi = A(ai _ pi),
Putting these equations into the conditions for null separations we obtain
Adij[(qi - pi)(qi _ pi) + (a i _ pi)(ai - pi) - 2(qi - pi)(ai - pi)] = 0,
dij[(qi _ pi)(qi _ pi) + A2(a i _ pi)(ai - pi) + 2A(qi - pi)(a i - pi)] = O.

Adding the above equations and cancelling the factor (1 + A) we get
dij(qi _ pi)(qi _ pi) = _ Adij(a i _ pi)(ai _ pi),
or [S(y)]2 = A[S(Y2)]2 = [S(yd] [S(Y2)].
The equation (2.2.12) suggests an operational method of measuring a space-

like separation by a standard clock and emission, reflection, and absorption
of photons.
Now we shall define the M-orthogonal intersection between two straight
world-lines YI' Y2 in M 4 Suppose that the corresponding straight lines r l , r 2
in 1R4 intersect at a (see Figure 10). The intersection at a is defined to be
M -orthogonal provided
(2.2.13)
where b, c are arbitrary points on r l , r 2 respectively and all three points are
distinct.
FIGURE to. M-Orthogonal in- 0

tersection.
In the case where the straight world-lines Yl' Y2 are both spacelike we have
dij(b i - ai)(b i - a i ) > 0,
dii(c i - ai)(c i - a i ) > O.
Using (2.2.13) we derive

diic i - bi)(C i - b i )
= dii[(c i - a i) - W - a i )] [(c i - a i ) - (b i - a i )]
= dii ci - ai)(c i - a i ) + dij(b i - ai)(b i - a i ) > 0;
that is, the straight world-line Y3 is spacelike. Furthermore"
[S(Y3)]2 = [S(yd]2 + [S(Y2)]2,
which is the generalized theorem of Pythagoras. In case one of the straight
world-lines, say Yl' is spacelike and the other is timelike, we have
dij(b i - ai)(b i - a i ) > 0,
dij(c i - ai)(c i - a i ) < o.
Using the M-orthogonality condition (2.2.13) we again get
dij(C i - bi)(C i - b i ) = dij(c i - ai)(c i - a i ) + dii(b i - ai)(b i - a i ),
or
[S(Y3)]2 = [S(Yl)]2 - [S(Y2)Y
Choosing Y3 to be a null straight world-line we have S(Y3) = 0 and
S(Yl) = S(Y2) (2.2.14)
FIGURE 11. Distance between two straight

lines.
t'
This equation can be used to determine the M-orthogonal intersection of a

timelike and a spacelike curve.
Now we shall derive the expression for the "distance" between two timelike
straight world-lines y, y' (see Fig. 11).
In terms of proper time parameters sand s', the equations of two straight
world-lines y and y' are
dutitj = -1,
(2.2.15)
dij(i(i = -1,
The two straight world-lines y, y' are defined to be parallel if t i = t d . (In
general, parallel lines can be defined only for a flat differentiable manifold.)
Assuming that the straight lines y, y intersect M-orthogonally, we have
dkix,k - Xk)tj = dkj(t,ks' + b'k - tks - bk)tj = 0,
(2.2.16)
s' = (dkjtjt'k)-l [dQctQ(b C- b'C) - s].
We can recall Theorem (1.1.4) to note that dkjtjt'k =f. 0 and thus conclude that
s' always exists. From (2.2.15), (2.2.16) we can express
X'k - Xk = _[tk + (di}it,jr1t,k]s + f3\
(2.2.17)
f3k == b'k - b k + (dijti(j)-l [dQctQt'k(b C_ b'c)].
We can obtain the expression for the distance D(s) along the straight world-
line y by the equation
[D(s)]2 == [S(y)]2 = dk/x'k - xk)(x'j - xi)
= [1 - (d i}i t ,jr 2 ]s2
- {2dijf3i[t i + (dk1tktd)-lt'j]}S + dijf3if3j, (2.2.18)
which is a second-degree polynomial in the variable s.
Theorem (2.2.2) (Synge): Let y and y' be two timelike straight world-lines in
M 4 , given by equation (2.2.15). Then the world-lines y and y' are parallel if and
only if D(s) is a constant.
Proof: (i) Assume that y and y' are parallel. Then t k = t 'k and by (2.2.15) and
/t
(2.2.18) we have dkltkt'l = -1, 1 - (d i 'i )-2 = 0, t i + (dkltkt!l)t'i = 0, and
[D(S)]2 = dij/Ji{3 i . Thus D(s) is a constant.
(ii) Assume that D(s) is a constant. Then equation (2.2.18) gives
d2
ds 2 {[D(S)]2} = 0, 2[1 - (d ij tit'i)-2] = 0, d,}it'j = l.
Recalling Theorem (1.1.5) we must have dijtit'i = -l. Then by Corollary
(1.1.2) we have t 'i = ti. Thus y and y' are parallel.
Let y: [t l , t 2 ] --+ M4 be a straight world-line. Among all the coordinate

charts, a Minkowski coordinate chart (X, M 4 ) will give the separation S(y)
along the straight world-line y by the simple formula [see equation (2.2.8)]
[S(y)]2 = l(xl2) - xll / + (Xf2) - xfl/ + (X{2) - X{I2 .- (Xt2) - xtl21,
(2.2.19)
where xii) are the coordinates ofy(td and Xt2) are the coordinates ofy(t 2 }.
A physicist, informed of the mathematical existence of such a Minkowski
coordinate chart, would like to know how to assign such coordinates to
events by the process of physical measurements. This is the problem of opera-
tional aspects of the Minkowski chronometry. We can assign coordinates of
an event by emission, reflection, and absorption of photons and the measure-
ments of proper times on a standard clock. The equations (2.2.12), (2.2.14),
the generalized Pythagoras theorem, and Theorem (2.2.2) were all derived for
the purpose of measurements in Minkowski chronometry. These equations
and theorems tell us how to measure a spacelike separation by measurements
of proper times; how to conclude M -orthogonality of intersections of straight
world-lines by measurements of separations; and how to find that two straight
world-lines are parallel. We shall use all these results to construct an opera-
tional Minkowski coordinate chart. We take an observer, who is not sub-
jected to any net external force and thus has a straight timelike world-line. At
an arbitrary event e (see Figure 12) on his history, he sets his standard clock
FIGURE 12. Operational construction of Minkowski coordinates.

to zero and his spatial coordinates to (0,0,0). After a while, his standard clock
reads unity and that event is denoted by d. By the measurements of proper
times and by using equations (2.2.12), (2.2.14) and the Pythagoras theorem,
the observer can identify three other events corresponding to a, b, c such that
each of the line segments connecting e to a, e to b, e to c, and e to d has unit
separation and all of them intersect M-orthogonally at e. The events a, b, c
may almost be chosen arbitrarily. Let the events a, b, c, d, e have coordinates
ai, bi, c i, d i, e i with respect to a Minkowski coordinate chart. These coordi-
nates exist in principle and may be unknown to the observer. But from
the measurements, the separations along four line segments must satisfy the
M-orthonormality conditions
dij(a i - ei)(a i - e i ) = dijW - ei)(b i - ei )
= dij(c i - ei)(c i - ei )
= - di)d i - ei)(d i - ei ) = 1,
dij(a i - ei)(b i - ei ) = dij(a i - ei)(c i - ei ) = dij(a i - ei)(d i - ei ) (2.2.20)
= dij(b i - ei)(c i - ei ) = dij(b i - ei)(d i - ei )
= dij(c i - ei)(d i - ei ) = O.
In the above equations there are ten equations for twenty unknowns. Thus
there are ten degrees of freedom in the solutions for ai, bi, c i , d i, e i. There are
four degrees of freedom for the event e and six degrees of freedom for the
events a, b, c, d. Now we choose a sixth event corresponding to x(1) (see Fig.
12). We can go from e to X(l) by going from e toward a, then going parallel to
e to b, e to c, and e to d in turn. (Going parallel is a measurable process, since
it involves keeping the "distance" or separation to a constant.) The separa-
tions in these four steps are measurable in chronometry, and let their values
be ~ll)' ~fll' ~tl)' ~(1)' By virtue of the construction just carried out, the mea-
sured separations ~tl)'s are related by the equations
xtl) = ~tl)ak + ~[l)bk + n)c k + ~(l)dk.
Similarly, for a seventh event x(2) = (xtZl' X[2)' xtZ)' X(Z, the measured and un-
known coordinates are related by
x~) = ~t2)ak + ~f2)bk + ~t2)Ck + ~(2)dk.
By subtracting the last two equations and using (2.2.20) we can get the sepa-
ration along the line segment corresponding to r as
[S(,,)]2 = Idi)xh) - Xtl)(X/2) - x/!) I
= 1(~t2) - ~tl2 + (~f2) - ~fl)f + (~t2) - ~tl2 - (~(2) - ~(121.
Comparing the above equation with (2.2.19) we realize that the observer
has succeeded in assigning to the events corresponding to x(l), X(2)' other
Minkowski coordinates, ~tl) and ~t2)' obtained by measurements in
Minkowski chronometry.
EXERCISES 2.2
1. A continuous, piecewise differentiable curve y in M4 is given by the
Minkowski coordinates
Xl = grl(t) == 0;
x 2 = gr2(t) == 0;
x3 = gr3(t) == {t,2, 0:::;; t :::;; 2,

2 < t:::;; 3;
X4 = gr4(t) == {O,t - : :; t :::;; 1,

1, I< t :::;; 3.
Obtain the separation s = S(y).
2. Let y, y' be two timelike straight world-lines given by the equations
Obtain all the possible conditions on tk, t'k, b\ b'k for the intersection of y and
y', by equating the distance function D(s) to zero.
2.3. Flat Submanifolds of Minkowski Space-Time M4

A differentiable submanifold a c M4 is a proper open subset of M4 such that
the assumptions of a differentiable manifold are satisfied for a. If further-
more a is flat, it is called a flat submanifold. We shall now define three flat
submanifolds.
(i) Using a Minkowski coordinate system, a 3-flat a3 c 1\14 is defined to be
a subset corresponding to
I:3 == {x: Aixi +B= 0, (iii AiAj > O} c /R 4. (2.3.1)
Here, A;'s are constants.
(ii) A 2-flat az is defined to be a subset corresponding to
I:2 == {x: A;Xi + B = 0, A;Xi + B' = 0, (iijAiAj > 0, (iijA;Aj > 0, A; #- AAi}'
(2.3.2)
(iii) A I-flat a l is defined to be a subset corresponding to
I:I == {x: AiXi +B= 0, A;Xi + B' = 0, A;'Xi + B" = O}. (2.3.3)
It is furthermore assumed that the rank of the matrix
is three. Therefore, solving the defining equations in (2.3.3), we get

Xl = C - At,
XZ = c' - A't,
(2.3.4)
x3 = c" - Alit,
X4 = t E 1R4
for some suitable constants c, c', c", A, A', A" so that (AZ + A'z + A"z > 0).
Thus the subset 0"1 can always be identified with a straight world-line.
In the case where 1: 3, 1:2 contain the origin x = 0, the defining equations
simplify to
1:3 == {x: Aixi = 0, c5 ijAiAj > O}
. ..... (2.3.5)
1: z == {x: Aix' = 0, A;x' = 0, c5IJAiAj > 0, c5'JA;Aj > 0, A; # .leA;}.
Let us solve the defining equations of 1: z , namely,
(2.3.6)
Since the rank of the coefficient matrix is assumed to be two, we have solu-
tions in the form
Xl = al x 3 + bl x 4 ,
(2.3.7)
x2 = a2x3 + bz x 4
for suitable constants ai' a z , bl , bz such that
ai + a~ > 0, bf + bi > 0,
We can rewrite the solutions as

xl=a1u+b1v,
XZ = azu + bzv,
(2.3.8)
where (u, v) E IRz. Since the solution space in (2.3.8) is two-dimensional, we

can also write the symmetrical form of the parametric representation
1: z == {x: Xi = aiu + biv, c5ija iaj > 0, (jijbibj > 0, ai # .lebi, (u, v) E IRZ}. (2.3.9)
A straight world-line y is M -orthogonal to a flat submanifold 0", provided
every straight world-line in 0" that intersects y is M-orthogonal to y.
Theorem (2.3.1): Let 1:3 == {x: Aixi = 0, (jijAiAj > OJ. Then the straight wor/d-
line y, given by the equations Xi = Ait, t E IR (Ai == d ij Aj), intersects the cor-
responding 3-j1at 0"3 M-orthogonally.
FIGURE 13. M-Orthogonal intersection of y

with 113.
Proof: Let Xfl)= A it(1)' as in Figure 13. Therefore, we have dijxix(l) =

dijx iA j t(1) = (AiXi)t(1) = O.
Example 1: Consider the straight world-line y corresponding to the x 4 -axis.

It is given by the parametric equations
The corresponding M-orthogonal 3-flat G"3 is given by

L3 = {x: X4 = 0,(Xl,X2,X3) E [R3}. D
Example 2: A 2-flat G"2 is given by the corresponding subset L z with pa-

rametric equations
The same subset L2 is also given by L z = {x: X2 = X3 = O}. 0
We can classify 3-flats corresponding to the value of dij AiAj. A 3-flat is

called spacelike, time like, or null in case dij AjAj < 0, d ij AiAj > 0, or dij AiAj =
o respectively.
We can classify 2-flats according to the value of
~ == (a ib i)2 - (aiai)(bjb j ),
with aib i == dijaib j.
(i) In the case ~ > 0, the 2-11at G"2 is called timelike.
(ii) In the case ~ < 0, the 2-flat G"2 is called spacelike.
(iii) In the case ~ = 0, the 2-11at G"z is called null.
Theorem (2.3.2): Let the null-cone corresponding to the points No ==

{x: dijxixj = O} intersect with a 2-flat corresponding to L 2 , which passes
through the origin.
(i) For a time like 2-j1at, the intersection contains two null-lines.
(ii) For a space like 2-j1at, the intersection is at one event corresponding to the
origin.
(iii) For a null 2-j1at, the intersection is a single null-line.
Proof: The parametric equations for L2 are Xi = aiu + biV, as in equation

(2.3.9). In the case aia i # 0, we shall retain this parametrization. In the case
aia i = 0, bib i # 0, we shall interchange parameters and coefficients by putting
U = v, V = u, iii = bi, hi = a i and dropping the bars subsequently. In the case
aia i = bib i = 0 (so that a i and b i are components of two null vectors not
M -orthogonal) we shall change to new parameters and coefficients by putting
U = u, V = v - u, iii = a i + b i, hi = b i (iiiiii = 2a ib i # 0) and dropping bars
subsequently. So, by proper parametrization we can always represent L2 by
equation (2.3.9), with aia i # O. Now, for the intersection with the null-cone,
we must have
dijxixi = dij(aiu + biv)(aiu + biv)
= (a ia i)u 2 + 2(ay)uv + (b ib i)V 2
= O. (2.3.10)
Solving for u we find
(i) For a timelike 2-flat d > 0 and u has two distinct solutions. Putting
these solutions back into (2.3.9) we get
Xi = ;?[i(V) == {ai(akakrl[ -(aibi) + fi] + bi}v == niv,
(2.3.11 )
Xi = ;?['i(V) == -{ai(akak)-l[(aibj) + ;-d] - bi}v == nfiv,
where v E IR. Clearly (2.3.11) yields two null-lines, since nini = n;n,i = O.
(ii) For a spacelike 2-flat d < 0 and u has complex roots unless v = O. In
that case u = v = 0, Xi = 0, which denotes the origin.
(iii) For a null 2-flat d = 0 and u has one real solution. Substituting this
solution into (2.3.9) we obtain
(2.3.12)
where v E IR. Obviously, (2.3.12) yields a single null-line (n(O)intO) = 0) passing
through the origin.
Examples: (i) An example of a timelike 2-flat is given by

L2 == {x: Xi = b i 1 U + bi 4 v, (u, v) E 1R2}.
It is clear that ~ = 1. The two null-lines of intersection with No as given by
equation (2.3.11) are
Xl = X4 = v,
Xl = _x 4 = -v,
(ii) An example of a spacelike 2-flat is given by

1:2 == {x: Xi = (ji1U + (ji 2V , (u,v) E \R 2 }.
= - 1. The intersection with No is only at the origin x = O.
It is clear that L\
(iii) An example of a null 2-flat is given by
1:2 == {x: xi = (jilU + W4 - (ji2)V, (u,v) E \R 2 }.
It is clear that L\ = O. The single null-line of intersection with No corresponds
to the equation
Now, we shall state and prove a counterintuitive theorem about a null

2-flat.
Theorem (2.3.3) (Synge): Let 1:2 and ro correspond to a null 2-j7at passing
through the origin and the single null-line of intersection of 1:2 with the null-
cone No respectively. Then every straight world-line in 1:2 that intersects ro is
M-orthogonal to roo
Proof: Suppose x E 1:2 and (Xo,1: 2 ) is a chart for 1:2 as in Figure 14. Then we
can write
(2.3.13)
For two points Xo. x# on ro we have

o = dij(xb - x#i)(x6 - x#j)
= (aia i ) [(u o - u#) + (ajbj)(akaktl(vo - v#)JZ. (2.3.14)
To check M-orthogonality of intersection between a line r 1 in 1:2 and the
v
Xo
-----+----u
(u,v)
FIGURE 14. Intersection of a line r 1 with the null-line roo

single null-line ro at x#, we have to compute

dij(xb - X#i)(X{ - x#i)
= [ai(u O - u#) + bi(v O- v#)] [ai(u l - u#) + bi(Vl - v#)]

= (aiai)(uO - u#)(u l - u#) + (bibi)(vo - v#)(v l - v#)
+ (ay) [(u o - U#)(Vl - v#) + (vo - v#)(u l - u#)]
bi 2 # #
_ (i)(
- aia
#)(
Uo - U Ul - U
#)
+ (ai ) (vo - V )(Vl - V )
(')
aja J
+ (ajbi)[(uo - u#)(v l - v#) + (vo - v#)(u l - u#)]
=( . i) [( _ #) (akbk)(vo - V#)] [( _ #) (alb')(Vl - V#)]

a,a Uo U + (j)
aja
U1 U + (ama m) .
By equation (2.3.14), the above expression vanishes. Thus r l and ro intersect

M-orthogonally at x#. Furthermore, r 1 could be ro itself.
We shall now consider the M-orthogonal projection from an event to a

straight world-line. Let r correspond to a nonnull straight world-line, given
by the parametric equations
t E lit
Let y correspond to an event from which an M-orthogonal projection is
made onto the line r as in Figure 15. We want to obtain the point z
on r corresponding to the projection. The M-orthogonality implies that
ai(zi - yi) = 0. Suppose that Zi = aitl + b i for some tl E IR. The M-
orthogonality yields tl = ai(aka k )-I(yi - bi). Substituting this value into the
expression for Zi, we finally get
(2.3.15)
Now, we shall obtain the M-orthogonal projection z from the point y to
the non null 2-flat corresponding to L2' given by
Xi = aiu + biv, A #- 0, (u, v) E 1R2.
JR4
FIGURE 15. M-Orthogonal projection on a straight

line.
From the M-orthogonality, we have aj(zi - yi) = MZi - yi) = 0. Let Zi =

+ biv i for some (u l , VI) E [R2. The M-orthogonality yields solutions
aiu l
UI = -(aiyiblb l - ajbibkyk)/A,
. I . k
VI = -(aia'bly - ajyJakb )/A.
Substituting these values, we get the expression
zr = -ar(aiyiblhl - ajbjbkyk)/A - br(ajaibll- ajyiakbk)/A. (2.3.16)
In the case aib i = 0, the equation (2.3.16) reduces to
zr = (araiyi/aka k) + (b'biyi/bkb k). (2.3.17)
EXERCISES 2.3
1. Suppose that a nonnull 2-flat
~2 == {x: Xi = aiu + biv, (jiiajaj > 0, (jiiVibi > 0, ai #- Abi , U, V E IR}.
The M-orthogonal 2-flat is defined by
~~ == {x: ajx i = 0, bix l = O}.
(i) Prove that the intersection ~2 n ~~ = {O}.
(ii) Prove that every straight line in ~~ passing through the origin is
M-orthogonal to every straight line in ~2 passing through the origin.
(iii) Prove that if ~2 corresponds to a null 2-flat then ~~ corresponds to
another null 2-flat.
2. Suppose that y and ~3 == {x: Aixi = 0, AiAi #- O} correspond to an event
and a 3-flat respectively (y ~3)' Obtain explicitly the point z that corre-
sponds to the M-orthogonal projection of y onto ~3'
2.4. Minkowski Tensor Fields on M4

Consider two charts (X, U) and (X, 0) in M4 that are ct'3- related and the
transformations are given by
Xk = gk(X) for x E Do,
Xk = Xk(X) for x E Do,
O(Xl x 2 x 3 x 4 )
o( .. 2' 3' 4) #- 0 for x E Do'
x,x,x,x
Let us consider two continuous vector fields u(p), v(p) for all p E UnO. If
both charts are Minkowski charts then the inner product is given in terms of
the components by
u(p)' v(p) = u[X-1(x)]' v[X- 1(x)] = diiui(x)vj(x)
= u[X-1(x)]' v[i-l (x)] = dijuj(x)Vi(x).
By the transformation rules in equation (2.1.9), we obtain

OXk(X) OXI(X)
[ dkl---aT aT - J' .
dij u'(x)vJ(x) = O.
Since the above equation is valid for every pair of continuous vector fields in
UnO, we must have
(2.4.1)
An inhomogeneous linear transformation

Xi = Xi(X) == a i + (ijxj, (2.4.2)
where a i E IR and [ij are the entries of a Lorentz matrix, is called a Poincare
transformation.
Theorem (2.4.1): The coordinate transformation from one Minkowski chart to

another is a Poincare transformation [equation (2.4.2)].
Proof: We start with the system of first-order nonlinear partial differential

equations (2.4.1). We denote partial derivatives by
X~ = OXk(X) ~k _ 0 2 Xk(X)
,- oxi' Xis = ox.ox i '
which are all continuous functions. Since Xk are ~3-functions,
~ 02 Xk(X) 02 Xk(x) ~
X~ == .= . = X:i
oxox' ox'ox
Differentiating equation (2.4.1) with respect to X S we get

+ Xi Xjs) = O.
~k ~I ~k ~I
dkl(XisXj (2.4.3)
Cyclically permuting the indices yields
+ Xj XsJ = O.
~k ~I ~k ~I
dkl(XjiXs (2.4.4)
Subtracting (2.4.3) from (2.4.4) and using Xt. = X:i we have
= O.
~k ~l ~k ~l
dkl(XjiXs - Xi Xj.) (2.4.5)
Interchanging sand j in the last equation we get
-k-'
dkl(XsiXj - Xi X.j ) = O.
~k ~l
(2.4.6)
Adding equations (2.4.3) and (2.4.6) we have 2dkI X:i X] = O. MUltiplying
the above equation with t(oXj(x)/oxQ) and recalling that (OXI(X)/oxj)
(oXj(x)/oxQ) = t5 IQ, we get
(2.4.7)
Multiplying the above equation by d ja we obtain

02 Xj(x)
. = O. (2.4.8)
ox'ox s
Integrating the above linear partial differential equations we have oXi(x)/ox j =
Iii' Xi = Xi(X) = a i + lii xi, where the a i and the lij are constants of integra-
tion. Since the equation (2.4.1) has to be satisfied, the matrix of the constants
L = [/ij] must satisfy L TDL = D.
The Lorentz matrix L, which satisfies the condition (1.2.8), has six indepen-
dent constants lij' In a Poincare transformation (2.4.2), the four constants a i
are independent. Thus the Poincare transformations involve ten independent
parameters. In the case where a i == 0, the transformation reduces to
LTDL = D. (2.4.9)
The above transformation, which involves six parameters, is called a Lorentz
transformation of coordinates. Note that a Lorentz matrix is involved in both
the Lorentz mapping of V4 and the Lorentz transformation of coordinates in
M4
Under a Poincare transformation (2.4.2), the components ui(x) of a vector
field will undergo the transformation
k OXk(X). k'
U (x) = -~-.- u'(x) = I jU'(x). (2.4.10)
ux'
The above equation reminds us of the equation (1.3.1). So we shall define
the transformation rules for the components of an (r + s)-order (or rank)
Minkowski tensor field under a Poincare transformation by analogy to
(1.3.8). Thus we assume
f i , :':~r(x) = Ii, .. '/ir kr am,11. ... am.Js rk, .....krms (x) '
11 Js kl nil
(2.4.11)
li ja \ = aijljk = (ji k
The apparent switching of the li k, amj in (1.3.8) and (2.4.11) is due to similar
switching in equations (1.3.1) and (2.4.10).
Example: Consider a scalar field rex) [a (0 + O)-order tensor field]. The trans-
formation of the scalar field by equation (2.4.11) is
f(x) = rex). (2.4.12)
Let this scalar field be defined by
rex) == (Xl)2 /2 + 3(X4)2
for (X l ,x 2,x3,x4) E [R4. In the hatted Minkowski coordinates, the scalar field
is given by (see equation (2.4.2
rex) = [alk( _a k + x k)]2/2 + 3[a\( _a k + Xk)y D
We shall generalize the theorems (1.3.2), (1.3.3), and (1.3.4) to create new
Minkowski tensor fields out of known Minkowski tensor fields.
Theorem (2.4.2): Let the components of scalar and Minkowski tensor fields
A' (x,
)
Jl (x,
)
rj""j. x,
i " " ir ( ) i''''ir ( )
(J'it"'j. X, x
k,"'k ( )
Pm,.""fq
be all defined in D c 1R4. Then

(i) the linear combination
produces the components of an (r + s)-order Minkowski tensor field in D;

(ii) the tensor product
gives the components of an [(r + p) + (s + q)]-order Minkowski tensor field in

D:
(iii) the single contraction
yields the components of an [(r - 1) + (s - 1)J-order Minkowski tensor field

inDo
The proof of this theorem is entirely parallel to that of theorems (1.3.2),

(1.3.3), and (1.3.4).
The raising and lowering of indices in a component of a Minkowski tensor
field is defined exactly as in equation (1.3.10).
The partial differentiation of tensor field components generates a new ten-
sor field. Let us denote a partial differentiation by a comma, that is,
Theorem (2.4.3): The partial derivatives rj: . . ..;:.a transform as the components
of a [r + (s + 1)]-order Minkowski tensor field.
Proof: Using equations (2.4.2), (2.4.11) and the chain rule we have
Example: Consider a Poincare transformation

i 1 = fJ(x 1 - vx 4 ),
i 2 = x2 - 1
(2.4.13)
i4 = fJ( - vx 1 + X4),
where fJ == (1 - V2)-1/2 and 0 < Ivl < 1.
Let a contravariant Minkowski vector field be defined by
for (xl, x 2 , x 3 , x 4) E 1R4. Then the transformed component M1(i)/ai 1 under

(2.4.13) is
fl,l = (- fJV)(fJV)T 4,4 = - fJ 2V2(X 4)2 = - fJ4V2(vi l + i4)2. D
We will now briefly discuss integration on the flat space-time manifold

M 4 Although elegant treatments involving differential forms and chains can
be used, we shall be content here with the usual multiple Riemann integral.
Suppose that (X, M 4 ) is a Minkowski chart for M 4 . Let U be a closed region
in M4 such that X(U) = D u aD, where aD is the boundary of Dc 1R4. Let
f: U c 1R4 --t IR be a continuous function. Then the volume integral
fuf(p)d4V == fn F(x)dx 1dx 2 dx 3 dx\ (2.4.14)
where F(x) == f[x- 1 (x)] = f(p) and the integral on the right is a multiple
Riemann -in tegral.
Example: Let F(x) == exp[ -(lxll + Ix 2 1+ Ix3 1+ Ix4 1)]'
f f(p)d4V = f exp[ -(Ixll + Ix 2 1 + Ix 3 1 + Ix41)]dx1dx2dx3dx4
t,
M4 II.
= [f: exp[ -(Ixll + Ix 2 1+ Ix 3 1+ Ix4 1)] dX 4]dX I dx 2 dx 3
=2 t2 [f: exp[ -(Ixll + Ix 2 1+ Ix 3 1)] dx 3 Jdx 1dx 2
= 16. D
To evaluate the above integral, we have used Fubini's Theorem and the
definition of an improper Riemann integral of the first kind.
We now state a very useful theorem involving integration. It is called
Gauss's Theorem or Green's Theorem on a four-dimensional manifold.
FIGURE 16. The region D u aD of a

four-dimensional integration.
Theorem (2.4.5): Let 15 = D u aD be a four-dimensional region of [R4 such that

the boundary aD is continuous, piecewise differentiable and orientable. Let ni(x)
be the components of a unit (ldijninjl = 1), outward normal vector to aD (as in
Fig. 16). Let ri(x) be a differentiable field in D. Then
f- ri.idxldxzdx3dx4 =
D
f oD
dijr i(x)n j(x)d 3 (J, (2.4.15)
where d 3 (J denotes the invariant, threedimensional hypersurface element

of aD.
The proof of this important theorem can be found in references at the end
of the chapter.
Example: See Figure 16. Consider the domain
D == {x: (xl)Z + (X 2 )2 + (X 3)2 < 1, tl < X4 < tz }

with boundary
aD = aD I v aDz vaD 3 ,
aD l = {x: (Xl)l + (X 2 )l + (x 3 )Z = 1, tl :-=::; X4:-=::; t l },
aDl = {x: (Xl)l + (Xl)2 + (X 3 )l < 1, X4 = td,
aD 3 = {x: (Xl)l + (x 2 )Z + (X 3 )2 < 1, X4 = t z }.
On aDz and aD 3 the invariant hypersurface element is d 3 (J = dx l dx z dx 3 ,

whereas on aD I the hypersurface elements is
d 3 (J = dx 1 dxzdx4/.jl=(~I)Z - (Xl)l. 0
EXERCISES 2.4
1. Prove that, if the 4'+5 components fj: ...... j:Cx) of a Minkowski tensor field
are zero everywhere, then in another coordinate system related by a Poincare
transformation rj: ...... j;(x) == O.
2. Suppose that a twice differentiable scalar field -p(x) satisfies the wave
equation D-P(x) == dij-P,ii = O. Prove that in another Minkowski coordinate
b '"
system D-P('x) == d a -P,ab = 0,
..... A
3. Suppose that Aii(x) = - Aiix) is an anti symmetric, twice differentiable

second-order covariant Minkowski tensor field. Prove that A ii .k + A jk i +
A ki j transforms as a totally antisymmetric (0 + 3)-order Minkowski tensor
field.
4. Prove that the four-dimensional divergence equation /i = 0 can be
solved by the expression / = AU. i , where Aij(x) is an arbitrary anti symmetric
twice differentiable (2 + O)-order Minkowski tensor field.
References
I. N. J. Hicks, Notes on differential geometry, Van Nostrand, London, 1971. [pp. 27,
30]
2. D. Lovelock and H. Rund, Tensors, differential forms, and variational principles,
John Wiley and Sons, New York, 1975. [po 61]
3. W. Noll, Notes on Tensor Analysis prepared by C. C. Wang, The Johns Hopkins
University, Mathematics Department, 1963. [po 61]
4. J. L. Synge, Relativity: The special theory, North-Holland, Amsterdam, 1964. [pp.
35,37,38,40,41,44,45,48,50,53,55,59]
5. J. L. Synge and A. Schild, Tensor calculus, University of Toronto Press, Toronto,
1966. [po 61]
3
The Lorentz Transformation
3.1. Applications of the Lorentz Transformation

We shall derive the Lorentz transformation by physical arguments. Let us
pretend for a short while that we do not know about Minkowski space-time
and Minkowski coordinates. Instead, we are aware of space and time and
inertial frames of reference. An inertial observer, an idealized point observer
subject to no forces, is assumed to follow a straight line in Euclidean space
E 3 In a suitably chosen Cartesian coordinate system, the straight line of the
inertial observer can be given by
x = E(<X4) == 0,
X4 = E(4(X4) == X4 E IR.
The above equation may also be interpreted in an inertial frame of reference
as the equation of a straight world-line in space and time. Similarly, a second
inertial observer, in relative motion to the first one, can be characterized in
another Cartesian coordinate and time system as
xa = a(x 4 ) == 0,
X4 = 4(X 4) == X4 E IR.
The transformation between these two coordinate systems in space and time
can be given by
x = ga(x 1 ,x2 ,x\x4 ),
(3.1.1)
X4 = g4(X\X 2,X\X4),
At this stage the functions gk are unknown. Now let both inertial observers
locate another massive free particle traversing on a straight line in space or a
straight world-line in space and time. The parametric equations for this line,
48
3.1. Applications ofthe Lorentz Transformation 49
in two coordinate systems, are given by

(3.1.2)
where v", x~, v", .x~ are constants. In the case (3.1.1) is given by a linear
transformation (not necessarily homogeneous)
(3.1.3)
the two equations in (3.1.2) are mutually consistent. The converse is also true.
For the sake of simplicity, assume that two inertial observers meet at some
initial time at the origin of space. This assumption simplifies (3.1.3) to
Xk = .fk(X) == ).kjx j (3.1.4)
Finally we have to incorporate the results of the Michelson-Morley experi-
ment, which implies the constancy of the speed of light relative to inertial
observers. Thus we have
speed of light = 1
J(X 1 )2 -I- (X 2)2 -I- (X 3)2 J(X 1 )2 -I- (X2)2 -I- (X3?
IX41 Ix 4 1
(3.1.5)
or
for all events on the null-cone corresponding to .#'0' The equations (3.1.4)
together with (3.1.5) do not yield the Lorentz transformation (see problem
(3.1. # I. However, it is usually assumed that
(3.1.6)
for all events in space-time. In that case, using (3.1.4) and the fact that {Xixi:
1 ::;; i ::;; j ::;; 4} is a basis for the second-degree homogeneous polynomials in
xl, x 2 , x 3 , X4, it follows that
(3.1.7)
where [/kj ] is a Lorentz matrix. Historically, the Lorentz transformation
(2.4.9), (2.4.10) was derived from such a sequence of arguments.
Now we shall derive a particular Lorentz transformation and interpret it
in physical terms. For that purpose we take inertial observers in a 2-flat
1:2 == {x: x 2 = x 3 = O}, as Figure 17.
We restrict the Lorentz transformation so that X2 = x 2, X3 = x 3. There-
fore, the resulting Lorentz transformation and the Lorentz matrix can be
written as
X2 = x 2,
(3.1.8)
x3 = x 3 ,
50 3. The Lorentz Transformation
'
~
----+-------1 ~' --~~~--,---
x
FIGURE 17. The special Lorentz transformation.
[11 0 0 [14]
L= r o
0
1 0
0 1
0
0
141 0 0 1\
From the condition LTDL = D, we obtain three independent nonlinear cou-
pled algebraic equations:
Wd 2 - (1\)2 = 1,
(144)2 - W4)2 = 1, (3.1.9)
114/11 -/44 /\ = O.
The first two of the above equations can be solved by putting (3.1.10)
Wd 2 = cosh 2 oc',
Wtf = sinh 2 oc',
(3.1.10)
(l44f = cosh 2 oc,
W4)2 = sinh 2 0c
for some real parameters oc, oc'. Now in Newtonian mechanics, which should
be a good approximation for low relative velocities, the Galilean transforma-
tion for the two inertial systems is given by
(3.1.11)
3.1. Applications ofthe Lorentz Transformation 51
Comparing the relative signs of the coefficients in (3.1.8), and (3.1.11) we

choose [11 > 0, [44 > 0. These inequalities yield from (3.1.10), the relations
[11 = cosh rx',
[41 = sinh rx',
(3.1.12)
[44 = cosh rx,
[14 = sinh rx.
The third condition in (3.1.10) implies that sinh(rx' rx) == 0, so rx' = rx.
Thus we can conclude that
[11 = cosh rx,
[41 = sinh rx,
(3.1.13)
[44 = cosh rx,
[14 = sinh rx.
Consider an event corresponding to e = (0,0,0, .i4) on the hatted obser-
ver's world-line (see Fig. 17). The coordinate .i4 is the proper time of the event
e measured by the moving standard clock. For this event, the coordinates of
the first observer are e = (Xl, 0, 0, x 4 ) such that
which is equivalent to
speed = Ivl = distance/time = Ixl/x41 = Itanhrxl. (3.1.14)
Since Itanh rxl < 1 for all (X E ~, we have Ivl < 1. This inequality yields a pro-
found physical consequence: The speed of any moving observer or any massive
particle must be strictly less than the speed of light. From (3.1.8), (3.1.12),
(3.1.14) we have
[II p,
=
[\ = vp,
[44 = p,
[14 = vp,
where p == 1/j1=~ and Ivl < 1. The ambiguity in the sign can be removed
by comparison with equation (3.1.11) for small v. Thus we have obtained a
special Lorentz transformation (boost)
.i 1 = P(x 1 - vx4 ),
(3.1.15)
where p == 1/~ and Ivl < 1. This is the relativistic generalization ofthe
Galilean transformation (3.1.11).
In general the transformation from the Minkowski coordinates associated
with an inertial observer to those associated with another is given by the
Poincare transformation (2.4.2)
We shall derive some of the physical consequences of the special Lorentz

transformation (3.1.15).
(i) First we shall investigate the simultaneity of events. Suppose that in
a Minkowski coordinate system associated with the first inertial observer
two distinct events X(l) = (x(\)' 0, 0, xil)' X(2) = (Xl2)' 0, 0, x~ are simultane-
ous. Thus we must have xi1) = x~)' xl1) # xl2 ). By the Lorentz transforma-
tion (3.1.15) we get
Xi2) - Xil) = vP(xlt) - xl2 + P(Xi2) - Xil

= vP(xll ) - xb) #
for v # 0. Therefore, the moving observer does not view these events as
simultaneous. The notion of simultaneity is thus seen to be an observer-
dependent relationship between events.
(ii) Secondly, we consider the apparent contraction of a moving rod. The
space-time history of a straight rod or string is a world-sheet as in Figure 18.
Consider the inertial observer with respect to whom the rod is at rest. Let the
FIGURE 18. Lorentz-Fitzgerald contraction of length.

3.1. Applications of the Lorentz Transformation 53
r r
edges l , 2 of the world-strip of the rod be given by
r l = {x: Xl = x2 = x3 = 0, X4 = s},
~
r 2 = {A.Al
X. x = 10' xA2 = xA3 = OA4
,X = A)
Sf
The length of the rod, as measured by the inertial observer relative to whom
the rod is moving with constant velocity v along the xl-axis, must be
x(1) = (x (I!) , 0, 0, X(!)),
X(2) = X(l).
By the Lorentz transformation (3.1.15) we have
10 = Xl2) - X(\) = [3[Xl2) - xl!) - V(X(2) - X(!))]

= [3(Xl2 ) - xlI = [31, (3.1.16)
1=lo~.
This effect is known as the Lorentz- Fitzgerald contraction of the moving rod,
since, for v =1= 0, 1< 10 . As the speed Ivl approaches 1, the speed of light, the
moving length I tends to zero!
(iii) We now discuss the time-retardation or time-dilation of a moving clock.
We consider two inertial observers in relative motion with constant velocity
v in the direction of the xl-axis as in Figure 19.
We note two distinct events X(2), x(1) on the standard clock of the relatively
moving observer. The proper time elapsed between these events is
~4
~3
FIGURE 19. Time-Retardation of a moving clock.

The difference in the readings of time for these two events for the stationary
observer is T = xti) - X(l)' The inverse Lorentz transformation of (3.1.15) is
given by
x2 = x2 ,
x3 = x3 ,
X4 = P(vx l + X4),
where P== 1/~v2 and Ivl < 1. So we can obtain
T = P[X(2) - X(l) + v(xl2) - xh]
= P(X(2) - X(!) = PTo
=To/~. (3.1.17)
Thus, for v # 0, To < T, and time goes at a slower rate on the moving clock
according to the judgement of the stationary observer. As the speed Ivl ap-
proaches the speed of light, the moving clock tends to stop!
(iv) Now we discuss the composition of relative velocities. Suppose that,
relative to the first inertial observer, the second one moves with velocity VI
along the xl-axis, so that from (3.1.15) we get
X2 = Xl,
(3.1.18)
X3 = x 3 ,
X4 = - Xl sinh IXI + X4 cosh lXI'
where tanh IXI = VI' Let a third inertial observer move, relative to the second
inertial observer, with velocity V 2 along the same direction. Then we have
(3.1.19)
where tanh IX2 = v2 .

Combining (3.1.18) and (3.1.19) we get
~2 = x2,
~3 = x3,
3.2. The Lorentz Group 24 55
The above transformation is a Lorentz transformation, and the correspond-

ing relative velocity v must be given by
v = tanh (X =tanh(Xl + (X2)
tanh (Xl + tanh (X2 VI + v2
(3.1.20)
1 + tanh (Xl tanh (X2 1 + VI V2
The above equation gives the velocity V, which is the composition of velocities
VI and V2. Suppose that 0 < IVll < 1 and V2 tends to thc~ velocity of light.
Then the limiting composite velocity is
V=

11m VI + V2 = 1.
vz~l_l + V 1 V2
Therefore, the velocity of light cannot be exceeded even by combining many
relative velocities. Many more physical problems will be di:.cussed in Chapter
5 on relativistic mechanics.
EXERCISES 3.1
1. Let Xi = ).ijxj, det[).ij ] #- O. Furthermore, assume that dijXixj = 0 if and
only if dijxix j = o. Then prove that
).ij = J1.l i j'
where J1. #- 0 and [Iij] is a Lorentz matrix.
2. Consider the equation of a null curve
d d.i(t) d.j(t) _ 0
ij -----at -----at - .
Prove that this equation remains invariant under the transformation of coor-
dinates Xi = xi/(dk,XkX'), where x rf. No.
3. Assuming IVII < 1, IV21 < 1, IV31 < I, prove that
I1VI++VlV2
V2 + V3 + VI V2V31
+ V2V3 + V 3 V 1
< 1.
Interpret the above inequality in physical terms.

4. Derive the Lorentz transformation from one inertial observer to another
such that the components ofthe relative three-velocity are V 1 , V 2 , v3
3.2. The Lorentz Group .!4

Let us start with the definition of a group. A group G is a SC!t of elements with
a composition rule which satisfies the following axioms:
G.l. gl 0 gz E G whenever gl, g2 E G (closure).

G.2. gj 0 (gz 0 g3) = (gj 0 g2) 0 g3 (associative).
G.3. There exist an element e E G such that (3.2.1)
eo g = g 0 e = g for all g E G (identity).
G.4. Given agE G there exists g-l E G such
that g 0 g-l = g-l 0 g = e (inverse).
A nonempty subset H ~ G is called a subgroup of G provided that hI 0
h2 -j E H whenever hj, h2 E H. In case commutativity holds, i.e., gl 0 g2 =
g 2 0 g I for all g I' g z E G, then the group G is called abelian. We shall give
some simple examples of groups.
Example 1: The real number system IR is an abelian group under the compo-
sition rule of addition. The number zero is to be identified with the identity
element, and the negative of a number is to be identified with its inverse
element. The number of elements in IR is infinite. The subset 7L c IR forms a
subgroup, which contain a denumerable infinity of elements. 0
Example 2: The set of permutations of integers (1, 2, ... , n) forms a nonabelian

group of permutations Sn called the symmetric group on n elements when
n > 2. It contains n! elements. The subset of even permutations is a subgroup.
o
It has been proved in Theorem (2.4.1) that the transformation from one
Minkowski coordinate system to another is given by a Poincare transforma-
tion PI: [R4 ..... [R4, where
(3.2.2)
where L j = [l(1/j] is a Lorentz matrix. A successive Poincare transformation
P2 can be written as
*i i l i ~j
X = a(2) + (2)jX, (3.2.3)
where L z = [l(2/J is a Lorentz matrix. The composite transformation P2 0 PI
is given by
~ = (P2 0 Pj)(x), ~i = at2) + l(2/jall) + l(2)ijl(1/kXk. (3.2.4)
Theorem (3.2.1): Under the rule of composition P2 0 PI' as given in (3.2.4), the
set of all Poincare transformations &4 == .1"(9(3,1) forms a group.
Proof: The equation (3.2.4) can be rewritten as
(3.2.5)
3.2. The Lorentz Group 24 57
Defining the matrix A == [).i k ], we have
ATDA = (L2Ld TD(L2Ld = L1 T(L2 TDL z )L 1 (3.2.6)

= L1TDL1 = D.
Thus A is a Lorentz matrix and the transformation in (3.2.5) is a Poincare
transformation. This proves the closure axiom G.l in (3.2.1).
The equations (3.2.2)-(3.2.4) can be expressed as matrix equations:
x = a(1) + L 1x,
(3.2.7)
~ = a(2) + Lzx = a{Z) + L 2a(1) + L z L 1x.
Here we have viewed x, X, ~ as column vectors.
We introduce a third Poincare transformation
Computing x# = P3(~) and x# = (P3 0 Pz)(x) we get

x# = a(3) + L 3a(2) + L 3[L za(1) + L z L 1 x],
(3.2.8)
x# = a(3) + L 3a(2) + L3L2[a(1) + L 1x],
which are identical due to the associativity of matrix multiplication. Thus
P3 0 (P2 0 P1) = (P3 0 Pz ) 0 P1 and the axiom G.2 of(3.2.1) holds.
The identity matrix I is a Lorentz matrix since ITDI = D. We choose the
identity Poincare transformation P1 = I by putting a/l) == 0 and 1(1)ij = b ij
Then equations (3.2.2), (3.2.3) reduce to
~i = a/2) + 1(2/jx j = ab + 1(2/jx j.

The above proves Pz 0 1= Pz and similarly I 0 Pz = Pz' Thus the axiom G.3
of (3.2.1) holds.
In equation (1.2.11) it was proved that detL = 1 for any Lorentz matrix.
Thus every Lorentz matrix has an inverse; furthermore, L -1 is a Lorentz
matrix (see problem 2 of Exercise 1.2). Recalling the Poincare transforma-
tions in (3.2.2)-(3.2.4), we define the inverse Poincare transformation P2 ==
p;-1 by putting L2 == Li 1, and a 2 == - Li 1a 1.
Thus equation (3.2.4) reduces to ~i = Xi. Therefore, p;-1 0 P1 = I and simi-
larly P1 0 p i l = I. This shows the validity of G.4 in (3.2.1).
The group &4 of Poincare transformations has many subgroups. The

translation subgroup associated with the set of Poincare transformations is
characterized by 1(1/j = b ij For such a transformation
(3.2.9)
For a second successive translation,

=
i i i
X a(Z) + i,
and the composite transformation is
*i
X = a(l) + a(Z) + x .
i i i
Since ail) + ab = aiz) + ail)' the two translations commute. Thus the sub-
group of translations is an abelian subgroup of &4. In the case that the
translation parameters a i == 0, the Poincare transformation reduces to the
Lorentz transformation ii = lijX i , where L = [lij] is a Lorentz matrix. It is
easy to prove that the set of all Lorentz transformations forms a subgroup of
the Poincare group &>4. This subgroup is called the general Lorentz group
~ == 0(3,1). The general Lorentz group 24 has two important subsets. The
orthochronous subset of ~ (characterized by 14 4 ~ 1) is denoted by 2/. The
proper subset of ~ with det[lij ] = 1 is denoted 2 4+. Furthermore the proper
orthochronous subset of ~ with 14 4 ~ 1 and det[/ij ] = 1 is denoted by 24~.
Theorem (3.2.2): (i) The subsets 24+ , 24+ are subgroups of 2 4 .

(ii) 24~ is a subgroup of both 2/ and 24+ .
Proof: (i) From the (4,4)-entries of the matrix equations

LIL21 == LIAz = L,
L!DLJ = D = AIDA z ,
we obtain
l(1j!a(z\4 + l(1jza(z~4 + l(1j3 ad4 + /(1j4a(Zj4 = 144,
(l(ljd + (l(ljzf + (l(1j3f - (l(1~4)2 = -1,
(ad)4)Z + (a(Z~4)Z + (a(Z~4)2 - (a(zj4)Z = -l.
Using the Schwarz inequality (1.1.13) and l(1j4 ~ 1, a(2j4 ~ 1 (see problem
4 of Exercise 1.2), we get
-J[(l(1jl)Z + (l(1jz)Z + (l(1j3)Z] [(a(2\4)2 + (a(z~4f + (a(Z~4)Z]
~ l(1j! a(Z\4 + l(1jza(z~4 + l(1j3a(z~4'
or equivalently
o ~ II(1j41Ia(2j41 [1 - J(1--=- II(1j41 z)(1 -la(zj41 Z)]
~ l(1j! a(Z\4 + l(1jza(z~4 + {(lj3a(Z~4 + {(1j4a(Zj4 = {44
Therefore, {44 ~ 1 and L! 0 L"2! E 24+. Thus 24+ is a subgroup. The subset
24+ c 24 is characterized by det L = 1. Suppose that L I' L z E L 4 +. Then
detL I L 21 = detL!/detLz = 1. Thus L! 0 L2! E 24+ and 24+ is a subgroup
of 2 4 .
(ii) The straightforward proof is ommited.
3.3. Real Representations of the Lorentz Group 24 59
The subset of improper Lorentz transformations is characterized by

det L = - 1. Examples of the improper Lorentz transformation are the
space reflection P (1.2.14) and the time reversal T (1.2.15). The subset of im-
proper Lorentz transformations is not a subgroup.
EXERCISES 3.2
1. Prove that the subset of special Lorentz transformations in (3.1.15), form
an abelian subgroup of 2 4++,
2. Consider the subset of Lorentz transformations for which 1\ :-:; -1.
Determine whether or not this subset is a subgroup.
3. Let L + E 24+ . Prove that L + E 24++' or L + = PoL!, where P denotes
the space reflection (1.2.14) and L! is an element of 2 4++,
3.3. Real Representations of the Lorentz Group 24

In this section we shall define real matrix representations of a group G and
apply this theory to the Lorentz group 2 4 , Suppose that the set of n x n
matrices with real entries is denoted by Mnn (note that there is no sum on n).
A mapping p: G -+ Mnn is called a real matrix representation of G if
pee) = I,
(3.3.1)
P(gl 0 gz) = p(gdp(gz),
for all gl' gz E G. The real vector space of column vectors v" == M.l is called
the n-dimensional space of representations p of G. The integer n is called the
degree of the representation.
We shall be interested in real representations of the Lorentz group 2 4 ,
The simplest representation is the one-dimensional scalar representation.
The next simplest is the four-dimensional self-representation f: 24 -+ M 44 ,
where .~(L) == L. For more general representations we have to introduce the
concept of the Kronecker product (also called the direct product) of two
matrices A, B of sizes m 1 x n 1 , m2 x n 2 respectively. The Kronecker product
C = A x B, which is of size m l m2 x n 1 n2' is defined to be
[aUR
a B
21
a 12 B
azzB
a,.,R
a .,B
2
J
C= .
am,l B am ,2 B am,.,B
[ a"b"
a ll b12
a,.,b,., J
= a .b ll 21 all b22 a 1 ., b2nz
(3.3.2)
am,l bm z l a m ,l bmz2 am,.,:bmz z

We will show some elementary examples now.
Example 1:
-5 -4
[~ -lJ
n x [4 - 5] = [48 -10 4n D
Example 2:
a 1 b1
a 1 b2
[::J G:J ~
x
a 1 b3
a2 b l
a 2 b2
D
a 2b3
The main theorem regarding the computations of Kronecker products will
be stated below.
Theorem (3.3.1): Let A, B, C be matrices of sizes m 1 x n 1 , m 2 x n 2, m3 x n3

respectively. Then the following equations must hold:
(i) A x (B x C) = (A x B) x C,
(ii) (A x B)T = AT X B T,
(3.3.3)
(iii) If A, B have sizes n 1 x m 1 , n 2 x m 2 respectively then
(A x B) (A x B) = (A A) x (BB).
Proof: The proof is left as problem 1 of Exercises 3.3.
Suppose that we have a Minkowski tensor field

tk, ... kr(x) =vtMx)vtl)(X) vt,.')(x),
where v~;ix) are vector fields. The transformation rule (2.4.11) of this tensor
at a special point Xo is
(3.3.4)
where the li k are the entries of a Lorentz matrix L, the hatted quantities are
evaluated at xo , and the unhatted quantities are evaluated at Xo. Defining a
4 x 1 column matrix
vti)(XO)l
vei)
_
= r vt)(xo)
3 (
VCi) Xo
) ,
v~)(xo)
and using equations (3.3.2), (3.3.3) we can rewrite (3.3.4) in matrix language
3.3. Real Representations of the Lorentz Group 24 61
as:
(3.3.5)
We define A == Lx' x L, and it is a square matrix of size 4' x 4'. We
also define W == v(l) x ... X v(')' and it is a column matrix of size 4' x 1. So
the equation (3.3.5) can be expressed as
W=AW. (3.3.6)
Theorem (3.3.2): The mapping p: .'4 --. M 4 '4" defined by A =" peL), is a matrix
representation of degree 4' of the general Lorentz group .'4'
Proof: Let the identity matrix of size 4' x 4' be denoted by I(T) so that
I(T) = I x ... x I (r factors), where I is the 4 x 4 identity matrix. The matrix I
is the self-representation of the identity I E .'4' We define 1(,) == p(I), A ==
peL). Using Theorem (3.3.1), we obtain
p(L I )p(L 2) = AIA2 = (LI X ... X Ld(L2 X ... x L 2)
= (LI L 2) x ... X (LI L 2 ) = P(LI L 2 }.
Thus by (3.3.1), p is a matrix representation of .'4 and the corresponding
degree is 4'.
We can conclude that the set of (r + O)-order Minkowski tensor fields at an

event is isomorphic to the 4'-dimensional space of representation of the gen-
eral Lorentz group .'4'
Let us go back to the matrix equation (3.3.6), viz. W = AW. Let the set of
all column vectors of the size 4' x 1 be denoted 1fI. It may so happen that,
for a subspace 1fI' c 1fI, the column vector Aw' E 1fI' for every w' E 1fI'. Such
a subspace 1fI' is called an invariant subspace of the particular linear mapping
corresponding to A. In the case 1fI' is invariant under every A, we call1fl' an
invariant subspace of the matrix representation of the group .'4' The matrix
representation is called irreducible in case it does not have any invariant
subspace other that 1fI or {O}.lfthe matrix representation has two nontrivial
invariant subspaces 1fI1, 1fI" such that 1fI' u 1fI" = 1fI, 1fI' n 1fI" = {O}, and
dim 1fI' > 0, dimir" > 0, the representation is called completely reducible.
For a completely reducible representation every matrix A ,= peL) assumes a
block-diagonal form:
(3.3.7)
Example: Consider a (2 + O)-order Minkowski tensor field rab(x) and its

transformation properties (2.4.11) at an event Xo. Thus
(3.3.8)
The matrix representation A inherent III the above equation IS sixteen-

dimensional. But
smn == [rmn(xo) + rnm(xo)]/2 = snm,

(3.3.9)
a mn == [rmn(xo) _ r nm (xo)]/2 = _ a nm ,
It can easily be proved that sij = ji, aij = _a ji . Therefore, the symmetric
tensors are transformed into symmetric tensors and antisymmetric tensors
are transformed into antisymmetric tensors. The only tensor that is both
symmetric and antisymmetric is the zero tensor. Furthermore, there are ten
and six independent components of smn and a mn respectively. Thus the dimen-
sion of the vector space associated with the symmetric tensors is ten and the
dimension of the vector space associated with the anti symmetric tensors is
six. So the 16 x 16 matrix representation A is completely reducible. To ob-
tain this reduction explicitly we label the tensor components as
(3.3.10)
The transformation equation (3.3.8) is equivalent to the matrix equation
(3.3.11)
where Wi and Wi are 10 x 1 column vectors, w" and w" are 6 x column
vectors, and A' and A" are given by large matrices, which have been left as an
exercise.
The matrix representation AU of the Lorentz group is irreducible, whereas
the representation A' is completely reducible. The reason behind the reduci-
bility of A' is that for a symmetric tensor sij one has
sij = s~ + d ii (S/4),
(3.3.12)
Under a Lorentz transformation, s = s, s!j = li)jbsob = s6i , djjs!j == O. There-

fore, the one-dimensional subspace of the scalar s and the nine-dimensional
subspace of the trace-free symmetric tensor s!j are invariant subspaces of the
3.4. The Lie Group :f!4++ 63
matrix representation of the Lorentz group 2'4' Thus
where A' is a 10 x 10 matrix, A'l is a 9 x 9 matrix, and A'l' A~ are irreducible

representations. 0
Higher-order tensor fields may exhibit intermediate symmetries called

Young symmetries, which are between total symmetry and total antisymme-
try. These intermediate symmetries also produce irreducible representations.
EXERCISES 3.3
1. (i) Prove the Theorem (3.3.1).
(ii) Let A and B be two square matrices of sizes m x m, n x n respectively.
Then prove that trace(A x B) = trace A . trace B.
2. Let A == Lx' .. x L, !?2J == D x ... x D, where there are r factors in each
of the definitions and L is a Lorentz matrix. Show that A T!?2J A = g.
3. Let Tijk(X) be an arbitrary Minkowski (3 + O)-order tensor field with
sixty-four components. One can write the identity at any point Xo
Sijk == [Tijk + Tjki + Tkij + Tikj + Tjik + T kj i]/6,
Aijk == [Tijk + Tjki + Tkij _ Tikj _ Tjik _ T kj i]/6,
yijk == [Tijk _ Tjki + Tjik _ T kj i]/3,
Y'ijk == [Tijk + Tkji _ Tjik _ T ki j]/3.
From the above identities prove that the sixty-four-dimensional represen-

tation of the Lorentz group decomposes into the sum of three twenty-
dimensional representations and one four-dimensional representation of the
Lorentz group.
4. Explicitly find the matrices A' and A" in the block decomposition
(3.1.11 ).
Let us obtain the Lorentz matrix of an arbitrary proper orthochronous

Lorentz mapping in terms of the six canonical parameters. In case of a three-
dimensional rotation, the rotation matrix can be conveniently expressed in
terms of three Eulerian angles. In this scheme one orthonormal triad is ro-
tated into another in three steps, each step being a rotation about an axis.
In expressing a proper orthochronous Lorentz matrix, we shall carry out a

scheme closely analogous to the Eulerian scheme, using six steps instead of
three. We first choose an orthonormal tetrad {e l , e 2 , e 3 , e4 }, satisfying equa-
tion (1.1.7). Changing the notation, we shall write
i i = j-j = k k = -1'1 = 1, (3.4.1 )

i .j = i . k = i . I = j . k = j . I = k . I = O.
The equation (1.2.16) can be expressed as
1= i cos 8 + j sin 8
j = - i sin 8 + j cos 8,
(3.4.2)
k= k,
i= I,
where - n < 8 < n.
The above proper orthochronous Lorentz mapping can be called a space-
like rotation by an angle 8 of the 2-lIat spanned by {i,j} about the fixed 2-lIat
spanned by {k, I}. Another proper orthochronous Lorentz mapping can be
given as
j = j,
(3.4.3)
k = k cosh 0: + I sinh 0:,
i = k sinhct. + I cosh ct.,
where 0: E IJt
The above Lorentz mapping can be called a timelike pseudorotation by
the pseudoangle 0: of the 2-lIat spanned by {k, I} about the fixed 2-flat
spanned by {i,j}.
The six steps of Lorentz mapping from the basis {i,j, k, I} to {i,J, k, i} can
be schematically exhibited by
{i,j,k,l} ~ {il,L,kl,ld ~ {i 2 ,h,k 2 ,1 2 } ~ {i 3 ,h,k 3 ,1 3 }
9, {' . k I} 9, {' k I} 93 {~ ~ kAAI}
-+ 14.14' 4, 4 -+ 15.15' 5, 5 -+ I,J, , .
We shall employ the following abbreviations in the sequel:

Sp == sinh O:p'
for p = 1,2,3. The six mappings can be expressed as
i l = C1 i + Sl l,
j1 = j, i2 = C2i1 + S2 11' h=h,
k1 = k, k2 = k l , k3 = C3 k2 + S3 12'
11 = Sl i + C11, 12 = S2L + C2 11 , 13 = S3 k 2 + C312'
3.4. The Lie Group 24++ 65
i4 = + Slj3'
c 1i3 is = c Zi4 - S2 k 4' i = C3iS + S3jS'
j4 = -Sl i 3 + cd3' is = j4, j = -s]is + c 3 js,
k4 = k 3, ks = szi4 + c Z k 4 , k = k s,
14 = 13 , Is = 14 , i = Is. (3.4.4)
The above six mappings can be combined into
i= [(C I C 2 C 3 - SlS3)C1 + (SlCZC3 + C I S 3 )S I SZ - S2C3S1 C 1 S 3 ]i
+ [(SIC1C3 + C I S 3 )Cl - SlC3S1S3]j - SZ C3 C3 k

+ [(e I CZC3 - SIS3)Sl + (S I CZC3 + C 1 S3)C1 S Z - SZC 3 C 1 Cz S 3 ]1,
j = [-(C I C2 S 3 + Sl C3)CI + (e I C3 - SICZS3)SIS2 + SlS3 S1C2S3]i
(3.4.5)
+ [(C I C3 - S I C 2 S 3 )CZ + SlS3SZS3]j + SZ S3 C 3 k
+ [-(C 1 CZS 3 + SI C3)Sl + (C 1 C 3 - SICZS3)ClS1 + SZ S 3 C l C zS 3 ]1,
k = [C 1 S 2 C 1 + SISZSISZ + CZS 1 C z S 3 ]i + [SISZCZ + CZS2S3]i
+ c z C 3 k + [C 1 S 2 S 1 + SI S Z C l SZ + C Z C 1 Cz S3 ]1,
i = S1CZC3i + SzC3 j + S3k + C 1 C1 C3 1.
Using equations (1.2.2) and (3.4.5) one can obtain the entries of the corre-
sponding Lorentz matrix L = [Iij] as
1\ = (C I C2 C3 - + (SIC2C3 + C1S3)SIS2 - S2C3S1C2S3'
SIS3)C1
[12 = -(C I CZS 3 + SIC3)CI + (C I C3 - SIC2S3)SISZ + SZS3S1 C2 S 3 ,

[13 = C 1 S 2 C 1 + SIS2SISl + C2 S 1 C1 S 3 ,
[1 4 =SI C2 C 3'
[11 = (SIC1C3 + C I S 3 )C2 - SlC 3 S 2 S 3 ,
[22 = (C 1 C3 - SIC2S3)Cl + S2S3S2S3'

[23 = S1 S2 CZ + CZSZ S3 ,
[24 = S2 C3'
(3.4.6)
[3 1 = -SZC 3 C 3 ,
[3 Z = SZ S 3 C 3,
[3 3 =C 1 C3 ,
[3 4 = S3,
[41 = (C 1 C 1 C 3 - + (SIC1C3 + C1S3)C1S1 - SlC 3 C 1 C 1 S 3 ,

SIS3)Sl
[\ = -(C I C1 S 3 + SI C3)SI + (C 1 C 3 - SI C1 S 3)C 1 S Z + SZ S 3 C I C 2 S3'
[43 = C 1 S Z S1 + S1S2ClS1 + C 2 C I C2 S3 ,
[44 = C I C2 C3 ;
rxp E IR, - n < 81 < n, 0 < 82 < n, - n < 83 < n.
The above equations provide a convenient parametrization of an arbitrary

proper orthochronous Lorentz matrix, in terms of six parameters r:t. p ' ()p.
Let us consider the domain of the six parameters
r:t.p E 1R 3 , -n < ()1 < n, 0 < ()2 < n, -n < ()3 < n}. (3.4.7)
The domain D# c 1R6 may be considered as the image X#(U#), U# c .A6 , a

six-dimensional differentiable manifold. But all possible proper orthochro-
nous Lorentz matrices do not correspond to the points contained in D#. To
accomplish that we first note that D# = 1R3 x Dt, where Dt == {(()1' ()2' ()3):
-n < 01 < n, 0 < O2 < n, - n < 03 < n}. The domain Dt is shown in Figure
20.
We add boundary points corresponding to ()2 = 0 and ()2 = n to Dt. Then
we add boundary points 0 1 = n, 0 1 = -n and topologically identify these
sets of points. Similarly we add points ()3 = n, 03 = -n and identify. The
resulting region, denoted by lit, is homeomorphic to the projective space 1P 3 .
The differentiable manifold .A6 = X- 1 [1R3 x lit] is called the group manifold
for !f>4++. Such a group, which is associated with a group manifold, gives rise
to a Lie group. In the present context, we shall define a Lie group of coordi-
nate transformations. Consider a differentiable coordinate transformation
f: D x D# C IRn x IR r -+ IR" given by:
(i)
(3.4.8)
FIGURE 20. The three-dimensional domain Dt.

Here x = (Xl, ... , x") E Dc jR", rx = (rxl, ... , rx') E D# c jR', such that the func-
tions p are real analytic functions of the variables rx l , ... , rx . Recall that a real
analytic function F allows a power series expansion of F(rx l , .. . , rx.) in a con-
vex domain D#.
(ii) The Composition rule of two successive transformations is
{i = Pcxl, ... , x"; f31'" ., f3.)
= fi[f I (x; rx), ... ,f"(x; rx); f31" .. , f3.J
where real analytic functions t/Jp. must exist such that

Yp. = t/Jp. (rx I , ... , rx.; f31"'" f3.) == t/Jirx; f3). (3.4.9)
(iii) The associativity of three successive transformations characterized by
parameters rx, f3, Y imply that
(3.4.1 0)
(iv) For the identity transformation there must exist parameters (rx~, ... , rx~)
E D# such that
(3.4.11)
(v) For the inverse transformation there must exist parameters (rxl , ... , rx.-)
E D# such that
(3.4.12)
Example: Consider a rotation in jR2 given by the coordinate transformation

Xl = fl(X;rx) == Xl cosrx - x 2 sinrx,
x2 = f2(X; rx) == Xl sin rx + x 2 cos rx,
where rx E D# == {rx: -TC < rx < TC} and XED == jR2.
To obtain every transformation of the rotation group, the end points rx =
TC have to be added to D# and then the end points are identified. In that
case jj# and Al are homeomorphic to the unit circle Sl. The Jacobian of the
transformation is
O(xl,x 2 )/O(xl,x 2 ) = 1.
Since cos rx and sin rx are real analytic functions of rx E D #, it can be concluded
that fi(X; rx) are real analytic functions of rx. The composition of two transfor-
mations is
~l = Xl cos f3 - x2 sin P =+ rx) - x 2 sin(f3 + rx),
Xl cos(f3
~2 = Xl sin f3 + x2 cos f3 = Xl sin(p + rx) + x 2 cos(P + rx);
thus, t/J(rx; f3) == rx + f3. More precisely t/J(rx; f3) == (f3 + rx)mod 2TC - TC so that
- TC ::; t/J(rx; f3) ::; TC. Note that t/J(rx; f3) is a real analytic function of rx, P in a
neighborhood of the origin. The associativity can be checked by noticing

that
;(;(rt.;[3);y) = (y + (f3 + rt.mod2n - n
= y + [3) + rt.)mod2n - n
= ;(rt.; r/J([3; y.
The identity transformation corresponds to rt. 0 == O. The inverse transforma-
tion corresponds to rt. - == - rt., so that
We are ready to state a theorem on the structure of the Lie group 2 4++,
Theorem (3.4.1): The set 24++ of all proper orthochronous Lorentz transforma-
tions of coordinates on ~4 given by the matrix (3.4.6) constitute a six-parameter
Lie group.
Proof: We shall give a partial proof of this theorem. Consider the proper
orthochronous Lorentz transformation
where the entries are explicitly given in equation (3.4.6). Clearly we have
O(AI A2 A3 A4)
X ,X ,X ,X = det L = 1
o(x!, x 2 , x 3 , x 4 )
for all x = (x 1, x 2 , x 3 , x 4 ) E ~4. The entries lij are the values of the functions
gij of the parameters in (3.4.6) so that
lij = giirt.!, rt.2, rt. 3 ,(}1' 82 , ( 3 ),
where (rt. 1 ,rt. 2 , rt. 3 , 81 , 82 , ( 3 ) E D#. The domain D# and the region 15# have
been determined in equation (3.4.7) and the subsequent discussions.
Consider, for example, the entry
13 3 = g\(rt. 1 , rt. 2 , rt. 3 , 81 ,82 , ( 3 ) == cos 82 cosh rt.3'
Since the expansions
(8 2)2 (82)4
cos 82 = 1- T + 4! - ... ,
(rt. 3 )2 (rt. 3 )4
cosh rt.) = 1+T + 4! + ...
are absolutely convergent in D#, we can write the product
(82 )2 (rt.3)2 (8 2 )4 (rt.3)4 (82 rt.3)2
cos 82 cosh rt.3 = 1 - T +T + 4! + 4! - (2!)2 - ...
3.4. The Lie Group 24++ 69
as an absolutely convergent series in D#. Therefore, the function g33 is a real

analytic function ofthe six parameters. Similarly all of the gi j are real analytic
functions in D#.
The composition of two Lorentz transformations can be written as
(3.4.13)
Considering a special case of (3.4. t 3) and using (3.4.6), we can obtain
p = [Jx l + [Jx 2 + ["'Jx 3
= -(sin "'2 cos "'3 coshY3)x l + (sin "'2 sin "'3 coshY3)x Z
+ (cos "'2 cosh Y3)X 3 + (sinh Y3)X 4.
Comparing the coefficients of X4 in the above equation we have
So we find by inversion of sinh that
== Argsinh [ - sin e~ cos e ~ cosh a~ sinh al cosh a2 cosh a3

+ sin e~ sin e~ cosh a; sinh az cosh a3
+ cos e; cosh a; sinh a3
+ sinh a~ cosh al cosh a2cosh a3].
The above function tP3 is real analytic in the twelve variables. Similarly the
other five real analytic functions
YI = tPI(al,a2,a3,el>e2,e3;a~,a;,a;,e~,e2,e3)'
Y2 = tP2(al,a2,a3,el,e2,e3;a~,a;,a3,e~,e2,e3)'
"'I = tP4(al,a2,a3,el,e2,e3;a~,a;,a3,e~,e2,tl;),
can be found. These functions must satisfy the associative property (3.4.10) by
Theorem (3.2.2).
The identity transformation corresponds to parameters
(a?, a~, ag, er, e~, ag) = (0,0,0,0,0,0).
The inverse of a transformation corresponds to parameters that are the solu-

tions of six equations. One of these six equations is
or equivalently
- sin 8zcos 8; cosh IX~ sinh IX\ cosh IX2 cosh IX3
+ sin 8z sin 8~ cosh IX3 sinh IX2 cosh IX3
+ cos 8zcosh IX3 sinh IX3 + sinh IX3 cosh IX\ cosh IX2 cosh ()(3 = o.
EXERCISES 3.4
1. Consider the set of projective transformations given by
A
X=---
ax +b
ex + d'
where ad - be =I 0 and x E IR. Prove that this set of transformations consti-
tutes a Lie group.
2. Consider a subset of 24++' characterized by
Xl = xlcos8 - x 2 sin 8,
x2 = Xl sin 8 + x 2 cos 8,
x = x cosh ()( - sinh ()(,
3 3 X4
X4 = - x 3 sinh ()( + X4 cosh ()(.

Prove explicitly that this subset is an abelian Lie subgroup and its group
submanifold is homeomorphic to Sl x IR.
3. Consider the one-dimensional wave equation
fPr/J 02r/J
ox 2 (x, t) - ot 2 (x, t) = O.
The characteristic or null coordinates (or light-cone coordinates) are given by
u = x - t, v=x + t.
(i) Show that the wave equation reduces to the form
o2
ouov (u, v) = 0,
where (u, v) == r/J(x, t).
(ii) Prove that the wave equation remains invariant under a coordinate
transformation
(j = e -(a2)3 v,
(iii) Prove that the set of transformations in part (ii) constitutes a Lie
group.
References
1. Y. Choquet-Bruhat, C. De Witt-Morette, and M. Dillard-Bleick, Analysis, mani-
folds and, physics, North-Holland, Amsterdam, 1977. [po 96]
2. A. Einstein, Ann. Physik 17 (1905), 891-921. [po 69]
3. I. M. Gelfand, R. A. Minlos, and Z. Y. Shapiro, Representations of the rotation and
Lorentz groups, and their applications, The MacMillan Co., New York, 1963. [po 81]
4. M. Hammermesh, Group theory, Addison-Wesley, MA, 1962. [pp. 84,97,89]
[pp. 91, 93]
4
Pauli Matrices, Spinors, Dirac
Matrices, and Dirac Bispinors
4.1. Pauli Matrices, Rotations, and

Lorentz Transformations
A rotation in the Euclidean plane can be characterized by the following
matrix equation:
~1] = [ co~ f/J sin rp] [Xl],
[ X2 - Sill rp cos f/J X2
where rp E ( -n, n) and Xl, x 2 are Cartesian coordinates. Introducing complex
variables z == Xl + ix 2 , i == Xl + ix 2 , the above transformation becomes the
conformal transformation in the complex z-plane,
i = uz, (4.1.1)
where u == e- i , so that uti = 1. Thus u is unimodular. Here we have used a bar
to denote complex conjugation.
A general rotation in three-dimensional Euclidean space can be character-
ized by two systems of Cartesian coordinates (X 1,X 2 ,X 3 ), (XI1,X'2,X'3), and
Euler angles e, f/J, l{I. This rotation can be obtained as a composition of three
rotations: the rotation rP about the x 3 -axis, the rotation e about the new
xl-axis, and the rotation l{I about the new .*3- axis. Thus the composite rota-
tion is given by the matrix equation
[ :::]=[~:~!l{I ;~:~ ~][~ CO~8 Si~8][~:::f/J ;~:~ ~][::].

X'3 0 0 1 0 -sine cos8 0 0 1 X3
(4.1.2a)
We can explicitly carry out the matrix multiplication to get the composite
matrix equation
Xd] [ COS IjJ cos <fJ-cos II sin <fJsin 1jJ, cos IjJ sin <fJ + cos II cos <fJ sin IjJ sin IjJ sin II ][Xl]
[ X'2 = -sin IjJ cos <fJ-cos II sin <fJcos 1jJ, - sin IjJ sin <fJ + cos II cos <p cos IjJ cos IjJ sin II X2
X'3 sin II sin <p -sin IIcos<fJ cosO X3
(4.1.2b)
72
4.1. Pauli Matrices, Rotations, and Lorentz Transformations 73
FIGURE 21. Stereographic projection of S2 onto 1E 2
A three-dimensional rotation can correspond to a complex conformal

transformation by the following device. Project S2, the surface of the unit
sphere given by (XI)2 + (X 2)2 + (X 3)2 = I, from the point (0,0, \) onto the
Euclidean plane 1E2 given by x 3 = O. This projection is called a stereographic
projection. Let (x,y) be the projection of(xl,x 2 ,x 3 ) as in Figure 21.
By the similarity of triangles, it can be shown that
Xl/X = x 2 /y = (1 - x 3 )/1. (4.1.3)
One can introduce a complex variable
(4.1.4)
A rotation of angle about the x 3 -axis can be expressed as the complex
conformal transformation
( = (Xl + ix 2 )/(l - x3 ) = (Xl + ix 2 )/(l - x3)
= e-i~(xl + ix 2)/(l - x 3)
= e-i~, = (e-i~!2, + 0)/(0 + ei~!2). (4.1.5)

To treat a rotation by an angle e about the xl-axis, we can introduce the
complex transformation
fj == (x 2 + ix 3 )/(1 - Xl)
= e- i9 (x 2 + ix 3 )/(1 - Xl)
(4.1.6)
We see now that
(I} + i)/(fj - i) = [ _(Xl + ix 2 ) + (I + x3 )]/[(X I - ix 2 ) - (1 - x3 )]
= [-((1 - x3 ) + (I + x3)]/[()-I(l + x3 ) - (I - x3 )]
= e, (4.1.7)
74 4. Pauli Matrices, Spinors, Dirac Matrices, and Dirac Bispinors
or equivalently we have ~ = iCe + 1)1(' - 1). Therefore, (4.1.6) yields (e + 1)/

(e - 1) = e- i8 ( + 1)/( - 1). Hence,
e= [cos(8/2)( - i sin (8/2)]/[ - i sin(8/2)( + cos(8/2)]. (4.1.8)
Both of the transformations (4.1.5) and (4.1.8) belong to the subset of Mo-
bius transformations (also called fractional linear conformal transformations).
These special Mobius transformations are characterized by the equation
, = (a( + fJ)/(y( + 15), (4.1.9)
where aJ - {Jy = 1, lal 2 + 1'112 = 1{JI 2 + I(W = 1, and a.{J + yJ = 0. With a
Mobius transformation (4.1.9), one can associate a 2 x 2 complex unimodular
unitary matrix
(4.1.10)
such that det U = 1, and uut = UtU = I where the dagger denotes the her-
mitian conjugation.
The two-dimensional complex vector space V2 , for which the matrix U
represents a linear mapping, is called the spinor space. We shall now intro-
duce three remarkable 2 x 2 matrices called the Pauli matrices. The Pauli
matrices are defined by
Theorem (4.1.1): The three Pauli matrices (1a satisfy

(i) (1:= (1a'
(ii) det (1a = - 1,
(iii) Trace (1a = 0,
(4.1.12)
(iv) (1a(1p = JapI + i11aPy(1y,
(v) (1a(1p + (1p(1a = 2bap I,
(vi) (1a(1p - (1p(1" = 2i11"py(1y,
where we let the Greek indices take values {1,2,3} and the totally antisym-
metric pseudotensor 11"py in Cartesian coordinates is defined analogously to
(1.3. 13b).
Proof: The proof of this theorem follows from direct computations, so we

will just prove (iv). (Note that parts (v) and (vi) follow directly from part (iv).)
To prove part (iv) we notice that
(11(11
= I = I + = bill + i1111y(1y,
(11(12 = [~ ~J = i(13 = b12 1 + i1112y(1y'

The other equations can be proved similarly.
The next theorem about the Pauli matrices is instructive.
Theorem (4.1.2): The set of Pauli matrices {O"l' 0"2' 0"3} is a basis for the vector
subspace of 2 x 2 complex matrices with zero trace.
.
Proo}: Let M = [}ov J.1.] en 'A + T = 0,so M =
T such that Trace M = O. Th
[ ).v J.1..] = [(J.1. + v)/2]0"1 + [i(J.1. - v)/2]0"2 + A0"3' This equation proves
-A
that the set {O" 1,0"2,0"3} spans the vector space of complex traceless 2 x 2
matrices. To investigate linear independence, we examine the matrix equa-
tion
This equation has only 0(1 = 0(2 = ::l3 = 0 as solution. Thus the set {0"1, 0"2' 0"3}
is linearly independent and, hence, is a basis.
Every 2 x 2 hermitian matrix with zero trace is a linear combination of O"a

with real coefficients. Thus a point (Xl, x 2 , x 3 ) in three-dimensional Euclidean
space can be associated with a 2 x 2 hermitian matrix P with zero trace in
the following manner:
Consider a 2 x 2 unitary matrix U and a similarity transformation
(4.1.14)
Since Trace P = Trace U PU- l = Trace UU- l P = Trace P = 0 and also pt =

(UPUt)t = UPU t = P, we can write
(4.1.15)
for some real numbers xa. Thus det P = det U det Pldet U = det P. There-
fore, we must have (Xl)2 + (X 2 )2 + (X 3 )2 = (Xl)2 + (X 2 )2 + (X 3 )2 under the
transformation (4.1.14). So we see that the transformation (4.1.14) induces an
orthogonal transformation (which can be proved to be a rotation) in the
three-dimensional Euclidean space.
Example 1: Let us choose the unitary unimodular matrix

where ; E (-n, n). The transformation (4.1.14) yields P = VIPvt, so

Xl -
_x 3
iX2] = [ei~/2
0
0] [
e-'<I/ 2 Xl
X3
+ iX2
Xl - iX2]
_X3
[e -i/2
0
X3 ei~(xi - iX2)]
[ (4.1.16)
= e-i(x i + ix 2 ) _x 3 .
The above equation implies that

Xl = + x 2 sin ;,
Xl cos ;
x2 = _Xl sin; + x 2 cos;,
x3 = x 3
This is a rotation by an angle ; about the x 3 -axis. The unitary matrix VI
corresponds to the transformation (4.1.5). 0
Example 2: Consider the unitary unimodular matrix

=[cos(e/2) iSin(e/2)]
V2 - ,
i sin(e/2) cos(e/2)
where e E (0, n). This is the matrix in equation (4.1.10) corresponding to the
transformation (4.1.8). The matrix equation = V2 PV; implies that P
.:P = x 2 cos e + x 3 sin e,

i3 = - x 2 sin e + x 3 cos e.
The above equation yields a rotation about the xl-axis. 0
Three successive rotations by Euler angles ;. e, t/J as in (4.1.2) can be
accomplished by the following unitary unimodular transformations
P~ = VIPV It , P'" ~t
= V 2 PV2, P' "'t
= V 3 PV t
3 = VPV, (4.1.17a)
where V is given by
V == V3 V 2 VI
o ] [cOS(e/2) i Sin(e/2)] [e i /2
e- iofr / 2 i sin(e/2) cos(e/2) 0
= [<ofr+)/2 c~s(B/2) ie-.i(<I- ofr)/2 Sin(e/2)].

(4.1.17b)
ie'(ofr-~)/2 sm(B/2) e-'(ofr+(fJ)/2 cos(B/2)
Note that another unitary unimodular matrix U' == - U induces exactly the
same rotation. In general, the same rotation in three dimensions is induced
by two unitary unimodular matrices V, which represent transformations in
spinor space.
We shall generalize these techniques for the Lorentz transformation by

adding to the set of three Pauli matrices in (4.1.11) a fourth matrix, namely,
0'4 == [~ ~] = I = -0'4' (4.1.18)
Theorem (4.1.3): The set of four matrices {0'1,0'2,0'3,0'4} is a basis for the
vector space of complex 2 x 2 matrices. Furthermore every 2 x 2 hermitian
matrix is a linear combination of {0'1,0'2, 0'3, 0'4} with real coefficients.
Proof: For an arbitrary 2 x 2 complex matrix
[
). jJ.] _ 0'1(jJ. + v) i0'2(jJ. - v) 0'3(,1, - r) a4.(). + r)
vr- 2 + 2 + 2 + 2'
This equation shows that {0't,0'2,0'3,0'4} spans the space of 2 x 2 complex

matrices. For linear independence of {O'l, 0'2, 0'3, 0'4} we investigate the matrix
equation
The only solution of the above equation is 1J(1 = 1J(2 = 1J(3 = 1J(4 = 0, which
proves the linear independence of {O'l, 0'2,0'3,0'4}. Thus {ITt, 0'2,0'3,0'4} is a
basis for the vector space of 2 x 2 complex matrices.
For a hermitian matrix H = Ht we see the matrix of H has the form
H = [p ~J, where a, c E R From this we get
[
a P] = O'I(P + fi) i0'2(f3 - fi) 0'3(a - c) 0'4(a + c)
pc- 2 + 2 + 2 + 2'
The coefficients on the right-hand side are obviously real.
The Minkowski coordinates (Xt,X 2,X 3,X4 ) of an event in space-time can

be associated with a 2 x 2 hermitian matrix X in the following way
X3 - X4 Xl - ix 2 ]
X == XkO'k = [
X
1
+ IX,
. 2
-(x
3
+ x4
)
' (4.1.19)
In analogy to equation (4.1.14), we define a transformation

X= AXAt, (4.1.20)
where det A = 1. The matrix A is assumed to be unimodular but not nec-
essarily unitary. Clearly det X = det A det X det At = det X. Therefore by
(4.1.19) we must have
(Xl)2 + (X2)2 + (X3)2 _ (X4)2 = (Xl)2 + (X 2)2 + (X 3)2 _ (X4)2.
The transformation (4.1.20) induces a Lorentz transformation (which is prop-
er and orthochronous).
Example 1: Consider either of the two matrices
= [e (x x
X 3 _ 4)
Xl + iX2
The above equation implies that
Xl = xl,
X2 = X2,
X3 = X3 cosh X- X4 sinh X,
X4 = _X3 sinh X + x 4 cosh X.
These equations define a well-known proper orthochronous Lorentz trans-
formation. 0
Example 2: Consider either of the two matrices

A == +[ J(coshX 1)/2 + -.;r-(C-OS-h-X---1)/-2].
- - J(cosh X - 1)/2 J(cosh X + 1)/2
Equation (4.1.20) produces
[
X3 - X4 Xl - ix 2 ]
Xl + iX2 _(x 3 + x4)

= [- X 1 sinh X + X3 - X4 ~osh X X 1 co~h X - iX2 + X4 sinh X ].
X 1 cosh X + ix 2 + X4 smh X - Xl smh X - x 3 - X4 cosh X
The above matrix equation yields the proper orthochronous Lorentz trans-
formation
EXERCISES 4.1
1. Let P == x~(J~, P == x~(Ja, where the (Ja are Pauli matrices. Consider a
unitary unimodular matrix U and the matrix equation P = UPU t . Prove
that it is impossible to obtain the transformation x = - x from the above
matrix equation.
2. Consider the matrix equations
where det A = 1. Construct the Lorentz transformation inherent in the above

equations in terms of the entries of the matrix A. Furthermore, prove that
this Lorentz transformation has to be proper and orthochronous.
3. Consider the six-parameter Lorentz transformation given by (3.4.6). Ob-
tain explicitly the two 2 x 2 complex matrices A that induce that Lorentz
transformation.
4.2. Spinors and Spinor-Tensors

The spinor space Vz has already been defined as a complex two-dimensional
vector space. Let {e1,e Z} be a basis (dyad) for Vz, and let ~'t E V2 Then 1/1 =
L~=l tf;AeA = I/IAe A for some complex components I/I A. An invertible linear
mapping :E: Vz -+ Vz can be characterized by the equation
(4.2.1)
where det:E "# O. In this case the components of 1/1 = I/IAeA = ~AeA undergo
the transformation [recall Theorem (1.2.2)]
(4.2.2)
We shall subsequently assume the unimodular conditions
det:E = detA = 1. (4.2.3)
The complex conjugate components transform under the rule:
~A == ~A = AABI/I B == AABI/IB. (4.2.4)
Here the repeated barred capital indices and the repeated capital indices are
summed disjointly over the range {1,2}. The components of a covariant
complex vector (which is a spinor) and its complex conjugate components
transform as follows:
(4.2.5)
Higher-order spinor components transform by the rule
.f,A,Ak!,i!! = AA, .. 'AAk AB,_ ... AB,_ :E G , :E Grn
'+'CtoCmDtoDn El Ek Fl Fl C1 C nt
(4.2.6)
x :EH,_ ... :E Hft _ .I,E, EkF,:.:.F, _
Dl Dn'Y Gt"'GmHt"'Hn '
The totally antisymmetric permutation symbol is defined by
(4.2.7a)
Thus in matrix form we have
(4.2.7b)
It follows that
eACe CB = -b BA,
eAce CB = -b BA, (4.2.8)
GACG
AC = GAC eAC 2
=.
The transformation properties of the permutation symbol (compare the
equation (1.3.13a
(4.2.9)
Under a unimodular transformation eAB' GAB behave as numerical spinors,

with unchanged values. The antisymmetric bilinear combination
GABI/1 AXB = 1/11X2 - 1/12X1 = det [1/1 1

1/1 2
Xl]
"1..2
remains invariant under a unimodular transformation. This quantity is cer-
tainly not an inner product between two complex vectors 1/1, X, but it is
formally the "complex area" between two complex vectors 1/1, X. Therefore,
the spin or space V2 is not an inner product space and GAB is not the metric
spinor as it is often loosely called. Since V2 is endowed with an antisymmetric
bilinear form e(l/1, X) == GABI/1 AXB, it is a symplectic space. In such a space, a
vector 1/1 E V2 has a dual covector naturally associated with it via the anti-
symmetric bilinear form e [compare (1.3.15)]
1/11 == eABI/1B,
(4.2.10)
./,'Ii< - ,, __ .},B
'I'A = "AB'I' .
In the literature the star is usually dropped and the above equations appear
as
I/1A == eABI/1B,
(4.2.11)
I/1A == eABI/1 B.
We shall follow the same notation as in the literature. Inverting the above
equations via (4.2.8) we get
I/1c = eAC I/1A = -e CA I/1A,
(4.2.12)
I/1c = eACI/1A = -G CA I/1A'
Equations (4.2.11), (4.2.12) indicate how to "raise or lower indices" for spi-
nors. But one should be careful to note that
(4.2.13)
The 2k+l+m+n+2(r+s) components of a mixed spinor-tensor transform, under

the combined unimodular and Lorentz transformations, according to
(4.2.14)
We choose a special set of numerical mixed tensor-spinors by the following
choice:
- [01 oIJ = ~1 '

[~lAB] [ 111
= ~_ ~1~2J =
~121 ~122
[211
[~2AB] = ~_ ~2~2J == [0 -ioJ -
- 2 ~,
~221 ~222 i
[311 (4.2.15)
[~3AB] = ~_
~321
~3~2J =
~322 - [1 -1OJ = ~3'
- [1 OJ1 4
[~4AB]
[ 411
= ~_ ~4~2J = = ~ .
~421 ~422
Lowering the spinor indices by the rule [see (4.2.11)]
we obtain
[~lAB] = -~\
[~2AB] = ~2,
(4.2.16)
[~3AB] = _~3,
[~4AB] = ~4.
Recall Theorem (4.1.1), which is applicable to some of the matrices in (4.2.15).

The appropriate generalization of this theorem is stated next.
Theorem (4.2.1): The components ~kAB of the numerical spinor-tensor satisfy

(i) ~kBA = ~kAB,
(4.2.17)
().).) ~a
BA ~bBC = - dabV~ AC +!
(/2) IJabkl~ kBA ~ IBC,
where IJabcd is the totally antisymmetric pseudotensor [equation (1.3.13b)].

Proof: (i) By the hermitian properties (Jkt = (Jk the equations (JkBA = (JkAB
follow.
(ii) The proof emerges by direct computations. For example,
= 0 + (-1) = (-1) +0
= 0 + (i) = 0 + (i/2)[(1)(1) - (1)( -1)]

= 0+ (i!2)[(J3BI(J4ih - (J4BI(J3BIJ
= -d 12 f5\ + (i/2)'112kl(JkBI(JI BI .
Similarly all the other equations follow.
From (4.2.17) the following three other useful equations emerge:

(Ja BA (JbBC + (Jb BA (JaBC = - 2dabU,A C,
.
(Ja BA (JbBC - (Jb BA (JaBC = l'1abkl(J kBA (J I BC, (4.2.18)
(JaBA(JbBA = -2dab
Now we shall discuss the transformation property of a second-order spinor
TAB.
Theorem (4.2.2): If TAB are the components of a second-order spinor then

det[TABJ is invariant under a spinor transformation (4.2.6).
Proof: Under (4.2.6) one has fAB = AACABDTcD. Thus taking the determi-
nant of both sides one obtains
det[fAB] = fIlf22 _ fI2f21 = (1/2)GCAGDBfABfcD
= (1/2) f CD fCD = (1/2) TCDTcD = det[TABJ
In the special case where the matrix [XABJ is hermitian, then by Theorem
(4.1.3) we can write
(4.2.19)
for some real numbers Xk, and the above equation is equivalent to equation
(4.1.19). Under a spinor transformation
gAB = AAcABDXCD,
the matrix gAB remains hermitian. Thus we have the linear combination
(4.2.20)
for some real numbers Xk. By equation (4.1.19) and Theorem (4.2.2), we have
under a spinor transformation
-dk1xkx' = det[XAB] = det[XAB] = -dk1xkx'. (4.2.21)
Since the above equation is also a consequence of a Lorentz transformation,

we can link a Lorentz transformation with a spinor transformation.
Theorem (4.2.3): Let a spinor transformation be given by the unimodular ma-

trix [AAB]' There exists a corresponding proper orthochronous Lorentz matrix
[lab] given by
(4.2.22a)
Proof: Recall that in (4.2.18) we had (JkAB(Jb AB = -2i5 bk Using this we get
(JbABXAB = (Xk(JkAB)(JbXB = _2 Xb;
thus, x b = (-1/2)(Jb AB XAB. Similarly we have

xa = (-1/2)(Ja AB xAB = (-1/2)(Ja AB (AAcABDXcD)
= (-1/2)AAcABD(JaAB(J/Dxb).
But under a Lorentz transformation we have x a = la bXb. Thus we see

[lab + (1/2)AAcABD(JaAB(JbCD]xb == 0,
where Xk can take arbitrary real values. Differentiating with respect to Xb, the
above identity yields the equation (4.2.22a). It can be proved that the Lorentz
transformation in (4.2.22a) is proper and orthochronous (see Exercises 4.2,
problem 3).
The equation
l ab(J bAB = ACAADB(J aCD (4.2.22b)
is equivalent to (4.2.22a).
Example: Consider the spinor transformation given by the unimodular

matrix
XE U;~.
By (4.2.22a)
[3 4 = -(1/2)AAcABD(J3AB(J/D
= -(1/2)[A\A 1 1 (J3 11 (J/l+ A22A22(J322(J422]
= _(1/2)[(e z/2)2( -1)( -1) + (e~z/2hl)( --1)]
= -sinh X.
This result agrees with the previous example worked out after equation
(4.1.20). 0
Consider now a higher-order spinor I/I AI A}il Ii. that is totally symmet-
ric with respect to the indices A I ' ... , Aj and the indices HI' ... , Hk . The
components of such a spin or correspond to a vector in a complex vector
space of dimension (2j + 1)(2k + 1). Precisely such vector spaces constitute
all irreducible representations for the proper orthochronous Lorentz group
!l'/+. The proof of this important statement is beyond the scope of this book.
(See the book by Gelfand et al in the references after this chapter.)
It has been mentioned before that improper Lorentz transformations are
not induced by spinor transformations. However, when we consider the con-
jugate linear mappings of the spinor space Vz , the improper Lorentz transfor-
mations can be accommodated. Recall that for a unimodular conjugate linear
mapping L of Vz we have
I:(AII/II + ).21/12) = AII:(I/Id + ,{lL(l/Iz)
In terms of the components, we can express a conjugate linear mapping as
$A = AABI/IB, $A = AAIiI/lIi,
det[AAB] = det[AAIi] = det[I: AB] = det[I: A Ii ] = 1.

Consider again the hermitian matrix X AB and the equation (4.2.20). Under an
improper Lorentz transformation we have
labxb = xa = (-1/2)(1a AB XAB
= (-1/2)(1aABAAcABvXcv
= (-1/2)(1aABAAcAB[iXvc
= -[(1/2)AAcABv(1aAB(1bVC]xb.
From the above equation we can derive

lab = (-1/2)AAcABv(1aAB(1b[iC. (4.2.24)
Example: As a special case consider the unimodular matrix [AAB] =
[-01 -1OJ . According to (4.2.24), (4.2.20), (4.2.16) we have

Xl = (-1/2)[ATIA2Z(1ITzXZI + AZzAIT(112IXT2]
= + ix 2 ) + (_I)(XI - ix 2 )]
(-1/2)[( -1)(x l = xl,
x2 = (-1/2)[(12T2X21 + (122IXT2] = _x 2 ,
x3 = (-1/2)[(1\I XTI + (13 22 X22] = x 3,
X4 = (-1/2) [(14Tl X TI + (1422XZ2] = X4.
The above set of equations denotes a special improper Lorentz transforma-

tion. D
EXERCISES 4.2
1. (i) Prove that for an arbitrary symmetric spinor SAB one has SAA == O.
(ii) Prove that for an arbitrary spinor !lAB one must have eL AB - eLBA -
eABeL c C == O.
2. Prove that (J"1iC(JaDF = - 2E1iDEcF'
3. Prove explicitly that the Lorentz transformation

lab = (-1/2)AAcABD(JaAB(J/D,
where ADB is a spinor transformation, has to be proper and orthochronous.

4. Obtain the conjugate linear unimodular transformation AAB , which
induces the improper Lorentz transformation of space reflection P.
4.3. Dirac Matrices and Dirac Bispinors

We shall define a Dirac bispinor as a vector in a four-dimensional complex
vector space V:. Note that in the literature a Dirac bispinor is commonly
called a Dirac spinor or simply a spinor. We shall adopt the simple conven-
tion that a bispinor refers to a Dirac bispinor. Relative to a basis of the V: ,
four components of a bispinor are expressed as the column vector
1
" t/J2 ]
t/J (4.3.1)
t/J = [t/J ] = [ ~: .
The indices u, v, w, t will be used for bispinors, and each will take values in
the set {I, 2,3, 4}. Furthermore, the bispinor components are identified with
the components of two spinors by the rule
(4.3.2)
The components of a bispinor transform under a unimodular transforma-

tion as
(4.3.3)
where det[ff"v] = det[Q"vJ = 1. Here the summation convention is adopted.

The components of a higher-order mixed bispinor-tensor transform under

a combined Lorentz and bispinor transformation according to
(4.3.4)
Example: Consider the transformation of the components cti"v of a mixed

bispinor-tensor:
&i"v = aik.:1"wQtvctkWt. 0 (4.3.5)
l n
Let us consider a special numerical mixed bispinor-tensor with components
relative to a suitable basis given by
~ ~~
0 0
0
" [,'".J 1 0
0 0
~ ~~ -~]o '
0 0
0
" [,'J [ -i 0
!
0 0 0
(4.3.6)
~ ~ -~]o '
0
0 0
" [,'J [ 0 0
~!
-1 0 0
~ -~]o .
0 -1
0 0
" [,'.J [ 0 0
0 0
The matrices ya are known as Dirac matrices. In terms of the 2 x 2 sub-
matrices (Jk, the Dirac matrices can be expressed as
-IJ
o. (4.3.7)
In the literature various authors make various choices for the Dirac matrices.
All these choices, however, must satisfy the following equations:
(4.3.8a)
Trace yU = 0, (4.3.8b)
(4.3.8c)
(4.3.8d)
Equation (4.3.8a) follows from (4.3.7) by blockwise matrix multiplication, and

equations (4.3.8b), (4.3.8c), (4.3.8d) follow directly from (4.3.6).
We can introduce another 4 x 4 matrix
(4.3.9)
The special properties of yS are summarized in the following equations:
(4.3.10)
The transformation properties of the 'Y" matrices are given by the equation
(4.3.5), i.e.,
(4.3.11 )
However, given a Lorentz transformation a ij , there exist unimodular matri-

ces :TU wand nu wrepresenting bispinor transformations, such that the numer-
ical components ,,/u v remain unchanged. For such a case
(4.3.12)
Instead of proving the equation (4.3.12) in general, we shall give some exam-
ples. Note that the last equation above generalizes equation (4.2.22a) to the
4 x 4 matrix form.
Example: Let us choose
~ [~
0 0
[I',]
1
0
0
0
cosh 1.
-sinhX
-sLx}
cosh X
One of the equations in (4.3.12) yields
o
~ ~ re~,
eX:
o 0
1',,' ,'coshX - ,'smhX o 0
-eX: 0
o 1
o o
o o
-1 o
All other equations in (4.3.12) for this case are also satisfied with the same
choice for Y. 0
Example 2: Consider the Lorentz matrix with entries lab = -Jab' The corre-
sponding Lorentz transformation is proper but not orthochronous. In this
case we have
Therefore, in this case we obtain the unimodular 4 x 4 matrix Y = y5. 0
EXERCISES 4.3
1. Consider the sixteen 4 x 4 matrices I, ya, yayb (a < b), y5, y5 y a.
(i) Prove that except for I all other fifteen matrices have a zero trace.
(ii) Prove that these sixteen matrices form a basis for the vector space of 4 x 4
complex matrices.
2. Consider the (improper) time reversal Lorentz transformation, viz.,
Obtain the two 4 x 4 unimodular matrices Y, corresponding to bispinor

transformations, which induce this time reversal.
References
1. W. L. Bade and H. Jehle, Rev. Mod. Phys. 25 (1953), 714-728. [pp. 104, 112, 115J
2. E. M. Corson, Introduction to tensors, spinors, and relativistic wave-equations,
B1ackie and Son, London, 1955. [pp. 112, 115, 121J
3. 1. M. Gelfand, R. A. Milnos, and Z. Y. Shapiro, Representations of the rotation
and Lorentz groups, and their applications, The MacMillan Co., New York, 1963.
[pp. 102, 103, 235J
5
The Special Relativistic Mechanics
5.1. The Prerelativistic Particle Mechanics

We shall start with a short review of the prerelativistic mechanics of Newton,
Lagrange, and Hamilton. For simplicity we restrict ourselves to systems of a
single-point particle having (constant) mass m > O. Let the parameterized
motion curve be given by x a = ,qra(t), in the Euclidean space 1E 3 . Let the three
components of the force vector be given by r(t, x, v), which are functions of
seven real variables. Here t stands for the time variable, x == (Xl, x 2 , x 3 ) are
the spatial coordinates in a Cartesian coordinate system, and v == (vt, v2 , v3 )
are the velocity variables. Newton's equations of motion for a single particle
are given by
(5.1.1)
These equations imply that
~ [(1/2)m5 d,qra(t) d,qrfJ(t)]

dt all dt dt
(5.1.2)
The above equations can be interpreted physically as "the rate of increase of
kinetic energy of a particle is equal to the rate of work performed by the
external force."
In the special case of a monogenic force, there exists a work function
Vet, x, v) such that
5 fll( x ) = aV(t, x, v) _~{[aV(t,x'V)J }

all t" V I a aX I..
dt aV a I
. (5.1.3)
In such a case, the equations of motion (5.1.1) are the Euler-Lagrange

equations derived from a variational principle. To explain such a statement
we define a Lagrangian function L: [t l ,t 2 J x D6 C 1R7 -.IR such that LE
Ij&'2([t l ,t 2J x D6)' and in [t l ,t 2] x D6 we have det[a 2L(t,x,v)!av aav il ] > O.
89
90 5. The Special Relativistic Mechanics
Consider the twice differentiable parameterized curve y in 1E3 given by x a =

gra(t), t E [t 1,t2]. The set of all such curves will be denoted bylj&'2[t1,t2]. The
action functional is a mapping J: lj&'2 [t l' t 2] --+ IR such that
J[y] == f l2
[L(t,x,v)lx'~.'(I),v'~d:P(I)ldt]dt. (5.1.4)
I,
A "slightly varied" curve y in 1E 3 , having the same end points, is given by

xa = gra(t) + 8h lZ (t),
ha(t 1) = h a(t 2) = 0,
where h a E lj&'2[t1,t2] and 8> 0 is a small positive number.
The "slight variation" in the action integral in (5.1.4) is given by
OJ[y] == J[y] - J[y]
= f l2
I,
[L(t, x, V)lx'= ''(I)+eh'(t). v'=d'(I)/dl+e(dh'(I)/dl)
- L(t, X, V)lx'~'(I). v'=d.'(I)/dl] dt
=e f l
2 {hlZ(t) [OL(t, x, V)]
I, OXIZ Ix'~'(I). v'~d'(I)fdl
+ dhlZ(t) [OL(t, x, V)] } dt + (D(e2)

dt ov lZ Ix'~ :'(1). v'=d:1"'(I)/dl
=e f l
2 {hlZ(t) [OL(t,~, V)] }dt + ehlZ(t) [OL(t,:, V)] 112
aX
f
I, I.. OV I.. I,
-8
l
2 {ha(t): [OL~':' V)] }dt + (D(e2)
f
I, t V I..
=e l
2 h'(t) {[OL(t,~, V)] _ ~ [OL(t,:, V)] }dt + (D(e2).
I, aX I.. dt OV I..
If we impose that stationary condition on the action integral of the curve y,

then we have lime~o+ (bJ Ie) = O. In that case we obtain
f I,
l
2 hlZ(t) {[OL(t, ~'
aX
V)] _ ~ [OL(t,:, V)] }dt = O.
I.. dt OV I..
Since the ha(t) are arbitrary twice differentiable functions in [t l' t 2] for IX = 1,
2, 3, we obtain by the lemma of Dubois- Reymond (see the book by Gelfand
and Fomin referred at the end) that
o = [OL(t,~, V)] _ ~ {[OL(t,:, V)] }. (5.1.5)

aX I.. dt OV I..
The above equations are called the Euler-Lagrange equations.

If we can choose the special Lagrangian
L(t, x, v) = T(x, v) + U(t, x, v), (5.1.6)
then the Euler-Lagrange equations (5.1.5) yield
liW(t,:, V)] _ ~{laU(t,:, V)] }_ ~lmc5af3 d:T~(t)] = o.

ax dt I..av dt dt I..
The above equations are precisely the equations of motion (5.1.1) for the case
of a monogenic force given by (5.1.3).
We define generalized or canonical momentum variables by
(5.1.7)
where det[a Z L(t, x, v)/ovaovJl] = det[oPa/ovP] > O.

In the next step we introduce the Hamiltonian function by a partial Legen-
dre transformation:
H#(t, x, v, p) == Pava - L(t, x, v), (5.1.8)
where p == (PI' Pz, P3)' But we notice that
Therefore, we can take the Hamiltonian function to be
H(t, x, p) == Pava - L(t, x, v). (5.1.9)
It follows that
oH(t, x, p) _ ,
oP, - v,
oH(t, x, p) oL(t, x, v)
(5.1.10)
ox'
oH(t,x,p) oL(t, x, v)
ot
The momentum components along the motion curve y is given by (5.1.7)
as
oL(t, x, v)
P (t ) = --,-----
, av Ix'= :i"(t). v, =d.''(t)jdt
By Euler-Lagrange equations (5.1.5) and equation (5.1.10) we have along the

corresponding motion curve y# in the phase space [six-dimensional manifold
coordinatized by (X 1,X 1 ,X 3,Pl,Pl,P3)] the equations:

d aL(t, x, v)
- Pa (t) = ------::----
dt ax a Ix'= .'f'(t), v'=d.'f'(t)!dt
aH(t, x, p)
(5.1.11)
ax' Ix'= .'f'(t).P,=P,(t)
~:!ta(t) = aH~,x,p)
dt oPa Ix'= .1"(t), p,=P,(t)
The above six equations are called Hamilton's canonical equations (for a sin-
gle particle), The total time derivative of the Hamiltonian function [using
(5.1.1 1)] is
d ( ) _ aH(t, x, p) d:!t'(t) aH(t, x, p) dPa(t) aH(t, x, p)
dHt,x,PI"- a +~ a' +-d------=-a--
t t l " x I.. t Pa I..
aH(t, x, p) d?'(t) dPa(t) d:!t'(t) dP,(t)
-----+----
at I.. dt dt dt dt
aH(t, x, p) aL(t, x, v)
(5.1.12)
at I.. at I ..
In case aL(t, x, v)/at l.. == 0, we have the integral

H(x,p) = E, (5.1.13)
where E is the integration constant representing the total energy of the
particle.
A conservative system is characterized by the work function satisfying
aU(t, x, v) = 0
at -,
aU(t, x, v) = 0
ava -, (5.1.14)
U(t, x, v) = - V(x),
where V(x) is called the potential energy of the particle. For such a system
[using (5.1.6), (5.1.8), (5.1.13), and Pa = mJapvP] we obtain
H(x, p) = Pava - [(l/2)mJap v av p - V(x)]
= (l/2m)(pi + p~ + p~) + V(x)
= (l/2m)Ja Pp,pp + V(x)
=E. (5.1.15)
The above relation yields a five-dimensional submanifold in the six-dimen-
sional phase space, Such a submanifold is called an energy hypersurface (or
energy shell). The motion curve y# of a particle in a conservative system with

total energy E must lie wholly in the corresponding energy hypersurface. We
can also express the constraining energy hypersurface by
W(x,p) == H(x,p) - E = O. (5.1.16)
The variational principle satisfying a constraint can be accommodated by the

method of Lagrange multipliers. To illustrate this technique, the canonical
action integral for the motion in the constraining hyper surface (5.1.16) is
taken to be
Jh[y] = f t2
I,
[Lh(x, p, v, 1!)lx'= f'(t), p,=P,(t), v'=d:,(t)/dl, It,=dP,(I)/dIJ dt
(5.1.17)
Here y is the motion curve in the twelve-dimensional manifold, which is

coordinatized by (xi,x2,x3,Pi,P2,P3,vi,v2,v3,ni,n2,n3)' such that its pro-
jection onto the phase space is the curve y #. The corresponding Euler-
Lagrange equations derived from (5.1.17) are (5,1.18)
[ CLh(X'P:V'~)J
aX I..
_ dt{[OLh(X'~V'~)J
d OV I"
}.
, ow(x, p) dP.(t)
= -)., ---
aX' I.. dt
, oH(x, p) dP.(t)
= -)., ---
aX' I" dt
= 0,
[ OLh(~P'V'1!)J
P. I..
_~{[OLh(X'P'V'1!)J }.
dt an. I..
_ _Aow(x,P)
- v I.. a
P. I"
= d~'(t) _ A oH(x, p)
dt oP. I..
= 0, (5.1.18)
These equations are equivalent to Hamilton's canonical equations (5.1.11)
for a conservative system provided that the Lagrange multiplier A is chosen
to be unity.
Example: The problem is to find the prerelativistic motion curve of a three-

dimensional oscillator with mass m > 0 and spring constant k > O. The cor-
responding Hamiltonian function is
For the case of strictly positive energy, the energy hypersurface is given by
The Hamilton's canonical equations are given by:
cH(x, p)
ex'
~ ret) = oH(x, p) = (l/m)P.(t).
dt CP. I ..
Putting the second system of equations into the first, we get a system of
second-order decoupled ordinary differential equations. The general solu-
tions of this system, yielding all possible y#, are given by
.Uf(t) = A'cos[J(k/m)t + c'], p.(t) = -J;kA" sin [J(k/m)t + c'],
where A', c' are six constants of integration. The total energy of the oscillator
is constant and is given by
E = (1/2m) [b' Pp.(t)Pp(t)] + (k/2) [b.p.Uf(t).UfP(t)]
= (k/2)o.pA AP ~ 0.

In case E > and the energy-hypersurface Ls is prescribed, three constants
of integration A' must satisfy the above condition. Therefore, two of the three
A' and the three c' can be adjusted to five prescribed initial conditions to
obtain a unique motion curve y# on the prescribed energy hypersurface in the
phase space. D
EXERCISES 5.1
1. For a plane pendulum with mass m > 0, length I > 0, the Lagrangian in
polar coordinates is given by
L(e, EI) = (1/2)mJ2{)2 - mgl(l - cos e),
where g is the constant of gravity.
(i) Obtain the corresponding Hamiltonian as given in (5.1.9).
(ii) Solve the corresponding canonical equations of Hamilton as given in
(5.1.11).
2. Suppose that the coordinate x 3 is ignorable, i.e., OW (x, p)/cx 3 == 0, and
thus P3 = C3' a constant by the canonical equations (5.1.18). Prove that the
remaining canonical equations can be derived from the reduced action inte-
5.2. Prerelativistic Particle Mechanics in Space and Time [3 x ~ 95
gral (Routh's procedure):
where Yo is the curve represented by

. df![(t)
x = f!l (t), p. = p.(t), v = ~ for r:J. =, 1, 2.
5.2. Prerelativistic Particle Mechanics in Space and

Time 1E3 x ~
Before passing to the relativistic mechanics, we shall discuss an intermediate
stage, where the preceding prerelativistic mechanics is developed in terms of
a mathematical parameter L as the independent variable instead of the usual
time parameter t. The motion curve y in the product manifold 1E3 x ~, which
represents space and time, is parametrized by
t = Y(r), x = f![(t) == ~(L),
(5.2.1)
~Y(L) # 0 df![(t) _ d~.(r)[dY(r)J-l
dt ' ~-~~
The action integral in (5.1.4) yields
i '2
I,
[L(t, x, v)lx'= '(1). v'=df'(t)/dt] dt
=JIt> [dY(L)
t,~ L(t, x, v)ll= .T(t). x'=~'(t).
]
v'=(d~'(t)/dt)/(d.T(t)/dt) dL
i
t~
== [L'(t, x, t', x')lt= .T(t). x'=~'(t). t=d.T(t)/dt. x"=d~'(t)/dt] dL.

t,
The modified Lagrangian is

L'(t, x, t', x') == t' [L(t, x, v)lv'=(x")/t']. (5.2.2)
The Euler-Lagrange equations derived from both Lagnmgians should be
equivalent. We can check that explicitly using (5.2.2) in the following steps.
oL'(t, x, t', x') ,oL(t, x, v)
- - - , - - - = t --,------
ox ox IV'=(x")/I'
oL'(t, x, t', x') oL(t, x, v)

---,------- = t' -~-
ot ot IV'=(x")/t'
,
p, ==
oLI(t,X,t',X')
iJ" = t
I [OL(t, X, V)
iJ
J 1-1
(t )
x V IVl=(X'I)/t'
iJL(t, X, v)
- ~ av' --IV'=(X")/I'
oL'(t x t ' X')

p; == 'at',' (5,2.3)
- L(t, X, v)lv'=(x")/"
_
+ t [OL(t,
a x, v)
I ]
[ - x '.(t)
1-2
]
V Iv'=(x")/,'
= [ L(t, X, v) - v'
oL(t, x,
0"
V)] = - H(t, X, p').
~ Iv'=(x")/"
The modified Euler-Lagrange equations are

OLI(t,x,t',X') _ {dT[iJLI(t,X;:',x l
)] }
iJx' I.. d iJx I ..
= dff{T) {~L(t,_x,v2 _ ~ [iJL{t, x, V)] } = 0

dT oX' I.. dt iJv' I.. '
and thus the modified Euler-Lagrange equations hold. However, there exists
the fourth Euler- Lagrange equation:
oC(t, x, t', x') _ ~ {[OLJt, ~', x-'1] }

at I.. dT at' I ..
_ dff(T} {OL(t, X, v) _ ~[ _ OL(t'~l] }

- d a
Ttl"
d
t
L(t, x, v) va. v I..
= dff(T) {OL(t, X, v)
dT tal " +~ [( )]} = 0
d H t, x, P I.. - .
t
The last expression vanishes by (5.1.12). Thus the two systems of Euler-
Lagrange equations are equivalent.
Now we notice an interesting property of the modified Lagrangian defined
in equation (5.2.2). For an arbitrary v =I 0,
L'(t, x, vt ', vx ' ) = vt' L(t, X, v)IV'=(vX")Jv"=(X"J/t'
= VL'(t, x, t', x'). (5.2.4)
Thus the modified Lagrangian function L' is homogeneous of degree one in
the generalized velocity variables t', X'. By Euler's theorem on homogeneous
functions we must have
t
I oL'(t, X, t', x')
at I +X
I. oL'(t,a X,
I
t', x') _
-
I
L (t, x, t , x .
I ')
(5,2.5)
x
5.2. Prerelativistic Particle Mechanics in Space and Time 1E3 x IR 97
But the above equation implies that the generalized Hamiltonian

H'(t,x,p;,p') == p;t' + p~x'a - L'(t,x,t',x'}
= t' oL'(t,x, t', x') + x' . oL'(t, x, t', x')

ot' ox'
- L'(t,x,t',x')
== o. (5.2.6)
Hence the generalized Hamiltonian is identically zero. But we need some other
function to replace the generalized Hamiltonian in order to obtain the gener-
alized canonical equations. We can treat the relation
w'(t,x,p;,p') == p; + H(t,x,p') = 0 (5.2.7)
as a seven-dimensional constraining hypersurface in the eight-dimensional
extended phase-space manifold coordinatized by (t, x, p;, p'). Now we can gen-
eralize the canonical action integral (5.1.17) (involving the Lagrange multi-
plier ),) to
(5.2.8)
where I.. denotes evaluation along the curve y in the sixteen-dimensional

generalized extended phase space. The corresponding Euler- Lagrange equa-
tions are given by
[OL~(oo)]
ot I..
_ ~ [OL~(,:J ]
dr ot I ..
= _A[OW'(' .)]
::l
- ~[ 'J
d p, I
ut I.. r
= _{dP;(r) + A [OH(oo) ]} = 0,
dr ot I ..
[ OL~(~.)] _~{[OL~(,~.)]
ox I.. dr oX I..
= _A[OW'(~.)] dp~
oX I.. dr I..
= _{dP~(r) A[OH( . . )] } = 0
dr + O'
x I ..
'
[ a~~(;.)] _~{[aL~(;.)] }
CPt I.. dr ant I.
,
= t - A ~-,-
. [aw'(' .)] -
,
0 = [t - )']1"
apt I
d.aT(r)
= ~ -l(r) = 0,
[ aL~(;.)] _~{[aL~(;.)]
apa I.. dr ana I..
= {x,a - A[130.:'(;' )]} - 0

oPa I
= d~a(r) _ A[aH( .. )] = 0 (5.2.9)

dr ap~ I.. .
As a special choice we can have t = .aT(r) == r. In that case A = l(r) = 1 and

Hamilton's canonical eq uations (5.1.11) and (5.1.12) are exactly recovered in
(5.2.9).
Example: Consider the prerelativistic mechanics of planetary motions. The

Lagrangian is given by (outside the origin)
L(t, x, v) = (m/2)bafJ v"v fJ + Mm(bl'vxl'xVfl/2,
where m > 0 is the planetary mass and M > 0 is the solar mass times the
Newtonian gravitational constant. The corresponding Hamiltonian is
H(t, x, p) = (l/2m)c5 aP paPp - Mm(bl'vxl'xTl!2.
Therefore, the energy constraint is
w'(t,x,p;,p') == P; + H(t,x,p')
= P; + (1/2m)c5aPp~pp - Mm(c5l'vxl'xTl/2 = 0,
aw'( .. ) _ aH( .. ) _ 0
~a-t - - ------at
= ,
aw'( .. ) _ aH( .. ) _ c5 P(c5 fJ. v)-3/2

~-~-Mm apx I'Vx X ,
aw'( .. ) _ aH( .. ) _ (1/ )~aP ,

ap:--~- mu pp.
Therefore, the Euler-Lagrange equations derived from the corresponding

5.2. Prerelativistic Particle Mechanics in Space and Time 1E3 x IR1 99
action principle [see equation (5.2.8)] are
dP;(r) = 0
dr '
dP~(r)
dT
+ )Mmb
' ap
xP(b Xl'x v)-3/2
I'V
= 0
,
d.r(r) _ A = d.r(T) _ l(r) = 0,

dr dr
d~:;r) _ I(T)(1/m)bapPp(r) = o.
Integrating the differential equation on .r(r) we get
t = f f
dt =
d.r(T)
~dr = fI(T)dr + C,
where C is a constant. Recalling ~a(r) = xa(t), P~(T) = PaCt) and noticing that
P;(r) = -E, the negative energy constant, we get
d~t(t) + MmbapXP(t)[bl'vq-l'(t)Xv(t)r3/2 = 0,
- E + (1/2m)b ap Pa(t)Pp(t) - Mm[bl'vXI'(t)q-V(t)r1/2 = 0
Eliminating PaCt) we get the following second-order equations of motion and

the first integral (indicating the conservation of energy),
d 2q-a(t)
~ + MXa(t) [bl'v.'?{I'(t)XV(t)r 3 / 2 =, 0,
(m/2)bap ~~:(t) dX~(t) _ Mm[bl'vXI'(t)XV(t)r1/2 =, E.
The second-order equations of motion could have been obtained directly

from the original Lagrangian L(t, x, v) as the Euler-Lagrange equations. This
fact shows the inner consistency of this mathematical method. The equations
of motion permit three other first integrals (indicating the conservation of
angular momentum), namely,
q-a(t) dXP(t) _ q-p(t) dXa(t) = hap = _ hPa = const.

dt dt
The relativistic version of the planetary motion will be completely solved in
the next section. 0
EXERCISES 5.2
1. Consider the motion curve on the energy hypersurface
H(x, p) == (1/2m)o~fJp~Pfl + V(x) = E.
Prove that the same motion curves are obtained by the variation of the
action integral (Jacobi's Principle)
" ft
J [y ] = t,
2
{
[E - V(x)] IX'=~'(t) oafJ
fJ
de (r)de (r)}l/2
a
----;rr ----;rr dr.
5.3. The Relativistic Equation of Motion of a Particle

In this section we shall discuss the special relativistic mechanics of a single
particle. For that purpose it is convenient to use the proper time s along a
motion curve as in Figure 22.
The parametric equations of a motion curve are given by
Xk = (n k 0 X 0 y)(s) = q"k(S),
(5.3.1)
t == X4 = q"4(S),
where s E [O,Sl]. It is assumed that the functions q"k E rt'2[0,Sl]. Further-

more, it is assumed that the motion curve in M4 is future-pointing and time-
like, i.e.,
dt dq"4(S)
-=-->0 (5.3.2)
ds ds '
From the first inequality it follows that the function [l'4 is invertible in [0, s 1];
hence, we can write
(5.3.3a)
R
o s
FIGURE 22. Proper time of a motion curve.

for t E [to, t 1]. We define the four-velocity vector along the curve as
(5.3.3b)
which is also the unit tangent vector along the curve. It follows from (5.3.3b)
and (5.3.2) that
(5.3.4)
But the three-velocity components, which are actually measured, are different
from these components. To express the measurable components of velocity,
we reparameterize the motion curve by
x" = .or(s) = ;l'[Y'(t)] == ;l'#O(t), X4 = ?[#4(t) == t,
(5.3.5)
d;l'#O(t)
VO(t) == .
dt
Note that these components of the measurable velocity are: not components
of any Minkowski vector.
Theorem (5.3.1): The components of a Minkowski velocity vector and the cor-
responding components of a measurable velocity vector satisfy the inequality:
Iv(t)1 ~ lu"(s)l (5.3.6a)
Proof: Equation (5.3.4) yields
[d~:(S) J- bop ~~:(s) d:r;}s) = 1,

so
Thus
(S.3.6b)
Hence,
IvO(t)1 Ia!- II d;l'4(s) 1-1 =

= d;l'"( ) -----;J;- luO(s)[1 - b"pvO(t)vP(t)] 1/2 I,
which gives Iv"(t)1 ~ luO(s)l.
For a photon, one can choose a special parameter T along the null world-
line such that

Xk = ~k(r), r E [rl' r 2],
uk(r) == d~k(r), (5.3.7)

dr
dkIUk(r)ul(r) = O.
The proper (rest) mass of a massive particle m = M(s) > 0 for S E [0, S1]'
For a photon it is assumed that m == O.
The Jour-momentum vector of a free massive particle (in this case the mass
is taken to be a constant) is defined as
Pk(S) == M(s)dk/u'(s) = mdk,u'(s).
By (5.3.4), one can obtain
dk'Pk(S)P,(S) = _m 2 (5.3.8a)
For a photon one assumes that
dk1pk(r)p,(r) = O. (5.3.8b)
The four-dimensional space of all possible four-momenta is the dual Min-
kowski space M 4 , which allows the Minkowski coordinate chart (X, M 4 ) such
that X: M4 -+ ~4. In ~4 all possible timelike four-momenta for a given parti-
cle of proper constant mass m "# 0 are constrained into the mass hypersurface
in four-momentum space given by
f3 == {p: dk'PkP, + m2 = O}. (5.3.9)
This hypersurface is called a pseudo-three-sphere and is plotted in Figure 23.
FIGURE 23. Pseudo-three-sphere representing a mass hypersurface.

This hypersurface has two disconnected sheets. For the upper sheet we
have P4 = JPr+pi + p~ + m 2 > 0, and for the lower sheet we have P4 =
-J pi + pi + p~ + m 2 < O. Usually the upper sheet is ignored from physical
considera tion.
Along a twice differentiable world-line of a massive particle, four-accelera-
tion vector components are defined to be d 2q'k(S)/ds 2 Note that
dk.~q'k(S)d2q'j(S) = ~[(1/2)dk.dq'k(S)df![j(s)J
J ds ds 2 ds J ds ds
d
= -( -1/2) == O.
ds
Therefore, the four-velocity is always M-orthogonal to the four-acceleration
vector.
Now we shall postulate the relativistic equation of motion. For a particle
under external force, the axiom of four-vector equation of motion (in terms of
components) is
d[ df![k(S)] k (5.3.10)
ds M(s)~ = F (x,u)lxk=;k(S),uk=d;k(S)/ds,
where x == (X 1,X 2 ,X 3 ,X 4 ), U == (ul,u 2 ,u 3 ,U 4 ), and Fk(x,u) stand for the com-

ponents of the external four-force field. This law is the natural generalization
of Newton's equations of motion (5.1.1).
Theorem (5.3.2): The proper mass M(s) is a constant along the massive world-
line if and only if
(5.3.11)
Proof: From (5.3.10) we get

dM (s) dq'k(S) ( ) d 2 f![k(S) _ k( )
-~~ +M s ------;[T - F x,u I.. '
Multiplying the above equation with dk1(dq'I(S)/ds) and summing we have

dM(s) dkl ~f![k(S) dq'I(S) + M(s)d k1 d 2 fZk(S) df![l(.~
ds ds ds ds 2 ds
k df![I(S)
= dk,F (x, U)I .. --as'
which reduces to [via (5.3.4)]
dM(s) (-1) + (1/2)M(S)~[dk' dq'k(S)df![I(S)]

ds ds ds ds !
df![k(S) I
= dkl~F (x,u)I'"
Thus
_ d~S(S) = dkl d~:(S) F'(x, u)I .. '
Therefore, dM(s)/ds == 0 if and only if dkl(dq['k(s)/ds)FI(x, u)I" == O.
Corollary (5.3.1): If M(s) is a positive constant along the world-line of a parti-

cle and the external force components rex, u)I" are not identically zero, then
Fk( .. )1" represent a space like vector along the world-line.
Proof: From the preceding theorem we conclude that

dg{k(S) I _
dkl~F (x,u)I .. = O.
Therefore, dg{k(S)/ds and F'(x, u)I" are M-orthogonal along the world-line.
But dg{k(S)/ds for a massive particle is timelike. Thus by Theorems (Ll.4) and
(1.1.6) it follows that F'(x, u)I .. is spacelike.
Now we shall physically interpret the relativistic equation of motion

(5.3.10) by comparing and contrasting it with the Newtonian equation of
motion (S.1.1). Suppose that we have a differentiable function f along the
world-line. By (5.3.5), (S.3.6b), and the chain rule, we get
f(s) = f[Y(t)] == f#(t},
(5.3.12)
Putting the above equation into the relativistic motion law (S.3.10) we can
have
d
dt [M#(t)va(t) [1 - Dpy v P(t)V Y(t)]-1/2]
= [1 - DpyVP(t)vY(t)]1/2Fa(x,u')I ..
== f#a(t, X, v')lxl'= :V1'(t), vl'=d:I""(t)/dp (S.3.13a)

and also
d
dt [M#(t)[1 - DpyVP(t)VY(t)]-1/2]
= [1 - DpyvP(t)vY(t)]1/2F4(x, u')I'" (5.3.13b)

To interpret physically (S.3.13a) we introduce the following concepts:
Moving mass == Mv(t) == M#(t)[1 - DpyVP(t)vY(t)rl/2,
Relative three-force == f#a(. ')1" (5.3.14)
== [1 - Dpy v p(t)V Y (t)]1/2p(")I ...
Therefore, equation (5.3.13a) can be written as
:t [Mv(t)va(t)] = f'~a( .. )t .
which can be recognized as the appropriate relativistic generalization of the
Newtonian motion law (5.1.1). To interpret the other equation (5.3. 13b), let us
restrict the case in which M(s) = M#(t) = m, where m is constant. By Theo-
rem (5.3.2), we have
c5ap u a(s)FP( . . )t .. - u4(s)F4(. )t .. = 0,
so
Hence,
[1 - c5pyVP(t)VY(t)]1/2F4(. )t .. = bapva(t)f#P(. )t
Putting the above equation into (5.3.13b) we get
m :t [1 - c5pyVP(t)vY(t)rl/2 = c5ap va(t)f#P()t ,.. (5.3.15)
Comparing the above equation with (5.1.2), we can interpret (5.3.15) as the
statement "the rate of increase in the relative energy equals the rate of work
done by the external relative force." As a result of the preceding statement, it
is reasonable to define the following quantities:
Relative energy == E(v) == Mv
== m[1 - bpy vPv rrl/2
~m,
Relative kinetic energy == T(v) == E(v) - E(O) (5.3.16)

= m{[I- bpy v Pv rrl/2 -I}
= (1/2)mbapvavp + ~1(lvI4),
Rest mass == E(O) == Mo = m.
Modern experimental physics confirms the notions inherent in (5.3.16) and
(5.3.14). Note that, if we reinstate c as the speed oflight, then equation (5.3.16)
can be expressed as E(v) = MvC2 so that E(O) = M oc 2 , or simply E = Mc 2.
This last formula is perhaps the most popularized formula in the history of
science.
Example 1: We choose a free massive particle. For this case FP( .. )t . == O.

Therefore, by Theorem (5.3.2), M(s) = m is a positive constant. So the relativ-
istic equations of motion (5.3.10) yield that the acceleration four-vector is
zero, or
(5.3.17)
This equation was assumed in (2.2.2) (with t = s). Integrating (5.3.17) we have
Xk = 2lk(S) = tks + b k where the eight constants of integration tk, b k satisfy
dk1tkt l = -1, t 4 > 0, and otherwise are arbitrary. The constants t k corre-
spond to an arbitrary point on the upper sheet of a hypersurface like that in
Fig. 23. 0
Example 2: Consider the case of a constant relative force K > 0 acting on

a massive point particle in the direction of the xl-axis. We make further
simplifying assumptions, namely, M(s) = m, and v 2(t) = v3(t) = f#2(. ')1 .. =
f#3( .. )1" == O. The equations of motion (5.3.13a) yield
d
([i [v 1 (t)[1 - (v 1(t))2r lI2 J = (K/m) > O.
Integrating, with the initial condition vl(O) = (d21# 1 (t)/dt)I'=O = 0, we get

v l (t)[1 - (v 1(t2rlI2 = (Kt/m).
Thus
vl(t) = Kt[m 2 + (Kt)2rl/2.
It can be noted 0::; v1(t) < I for t E [0,00) and limt~oo vl(t) = 1, the speed of
light. Therefore, even a constant forward thrust on a massive particle does
not allow it to actually attain the speed of light! Integrating once more with
the initial condition ,q[# 1 (0) = 0 we have for 0 ::; x and 0 ::; t
[(K/m)x l + IJ2 - [(K/m)tJ2 = 1.
The above conditions yield a part of a hyperbola in the 2-flat given by
x 2 = x3 = O. The world-line of the motion curve approaches a null-line
asymptotically. 0
The equation of motion of a massive charged particle in an external elec-

tromagnetic field is governed by the Lorentz equations of motion. The
Lorentz four-force field is given by
Fi(X, u) == eFij(x)u j , (5.3.18)
where Fij(x) = - FJJx) and e is the charge parameter. Note that
uiFi(x, u)l .. = euiujFij(x)I" == O.
Thus by Theorem (5.3.2) the proper mass M(s) = m is a constant along the
motion curve. The Lorentz equations of motion by (5.3.10), (5.3.18) are
d 2,q[k(S) . k
m---;jT = euJF iX)I" . (5.3.19)
The three components of the above equations, by (5.3. 13a), yield (5.3.20)
d
m([i [v a (t)[1 - bpyVP(t)vY(t)rl/2J
(5.3.20)
where Ea(x) == F a4 (x), Ha(x) == (1/2)sapyFPY(x), and Sapy is the totally antisym-
metric permutation symbol with SI23 = 1 [compare (1.3.12)]. The compo-
nents P(x) and Ha(x) are those of the electric and magnetic vector fields
respectively. Equation (5.3.l3b) in this case yields
d
m dt [[1 - OpyVP(t)vY(t)rl/2]
= eOa/lva(t)EP(x)lx>=P>(tJ.x4=t. (5.3.21)
The right-hand side is the rate of work done by the electric field on the
charged particle. The rate of work done by the magnetic field is exactly zero.
Example: We shall work out a particular problem for motion curves of a

charged elementary particle in a cyclotron. There exists a constant magnetic
field along the x 3 -axis. Only uniplanar motions in the (x l -x 2 )-plane are con-
sidered. Thus we choose
HI(X) = H2(X) == 0, H3(X) = h, a constant,
P(x) == 0, v3 (t) == O.
Putting the above conditions in (5.3.20) and defining the speed
vet) == {[V I (t)]2 + [V 2(t)]2}1/2,
we obtain the following two equations:
m :t[vl(t)[l - (v(t)fr l /2 ] = ehv 2 (t) = ehv(t)sin0(t),
d
m dt[v 2 (t)[1 - (v(t))2rl/2] = -ehvl(t) = -ehv(t)cos0(t).
Here tan 0(t) == v 2 (t)/V 1 (t). Multiplying the first equation by cos (t) and the
second by sin 0(t) and adding we get
d
cos0(t) dt [v(t) cos 0(t)[1 - (v(t))2r l / 2 ]
+ sin0(t) :t[v(t)Sin0(t)[1 - (v(t))2rl/2] = o.

Hence,
which implies that vet) = vo, a constant. Putting the above result in either of
the equations of motion we obtain
Thus the particle describes a circular orbit in the (x l -x 2 )-plane with con-
stant angular velocity w. Measuring experimentally the radius of the circular
orbit ro == (vo}/w and the time period To == 2nlw of a revolution, one can
determine an important property of the elementary particle:
(elm) = -(wlh) [1 - (vo)2rlI2
= -(2nlh)[(To)2 - (2nro)2rlI2. 0
EXERCISES 5.3
1. (i) Express the equation of motion (5.3.13a) for M#(t) = m, a constant, in
the form
where v == (V l ,V 2 ,V 3 ).
(ii) Obtain the eigenvalues of [Mafi(v)], and discuss the physical meaning of
these eigenvalues.
2. Consider the Lorentz equations of motion (5.3.20) in absence of any
electric field (P(x) == 0). Prove that these equations can be reduced to
e m)eafiyv fi( t )HY(X)Ix'= ,f#'(t).

[1 2]-112 dva(t) -_ (I
- V ----;Jt X4=t>
5.4. The Relativistic Lagrangian and Hamiltonian

Mechanics of a Particle
We shall now study the relativistic mechanics of a particle in both the La-
grangian and the Hamiltonian formulations. For that purpose we cannot use
the Lagrangian L'(t, x, t', x') from equation (5.2.2), since it is not Lorentz
invariant (a scalar under Lorentz transformations). We need a new Lagran-
gian A(x, u) in terms of the space-time variables x == (Xl, x 2 , x 3 , x 4 ) and u ==
(u 1 , u 2 , u 3 , u4 ). A(x, u) is assumed to have the following properties:
(i) A E ,2(D), where D c IRs is a domain corresponding to that of an ex-
tended space.
(ii) A is a Lorentz invariant, i.e., a scalar under Lorentz transformations.
(iii) A is a positive, homogeneous function of degree one in the variables Uk.
We may define a metric gij(x, u) on the extended space, which is associated
with the Lagrangian A(x, u), by the requirement
8 2A2(X,U)
gij(x,u) == -(1/2)-au i8u j = gji(X,U).
5.4. The Relativistic Lagrangian and Hamiltonian Mechanics of a Particle 109
It is assumed that det[gij] =I 0 [compare the inequality in (5.1.7)]. Note that

from property (iii) of A(x, u) one can show that gij(x, u) is a Finsler metric (see
the book by Lovelock and Rund referred at the end).
By property (iii) and Euler's theorem on homogeneous functions we obtain
[cf. equation (5.2.6)]
koA(X,U) _ A( ) (5.4.1 )
U OUk - x,U.
The relativistic invariant action integral for a smooth curve y is
= 1<2
<,
[A(x, u)I'.] dr. (5.4.2)
This action integral is invariant under the reparametrization
dyer) 0
i = Y(r), ~>, (5.4.3)
The Euler-Lagrange equations from (5.4.2) are
[OA(X~
ax
u)] _ ~ {[aA(X~ u)] } = o.
I.. dr OU I..
(5.4.4)
For a timelike curve dkl[d~k(r)/dr] [d~l(r)/dr] < 0, one can choose r = s, the
proper time parameter, in (5.4.4). The four equations (5.4.4) are not indepen-
dent as there exists one identity among them. Using (5.4.1) we find
d~k(r) {[OA(X~ U)] _ ~ [OA(X~ u)] }

dr oX I.. dr OU I
= d~k(r)[oA(X'U)] d2~k(r)[oA(X'U)] _ ~[- kOA(X,U)]

dr ox k I.. + dr 2 auk I.. dr _U OUk I
dA(x,u) dA(x,u)
dr I dr I
=0.
Example: Consider the simple case of a massive (m > 0) free particle. The
relativistic Lagrangian is chosen to be
A(x,u) =-m/dklukul/1/2.
(5.4.5)
Thus
oA(x, u) = 0
OX k -
oA(x, u) _ kid j '1- 1/2

OUk - mu j'U U ,
_ o2A2(X,u) 2
gk'(X, u) = -(1/2) oukou' = m dkh
det[gk,] == _m 8 f= O.
The Euler-Lagrange equations (5.4.4) yield
d {[mu kidj,U j U '1- 1/2 ] I }

O - d! -- 0.
After putting! = sand [dk,UkU']luk=d.:f"<S)/dS = -1, we obtain

d2 f![k(S)
m~=O.
which are the same equations as in (5.3.17). D
A very general class of the relativistic invariant Lagrangians is provided in

the following theorem.
Theorem (5.4.1): Let Vex), Ai(X), Sij(x), rPijk(X), I/Iijk'(X) be various totally sym-
metric M inkowski tensor fields of class C(/2 on their domains. Then, for arbi-
trary constants co. c l , C2, C 3 C4 , the Lagrangian
A(x, u) == [1 + Co V(x)]
x [clIAi(x)uil + czISij(x)uiujI1/2
+ c3IrPijk(x)uiuiukI1/3 + c411/1ijk,(x)Uiujuku'11/4], (5.4.6)
is a relativistic invariant and a positive homogeneous function of degree one in
the variables Uk.
Proof: The proof is straightforward and is left to the reader.
Now we shall develop the relativistic Hamiltonian particle mechanics.

First we define the generalized four-momentum variables by
oA(x,u)
Pi == OUi (5.4.7)
The coresponding relativistic "Hamiltonian" is

. .oA(x,u)
PiU' - A(x, u) = U OUi - A(x, u) == 0,
by (5.4.1) [analogous to the identity (5.2.6)]. So we need another function in

the place of this "Hamiltonian." To obtain that function, we first notice that
by differentiating (5.4.1) we get
.0ZA(x,u) OUi oA(x,u) oA(x,u)
u ' - -.-.--+ -~'---.- = ..
ouJou' ou J ou' ou J
This is equivalent to
OZ A(x, U)] i_
[ . . U - 0
ouJou'
for all u E D C ~4. Thus we see that for nonzero solutions of above linear
homogeneous equations on ui's we must have
. .U)] = de[OPi]
d et [ 01 A(x, t . = 0.
ouJou' OU'
Thus all the Pi are not independent and there must exist at least one Junctional
relationship among them, i.e.,
O(x,p) = 0, (5.4.8)
where we write p == (Pl,P1,P3,P4)' The above relationship denotes a con-

straining seven-dimensional mass hypersurJace in the eight-dimensional ex-
tended phase space M 8 . The function 0 is called the "Super-Hamiltonian"
function.
The relativistic canonical action integral that generalizes (5.1.17) and (5.2.8)
can be expressed involving a Lagrange multiplier as
(5.4.9)
The corresponding Euler-Lagrange equations yield
dPJc) = -[A OO(X;P)] = -1(T) [on(x;pfl '

dT oX I.. oX _I ..
(5.4.10)
d:ri(T) = [A on(x,p)] = I(T)[~n(x,p)] .
dT 0Pi I.. 0Pi I ..
If we reparameterize the curve by the special parameter (arbitrary up to addi-
r
tion of a constant) defined by the indefinite integral
() = SeT) == l(~)d~,
~i(() == ~i(r), (5.4.11)
iWJ) == Pier),
then the equations (5.4.10) transform into the relativistic canonical form:
diWJ) = _[on(x: p)]
de oX' I..
(5.4.12)
d?[i(e) = [on(x,p)]
de op; I.. '
Now we shall illustrate this method by some examples.
Example 1: We shall consider the simple case of a free massive particle with
the Lagrangian given in (5.4.5) [which is a special case of (5.4.6)]. In that case
the four-momentum variables are given by
Pi = mdijujldklukull-l/2,
Pa = mbapvflll - b,uv'v ul- 1 / 2 ,
(5.4.13)
P4 = -mil - (j",v'v ur 1 / 2,
va == u a/u 4.
The Pi and the va are homogeneous functions of degree zero of the Uk. Elim-
inating three ratios va we obtain
dk1pkPI = (ja Pp pp - (P4)2
= m2 11 - (j,uv'vurl[(japvvfl- 1]
(5.4.14)
n(x,p) == (1/2m) [dklpkPI + m2] = O.

The corresponding canonical equations (5.4.12) yield:
dPi(e) = 0
de - ,
d?[i(e) .. _
~ = d'JPie)/m,
d2 ?[i(e) = (1/ )dijd~(e) = 0

de2 m de .
Comparing the above equations with (5.3.17) we conclude that

e = s + c, (5.4.15)
where s is the proper time parameter along the curve y and c is an arbitrary
constant. 0
Example 2: We choose the case of a massive charged particle in the external

electromagnetic field. In this case, the Lagrangian is assumed to be
A(x, u) = eAk(x)u k - mldklukulll/2, (5.4.16)
which is a kind of special case of (5.4.6).
The corresponding four-momentum components are
Pi = eA;(x) + mdiiuildkIUkUII-l/2,
Pa = mbapvPII - b,O'v'vO'I- 1 / 2 + eAa(x),
(5.4.17)
P4 = -mil - b,O'v'vO'I- 1/ 2 + eA 4 (x),
Eliminating three ratios va we get
dkl[Pk - eAk(X)] [PI - eA/(x)]
= m2[bapVaVP - 1]/[1 - b,O'v'vO'] = _m 2,

(5.4.18)
n(x,p) == (1/2m){d k1 [pk - eAk(x)][PI- eA1(x)] + m2 } = O.
The corresponding canonical equations (5.4.12) yield
dPa(e) = _[an(x,p)] = (/ )A [k _ eAk(x)]

de a a X I..
e m k.a P I .. '
(5.4.19)
dlJiace)
dO
= [an(x,p)]
amp
= (1/ )[ a _ eAa( )]
x I.. '
Pa I ..
Eliminating Pace) in the above equations we obtain

dpa(e) d2lJiace) dlJik(e)
----;re = m d02 + e -d8 Aa kl
_ ab dPZk( e)
- ed ------;w- Ak,bl'"
Thus we find
d 2 pza(e) ab dPZk(e)
m de 2 = ed ------;w- [Ak,b - Ab.k] I .
The above equation is equivalent to (5.3.19) provided that

e = s + c,
where s is the proper time parameter and c is a constant. 0
Example 3: In this example we study the motion of a massive particle (m > 0)

in a gravitational field according to a special relativistic model equivalent to
Einstein's linearized gravitational theory. In this theory, the gravity is gener-
ated by a symmetric Minkowski tensor field Yab(X) such that
YabCX) = Yba(X),
IYab(X) I < 1, (5.4.20)
det[dab + Yab(X)] < O.
From the last condition we can conclude the existence of the inverse matrix
[dab + Yab(X)]-l. Thus there is a contravariant symmetric Minkowski tensor
field yab(x) = yba(x) such that
Iyab(x) I < 1,
(5.4.21)
[dab + yab(X)] [d bc + Ybc(X)] = {jac '
The relativistic Lagrangian is assumed to be
(5.4.22)
hence, we have the Finsler metric
with det[gab(x, u)] < O. The above equations are again a special case of equa-
tion (5.4.6).
For brevity we now introduce a new function S by the solution of the
ordinary differential equation
dS(r) _ b 1/2
----;rr- - [I[dab
,a
+ Iab(X)]U U I ] Ixa=cra(r),ua=d,P(r)/dr
(5.4.23)
> O.
With this notation we find the Euler-Lagrange equations to be
Thus
The above equation is equivalent to
d2~:(r) _ (1/2){[dCk + yCk(X)] [Yab,k - Yka,b - Ykb,a]UaU b} I..
_ d:!l'C(r) d 2S(r) [dS(r)]-l

- ~~ ----;rr- (5.4.24)
Now we choose the parameter r such that
d2 S(r) = 0
dr2 '
(S.4.2S)
s= S(r) = C1 r + c2 ,
where c 1, C 2 are constants of integration. This class of parameters is a special
class. The simplest choice of the special parameters is r = s. Using this pa-
rameter, the equations of motion (S.4.24) become
d2~;(S) _ (1/2){[dCk + yCk(X)] [Yab.k - I'ka.b -l'kb.aJuaubk = O. (S.4.26a)
A first integral of (S.4.26a), which follows from (5.4.23) and d!~r) = 1, is
dPr (s) d?l'b (s)

[dab + Yab(X)]lxa= 1'"a(';) ------;[ ------;[ = -1. (S.4.26b)
We note that the Euler-Lagrange equations (S.4.26a) can also be derived

from the squared Lagrangian (using r = s)
A 2 (x, u) == -(112m) [A(x, U)]2
(S.4.27)
Example 4: Now we shall work out a specific problem in tht: linearized gravi-
tation, namely, the problem of planetary motion in the gravitational field
of the sun. For that purpose, we use the squared Lagrangian (S.4.27) corre-
sponding to the linearized Schwarzschild field:
A 2 (x, u) == (mI2){ [<5ap + Yap(x)]uau P - [1 - Y44(X)] (u 4 f},
Yo;p(x) == 2M<5o;p[<5IlV XIlX vr 1/2 , (5.4.28)
Y44(X) == 2M[<5llvxllxvr1/2.
Here m > 0 denotes the planetary mass and M > 0 denotes the solar mass
multiplied by the Newtonian gravitational constant.
To make the problem easier to solve, we transform from the Minkowski
chart to spherical polar and time coordinates by
Xl = X1(X) == x1sinx2cosx3 == rsinecos~,
x 2 = X2(X) == Xl sin x 2 sin x 3 == r' sin e sin 4',

(S.4.29)
x3 = X 3 (x) == Xl cos x 2 == r' cos e,
X4 = X4(X) == X4 == t.
Using (S.4.29), the squared Lagrangian (S.4.28) transforms (see problem 2 of

Exercises S.4) to
A 2 (x, u) = A2 (r, 0, </1, t, u r , u B, U if1, u t )
= (m/2){(1 + 2M/r) [(U r )2 + r2(u B)2 + r2 sin 20(u if1 )2]
- (1 - 2M/r)(u t )2}. (S.4.30)
The Euler-Lagrange equation corresponding to the O-coordinate is
[CAcO 2 ( . )] _ ~ {[CA2~ .)] }

I.. d CU I..
= m{[(r2 + 2Mr)sinOcosO(u 9)2]1"

2 d 2 0() d0() d 2 }
- (r + 2Mr)I" ----;[2 - -----;[ d (r + 2Mr)I
= o. (S.4.31)
The above differential equation yields a particular solution
o= 8() = n/2, (S.4.32)
which corresponds to a special uniplanar motion in space. We shall choose
this solution subsequently. We note that coordinates </1 and t do not appear
explicitly in (S.4.30). So the coordinates </1 and t are ignorable. The momenta
corresponding to </1 and t are conserved. We can deduce explicitly from the
Euler-Lagrange equations
d{[CA
- d ~
2 ( )]
I..
} _
- -
d{
d m[(l + 2M/r)r2 sm
. 2 q,
Ou ]1"
}
=0
that
2 d<ll(s)
mEr (1 + 2M/r)]I"Ci. = mh, (S.4.33a)
where h is a constant. Similarly we have
_~{[CA2(")J
d out I..
}= ~{
d m
[(1 _ 2M/)]
r I
. dl"4()}
d
=0;
thus,
(S.4.33b)
a constant. Now instead of a second-order equation of motion for the r-

coordinate, we can use the first integral (S.4.26b) together with (S.4.33a) to
get
(I + 2M/r)I .. [d~~s)J + [h 2/(r2 + 2Mr)I'.J
- [E2/m2(1- 2M/r)I'.J = -1. (5.4.34)
Now we shall assume the constant "areal velocity" h > 0, which implies by
(5.4.33a) that d<l>(s)/ds > O. Therefore, there exists an inverse function .<i'such
that
I
s=!)(J),
(5.4.35)
d~(s! = d<l>(s) d~(!)(J = h d:7t(1) (2 2 r)
ds ds dJ dJ r + M. I'
Putting (5.4.35) into (5.4.34), we obtain
[h 2 /(l + 2M/r)I ..J {[:J (1/:7t(JJ + [1/:7t(J)] 2 }

- [E2/m2(1 - 2M/r)I'.J = -1. (5.4.36)
Making the inversion
u = I/r > 0,
(5.4.37)
u = U(J) == 1/:7t(J),
equation (5.4.36) transform into
{[I - 2MU(J)]/[1 + 2MU(J)]} {[d~~J)J + [U(J)]2}

+ [1 - 2MU(1)]jh 2 = E2/m 2h 2. (5.4.38)
We can solve this nonlinear first-order equation exactly in terms of elliptic
functions. But it is easier to solve by the perturbation method, the second-
order equation obtained by differentiating (5.4.38). It is not difficult to extract
the physically relevant effects from the second-order equation
un + u = (M/h2) [1 + 2Mu]/[1 - 2Mu] + 2M[(u')2 + u 2 ]/(l - 4M 1 u1 )
= (M/h2) + (4M 2u/h2) + 2M(1 + 4M 2 u 2)[(u'f + u1 ]
(5.4.39)
For brevity we have written u' == dU(J)/dJ, un == d 2U(J)/dJ2. For all practi-
cal purposes, we can ignore the higher-order terms since (M/h) is a small
number for all the planets in the solar system. We shall solve (5.4.39) by the
method of perturbations, namely,
u = U o + U 1 + ... ,
u~ + U o = M/h2,
(5.4.40)
u'{ + U 1 = 4M 1 uO /h2 + 2M[1 + 4M2U~] [(U~)2 .... un
The general solution for U o (the Newtonian approximation) is

Uo = llo(rjJ) = (Mjh2) [1 + e' cos(rjJ - ill)]. (5.4.41)
Here e, ill are constants of integration and represent the eccentricity and
angle of perihelion of the orbit respectively. The orbit is circular, elliptic,
parabolic, or hyperbolic if e = 0, 0 < e < 1, e = 1, 1 < e respectively. We
shall consider the elliptic case 0 < e < 1, which corresponds to the planetary
orbit. Putting (5.4.41) into (5.4.40) we get the differential equation for the
first-order correction as
U'{ + U 1 = 2(3 + e2)M 3 h- 4 + 8M 3 h- 4 e'cos(rjJ - w) + (!)(M 7 h- S ). (5.4.42)
Neglecting the (!J(M 7 h -S) term, the particular integral of (5.4.42) is
U1 = !l/il(rjJ) = 2(3 + eZ )M 3 h- 4 + 4M 3h- 4 erjJsin(rjJ - w). (5.4.43)
Putting (5.4.41), (5.4.43) into (5.4.40) we have (neglecting higher-order terms)
= (MjhZ)[l + 2(3 + eZ)M 2 h- 2 + e'cos(rjJ - w) + 4M 2 h- 2 erjJsin(rjJ - w)]

= (MjhZ)[l + 2(3 + e1)MZh-l] [1 + ecos(rjJ - w - bw)], (5.4.44)
where
e == e[l + (4M2h-ZrjJ)2] 1/2 [1 + 2(3 + e1)M1h-1rt.
bw == Arctan(4M 1h- zrjJ) = 4M 2 h- zrjJ + (!)[(4MZh-zrjJ)3].
The elliptic orbit is distorted in the sense that the eccentricity e and the
perihelion angle wchange as the planet moves around the orbit. The change
of the eccentricity is too small to measure. The perihelion angle ill changes at
the rate bw per radian of revolution. Therefore, the perihelion shift per full
revolution is
(5.4.45)
where
a == semimajor axis of the orbit,
e == eccentricity of the orbit,
To == period of revolution of the orbit (planet's year).
For the planet Mercury, the above amount is a little over 57" per century,
which is slightly larger than the observed effect. Einstein's theory of general
relativity, without linearization, yields the exact perihelion shift 43" per cen-
tury (within the limit of present day observations). 0
EXERCISES 5.4
1. Consider the relativistic Lagrangian
A(x,u) = -ml[d ab + Yab(X)]u au b I1 / 2

Set up explicitly the corresponding relativistic action integral JH[y] as given

in equation (5.4.9).
2. Consider a general coordinate transformation from one chart to another
given by
8(xl, ... ,X4) 0

8(x , ... ,x ) i= .
1 4
Suppose that the Lagrangian remains invariant, i.e., A(x, 12) = A(x, u). Prove
that the Euler-Lagrange equations from both Lagrangians are equivalent.
3. Consider the relativistic canonical equations
dPk(s) = _[8n(X,p)] ,
ds 8Xk I..
d~:~ = [8n8~~~1 ..
Consider a general coordinate transformation in the extended phase space
Ms ,
Xk = ~k(X, p),
8x k /8 pr j
pk = I'/k(X, p), 8pk/8pl_
for all (x, p) E D c. IRs. Prove that the relativistic canonical equations trans-
form covariantly provided that JTrJ = r, where r is the 8 )( 8 matrix whose
block decomposition is
References
1. P. G. Bergman, Introduction to the Theory of Relativity, Prentice-Hall, New Jersey,
1942. [pp. 134, 158J
2. P. A. M. Dirac, Canad. J. Math. 2 (1950),129-148.
3. I. M. Gelfand and S. V. Fomin, Calculus of variations, PrenticeHall, New Jersey,
1963. [po 128J
4. C. Lanczos, The variational principles of mechanics, University of Toronto Press,
Toronto, 1977. [pp. 126, 128, 131, 134, 136, 137]
[pp. 141, 145, 147, 150J
6. - - , Classical dynamics, Reprint from Handbuch der Physik, Springer-Verlag,
Berlin, 1960. [po 157J
6
The Special Relativistic Classical
Field Theory
6.1. Variational Formalism for Relativistic Classical

Fields
We shall define the Lagrangian function for a relativistic classical field. Let
15 c: 1R4 be a bounded region corresponding to a region [J in space-time M 4
Let F: 15 --+ W be a twice differentiable function, i.e., FE 1{f2(15; W) (see Fig-
ure 24).
Let r/JR == n R 0 F, where R is a multi-index taking values in {l, ... ,p}. We
shall use capital indices R, S, T, U, V, W as multi-indices in the sequel. A
multi-index is simply the concatenation of some finite number of indices
taking values in some fixed range. Since r/JR: D --+ ~, we can associate yR =
r/JR(x) with components of a Minkowski tensor field.
Let the Lagrangian function .P: 15 x ~p X ~4p --+ ~, and furthermore as-
sume that .P is twice differentiable. The real value of this function is denoted
by .P(x, yR, yRJ Let the action functional d be a function d: 1{f2(15; ~P) --+ ~
such that
d(F) == IIi .P(x,yR,yR j )lyll=;R(X),yR;=q1R.;d 4 x. (6.1.1)
It is further assumed that d is a totally differentiable function. A "slightly

varied" function F + eh, which has the same boundary value, is defined by
yR = JR(x) + ehR(x),
(6.1.2)
hR(x)IOD == O.
Here hR is assumed to be twice differentiable and e > 0 is a small positive
number. The variation of the action integral (6.1.1) is given by (summation
convention on index R is used),
c5d(F) == d(F + eh) - d(F) = IIi [.P(x, yR, yRi)II - .P(x, yR, yRi)l~] d 4 x
(6.1.3)
120
6.1. Variational Formalism for Relativistic Classical Fields 121
FIGURE 24. The mapping corresponding to a tensor field t,6R(X).
where f. is the submanifold yR = (JR(X) + 6h R(x), yRi = IpR,i + 6h R,i' and

where ~ is the submanifold yR = r(x), yRi = (JR,i' We introduce now the
notation for a total partial derivative:
d R R
dxJf(x,y ,y j)II]
== of(. :)
'" 1
uX II
+.l.R .
'I' ,I
[Of(. ')J
'" R
uy II
+.l.R
'I' "
).[of(")J
'" R
uy j II '
(6.1.4)
where ~ is the submanifold yR = r(x), yRi = (JR,i' With this notation, (6.1.3)
becomes
+6 f [8.P( .. ) ]
aD
~
UY i II
R'
h (x)n'd 3 0' + &(6 2 ).
Here we have used Gauss's Theorem (2.4.5). Thus we find that
dd(F) = 6f- hR(X){[O.P(~')J - ~[o.P~')J }d

D oy II dx oy i II
4 x + &(8 2) (6.1.5)
since hR(x)laD == O.
122 6. Special Relativistic Classical Field Theory
If we demand the stationary (or critical) value of the action integral d(F),
then we must have lim.~o+ [bd(F)/e] = O. Thus we obtain
f-hR(X){[B!l'(~')J
D By 11:
- ~[B!l'~')J
dx By i 11:
}d4-x = 0 (6.1.6)
for all hR(X) such that hR(x)laD == O. To obtain the Euler-Lagrange field equa-
tions from (6.1.6) we need the following lemma.
Lemma (6.1.1): Let orc 2 (i5; IRP) == {H: HE rc 2 (i5; W), H laD = O}. Let G E
rc 2 (15; IRP) such that fngR(X)h R(x)d4- x = 0 for all HE orc2 (15; W). Then each
of the gR(X) == 0 on 15.
Proof: Choose HE orc 2(15; W) so that h2(X) = h3(X) = ... = hP(x) == 0 in 15.
Assume contrary to the lemma that fngR(X)hR(x) d4-x = 0, but g(l)(~) # 0 for
some ~ E 15. Suppose that g(l)(~) > 0 (the case of a negative value can be
treated similarly). From the assumption of continuity, there exists a neigh-
bourhood N~(~) c D such that g(l)(~) > 0 in N~(~). The neighbourhood N~(~)
is exterior to the hypercube
C(4-) == {x: -b/2::; Xk - ~k ::; b/2, k = 1,2,3, 4}.
We define a twice differentiable function h(1)(x) on 15 by
h(l)(X) == {O 4-k=l (k
X -
k
~ -
b)4(
-2 x k- ~
k+ -2b)4 c
lor
.
x 1D C(4)'
o for x in 15 - C(4)'
Therefore, the resulting test function H belongs to orc 2(15; [RP). Thus
In gR(x)h R(x)d4- x = In g(l)(x)h(I)(x)d4-x
= f.
C(4)
g(l,(x)h(1)(x) d4-x
>0.
The above inequality contradicts the starting hypothesis that the above inte-
gral is zero.
Theorem (6.1.1) (Euler-Lagrange): Let the action integral d(F) be given by

(6.1.1). If the stationary condition lim.~o+ [bd(F)/e] = 0 holds, then the Euler-
Lagrange equations:
[B!l'(~')J _~{[B!l'~')J }=O (6.1.7)

By I" dx By I" i
must be satisfied in the domain D c 1R4-.
Proof: The proof is straightforward from equation (6.1.6) and the preceding
lemma.
Example: We consider the case of a real scalar wave field (x). The relativistic
Lagrangian f: 15 x IRs -+ IR is defined by
f(X,Y,Yi) == -(1/2)d kI YkY"
of(x, y, YJ = 0
oY -,
of(x, Y, yJ dkl
Oh = - y"
det [02 f(x, y, yaJ = _ 1 # O.

hOY,
The Euler-Lagrange equations (6.1.7) in D yield
[Of~;'Y)l=(x),Yj=q\,j - fxi{[o~~'Y)l=q\(x).y-,=J
= -~
dx {_dijJ.
l
-} = d ij _ = OJ.(x)
'f',j 'I'
.I}
= O.
The complex fields ",R(X) describe the wave mechanics of I~harged particles
in nature (like Dirac's bispinor field). The variational method has to be
slightly extended for such a field. Suppose that
",R(X) = rex) + iI1 R(x),
R(X) == Re[",R(x)],
I1R(X) == Im[",R(x)].
We introduce the Lagrangian function f: 15 x IRP x IRP )( 1R 4 p X 1R 4 p -+ IR
so that the Lagrangian field f(x, yR, w R, yR k' WRk) is real-valued.
We introduce complex-conjugate coordinates so that
(R = yR + iWR,
(R=yR_iwR,
yR = (1/2)[(R + (R],
w R = _(i/2)[(R _ (R],
_0_ ==
oe
(1/2)[~
oyR
- i~J,
ow R
(6.1.8)
Consider the Euler-Lagrange equations derived from the Lagrangian func-

tion ft',
namely,
[0ft'(~
oy
.)J - ~
I..dx
{[o ft'~
oy
.)J } 0, i I..
=
[ Oft'(~.
oW
)J - dx~ {[Oft'(~.
I.. ow
)J } = o. i I..
A complex combination of the above equations can be written as
{[ Oft'(~')J
oy I..
- i[Oft'(~')J
oW I ..
}
- dx~{[Oft'~')J
oy
- i[Oft'~')J
k oW } = O.
I.. k I ..
Therefore multiplying by 1/2 and using the definitions above we get
[o.\')J - ~{[02~')J }=
0, I.. dx 0, k I ..
O. (6.1.9)
Example: Consider a relativistic Lagrangian function for a complex scalar

field I/t(x) such that
where 9 is a real constant and n is a positive integer. Thus we have
o~~..) = gn("r 1"
02( .. ) = _dkl"f
O'k "I'
Consequently the Euler-Lagrange equations (6.1.9) yield
[02(")J
0, - ~{[02(")J }
I.. dXk O'k I..
= gnll/t(x)1 2 (n-l)i//(x) + oi//(x)
=0.
Taking the complex conjugation of the last equation, we have the nonlinear
wave equation
Ol/l(x) + gnll/l(x)1 2 (n-l)l/I(x) = o. 0
Subsequently the hat on a Lagrangian 2( .. ) for complex fields will be
omitted.
Now we shall discuss the invariance of the action integral (6.1.1) under
a coordinate transformation and the consequent conservation equations of
IR
FIGURE 25. Two coordinate charts and a tensor field mapping.
Noether. For that purpose consider two charts in M4 and the corresponding
two expressions of the action integral (6.1.1) (see Figure 25).
From these mappings we conclude that
Xk = [nk 0 X0 X-I] (x) == gk(X) == Xk + ~k(X),
yR = [n R 0 F] (x) == q>R(x),
kl' "(x) == R(X)
= [n R 0 Fox 0 X-I] (x)
~k
a~x~;'
]
= [ q>il:::(X) (6.1.10)
= [(bk1j, + ~kli,)'" ]q>i! :::(x)

= q>k,:::(X) + hk,:::(x)
= q>R(X) + hR(X).
d(F + h) - d(F) = Ii5 2(x, yR, yR;)lf d4X

- ID 2(X,yR,y RJII:d 4x,
where f is the submanifold yR = q>R(X) + hR(X), yRi = q>R,i + hR,i' and ~ is
the submanifold yR = q>R(X), yRi = q>R,i'
We have to consider an "infinitesimal" coordinate transformation i 0 X- 1 =

1+ 3, where I is the identity mapping and 3 E rt2(D; 1R4) is in some sense
small. This will induce a mapping hE rt 2 (D;RP) such that h is "small." We
can make precise the notion of "smallness" in the following equations:
ek(x) = t:(l' letl'l(x) + r~(t:l, . .. , t:'), (6.1.11)
where J1. is summed from I to rand 1t:(l'l l are sufficiently small positive num-
bers. The remainder term r~(t:1, ... , t:') must satisfy
lim r~(e1, ... , er)/e = 0,
.t:-+O+
with t: == [(t: 1)2 + '" + (t: r)2J1/2. The small variations are summarized in the
following equations:
c5x k == :x k - Xk = e k = t:(l' letl'l(x) + r~(t:1, . .. , t:'),

:1(A1 .~2 .~3 A4)
u X ,A ,A , X _ (>: i) k( 1
o( 1 2 3 4) - 1 + uX ,j + ro t: , ... , t: ,
')
X,X,X,X
c5q>R(x) == ~R(:x) - q>R(x) = hR(X)
= t:(l'lh~l(x) + rR(t:\ ... , 1>'),
[fJq>R(X)J.i = B(/J)h~l,i + rRi(Bl, ... , B'),
(6.1.12)
fJLq>R(X) == ~R(X) - q>R(x)
= [~R(:x) _ q>R(X)] _ [~R(:x) _ ~R(X)],
fJ[q>R,J = c5L[q>R,J + q>R,iic5xi + pRi(B 1 , ... , Br).

Here all the remainder terms must satisfy an equation like
lim r(t: 1, ... , B')/e = O.
Now we are ready to discuss the Noether's theorem.
Theorem (6.1.2) (Noether's theorem): Suppose that the action integral d(F), as
defined in (6.1.1), is invariant under the transformation of coordinates in every
domain D c: D. c: 1R4. Furthermore, let the field functions q>R == n R 0 F satisfy
the Euler-Lagrange equations (6.1.7). Then the following partial differential
equations will hold in D:
d { e(l'l(X)
(1l i
[!!'(")]I" + [O!!'(
~
.. )] R
[h(I')(x) J
- e(/Jl(x)q> R
)x)] } -_ O. (6.1.13)
x Y, I..
Proof: From the in variance of the action integral and equations (6.1.10),
(6.1.12), it follows that
o= d(F + h) - d(F)
= f{_
D
R
2'(x+Jx,y +Jy,y ;+JY;)I"
R R R 0(X 1 ,X 2 ,X 3 ,X 4 )
o( 1
X,X,X,X
2 3 4)
- 2'(X,yR,yR;)I .. }d 4X
= Iv {[2'(x + JX,yR + JyR,yR; + JyRJ 1.'] [1 + (Jxj),J

- 2'(X,yR,yRJI .. }d 4x + rl(el, ... ,e')
= Iv {2(x + JX,yR + JyR,yR; + JyRJ 1..
- 2'(X,yR,y RJ, .. + (Jx j ),j2'(X,yR,yR;)I .. }d 4x + rl(e\ ... ,e')

= f-D
{[02'(;
ox
')J,.. Jx; + [02'(~')J
oy I ..
JtPR(X)
02'( ..
+ [ ~- )J J[tP R ,J + (Jx j ),j2'(X,y R ,y R J 1.. } d 4x .+ I\(e
' I r
, ... ,6)
y, I"
= f {[Jx ;2'( .. ),; + [02'

_
D
~
y
(")J I ..
[JLtP R + Jx itP R ,J
+ [ 02'(.
~ ')J i} R
[JL(tP') + tP R ,jiJX] d4 x ~ I , ... ,6),
+ rl(e
f{
Y J I ..
= _
d.
dx;CJx'2'( .. )] + [02(")J R+ [02'(")J
~ JLtP ~ JL(tP
R}
,J d4x
D Y I.. y, I ..
+ f1 (e\ ... ,e')

-f {d (
- _ d~ [bx 2( .. )]
D x
;. + [02'(")J~
uy, ,..
JLtP R)
+ ([02'(")J
~
uy I..
-d~ ~
d
x
[02(")J) uy, I ..
JLtP R} dx+rl(6,
4 ~ 1... ,e)., (6.1.14)
Using the Euler-Lagrange equations (6.1.7) we have
0= e(/l) Iv {d~i([~t)(X)2'( n. ..
From the assumption that the action integral .Pi is totally differentiable, it
now follows that
0= lim [13(1')13- 1
,-0+
f- D
{dd
X
i([~:I')(x).:t'c .. n..
+ [a:~
y,
. )] I
[ht)(x) _ ~/1')(x)R.J)}d4x
+ f2 (e 1, ... ,6 )/e], 7
or equivalently
In L~i( [~fl'/x).:t'( .. )l
(6.1.15)
for J1 = 1, ... , r. Since l5 is an arbitrary contractible domain, we must con-

clude that equations (6.1.13) hold.
Now we shall discuss three examples on Noether's theorem.
Example 1: Let us consider the "infinitesimal" translation of Minkowski co-

ordinates, that is,
a(x 1 ,X 2 ,X 3 ,X4 ) k
a( 1
X,X,X,X
2 3 4) = det[J j + 0] = 1,
(6.1.16)
JR(X) = rcx), Jr == 0, hRk(X) == 0,
Noether's theorem summarized in (6.1.13) yields:
d i { J ij[.:t'cn ..
-d + [a.:t'(
-a .. )] [0 - J kj R ,k] } = 0.
R.
x y, I ..
Thus Tij,i = 0, where we define
i
T j(x)
_
= i
J j[.:t'(")]I" -
[a.:t'c
~
.. )] R
,j' (6.1.17)
Y I I ..
The second-order tensor field Tiix) for the field R(X) is called the canonical
energy-momentum-stress tensor density. It is not necessarily a symmetric ten-
sor. Equation (6.1.17) physically represents the differential conservation of
energy and momentum of the field R(X). 0
Example 2: Let us consider an "infinitesimal" proper, orthochronous Lorentz

transformation:
[ij = l5 ij + eij'
From the condition (1.2.10) we obtain
dii15ik + eid(l5 jm + ejm) = dkm ,
so that
Therefore,
emk + ekm = -dijeikejm == rkm(e ab )
Therefore, we can write
Jx k = ekjxj = eabl5\djbxj
= (1/2)e ab [(l5 kAb -15\dja )x j ] + r~(ecd)'
~;b(X) = (1/2)(Jk Ab - J\dja)x j = - ~\a(x),
O(X 1 ,X 2 ,X 3 ,X4 ) k k
o( 1 2 3 4)
x ,x ,X,X
= det[15 j + e j ] = 1,
~k:::(X) = (15 k ; + e\)"'(,V::(X),

Ml::: == e\"'(,V::(X) + ... (6.1.18)
= eabi5\dbi",pi:::(X) + ...
= (1/2)eab[(Jkadbi - d\da;} .. ]Pi::: + r~(ecd)
== (1/2)e ab [S;bi:::p i :::] + r~(ecd)'
hkab(X) == (1/2)[S;b;:::p i:::] = -hkba(X),
JLr/Jk::: = Jr/Jk - JXir/Jk:::.i - fk(ecd)
= (1/2)e ab [S;bi:::r/J i :: - (djbr/Jk:::.a - djar/Jk:::.b)Xi] + rk(eCd )'
Putting the above equations into (6.1.13) we obtain the following six equa-
tions:
(6.1.19)
Here we define ,Iiab(X) by
,Iiab(X) == {Tia(x)dbixi - Tib(x)daixi} + {[aa~('::~] [S!b/::(V::(X)}

y ... , ,..
= - ,Iiba(X).
The third-order tensor field ,Iiab(X) represents the relativistic angular-
momentum tensor density for the field tPi:::(x). It contains the orbital angular-
momentum tensor density plus the spin angular-momentum tensor density of the
field [each enclosed in curly parentheses in the definition of ,Iiab(X)]. D
Example 3: This example involves a complex scalar field ",R(X). We consider

the "infinitesimal" global gauge transformation characterized by the follow-
ing equations:
a( XA1 ,xA2 , ....-,,3 ,xA4) _ 1

a(xl,x 2 ,x 3 ,x4 ) - ,
(6.1.20)
~R(X) = eiO",R(x) = ",R(X) + is",R(x) + rr(s),
b",R(X) = '1R(X) + rr(x),
'1R(X) == - is",R(x).
The complex generalization of Noether's equations (6.1.13) for this case is
d{ k
0= dx k ~ (x) [.9'( .. )], .. + [0.9'( .. )] R
aCRk ,.. ['1 (x) -
.
~J(x)'" ,ix)]
R
+ (c.c) } ,
where "c.c" means the complex conjugate terms corresponding to the previ-
ous terms in this expression. Putting (6.1.20) into the above equations we
obtain
/,k = 0,
(6.1.21)
lex) ==
ac )]
- ie {[0.9'~.
k ,..
",R(X) - [a~~, )]
ac k ,,,
",R(X)},
where e is the charge parameter and lex) represent components of the elec-
tric charge-current vector density for the complex scalar field ",R(X). D
Now we shall discuss the concepts of integral or total conservation of

various properties of a field. For that purpose we express Noether's equa-
tions (6.1.13) in the following form:
Jill),i = 0, (6.1.22)
t2 ------------ 1'-";;"-"'"
FIGURE 26. A world-tube con-

taining the field <6 R (x).
Theorem (6.1.3) (Integral conservation laws): Let -,<~,(x) 0 inside a world-

tube (see Figure 26) and J/p,(x) == 0 outside the world-tube. Let X4 = tl and
X4 = t2 be three-dimensional cross-sections of the tube denoted by 01 D and 02D
respectively and 0'3 the "cylindrical wall" around the domain D of the tube.
Furthermore, let J(~,nila, = 0, where ni denotes the unit outer normal on the
continuous, piecewise smooth, orientable boundary aD. Then the following inte-
gral conservation law holds:
J(P' == f J(!,(Xl,X2,x3,tddx!dx2dx3
f
o,D
= J(!,(X 1 ,X 2,X 3, t 2 )dx 1 dx 2 dx 3. (6.1.23a)

02D
Proof: From Fig. 26 it is clear that aD = aiD U 02D U 0'3' Applying the gen-
eralized Gauss's theorem (2.4.5) and equation (6.l.22) we have
0= In J/",.i{x)dx dx 2dx dx4

1 3
= folD
J/",{x)n l dx! dx 2 dx 3 + f J(~,{x)ni dx! dx 2dx 3
02D
+ fa, J(~,(x)ni d30'

= f J(!,(x)n4 dx 1 dx 2dx 3 + f J(!iX)n 4 dx! dx 2 dx 3 + 0
f
olD 02D
= J(!,(Xl,X2,x3,t2)(1)dxldx2dx3.
f
o,D
+ J(!,(xl,x2,x3,t2)(-1)dxldx2dx3.
o2 D
For the relativistic fields it is usually assumed that the fields rjJR(x) and
J(!)(x) are nonzero almost everywhere. Furthermore, it is assumed that the
improper space-integrals
J(Il) == f J(!)(x!,X 2,x 3,t)dx!dx 2 dx 3
f
R3
= J(!)(X!,X2,X3,0)dxldx2dx3 (6. 1.23 b)

R3
converge. Since J(Il) is independent of time, it is a conserved quantity of the

field rjJR(X). From the differential conservation equations (6.1.17), (6.1.19), and
(6.1.21) we can construct the following conserved quantities:
Pa == f R3
T4a(x!,x2,x3,0)dx!dx2dx3,
Jab == f J"\b(XI,X2,X3,O)dx! dX2 dX3 = -Jba , (6.1.24)
f
R3
f2 == j4(X!, X2, X3, 0) dx! dX2 dx 3.

R3
The quantities Pa represent components of the total energy-momentum

Jour-vector of the field. The quantities Jab represent the components of the
total relativistic angular-momentum tensor of the field. The quantity f2 repre-
sents the total electric charge of the field.
EXERCISES 6.1
1. Consider the relativistic sine-Gordon field equation:
Ol/t = ksinl/t,
where k #- 0 is a real constant. Construct an action integral such that the
corresponding Euler-Lagrange equation is the sine-Gordon equation.
2. Let
2(X,yR,yRJI" == j~ d~j {Sj(x,yR)I.}.

Show that the Euler-Lagrange equations vanish identically. Explain this
result.
3. Consider the action integral
d(F) == Lj 2(x, yR, yRi' yRij)ly';~R(X)'Y'i;+"i'Y'ij;~R'ij d 4x,

(i) Obtain the requisite equation for d to have a critical value for yR =
rjJR(X).
(ii) Prove that, if the above action integral is invariant under a general
coordinate transformation
Xi = Xi(x),
"'( Al A2
A3 A4)
uX,x,x,x 0
0(xl,x 2 ,x 3 ,X4 )"# ,
for all xED C 1R4, then :t' must satisfy

o:t'( .. ) = 0
iJx u - ,
O:t'(")yR - 0
~ k='
uy ij
4. Suppose that the parameterized world-sheet of a string in M4 be given
by the twice differentiable functions ~i:
A, B E {1,2}.
Assume the metric tensor on the world-sheet is differentiable and satisfies:

gAB(U) = gBA(U),
g(u) == det[gAB(u)] < O.
Furthermore, the action integral for the string is assumed to be
f 15
AB( ) O~i(U)
d.}g U iJ A
U
iJ~j(u) [_
'" B
uU
g
(
U
)J 1/2 dId 2
U u,
(Here the summation convention operates for lower-case and capital indices).
Show that the Euler-Lagrange equations are
~[[_ ( )Jl/2 AB( )iJ~k(U)'] = 0

iJu A g U g U iJuB '
6.2. The Klein-Gordon Scalar Field

Let us consider a complex scalar field I/J(x) that satisfies the Klein-Gordon
equation:
DI/J(x) - J121/J(X) = dkll/J.kl - J121/J(X) = 0, (6.2.1)
where J1 is a real mass parameter and I/J E <'(i'2(M4; q.
In this section and subsequently we choose units such that c = h = 1, where

h is the Planck constant divided by 2n.
The real-valued Lagrangian function !': (; x (; X (;4 X (;4 -+ ~ is given by:
!'(, (, (k' (,) =- dk'(k(, - Jl2((,
a!'~.. )= _ Jl2(,
a( (6.2.2)
a!'~ ..) = _dkl("

a(k
The Euler-Lagrange equation is
[ a2'~
a(
.. ~] _ ~ {[a!'~ .. )] } = _ Jl21/t(x) + dkll/t.kl = 0,
I.. dx a(k I ..
(6.2.3)
which is the Klein-Gordon equation (6.2.1).
Example: We shall obtain a special class of solution of (6.2.1). Assume that

the solution is nontrivial and separable, that is,
I/t(Xt,X 2,X 3,X 4) = ~(Xl)'1(X2)((X3).'1(X4) =1= O.
Equation (6.2.1) yields
~"(Xl )/~(Xl) + '1"(x 2)/'1(X 2)
+ ("(X 3)j(X 3) - .'1"(x4)/.'1(X 4 ) - Jl2 = O.
From the separation of variables it follows that
~"(Xl )/~(Xl) = ex,
r(x2)/'1(x2) = /3,
= y,
C'(X 3 )j((X 3 )
.'1"(X4);.'1(X4) = b.
Here ex, /3, y, b are complex constants that satisfy the relations:
Re(ex + /3 + y - b) = Jl2,
Im(ex + /3 + y - b) = O.
These constants are otherwise arbitrary. The most general separable solution
of (6,1.1) is
I/t(x) = exp{Jax l + Jl3x2 + JYx 3
+ [Re(ex + /3 + y} - Jl2 + iIm(ex + /3 + y}]1/2X4}, (6.2.4a)
where ex, /3, yare arbitrary complex constants. In case we choose ex, /3, y to be
nonpositive real numbers, we obtain the special case of (6.2.4a), namely,

(6.2.4b)
where w(k) == W(kl,k2,k3) == [(k 1 )2 + (k 1 )2 + (kd 2 + 1l2r/2. The real and
the imaginary parts of the solution in (6.2.4b) represent simple harmonic plane
waves with frequency w(k) and wave number k == (k 1 , k2' k3)' D
The Lagrangian (6.2.2) is invariant under Poincare transformations as well

as the global gauge transformation (6.1.20). Thus the Noether theorems
(6.1.17), (6.1.19), and (6.1.21) hold. It will be instructive to work out explicitly
the tensors Tk/(x),,fkab(X), lex) from the Lagrangian (6.2.2). These are com-
puted to be
= dk'(Iii"t/I,j + lii,jt/l,,] - Jkldablii,at/l,b + 1l 2lii(x)t/I(x)],

T,.i x ) = lii.kt/l,j + lii,jt/l,k + dkj 5l'(")I" = 1jk(X), (6.2.5)
"\b(X) = [TkAx)dbj - T\(x)daj]x i = - ,1\a(x),
lex) = _ie{[05l'( .. )] t/I(x) - [O5l'~ .. )] lii(x)}

O(k I.. O(k I..
= - iedka[lii(x)t/I,a -lii,at/l(x)].
Suppose that t/I is a complex-valued function such that the improper inte-
gral JR4It/1(x)1 dx 1dx 2 dx 3 dx 4 converges. Then t/I possesses the Fourier
transform
t/I(x) = O/4n2) fR4 r(kl' k2' k3' k4)eikaXa dkl dk2 dk3 dk 4
(6.2.6)
where r is an appropriate complex-valued function. We shall now seek the

class of general solutions t/I(x) of the Klein-Gordon equation (6.2.1) such that
t/I admits a Fourier transform.
Theorem (6.2.1): Let the complex-valued field t/I(x) be given by the uniformly
convergent improper integral (plane wave decomposition):
t/I(x) = (1/2n)3/2 f R3
[2w(k)]-1/2{IX(k)e i [k.X.-ro(k)x 4 ]
+ p(k)e- i[k.x.-ro(k)x 4 ]} d 3 k, (6.2.7)

w(k) == W(kl,k2,k3) = [kak a + 112]1/2,
d 3 k == dk 1 dk2 dk 3
Furthermore let the following improper integrals converge uniformly
f 1;l3
kl'[2w(k)TlI2 {a(k)e i[k,x'-w(k)x 4 ] + p(k)e- i[k,X'-W(k)x 4 ]} d 3k,
f 1;l3
[w(k)/2] 112 {a(k)e i[k.x'-w(k)x 4 ] + p(k)e-i[k.X'-W(k)X']} d 3 k,
f 1;l3
kl' k/l [2w(k)T lI2 {a(k)e i[k.X'-w(k)x 4] + p(k)e- i[k,X'-W(k)x 4]} d 3 k,
f j,l3
{[w(k)]3/2J!12{a(k)ei[k.x'-W(k)X4] + p(k)e- i[k.X'-W(k)X 4 ]}d 3 k.
Let the continuous complex-valued functions a, fJ be otherwise arbitrary. Then

the equation (6.2.7) yields a class of general solutions of the Klein-Gordon
equation (6.2.1).
Proof: By the assumption that the above improper integrals converge uni-
formly we may differentiate under the integral signs. Thus
ot/l(x) - /12t/1(X)
= (1/2n)312 f 1;l3
[2w(k)T l I2 { - kvkv + [W(k)]2 - /1 2 }
=0.
Example 1: With the general solution in (6.2.7) we shall compute some of

the integral constants given in (6.1.24). Let (x,O) = (xl,x 2 ,x 3 ,0) and d 3 x =
dx 1 dx 2 dx 3 By (6.2.5), (6.2.7), and (6.1.24) we have for the total electric charge
of the field
:!2 = -ie f
1;l3
d 44 {i]/(X, 0)t/I,4(X, 0) - i]/,4(x,0)t/I(x,0)}d 3 x. (6.2.8)
By (6.2.7), we get
i]/(X,0)t/I,4(X,0) = (-i/16n 3 )
x ff
1;l3 1;l3
[w(k)/w(k')] lJ2 {a(k)eik,X' - p(k)eik,x'}
x {iX(k')e-ik~x' + fJ(k')eik~x'} d3 k' d 3k.

Putting the above expression in (6.2.8), we obtain
fl = (e/16n 3 ) f f f [w(k)/w(k')] 1/2

R3 R3 R3
x {[a(k)eik.XI' - p(k)e- ikX.]. [ii(k')e-ik~X. + p(k')eik~X.]

+ (C.C.)} d 3 xd 3 k' d 3 k
= (e/16n 3 ) f f f [w(k)/w(k')] 1/2
R3 R' R3
x {[IX(k)ii(k')ei(k.-k~)x. - P(k)p(k')e-i(k.-k~Jx"
+ a(k)p(k')ei(k.+k~)x. - ii(k')p(k)e-i(k.+k~)x'] + (C.C.)} d 3 x d 3 k' d 3 k.

(6.2.9)
We shall digress slightly to introduce the Dirac "delta function," which will
be used to simplify the nine-fold integral in (6.2.9). Let f(k, k') be a real-valued
continuous function from 1R2 such that the Cauchy principal value of the
improper integral C.P.V. J~':J(k, k)dk converges. In that case
c.P.V. (1/2n) f +OO f+oo f+oo f(k, k')ei(k-k')x dx dk' dk

-00 -00 -00
== lim lim lim (1/2n)
x f +L3 f+L2 [f(k, k') f+Ll ei(k-k')x dx ] dk'dk

-L3 -L2 -L1
== lim lim lim (l/n)
x t+2 t+:'2 [f(k, k') sin[L1 (k - k')]/(k - k')] dk' dk.

3
(6.2. lOa)
Thus we have the delta function defined by the equation

+OO f+oo f(k, k')(j(k -
c.P.V. f -00 -00 k') dk' dk
== lim lim lim (l/n)
X
f +L3 f+L2 [f(k, k') sin[L1 (k -
-L -L k')]/(k - k')] dk' dk
f
3 2
+OO
= c.P.V. -00 f(k,k)dk.
We give a proof of the last step under stronger assumptions. Let g(x) be a
smooth function that vanishes outside a finite interval. It is well known that
c.P.V. J~:(sinLx/x)dx = n for every positive L. Since g is continuous, one
has Ig(x) - g(O) I < e whenever Ixl < 15. Thus we have
c.P.V. t+: Ig(x) - g(O)1 (sin Lxlx)dx
= r:, Ig(x) - g(O)1 (sin Lxlx)dx + t-~' Ig(x) - g(O) I (sin Lxix) dx
+ l~ Ig(x) - g(O)1 (sin Lxix) dx.
Integration by parts shows that the limit of the last two integrals is zero for
every J. > 0 when L increases without bound. Furthermore, we see that the
first integral is bounded by 2eLJ. So choosing e = L -2 we can prove that
the limit of the first integral is zero as L increases without bound. Thus we
have shown that c.P.V. limL .... oo (lIn) J~~ g(x)(sinLxlx)dx = g(O). (There is a
weakness in this proof on account of switching two limiting processes.)
We generalize (6.2.10a) to the three-dimensional case as follows:
c.P.V. ff ~3 IP
f(k,k')b 3(k - k')d 3k' d 3k
== c.P.V. (1/2n)3 f f f
~3 ~3 ~3
f(k, k')ei(k,-k;)x' d 3x d 3k' d 3k
= c.P.V. f R3
f(k,k)d 3 k. (6.2. lOb)
Note that in the sequel we shall suppress explicit mention of the Cauchy
principal value of an integral; i.e., "c.P.V." will be omitted. Using (6.2.l0b) in
(6.2.9) we get
g. = (eI2) ff IP R3
[w(k)/w(k')] 1/2 ([o:(k)iX(k') - P(k)fJ(k')]J3(k - k')
+ [o:(k)fJ(k') - iX(k')P(k)]J3(k + k') + c.c.} d 3 k' d 3 k

= (eI2) f ~3
([Io:(kW - IfJ(k)12]
+ [o:(k)fJ(-k) - iX(-k)P(k)] + (c.c.)}d 3 k

=e fR3
[lo:(kW - IfJ(kW] d 3 k. D (6.2.11)
Example 2: Next we shall compute the total energy of the field from (6.2.24),
(6.2.5), (6.2.7), and (6.2. lOb). Thus we have
H == -P4 = -f T 4 4 (X,0)d 3 x
-f
R3
- -
R4
{44
d (21/1.41/1.4)
-
- (1/1,.1/1,.
-
- 1/1,41/1,4
-
+ Jl 2 11/11 2}
) 1(><,0) d 3 X
x {[kvk~ + w(k)w(k')] [a(k)eik,xY - p(k)e-ik,X']

x [Ci(k')e-ik~x" - p(k')eik~x"] + .u 2 [a(k)e ik ,x' + p(kJie-ik,x']
x [Ci(k')e-ik~x" + p(k')eik~x"]} d 3 xd 3 k' d 3 k
= 0/2) f ~3
[w(k)rl
x {[kvkv + (W(k2 + .u 2 ] [la(kW + IP(kW]

+ [k,.kv - (W(k2 + .u 2][a(k)p(-k) + P(k)Ci(--k)]}d 3 k
= f ~3
[la(k)1 2 + IP(kW]w(k)d 3 k. D (6.2.12)
From equations (6.2.11) and (6.2.12) we can interpret physically la(kW and
IP(kW as densities (or distributions) of the positively charged and the nega-
tively charged scalar fields in the space of wave-numbers (three-dimensional
momentum space).
In this section if we put the mass parameter .u = 0 in all the equations then
the corresponding equations for a complex wave field obeying DI/t(x) = 0 are
obtained.
EXERCISES 6.2
1. Show that by equations (6.1.24), (6.2.5), (6.2.7), and (6.2. lOb) the compo-
nents of the total three-momentum of a complex scalar field are given by
Pv = f ~3
[la(kW + IP(kW]kvd 3 k.
2. In the spherical polar and time coordinate chart, the Klein-Gordon
equation (6.2.1) becomes
r-2(r2X.r),r + (l/r2 sin e)(sin e X,e),e
+ (l/r2 sin 2 e)x, - X,It - .u 2 x( .. ) = 0,
x(r,e,I/J,t) == I/t(X 1,X 2 ,X 3,X4 ).
Obtain the most general separable solution
x(r, e, I/J, t) = p(r)0(e)<l>(I/J).Y(t)
of this equation.
3. Consider the relativistic wave equation for the "spinless" electron in the
external electrostatic field of a positively charged nucleus. This equation is
the modified Klein-Gordon equation:
r-2(r2X.r).r + (1/r2 sin e)(sin e X.o),o + (l/r2 sin 2 e)x,

+ [ -i(a/at) + Ze 2 /r]2 x( .. ) - Jl2 X( .. ) = 0,

where - e < is the charge of the electron and Z is the number of protons in
the nucleus. The stationary solutions are characterized by the special sepa-
rable structure:
x(r, e, t/J, t) = p(r)0(e)(t/J)e- iEt ,
where E is the parameter representing the energy eigenvalue. Show that for
the case of square-integrable solutions, i.e.,
the energy eigenvalues are given by
E = Jl(n + s)[(n + S)2 + Z 2 e4 TI/2,

s == (-1/2) + (1/2)[(21 + 1)2 - 4Z 2 e 4 ]1/2,
n E {I, 2, .. , }, I E {O, 1, ' .. }.

(Note that these values have small discrepancies with the experimental
values. The Dirac equation in sec. 5 can remove these discrepancies.)
4. Suppose that Q: [R3 -. C, s: [R3 -. C are such that
w = Q(xi, x 2 , x 3 ) == Xl cos Cl: + x 2 sin Cl: + ix3,
where Cl: is an arbitrary real parameter. Let h: Dee x C -. C be represented

by '1 = h(w, ~). Let h be an arbitrary holomorphic function of two complex
variables w, ~. Then show that the complex-valued function t{!(x l , x 2, x 3, x 4) ==
h[Q(xi,x2,X3),S(X2,X3,X4)] satisfies the wave equation Dt{!(x) = 0.
6.3. The Electromagnetic Tensor Field

The special theory of relativity emerged from investigations into the speed of
light propagation, which is also the speed of electromagnetic wave propaga-
tion. Therefore, the electromagnetic field equations are of particular interest
in the special theory of relativity. Maxwell's equations, as three-dimensional
vector field equations, are:
t7 H() = aE( . . )
v x.. at'
VE( .. ) = 0,
(6.3.1)
VH( .. ) = 0,
oH( .. )
V x E( .. )= -~'
where t is the time variable, V is the gradient operator, and E( .. ) and H( .. )

are the electric and magnetic fields respectively. The equations (6.3.1) hold
outside the charged sources. In the language of three-dimensional Cartesian
tensor analysis, (6.3.1) can be expressed as:
EIl,Il=O,
(6.3.2)
HIl,Il=O,
aH/A(")
EIlVAEv,A = -a-t-'
where s/Av" is the totally anti symmetric permutation symbol. The above equa-
tions can be cast into Minkowski tensor field equations. To achieve that we
have to identify electric and magnetic three-vectors with an antisymmetric
Minkowski tensor (see (5.3.20:
H 3 (x)
o
The above definition unifies both electric and magnetic vector fields into a
single Minkowski electromagnetic tensor field. Transformation properties of
the tensor Fab(X) under a Lorentz transformation have physical conse-
quences. In the example following equation (1.3.7) in Chapter 1, we find that
a purely electric field in one inertial frame appears to be a mixed electromag-
netic field to another inertial observer.
Maxwell's equations (6.3.2) can be written as the equations
FIlII,II = - F1l4,4'
F 41l,Il =0,
(6.3.4)
Fllv,.l. + FVA,/A + FAll,v = 0,
Fv4,.l. + F4.l.,v = -F"v,4
The above equations can be cast into the following Minkowski tensor field
equations:
Ma == F ab .b = 0,
(6.3.5)
Mabc == Fab,c + Fbc.a + FCa.b = O.
Since Mabc is totally antisymmetric, there are four linearly independent equa-
tions Mabe = O. The eight linearly independent equations in (6.3.5) have two
differential identities (assuming that Fab are twice differentiable):
Ma,a == 0,
(6.3.6)
"abed M bCd ,a == 0,
where "abed is the totally antisymmetric pseudotensor,
Example 1: We shall derive the wave equation for the electromagnetic tensor
field Fab(X) assuming that the Fab are twice differentiable functions. From
(6.3.5) we obtain
0= M abc .a = Fab,ca + dakFbc.ka - Fae,ba
= [Fab,ale + DFbe - [Pe.al b
= 0 + OFbc - 0
= DFbe
The above equation shows that both the electric and magnetic fields travel
with the speed of light. D
Example 2: Introducing a complex-valued antisymmetric field [see equation

(1.3.15)],
l/Iab(X) == Fab(x) + (ij2)"abed Fed (x) = Fab(x) + iFat(x),
we can express Maxwell's equations (6.3.6) in a neat form:
I/Iab,b = 0, D (6.3.7)
The equations Mabe = 0 can be solved by putting
( ) = oAb(x)
Fab x oxa
_ oAa(x) = A _ A
oxb - b.a a.b' (6.3.8)
where the Aa are arbitrary twice differentiable functions. The functions Aa are
called the components of the four-potential field. We note that the four-
potential is not uniquely determined by (6.3.8). We can make a local gauge
transformation (involving an arbitrary twice differentiable function A.):
A~(x) = Ak(X) - A.. k,
(6.3.9)
Inserting (6.3.8) into Maxwell's equations (6.3.5), we obtain four equations:

(6.3.10)
Now we make a local gauge transformation (6.3.9) such that ). is thrice differ-
entiable and satisfies
(6.3.11)
Thus from (6.3.9) we have
A'b ,b = Ab ,b _ dbaA. ,ab
= Ab,b - Ab,b
=0. (6.3.12)
This special condition on A,b(X) is called the Lorentz gauge condition and is
preserved under the restricted gauge transformations:
A:(x) = A~(x) - A.. a ,
OA.(x) = O.
Maxwell's equations (6.3.5) with Fab(X) = F~b(X) = A~,a - A"a,b become
(6.3.13)
Subsequently, we shall assume the Lorentz gauge condition and drop the
prime on the four-potential and write simply
OAk(X) = 0,
(6.3.14)
Example: We shall obtain simple harmonic plane-wave solutions of (6.3.14).

We assume the solutions of the form:
(6.3.15)
where
and
[kd 2 + [k 2]2 + [k3]2 + [k 4]2 > O.
The wave equation (6.3.14) implies that
[k 4]2 = [k 1 ]2 + [k2]2 + [k3]2 = kllkll > 0,
(6.3.16)
k4 = v(k), v(k) == [k ll kIl Ji/2.
Therefore, the four-vector ka must be a nonzero null vector. The Lorentz
gauge condition Ab.b = 0 implies that
kbab(k) = kllall(k) + k4a4(k) = O. (6.3.17)
Using Theorem (1.1.6), equation (6.3.17) implies that ab(k) cannot be time-
like. Thus we must have either ab(k) is null and
ab(k) = )'(k)k b,
for some )'(k) = 0, or else ab(k) is spacelike.

In the case ab(k) is null we have Fab(X) == and we shall not consider this
case further. In the case ab(k) is spacelike we get
~ [a 4(k)]2 < a/l(k)a/l(k).
Using the Lorentz gauge condition we obtain
[k/la/l(k)]2 < [k.kva/l(k)a/l(k)].
By the Schwartz inequality (1.1.13) we conclude that
a/l(k) "# )"(k)k/l
for any )'(k). The three functions are otherwise arbitrary. The function a4 is
given by a4(k) = -k/la/l(k)jk4' In this case
- F/l 4(x) = [k/lkvav(k)j(k4)2 - a/l(k)]k4 sin [kaxa + 8(k)] = 0,
- F/lv(x) = [k/lav(k) - kva/l(k)] sin[kax" + 8(k)] = 0,
k/lF/l 4 (x) = kA.BA./lvF/lv(x) = FA.4(X)BA./lvF/lv(x) == 0.
The above equations show that the propagation vector k == (kl> k2' k 3 ), the
electric field vector E(x), and the magnetic field vector "(x) are all mutually
perpendicular. Thus the electromagnetic wave is propagated in a transverse
fashion. D
Now we shall derive Maxwell's equations from a variational principle. We

can choose the Lagrangian 2: ~4 x ~4 X ~16 -+ ~ as
.5t'(x, Ya' Yak) == - (lj4)d ab dkl (Ylb - Ybl)(Yka - Yak)' (6.3.18)
Therefore,
.5t'( .. )IYa=A.(X),Yak=A a.k = -( Ij4)F"b(x)Fab (x),
02( .. ) =
OYa -

,
Thus the Euler-Lagrange equations are
(6.3.19)
which are Maxwell's equations (6.3.10). The canonical stress-energy-momen-

tum tensor, computed from (6.1.17) and (6.3.18) is
i i [82"( .. )]
T j(x) = b j2"(. ')1" - -8-, Aa,j
Ya. I ..
(6.3.20)
The above tensor is neither symmetric nor gauge-invariant. We define the

symmetrized gauge-invariant stress-energy-momentum tensor by
Oij(X) == I;ix) + [Pi(x)Aix)L

= Pi(x)Faix) - (1/4)d uP b(x)Fab (x). (6.3.21)
It can be verified (see problem 4 of Exercises 1.3) that
()ii == 0,
()ij , = Tii , =
,J ,J
0'
The last improper integral may be replaced by a surface integral by Gauss's

Theorem, which converges to zero by the usual hypothesis. The total energy-
momentum of the electromagnetic field is given by
Pk = f GP
()4 k(X, 0) d 3 x. (6.3.22)
There is a simpler Lagrangian (which is not gauge-invariant) from which

Maxwell's equations (6.3.14) can be derived. It is given by
2"0 (x, Ya' Yak) = - (1/2)d Ud kIYkiYlj, (6.3.23)
where we have
82"0(.') = - (1/2)dijd k, [s;a

~Y ---
s;b
U k U iYlj + s;a s;b
U IU jYki ] = - Yab .
ab
Thus the Euler- Lagrange equations are
The Lorentz gauge condition has to be put as an additional equation. The

canonical stress-energy-momentum tensor constructed from 2"0(' .) is
(6.3.24)
so that 1(Ojij(x) = 1(Ojji(x), The total energy of the electromagnetic field ac-
cording to (6.1.24) and (6.3.24) is
H= -P4
= -f 1(0)44(X,O)d 3x
~ t,
R3
= [A a.4 Aa,4 + b~f1 Aa,~Aa,f1] d 3 x. (6.3.25)
Example: A class of general solutions of Maxwell's equations can be ex-

pressed as a Fourier integral [compare the plane-wave solution in (6.3.15)]:
Ab(X) = (2nr3/2 f R3
[2v(k)rl/2
x [O:b(k)e i[k,x"-v(k)x 4] + iXb(k)e- i[k,x'-v(k)x 4 ]] d 3k (6.3.26)
where v(k) == [kl'kl']1/2. Putting the above solution into (6.3.25) we obtain
H = -(32n 3 r 1 fff
R3 R' R'
[v(k)v(k')r 1/2
X {[v(k)v(k') + kvk~] [o:b(k)eik,X' - iXb(k)e-ik,X']

x [O:b(k')eik;XI - iXb(k')e-ik;X']} d 3x d 3k' d 3k
= -(1/4) fR'
[v(k)rl[[v(k)2 - kvkv] [o:b(k)O:b(-k) + iXb(k)iXb(-k)]
(6.3.27)
We can interpret lo:b(kW physically as the electromagnetic field density

(distribution) in the three-momentum space. Then from (6.3.27) we can con-
clude that the energy density of the electromagnetic field is the product
of frequency with the sum of the field densities (in different modes of
polarization). D
EXERCISES 6.3
1. Let the complex electromagnetic field be given by [see (6.3.7)]
l/Iab(X) == Fab(X) + iFat(x).

6.4. Nonabe1ian Gauge Fields 147
Prove that the symmetrized energy-momentum-stress tensor in (6.3.21) can

be expressed as
2. Consider the eigenvalue problem:
where 8ab (x) is the symmetrized energy-momentum-stress tensor of the pre-

ceding problem. Prove that there exist two invariant eigenvalues
A.(x) = (1/4){ [Fab(X)Fab(X)] 2 + [Fab (x)F*ab(X)]2 P/2.
3. Suppose that f is an arbitrary holomorphic function of two complex
variables w, ~ in a domain of 1[;2. Let
1f'"(x) == Re[f(w, ~)I(Q=xl cosk+x2sink+ix'.~=x2+ix3sink+x4cosk]'
where k is an arbitrary real parameter. Let Cba = - Cab be an arbitrary con-
stant Minkowski tensor field. Then show that
Aix) == C/1JI:k
solves Maxwell's equations DAix) = 0, Aj.i = O.
4. Consider 1(O)ij (X) as given in (6.3.24) and the class of solutions in (6.3.26).
Prove that the components of the total three-momentum vector of the elec-
tromagnetic field is
PI' == f~3
1(O)\(x,0)d 3 x = f
~3
[a b(k)iXb(k)]kl'd 3 k.
6.4. N on abelian Gauge Fields

Gauge fields are natural generalizations of the electromagnetic field, which is
an example of an abelian gauge field. The theory of gauge fields starts with a
Lie group G [see equation (3.4.8)]. Suppose that the group G involves r
parameters a l , .. , ar . We can associate an r-dimensional group manifold with
G. The group manifold is a differentiable manifold so that there exists a
tangent space at every point of the manifold. The tangent space consists of the
tangent vectors that are taken to be directional derivatives [see equation
(2.1.6)]. There exists a basis {rA(p): A = 1, ... ,r} for the tangent space at a
point p. The basis vectors rA(p) are called generators of the corresponding
Lie algebra. The generators satisfy the commutation relations (assuming the
summation convention for capital label indices)
[rA(p),rB(p)] == rA(p)rB(p) - rB(p)rA(p)
= CEABrdp), (6.4.1)
where the constants CEAB are called structure constants. The structure con-
stants can be either real or complex numbers according to the choice of the
original Lie group G. From equation (6.4.1) and Jacobi's identities
[['A(P), 'B(P)], 'dp)] + [['B(P), ,dp)], 'A(P)J

+ [[,dp), 'A(P)], 'B(P)] == 0,
it follows that the structure constants must satisfy
CEBA =- C EAB ,
(6.4.2)
CEABCFED + CEDACFEB + CEBDCFEA = 0.
It can be proved that for an abelian Lie group G all the structure constants
vanish. For a nonabelian Lie group G, at least one of the structure constants
is nonzero. We assume that Gis nonabelian in the sequel.
Now we define
(6.4.3)
It follows that
For a semisimpie (the invariant subgroups are nonabelian) Lie group it can be
proved that det[YABJ #- 0. In such a case YAB can be treated as a group metric
for which the contravariant tensor yAB exist to raise capital indices. We as-
sume that the Lie group G is semisimple.
Example: Let us consider the semisimple Lie group SU(2), the group of uni-
tary unimodular transformations of two-dimensional complex vectors (see
Chapter 4, Sec. 1). This group has three generators ,~, which satisfy the
commutation relations:
['I,'2J = -'3'
['2"3J = -'1'
['3,'IJ = -'2
Thus the structure constants are given by
where B. p" is the three-dimensional totally antisymmetric permutation sym-

bol. The group metric components are given by
(The metric tensor Yp<1 is negative definite since the group SU(2) is compact.)
The two-dimensional self-representation of SU(2) induces a 2 x 2 complex
matrix representation of the corresponding Lie algebra. This representation
6.4. Nonabelian Gauge Fields 149
is given by the 2 x 2 complex matrix representation of the generators

[rIl] = (ij2)aIl'
where all are Pauli matrices. The exponentiation of such a matrix representa-
tion is instructive. For example,
L (n!fl( -i>j2ta3n
00
==
n=O
= cos(>j2)/ - i sin(>j2)~
= [e-;/2 ei~/2l
This unitary unimodular matrix has been considered before in Chapter 4
[preceding equation (4.1.16)]. D
Next we introduce the gauge potentials BAk(X). We assume that the func-
tions BA k are thrice differentiable in the domain D c 1R4 of consideration.
Nonabelian gauge field are defined by
FAij(x) == B\i - BAi,j + CABCBBiBCj = -FAjj(x), (6.4.4)
where C A BC . O.
We note that
.. k + F AJ'k, l. + F Akl,J
FA IJ. .
= CABc[[BBiBCjlk + [BBjBCkL + [BBkBCJ,j]
= -CABc{BBiFCjk + BBjFCki + BBkFCij}' (6.4.5)
The Lagrangian f': D X 1R4r X 1R 16r -+ IR for the gauge field is defined to be
f'(x, y\, yAu) == (lj8)YABdikdjl
X [[yAjj - yAij + CACDyCiyDj]
X [yB'k - yBkl + CBEFyEkyF,]], (6.4.6)
f'(X, yAk' yAij)lyAk=BAk(X).yAij=BA,.j(X) = (lj8)FAiix)}~ij(x).
From the definition it follows that
of'( .. ) _ (1/8) dikdjl[[CA ~s D

~ - YAB GDU iY j + CA CGU~sjY C]i
Y s
X + CBEFyEkyFl]
[yB'k - yBkl
+ [yAji - yAij + CACDyCiyDJ
X [C BGF{)\yF, + C BEG{)S/yEkJ]
= (1/4)YAB[yA ji - + CACDyCiyDj] [C BGFdiSyFj + CBEGdjSyEi]

yAij
= (1/4)C AGE[yAji - yAij + C ACDyC,yDj ] [dlSyEj _ djSyEi]
= -(1/2)C AEG[yAj, - yAij + C ACDyC,yDj]diSyEj,
a:y~") = (1/8)yABdikdjl{ bA G[bSjbbi _ bSibbj ]

sb
X + CBEFyEkyF,]
[yB 'k - yBkl
+ [yAj, _ yAij + CAcDyC,yDj]bBG[bS,bbk _ b\bb,]}
= (1/4)(d bkd SI - dSkd bl ) [YGlk - YGkl + CGEDyEkyD,]
= (1/2)d bkd SI [YG'k - YGkl + CGEDyEkyD,]. (6.4.7)
Therefore, the corresponding Euler-Lagrange equations are
[ a!l'( .. )] d {[a!l'( .. )] }
ayGs I.. - dXb ayGSb I
= -(1/2)CAEGFAsj(x)BEj(x) - (l/2)FGb. b = 0,
which are equivalent to
(6.4.8)
The canonical energy-momentum-stress tensor, as given in (6.1.17), can be
computed from (6.4.6) as
i i [a!l'( . . )] G
T j(x) = 15 j!l'( . . )1" - ~ B s.j
Y.I I
(6.4.9)
The canonical tensor T;ix) is not symmetric. The symmetrized energy-
momentum-stress tensor e,j(x) can be constructed by subtracting a "diver-
gence term" from T;j. More precisely,
e,ix) == T;ix) - (1/2) [F}i(X)BA/x)h
= (1/2)[FAbi(x)BAb,j + (l/4)dijFAkI(X)FAk'(X)]
- (1/2) [FAbi(X)B\b + F}i.bBAj(x)]
= (1/2)F}i(x)[B\,j - BAj,b] - (l/2)CBDAFBbi(X)BDb(X)BAj(X)
+ (1/8)dijFAkl(X)FAkl(X)
= (1/2)F}i(x) [BAb,j - BAj,b + CAEDBEj(x)BDb(X)]
+ (1/8)dijFAkl(X)FAkl(X)
= -(1/2)F}i(x)F\j(x) + (1/8)dijFAkl(X)FAkl(X)
= eji(x). (6.4.10)
Note that 8ij(x) is expressible algebraically in terms of the gauge fields

rather than the gauge potentials.
Example: We shall obtain the Yasskin class of exact solutions for the gauge
field equations (6.4.8). Let Ai(X) be thrice differentiable components of the
electromagnetic four-potential, which satisfy Maxwell's equations (6.3.14).
We assume for the gauge potential the special form
BAj(X) = k AAj(x), (6.4.11)
where the k A are r real constants. Thus the gauge field components become
FAij = B\i - BAi,j + CAEFBEiBFj
= kA(Aj,i - Ai,) + [CAEFkEe] [AiAj]
= kA(Aj,i - Ai,)
= kAFij.
Therefore, we obtain
FGsb,b - CADGFASbBDb = kGFsb,b - CADGkAkDFsbAb =O.
The last term vanishes identically due to the anti symmetry CDAG = - CADG
(see problem 1 of Exercises 6.4). Thus by the ansatz (6.4.11), the gauge field
equations (6.4.8) are exactly satisfied. D
The gauge Lie group G is assumed to be SU3 for the strong interactions in
quantum chromodynamics, whereas the gauge group is taken to be SU 2 x
U 1 for electroweak interactions. The interacting gauge field will be discussed
in sec. 6.
EXERCISES 6.4
1. Let CABD =YAECEBD' Prove that CBAD = -CABD .
2. Let the gauge potentials BAi be twice differentiable functions in the
domain D c 1R4 of consideration. Prove the differential identities
[CADEBDj(x)FEij(X)J.i =O.
3. Consider the symmetrized energy-momentum-stress tl:nsor 8i)x) given
in equation (6.4.10). Assuming that BAi(X) are twice differentiable functions
=
in D c 1R4 , prove the differential identities 8ij,j 0 in D by explicit
computation.
6.5. The Dirac Bispinor Field

In this section we shall use many definitions and equations from Chapter 4.
Thus it is advisable to quickly review its contents before proceeding. Let us
consider a twice differentiable spinor field ,pA(x). A logically simple relativis-
tic field equation for such a field is
(6.5.1 )
where m ~ 0 is the mass parameter. It is known that a class of solutions of

the (simple) second-order factorizable ordinary differential equations y" -
m2y = (d/dx + m)(d/dx - m)y = 0 is provided by the solutions of the first-
order ordinary differential equation y' = my. Similarly the second-order
partial differential equations (6.5.1) allow a class of twice differentiable
solution functions t/JA which satisfy the system of first-order partial differen-
tial equations
aaBAt/JB.a = mt/JA(x). (6.5.2)

The reason for this assertion is (6.5.2) implies
b_ aBA", _ b ",A _ b ",A
a CAa 'f'B.ab - ma CA'f' .b - ma AC'f' .b
Thus
or equivalently we get
-
d ab~B_"'- _ 2",_()
U C'f'B.ab - -m 'f'C x .
Hence, we have
and finally we get
Here we have used equation (4.2.18) in the above computations. Note that for
the zero rest mass case (m = 0), equation (6.5.2) reduces to aaBA<Pi.a = O. This
equation is known as the Weyl equation of a spinor field.
Equations (6.5.1) and (6.5.2) are mixed spinor-tensor equations and co-
variant under proper, orthochronous Lorentz transformations. But these
equations are not covariant under improper Lorentz transformations. To
obtain spin or equations that are covariant under general Lorentz transfor-
mations we have to consider two spinor fields XB' and ,pA. The coupled spinor
field equations are taken to be
akABxu =- m,pA(x),
(6.5.3)
akBA,pB.k = mXA(x).
We shall now introduce bispinor fields [equation (4.3.2)] and 4 x 4 Dirac

matrices [equation (4.3.6)] by
(6.5.4)
where the bispinor indices u, v take values in {l, 2,3, 4}. The coupled spinor
field equations (6.5.3) can be cast into the bispinor field equation
ykuvt/lv. k + mt/lU(x) = 0,
or more abstractly
(6.S.Sa)
The above bispinor field equations for m > 0 is the well-known Dirac equa-
tion, which describes the wave field of a free spin (1/2) massive particle.
The hermitian conjugation of (6.S.5a) can be written as
t/I.tkykt + mt/l(x)t = ot. (6.S.5b)
Thus
t/I.t~ya - t/I.t4y4 + mt/l(x)t = ot.
Other ways of writing the above equations are
it/l.t~yay4 _ it/l.t4[y4]2 + imt/l(x)ty4 = ot,
$.kyk - m$(x) = 0\ (6.S.Sc)
$(x) == it/l(x)ty4.
We shall now state and prove the general Lorentz covariance of the Dirac
equation (6.S.Sa).
Theorem (6.5.1): The Dirac equation (6.S.Sa) is covariant under a general

Lorentz transformation Xi = aijxj and the induced unimodular bispinor trans-
formation r/i(x) = .'Tt/I(x).
Proof: Suppose that we start with the Dirac equation in the hatted
Minkowski coordinates:
b or/i(x)
+ mt/l(x) = O.
A
y ox b
By equations (4.3.3) and (4.3.12) we obtain
ybl\ffot~:) + mfft/l(x) = O.
Thus
or
ff[ykt/l.k + mt/l(x)] = O.
Since the 4 x 4 unimodular matrix ff is invertible, we can conclude
ykt/l,k + mt/l(x) = O.
Thus the covariance of the Dirac equation (6.5.5a) is proved.
The Dirac matrices, satisfying equations (4.3.8a) and (4.3.8b), are deter-
mined only up to a similarity transformation. There exist infinitely many
choices of Dirac matrices, and each choice is related to another by a similar-
ity transformation. A popular choice of Dirac matrices (in signature + 2) is
the following:
y'" = y",
y'4 .[1
= -I 0
y'S == y'ly'2y'3y'4 = [~ -/J

o '
(6.5.6)
S"lS-l = y'\
il
where .j2s == [ 1
IJ
il'
Writing t/I'(x) == St/I(x) and multiplying (6.5.5a) from the left by S, we can
obtain an equivalent form of the Dirac equation as
(6.5.7)
In the sequel we shall drop the prime so that the Dirac equation will be
formally given by (6.5.7), but the Dirac matrices are given by (6.5.6).
Example: We shall derive the plane-wave solutions of the Dirac equation

(6.5.7). For that purpose we shall seek a solution in the form
t/I(x) = (p)e iPkX\
where (p) is the unknown bispinor column vector. The above form implies,
by the Dirac equation, the matrix equation
[iPk yk + ml] (p) = O. (6.5.8)
The above equations yield a system of homogeneous linear equations on the

unknown components ~U(p). The necessary condition for the existence of a
nontrivial solution is
(6.5.9)
Thus reducing (6.5.9) to (dk1pkPI + = 0, we find that P4 = E(p), where
m2)2
E(p) == [(pd 2 + (P2)2 + (P3)2 + m2 ]1/2. So we have to investigate the follow-
ing two classes of solutions:
tjJ(x) = ~ [p, - E(p)] ei [p_x--E(p)x 4]
== U(p)e i [p_x'-E(p)x4 ];
(6.5. lOa)
tjJ(x) = ~[_p,E(p)]e-i[p_x--E(p)X4]
== V(p)e- i [P.x--E(P)x 4 ].
(6.5. lOb)
These two classes represent solutions of positive and negative energy states
respectively. Putting (6.5.l0a) into the Dirac equation we obtain
[iPIl,),1l - iE(p)l + mI]U(p) = O.
Defining U(p) == [X(P)], where X(p) and ~(p) are column vectors of size 2 x 1,
~(p)
the last matrix equation yields the following two matrix equations:
+ ipll(JIl~(p) = 0,
[m - E(p)Jx(p) (6.5.l1a)
iPIl(JIlX(p) + [m + E(p)]~(p) = O. (6.5.l1b)
By multiplying with the matrix i[m + E(p)r 1pv(JY on the left of (6.5.l1b) and
using (4.1.12) and (6.5.9) we derive (6.5.11a). Thus the matrix equations (6.5.11a)
and (6.5.11b) yield a linearly dependent system and we have to solve either
(6.5.11a) or (6.5.11b). Let us solve (6.5.11b), which can also be expressed as
[I,i(m + Erlpll(Jll] [~i:~] = o.

It is clear that the 2 x 4 coefficient matrix [I, i(m + E)-IPIl(Ju] is ofrank 2. It
follows that the original coefficient matrix [iPkyk + mI] ofthe Dirac equation
(6.5.8) is of rank 2 for P4 = E(p), where p E ~3 is an arbitrary point. The
general solution of (6.5.11b) can be obtained by putting
~(p) = - i(m + E)-I PI1 (JI1 X(p),
U(p) = [X(P)] = [ X(p) ]
~(p) - i(m + E)-1 PI' (JIlX(p)
~ "(PJ l-i(m
-i(m} El-'p,
+ Erl(pi + iP2)
l "(PJl
+ -i(m
-i(m
+ El~'(P' ip'll'
-
+ E)-l p3
where Xl' X2 are arbitrary complex-valued functions of p. Therefore, a basis

for this two-dimensional solution space of positive energy is {U1 (p), U2(P)}
where
(6.5. 12a)
Here we have chosen a suitable normalization. We follow exactly analogous
l
steps for the negative energy [p4 = - P4 = - E(p)] solutions to derive the
other two basis vectors of the solution space as
+ Eflp3 ] i(m
v,(P) ~ [(m + E)/2m]'" '(m + E)-t' + 'p,) ,
(6.5.12b)
By direct computations it can be verified that the above basis vectors satisfy
formal "orthogonality" conditions:
ur(p)tu.(p) = (l/m)E(p)c5r.,
vr(p) tv .(p) = (l/m)E(p)c5rs>
(6.5.13)
Ur(p)tv.( - p) = 0,
v.(p)tu.( - p) = o.
We shall now state and prove a theorem regarding the superposition of
plane-waves.
Theorem (6.5.2): Let a Fourier integral be given by
"'(x) = (2nf3 /2 t3 {[m/E(p)] 112 rt [lXr(p)u r(p)e i [p,X'-E(P)x 4)
+ Pr(p)vr(p)e- i [P,X"-E(P)x4)] } d 3 p,
(6.5.14)
such that the above improper integral converges uniformly for otherwise arbi-
trary choices of the function OCr and Pro Furthermore, assume that the following
improper integrals also converge uniformly:
L3 {[mI E(p)r/2 P. rt [ocr(p)ur(p)ei[p.X'-E(P)x41
_ Pr(P)v r (p)e- i [p.X.-E(P) X4 1]} d 3p,
L, {[mE (p)] 1/2 rt [ocr(p)ur(p)ei[P.X.-E(P)x41
_ Pr(p)vr(p)e- i [P.X'-E(P) x4 1] } d 3p.
Then equation (6.5.14) provides a general class of exact solutions of the Dirac
equation (6.5.5a).
Proof: Under the assumption of uniform convergence of the above improper
integrals we are allowed to differentiate under the integral sign in (6.5.14).
Thus using (6.5. 11 a), (6.5. 12b), and (6.5.8) we obtain
ykt{!.k + mt{!(x)
= (21t)-3/2 L3 {[mIE(p)] 1/2
2
x L [ocr{p) [iP.Y - iE{p)y4 + mI]ur{p)eiIP.x'-E(p)x41
r=1
+ Pr{P) [ -iP.Y + iE(p)y4 + mI]Vr{p)e-iIP.X.-E(P)X41}d3p = O.

Now we derive the Dirac equation by a variational principle. We introduce
the notation:
(= [(1,(2,(3,(4]
j]
0 0
=[i", (',(',,'] [~ 0
1
-1
0
0 0
= ie t y4
= [,"1,,2, _,3, _,4]. (6.5.15)
Here (" = s(u)''', s(l) = s(2} = -s(3} = -s(4} = 1. The Lagrangian function
is defined by
!l'W,(", ':,("a) == (1/2) [,,,yau .eva - , ..ayau ve V
] + m'.. eu
= (1/2)'Y"'a - (1/2),"y"e + m,e,
!l'(. ')1" = (1/2)if,(x)y"t{!,a - (1/2)if" ayat{!(x) + mif,(x)t{!(x). (6.5.16)
Note that 2 is a real-valued function from ([4 x ([4 X ([16 X ([16. It follows
that
arv
02( .. ) -
= -(1/2KuoY ou v + m'v,
-
02( .. ) -
~ = (1/2),uY"v.
0= [02(")J
oC
- ~{[02(")J
o'va
I..
} dx I ..
= -(1/2)~u,oY"v + m~v(x) - (1/2)~u,aya"v

= -[~u,oyOUv - m~v(x)]. (6.5.17)
Thus we have
or equivalently
Yt/l,o + mt/l(x) = O.
Example 1: The canonical energy-momentum-stress tensor is computed from
(6.S.16) and (6.1.17) as'
T k j(x) -_ b kj2(")I" - [02(")J

----arv
.. k I..
- [02(")J
t/I,jv - t/lu,j ~--
O'uk I..
= (1/2) [~u.jykU vt/lV(x) - ~u(X)yk"vt/lV.j]
= (1/2) [~.j(X)ykt/l(X) - ~(X)ykt/l.j]. (6.S.18)
The symmetrized energy-momentum stress tensor is
o -
8kj(x) = T"j(x) + (1/4) ax O{t/I(x) [yoYjYk - djk yO - bajYk + b OkYj]t/I(x)}
= (1/4) [~.jYkt/l(X) + ~,kYjt/l(X) - ~(X)Ykt/l,j - ~(X)Yjt/l.k]

= (1/2) [T"ix) + 1jk(X)]. 0 (6,S.19)
Example 2: We can compute the charge-current vector density lex) by equa-

tions (6.S.16) and (6.1.2):
lex) = -ie{[02~")J t/lV(x) - ~u(X)[02~(")J }

o( k I.. a'uk I..
= -ie~u(x)ykuvt/lV(x)
= -ie~(x)ykt/l(x), 0 (6.S.20)
Example 3: We shall compute the total electric charge of the Dirac field t/t(x)
using (6.1.24) and the class of solutions given in (6.5.14) and the equation
(6.5.20).
:!2 = f r(x,O) d 3x = -ie f ~(x,O)'lt/t(x,O)d3X

t3 t3 t3
u;l3 u;l3
= -(iem/8n 3 ) ,tl ,~ {-i[E(p)E(p')rl/2
x [iir(p)ur(p)te-iP,x' + Pr(p)vr(p)eip,X']
x [a,(p')us(p')eip;X'
+ .Bs(p')v,(p'f e-iP;X']} d 3x d 3p' d 3 p
= -(em) r~ s~ L3 ([1/E(p)]
x [iir(p)as(p)ur(p)tus(p) + Pr(P).Bs(p)vr(p)tvs(p)
+ iir(p).Bs( -p)ur(p)tv,( -p)
+ Pr(p)a.( -p)vr(p)tus( -p)]} d 3 p.
Using equations (6.5.13) we finally obtain
}2 = -e r~ t3 {lar(pW + IPr(PW} d 3 p. (6.5.21)
The total charge :!2 is negative for e > 0 and ar(p) =1= 0, Prep) 'f. O. This is physi-
cally unacceptable since t/t(x) should describe electrons as well as oppositely
charged positrons. The situation can be remedied by the process of the second
quantization whereby :!2 can be of both signs. 0
EXERCISES 6.5
1. (i) Construct a Lagrangian function !l' such that the corresponding
Euler-Lagrange equations are
akABxB.k = -mtftA(x),
akBAtftB.k = mXA(x).
(ii) From the above Lagrangian derive the canonical energy-momentum-
stress tensor Tkix) and the charge-current vector lex).
2. Consider the basis vectors of the solution space of plane-waves in equa-
tions (6.5.12a) and (6.5.12b).
(i) Prove the "orthonormality" conditions
iir(P)us(p) = -vr(p)v,(p) = b",
iir(p)vs(p) = vr(p)u,(p) = 0.
(ii) Energy-projection matrices are defined by

A(p) == (1/2m)[=Fi(pI'YI' - E(p)y4) + mI].
Show that
A+(p)ur(p) = ur(p),
A+(p)vr(P) = 0,
A_(p)ur(p) = 0,
A-(p)vr(P) = vr(P)
3. Using the class of solutions in equation (6.5.14), and the equations (6.5.18),
(6.1.24), show that the total energy and momentum components of the Dirac
field are
H = -P4 = rt t, {lor:r(p)12 -IPr(p)I2}E(p)d 3 p,
PI' = rt L, {lor:r(P)1 2 - IPr(PW}pl' d 3 p.
6.6. Interaction of the Dirac Field with Gauge Fields

The electromagnetic field is an abelian gauge field, and it is the simplest case
to consider. The Lagrangian for the Dirac field interacting with an electro-
magnetic field is taken to be [see equations (6.3.18) and (6.5.16)]
2", 'u, CVa' 'ua' Ya' Yak)
== (1/2) ['uyauvWa + ieYaCV) - ('ua - ieyJu)yauvC v] + m,uCu
- (1/4)d ab dkl (Ylb - Ybl)(Yka - Yak)
= (1/2) [,ya(Ca + ieYaO - (t - ieYaOyan
+ mC' - (1/4)d ab d kl (Ylb - Ybl)(Yka - Yak)' (6.6.1)
Thus we may write
2( .. )1" = (1/2l{ ~(x)ya[t/t.a + ieAa(x)t/t(x)] - [~.a - ieAaCx)t/t(~)]yat/t(x)}
+ m~(x)t/t(x) - (1/4)Fab (x)Fab(x).
Note that 2( .. ) is a real-valued relativistic invariant. Furthermore, 2( .. ) is
invariant under the combined local gauge transformations:
t/t'(x) = t/t(x)eieA(x),
(6.6.2)
From (6.6.1) it follows that

02( .. ) - - - -
~ = (ie/2)'uyauvYa - (1/2)('ua - ieYaUyau v + m'v,
= [ -(1/2)'ua + ieyi.]ya"v + m'v,
~~U
oG = (1/2)1'~uy au v'
(6.6.3)
02( .. ) -
- 0 - = (ie)'uya"v,v,
Ya
02( .. ) = (yba _ yab).
OYab
0= [02()J - ~{[02~")J }
0' I.. dx O( a I..
= [ -(1/2)~".a + ie~u(x)Aa(x)]yauv + m~v(x) - (1/2)~u,aya"v
(6.6.4a)
and
0= [02(")J - ~{[02(")J }
OYa I.. dx 0Yab I..
= ie~u(x)yau vt/JV(x) - Fab,b' (6.6.4b)
These equations may be written in matrix form as

[~,a - ieAa(x)~(x)]ya - m~(x) = ot,
or
ya[t/J,a + ieAaCx)t/J(x)] + ml/J(x) = 0,
and
Fab,b = ietj,(x)yat/J(x).
If we define the gauge-covariant derivative as
Dal/J(x) == I/J,a + ieAa(x)l/J(x), (6.6.5)
then we can write the combined Maxwell-Dirac equations (6.6.4a) and
(6.6.4b) as
yaDat/J(x) + mt/J(x) = 0,
jja~(x)ya - m~(x) = ot, (6.6.6)
pab,b = ie~(x)yat/J(x).
Example 1: Suppose that the Dirac wave functions 1jI" are twice differentiable.
In that case from (6.6.4a) and (4.3.8a) we obtain
0= [yb(Ob + ieAb(x)) - ml] [ya(ljI.a + ieAa(x)ljI(x + mljl(x)]
= ybya{IjI,ab - e2Aa(x)Ab(X)IjI(X) + ie[Aa,bljl(X) + Aa(X)IjI,b + Ab(x)IjI,a]}
- m21j1(x)
= [dabl + (1/2)(yb ya _ yayb)]
X [IjI,ab - e2Aa(x)Ab(X)IjI(X) + (ie/2)(A a,b + Ab,a)IjI(X)

+ (ie/2)Fab (x)IjI(X) + ie(Aa(x)IjI,b + Ab(x)IjI,a)] - m2 1j1(x)
= DaDaljl(x) + [(ie/4)(yayb - ybya)Fab(x) - m21]IjI(x). (6.6.7)
The term (ie/4)(yayb - ybya)Fab(x) represents the magnetic moment of the
Dirac particle (an electron or Jl-meson) and the corresponding gyromagnetic
ratio 2 agrees well with experimental results. D
Example 2: Let us consider the coupled Maxwell-Dirac field equations

(6.6.4a), (6.6.4b) for the zero rest mass case (m= 0). We pose the question
whether or not the spin (1/2) particle can acquire a nonzero mass due to its
own electromagnetic self-interaction. We will try to obtain such a solution
with the following assumptions:
AI(x) = A 2(x) == 0,
A3(X) = a(xl,x2),
A4(X) = W(xl,x 2),
(6.6.8)
~(x) ~ -',;:,)1
[ ,(x') # o.
The Dirac equation (6.6.4a) for the case m = 0 reduces to one independent
equation:
d
e[a(xl,x 2) + W(X 1,X 2)] = i dx4Iog[e(x4)],
By separation of variables we obtain
a(xl, x 2) + W(xl, x 2 ) = M/e,
(6.6.9)
e(x 4 ) = a.e- iMx',
where ME IR and a. E I[: are arbitrary constants. The electromagnetic equa-
tions (6.6.4b) reduce to two independent equations:
a,l1 + a,22 = -2ela.12,
~ll + ~22 = 2ela.1 2,
The general solutions of the above equations, consistent with (6.6.9), are
a(xl,x 2) = _(e/2)laI 2[(xl)2 + (X 2)2] - h(x\x 2),
(6.6.10)
W(X 1 ,X 2 ) = (e/2)laI 2[(xl)2 + (X 2 )2] + h(X 1 ,X 2 ) + (M/e),
where hex!, x 2 ) is an arbitrary real harmonic functon. Equations (6.6.8),
(6.6.9), and (6.6.10) constitute a special class of local solutions of (6.6.4a) and
(6.6.4b) with m = O. By applying a gauge transformation, the solution can be
simplified into the case where M = O. D
We shall now formulate the Lagrangian for a Dirac field in interaction

with a general nonabelian gauge field. Recall that there exist a gauge group
G and the generators TA of the corresponding Lie algebra that satisfy the
commutation relations (6.4.1). An n-dimensional representation [see equa-
tion (3.3.l)] of the group G induces n x n matrix representations [TA KJ] of the
generators. Here the indices A,B .... ,F takes values in {l, ... ,r} and indices J,K
take values in {l, ... ,n}. We will apply the summation convention to both
types of indices. The matrices
[1~/J] == -i[TAKJ ], (6.6.11)
can be chosen to be hermitian whenever G = SU(N)for some positive integer N,
The Lagrangian for the interacting fields is defined by
if(KV, (Ku, (Kv a' (Kua, yAj, yAlk)
== (Kuya\[(Kv a + igTAKJyAaeV]
+ (1/8)YABdbkdjl[yAjb - Y\j + gC ACDyCbyDj ]
X [yBlk - yBkl + gCBEFy\yF 1], (6.6.12)
(K == i(K)ty4,
FAk1(X) == [yAlk - y\l + gCAEFyEkyFalyA/=BA/(X),Y=BA/.k'
if(")I" = ~K(X)ya[",K.a + ig1~/JBAa(x)",J(x)] + (l/8)YABPAkl(X)FBkl(X),
where g#-O is the coupling constant (or charge) of the gauge field interaction,
It follows from (6,6.12) that
aif( .. ) . - au K A
ar" = Zg(KuY v ~ JY a'
aif( .. ) - au
a(Jva = (JuY V'
aif( .. ) . - su K Jv
aes = Zg(KuY v ~ J(
- (g/2)CEDAdSb[yEjb - yEbj + gCECFyCbyFj]yDj,

aif( .. ) sa rb D E
~ = -(1/2)d d [YAba - YAab + gCADEy aY b].
y sr
Thus the Euler-Lagrange equations are
o = [o_!!~v )] - ~ {[o!t'~~ .)] }

o( I.. dx 0(. I..
= ig~K"(X)Y"vTAKJB\(x) - ~J",.y"v'
Hence we have
or equivalently
(6.6.13a)
Furthermore, the Euler- Lagrange equations for the gauge fields are
o- [o!t'( .. )] d {[o!t'( .. )] }
- oyAs I.. - dx r oyAsr I ..
= ig~K"(x)ys"v ~KJIjIJv(X) - (g/2)CEDAFESj(x)BDj(x) + (1/2)FAsr,r'

The above equations are equivalent to
FAar,r = gCEDAFEaix)BDj(x) - 2ig~K(u)(x)y~KJIjIJ(x). (6.6.l3b)
The Lagrangian in (6.6.12) is not only invariant under Poincare trans-
formations, it is also invariant under some gauge transformations. To in-
vestigate the local gauge invariance, we define the following local gauge
transformations:
00
exp[M] == I
n~O
[Mnn!,
U(x) == exp[ieA(x)~],
IjIV(x) == [ljIf:(X)] ,
t/J;(x) (6.6.14a)
~"(x) == s(u)IjI"(x)t.
1jI"(x) = U(Xflljl'V(X),
~u(x) = ~~(x)U(x),
~B'A .(x) = U(x) 1A U(X)-l BA a(X) - (i/g)U(x) [U(X)-l la, (6.6.14b)

where we have U(x)t = exp[ - iOA(x) TA] = U(Xfl and s(l) = s(2) = - s(3) =
-s(4) = 1.
The transformation of gauge field components is described in the following
lemma.
Lemma (6.6.1): Under the local gauge transformation (6.6.l4b) the gauge field
components transform by the rule
(6.6.l4c)
Proof: From equation (6.6.14b) it follows that

~B'\.l = U(x)TAU(xr1B\.1 + U.I~U(xrlBAk(X)
+ U(x)~[U(x)-ll,BAk(x)
- (i/g) {U.I[U(xr1lk + U(x) [U(x)-llkl}'
From the definition of the gauge field components in (6.6.12) we obtain
~F'\l = ~B'Al.k - TAB'Ak,l + gC AEF ~B'Ek(X)B'Fl(X)
= ~B'\k - ~B'\,l + ig[TEB'Ek(X), TFB,FI(X)]
= U(x)~ U(X)-l F\I(X)
+ {U,k ~ U(X)-l BA1(x) - U,l ~ U(X)-l BAk(X) + U(x) [U(xr1lk

x U(x)~ U(x)-lBAI(X) - U(x)[U(xrlllU(x}~ U(xrl BAk(X)}
+ (ilg) {U,I[U(x)-llk - U.k[U(xr1l,
- U(X)U-1,kU(X)[U(xr1ll
+ U(x)U- 1,I U(x) [U(x)-lld
Using [U(x)U(xr 1L = 0, so that [U(X)-l L = - U(X)-l U,a U(xr\ we see
that both of the terms enclosed in braces vanish and equation (6.6.l4c)
holds.
Theorem (6.6.1): The Lagrangian in (6.6.12) is invariant under the gauge

transformations in (6,6.l4a) and (6,6.l4b),
Proof: We use equations (6.6. 14a)-(6.6.l4c) and assume that YAB =

- c Trace[~ TB ] where c is a suitable real constant. Then the Lagrangian can
be expressed as
2'(")1" = ~u(x)yauv[t/lv.a + igTABAa(x)t/lV(x)] + (l/8)yAB F\I(x)F Bkl (X)
= ~~(x)U(x)yauv[U(x)-lt/l'V,a + (U(xr1),at/l'V(x)
+ ig~BAa(x)U(x)-lt/l'V(x)]
- (c/8) Trace[~ TB]FAkl(X)FBkl(X)
= ~~(x)ya\[t/I'v.a + ig(U(x)~U(x)-lBAa(x)
- (ilg)U(x)(U(x)-l).a)t/I'V(x)]
- (c/8) Trace {[U(X)-l ~ F'\I(X)U(X) U(X)-l TBF'Bkl(X)U(X)]}
= ~~(x)yauv[t/I'v,a + igTAB'\(x)] - (c/8) Trace[TA T~]F'Akl(X)F'Bkl(X)
= ~~(x)yauv[t/I'v,a + igTAB'\(x)] + (1/8)YAB F,A k1 (X)pBkl(X).
We can define a gauge-covariant derivative by

Dal/lK(x) == IV.a + igTAKJBAa(x)I/IJ(x). (6.6.15)
With the above definition, the Lagrangian and the Euler-Lagrange equa-
tions can be written as
.!l'(. ')1" = ~K(x)yaDal/lK(x) + (1/8)yAB F\,(x)FBk'(x),
yaDal/lK(x) = 0, (6.6.16)
FAab.b = gCEDAFEoj(x)BDj(X) - 2ig~K(x)yOTf/ JI/IJ(x).
Now we shall discuss a special example of the interacting gauge field and
the Dirac field.
Example: Let us choose the semisimple gauge group G = SU(2). It is the

group of unimodular unitary group of transformations of the complex space
"Y 2' The group has three generators til = iTIl such that [T,., Tp] = ie~py I'y. The
group metric is Ypa = - 2c;pa' The 2 x 2 complex matrix representation of T"
is given by I;. = (l/2)er", where erl' is a Pauli matrix. The Lagrangian (6.6.12) is
.!l'(")I" = ~K(X)ya[I/IK.a + (ig/2)er/JB"a(x)I/I J(x)]
- (l/4)c;pF\,(x)F Pk'(x)
= ~u(x)yauv[I/Iv,o + (igI2)erBo(x)I/IV(x)]
- (1/4)c;"IIF kl(X)FPk'(X).
The field equations (6.6.13a) and (6.6.13b) yield in this case
yO[I/IK,Q + (ig/2)er/ JBa(x)I/IJ(x)] = 0,
Fab,b = -ge".~paj(x)Bj(x) - ig~K(x)yaer/JI/IJ(x). D
In the electroweak field theory the gauge group is taken to be G = SU(2) x

U(I). Furthermore in quantum chromodynamics (the theory of the strong in-
teraction of quarks and gluons) the gauge group is chosen to be SU(3).
EXERCISES 6.6
1. Consider the Lagrangian function given in equation (6.6.1). Prove that
2. Obtain the symmetrized energy-momentum-stress tensor lJ,ix) from the

Lagrangian (6.6.12) such that
lJij(x) - T,j(x) = }:.;/,a,
where }:.i/(X) is a 12-tensor field satisfying
}:.ija,ai(X) == O.
3. Using Euler-Lagrange equations (6.6.13a), (6.6.13b), and appropriate
conditions on differentiability, prove the differential identity
[CEDAFEaj(x)BDj(x) - 2itiiK(X)}'a1~/J",J(x)L == O.
References
1. Y. Choquet-Bruhat, C. Dewitt-Morette, and M. Dillard-Bleick, Analysis. manifolds.
and physics, North-Holland Pub!. Co., Amsterdam, 1977. [pp. 154, 168, 174]
2. E. M. Corson, Introduction to tensors. spinors. and relativistic wave-equations,
Blackie and Son Ltd., London, 1955. [pp. 72, 176, 179, 213]
3. K. Huang, Quarks, leptons. and gauge fields, World Scientific Pub!. Co., Singapore,
1982. [pp. 204, 227, 231]
4. D. Lovelock and H. Rund, Tensors. differential forms. and variational principles,
John Wiley and Sons, New York, 1975. [po 176]
5. Y. Takahashi, An introduction of field quantization, Pergamon Press Ltd., Toronto,
1969. [pp. 183, 186, 195,213,217]
6. P. B. Yasskin, Phys. Rev. D 12 (1975), 2212-2217.
7
The Extended (or Covariant) Phase
Space and Classical Fields*
7.1. Classical Fields

In Hamiltonian mechanics and the subsequent quantum (particle) mechanics
the basic equations remain covariant under canonical transformations (see
Chapter 5). A particular example of canonical transformation is given by
(7.1.1)
Born and Lande called this transformation the reciprocity transformation.
The usual quantum field theories do not allow covariance under reciprocity.
Yukawa and his students incorporated the idea of reciprocity in their formu-
lations of nonlocal fields. In such formulations, fields had to be defined on the
eight-dimensional extended phase space (or covariant phase space) Mg (see
Sect. 5.4).
The present author constructed field theories in the four-dimensional com-
plex space-time in order to quantize space-time and at the same time to
utilize the larger group of covariance to interpret internal groups.
However, if complex coordinates are interpreted as z" = qa + ip" accord-
ing to Bargman's representation of quantum mechanics, then fields are auto-
matically defined over the extended phase space (or the covariant phase
space) Mg. In this space a subgroup of transformations can be defined as
qa = (cos rP)qa - (sin rP)pa,
(7.1.2)
p" = (sinrP)qa + (cosrP)p",
where rP E ( - n, n) is the group parameter. For the special value rP = n/2,
(7.1.2) reduces to (7.1.1). This transformation is called a generalized reciproc-
ity. It is a subgroup of canonical transformations (problem 3 of Exercises 5.4)
as well as that of the 36-parameter extended Poincare group. This subgroup
naturally leads to a flat extended phase space Mg with the "line element"
(dt)2 = dij(dqidqi + dpidpj). (7.1.3)
Thus Ms is of signature + 4.
* This chapter is a review of a research topic pursued by several scientists.
168
7.1. Classical Fields 169
The special theory of relativity presupposes that the maximum speed of

propagation of action is the speed of light. Similarly, the: metric in (7.1.3)
implies a "maximal proper acceleration." This point of view has been ex-
plored by Caianiello and others.
Notation will be elaborated now. The physical units are so chosen that
a = b = c = 1, where a, b are the fundamental length and momentum respec-
tively (ab = h) and c is the speed oflight. All physical quantities are expressed
as (dimensionless) numbers. A point in the extended phase space is denoted
by
~ = (~1,~2,~3,4,~5,6,C,~8)
=(ql,q2,q3,q4,p\p2,p3,p4) = (q,p).
Greek indices take values from {I, 2, 3}, lower case Roman indices take from
{l, 2, 3, 4}, and capital Roman indices take from {l, 2, 3,4, 5, 6, 7, 8}. (The
last convention differs from those of the preceding chapters.) In all cases sum-
mation convention is followed. Furthermore, in expressions like ajAqJ, qifjJ,ph
and ~jT44+j' the index j is automatically summed from 1 to 4. Partial deriva-
tives are denoted by
A,." = A,.': = i3fjJ"(q,p)

'I',qJ - 'I'.J - oqj ,
A,. " . = A,." = ofjJ"(q,p)

'I',pJ - 'I',j+4 - opj .
In the eight-dimensional notation, the "line element" (7.1.3) can be expressed

as
(dt)2 = DABded~B,
(7.1.4)
[DAB]=[d~j I ~J.
The nonhomogeneous linear transformations that leave the above line
element invariant is given by
(7.1.5)
pi = di + cijqj + eiipi,
DABttt~ = Den,
dij(aika jl + cikCjl) = dij(bikbjl + eike j l) = dkl ,
dij(aikb jl + cikeil) = O.
170 7. Extended Phase Space and Classical Fields
The above equations yield the 36-parameter extended Poincare group. In

case d i == 0, b ij = c ij == 0, a ij == eij, the usual ten-parameter Poincare group
is recovered. Furthermore, in case b i = d i == 0, a ij = eij = (cos rjJ)(jij' Cij =
_b ij = (sinrjJ)(jij' the generalized reciprocity (7.1.2) is obtained.
Now the tensor transformation rules under an extended Poincare transfor-
mation will be discussed [compare the equation (2.4.11)].
fAI"A r ... (~)=tAI tAr aDI "'aDs TCI'''Cr ... (~),
Bl B, Cl Cr Bl B, D, Ds (7.1.6)
A
[a B] == [tABr
1
Two simple examples will be dealt with. Consider a scalar field rjJ(~). The
transformation is given by simply
$(~) = rjJ(~),
(7.1.7a)
$(4, p) = rjJ(q, p).
Now consider a vector field AB(O. The transformation rule is the following
equations:
All(~) = tBcAC(~),
Aq'(q,p) == Ai(o,
AP'(q,p) == Ai+4(~), (7.1. 7b)
AQ'(4,p) = aijAqi(q,p) + bijApi(q,p),

AP'(4,p) = cijAqi(q,p) + eijApi(q,p).
Note that the summation convention is operating on the index j above.

The contravariant metric tensor components are given by
(7.1.8)
The raising and lowering of indices can be performed as follows:
~(~) == DABTB(~), TA(~) = DABTB(O
~,(q, p) == dij pi(q, p), P'(q, p) = dij~J(q, p), (7.1.9)

Tp,(q, p) == d u pi(q, p), P'(q, p) = dijTpJ(q, p).
Expressions like
1A(~)TA() = dij[P'(q,p)pJ(q,p) + P'(q,p)pJ(q,p)],
UAB(~)VAB(~) = Uq'qJ(q,p)Vq'qJ(q,p) + UqipAq,p)Vq'pJ(q,p) (7.1.10)
+ Up'qJ(q, p) VpiqJ(q, p) + UpipAq,p) VPipi(q, p)

are invariant under the extended Poincare group.
Now Sect. 6.1 will be generalized for the extended phase space
Ms. The invariant action integral for a complex-valued tensor field
7,1. Classical Fields 171
rjJA'" E C 2 (D eMs; q can be defined as
d(rjJ") == III ['(~, "'''''; (A' ".4.')]1(" =qfW,.,.G =4i"~ dS" (7.1.11)
where D is a bounded, simply connected domain in Ms with a piecewise

smooth, orientable, closed boundary aD.
The variational principle
8d(rjJ") = 0,
yields the Euler-Lagrange equation [compare the equation (6.1.7)]
a'( .. ) _ ~{[a~UJ
a(" I" d, a(;.j I..
}= O. (7.1.12)
The infinitesimal version of the extended Poincare transformation (7.1.5)

can be summed up as
(7.1.13)
eBA = -eAB + rAB(eCD),
whereleAI, leABI are arbitrarily small positive numbers. Variations and the Lie
variations of a complex tensor field rPA .. (~) are furnished by
8rjJA .. == JA"(~) _ rP A.. (,)
= (1/2)e cD ScD A"B .. rjJB .. (~) + rA"(e"),

8Lr" == JA .. (,) - r"(,) (7.1.14)
= -eBr",B + (1/2)e CD [('cr",D - 'DrjJA".C>
+ SCDA"B .. rPB .. (~)] + fA"(e"),
where SCDA"B .. is a numerical tensor representing the "extended spin" and
remainder terms r A.. , fA .. are of order 0(e 2 ).
The invariance of the action integral with respect to transformations
(7.1.13) leads to Noether's theorem (compare Theorem 6.1.2):
d {
d;A
.,
[2'(. ')]I,,(j~ A + [a'(.
~
~ A
')J . + [a2'
I..
(")J -..} o.
--.-.
(jLrP
a(A I"
8LrP =
(7.1.15)
By considering different possible cases for (jLrP", one obtains various differ-
ential conservation laws. For example, in the case eA =F 0, eAB == 0, using
(7.1.13) and (7.1.14), the equation (7.1.15) yields
TAB,A = 0,
(7.1.16)
In case eA == 0, eAB i= 0, the equation (7.1.15) provides another conserva-

tion law, namely,
jABc,A = 0, jABd~) == MABd~) + 9'ABd~) = _jACB(~)'
MABde) == ~CTAB(~) - ~BTAd~), (7.1.17)
A - [OlE( . . )] ....
9' Bd~) = ~ SBC"tP (~) + (c.c.).
'>A i .
The infinitesimal version of the generalized reciprocity transformation
(7.1.2) is given by
(7.1.18)
As a consequence of this invariance the following conservation law emerges:

j '(A) .A = ,
T(A)(~) == ~jTAj+4C~) - ,j+4 TAi ,),

B (A)C):)_[OlE(
'> = ~
.. )] dijCSi+4,i.. .. - S"i,i+4 .. ),
'I'
.. ():)+(
'>
)
C.c. ,
'>A i..
(7.1.19)
T(4)(q, p) = qiT\j(q, p) - pjT\J(q, p),
B(4)(q,p) = [O~r\:')] dii(Sp;qj:. - Sq;pj:.)tP'"(q,p) + (c.c.).

'>4 i ..
In case the action integral (7.1.11) is invariant under an infinitesimal phase

transformation (nongeometrical!)
~(~) = tPC,)exp(ie),
(7.1.20)
~(O = ~(Oexp(-ie),
where lei> 0, a conservation rule follows, namely,
VA,A = 0,
(7.1.21 )
vAC~) == i {[alEC .. )]
-_.-. -
tP"C,) - Cc.c.) } .
O(A i..
One would consider ,4
= q4, the time coordinate as the preferred one to
obtain integral conservation laws. Furthermore, a hypercylindrical, bounded
domain Dc Ms will be picked up such that D lies between two distinct
~4 = constant hyperplanes and surrounded by a wall 17 (see Figs. 26, 27). It is

assumed that TAB(~) = v A(,) == outside the phase space tube containing D.
Across the wall 17, the usual jump conditions
TABnAiO' = jABcnAia = vAnAiO' = 0
are imposed, where n A is the unit outward normal to 17. Applying the eight-
7.1. Classical Fields 173
FIGURE 27. An extended world-tube in IRs.
dimensional Gauss theorem to the differential conservation laws, we derive

the following integral constants:
KA == f- T4A(~)d7~,
f- rBd~)d7~
Do
JBC == = -JCB ,
f- [T(4)(~)
Do
J == + (1/2)B(4)( ~)] d7~, (7.1.22)

Do
d 7 ~ == d~l d~2 d~3 d~5 d~6 de d~8,

where Do is a seven-dimensional bounded domain for which ~4 = q4 = const.
(see Figure 27).
These integrals are independent of the time coordinate ~4 == q4. Therefore,
K A, JBC , J, N are called constants of motion (or evolution) of the field, or, in
short, integral constants. The domain Do is usually extended to the whole of

seven-dimensional space 1R7 == g E IRs: ~4 = O}, and it is assumed that these
improper integrals converge uniformly. With these assumptions and equa-
tions (7.1.16)-(7.1.19), (7.1.21), (7.1.22) we can express the constants of motion
of the field q>"(~) as
Ka - - f {[8!l'(")J"
= K q, - - ~ rP.q, + (c.c.) } d3 qd 4 p,
- -f ')J .
H;!7 '>4 I.. Iq4=0
H = -K4 - {[8!l'(.
~ rP,q4 + (c.c.) - [!l'( .. n.. } d 3qd 4 p,
- -f
[R7 .. 4 I.. Iq4=0
Xj = Kj+ 4 - - {[8!l'(
~ . . ) ] .rP,pi
. + (c. c.) } d 3 q d 4 p,
H;!7 ':.4 I.. Iq4=0
(7.1.23)
X [-8!l'8
(")Jr"
.. 4 I..
+ (c.c.) - p4 [!l'( .. n.. }
Iq4=0
d3 qd4 p,
Q == eoJ,
N -= . f
I
[R7
{[8!l'(")J
---
..-
8(4 I..
'J:.. -
'I' (c. c.) }
Iq4=0
d3 q d4 p,
where eo is the charge parameter.

Now, physical interpretations of various constants in (7.1.23) are in order.
With regards to the r/J"-field distribution in the extended phase space Ms:
(i) Ka are components of the total three-momentum;
(ii) H is the total energy;
(iii) Xa are components of spatial center of the distribution;
(iv) X 4 is the time center of the distribution;
(v) Jij are components of the total "relativistic" angular momentum;
(vi) Lij are components of the total "angular momentum" in the four-
momentum space;
(vii) Q, which arises from the sum of isotropic contribution eo T(4)(~) and
baryonic contributions (eo/2)B4( e), is the total charge;
(viii) N is the total content of the field (which can represent the total num-
ber of quanta after the second quantization).
To contrast with the usual relativistic field theory, we may notice that only
integral constants that emerge in that arena are K a , H, Jij , Q == eN [see
equation (6.1.24)]. Therefore, in the present formulation, we have many more
integral constants.
7.2. The Generalized Klein-Gordon Equation

In this section the generalized Klein-Gordon equation for a scalar field de-
fined over Ms will be discussed. Yukawa pursued Born's reciprocity covari-
ance in his theory of the nonlocal scalar field. He advocated two equations
(one for the space-time variables, another for the four-momentum variables)
for this scalar field. In this section, the generalized Klein-Gordon equation is
obtained by "gluing" his two equations together, so to say. The first advan-
tage of this procedure is that group of covariance is enlarged to the extended
Poincare group, and, secondly, the scalar field thus obtained can describe an
arbitrary number of mesons in a unified manner.
The Lagrangian function in this case is chosen to be 2: C x C X C B X
C -+ IR such that
B
2(',C"A,CA) == _[DABCA'B + jJ 2 Cn,

(7.2.1)
02~.. ) = 02! .. ) = _DABrB
0' ,. "',
_,,2r
O'A '" ,
where jJ is the (real) mass parameter.
The Euler-Lagrange equations (7.1.12) [compare equation (6.2.3)] yield in
this case
[02~.. )] __ dA{[02! .. )] }
a, I.. de a'A I..
..
2A, ):
= -jJ ."(,,,) + DABrP,AB = d'J(rP,q;qj + "',p;pJ) -
A,
jJ
2
rP(q,p) = o. (7.2.2)
The above equation is obviously invariant under the extended Poincare
group. From the transformation rule ~(e) = rP(e) and the equation (7.1.14), it
is clear that
SCD == 0, r(e) = P(e") == O.

The various tensor densities obeying differential conservation rules can be
computed from (7.2.1), (7.1.16), (7.1.17), (7.1.19), (7.1.21). These are furnished
by
TAB(e) = JA B[2(. ')]1" - [a~~")J rP.B - [02!"~J (fi.B

"'A I.. a'A I ..
= DAC[(fi,CrP,B + (fi,BrP,C] - JAB[DCE(fi,crP,E + jJ 2 IrP(eW],

jABd~) = MABd~)
= ~dDAE[;P,E,B + ;P,Br/>,E] + b AB[2'( .. )] I.. }

-~B{DAE[;P,E.C + ;P,C,E] + b AC[2'(")I .. },
gABd~) =0, (7.2.3)

/A)(~) = T(A)(~)
= ~J D
'{ AC (,C,j+4
- -
+ ,j+4r/>,d + b Aj+4[2'( ')1" }
_~j+4{DAC(;P,C,j + ;P,j,d + b Aj[2'( .. H .. },
BA(~) = 0,
VA(~) = iDAB[;P,B(~) - ;P(Or/>,B]'
The constants of motion (or integral constants) for the scalar field follow
from (7.2.1) and (7.1.23). These are given by
Ka =K q, = - f [;P,q4,q'
~7
+ (c.C)]lq4=od 3qd4p,
+ ;P,p',p' - ;P,p4if;,p4 + 1l 21if;12] Iq4=O d 3q d4p,
Xi =K i+4 =Kpi = - f ~7
[;P,q4,pi + (C.C)]I .. d3 qd 4p,
Jij = f {(q,;P,qi - %;P,q.),q4 + (C,C.) + (qjb4, - qib4j)[2'( .. )h.h.d3qd4p,
f - -
~7
(7.2.4)
Lij = [(P,,pi - Pj,pi),q4 + (C.C.H .. d 3 4
q d p,
~7
Q = eoJ = eof {(pi;P,qi - q';P,p.),q4 + (c.c,) - p4[2'( .. H .. h.d 3qd 4p,

~7
B = 0,
N = if [;P(q,P)'if;,q4 - ;P,q4(q,pn .. d3 qd 4p,

~7
Now we shall introduce the Fourier dual of the seven-dimensional space of

(ql,q2,q3,pl,p2,p3,p4). A subset D7 of this dual space will be essential in
subsequent computations. It is defined by
In this domain the following function satisfies the inequality

O(k, X) == Jkaka + XaXa + Jl2 - (X4)2 > 0,
w = W(k,x) == Jkaka + XaXa + Jl2 > 0, (7.2.6)
w' = W(k', x') == Jk~k~ + x~x~ + Jl2 > 0.
A class of solutions of the generalized Klein-Gordon equation (7.2.2) can
be furnished by [compare equation (6.2.7)]
fjJ(q,p) = (2n)-7/2f [20(k,x)]-1/2{tX(k,x)exp[i(kaq a + X"pa)]

D7
+ P(k, x) exp[ - i(k"q" + XaP")]} dlk d4x,

iP(q,p) = (2n)-7/2f [20(x, X)]-1/2 {tX(k,x)exp[ -i(kaqa + xapa)]
D7
+ P(k, x) exp[i(kaqa + xapa)]} d 3 k d4x,
k4 == -O(k, x) = -Jka.ka. + Xa.Xa. + Jl2 - X~,
fjJ,q4 = - i(2nt7/2 f (0/2)1/2 {tX exp[i(kaqa + xapa)] (7.2.7)

D7
- Pexp[ - i(k"qa + XaP")]} d 3 k d4x,
iP,q" = -i(2nt7/2 f D7
(20 t 1/2ka.{iiexp[ -i(k"q" + x"p")]
-pexp[i(kaq" + XaP")]}d l kd4x,
iP,p. = -i(2nt7 /2 fD7
xa{iiexp[ -i(k"qa + XaP")]
- P exp[i(k"qa + xapa)]} d 3 k d4x.
Here we have assumed that the above seven-dimensional improper integrals
converge uniformly and that differentiations under the integral signs are
permitted. The complex-valued functions tX, P are otherwise arbitrary. We
shall compute the constants of motion (7.2.4) using this class of solutions
(which may be considered as plane wave superposition). The easiest one to
calculate is N, and it will be accomplished presently. Using (7.2.4), (7.2.7) we
have
N = I - P)fjJ,q4]
'f [fjJ(q,
OF
Iq4=O d l q d4 p + (C.C.)
= (2)-1 f {f
D7 D7
(n/n,)1/2[(ii'tX - P'P)Dl(k - k')D 4 (X - x')
+ (P'tX - ii'P)Dl(k + k')D 3(X + x')] d3 k' d4x'} d3 k d4x + (C.C.)

= (2fl f {f
D7 1R6
f""
-W'
[(iX'a - fJ'P)b 3 (k - k')
X b 3 (x - x')b(x 4 - X 4 ') + (f3'a - iX'P)b 3 (k + k')

x b 3 (x + X')b(X4 + X 4 ')] d3 k' d 3 x' dx~ } d 3 k d4 x + (C.C,)
= (2)-1 fD7
{(laI 2 -1f31 2 )
+ [fJ( -k, -x)a(k,x) - iX( -k, -x)P(k, X)]} d 3kd 4x + (c.c.), (7.2.8)
Now let us consider the second part of the integral (with a substitution k' =
-k, x' = -x)
/ == f~oo"'" f~oo f~", [f3( - k, - x, - x4)a(k, x, X4)
- iX( - k, - x, - x 4)P(k, x, x 4 )] d3 k d4 x + (c. c.}

= - f~oo ..... f~oc f: w
[fJ(k', x', x~}a( - k', - x', - x~)
- iX(k, x', x~)P( - k', - x', - x~)] d3 k' d4 x' + (c.c.)
= -/ = O. (7.2.9)
Therefore, from (7.2.8) we obtain
N = f D7
[la(k,xW -1f3(k,x)1 2 ]d 3 kd 4 x. (7.2.10)
Now let us compute another important integral constant, namely, the total
energy H from equations (7.2.4), (7.2.7):
H = (2)-1 (2nf7 fff

j;!7 D7 D7
(00')-1/2
X {(O'O + k;ka + x;xa - X~X4)

x [iX'exp[ -i( .. )'] - f3'exp[i( .. )']]
x [aexp[i( .. )] -pexp[ -i( .. )]]
+ jl2[iX'exp[ -i( .. )'] + f3'exp[i( .. )']]
x [aexp[i( .. )] + pexp[ -i( .. )]]} d 3 qd4 pd 3 k' d4 x' d3 kd 4 x
= (2)-1 f
D7
f D7
(rul')1/2
x [(0'0 + k~ka + x~xa - X~X4 + jl2)

x (iX'a + fJ'lJ)b 3 (k - k')b 4(x - x')
- (0'0 + k~ka + x~xa - X~X4 _ jl2)
x (iX'P + f3'a)b 3 (k + k')b 4 (x + x')] d 3 k' d4 x' d 3 k d4 x. (7.2.11)
Now let us digress slightly to consider an integral of the type
t'[f~~/(kl, k,X',X,X~,X4)5(X4 - x~)dx~ ] 53(k - k ' )53(X - xl)d 3 k' d 3 x '
= {!(k,k,x,X,X 4,X 4) for X4"E (-w,w),

o forx 4 (-w,w).
Substituting the above result into the equation (7.2.11) we obtain
H = (2)-1 f (Q)-l {[Q2 + k~k~ + x~x~ - xi + J12] [1~12 + IPI 2]

D7
- [Q 2 _ + xi - J12]
k~k~ - x~x~
X [a(-k, -x)P(k, x) + a(k,x)p(-k, -x)]}d 3kd 4x
= fD7
[la(k,x)1 2 + IP(k,xW]Q(k,x)d 3 kd 4 x. (7.2.12)
Now we shall summarize computations of many constants of motion:
K~ = f D7
[f+(k,x) + !_(k,x)]k~d3kd4X,
H = -K4 = f [f+(k,x) + !_(k,x)]Q(k,x)d 3 kd 4x,
f
D7
Xa = [f+(k,x) + !_(k,x)]xad 3 kd 4 x,
D7
(7.2.13)
B= 0,
N = f D7
[f+(k,x) - !_(k,x)]d 3 kd 4x,
!+(k,x) == la(k,x)1 2 :?: 0, !_(k,x) == IP(k,xW :?: O.

Some physical interpretations can be reached now. The functions !+ (k, x),
!_(k, x) can be naturally interpreted as a Boltzmann type of distribution
function for the similarly created particles and antiparticles. This view point
supports the statistical interpretation of quantum mechanics for Bosons. All
the integral constants belong to the scalar field representing the unified me-
son field ensemble. In the equation (7.2.13),
(i) K~ represent the total three-momentum components;
(ii) H represents the total energy;
(iii) X are components of the first moment for the distribution;
(iv) X 4 is the center of time for the distribution;
(v) B = 0 is the baryonic content;
(vi) N is particles minus antiparticle content.
Computations of Jij, L ii , J are extremely cumbersome. However, for a physi-

cal theory we do need the integral constant J to evaluate the total charge Q.
We shall make only a partial attempt to perform this calculation now. We
note that
J=1 1 +1 2 ,
11 == f [(piifj,q' - qiifj,p'] Iq4=O d 3q d4p + (c.c.),
-f
GO!7
(7.2.14)
12 == {p4[2"(n}lq4=od 3qd 4p.
GO!7
We further simplify the situation by putting the antiparticle amplitude

f3(k, x) == 0. In that case
12 == (2r1(2nr7 ff GO! 7 D7
f
D7
p4(nnT1/2
{(k~k, + X~Xa - n'n - X~X4 + /1 2 )(a'a)

x exp[ik a - k~)q' + (xu - x~)pa)]} d 3qd4pd 3k' d4x' d 3kd 4x
= (4n)-1 f fro f p4[n(k, x, x~)n(k, x, X4 )]-1/2

GO! -ro D7
x {(k,k, + x,x, + /1 2 - n'n - x~x4)a(k, x, x~)

x a(k,x,x 4)exp[i(x 4 - x~)p4]}dp4dx~d3kd4X.
Since 12 is real we can write
x {(k,k, + x,xa + /1 2 - n'n - X~X4)

[a(k, x, x~)o:(k, x, X4) exp[i(x4 - X~)p4]
+ o:(k,x,x4)a(k,x,x 4)exp[ -i(X4 - X~)p4]]} dp4dx~d3kd4X. (7.2.15)
Let us work out the Cauchy principal value of the integral:
c.P.V. f~oo p4 {iX(k, x, x4)o:(k, x, x 4 ) exp[i(x4 - X4)p4]
+ o:(k, x, x~)a(k, x, x 4)exp[ - i(X4 - X~)p4]} dp4
{ofor X4 = x~,
= i2(X4 - x4)-llimL _ oo [a(k, x, x 4)iX(k, x, x 4 ) - (c.c.)]
x [LCOSX4 - x~)L) - (X4 - x~)-lsin(x4 - x 4)L)]
With the further simplifying assumption

Q!(k, x, x~)iX(k, x, X4) = iX(k, x, x~)Q!(k, x, X4), (7.2.16)
the integral 12 = O. The computation of 11 in (7.2.14) will be done later.

Now, to develop the unified meson field theory, we transform coordinates
qq, pa of each phase plane to the corresponding cylindrical coordinates by the
following equations:
qa = pQ cos oa, pa = pa sin oa;
(7.2.17)
pQ = J(qaf + (pa)2, oa = arc(qQ,pa).
Here the subscript 'a' is not summed, and furthermore coordinates (pa)2, oa
are a canonically conjugate pair. To make the transformation one-to-one, we
restrict pQ > 0 and loai < 7l for each a E {l, 2, 3, 4}. We define the new scalar
field as ~(p, 0) == iP(q, pl. The generalized Klein-Gordon equation goes over
to
(7.2.18)
Separating variables and demanding that angular functions are (properly)
defined, we obtain the basic solutions as
(7.2.19)
where (t) == (tl' t 2, t 3, t 4) and ta E Z. The function x1t) satisfies the partial differ-
ential equation
(7.2.20)
If we define the (mass)2 operator associated with the meson field X(t)(p) or
~(t)(p, 0) as
then the equation (7.2.20) implies that

M 2iP(t)(p,0)
= [J.l2 + (tdpl)2 + (t 2Ip2)2 + (t3/p3)2 - (t4/p4)2]iP(t)(p,O). (7.2.22)
The above equation shows some analogy with Okubo's mass formula pro-
vided t 1 , t 2 , t3 are associated with isotopic numbers, and 2t4 is associated
with the strangeness quantum number.
The basic solution iP(t)(p, 0) can be expressed in terms of Bessel functions of
the first kind [see the equation (7.2.41-iv)]
iP(t)(p,O) = [D Jt.(KbPb)] exp[ita Oa ],

(7.2.23)
where the index b is not summed. The asymptotic version of (7.2.23) for large
values of pb'S is
J(/)(p,O) = (2nf2 [il ( Kbp b)-1/2 COS(Kbpb - 2- 1ntb - 4-In) }XP[ita oa ]
+ O Kbp b)-l). (7.2.23')

Bessel functions of the second kind are avoided because of the singular behav-
ior at the origin. A general solution (but not the most general solution) of
(7.2.18) can be constructed by linear combinations of basic solutions. We can
arrive at such a class of general solutions directly from the expression (7.2.7).
For that purpose we make the coordinate transformation
(7.2.24)
(Please distinguish between the scalar field J and angles J.) The domain of this
transformation is D7 in (7.2.5) (minus some points) and the range is given by
157 == {(~,~,X4) E [R7: 0< Ka, lJal < n, IX41 < w},
(7.2.25)
w = W(k, x) = W(~) = KaKa + /1.2.
Now using (7.2.17), (7.2.24), and Jacobi-Anger's formula (7.2.41-x) for
Bessel functions we obtain
= L
<Xl
(i)/'-",( K1p 1)exp[it 1(Ol - J1)]. (7.2.26)

t1 =-00
Utilizing the above and similar expressions, the scalar field in (7.2.7) can be
written as
~(p,fl,q\p4) = (2n)-7/2
-
fD7
[2Jw 2 - x~r1/2
x 1;~<Xl {a(~,~,x4)LQ (i)'PJ1,(KPpfJ)]

x exp[i(t,,(Oa - J") + k4q4 + q4p4)]
+ P(!~'~'X4{}] (-i)'fJ-",(KppfJ)]
x exp[ -i(ta(O'1 - Ja) + k4q4 + q4P4)]} K1K2K3 d3~d3~dx4'

~(e., fl, q4, p4) == J(q, p),
a(~, J, x 4 ) = o:(k, x), p(~, J, X4) == P(k, x),
(7.2.27)
k4 == -Q(k,x) = -Jw2 - x~.
Now we make a coordinate transformation for which the inverse transfor-

mation is given by
L
00
exp[ -i(k4q 4 + X4p4)] = (i)'4Jt.(O)p4)exp[it4(e 4 + ,p4)]. (7.2.28)
t 4 =-00
The usual domain of this transformation is the (k4-x4)-plane minus the non-
negative k4 half-line (0) > 0, 1,p41 < n). If we insist to keep k4 < 0 (which
is mathematically not imperative) then we have to choose 0) > 0, 1,p41 < n12.
Logically speaking, we can also try multiply connected Riemann-like sur-
faces for which 0) > 0, l,pl < mI:, n E {2, 3, 4, ... }. Out of all these choices, the
usual choice 0) > 0, l,pl < n yields the neatest expressions for the integral
constants. On that account, we shall adopt the preceding choice so that the
integration has to be performed over the subset
Dt == {(~,,p) E [R7: 0< K" l,pil < n}. (7.2.29)

The scalar field in (7.2.27) goes over to
,p#(p, e) = (2n)-7 /2 fD~

[(0) cos ,p4)/2] 1/2
X t=~oo {1i(.!f,~,O)sin,p4{t' (i)'PJt#(KppP)]

x (-i)'4J. 4 (O)p4)exp[i(t,(e - ,p') - t4(e 4 + ,p4]
+ b(.,., .)[fI (-i)PJt (KfJPfJ)] (i)' 4Jt.(O)p4)

P=l #
(7.2.30)
x exp[ -i(t,(e' - ,p') + t4(e 4 + ,p4))] }K1K2K3d3~d4,p,
,p# (p, e) == J(e.,~, q4, p4).

Let us consider the integrals
IX#(~,~, (4) == (4n)-1/2 f:" Ii(.!f,~, 0) sin ,p4)
x (0) cos ,p4)1/2 exp[ - it4,p4] d,p4,
f:"
(7.2.31 )
P#(.,., .) == (4n)-1/2 b(.,.,.)
x (0) cos ,p4)1/2 exp[it4,p4] d,p4.
This equation can be inverted (under usual conditions) by the theory of

Fourier series to obtain

[(w cos 4)/2r/2a(~, , w sin 4)
L
00
= (2n)-1/2 C(#(~,,t4)exp[jt44],
(7.2.32)
L
00
= (2n)-1/2 Jj#(.,., .)exp[ -it44].
Now we shall define the following functions:
y(~,t) == 21/2(2n)-3/2 f,,' f" C(#(~,,t4)exp[ -itpP]d 3,
v(~, t) == 21/2(2nr3/2 f,,' f:" Jj#(~, , t4)exp[itpP] d3t

(7.2.33)
These formulae can be inverted to obtain
L
00
C(#(~, , t 4 ) = r1/2(2n)-3/2 y( .. ) exp[itllII],

(7.2.34)
L
00
Jj#(., .,.) = 2- 1/2(2nr 3 / 2 v(., .)exp[ -itllII],

l;;-oo
Let us denote D! == {(~, ) E [R6: 0 < K~, IIII < n}. Then utilizing (7.2.31),
(7.2.33), the equation (7.2.30) (assuming that interchange of integrations and
summations is valid) yields
#(p, 8) = (2n)-3 L
t
fD: {C(#(~, , t4)
11=1
[n
(i)'11 J,P(KIIPIl)]
x exp[it~(8~ - ")]( -i)'4Jt.(wp4)exp[ -it484 ]
+ Jj#(~, , t4>[A (- i)'11Jtp(Kppll)]
x exp[ -it,,(8~ - ")](i)'4Jt.(wp4)exP[it484]}
x KIK2K3d3~d3
(t)(p,8) == r1/2(2n)-3/2 fooo. ..{y(~, t) [rl

P=l
(i)'P Jtp(KppP)]
x ( - i)'4Jt4 (Wp4) exp[i(t~8~ - t 484 )]
+ v(~, t>[ll (- i)'11Jtp(Kppll)] (i)' 4Jt'( Wp 4)
x exp[ -i(t,,8" - t 484 )] }KIK2 K3 d3~.

(7.2.35)
Assuming that above integral converge uniformly and differentiations under

integral signs are permitted, the fields rp#(p,O) and rp(r)(p,O) satisfy partial
differential equations (7.2.18) and (7.2.22) respectively. We postulate that
rp(r)(p,8) represents a particular meson field. In that sense the equation
rp#(p, 8) = I~-oo rp(t)(p, 8) stands for the unified meson field.
Now we shall compute many constants of motion in terms of meson fields
rp(r)(p,8). In these calculations we shall put antiparticle amplitude f3(k, x) == 0
for the sake of simplicity. Consider the constant N in the equation (7.2.10)
(with f3 == 0). It is given by
N = fD7
IIX(k,xW d 3 kd 4 x
= f la(~,p,wsinrp4WwcoSrp4KlK2K3d3!id4rp.
D7
# (7.2.36)
Now using (7.2.32), the integral
2 f:" I[w cos rp4)/2] 1/2a(~, p, w sin rp4W drp4
= (nfl t t {IX#(~' t4)ii#(~,

p, p, t4)
[f"
x exp[i(t4 - 4)rp4] dIV4]}
Therefore, from (7.2.36) we obtain
N = 2 If. 11X#(~'P,t4WKIK2K3d3~d3p
'4 D6
x [(2nf3 J:" exp[i(tp - fp)rpP]d3p]}KIK2K3d3~

= I foo . ly(~,tWKIK2K3d3~.
t 0+
(7.2.37)
Let us consider the total energy H from the equation (7.2.12) (with f3 == 0):
H= f D7
1Ck,xWQ(k,x)d 3kd 4x
= f- 1&(~,t,X4)12(W2 - X~)1/2"1"2"3d3~d3tdx4
D7
=f # 11X(~,twsint;64)12(wcost;64)2"1"2"3d3~d4t;6. (7.2.38)
D7
Now the integral (with the equation (7.2.32))
w2 f:" &(~, t.I w sin t;6412 cos 2 t;64 dt;64
= f:" 1[(wcost;64)/2]1/21X(~,t,wsint;64Wcost;64dt;64
2w
= w L L {a#(~'t,t4)C(#(~'t,f4)
t4 ;4
x [(21t)-1 f:" (exp[i(t4 - t4 + 1)t;64]
+ exp[i(t4 - t4 - 1)]t;64)dt;64]}
= W L {iX#(~.t;6,t4)[C(#(~,t;6,t4 - 1) + C(#(~,t;6,t4 + I)]}.

'4 - - -
Therefore, from (7.2.38), (7.2.33) we have
H =L
I.
f {iX#(~,~,t4)[C(#(~,~,t4
#
D.
- 1)
+ C(#(~,~.t4 + 1)]}w"1"2"3d3~d3t
= (2)-1 L L Li fOO .. {y(~. i,t4)[y(~,t,t4 -
10 t 0+
1) + y(~.t.t4 + I)]}
X {(21t)-3 f:" . exp[i(tp - tp)t;6I1] d3t} W"1 "2"3 d3~

= (2)-1 L fOO .. {y(~, t) [1'(~. t, t4 - 1)
1 0+
(7.2.39)
The reality of the right-hand side can be proved.
Now we shall compute the integral constant J from the equation (7.2.14)
with assumptions p == 0 == v and equation (7.2.16). In that case
J = 11 = f [(pj~,qj
Oil'
- qj~.pj)t;6.q']lqo=o d q d p + (c.c.).
3 4 (7.2.40)
We shall compute J using equation (7.2.35) involving Bessd functions. For

that reason, we digress to list some relevant formulae for Bessel functions,
namely,
L,(p) = (-I)'J,(p), (7.2.41-i)
(djdp)J,(p) = (lj2) [J,-l(P) - J,+l(P)], (7.2.41-ii)
J,-l(P) + J,+l(P) = (2tjp)J,(p), (7.2.41-iii)
[(d 2 jdp2) + (ljp)(djdp) + 1 - (t2jp2)Jlr(p) = 0, (7.2.41-iv)
(8j8q) [J,(kp)ei'B] = (k/2) [J,_l(-)e+ iB - J'+l()eiB]ei'B
= [(lj2)kcos8(J'_1 - J,+tl =+= (itjp) sin 8J,]eitB,

(7.2.4I-v)
(8/8p)[J,(kp)ei'B] = (ik/2) [J'_l(-)e+ iB + J'+1(-)]ei,e

= [(1j2)ksin8(J'-1 - J'+1) (itjp)cos8J,]eitB,
(7.2.41-vi)
(7.2.41-vii)
f oo
0+
[J,(wp)]2(pfl dp = (2tft. (7.2.41-viii)
i foo J,(p)Ji(p)(p)-l dp
0+
for i = 0,
[sgn(i)]lil[sin(lil nj2)] (inj2) for t = 0, i if. 0,
= Set, i):= r 1 sgn(t) for i # 0, It I = Iii,
2(nfl [sgn(t)] It I[sgn(i)] lil i[li12 - Itl 2 r 1 sin[(1 il - Itl)nj2]
for f # 0, t # 0, It I # IiI,
(7.2.41-ix)
I
00
exp[ipcos8] = (i)'Jt (p)exp[it8]. (7.2.41-x)

t=-oo
We use now the equation (7.2.35) (with v == 0) and equations (7.2.41-v),

(7.2.41-vi) to obtain
-;jJ#(p, 8) =r 1/2(2n)-3/2 L foo .. Y(!f, t)[ TI (- i)'tJJ,p(KtJ PP )] (i)'4J,.(Wp4)

I 0+ tJ=l
ifJ,~. = rl/2(2n)-3/2 L foo .. y(.,

, 0+
.)[TI P=l
(i)'PJ"(')](_i)'4
x [(wI2)(cos8 4)(J'4_ 1 (') - J,.+1('

+ it4(p4rl(sin 84 )J,.(.)] exp[i(taW - t 484)] Kl K2K3 d3~,
qj~,pj - pj~,qj = pj cos 8 j [(sin 8j)ifJ,~j + (cos 8j)(pjrl~,~j]
- pi sin 8i[(cos 8i)~,~j - (sin 8j)(pi)-I~,~j]
4
= L ~,~j = _i2- 1/2 (2n)-3/2
i=1
x L (tl + t2 + t3 -
,
t4 ) fOO .. y(.,
0+
.)[TIP=1
(-i)'fJJ,p(')]
X (i)'4J'4<- )exp[ - i(taW - t 404)]Kl K2K3 d3~.

(7.2.42)
Now, q4 = 0 implies that either 8 = nl2 or else 8 = - nl2 [by equation
4 4
(7.2.17)]. For 84 = nl2 we have p4 > 0, whereas for 84 = - nl2 we get p4 < O.
Let us compute a part Ii of J from (7.2.14), (7.2.42) at 84 = n12.
x f foo ... y(~,

D~ 0+
t)y(g, i) [TI (-
fJ=1
i)'f!(i)'fJJ,p(' )Ji,()]
X (i)"( _i)'4-'r.(wp4)-h4(wp4)exp[ita - taW" - (t4 - t 4)(nI2))]

x plp2p3(p4flKIK2K3KIK2K3 d4pd3fld3~d3g + (c.c.). (7.2.43)
Using the 8-integrations and the equation (7.2.41-vii), (7.2.41-ix) the above
equation yields
Ii = _(2)-1 L L L (-1)'4+i4(tl + t2 + t3 - t 4 )t4
t 14 i4
X fOO foo .. y(~, t, t4)Y(~' t, (4)J,.(wp4)Ji.(wp4)(p4rl

0+ 0+
= _(2)-1 L L L (-1)/4+i4(tl + t2 + t3 - t4 )
t t4 i4
(7.2.44)
Similar computation yields for 8 = - n12,
Ii = (2)-1 L L L (tl + t2 + t3 - t4 )
t '4 i4
X fOO .. y(~, t)y(~, t,t4)S(t 4,t4)K 1 K2 K3 d3~ + (c.c.). (7.2.45)
0+
Therefore, the integral constant
J = Ii + 11 = r l I I [1 - (-1Y4+i 4](t l + t2 + t3 - t 4 )
t ;4
We summarize many constants of motion in the following:
2H =r l I (00 .. {y(~, t)[y(~, t, t4 - 1) + y(~, t, t4 + I)]}

t Jo+
B:=O,
Q = eoJ = 2- le o I I [1 - (_1)/4+i4] (t l + t2 + t3 - t4 )S(t4, (4)

t i4
(7.2.47)
Suppose that we consider a mixture of four kinds of elemental meson fields

,p(T,. T 4)(p, 0), ,p(T,. T 4+ 1)(p, 0), ,p(T,+l. T 4)(p, 0), ,p(T,. T 4+l)(p, 0), ,p(T,+l. T 4+ 1 )(p, 0).
Furthermore, let the meson field amplitude y(~, t) satisfy the special condi-
tions characterized by
JF(K' T3, T4)exp[ig(~, T3, T)]

(2Kl K2K3)1/2y(~, t) := { for tl = t2 = 0; t3 = T3, T3 + 1; t4 = T4, T4 + 1;
o otherwise.
(7.2.48)
This condition physically means that the special four modes are equally ex-
cited and other modes are absent. In this case, the constants of motion (7.2.47)
reduce to
KI = K2 = 0,
K3= foo .. F(!i, T3, T4)"3 d3!i,
0+
H = foo .. F(!i, T3, T4)wd!i,

0+
Xl = X 2 = X3 = 0,
B=O,
N = 2 foo .. F(!i, T3, T4) d3!i,

0+
Q = 4(nf 1 eo(T3 - T4 ) foo .. F(!i, T3, T4) d 3k = e(T3 -

0+
T4 )N,
e == 2(nf l eo. (7.2.49)

In this case F should be interpreted as a statistical distribution of mesons in
"momentum and charge" space. The expression for the charge resembles
Gellman-Nishijima's formula for the meson charge provided T3 is identified
with isotropic quantum number and 2T4 is identified with strangeness quan-
tum number. In (7.2.49) each of the momentum component is integrated over
IR+ instead of IR in the usual theory. This difference can be reconciled by
noting that for an absolutely integrable, piecewise differentiable real-valued
function f, the Fourier integral is
f(x) = too [a(k) cos kx + b(k) sin kx] dk = too [y(k)e ikx + y(k)e- ikX ] dk,
y(k) == 2- 1 [a(k) - ib(k)].

By extending the domain of definition for y from IR+ into IR by y( - k) == y(k),
the usual formula S~oo y(k)e ikx dk results. In this model Kl = K2 = in
(7.2.49). To obtain all nonzero components we need a mixture of sixteen

elemental meson fields satisfying
for to = r;, and T" + 1,
(8
"1"2"3 )
112 (
Y ",t--
) = {JF(l!, T) exp [ig(l!, T)]
.
otherWIse.
(7.2.50)
7.3. Spin-i Fields in the Extended Phase Space

The relevant matrix algebra for a spin-t in the extended phase space is the
Clifford algebra of eight generators r A satisfying
(7.3.1)
7.3. Spin-! Fields in the Extended Phase Space 191
where I stands for the unit element. A convenient basis set for this 256-
dimensional algebra is given by
(7.3.2)
Here nA E {O, l}. The only irreducible representation of this algebra is sixteen
dimensional. A possible 16 x 16 irreducible representation of the generators
is furnished by
r 1 = (1'1 X I x I x I = (r 1 )t, r 2 = (1'3 X (1'1 X I x I = (r2)t,
r 3 = (1'3 X (1'3 X (1'1 X 1= (r3)t, r 4 = ;(1'3 X (1'3 X (1'3 X (1'1 = _(r4)t,
r S = (1'2 X I x I x 1= (rS)t, r 6 = (1'3 X (1'2 X I x I = (r6)t,
r 7 = (1'3 X (1'3 X (1'2 X I = (r7)t, r 8 = ;(1'3 X (1'3 X (1'3 X (1'2 = _(rS)t.

(7.3.3)
Here, (J' are the Pauli matrices, I is the 2 x 2 unit matrix, and the x denotes
the Kronecker product [equation (3.3.2)]. There exists a nonsingular hermi-
tian matrix n such that
(7.3.4)
The Lagrangian of a spin-! field t/I(~) is given by
!f(. ')1" = H~(~)rAt/I.A - ~.ArAt/I(~)] + m~(~)I"(O,
(7.3.5)
~(~) == [t/I(~)]tn.
Here, t/I(~) is a 16 x 1 column vector field. The corresponding Euler-
Lagrange equations yield
rAt/I.A + mt/l(~) = 0,
(7.3.6)
~.ArA - m~(~) = ot,
where the 0 is the 16 x 1 zero column vector.
The various tensor fields obeying differential conservation laws can be
constructed from the Lagrangian in (7.3.5). These are given by
TAB(~) = (1/2)~.BrAt/I(~) + (h.c.),
jCAB(~) = (1/2)(~.Arct/I- ~rct/I.A)~B - (1/2)(~.BrcI/I - ~rcI/I.B)~A
+ (1/8)[~rC(rArB - rBrA)I/I + (h.c.)],
(7.3.7)
j<A)(~) = T(A)(,) + (1/2)B(A)(,)
= (1/2)[(qi~.pj - pj~.qj)rAI/I _ ~rA(qil/l.pj _ pil/l.gj)]
- (1/16) [~rAdij(r)j+4 - 1j+4ri - r i+41j + Ijr~+4)1/1 + (h.c.)],

v A(,) = _ i~(,)rAI/I(').
Here, (h.c.) stands for the hermitian conjugation of the precc:ding terms.
Some of the constants of motion, which can be constructed out of cor-

responding densities in (7.3.7), are furnished as:
Ka = (1/2) f [~,q,r4t/1 (h.c')]lq4~O

+ d 3q d 4p,
f -
R7
H= -K4=(1/2) [t/lr4 t/I,q4+(h,c,)1.d 3 qd 4 p,

R7
Xa = (1/2) - t/I + (h.c)l.d qd p,

f [t/I,par 4 3 4
R7
(7.3.8)
Q = eo[T + (1/2)B]
= (1/2) f {qj~,pj
R7
- pj~,qj)r4t/1
- (1/8) [t/I- r 4 d ab (rarb+4 - r b+4r a - r a +4r b + r br a+4 )t/I]
+ (h,c)k d3 qd 4 p,
N =- i f (~r4t/1)l .. d q d p,
R7
3 4
where eo is the charge parameter.

We can write the first-order matrix differential equation (7.3.6) in the form
V"t/I,qa + abt/l,ph + mt/l(q,p) = 0,
(7.3.9)
v"v b + vbv a = 2d ab ] = aaa b + aba",
vaa b + abv" = O.
The v" matrices have the significance of the four-velocity components and a b
matrices have the meaning of the four-acceleration components. The equa-
tion (7.3.9) has an analogy to the Boltzmann's transport equation for the
statistical distribution function in the extended phase space. We shall call this
equation the Boltzmann-Dirac-Yukawa or, in short, B.D.. equation.
Now we shall derive the unified spin-t fields out of the B.D.. equation
(7.3.9). For that purpose we introduce polar coordinates [also see equation
(7.2.17)]
qb = pb cos (}b,
where b is not summed. We recall that

a b a b -1' b a
aqb = cos () apb - (p) sm () a()b'
a . b a b -1 b a
apb = sm () apb + (p) cos () a(}b'
t/J(p, () == t/I(q, pl
7.3. Spin-t Fields in the Extended Phase Space 193
Therefore, the B.D.Y. equation goes over into

b+ [
e -i6 (S, .( Pb)-I,i; )]
b
IX 'I',pb - I 'I',6b
+ IX b - [e i6b (tfr,pb + i(pb)-ltfr,6b}] + mtfr(p, () = 0, (7.3.10a)
We can more explicitly write 16 x 16 IX b matrices in the following way:

1
1
1
1 1-
IX =
1
1
1
1
1
1
1
1
, 1X 2 - --
-1
-1
-1
-1
1 1
-1
3- -1
,IX =
-1
-1
1
1
-(u,-iud
uZ-iu1
-(0-2 +io-d
-(0-2+ io-,)
-(0-2 + io-d
0-2 + io-,
-(0-2 +io-,)
(7.3.1Ob)
where I stands for the 2 x 2 identity matrix and blank spaces contain zeros
only.
We shall seek basic, separable solutions of the polar B.D.Y. equation
(7.3.10a) by putting the spin-t field components as
~L(p, () = XL(p) exp(itt()b),

(7.3.11)
LE{1,2, ... ,16}.
In the sequel, L, M, NE{1,2, ... ,16}, and summation convention will be
carried out between a pair of dummy superscript and subscript. According to
our convention, the index L in (7.3.11) is not summed.
Substituting (7.3.11) into (7.3.10) we can derive the following system of
linear partial differential equations:
[X:p' + (tilpl)x9]exp[i(t~()b - ()l)] + [/p2 + (tVp2)x5]exp[i(tg()b _ ()2)]
+ [X~p3 + (tVp3)X3] exp[i(t~()b _ ()3)]
+ i[X~p4 + (d/p4)X2] exp[i(t;()b - ()4)] + mxl exp[itt()b] = 0,
[X,lp~ + (t~O/pl)xlO]exp[i(ttO()b - ()l)] + [X~p2 + (tVp2)l]exp[i(tt()b _ ()2)]
+ [X~p3 + (tjjp3)x 4]exp[i(tt()b - ()3)] + i[X,\,4 - (ti/p4)l]

x exp[i(tt()b + ()4)] + mx 2 exp[it;()b] = 0,
[X,lp\ + (t~l/pl)xll]exp[i(ttl()b - ()l)] + [X> + (ti/p2)x 7]exp[i(tZ()b _ ()2)]

+ [X~p3 - (ti/ p3)X l ] exp[i(tl()b + ()3)]
- j [lp4 + (t!/ p4)x4] exp[i(tt()b - ()4)] + mx 3 exp[jt~()b] = 0,
[X,lp2, + (t~2Ipl)x12]exp[i(tt2()b _ ()t)] + [X~p2 + (tVp2)x8]exp[i(t:()b _ ()2)]

+ [X;p3 - (tV p3)x2] exp[i(t;()b + ()3)]
- j[X~p4 - (tV p4)X3] exp[i(t:()b + ()4)] + mx4 exp[it:()b] = 0,
[X,l;, + (t~3Ipl)xl3]exp[i(tt3()b - ()l)] + [X~p2 _ (t~/p2)xl]exp[i(tt()b + ()2)]
- [X:p3 + (tVp3)x7]exp[i(tZ()b - ()3)]
- i[X~p4 + (tV p4)X6] exp[i(t~()b - ()4)] + mx5 exp[itg()b] = 0,
7.3. Spin-t Fields in the Extended Phase Space 195
[X~p~ + (t~4Ip1 )X14] exp[i(tt 4o b - 0 1)] + [X~p2 - (dl p2)x2] t:xp[i(tlO b + 02)]
- [X~p3 + (tV p3)x8] exp[i(tgO b - 03)]
- i[X:p4 - (tV p4)x5] exp[i{t~Ob + 04 )] + mx6 exp[it~t,b] = 0,
[x~i' + (tPlp1)X 15 ] exp[i(tt 50 b - 0 1)] + [X~p2 - (tVp2)X3] exp[i(t~Ob + 0 2)]
- [X:p3 - (tV p3)X 5] exp[i(t~Ob + 03)]
+ i[X~p4 + (tllp4)x 8]exp[i(tgob - 04 )] + mx7 exp[itZO b] = 0,
[x~p61 + (t~6Ip1)X16] exp[i(tt 60 b - 0 1)] + [X:p2 - (tilp2)x 4 ]exp[i(ttO b + in]
- [X~p3 - (tV p3)X 6 ] exp[i(t~Ob + 0 3)]

+ i[X:p4 - (tIl p4)X 7] exp [i{tZ Ob + 04)] + mx8 exp[it:O b] = 0,
[X~pl - (tUp3)x1]exp[i(ttO b + 0 1)] - [X,\;2 + (t~3Ip2)x13]exp[i(tt30b - 04 )]
- [X~:3 + (tj 1I p3)xll] exp[i(tt lOb - 0 3)]

- i[X,lp04 + (t!Olp4)x10]exp[i(ttOob - 04 )] + mxgexp[it~Ob] = 0,
[X~pl - (ti/ p1 )x2] exp[i(tlO b + 0 1)] - [X~p42 + (t~4Ip2)X14] exp[i(tt 4ob - 0 2)]
- [X~;3 + (t;2/p3)x12]exp[i(tt 2ob - 03)]

- i[X:p4 - (t~/p4)x9]exp[i(t~Ob + 04 )] + mx 1exp[itiOO b] = 0,
[X~pl - (tVp1)x3]exp[i(t~Ob + ( 1)] - [x,li2 + (t~5/p2)xI5]exp[i(tt50b - ( 2)]
- [X:p3 - (t~!p3)x9] exp[i(t~Ob + 03)]

+ i[X,1;4 + (t!2/p4)xI2]exp[i(tt 2ob - ( 4 )] + mxllexp[itt 10b] = 0,
[X:pl - (ti/p1)x 4]exp[i(ttO b + ( 1)] - [X~p62 + (t~6Ip2)x16]exp[i(tt60b - ( 2)]
- [X,1~3 - (tjO/p3)X 10 ] exp[i(ttOO b + 03)]
+ i[X,1;4 - (t!1/ p4)xll]exp[i(tpO b + ( 4 )] + mx I2 exp[itt 20 b] = 0,
[X:pl - (ti!p1)x5]exp[i(tgO b + ( 1)] - [X:p2 - (t~/p2)x9]exp[i(t~Ob + ( 2)]
+ [x~i3 + (tj5/p3)X 1S ] exp[i(ttSO b - ( 3)]
+ j[X,lp4:. + (t!4/p4)x14]exp[i(tt 4ob - ( 4)] + mx13exp[itt30b] = 0,
[X~pl - (tV pI )x6] exp[i(t~Ob + ( 1)] - [X~p02 - (dO I p2)X10] exp[i(ttOO b + ( 2)]
+ [X,I;3 + (tj6/ p3)X 16 ] exp[i(tt 6o b _ ( 3)]
+ i[X~;4 - (tp/p4)x13] exp[i(tt 30 b + ( 4 )] + mx 14 exp[itt 4 0b] = 0,
[X:pl - (tI/pl)x7]exp[i(tlO b + 0 1)] - [X~pI2 - (t~l/p2)x11Jexp[i{ttlob + 02)]
+ [X,I;3 - (tj3/p3)x13Jexp[i(tt30b + ( 3)]

- j[X,lp64 + (t!6/p4)xI6]exp[i(tt60b - ( 4 )] + mx 15 exp[itt SOb] = 0,
[X~pl - (tVp l )x 8 ]exp[i(tgO b + ( 1 )] - [X.I;> - (ti 2jp2)x12]exp[i(tt 20 b + ( 2)]

+ [x~p4, - (t~4/p3)x14] exp[i(tt 4ob + ( 3)]
- i[X,I}4 - (d 5 j p4)x15] exp[i(tt 50b + ( 4 )] + mx16 exp[itt 6 0b] = o.
(7.3.12)
Let us analyze the first of these equations. It can be written in the form
G1(p;02,03,04)exp[i(ti - 1 - t~)OI]
+ G2(p; 02, 03, (4)exp[i(ti - d)Ol] + G3(P;02,03,(J4)exp[i(t~ - t~)OI]
+ G4(p; 02, 03, ( 4) exp[i(ti - tDOI]

+ GS (p;02, 03 , ( 4 ) = o. (7.3.13)
Here it is assumed that at least one of the Ga ( ) =I: 0 (say, G1 ( .. ) =I: 0). Other-
wise, the equation becomes an identity. We differentiate the left-hand side of
(7.3.13) with respect to 0 1 and divide by i exp[i(ti - d)Ol]. Then again we
differentiate with respect to 0 1 and divide by i exp[i(t~ - ti)Ol]. Next we
differentiate and divide out i exp[i(ti - t~)OI]. The resulting equation yields
(since G1 ( .. ) =I: 0)
(ti - 1 - tD(ti - 1 - tiHti - 1 - ti)(ti - 1 - i) = O. (7.3.14)
Similarly, many more equations can be derived from the sixteen equations of
(7.3.12). Invoking separation of variables, the solution of these sixteen equa-
tions is
d = ti = ti = t1 = ti = t~ = t I= t~ = ti - 1
= dO - 1 = d1 - 1= tF - 1 = tP - 1 = d4 - 1
= tP - 1 = d6 - 1. (7.3. 14a)
Similarly, from differentiations of 0 2 , 0 3 , 04 the following relations emerge:
~=~=~=~=~-1=~-1=~-1
= t~ - 1 = t~ = ~O = ~ 1 = ti 2 = tP - 1 = ti 4 - 1
= ti 5 - 1 = ti 6 - 1,
d= t~ = t~ - 1 = tj - 1 = t~ = t~ = t~ - 1
= t~ - 1 = t~ = t~O = t~1 - 1= t~2 - 1= tP = t~4
= t~5 - 1= t~6 - 1,
tl = t~ - 1 = tI = t! - 1 = tl = t~ - 1 = t1
= t: - 1 = t~ = tlO - 1 = tl l = t1 2 - 1 = t1 3 = d 4 - 1
(7.3.14b)
The above sixty linear equations hold among sixty-four integers tf;. It is clear
that four of tf;'s can be chosen arbitrarily and that the remaining sixty can be
obtained through (7.3. 14a,b). The following tf; are chosen freely:
t1 = tL t2 = d, t3 =d- 1, t4 = t1 - 1. (7.3.15)
With these choices of tf;, the functions of OJ's are all cancelled and the polar
B.D.Y. equation (7.3.10a) or (7.3.12) goes over to
(cxb+)nX~pb + (tf;/pb)x L] + (cxb-)nX~b - (tf;/pb)x L]
+ mxN(p) = O. (7.3.16)
We want to solve the above system of equations. We notice that, if we
operate from the left by rA%~A - mI the first of equations (7.3.6), then we
obtain the decoupled second-order system DABt/I.AB - m 2t/1() = O. Similarly,
the system (7.3.16) yields (under the assumption that the XL'S are twice differ-
entiable) the decoupled system:
M2XL(p) = dab[X~.pb + (pa)-lX~b]
= [m 2 + (dabt~tf;/papb)]XL(p). (7.3.17)
(Here L is not summed.) This equation indicates an Okubo-like mass split-
ting [compare the equation (7.2.22)]. From the knowledge of Sec. 2, we try to
solve (7.3.16) in terms of Bessel functions of the first kind. Let us assume for
separable solutions
(7.3.18)
Here j is not summed, tf; are given by (7.3. 14a,b), and UL(K) are undetermined
amplitude functions. We substitute (7.3.18) into (7.3.16). The first equation
(for N = 1) yields
[Jr'I' pI + (ti!p1)Jr.]Jr.Jr.Jr.u9 + [Jr, p2 + (ti! p2)Jr,].",Jr,)t,U 5
1 2 3 <4 2' 2 1 3 4
Using the rules [op + (t/p)]Jr(Kp) = +KJ,l(Kp) of the equations (7.2.41-ii,iii)

and the equations (7.3.14a,b), we observe that Bessel functions in (7.3.19) can
all be cancelled out. Therefore, (7.3.19) yields the algebraic e:quation
(7.3.20)
Similar simplifications occur in other fifteen equations of (7.3.16). The result-
ing algebraic system of equations can be neatly summarized into
(7.3.21)
Here the 16 x 16 matrices a b have been defined in the equation (7.3.9) and
U(K) is the unknown 16 x 1 column vector with components UL(K). Multi-

plying the above equation from the left by [ - iKcac + mI] we obtain
[dbCKbKc + m 2 ]u(K) = O.
For a nontrivial solution we must have the necessary condition:
dbcKbKc + m 2 = 0,
K4 = E(!S), (7.3.22)
E(!S) == J K~Ka + m 2
The present representation of the a b matrices in (7.3.9) is too complicated to

solve (7.3.21) neatly. We notice, however, that abac + acab = 2dbJ, which im-
plies that the ab-matrices are reducible to four direct sums of Dirac matrices
by a similarity transformation. Such a similarity matrix, which is unitary as
well, is explicitly furnished:
[+(12 1_(12 (13+i(11 tT 3 _ia l i(1-a') i(/+o') _(0'1+i0'3) (11_i(13
_(O'l+I( 1) o"~ -ja l I+a 2 -(I-a') (11+i(13 0'1_;a 3 i(l-a') -i(I+o')
(]3_jol 113+ia1 _(1 __ 0') -(l+a') 0'1_i 11 3 _(0'1+i(13) -i(1+a') -i(l-a')
_(I_a') 1+(12 _(0-3_;a 1) (J3+iu 1 -i(l +a') i(l-a') _(a'_j(73) _(a1+ia 1 )

4S '"
-i(l+a') i(1-a') _(a 1 _iu 3 ) _(q'+iu 1 ) -(I-a') 1+(12 _(a 3 _ja 1 ) {JJ+ier l
(Tl_ia 3 _((11+IU]) -i(l+a') -i(l-a') (TJ_ia l (]3+iu 1 -(I-a') -(l+a')
_(a 1 +iu J ) -(al-ja J ) -i(l-o') i(l+a') 0- 3 + io- _(0"3_;(11) -(I+a') l_a 2
-i(l-a') -I(I+a') (Tl+ia 3 _{(Tl_j q J) -([+a') -(I-a') _13+ 10'1) _(a 3 _iO'I)
(7.3.23)
We can verify by tedious computations (which involve block by block matrix
multiplication) that
sts = I,
(7.3.24)
Here the same symbol I has been used for unit matrices of different sizes. The
pb matrices are representations of Dirac matrices that are different from those
in either (6.5.4) or (6.5.6). Multiplying the equation (7.3.21) by S from the left
and using (7.3.24), we can derive
[I x (iKbpb + mI)]u = 0,
(7.3.25)
u = Su, u = stu.
If we put
[1]T == (1, 1, 1, 1), (7.3.26)
where U v is a 4 x 1 column vector, then (7.3.25) reduces to
I x [(iKbPb + mI)u v ] = 0,
(7.3.27)
[iKbyb + mIJuv = O.
The last equation is equivalent to (6.5.8), the Dirac equation for the plane
wave amplitude. Four basic solutions of (7.3.27), which are analogous to
those given in (6.5.12a,b), are listed in the following:
v[elJ T = (I,O,(E + m)-lK 3 ,(E + m)-l(K l + jK l ,
v[e 2 J T = (0, 1,(E + m)-l(Kl - jK l ), -(E + m)-lK ),
3
v[e~JT = (-(E + mf K3' -(E + mfl(K l + jKl)' 1,0),

l (7.3.28)
v[e;J T = (-(E + m)-l(Kl - iK2)' (E + m)-lK3'0, 1),
v- l == J(E + m)/2m.
Here T denotes the transposition. We can construct sixteen basic solutions of

(7.3.25) by using (7.3.28). These are furnished by
aill) = [eL 0, 0, OJ, ai12) = [e~,O,O,OJ,
ai21) = [0, eL 0, 0], ar22) = [O,e~,O,OJ,
ai3l) = [O,O,eLO], arll) = [O,O,e~,OJ,
ai41) = [0,0,0, en. 12r42) = [0,0,0, eD,

(7.3.29)
(Jill) = [e?,O,O,O], (Jr12) = [e;T, 0,0,0],
(Jill) = [O,e?,O,O], (Jil2) = [0, e;T, 0, OJ,
(Jbl) = [O,O,e~T,O], (Ji32) = [0,0, e;T, 0],
ei41) = [0,0,0, e?J, (Ji42) = [0,0,0, e?J.

Here, each entry represents an 1 x 4 row matrix. By the linear independence
of el , el , e~, e;, the linear independence of a(..~), V(bfJ) (where IX, f3 E {I, 2})
follows. The sixteen (linearly independent) basic solutions of (7.3.21) will be
given by [using equations (7.3.29), (7.3.25), (7.3.24)]
u(a~) == sta(aa), V(bfJ) == St{J(bfJ). (7.3.30)
The orthogonality and normalization conditions among basic solutions
can be computed directly, and they are listed as follows:
a~a)(~)a(bfJ)(~) = llla~)(~)V(bfJ)(~) = u~a)(~)U(bfJ)(!!:)
= v~a)(!!:)V(bfJ)(!!:) = (E/m)O(ab)O(~fJ)'
(7.3.31)
v~~)(~)a(bfJ)(~) = ula~)(~)V(bfJ)(~) = vlaa)(~)UbfJ(~)
= ula~)(~)V(bfJ)(~) == o.
Every solution of (7.3.21) can be obtained as a linear combination (or
superposition) of basic solutions in (7.3.30). Therefore,
4 2
uL(~) =L L [f3(ra)(~)U~a)(~) + y<r")~)V~..)(~)J (7.3.32)
r=l a=l
for some complex coefficients p(ra)(Js), y(ra)(~). A general class of solutions of

the polar B.D.Y. equation (7.3.10) is given by the Fourier-Bessel integral [the
equations (7.3.11), (7.3. 14a,b), (7.3.18), (7.3.22), (7.3.32) are assumed]:
~L(p, (J) = (2nf3/2 vim
x rt at1 I=~OO {L~ . [p(ra)(~, t)u~a)(~)

+ rra)(~, t)v~a)(~)]
x [Jj .lrj(Kjpj)JeXp[i(tt(Jb)]K1K2K3d3~}.
(7.3.33)
Here, tt satisfy (7.3.14a,b) and K4 == E(~). Furthermore, we assume that the
above improper integral converges uniformly and partial differentiations with
respect to pa, (Jb are permitted under the integral signs. Otherwise, ~ra), y(ra)
are arbitrary complex-valued functions. The spin-! field in (7.3.33) can also be
expressed as
L
00
~L(p,(J) = ~L(p,(J;t1,t2,t3,t4)'
t=-oo
~L(p, (J, t) = (2n)-3/2 vim

~ f, {(OO () (7.3.34)
rfl /='1 J0+ .. [P rlZ (~, t)U(ra)(~) + y ra (~, t)V(ra)(~)]
L ;;( ) L
x
x [Jj Jlj(KjPj)}XP[i(tt(Jb)]K1K2K3d3~}.
This equation resembles the unified meson field equation (7.2.35). The con-
stants of motion in (7.3.8) have been calculated in our paper (referred after
this chapter,) using a single t-mode I/IL(p, (J; t 1, t 2, t 3, t 4 ). But those results may
not be physically reliable, since the complete set of separable solutions in
(7.3.33) has not been used. Such a calculation with the complete set will be
extremely difficult.
The functions p(ra)(~, t) and y(ra)(~, t) in (7.3.34) represent the particle and
antiparticle amplitudes. The index Q( {I, 2} represents the usual two degrees
of freedom for a spin-! wave field. The index r E {I, 2,3, 4} stands for four
families of spin-! wave fields. Furthermore, each family can have fourfold
infinitely many flavorlike quantum numbers t 1 , t 2 , t 3 , t4 {I, 2, 3, ... }. Thus,
the equation (7.3.34) stands for the unified spin-! fields on the extended phase
space Mg.
References
1. M. Born, Proc. Roy. Soc. London Ser. A 165 (1938), 291; 116 (1938), 552; Rev.
Mod. Phys. 21 (1949),463.
2. J. A. Brooke and E. Prugovecki, Nuovo Cimento A 78 (1984),17.
7.3. Spino! Fields in the Extended Phase Space 201
3. H. Brandt, Nuc!. Phys. B 6 (Proc. Supp!.) (1989),367.

4. E. R. Caianiello, Lett. Nuovo Cim. 25 (1979), 225.
5. E. R. Caianiello, M. Gasperini, and G. Scarpetta, Nuovo Cim. B 105 (1990), 259.
6. A. Das, J. Math. Phys. 7 (1966), 45; 21 (1980), 1506, 1513, 1521; Nuc!. Phys. B 6
(Proc. Suppl.) (1989); Z. Phys. C. 41 (1988), 505.
7. M. Gell-Mann, Cal-Tech. Lab. Report CTSL-20, 1961.
8. Y. S. Kim and E. P. Wigner, Phys. Rev. A 36 (1987), 1293.
9. S. Okubo, Prog. Theor. Phys. (Kyoto) 27 (1962), 949.
10. G. N. Watson, A treatise on the theory of Bessel functions, 2nd ed., Cambridge
Univ. Press, London and New York, 1966, pp. 248.
11. E. P. Wigner, Ann. of Math. 40 (1939), 149; in Proceedings of the First Interna-
tional Conference on the Physics of Phase Space (Y. S. Kim and W. W. Zachary,
eds.), Springer, Verlag, Heielberg, 1987.
12. H. Yukawa, Phys. Rev. 76 (1949),300; 77 (1950),899; 80 (1950), 1047.
Answers and Hints to
Selected Exercises
EXERCISES 1.1
1 (ii). <T(a) = j2, <T(b) = <T(e) = 0, <T(d) = 1.
EXERCISES 1.3
2 (i). Hint: Note that iij == (lj2)(iij + iji) + (lj2)(iij - i j ;).
EXERCISES 2.1
2. .?li(t) = sin(t - c i).
EXERCISES 2.2
1. S(y) = 2.
EXERCISES 2.4
EXERCISES 3.2
2. Hint: Note that the 4 x 4 identity matrix does not belong to this subset.
EXERCISES 3.4
3(') H" a a a a a a
I. Illt: ax = au + au' at = au - au'
EXERCISES 4.1
1. Hint: Assuming xa = _x a prove that UPU t + p = 0, or U<Ta + <TaU = O.
(The last equation leads to a contradiction.)
202
Answers and Hints to Selected Exercises 203
EXERCISES 4.2
l(i). Hint: Compare the problem 2(ii) of Exercises 1.3.
EXERCISES 4.3
EXERCISES 5.1
1. H(8,Pe) = (pN2mt2) + mgt(l - cos 8).
EXERCISES 5.3
EXERCISES 5.4
. oA(x, u) OXk(X) oA(x, a) b0 2 Xk(X) 0 A(x, U)
2. Hmt: 8xi = ~ 8Xk +u oxboXi oak '
8A(x, u) 8X k (x) 8A(x, a)

ou i - -Ox i 8a k
EXERCISES 6.1
EXERCISES 6.2
. 81jJ 8h 01jJ [. oh Oh]

4. Hmt: 8x1 = (cosO() ow' 8x2 = (smO() 8w + o~ ,
81jJ .[8h . Oh] 81jJ 8h
8x3 = I aw + (sm O() o~ , ox4 = (cos ct) 8~'
EXERCISES 6.3
2. Hint: Use special Minkowski coordinates such that
F34(X) = F12 (x) == O.
Index of Symbols
d(F) action functional of a relativistic field

Ak(X) electromagnetic field potential
a . b inner product between two vectors
A x B Cartesian product of two sets
M x N Kronecker product of two matrices
BAi(X) gauge potential components
riZ(D; IR); ri 2 (D; S) the set of twice differentiable functions
from D into IR, S respectively
C A BD structure constants of a Lie algebra
x(p) = x = (xl, X2, X3, x 4 ) local coordinates of a point p in the manifold
(X, U) chart in a differentiable manifold
D = [diJ == diag[l, 1, 1, -1] Lorentz metric; a domain D c [Rn
aD boundary of a domain D
D d'Alembertian or wave operator,
also denotes the completion of an example
Q.E.D.
Dat/lK(x) gauge covariant derivative
t5(k - k') Dirac delta function
t5 a b Kronecker delta
Ad) perihelion shift per revolution
8(x l , x2 , x3 , x 4 )/8(x 1 , x 2 , x 3 , x 4 ) Jacobian of a coordinate transformation
1E3 three-dimensional Euclidean space
{ei}i == {e 1,e Z,e3,e 4 } a basis set or tetrad
e i ' ej = dij an M-orthonormal tetrad
E(p) frequency associated with Dirac plane wave
E belongs to
Gijkl totally anti symmetric permutation symbol (Levi-Civita)
'1ijkl totally antisymmetric pseudotensor (Levi-Civita)
Fi(x, u) four-force components
Fij electromagnetic tensor components (field)
Fi1 Hodge dual of electromagnetic tensor components
FAij(X) gauge field components
204
Index of Symbols 205
t/JA spinor (field) components

G a group
glj metric tensor components
gij(X, u) Finsler metric components
1': Xi = ,2"i(t) a parameterized curve on the manifold
yk Dirac matrices
yku v entries of Dirac matrices, also numerical bispinor-tensor components
Yab(X) linearized gravitational ten-potential components
H(t, x, p) (prerelativistic) Hamiltonian function
1= [c5 iJ the identity matrix
J [I'] prerelativistic action functional
Jh[f.] prerelativistic canonical action functional
JL[y] relativistic action functional
JH[y] relativistic canonical action functional
,1iab (X) relativistic angular momentum tensor density
lex) charge-current four-vector density
Jab the total relativistic angular momentum tensor
L, L == [[iJ a linear (or Lorentz) mapping and the corresponding matrix
L T the transposed matrix
L(t, x, v) prerelativistic Lagrangian
L'(t, x, t', x') prerelativisitc Lagrangian in space and time
~(X,yA,yAi) the Lagrangian ofa field
A. a scalar (a real or complex number)
A(x, u) relativistic Lagrangian of a particle
~4 general Lorentz group
~4++ proper, orthochronous subgroup of ~4
M4 (flat) Minkowski space-time manifold
.;v"o Null cone with vertex at Xo
v(k) frequency associated with electromagnetic plane wave
w(k) frequency associated with Klein-Gordon plane wave
!lex, p) = 0 constraining mass hypersurface, Super-Hamiltonian
P space reflection mapping
Pa the total four-momentum vector
Pk four-momentum components
9i'4 Poincare group
f2 the total charge
IR the set ofreal numbers
IR" == IR x ... x IR (n factors) Cartesian product of the set IR
S(y) separation functional on the curve I'
SU(N) the set of unitary unimodular matrices
u(v) separation number of the vector v
u~ Pauli matrices
U kAB numerical spinor-tensor
"'(x) Klein-Gordon field
"'", "'"(x) bispinor, Dirac field components
206 Index of Symbols
1: 1 ,1: 2 ,1: 3 I-flat, 2-flat, 3-flat submanifolds

T time reversal mapping
Tij(x) canonical energy-momentum-stress tensor density
rJ; :::~: Minkowski tensor components
rj;:::~:(x) Minkowski tensor field components
BiAx) symmetrized energy-momentum-stress tensor density
u, U i covariant vector (or covector) and components
ui(s) four-velocity components
Vex) the potential energy
V4 Minkowski vector space
v, Vi Minkowski vector, components
z == Xl + ix 2 a complex variable
z == Xl - ix 2 the complex conjugate variable
7L the set of integers
7L+ the set of positive integers
Subject Index
abelian group (see group) bilinear mappings 14

acceleration four-vector 103 bispinors (Dirac) 85,86,152,153
action functional (or integral) 90, 93, bispinor-tensor 86
97,109,111 numerical bispinor-tensor 86
pre-relativistic mechanics 90 bispinor field 152, 153
canonical, prereIativistic mechanics Dirac equation 153
93 (see also spinors and Dirac)
canonical, prerelativistic mechanics in boost mapping 12,51
space and time 97 transformation 51
relativistic mechanics 109 (see also Lorentz)
canonical, relativistic mechanics 111
relativistic fields 120, 171
algebra, Lie (see Lie algebra) canonical equations
angle 3, 12, 64, 76 (see Hamilton)
Eulerian 72 Cartesian coordinates 25, 48, 89
pseudoangle 64 characteristic coordinates 70
angular momentum 99, 130, 132 equation 2
relativistic angular momentum tensor (see also null coordinates)
density 130 chart 20,21
spin and orbital tensor density 130 (see also differentiable manifold)
total relativistic angular momentum charge (electric) 106, 135
tensor 132 charge-current density 130
arclength 27 Dirac field 158
(see also separation) Klein-Gordon field 135, 175
area 80, 117 charged particle (see particle)
areal velocity 117 total charge 132
atlas (see differentiable manifold) chronometry, Minkowski 33, 34
axioms 1, 8, 27, 56 clock, apparent retardation 53
group 56 standard 28
linear mapping 8 commutation relations 147, 148
Minkowski vector space 1,2 complex numbers 72, 73, 123,
nondegeneracy 2,27 147
relativistic equations of motion 103 conformal mappings 73
conjugation 72
holomorphic functions 74,147
basis set of Minkowski vectors (or tetrad) matrices 75, 77, 78
2,3,8 vector spaces 79, 85
spinor vectors (dyad) 79 wave fields (see wave fields)
207
208 Subject Index
conservation equations 130, 131, 132, Hodge dual tensor 18, 19, 80, 142,
176 146, 147
differential, angular momentum 129 dynamics (see mechanics)
differential, charge-current 130
differential, energy-momentum-stress
128 eigenvalues 2, 108, 147
total, angular momentum 132, 176 metric tensor 2
total, charge 132, 176 energy-momentum-stress tensor 147
total, four-momentum 132, 176 Einstein ix, 71,113, 118
(see also Noether) general relativity 118
contraction, moving rod (see Lorentz- linearized gravity 113
Fitzgerald) philosophy ix
tensors (see tensors) summation convention 2, 3, 79, 85,
contravariant vectors (see vectors) 147, 163, 169
coordinates 20,22,25, 115, 123 electric 106, 108, 112
Cartesian 25 charge (see charge, electric)
charts (general) 20,21 field (see electromagnetic field)
complex conjugate 72,123 electromagnetic field 18, 140, 141
ignorable 94,116 complex version 142, 146
Minkowski 25, 26, 33, 34 energy-momentum-stress tensor 19,
transformation (see transformation) 145, 146, 147
constraints (or constraining hyper- electric field 107, 113, 141
surface) 93, 97, 111 four-potential 113, 142, 143, 161, 162
(see also Lagrangian) magnetic field 107, 108, 113, 141
curve, parameterized 22, 27, 28, 29, 30, Maxwell's equations 141
33,34 (see also Lorentz equations of motion,
arclength 27 photon)
separation 27, 28, 29, 30, 33, 34 electroweak theory 166
(see also motion curve, equations of elementary particles (see relativistic
motion) fields)
energy 89,92,96,105,132,174
hypersurface (or shell) 92,94, 100
diagonal 2,61,63 relative 105
Lorentz metric 2 rest 105
block diagonal matrix 61, 63 total 92, 132, 174
d'Alembertian operator 47, 123, 124, energy-momentum-stress tensor 128,
132,133 171,175
differentiable manifold 20,21,22,25, gauge field 150
115 Klein-Gordon field 135, 175
atlas 21 electromagnetic field 145, 146, 147
chart (or local coordinate system) 20, Dirac (bispinor) field 158
21,22,25, 115 equations of motion 89, 103, 106
Hausdorff topology 20 Newtonian (or prerelativistic) 89
projection mappings 20, 22 relativistic 103
subatlas 21 Lorentz 106
Dirac 85, 137, 151, 192 (see also Euler-Lagrange equations)
delta function 137 Euclidean space (or geometry) 48, 72,
equation, wave or field 151,153,160, 95
192 plane 7, 67, 73
matrices 86, 153, 154, 198 orthogonal 7
bispinors (see bispinors) Euler 63, 64, 72
dual 13,18,19,80,142,146,147 angles 63, 64, 72
vector (covariant vector) 13, theorem on homogeneous functions
80 96,109
Subject Index 209
Euler-Lagrange equations 90,93,95, (see also differentiable manifold)

97, 111, 123, 171 gravitation 98, 113
events 25, 26 Newtonian constant 98, 115
future 27 linearized theory 113, 118
past 27 linearized Schwarzschild field 115
present 27 planetary orbits 98, 115
separation 27, 28 Greek indices 4
simultaneous 52 Green's theorem (see Gauss's theorem)
group 55,56,63,66,147,163
Lie group 63, 66, 70
field 1, 120, 168, 191 Lorentz group (see Lorentz)
bispinor (or Dirac field) 153,160,191 Poincare group 42, 56, 170
complex numbers 72, 73, 123 manifold 66, 147
gauge field 147, 163 representation 59, 61, 149, 163
real numbers I SU (N) 163
real scalar field 43, 47, 123 subgroup 56, 57, 58, 59
tensor field 43,47, 120, 142, 170 symmetric 56
(see also relativistic fields) unitary, unimodular 74, 77, 78, 148,
Finsler metric 108, 109, 110 151,164,166
flat 20, 35, 41 (see also unitary, unimodular
space-time (Minkowski) 20 conditions)
submanifolds 35
I-flat 35
2-flat 35,36,37,38,39,41 Hamiltonian mechanics 91, 92, 93, 97,
3-flat 35, 37, 41 110, 112, 119
force 48, 89, 103 prerelativistic canonical equations
four-force vector 103, 104, 106 92,98
monogenic 89 relativistic canonical equations 112
relative three-force vector 104 hermitian 74, 75,77,82, 153
three-force vector 89 conjugation 74, 153
frames of reference (see inertial frames) matrix 75, 77, 82
frequency (with wave) 135 history of a particle 26
functions (see mappings) Hodge dual 18, 19,80, 142
functionals 27,90,93,97, 109, 111, 120, pseudotensor 18, 19, 142, 146, 147
171 spinor 80
action (see action) homogeneous function 96, 97, 109, 110
separation 27 hypersurface (see constraints)
Galilean transformation 50 inertial observer (or frame) 26, 33,48,

gauge fields (nonabelian) 147, 149, 163 52,53
energy-momentum-stress tensor 150, infinitesimal transformation 126, 128,
151 129,130,171,172
Yasskin class 151 inner product (or dot product) 1,2,3,5,
gauge transformations 142, 143, 160, 7,8, 125,80, 148, 169
164 integral (Riemann) 45, 135, 138, 146,
covariant derivative 161,166 200
global 130 Cauchy principal value 137, 138
local 142, 164 four-dimensional 45
Lorentz gauge condition 143, 147 improper 135,137,146,156,177,
Gauss's theorem 46, 121, 131, 173 185,200
general relativity 118 conserved quantities (see
geometry 20, 25, 108 conservation)
Euclidean (see Euclidean) intersection of curves 30, 31, 35
210 Subject Index
invariant (or scalar) 14,43, 108, 124, Lorentz equations of motion 106
127,171 Lorentz metric 2
invariance of action 108, 124, 171 gauge condition 143,147
inverse mapping 9,21,100,117 group 55,58
invertible linear mapping 9 general 58
isomorphic vector spaces 23 manifold 66
proper 58
proper and orthochronous 58, 59,
Jacobi 21,48, 100, 148 63
identity 148 orthochronous 58
Jacobian 21,24,41,66, 125 representations 59, 61, 63, 84
Jacobi's principle matrix 10, 13, 42, 43, 49, 58, 65
mapping 8, 10
proper 10
Kepler problem (see planetary motion) improper 10
Klein-Gordonfield 133,175 orthochronous 10
Kronecker 9, 59, 60 transformation 43, 48, 49, 58
delta 9, 16, 17, 57, 74, 81, 125, 148, applications 51,52,53,55
171 boost 12,51
product of matrices 59,60,63 infinitesimal 128, 129
Lagrange multiplier 93,95,97, 111 magnetic field (see electromagnetic field)

Lagrangian 89,95, 108, 120, 171 mappings 8,13,14,20,21,22,27,
prerelativistic mechanics 89 59,90,93,97, 109, 111, 120,
prerelativistic mechanics in space and 127
time 95 bilinear 14, 80
relativistic mechanics 108 conformal 74
relativistic fields 120, 123, 124, 134, conjugate linear 84
144,149,157,160,163,171,175, holomorphic 74,147
191 invertible 8, 9, 21, 100, 117
latin indices (see Roman indices) linear 8
Legendre transformation 91 Mobius 74
length 3,4, 7, 52 parameterized curve 22, 23, 25, 26,
moving rod 52, 53 27,29,31,32,35,55,89,90,93,
Levi-Civita 17,18 97,109
permutation symbol 17, 18, 79, 80, projection 20,21,40,41, 73, 121
81, 8~ 10~ 10~ 108 Lorentz 8, 10
pseudotensor 18,19,74,81,87,142 manifold, differentiable (see differentiable
(see also permutation symbols) manifold)
light cone (see null cone) group 66, 147
coordinates 70 mass 89,98, 103, 104, 105, 108, 109,
Lie 63, 147, 148, 151, 163, 166 113, 115
group 63,66,67,70,147,151,163, proper (or rest) 102, 103, 105, 106
166 relative (or moving) 104, 105
semisimple 148, 163, 166 hypersurface (or shell) 111
algebra 147, 148, 163 matrix 8,9, 10, 11, 12, 13,42,43,49, 58,
representation 148, 149, 163, 166 65,72,74,77,86,149,154,164,
linear 8, 10, 13, 14,75,77,79 191,193,198
conjugate linear 84 Dirac matrices 86, 154, 198
independence 10, 14, 75, 77 hermitian 75, 77, 82
mappings 8, 13, 79 Lorentz matrices 10, 13,42,43,49,
Lorentz 2, 53, 55, 58, 106, 143, 147 58,65
Lorentz- Fitzgerald contraction 53 Pauli matrices 74, 77, 78, 149, 166
Subject Index 211
rank 35, 155 equations of motion 89

representation 8 mechanics 89, 92
group 59,61,63,84 Noether's theorem 126, 128, 130, 171
linear mappings 8 nonabelian group 147, 148, 163
Lie algebra 148, 163, 166 gauge fields 149,151, 160, 163, 165,
trace 75, 165 166
tracefree 62, 75 nondegeneracy 2,27
transposition 10,60,63, 119 axiom 2
similarity transformation 154 curve 27
unitary 74,75,76,78, 148, 149, 163, null 26,27,49
198 cone (or light-cone) 26,27,49
unimodular 72,77,79,80,81,83,84, coordinates 70
85,87,88 2-flat 37,38, 39,41
unitary and unimodular 74, 75, 76, vectors 4,6,26, 102, 144
78,148, 149, 163 world-line 26, 55
measurement 33, 34 numerical 17, 19, 79, 80, 86
Minkowski coordinates 33, 34 bispinor (see bispinor)
proper time 28, 29 pseudotensor (see tensors)
spacelike separation 30, 32 spinor (see spinors)
mechanics (see equations of motion, tensors (see tensors)
Lagrangian, and Hamiltonian)
metric tensor 2, 108, 169
Lorentz metric 2, 10, 16,25 observer 26,33,48
Finsler metric 108, 109, 11 0 inertial (see inertial)
Michelson-Moreleyexperiment 49 open subset 20,21, 121
Minkowski 1,2,4, 10, 15,25,26,30,31, operational method 33, 34
33, 34, 41 orbit (see planetary)
chronometry 33, 34 orientation 46, 131
coordinates 25, 26, 33, 34 orthogonality 2, 4, 6, 7
M-orthogonality 2,4,6,30,31,33 Euclidean 7
M -orthonormality 2,4, 10, 11,25, 34 Minkowski (see Minkowski)
tensors (see tensors) solution vectors 156, 159
tensor fields (see tensor fields) orthonormality, Minkowski (see
vector space 1, 23 Minkowski)
momentum 91,102,110,132,160,174 M -orthonormal basis (or tetrad) 2,3,
four-momentum 102, 110, Ill, 112, 4,9,10,25
113 outer product (see tensor product)
three-momentum 91,92
total four-momentum of fields 132,
160,174 parallel 7, 32, 33, 34
total three-momentum of fields 132, parallelogram law 7
139,147,174 world-lines 32,33,34
motion curve 26,89,92,96,98, 103, 111 particle, point 25,89, 100, 106
equations of motion (see equations of charged, massive 106, 107
motion), in extended phase space free, massive 26, 102
98,111,112,113,119 photon, massless 26, 102
in phase space 92, 94 (see also motion curve)
in space 89 past, present, and future (see events)
in space and time 96 Pauli matrices 74, 77, 78, 149, 166
in space-time 103 perihelion shift 118
permutation 17,18,56,79,80,81,85,
106, 107, 108
Newtonian 89,92,98, 115 symbol (see Levi-Civita)
constant of gravitation 98, 115 group (symmetric) 56
212 Subject Index
planetary motions (or orbits) 98, 113, electromagnetic field 15, 18, 19, 140,
115,118 141, 142, 143, 146, 147
phase space 91,92,94,97, 111, 119, 168 gauge fields 147, 149, 163
extended 97, 111, 119, 168 interacting fields 160, 161, 162, 163,
photon (massless, electromagnetic field 164
quantum) 26, 30, 33, 102 Klein-Gordon field 133, 175
polar coordinates 22, 24, 25, 115, 181, wave equation 47, 70, 123, 124, 139
182, 183, 192 representation, matrix 8,59,61,63,84
spherical (see spherical polar) degree 59,61
Poincare transformation 42, 43, 45, 52, group 59,61,63,84
128,129,170,171 linear mapping 8
group 56, 57, 58, 170 irreducible 61,62,63
positive definite 2, 3 retardation, clock (see clock)
potential 92,98, 110, 113, 142, 143, 145, Riemannian manifold 25
146, 147, 149, 150, 151, 160, 161, rod (see Lorentz-Fitzgerald)
162, 163, 164, 165 Roman indices 2,4, 79,85, 120, 133,
electromagnetic four-potential 113, 147, 163, 169
142, 143, 145, 146, 147, 160, 161, lower case 2,4, 85, 169
162 capital 79, 120, 133, 147, 163, 169,
gauge potentials 149,150,151,163, 168
164,165,166,167 rotation 9,11,64,67,72,76,78,87,168
gravitational ten-potential 113, 114, group 67
115,116 matrix 9,11,72
projection 20,21,40,41,70, 73, 121 plane 11,67,72, 168
mappings 20,21,40,41,73,121 pseudorotation 64, 78, 87
M -orthogonal 40, 41 spatial 72, 76
transformations 70 Routh's procedure 95
stereographic 73
proper 11,28,29,33,53,58,63,64,68,
77,83,88, 102, 103, 105, 106 scalar 1,8, 14,21,72,73, 79, 84,123,
Lorentz mapping (or transformation) 133, 175
11,58,63,64,66,68,77,78,79,83, complex numbers 72, 73, 123
88 field (see Klein-Gordon)
mass (or rest) 102, 103, 105, 106 multiplication 1, 79, 84
time 28,29, 33, 53, 100, 112, 113 product (see inner product)
pseudoangle 64, 78 real numbers 1, 8, 14,21
Pythagoras, generalized theorem 31 Schwarz inequality 4, 5,6, 7, 8
Schwarzschild (see gravitation)
separable solution 134, 139, 140
quantum chromodynamics (QeD) 151, separation, along a curve 27,28,29, 30,
166 33, 34
functional 27
number 4, 5, 7, 8
rank 15,35,43, 155 signature 2, 168
matrix 35, 155 simultaneity 52
tensor 15, 43 similarity transformation 75, 87, 88
relative 48, 104, 105, 108 space 4,8,11,26,30,32,37,38,48,49,
energy 105 59,67,89
force 104 coordinates 48,49, 89
kinetic energy 105 reflection 11, 59
mass 104, 105, 108 spacelike separation 30, 32, 33
relativistic fields 120, 168 spacelike 2-flat 37, 38
Dirac field 151, 153, 154, 157, 158, spacelike 3-flat 37
160, 161, 162, 163, 164, 191, 192 spacelike vectors 4, 8, 26, 104
Subject Index 213
space-time manifold 20, 25 timelike 4,5,6,8,26,37,38, 102

diagram 27,29,30,31,32,33,37,39, curve 26
40,46,50,52,53,100 2-flats 37,38
speed of light (c = 1) 27,49 momentum 102
spherical polar coordinates 22, 25, 115 velocity 101
spin angular momentum 130,171 vectors 4, 5, 6, 8, 26
spin-(1/2) particle (see Dirac equation) transformation 11,13,15,20,21,22,24,
spin-O particle (see Klein-Gordon 40,41,43,67,72,74,76,79,84,
equation) 121, 168
spinors 79, 80, 84, 85 coordinates 20,21,22,24
spin or fields 152, 159 Galilean 50
spinor-tensors 81, 82 infinitesimal (see infinitesimal)
symmetric 84, 85 Legendre 91,92,97,110
straight lines in space-time 26, 29, 30, Lorentz (see Lorentz)
31, 32,40, 52, 53, 105, 110 Mobius 74
string 52, 133 Poincare (see Poincare)
structure constants 148, 149, 150, 151, projective 20,21,40,41, 70, 73,
163, 167 121
summation convention 2,3,79,85, 147, rotation 11,67, 72, 76, 168
163, 169 spinor components 79, 84
symmetry (group) 55 tensor components 15,43
Young symmetries 63 vector components 13, 43
symmetric group 55 translation, coordinates 57, 58
symplectic 80, 119 infinitesimal 128, 171
Synge's theorems 5, 32, 39 trace 75, 165
standard clock (see clock) tracefree 62, 75
stationary (or critical) values 90, 122 twin paradox 28, 29
subspace, vector 3,4, 8
subatlas (see differentiable manifolds)
subgroup (see group) unimodular condition 74, 75, 76, 78,
148, 149, 163, 198
unitary unimodular condition 74,75,
tangent 23, 147 76, 78, 148, 149, 163
vectors 23, 147 unit vector (see vectors)
space 23, 147 units 27, 115, 134
tensors 4,16,17,18,19,43,44,47,106, speed of light c = 1 27
108,120,120,142,170 c = G = 1 115
Cartesian 4, 89, 106, 108 c=h=I134
contraction 16, 44
Minkowski (mixed) 14,15,16, 18, 19
fields 43,44,47,120,142 variational formalism (see Euler-
numerical 17,19 Lagrange equations)
permutation symbol 17,18 vectors, Minkowski 1, 23
pseudo tensor 18, 19,81,82,87, future and past pointing 5, 8
142 null 4,6,8,26,102,144
raising and lowering indices 17, 18, spacelike 4,8,26, 104
44 separation number 4, 5, 7, 8
tensor or outer product 16,44 tangent vectors 23, 147
tetrad (see basis set) timelike 4,5,6,8,26, 101, 102
time 11,28,29, 33, 53, 59, 88, 100, 112, un~ 4,5,6,4~ 101, 131, 173
113 vector space, inner product 1,23
proper 28,29,33, 53, 100, 112, 113 four-dimensional, complex 85
retardation (see clock) Minkowski 1, 2, 23
reversal 11,59,88 subspace 3, 4, 8
214 Subject Index
vector space (cont.) wave equation 47, 70, 123, 124, 139
three-dimensional 3, 4 fields (see relativistic fields)
two-dimensional, complex 79, frequency 135
80 operator or d'Alembertian (see
velocity 50,51,89, 101, 113, 119 d' Alembertian)
composition 55 plane wave 135, 143, 154, 177
four-vector 101, 106, 108, 109, 110, world-line 26, 28, 30, 32, 33, 100
112, 113, 114, 119 (see also motion curve)
three-vector 89,91,93, 101, 104, 105, work 89,107
108, 112, 113 function 89, 91, 92
volume, integral 45, 46, 120, 131, 132, Weyl equation 152
135, 173
four-dimensional 45,46, 120, 135
three-dimensional 131, 132, 135 Yasskin class of gauge fields 151
Universitext (continued)
Sagan: Space-Filling Curves

Samelson: Notes on Lie Algebras
Schiff: Normal Families
Shapiro: Composition Operators and Classical Function Theory
Smith: Power Series From a Computational Point of View
SmoryJiski: Self-Reference and Modal Logic
Stillwell: Geometry of Surfaces
Stroock: An Introduction to the Theory of Large Deviations
Sunder: An Invitation to von Neumann Algebras
Tondeur: Foliations on Riemannian Manifolds

The Special Theory of Relativity. Anadijiban Das

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

The Special Theory of Relativity. Anadijiban Das

Caricato da

Copyright:

Formati disponibili

Universitext

Springer Science+Business Media, LLC

AksoylKhamsi: Nonstandard Methods in Fixed Point Theory

(continued after index)

The Special Theory

Printed on acid-free paper.

1993 by Springer Science+Business Media New York

9 8 7 6 5 4 3 2(Corrected second printing, 1996)

The material in this book is presented in a logical sequenc:e rather than a

axiom is called the Principle of Special Relativity. Many experiments in-

Chapter 1. Four-Dimensional Vector Spaces and Linear Mappings 1

Chapter 2. Flat Minkowski Space-Time Manifold M4 and

Chapter 3. The Lorentz Transformation 48

Chapter 4. Pauli Matrices, Spinors, Dirac Matrices,

Chapter 5. The Special Relativistic Mechanics 89

5.3. The Relativistic Equation of Motion of a Particle 100

Chapter 6. The Special Relativistic Classical Field Theory 120

Chapter 7. The Extended (or Covariant) Phase Space and

Answers and Hints to Selected Exercises 202

1.1. Minkowski Vector Space V4

The axiom 14 is called the axiom of nondegeneracy. It is a weaker axiom than

Sl. ).1 > 0, ).2 > 0, ).3> 0').4 < O.

over the repeated index is implied. For example, we write

Theorem (1.1.2): In terms of the Minkowski components, the inner product

By the axioms in (1.1.2) and (1.1.8) we have

Note that from (1.1.9)

(i) a vector U E V4 that satisfies U U > 0 is called a spacelike vector;

Example: Let {e l ,e 2 ,e 3 ,e 4 } be an M-orthonormal basis for V4 By (1.1.9),

The separation numher is a generalization of the concept of length and is

The M-orthogonality between two vectors in V4 is not always intuitively

Theorem (1.1.4): No two time like vectors in V4 can be M-orthogonal.

Proof: Let D, v be two timelike vectors. Thus we have

By the Schwarz inequality (1.1.13) we obtain

Suppose, contrary to the conclusion of the theorem, that D' v = 0. Then

Theorem (U.S) (Synge): Let i, t be two timelike, future-pointing unit vectors

Proof: For definiteness assume that i, t are future-pointing timelike vectors.

where X, XE 1Ft Therefore,

Theorem (1.1.6): A timelike vector cannot be M-orthogonal to a nonzero null

Proof: Suppose that t is a timelike vector and n is a nonzero null vector.

Now we shall prove a very counterintuitive theorem.

It is hard to plot the Minkowski vectors, since the concepts of length of a

FIGURE 1. Minkowski vectors e l ,

(i) Determine which of these are spacelike, timelike, or null vectors.

1.2. Lorentz Mappings of V4

The proof is left to the reader.

Example: Let us consider a linear mapping L such that

Now we shall define the Kronecker delta:

Example: Consider the sum

For an invertible linear mapping L, we denote the inverse mapping by A,

r ,i. (./4) cos (nj4) 0

Now we shall discuss the transformation properties of vector components

Proof: It is easy to see u = ukek = ujej = uj(l\ek) = Wjui)ek , so (Uk - Zkjuj)e k =

Theorem (1.2.3): A Lorentz matrix L must satisfy the matrix equation

Proof: From the definition of a Lorentz mapping it follows that

Corollary (1.2.1): The determinant of a Lorentz matrix L must satisfy

Proof: Taking the determinant of both sides of (1.2.8) we have

A Lorentz mapping L with det L = 1 is called a proper Lorentz mapping. A

T is an improper Lorentz mapping that is not orthochronous.

L(e 3 ) = e 3 = e3 , where -:rr < 8 < n.

The Lorentz mapping L is proper and orthochronous. The plane rotation is

FIGURE 2. Plane rotation by the

FIGURE 3. Boost mapping associated

1.3. The Minkowski Tensors

Theorem (1.3.1): The components Wi of a co vector W, under a Lorentz mapping