Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Editorial Board
(North America):
S. Axler
F.W. Gehring
P.R. Halmos
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
ISBN 978-0-387-94042-7
SPIN 10528157
Dedicated to Sri Gadadhar Chattopadhyaya
Preface
vii
viii Preface
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
xi
xii Contents
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
1.1. Minkowski Vector Space V4 3
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 }.
(1.1.9)
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.
Corollary (1.1.1): For two timelike vectors D, v such that u4 > 0, v4 > 0, we
have D'V < O.
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 (1.1.7): Two nonzero null vectors are M-orthogonal if and only if they
are scalar multiples of each other.
1.1. Minkowski Vector Space V4 7
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.
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.
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
8 1. Four-Dimensional Vector Spaces
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.
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
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:
~]
[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
10 1. Four-Dimensional Vector Spaces
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)
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.
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
-sin8 0 0]
cosll 0 0
o 1 0 .
o 0 1
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.
1.3. The Minkowski Tensors 13
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.
(1.3.4a)
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
= ( _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.
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
1.3. The Minkowski Tensors 17
raising indices:
(1.3.10)
r/ = dljr jk = dkj'lj. 0
(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
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
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),
20
2.1. A Four-Dimensional Differentiable Manifold 21
.
)
IR
IR
"
X
Us
X
" X- 1
Xo
" -1
XO X
1R4
IR
t--+-IR
~k
0(x , x2 "x3 x
1 4)
== det [OXi(X)]
--.- #0 0 (2.1.2)
o(xl, x 2 , x 3 , x 4 ) ox)
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
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 .
2.1. A Four-Dimensional Differentiable Manifold 23
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)
x 2 = f![2(t) == t,
24 2. Flat Minkowski Space-Time Manifold M4
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.
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
[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.
Jl
l 0 0
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
26 2. Flat Minkowski Space-Time Manifold M4
d .. dpi(t) df!"i(t) 0
'J 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
2.2. Minkowski Space-Time M4 and the Separation Function 27
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'[-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.
s= L 19t4+0+0-27t411/2dt=.j2. D
_ _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
2.2. Minkowski Space-Time M4 and the Separation Function 29
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
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)
30 2. Flat Minkowski Space-Time Manifold M4
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,
In the case where the straight world-lines Yl' Y2 are both spacelike we have
dij(b i - ai)(b i - a i ) > 0,
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.
2.2. Minkowski Space-Time M4 and the Separation Function 33
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.
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.
2.3. Flat Submanifolds of Minkowski Space-Time M4 35
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;
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.
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.
2.3. Flat Submanifolds of Minkowski Space-Time M4 37
(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.
(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.
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)
v
Xo
-----+----u
(u,v)
JR4
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'
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).
42 2. Flat Minkowski Space-Time Manifold M4
Since the above equation is valid for every pair of continuous vector fields in
UnO, we must have
(2.4.1)
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
44 2. Flat Minkowski Space-Time Manifold M4
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
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
2.4. Minkowski Tensor Fields on M4 45
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
where F(x) == f[x- 1 (x)] = f(p) and the integral on the right is a multiple
Riemann -in tegral.
t,
M4 II.
= 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.
46 2. Flat Minkowski Space-Time Manifold M4
f- ri.idxldxzdx3dx4 =
D
f oD
dijr i(x)n j(x)d 3 (J, (2.4.15)
The proof of this important theorem can be found in references at the end
of the chapter.
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 }.
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
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
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
X2 = x 2,
(3.1.8)
x3 = x 3 ,
50 3. The Lorentz Transformation
'
~
----+-------1 ~' --~~~--,---
x
[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
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)
52 3. The Lorentz Transformation
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)
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
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
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)
~2 = x2,
~3 = x3,
3.2. The Lorentz Group 24 55
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.
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
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.
(3.2.5)
3.2. The Lorentz Group 24 57
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
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~.
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
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++,
[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)
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.
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'.
(3.3.7)
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)
(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)
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
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, , .
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
[3 Z = SZ S 3 C 3,
[3 3 =C 1 C3 ,
[3 4 = 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.
66 3. The Lorentz Transformation
r:t.p E 1R 3 , -n < ()1 < n, 0 < ()2 < n, -n < ()3 < n}. (3.4.7)
(3.4.8)
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
68 3. The Lorentz Transformation
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
(3.4.13)
= -(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
"'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).
70 3. The Lorentz Transformation
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
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]
5. J. L. Synge, Relativity: The special theory, North-Holland, Amsterdam, 1964.
[pp. 91, 93]
4
Pauli Matrices, Spinors, Dirac
Matrices, and Dirac Bispinors
(4.1.2b)
72
4.1. Pauli Matrices, Rotations, and Lorentz Transformations 73
(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
(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 (4.1.2): 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.
(4.1.14)
(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.
V == V3 V 2 VI
o ] [cOS(e/2) i Sin(e/2)] [e i /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.
4.1. Pauli Matrices, Rotations, and Lorentz Transformations 77
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.
[
). jJ.] _ 0'1(jJ. + v) i0'2(jJ. - v) 0'3(,1, - r) a4.(). + r)
vr- 2 + 2 + 2 + 2'
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'
= [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
[
X3 - X4 Xl - ix 2 ]
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.
4.2. Spinors and Spinor-Tensors 79
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 '
(4.2.7a)
80 4. Pauli Matrices, Spinors, Dirac Matrices, and Dirac Bispinors
(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)
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-
4.2. Spinors and Spinor-Tensors 81
(4.2.13)
(4.2.14)
We choose a special set of numerical mixed tensor-spinors by the following
choice:
[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
we obtain
[~lAB] = -~\
[~2AB] = ~2,
(4.2.16)
[~3AB] = _~3,
[~4AB] = ~4.
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
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)
4.2. Spinors and Spinor-Tensors 83
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)
Proof: Recall that in (4.2.18) we had (JkAB(Jb AB = -2i5 bk Using this we get
(JbABXAB = (Xk(JkAB)(JbXB = _2 Xb;
= (-1/2)AAcABD(JaAB(J/Dxb).
The equation
l ab(J bAB = ACAADB(J aCD (4.2.22b)
is equivalent to (4.2.22a).
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.
84 4. Pauli Matrices, Spinors, Dirac Matrices, and Dirac Bispinors
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,
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.
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)
(4.3.3)
(4.3.4)
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)
4.3. Dirac Matrices and Dirac Bispinors 87
(4.3.9)
(4.3.10)
The transformation properties of the 'Y" matrices are given by the equation
(4.3.5), i.e.,
(4.3.11 )
(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.
~ [~
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
88 4. Pauli Matrices, Spinors, Dirac Matrices, and Dirac Bispinors
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
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.,
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
(5.1.1)
89
90 5. The Special Relativistic Mechanics
J[y] == f l2
[L(t,x,v)lx'~.'(I),v'~d:P(I)ldt]dt. (5.1.4)
I,
= f l2
I,
[L(t, x, V)lx'= ''(I)+eh'(t). v'=d'(I)/dl+e(dh'(I)/dl)
=e f l
2 {hlZ(t) [OL(t, x, V)]
I, OXIZ Ix'~'(I). v'~d'(I)fdl
=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..
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
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)
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
~:!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 ..
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
5.1. The Prerelativistic Particle Mechanics 93
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)
[ 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.
For the case of strictly positive energy, the energy hypersurface is 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
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~
,
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'
- 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")/"
and thus the modified Euler-Lagrange equations hold. However, there exists
the fourth Euler- Lagrange equation:
= 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
- 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)
[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 ..
'
98 5. The Special Relativistic Mechanics
[ 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..
dP;(r) = 0
dr '
dP~(r)
dT
+ )Mmb
' ap
xP(b Xl'x v)-3/2
I'V
= 0
,
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,
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.
R
o s
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
Thus
(S.3.6b)
Hence,
For a photon, one can choose a special parameter T along the null world-
102 5. The Special Relativistic Mechanics
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,
Theorem (5.3.2): The proper mass M(s) is a constant along the massive world-
line if and only if
(5.3.11)
df![k(S) I
= dkl~F (x,u)I'"
104 5. The Special Relativistic Mechanics
Thus
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.
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 ..
: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
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,
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
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)
5.3. The Relativistic Equation of Motion of a Particle 107
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.
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 <I>(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 ]
which implies that vet) = vo, a constant. Putting the above result in either of
the equations of motion we obtain
108 5. The Special Relativistic Mechanics
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
koA(X,U) _ A( ) (5.4.1 )
U OUk - x,U.
= 1<2
<,
[A(x, u)I'.] dr. (5.4.2)
dyer) 0
i = Y(r), ~>, (5.4.3)
[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
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)
110 5. The Special Relativistic Mechanics
Thus
oA(x, u) = 0
OX k -
_ 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
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.
. .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)
(5.4.9)
r
tion of a constant) defined by the indefinite integral
() = SeT) == l(~)d~,
~i(() == ~i(r), (5.4.11)
iWJ) == Pier),
112 5. The Special Relativistic Mechanics
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.. '
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)
d?[i(e) .. _
~ = d'JPie)/m,
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),
(5.4.19)
dlJiace)
dO
= [an(x,p)]
amp
= (1/ )[ a _ eAa( )]
x I.. '
Pa I ..
Thus we find
d 2 pza(e) ab dPZk(e)
m de 2 = ed ------;w- [Ak,b - Ab.k] I .
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)
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
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
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~,
= 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)
_~{[CA2(")J
d out I..
}= ~{
d m
[(1 _ 2M/)]
r I
. dl"4()}
d
=0;
thus,
(S.4.33b)
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'
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
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]
5. J. L. Synge, Relativity: The special theory, North-Holland, Amsterdam, 1964.
[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
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
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
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
= 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.
Proof: The proof is straightforward from equation (6.1.6) and the preceding
lemma.
6.1. Variational Formalism for Relativistic Classical Fields 123
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"
[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)
124 6. Special Relativistic Classical Field Theory
[0ft'(~
oy
.)J - ~
I..dx
{[o ft'~
oy
.)J } 0, i I..
=
[ Oft'(~.
oW
)J - dx~ {[Oft'(~.
I.. ow
)J } = o. i I..
{[ Oft'(~')J
oy I..
- i[Oft'(~')J
oW I ..
}
- dx~{[Oft'~')J
oy
- i[Oft'~')J
k oW } = O.
I.. k I ..
[o.\')J - ~{[02~')J }=
0, I.. dx 0, k I ..
O. (6.1.9)
02( .. ) = _dkl"f
O'k "I'
[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
6.1. Variational Formalism for Relativistic Classical Fields 125
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)
with t: == [(t: 1)2 + '" + (t: r)2J1/2. The small variations are summarized in the
following equations:
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)],
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..
6.1. Variational Formalism for Relativistic Classical Fields 127
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
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"
+ [ 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 ..
0= e(/l) Iv {d~i([~t)(X)2'( n. ..
128 6. Special Relativistic Classical Field Theory
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)
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,
d i { J ij[.:t'cn ..
-d + [a.:t'(
-a .. )] [0 - J kj R ,k] } = 0.
R.
x y, I ..
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
6.1. Variational Formalism for Relativistic Classical Fields 129
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,
(6.1.19)
130 6. Special Relativistic Classical Field Theory
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
t2 ------------ 1'-";;"-"'"
J(P' == f J(!,(Xl,X2,x3,tddx!dx2dx3
f
o,D
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
= folD
J/",{x)n l dx! dx 2 dx 3 + f J(~,{x)ni dx! dx 2dx 3
02D
f
olD 02D
= J(!,(Xl,X2,x3,t2)(1)dxldx2dx3.
f
o,D
+ J(!,(xl,x2,x3,t2)(-1)dxldx2dx3.
o2 D
132 6. Special Relativistic Classical Field Theory
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
f
R3
Pa == f R3
T4a(x!,x2,x3,0)dx!dx2dx3,
f
R3
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
(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 )"# ,
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}.
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
a!'~.. )= _ Jl2(,
a( (6.2.2)
[ a2'~
a(
.. ~] _ ~ {[a!'~ .. )] } = _ Jl21/t(x) + dkll/t.kl = 0,
I.. dx a(k I ..
(6.2.3)
t/I(x) = O/4n2) fR4 r(kl' k2' k3' k4)eikaXa dkl dk2 dk3 dk 4
(6.2.6)
t/I(x) = (1/2n)3/2 f R3
[2w(k)]-1/2{IX(k)e i [k.X.-ro(k)x 4 ]
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.
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.
:!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)ii(k')ei(k.-k~)x. - P(k)p(k')e-i(k.-k~Jx"
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
138 6. Special Relativistic Classical Field Theory
has Ig(x) - g(O) I < e whenever Ixl < 15. Thus we have
= r:, Ig(x) - g(O)1 (sin Lxlx)dx + t-~' Ig(x) - g(O) I (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')
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
6.2. The Klein-Gordon Scalar Field 139
= 0/2) f ~3
[w(k)rl
= 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.
140 6. Special Relativistic Classical Field Theory
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:
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:
where E is the parameter representing the energy eigenvalue. Show that for
the case of square-integrable solutions, i.e.,
t7 H() = aE( . . )
v x.. at'
VE( .. ) = 0,
(6.3.1)
VH( .. ) = 0,
oH( .. )
V x E( .. )= -~'
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
142 6. Special Relativistic Classical Field Theory
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
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)
6.3. The Electromagnetic Tensor Field 143
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)
144 6. Special Relativistic Classical Field Theory
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
02( .. ) =
OYa -
,
(6.3.19)
6.3. The Electromagnetic Tensor Field 145
i i [82"( .. )]
T j(x) = b j2"(. ')1" - -8-, Aa,j
Ya. I ..
(6.3.20)
()ii == 0,
()ij , = Tii , =
,J ,J
0'
Pk = f GP
()4 k(X, 0) d 3 x. (6.3.22)
(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
Ab(X) = (2nr3/2 f R3
[2v(k)rl/2
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
= -(1/4) fR'
[v(k)rl[[v(k)2 - kvkv] [o:b(k)O:b(-k) + iXb(k)iXb(-k)]
(6.3.27)
EXERCISES 6.3
1. Let the complex electromagnetic field be given by [see (6.3.7)]
PI' == f~3
1(O)\(x,0)d 3 x = f
~3
[a b(k)iXb(k)]kl'd 3 k.
= CEABrdp), (6.4.1)
148 6. Special Relativistic Classical Field Theory
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
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
(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
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
X + CBEFyEkyFl]
[yB'k - yBkl
+ [yAji - yAij + CACDyCiyDJ
X [C BGF{)\yF, + C BEG{)S/yEkJ]
150 6. Special Relativistic Classical Field Theory
[ 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)
6.5. The Dirac Bispinor Field 151
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.
(6.5.1 )
Thus
or equivalently we get
-
d ab~B_"'- _ 2",_()
U C'f'B.ab - -m 'f'C x .
Hence, we have
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).
(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).
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
154 6. Special Relativistic Classical Field Theory
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
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).
~ "(PJ l-i(m
-i(m} El-'p,
+ Erl(pi + iP2)
l "(PJl
+ -i(m
-i(m
+ El~'(P' ip'll'
-
+ E)-l p3
156 6. Special Relativistic Classical Field Theory
(6.5. 12a)
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.
+ Pr(p)vr(p)e- i [P,X"-E(P)x4)] } d 3 p,
(6.5.14)
6.5. The Dirac Bispinor Field 157
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:
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
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)
158 6. Special Relativistic Classical Field Theory
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 ..
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'
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).
x [iir(p)ur(p)te-iP,x' + Pr(p)vr(p)eip,X']
x [a,(p')us(p')eip;X'
= -(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
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.
160 6. Special Relativistic Classical Field Theory
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
~~U
oG = (1/2)1'~uy au v'
(6.6.3)
02( .. ) -
- 0 - = (ie)'uya"v,v,
Ya
02( .. ) = (yba _ yab).
OYab
Thus the Euler- Lagrange equations are
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)
yaDat/J(x) + mt/J(x) = 0,
jja~(x)ya - m~(x) = ot, (6.6.6)
pab,b = ie~(x)yat/J(x).
162 6. Special Relativistic Classical Field Theory
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) ~ -',;:,)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,
6.6. Interaction of the Dirac Field with Gauge Fields 163
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
aif( .. ) - au
a(Jva = (JuY V'
aif( .. ) . - su K Jv
aes = Zg(KuY v ~ J(
= 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 ..
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),
Lemma (6.6.1): Under the local gauge transformation (6.6.l4b) the gauge field
components transform by the rule
(6.6.l4c)
EXERCISES 6.6
1. Consider the Lagrangian function given in equation (6.6.1). Prove that
6.6. Interaction of the Dirac Field with Gauge Fields 167
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*
=(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
[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
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,
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)
d(rjJ") == III ['(~, "'''''; (A' ".4.')]1(" =qfW,.,.G =4i"~ dS" (7.1.11)
a'( .. ) _ ~{[a~UJ
a(" I" d, a(;.j I..
}= O. (7.1.12)
(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.. (,)
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)
172 7. Extended Phase Space and Classical Fields
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)
KA == f- T4A(~)d7~,
f- rBd~)d7~
Do
JBC == = -JCB ,
f- [T(4)(~)
Do
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,
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.
[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
jABd~) = MABd~)
= ~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,
Xi =K i+4 =Kpi = - f ~7
[;P,q4,pi + (C.C)]I .. d3 qd 4p,
f - -
~7
(7.2.4)
Lij = [(P,,pi - Pj,pi),q4 + (C.C.H .. d 3 4
q d p,
~7
B = 0,
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')
= (2fl f {f
D7 1R6
f""
-W'
[(iX'a - fJ'P)b 3 (k - k')
= (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)
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):
= (2)-1 f
D7
f D7
(rul')1/2
- [Q 2 _ + xi - J12]
k~k~ - x~x~
= fD7
[la(k,x)1 2 + IP(k,xW]Q(k,x)d 3 kd 4 x. (7.2.12)
K~ = f D7
[f+(k,x) + !_(k,x)]k~d3kd4X,
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,
-f
GO!7
(7.2.14)
12 == {p4[2"(n}lq4=od 3qd 4p.
GO!7
12 == (2r1(2nr7 ff GO! 7 D7
f
D7
p4(nnT1/2
{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)]
7.2. The Generalized Klein-Gordon Equation 181
(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
where the index b is not summed. The asymptotic version of (7.2.23) for large
values of pb'S is
= L
<Xl
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
+ P(!~'~'X4{}] (-i)'fJ-",(KppfJ)]
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
f:"
(7.2.31 )
P#(.,., .) == (4n)-1/2 b(.,.,.)
L
00
= (2n)-1/2 C(#(~,,t4)exp[jt44],
(7.2.32)
L
00
L
00
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 KIK2K3d3~d3
N = fD7
IIX(k,xW d 3 kd 4 x
= f la(~,p,wsinrp4WwcoSrp4KlK2K3d3!id4rp.
D7
# (7.2.36)
[f"
x exp[i(t4 - 4)rp4] dIV4]}
N = 2 If. 11X#(~'P,t4WKIK2K3d3~d3p
'4 D6
Let us consider the total energy H from the equation (7.2.12) (with f3 == 0):
186 7. Extended Phase Space and Classical Fields
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
= f:" 1[(wcost;64)/2]1/21X(~,t,wsint;64Wcost;64dt;64
2w
= w L L {a#(~'t,t4)C(#(~'t,f4)
t4 ;4
+ exp[i(t4 - t4 - 1)]t;64)dt;64]}
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)]}
(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)
7.2. The Generalized Klein-Gordon Equation 187
(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
x L (tl + t2 + t3 -
,
t4 ) fOO .. y(.,
0+
.)[TIP=1
(-i)'fJJ,p(')]
= _(2)-1 L L L (-1)/4+i4(tl + t2 + t3 - t4 )
t t4 i4
(7.2.44)
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+
7.2. The Generalized Klein-Gordon Equation 189
J = Ii + 11 = r l I I [1 - (-1Y4+i 4](t l + t2 + t3 - t 4 )
t ;4
B:=O,
(7.2.47)
reduce to
KI = K2 = 0,
K3= foo .. F(!i, T3, T4)"3 d3!i,
0+
Xl = X 2 = X3 = 0,
B=O,
f(x) = too [a(k) cos kx + b(k) sin kx] dk = too [y(k)e ikx + y(k)e- ikX ] dk,
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,
Here, (h.c.) stands for the hermitian conjugation of the precc:ding terms.
192 7. Extended Phase Space and Classical Fields
f -
R7
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
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
1
1
1
1
, 1X 2 - --
-1
-1
-1
-1
1 1
-1
3- -1
,IX =
-1
-1
1
1
-(u,-iud
uZ-iu1
194 7. Extended Phase Space and Classical Fields
-(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
[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]
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]
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)
7.3. Spin-! Fields in the Extended Phase Space 197
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
-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.
7.3. Spin-! Fields in the Extended Phase Space 199
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
L
00
~L(p,(J) = ~L(p,(J;t1,t2,t3,t4)'
t=-oo
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
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 '
EXERCISES 6.1
EXERCISES 6.2
EXERCISES 6.3
2. Hint: Use special Minkowski coordinates such that
F34(X) = F12 (x) == O.
Index of Symbols
204
Index of Symbols 205
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
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
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
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)