Spacetime and elds

arXiv:0911.0334v1 [gr-qc] 2 Nov 2009

Nikodem J. Poplawski
Department of Physics, Indiana University, Bloomington, IN 47405, USA

Contents
1 Spacetime 1.1 Tensors . . . . . . . . . . . . . . . . . . . . . . 1.1.1 Vectors . . . . . . . . . . . . . . . . . . . 1.1.2 Tensors . . . . . . . . . . . . . . . . . . 1.1.3 Densities . . . . . . . . . . . . . . . . . . 1.1.4 Contraction . . . . . . . . . . . . . . . . 1.1.5 Kronecker and Levi-Civita symbols . . . 1.1.6 Dual densities . . . . . . . . . . . . . . . 1.1.7 Covariant integrals . . . . . . . . . . . . 1.1.8 Derivatives . . . . . . . . . . . . . . . . 1.2 Ane connection . . . . . . . . . . . . . . . . . 1.2.1 Covariant dierentiation of tensors . . . 1.2.2 Parallel transport . . . . . . . . . . . . . 1.2.3 Torsion tensor . . . . . . . . . . . . . . . 1.2.4 Covariant dierentiation of densities . . 1.2.5 Covariant derivatives . . . . . . . . . . . 1.2.6 Partial integration . . . . . . . . . . . . 1.2.7 Geodesic frame of reference . . . . . . . 1.2.8 Ane geodesics and four-velocity . . . . 1.2.9 Innitesimal coordinate transformations 1.2.10 Killing vectors . . . . . . . . . . . . . . . 1.3 Curvature . . . . . . . . . . . . . . . . . . . . . 1.3.1 Curvature tensor . . . . . . . . . . . . . 1.3.2 Integrability of connection . . . . . . . . 1.3.3 Parallel transport along closed curve . . 1.3.4 Bianchi identities . . . . . . . . . . . . . 1.3.5 Ricci tensor . . . . . . . . . . . . . . . . 1.3.6 Geodesic deviation . . . . . . . . . . . . 1.4 Metric . . . . . . . . . . . . . . . . . . . . . . . 1.4.1 Metric tensor . . . . . . . . . . . . . . . 1.4.2 Christoel symbols . . . . . . . . . . . . 1.4.3 Riemann curvature tensor . . . . . . . . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 4 4 5 5 6 7 8 8 9 9 9 11 11 12 13 13 14 14 16 18 18 18 19 20 21 21 22 23 23 25 27

1.5

1.6

1.7

1.4.4 Properties of Riemann tensor . . . . . . . . . . . . . 1.4.5 Weyl tensor . . . . . . . . . . . . . . . . . . . . . . . 1.4.6 Metric geodesics . . . . . . . . . . . . . . . . . . . . . 1.4.7 Galilean frame of reference and Minkowski tensor . . 1.4.8 Intervals, proper time and distances . . . . . . . . . . 1.4.9 Spatial vectors . . . . . . . . . . . . . . . . . . . . . Tetrad and spin connection . . . . . . . . . . . . . . . . . . 1.5.1 Tetrad . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5.2 Lorentz transformation . . . . . . . . . . . . . . . . . 1.5.3 Tetrad transport . . . . . . . . . . . . . . . . . . . . 1.5.4 Spin connection . . . . . . . . . . . . . . . . . . . . . 1.5.5 Tetrad representation of curvature tensor . . . . . . . Lorentz group . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6.1 Subgroups of Lorentz group and principle of relativity 1.6.2 Innitesimal Lorentz transformations . . . . . . . . . 1.6.3 Generators and Lie algebra of Lorentz group . . . . . 1.6.4 Rotations and boosts . . . . . . . . . . . . . . . . . . 1.6.5 Poincar e group . . . . . . . . . . . . . . . . . . . . . 1.6.6 Casimir operators of Lorentz and Poincar e group . . 1.6.7 Relativistic kinematics . . . . . . . . . . . . . . . . . 1.6.8 Four-acceleration . . . . . . . . . . . . . . . . . . . . Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.7.1 Spinor representation of Lorentz group . . . . . . . . 1.7.2 Spinor connection . . . . . . . . . . . . . . . . . . . . 1.7.3 Curvature spinor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

29 30 30 31 31 33 35 35 36 36 37 38 39 39 40 41 42 44 46 47 52 52 52 54 55 57 57 58 60 60 60 61 61 63 64 64 64 65 66 67 69 72 80 84

2 Fields 2.1 Principle of least action . . . . . . . . . . . . . . . . 2.2 Action for gravitational eld . . . . . . . . . . . . . 2.3 Matter . . . . . . . . . . . . . . . . . . . . . . . . . 2.3.1 Metric dynamical energy-momentum density 2.3.2 Tetrad dynamical energy-momentum density 2.3.3 Canonical energy-momentum density . . . . 2.3.4 Spin density . . . . . . . . . . . . . . . . . . 2.3.5 Belinfante-Rosenfeld relation . . . . . . . . . 2.4 Symmetries and conservation laws . . . . . . . . . . 2.4.1 Noether theorem . . . . . . . . . . . . . . . 2.4.2 Conservation of spin . . . . . . . . . . . . . 2.4.3 Conservation of metric energy-momentum . 2.4.4 Conservation of tetrad energy-momentum . 2.4.5 Conservation laws for Lorentz group . . . . 2.4.6 Components of energy-momentum tensor . . 2.4.7 Mass and Papapetrou equations of motion . 2.4.8 Energy-momentum tensor for particles . . . 2.5 Gravitational eld equations . . . . . . . . . . . . . 2

2.6

2.7

2.5.1 Einstein-Hilbert action and Einstein equations . . . 2.5.2 Einstein pseudotensor and principle of equivalence . 2.5.3 Landau-Lifshitz energy-momentum pseudotensor . . 2.5.4 Utiyama action . . . . . . . . . . . . . . . . . . . . 2.5.5 Mller pseudotensor . . . . . . . . . . . . . . . . . 2.5.6 Einstein-Cartan action . . . . . . . . . . . . . . . . 2.5.7 Kibble-Sciama action . . . . . . . . . . . . . . . . . 2.5.8 Einstein-Cartan pseudotensor . . . . . . . . . . . . 2.5.9 Palatini variation . . . . . . . . . . . . . . . . . . . 2.5.10 Gravitational potential . . . . . . . . . . . . . . . . 2.5.11 Hydrodynamics . . . . . . . . . . . . . . . . . . . . Spinor elds . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6.1 Dirac matrices . . . . . . . . . . . . . . . . . . . . . 2.6.2 Dirac equation . . . . . . . . . . . . . . . . . . . . 2.6.3 Spinors in Einstein-Cartan gravity . . . . . . . . . Electromagnetic eld . . . . . . . . . . . . . . . . . . . . . 2.7.1 Gauge invariance and electromagnetic potential . . 2.7.2 Electromagnetic eld tensor . . . . . . . . . . . . . 2.7.3 Lagrangian density for electromagnetic eld . . . . 2.7.4 Electromagnetic current . . . . . . . . . . . . . . . 2.7.5 Maxwell equations . . . . . . . . . . . . . . . . . . 2.7.6 Energy-momentum tensor for electromagnetic eld 2.7.7 Lorentz force . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . .

84 85 87 89 89 90 92 93 94 95 96 97 97 98 100 102 102 104 106 106 108 111 112

Spacetime

Einsteins principle of general covariance states that all physical laws do not change their form (are covariant) under continuous coordinate transformations in four-dimensional spacetime.

1.1
1.1.1

Tensors
Vectors

Consider a coordinate transformation from old (unprimed) to new (primed) coordinates in a four-dimensional manifold: xi x j (xi ),

(1.1.1)

where x j are dierentiable and nondegenerate functions of xi and the index i can j x j has the nonzero determinant | | = 0, so xi are be 0,1,2,3. Thus the matrix x i x xi i j x dierentiable and nondegenerate functions of x . The matrix x j is the inverse of
x j : xi

x i xk k = j , xj x i 1 i=k 0 i=k .

(1.1.2)

where
i k =

(1.1.3)

The dierentials and derivatives transform according to x j i dx , dx = xi xi = . x j x j xi

j

(1.1.4) (1.1.5)

A scalar (invariant) is dened as a quantity that does not change: = . (1.1.6)

A contravariant vector is dened as a quantity that transform like a dierential: Aj =

x j i A. xi

(1.1.7)

A covariant vector is dened as a quantity that transforms like a derivative: xi B = j Bi . x

j

(1.1.8)

Therefore a derivative of a scalar is a covariant vector. The coordinates xi do not form a vector. 4

1.1.2

Tensors

A product of several vectors transforms such that each coordinate index transforms separately: x i x j xp xq m n A B . . . C p Dq . . . . xm xn x k x l A tensor is dened as a quantity that transforms like a product of vectors: A i B j . . . C k D l . . . =

(1.1.9)

ij...

kl...

x i x j xp xq = m n k l T mn... pq... . x x x x

(1.1.10)

A tensor is of rank (k, l) if it has k contravariant and l covariant indices. A scalar is a tensor of rank (0,0), a contravariant vector is a tensor of rank (1,0), and a covariant vector is a tensor of rank (0,1). A linear combination of two tensors of rank (k, l) is a tensor of rank (k, l). The product of two tensors of ranks (k1 , l1 ) and (k2 , l2 ) is a tensor of rank (k1 + k2 , l1 + l2 ). Tensor indices (all contravariant or all covariant) can be symmetrized: 1 T(ij...k) = T{ij...k} , (1.1.11) n! permutations or antisymmetrized: T[ij...k] = 1 n! permutations T{ij...k} (1)m , (1.1.12)

where n is the number of symmetrized or antisymmetrized indices and m is the number of permutations that bring Tij...k into T{ij...k} . For example, for two indices: 1 1 T(ik) = 2 (Tik + Tki ) and T[ik] = 2 (Tik Tki ), and for three indices: T[ijk] = 1 (Tijk + 3 Tjki + Tkij ). If n > 4 then T[ij...k] = 0. Symmetrized and antisymmetrized tensors or rank (k, l) are tensors of rank (k, l). Symmetrization of an antisymmetric tensor or antisymmetrization of a symmetric tensor bring these tensors to zero. Any tensor of rank (0,2) is the sum of its symmetric and antisymmetric part, T(ik) + T[ik] = Tik . (1.1.13)

The number 0 can be regarded as a tensor of arbitrary rank. Therefore all covariant equations of classical physics must be represented in the tensor form: T ij... kl... = 0. 1.1.3 Densities

The element of volume in four-dimensional spacetime transforms according to x i 4 d x. dx = xk

4

(1.1.14)

A scalar density is dened as a quantity that transforms such that its product with the element of volume is a scalar, d4 x = d4 x:

xi . x k 5

(1.1.15)

A tensor density, which includes a contravariant and covariant vector density, is dened as a quantity that transforms like a product of a tensor and a scalar density:

ij...

kl...

xi x i x j xp xq mn... = pq.... x k xm xn x k x l

(1.1.16)

For example, the square root of the determinant of a tensor of rank (0, 2) is a scalar density of weight 1:
|Tik |=

xl xm Tlm | = x i x k

xj xj 2 | | T = | | |Tik |. ik x n x n

(1.1.17)

The above densities are said to be of weight 1. One can generalize this denition of densities by introducing densitites of weight w , which transform like normal densities xi xi w except that | x For example, d4 x is a scalar density of k | is replaced by | | . x k weight -1. A linear combination of two densities of weight w is a density of weight w . The product of two densities of weights w1 and w2 is a density of weight w1 + w2 . Symmetrized and antisymmetrized densities of weight w are densities of weight w . Densities of weight 1 are simply referred to as densities. Tensors are densities of weight 0. 1.1.4 Contraction

We adopt Einsteins convention: if the same coordinate index i appears twice (as a contravariant index and covariant index) then we perform the summation i over a given tensor or density. Such a tensor or density is said to be contracted over index i. A contracted tensor of rank (k, l) transforms like a tensor of rank (k 1, l 1): x j xq p mn... x j xq mn... x i x j xp xq = T = T mq... . T il... = m n i l T mn... pq... pq... x x x x xn x l m xn x l (1.1.18) i For example, the contraction of a contravariant and covariant vector A Bi is a scalar (scalar product). A contracted tensor density of rank (k, l) and weight w transforms like a tensor density of rank (k 1, l 1) and weight w :
ij...

xi xi w x i x j xp xq mn... = il... = k pq... x xm xn x i x l x k xi w x j xq mn... = mq... . x k xn x l

ij...

w x j

xq p mn... pq... xn x l m (1.1.19)

Contraction of a symmetric tensor with an antisymmetric tensor (over indices with respect to which these tensors are symmetric or antisymmetric) gives zero. If contraction of two tensors gives zero, these tensors are said to be orthogonal. Two orthogonal vectors (one contravariant and one covariant) are said to be perpendicular.

1.1.5

Kronecker and Levi-Civita symbols

i The Kronecker symbol k (1.1.3) is a tensor with constant components: i k =

x i xj x i xl j i = = k . xj x k l xj x k

(1.1.20)

A totally antisymmetric tensor of rank (4, 0), T ijkl = T [ijkl] has 1 independent component T : T ijkl = T ijkl , where ijkl is the totally antisymmetric, contravariant permutation Levi-Civita symbol: 0123 = 1, ijkl = (1)m , (1.1.21)

and m is the number of permutations that bring ijkl into 0123 . The determinant of i a matrix Sk is dened through the permutation symbol as
r ijkl i j k l mnpq | Ss | = Sm Sn Sp Sq . i Taking Sk = x i k

(1.1.22)

gives
ijkl

xr x i x j x k x l mnpq = . x s xm xn xp xq

(1.1.23)

This equation looks like a transformation law for a tensor density with constant components: ijkl = ijkl . Accordingly, T is a scalar density of weight -1. We also introduce the covariant Levi-Civita symbol ijkl through:
i m j = m k m l m i n j n k n l n i p j p k p l p i q j q k . q l q

ijkl mnpq

(1.1.24)

Thus the covariant Levi-Civita symbol is a tensor density of weight -1 and its product with a scalar density is a tensor. The covariant Levi-Civita symbol is given by 0123 = 1, ijkl = (1)m ,
r n p q | Ss |ijkl = Sim Sj Sk Sl mnpq .

(1.1.25)

where m is the number of permutations that bring ijkl into 0123 , and satises (1.1.26)

Contracting (1.1.24) gives the following relations: ijkl mnpl

i i i m n p j j j n p , = m k k k m n p

i j i j ijkl mnkl = 2(m n n m ), ijkl i mjkl = 6m , ijkl ijkl = 24.

(1.1.27)

1.1.6

Dual densities

A contracted product of a covariant tensor and the contravariant Levi-Civita symbol gives a dual contravariant tensor density: iklm Am =
ikl

, iklm Blm =

ik

, iklm Cklm =

(1.1.28)

A contracted product of a contravariant tensor and the covariant Levi-Civita symbol gives a dual covariant tensor density: iklm Am =
ikl ,

iklm B lm =

ik ,

iklm C klm =

i.

(1.1.29)

Therefore there exists an algebraic correspondence between covariant tensors and contravariant densities, and between contravariant tensors and covariant densities.

1.1.7

Covariant integrals

A covariant line integral is an integral of a tensor contracted with the line diereni tial dxi : T j... i... dx . A covariant surface integral is an integral of a tensor contracted with the surface dierential df ik = dxi dx k dxk dx i (which can be geometrically ik represented as a parallelogram spanned by the vectors dxi and dx i ): T j... ik... df . A covariant hypersurface (volume) integral is an integral of a tensor contracted with dxi dx i dxi the volume dierential dS ikl = dxk dx k dxk (which can be geometrically rep dxl dx l dxl ikl resented as a parallelepiped spanned by the vectors dxi , dx i ) and dxi : T j... ikl... dS . A covariant four-volume integral is an integral of a tensor contracted with the fourvolume dierential dS ijkl, dened analogously to dS ikl . The dual density corresponding to the surface element is given by 1 dfik = lmik df lm . 2 The dual density corresponding to the hypersurface element is given by 1 dSi = klmi dS klm . 6 The dual density corresponding to the four-volume element is given by d = 1 iklm dS iklm = dx0 dx1 dx2 dx3 . 24 (1.1.32) (1.1.31) (1.1.30)

Covariant integrands that include the above dual densities of weight -1 must be multiplied by a scalar density, for example, by the square root of the determinant of a tensor of rank (0, 2). According to Gau and Stokes theorems, there exists relations

between integrals over dierent elements: , xk dfik dSi k dSk i , x x dSi d i . x dxi df ik 1.1.8 Derivatives (1.1.33) (1.1.34) (1.1.35)

A derivative of a covariant vector does not transform like a tensor: xl xm xl xm Am 2 xm Ak = A = + Am , m x i x i xl x k x i x k xl x i x k (1.1.36)

because of the second term which is linear and homogeneous in Ai , unless xi are linear functions of x j . This term is symmetric in the indices i, k so the antisymmetric part k of A with respect to these indices is a tensor: xi [i Ak] = xl xm xl xm A = [l Am] , l m x [i x k] x i x k (1.1.37)

where we denote i = x i . The curl of a covariant vector Ai is dened as twice the antisymmetric part of i Ak : i Ak k Ai , and is a tensor. We will also use ,i = x i to denote a partial derivative with respect to xi . Similarly, totally antisymmetrized derivatives of tensors of rank (0, 2) and (0, 3), [i Bkl] and [i Cklm] , are tensors. If Bkl = A[k,l] then [i Bkl] = 0, or conversely, if [i Bkl] = 0 then there exists a vector Ai such that Bkl = A[k,l]. The divergence of a tensor (or density) is a contracted derivative of this tensor (density): i T ...il... jk... . Because of the correspondence between tensors and dual densities, divergences of (totally antisymmetric if more than 1 index) contravariant densities are densities, dual to totally antisymmetrized derivatives of tensors:

= iklm [i Cklm] , k
ik

ik

= iklm [k Blm] , l
k

ikl

= iklm [l Am] .

(1.1.38)

For example, the equations dynamics, are tensorial. References: [1, 2].

,i

and F[ik,l] = 0, that describe Maxwells electro-

1.2
1.2.1

Ane connection
Covariant dierentiation of tensors

An ordinary derivative of a covariant vector Ai is not a tensor, because its coordinate transformation law contains an additional noncovariant term, linear and homogeneous in Ai . Consider the expression Ai;k = Ai,k ilk Al , 9 (1.2.1)

where the quantity ilk (in the second term which is linear and homogeneous in Ai ) transforms such that Ai;k is a tensor: Ai;k = xl xm xl xm (Al,m ln i k Al;m = m An ). x x x i x k (1.2.2)

On the other hand (1.1.36) gives Ai;k = Ai,k i lk Al = so we obtain 2 xn xn l xm xl A + An , A l,m n x k x i x k x i x l i k (1.2.3)

xl xm n 2 xn xn l = + . x l i k x i x k l m x k x i
x j xn

(1.2.4)

Multiplying this equation by i j k

gives the transformation law for ilk :

x j xl xm n x j 2 xn = + . xn x i x k l m xn x k x i

(1.2.5)

The algebraic object ilk , which equips spacetime in order to covariantize a derivative of a vector, is referred to as the ane connection, anity or simply connection. The connection has generally 64 independent components. The tensor Ai;k is the covariant derivative of a vector Ai with respect to xi . We will also use i =;i to denote a covariant derivative. The contracted ane connection transforms according to x i 2 xn xm l i (1.2.6) i k = k l m + n k i . x x x x The ane connection is not a tensor because of the second term on the right-hand side of (1.2.5). A derivative of a scalar is a covariant vector. Therefore a covariant derivative of a scalar is equal to an ordinary derivative: ;i = ,i . (1.2.7)

If we also assume that a covariant derivative of the product of two tensors obeys the same chain rule as an ordinary derivative: (T U );i = T;i U + T U;i , then
l k (Ak B k ),i = (Ak B k );i = Ak;iB k + Ak B k;i = Ak,i B k lk i Ak B + Ak B ;i .

(1.2.8)

(1.2.9)

Therefore we obtain a covariant derivative of a contravariant vector:

l B k;i = B k,i + lk iB .

(1.2.10)

10

The chain rule (1.2.8) also implies that a covariant derivative of a tensor is equal to the sum of the corresponding ordinary derivative of this tensor and terms with the ane connection that covariantize each index:
ij... nj... ij... ij... i j in... n n T ij... kl...;m = T kl...,m + n m T kl... + n m T kl... + . . . k m T nl... l m T kn... . . . . (1.2.11) A covariant derivative of the Kronecker symbol vanishes: k lk;i = jki lj lji j = 0.

(1.2.12)

The second term on the right-hand side of (1.2.5) does not depend on the ane connection, but only on the coordinate transformation. Therefore the dierence between two dierent connections transforms like a tensor of rank (1,2). Consequently, the variation ijk , which is an innitesimal dierence between two connections, is a tensor of rank (1,2). 1.2.2 Parallel transport

Consider two innitesimally separated points in spacetime, P (xi ) and Q(xi + dxi ), and a vector eld A which takes the value Ak at P and Ak + dAk at Q. Because dAk = Ak,idxi and Ak,i is not a tensor, the dierence dAk is not a vector, which is related to subtracting of two vectors at two points with dierent coordinate transformation laws. In order to calculate the covariant dierence between two vectors at two dierent points, we must bring these vectors to the same point. Instead of subtracting from the vector Ak + dAk at Q the vector Ak at P , we must subtract a vector Ak + Ak at Q that corresponds to Ak at P , so the resulting dierence (covariant dierential) DAk = dAk Ak is a vector. The vector Ak + Ak is the parallel-transported or parallel-translated Ak from P to Q. A parallel-transported linear combination of vectors must be equal to the same linear combination of parallel-transported vectors. Therefore Ak is a linear and homogeneous function of Ak . It is also on the order of a dierential, thus a linear and homogeneous function of dxi . The most general form of Ak is l i Ak = lk (1.2.13) i A dx , so
l i k i DAk = dAk + lk i A dx = A ;i dx .

(1.2.14)

k k Because Ak is not a vector, lk i is not a tensor. Because DA is a vector, A ;i is a tensor. The expressions for covariant derivatives of a covariant vector and tensors result from = 0, (T U ) = T U + T U. (1.2.15)

1.2.3

Torsion tensor

The second term on the right-hand side of (1.2.5) is symmetric in the indices i, k so the antisymmetric part of the connection with respect to these indices, S jik = [ijk] , is a tensor: x j xl xm n j S . (1.2.16) S ik = xn x i x k lm 11

This tensor is called the Cartan torsion tensor. The torsion tensor has generally 24 independent components. The contracted torsion tensor, S kik = Si , is the torsion trace vector. 1.2.4 Covariant dierentiation of densities (1.2.17)

A derivative of a scalar density of weight w , , does not transform like a covariant vector density: i = xl xj w xl xl xj w1 xr xj w = i k l + w i k l i l k x x x x x x x s xl xj w xl xj w1 xr x n xm l + w i = i x x k x x k x s xm xl x n xl xj w xj w x n 2 xm l + w = i . (1.2.18) x x k x k xm x n x i

Consider the expression

;i = ,i wi , (1.2.19) where the quantity i transforms such that ;i is a vector density of weight w : ;i =
xl xj x i x k
w

;l =

xl xj x i x k

(,l w l ).

(1.2.20)

On the other hand (1.2.18) gives

=
;i ,i

w i

xl xj = i x x k

xj l + w x k

w x n

2 xm xj w xm x n x i x k

i ,

(1.2.21)

so we obtain the transformation law for i : i x n 2 xm xl = i l + m n i , x x x x

(1.2.22)

which is the same as the transformation law for kki (1.2.6). Therefore the dierence i kki is some covariant vector Vi . If we assume that parallel transport of the product of a scalar density of any weight and a tensor obeys the chain rule: (T ) = T + T, (1.2.23)

so a covariant derivative of such product behaves like an ordinary derivative: (T );i = ;iT + T;i , (1.2.24)

12

then a covariant derivative of a tensor density of weight w is equal to the sum of the corresponding ordinary derivative of this tensor, terms with the ane connection that covariantize each index, and the term with i :
ij... nj... i j in... ij... kl...;m = kl...,m + n m kl... + n m kl... + . . . ij... ij... n knm ij... nl... l m kn... . . . w m kl... .

(1.2.25)

A covariant derivative of the contravariant Levi-Civita density is ijkl;m = ni m njkl + njm inkl + nkm ijnl + nl m ijkn m ijkl . (1.2.26)

In the summations over n only one term does not vanish for each term on the righthand side of (1.2.26), so ijkl;m = ni=i|m n=i|jkl + nj=j |mi|n=j |kl + nk=k|mij |n=k|l + nl=l|mijk|n=l m ijkl

= (nnm m )ijkl = Vm ijkl .

(1.2.27)

The Levi-Civita symbol is a tensor density with constant components, so it does not change under a parallel transport, = 0. Therefore ijkl;m = 0, so Vi = 0 and i = kki . 1.2.5 Covariant derivatives (1.2.28)

Totally antisymmetrized ordinary derivatives of covariant tensors, A[i;k], B[ik;l] and C[ikl;m] , are tensors because of antisymmetrization. Totally antisymmetrized covariant derivatives of tensors are clearly tensors because i is a covariant operation, and are given by direct calculation using the denition of a covariant derivative: A[i;k] = A[i,k] S lik Al , B[ik;l] = B[ik,l] 2S m[ik Bl]m . (1.2.29)

Divergences of (totally antisymmetric if more than 1 index) contravariant densities, ikl i ik ,i , ,i and ,i , are densities because of the correspondence between tensors and dual densities. Covariant divergences of contravariant densities are clearly densities, and are given by direct calculation:
i ;i

i ,i

+ 2 Si i ,

ik

;i

ik

,i

S kil

il

+ 2 Si

ik

(1.2.30)

1.2.6

Partial integration

If the product of two quantities (tensors or densities) T U is a contravariant density k then T U;k d = (T U ); k d T;k Ud = (T U ),k d + 2 Sk T Ud T;k Ud. (1.2.31)

13

The rst term on the right-hand side can be transformed into a hypersurface integral T UdSk . If the region of integration extends to innity and k corresponds to some physical quantity then the boundary integral T UdSk vanishes, giving T U;k d = 2
k If T = i then U = i

Sk T Ud

T;k Ud.

(1.2.32)

and
i ; i d

=2

Si i d.

(1.2.33)

1.2.7

Geodesic frame of reference

Consider a coordinate transformation 1 xk = x k + aklm x l x m , 2 (1.2.34)

where aklm is symmetric in the indices l, m. Substituting this transformation to (1.2.5) and calculating it at xk = x k = 0 gives xi i = k x k and Putting gives
j j i j k = i k + a ik .

(1.2.35)

(1.2.36) (1.2.37) (1.2.38)

aj ik = (ijk) |xl=0 (ijk) = 0.

Therefore there always exists a coordinate frame of reference in which the symmetric part of the connection vanishes locally (at one point). If the ane connection is symmetric in the covariant indices, ijk = kji (the torsion tensor vanishes) then (1.2.38) gives i j (1.2.39) k = 0. The coordinate frame of reference in which the connection vanishes (locally) is referred to as geodesic. 1.2.8 Ane geodesics and four-velocity

Consider a point in spacetime P (xk ) and a vector dxk at this point. Construct a point P (xk + dxk ) and nd the vector d xk which is the parallel-transported dxk from P to P . Then construct a point P (xk + dxk + d xk ) and nd the vector d xk which is the parallel-transported d xk from P to P . The next point is P (xk + dxk + d xk + d xk ) etc. Repeating this step constructs a polygonal line which in the limit dxk 0 k (where is a parameter along the curve) becomes a curve such that the vector dx d 14

tangent to it at any point, when parallely translated to another point on this curve, coincides with the tangent vector there. Such curve is referred to as an autoparallel curve or ane geodesic. Ane geodesics can be attributed with the concept of length, which, for the polygonal curve, is proportional to the number of parallel-transport steps described above. The condition that parallel transport of a tangent vector be a tangent vector is dxi dxi + d d = dxi dxk l dxi d2 xi kil dx = M + d , d d d d2 (1.2.40)

where the proportionality factor M is some function of , or M

l k d2 xi 1 M dxi i dx dx + = , kl d2 d d d d

(1.2.41)

from which it follows that M must dier from 1 by the order of d. In the rst term on the left-hand side of (1.2.41) we can therefore put M = 1, and we denote 1 M by ()d, so l k d2 xi dxi i dx dx + = ( ) . (1.2.42) kl d2 d d d If we replace by a new variable s() then (1.2.42) becomes
l k s s dxi d2 xi i dx dx + = , k l ds2 ds ds s2 ds

(1.2.43)

where the prime denotes dierentiation with respect to . Requiring s s = 0, which has a general solution s = d exp[ (x)dx], brings (1.2.43) into
l k d2 xi i dx dx + = 0, kl ds2 ds ds

(1.2.44)

where the scalar variable s is the ane parameter. The autoparallel equation (1.2.44) is invariant under linear transformations s as + b since the two lower limits of integration in the expression for s() are arbitrary. Dening the four-velocity vector ui = brings (1.2.14) into DAk dAk = Ak;i ui , = Ak,iui , ds ds so dui Dui = + kil uk ul = ui;j uj = 0. ds ds The relations (1.2.46) can be generalized to any tensor density T : DT dT = T;i ui , = T,i ui , ds ds 15 (1.2.46) (1.2.47) dxi ds (1.2.45)

(1.2.48)

| is a parallel translation of dx | . Because ds is a scalar, it is invariThe vector dx ds Q ds P ant under parallel transport, ds|Q = ds|P . Therefore the vector dxi |Q is a parallel translation of dxi |P , so ds measures the length of an innitesimal section of an ane geodesic. Only the symmetric part (ki l) of the connection enters the autoparallel equation k dxl (1.2.44) because of the symmetry of dx with respect to the indices k, l; ane ds ds geodesics do not depend on torsion. At any point, a coordinate transformation to the geodesic frame (1.2.34) brings all the components (ki l) to zero, so the autopari = 0. The autoparallel equation is also invariant under a allel equation becomes du ds projective transformation i kil kil + k Al , (1.2.49) where Ai is an arbitrary vector. Substituting this transformation to (1.2.47) gives dui + kil uk ul = ui uk Ak . ds If we replace s by a new variable s (s) then (1.2.50) becomes dU i uk Ak s + s dxi + kil U k U l = , ds s 2 ds where Ui = (1.2.51) (1.2.50)

dxi (1.2.52) ds and the prime denotes dierentiation with respect to s. Requiring uk Ak s + s = 0, which has a general solution s = s ds exp[ s Ak uk (x)dx], brings (1.2.51) into dU i + kil U k U l = 0. ds (1.2.53)

1.2.9

Innitesimal coordinate transformations

Consider a coordinate transformation x i = xi + i,

(1.2.54)

where i = xi is an innitesimal vector (variation of xi ). For a tensor or density T dene T = T (x i ) T (xi ), = T (xi ) T (xi ) = T k T,k . T For a scalar we nd For a covariant vector = k ,k . = 0, xk Ak Ai k,i Ak , x i i k Ak k Ai,k . A ,i Ai = 16

(1.2.55) (1.2.56) (1.2.57)

(1.2.58) (1.2.59)

The variation (1.2.58) is not a tensor, but (1.2.59) is: i = k Ak k Ai;k 2S j k Aj . A ik ;i as a Lie derivative of T , L T . For a contravariant vector We refer to T x i k B B i i,k B k , B = xk i i B k k B i = i B k k B i + 2S i k B j . B ,k ;k jk ,k ;k
i

(1.2.60)

(1.2.61) (1.2.62)

For a scalar density xi i = i 1 ,i , x i k ,k = i k ;k + 2Si i. ,i ;i (1.2.63) (1.2.64)

The chain rule for implies that for a tensor density of weight w (which includes tensors as densities of weight 0)
mj... j ij... ij... i m im... m ij... kl... ,m kl... + ,m kl... + . . . ,k ml... ,l km... . . .

w m,m ij... (1.2.65) kl... , ij... i mj... + j;m im... + . . . m ij... m ij... . . . kl... ;m kl... kl... ;k ml... ;l km...
j m ij... i m nj... m in... w m;m ij... kl... kl...;m + 2S nm kl... + 2S nm kl... + . . .

n m ij... m ij... 2S nkm m ij... nl... 2S lm kn... . . . + 2wSm kl... .

(1.2.66)

A Lie derivative of a tensor density of rank (k, l) and weight w is a tensor density of rank (k, l) and weight w . The formula for a covariant derivative of T can be written as i T, T;k = T,k + ijk C j is an operator acting on tensor densities: where C i = 0, C i Ak = i Aj , C i Bk = k Bi, C i = i , C j j k j j j j or generally m ij... = i mj... + j im... + . . . m ij... m ij... . . . w m ij... . C n n l n kl... k n kl... kl... nl... kn... kl... (1.2.69) Such dened operator also enters the formula for T : kT i . T = C i ,k (1.2.70) (1.2.68) (1.2.67)

17

1.2.10

Killing vectors (i;k) = 0 (1.2.71)

A vector i that satises is referred to as a Killing vector. Along an ane geodesic D i (u i) = uk (uii );k = uiuk i;k + iuk ui;k = 0. ds (1.2.72)

The rst term in the sum in (1.2.72) vanishes because of the denition of i and the second term vanishes because of the ane geodesic equation. Therefore, to each Killing vector i there corresponds a quantity ui i which is constant along an ane geodesic. References: [1, 2, 3].

1.3
1.3.1

Curvature
Curvature tensor

The commutator of covariant derivatives of a contravariant vector is a tensor:

l i i l [j , k ]B i = 2[j k] B i = 2[j k] B i 2[k j ] l B + 2l [j k ] B i i l m l i i m l i = 2([j |m | k ] + l [j |m| k ] )B + 2S jk l B = R mjk B + 2S jk l B , (1.3.1) i m l i i l i l m = 2[j (|m | k ] B ) + 2S jk l B + 2l [j k ] B + 2l [j |m| k ] B

where || an index which is excluded from symmetrization or antisymmetrization. Therefore Rimjk , dened as
i i i l i l Rimjk = j m k k m j + l j m k l k m j ,

(1.3.2)

is a tensor, referred to as the curvature tensor. The curvature tensor Rimjk is antisymmetric in the indices j, k and has generally 96 independent components. The commutator of covariant derivatives of a covariant vector is [j , k ]Ai = Rmijk Am + 2S ljk l Ai , and the commutator of covariant derivatives of a tensor is
i mn... n im... m in... m in... [j , k ]T in... lp... = R mjk T lp... + R mjk T lp... + . . . R ljk T mp... R pjk T lm...

(1.3.3)

. . . + 2S ljk l T in... lp... .

(1.3.4)

A change in the connection where T ijk is a tensor, results in the following change of the curvature tensor: Riklm Riklm + T ikm;l T ikl;m + T jkm T ijl T jkl T ijm . 18 (1.3.6) jik jik + T ijk , (1.3.5)

i For a projective transformation (1.2.49), T ijk = j Ak , so i Riklm Riklm + k (Am;l Al;m ).

(1.3.7)

The variation of the curvature tensor is Riklm = ( kim ),l ( kil ),m + jil kjm + jil kjm jim kjl jim kjl ljm kij + jil kjm + jil kjm jim kjl jim kjl Integrability of connection

j j j i i i i = ( kim );l jil kjm + kjl jim + m l k j ( k l );m + j m k l k m j l

= ( kim );l ( kil );m 2S nlm kin . 1.3.2

(1.3.8)

The ane connection is integrable if parallel transport of a vector from point P to point Q is independent of a path along which this vector is parallelly translated, or equivalently, parallel transport of a vector around a closed curve does not change this vector. For an integrable connection, we can uniquely translate parallelly a given vector hi at point P to all points in spacetime: hi = dhi , or hi,k = jik hj . Therefore
j j m i j i m i j (jik hj ),l (jil hj ),k = jik,l hj jik m l h j l,k h + j l m k h = R jlk h = 0, (1.3.11) so, because hi is arbitrary, (1.3.12) Riklm = 0.

(1.3.9) (1.3.10)

Spacetime with a vanishing curvature tensor Riklm = 0 is at. Consider 4 linearly i independent vectors hi a , where a is 1,2,3,4, and vectors inverse to ha :
i hi a hka = k . a

(1.3.13)

If the ane connection is integrable then (1.3.10) becomes

i l hi a,k = l k ha .

(1.3.14)

Multiplying (1.3.14) by hja gives

i jik = hja hi a,k = hja,k ha .

(1.3.15)

An integrable connection has thus 16 independent components. If the connection is also symmetric, S ijk = 0, then hja,k hka,j = 0, 19 (1.3.16)

which is the condition for the independence of the coordinates ya =

Q P

hia dxi

(1.3.17)

of the path of integration P Q. Adopting ya as the new coordinates (with point P = (0, 0, 0, 0) in the center) gives xi ya = hia , = hi a, i x ya so (1.3.15) becomes jik (xi ) = xi 2 ya . ya xk xj (1.3.19) (1.3.18)

The transformation law for the connection (1.2.5) gives (with ya corresponding to xj ) jik (ya ) = 0. (1.3.20)

A torsionless integrable connection can be thus transformed to zero; one can always nd a system of coordinates which is geodesic everywhere. If a connection is symmetric but nonintegrable then a geodesic frame of reference can be constructed only at a given point (or along a given world line). 1.3.3 Parallel transport along closed curve

Consider parallel transport of a covariant vector around an innitesimal closed curve. Such transport changes this vector by Ak = Ak = kil Ai dxl = 1 2 (kim Ai ) (kil Ai ) df lm xl xm

1 kim kil i n lm Ai + (kim in l k l i m )An df 2 xl xm 1 = Riklm Ai f lm , 2

(1.3.21)

where we use Stokes theorem (1.1.33) and Ak,l = kil Ai which is valid along the curve and thus is approximately valid (to terms of rst order in f lm ) inside this curve. The change of a contravariant vector due to parallel transport around an innitesimal closed curve results from (Ak B k ) = 0: 1 B k = Rkilm B i f lm , 2 (1.3.22)

and the corresponding change of a tensor results from the chain rule for parallel transport: 1 i jk... ij... j j k ik... ik... lm T ik... np... = (R jlm T np... + R jlm T np... + . . . R nlm T jp... R plm T nj... . . .)f . 2 (1.3.23) 20

1.3.4

Bianchi identities 1 j [k l] B i = j (Rimkl B m ) + j (S mkl m B i ) 2 (1.3.24)

Consider and

1 1 1 [j k] l B i = Rmljk m B i + Rimjk l B m + S mjk m l B i = Rmljk m B i 2 2 2 1 i (1.3.25) + R mjk l B m + S mjk l m B i + S mjk Rinml B n + 2S mjk S nml n B i . 2 Total antisymmetrization of the indices j, k, l in (1.3.24) and (1.3.25) gives 1 1 [j k l] B i = [j Ri|m|kl] B m + Rim[kl] j ] B m + [j S mkl] m B i + S mkl j ] m B i 2 2 (1.3.26) and 1 1 [j k l] B i = Rm[ljk]m B i + Rim[jk l] B m + S m[jk l] m B i 2 2 +S m[jk Ri|nm|l] B n + 2S m[jk S n|m|l] n B i , (1.3.27) so 1 1 [j Ri|m|kl]B m + [j S mkl] m B i = Rm[ljk]m B i + S m[jk Ri|nm|l] B n 2 2 m n i +2S [jk S |m|l] n B . (1.3.28) Comparing terms in (1.3.28) with B i gives the rst Bianchi identity or simply Bianchi identity: Rin[jk;l] = 2Rinm[j S mkl] , (1.3.29) while comparing terms with k B i gives the second Bianchi identity or cyclic identity: Rm[jkl] = 2S m[jk;l] + 4S mn[j S nkl] . For a symmetric connection, S ijk = 0, these identities reduce to Rin[jk;l] = 0, Rm[jkl] = 0. (1.3.31) (1.3.32) (1.3.30)

The cyclic identity (1.3.32) imposes 16 constraints on the curvature tensor, so the curvature tensor with a vanishing torsion has 80 independent components. 1.3.5 Ricci tensor

Contraction of the curvature tensor with respect to the contravariant index and the second covariant index gives the Ricci tensor: Rik = Rj ijk = ijk,j ijj,k + ilk ljj ilj ljk . 21 (1.3.33)

Contraction of the curvature tensor with respect to the contravariant index and the third covariant index gives the Ricci tensor with the opposite sign due to the antisymmetry of the curvature tensor with respect to its last indices. Contraction of the curvature tensor with respect to the contravariant index and the rst covariant index gives the homothetic or segmental curvature tensor: Qik = Rj jik = jjk,i jji,k , (1.3.34)

which is a curl. A change in the connection (1.3.5) results in the following changes of the Ricci tensor and segmental curvature tensor: Qik Qik + T jjk,i T jji,k . For a projective transformation (1.2.49) Rik Rik + Ak;i Ai;k , Qik Qik + 4(Ak,i Ai,k ). (1.3.37) (1.3.38) Rik Rik + T lik;l T lil;k + T jik T ljl T jil T ljk , (1.3.35) (1.3.36)

Therefore the symmetric part of the Ricci tensor is invariant under projective transformations. The variation of the Ricci tensor is Rik = ( ilk );l ( ill );k 2S jlk ilj , while the variation of the segmental curvature tensor is Qik = ( jjk ),i ( jji ),k . 1.3.6 Geodesic deviation (1.3.40) (1.3.39)

Consider a family of ane geodesics characterized by the ane parameter s and distinguished by a scalar parameter t. Dene the separation vector vi = so v i;k uk ui;k v k = v i,k uk ui,k v k 2S ikl uk v l = Therefore D2vi = (v i;j uj );k uk = (ui;j v j );k uk 2(S ikl uk v l );j uj 2 ds = ui;jk v j uk + ui;j v j;k uk 2(S ikl uk v l );j uj dui dv i 2S ikl uk v l = 2S ikl uk v l . dt ds (1.3.42) dxi , dt (1.3.41)

2(S ikl uk v l );j uj = (ui;k uk );j v j + Rijkl uj uk v l 2(S ikl uk v l );j uj D = Rijkl uj uk v l 2 (S ikl uk v l ) ds 22

= ui;kj v j uk Riljk ul v j uk 2S ljk ui;l v j uk + ui;j (uj ;k v k 2S jkl uk v l )

= ui;kj v j uk Riljk ul v j uk 2S ljk ui;l v j uk + ui;j v j;k uk 2(S ikl uk v l );j uj

(1.3.43)

D Dv i (1.3.44) + 2S ikl uk v l = Rijkl uj uk v l . ds ds This is the equation of geodesic deviation. If we replace ane geodesics by arbitrary curves then ui;k uk = 0 and (1.3.44) becomes D Dv i + 2S ikl uk v l = Rijkl uj uk v l + (ui;k uk );j v j . ds ds References: [1, 2, 3, 4]. (1.3.45)

or

1.4
1.4.1

Metric
Metric tensor

An ane parameter s is a measure of the length only along an ane geodesic. In order to extend the concept of length to all points in spacetime, we equip spacetime with an algebraic object gik , referred to as the covariant metric tensor and dened as ds2 = gik dxi dxk . The metric tensor is a symmetric tensor of rank (0,2): gik = gki. (1.4.2) (1.4.1)

The ane parameter s, whose dierential is given by (1.4.1), is referred to as the interval. Because ds does not change under parallel transport along an ane geodesic from point P (xi ) to point Q(xi + dxi ), ds|Q = ds|P , and dxi |Q is a parallel translation of dxi |P , gik |Q = gik |P + gik,j dxj is a parallel translation of gik |P : gik |Q = gik |P + gik , so Therefore a covariant derivative of the covariant metric tensor vanishes: Njik = gik;j = 0 or where Nijk is the nonmetricity tensor. The symmetric contravariant metric tensor g ik = g ki is dened as the inverse of gik :
k . gij g ik = j

(1.4.3) (1.4.4)

Dgik = gik;j dxj = dgik gik = gik,j dxj gik = 0.

(1.4.5) (1.4.6)

gik,j ilj glk kl j gil = 0,

(1.4.7)

A covariant derivative of the contravariant metric tensor also vanishes: g ik;j = 0. 23 (1.4.8)

The metric tensor allows to associate covariant and contravariant vectors: Ai = g ik Ak , Bi = gik B k , (1.4.9) (1.4.10)

because such association works for the covariant dierentials of these vectors which are vectors: DAi = D (g ik Ak ) = g ik DAi , DBi = D (gik B k ) = gik DB k (1.4.11)

(raising and lowering of coordinate indices commutes with covariant dierentiation with respect to ). For covariant and contravariant indices of tensors and densities this association is
j... gim ij... kl... = m kl... , ijm... g km ij... kl... = l... .

(1.4.12) (1.4.13)

The square root of the absolute value of the determinant = |gik | (1.4.14)

of the metric tensor is a scalar density, which we can use to multiply covariant integrands that contain dual densities of weight -1, since eiklm = | |iklm , eiklm = 1 | | iklm (1.4.15)

are tensors. Thus the relations (1.1.27) are also valid if we replace by e. The variation of the determinant of the metric tensor is = g ik gik = gik g ik . A covariant derivative of the determinant of the metric tensor vanishes:
;j

(1.4.16)

= 0.

(1.4.17)

A Lie derivative of the metric tensor is L g ik = 2 (i;k) 4S where

;i (ik ) l l ,

(1.4.18)

=;k g ik . The four-velocity vector (1.2.45) is normalized due to (1.4.1): u i u i = 1, (1.4.19)

thus having 3 independent components. The commutator of covariant derivatives (1.3.4) of the metric tensor gives R
(ij ) kl

= N[k

ij ;l ]

S mkl Nm ij = N[k 24

ij ,l] ,

(1.4.20)

so the segmental curvature tensor (1.3.34) is Qkl = N[k

ij ,l] gij .

(1.4.21)

Because the nonmetricity tensor (1.4.5) vanishes, the curvature tensor is antisymmetric in its rst two indices: Rijkl = Rjikl . (1.4.22) Thus the segmental curvature tensor also vanishes, and Rijkl g jl = Rik , (1.4.23)

so there is only one independent way to contract the curvature tensor, which gives the Ricci tensor up to the sign. 1.4.2 Christoel symbols

The condition (1.4.5) is referred to as metricity or metric compatibility of the ane connection, and imposes 40 constraints on the connection: gik;j + gkj ;i gji;k = gik,j ilj glk kl j gil + gkj,i kl i glj jli gkl gji,k + jlk gli +ilk gjl = gik,j + gkj,i gji,k 2(il j ) gkl 2S lkj gil 2S lkigjl = 0. Multiplying (1.4.24) by g km gives
m (im j ) = {i j } + 2S(ij ) , m

(1.4.24)

(1.4.25)

1 mk (gki,j + gkj,i gij,k ) {im j} = g 2 are the Christoel symbols, symmetric in their covariant indices:
k {ik j } = {j i }. k k Because ik j = (i j ) + S ij , the metric-compatible ane connection equals k k ik j = {i j } + C ij ,

where

(1.4.26)

(1.4.27)

(1.4.28) (1.4.29)

where

C kij = 2S(ij ) + S kij is the contortion tensor, antisymmetric in its rst two indices: Cijk = Cjik . The inverse relation between the torsion and contortion tensor is S ijk = C i[jk] . 25

(1.4.30)

(1.4.31)

The dierence between two ane connections is a tensor, so the sum of a connection and a tensor of rank (1,2) is a connection. Therefore the Christoel symbols form a connection, referred to as the Levi-Civita connection. Dene the covariant derivative with respect to the Levi-Civita connection analogously to (1.2.11), with {} k ik j replaced by {i j }, and denote it :i instead of ;i , or i instead of i . A covariant derivative with respect to the Levi-Civita connection of the metric tensor vanishes due to the denition of the Christoel symbols: gik:j = gik,j {ilj }glk {kl j }gil = 0, (1.4.32)

which gives the inverse relation between ordinary derivatives of the metric tensor and the Christoel symbols. The variation of the Levi-Civita connection is a tensor: 1 kl {ik j } = g ((gli ):j + (glj ):i (gij ):l ). 2 (1.4.33)

The covariant derivative over s of a tensor density with respect to the Levi-Civita connection is, analogously to (1.2.48), D {} T = T:i ui . ds One can show that the following formulae hold: {kki } = (ln | |),i , 1 ij {ik ( | |g ik ),i , j }g = | | 1 B i:i = ( | |B i ),i , | | 1 F ik:i = ( | |F ik ),i , | | B i | |dSi = B i:i | |d, (1.4.35) (1.4.36) (1.4.37) (1.4.38) (1.4.39) (1.4.40) (1.4.34)

Ai:k Ak:i = Ai,k Ak,i ,

where F ik = F ki. The Christoel symbols satisfy all formulae that are satised by i ik j in which S jk = 0. Because the Levi-Civita connection is a symmetric connection, it can be brought to zero by transforming the coordinates to a geodesic frame. In a geodesic frame, the covariant derivative with respect to the Levi-Civita connection, {} i , coincides with the ordinary derivative i . A Lie derivative of the metric tensor (1.4.18) can be written as L g ik = 2 (i:k) , L gik = 2(i:k) , where
:i

(1.4.41)

=:k g ik . A Killing vector (1.2.71) for the Levi-Civita connection satises (i:k) = 0, 26 (1.4.42)

thus becomes a generator of isometries, transformations that do not change the metric tensor. If the nonmetricity tensor does not vanish, the general formula for the ane connection (1.4.28) is 1 k k k k ik j = {i j } + C ij N ij + N(i j ) . 2 1.4.3 Riemann curvature tensor (1.4.43)

The curvature tensor constructed from the Levi-Civita connection is referred to as the Riemann tensor:
i i i l i l P imjk = j {m k } k {m j } + {l j }{m k } {l k }{m j }.

(1.4.44)

The commutator of covariant derivatives of the metric tensor vanishes: [j , k ]glp = P mljk gmp P mpjk glm = 0,
{} {}

(1.4.45)

so the covariant Riemann tensor Pimjk is also antisymmetric in the indices i, m. Substituting (1.4.26) in (1.4.44) gives 1 Piklm = (gim,kl + gkl,im gil,km gkm,il ) + gjn ({ijm }{knl } {ijl }{knm }), 2 (1.4.46)

which explicitly shows the following symmetry and antisymmetry properties: Piklm = Pikml , Piklm = Pkilm , Piklm = Plmik . Accordingly, the Riemannian Ricci tensor is symmetric: Pik = P jijk = Pki. (1.4.50) (1.4.47) (1.4.48) (1.4.49)

Substituting (1.4.28) in (1.3.5) and (1.3.6) gives the relation between the curvature and Riemann tensors: Riklm = P iklm + C ikm:l C ikl:m + C jkm C ijl C jkl C ijm . Contracting (1.4.51) with respect to in the indices i, l gives Rkm = Pkm + C ikm:i C iki:m + C jkm C iji C jki C ijm . Consequently, the Ricci or curvature scalar, R = Rik g ik , is given by R = P g ik (2C lil:k + C jij C lkl C lim C mkl ), 27 (1.4.53) (1.4.54) (1.4.52) (1.4.51)

where P is the Riemannian curvature scalar, P = Pik g ik . The variation of the Riemann tensor is, analogously to (1.3.8), P iklm = ( {kim }):l ( {kil }):m , and the variation of the Riemannian Ricci tensor is Pik = ( {ilk }):l ( {ill }):k . (1.4.57) (1.4.56) (1.4.55)

The Bianchi identities (1.3.29) and (1.3.30) contracted with respect to one contravariant and one covariant index give Rin[ik;l] = 2Rinm[i S mkl] , Rk[jkl] = 2S k[jk;l] + 4S kn[j S nkl] . Contracting these equations with the metric tensor gives Rnk;l Rnl;k + Rinkl;i = 2Rnm S mkl 2Rinmk S mil + 2Rinml S mik and the contracted cyclic identity: Rjl Rlj = 2Sj ;l + 2Sl;j 2S klj ;k + 4Sn S nlj . (1.4.61) (1.4.60) (1.4.58) (1.4.59)

Further contraction of (1.4.60) with the metric tensor gives the contracted Bianchi identity: 1 (1.4.62) Ril;i R;l = 2Rkm S mkl Rikml S mik . 2 The Bianchi identities (1.3.31) and (1.3.32) for the Riemann tensor are P in[jk:l] = 0, P m[jkl] = 0. Contracting these equations with the metric tensor gives Pnk:l + P inkl:i Pnl:k = 0, Pjl Plj = 0, (1.4.65) (1.4.66) (1.4.63) (1.4.64)

in agreement with (1.4.50). Further contraction of (1.4.65) with the metric tensor gives the covariant conservation, Gik:i = 0, (1.4.67) of the symmetric Einstein tensor, 1 Gik = Pik P gik . 2 28 (1.4.68)

1.4.4

Properties of Riemann tensor

In two dimensions there is only 1 independent component of the Riemann tensor, P1212 . The Riemann scalar is 2P1212 , (1.4.69) P = where is the determinant of the two-dimensional metric tensor ik :
2 = |ik | = 11 22 12 .

(1.4.70)

A surface near point x = 0, y = 0 is given by z= x2 y2 + , 21 22 (1.4.71)

where 1 and 2 are the radii of curvature. Substituting (1.4.71) to dl2 = dx2 + dy 2 + dz 2 = ik dxi dxk gives ik (x, y ), which then gives P 2 =K=
x=y =0

(1.4.72)

1 , 1 2

(1.4.73)

where K is the Gaucurvature. In three dimensions there are 3 independent pairs, 12, 23, and 31, so the Riemann 2 tensor has 6 independent components: 3 with identical pairs and 32 = 3 with dierent pairs (the cyclic identity does not reduce the number of independent components). The Ricci tensor has also 6 components, which are related to the components of the Riemann tensor by P = P P + P P + g = diag(1, 1, 1), P ( ). 2 (1.4.74)

Choosing the Cartesian coordinates at a given point, dened by the condition (1.4.75)

and diagonalizing P , which is equivalent to 3 rotations, brings P to the canonical form with 6 3 = 3 independent components. Consequently, the Riemann tensor in three dimensions has 3 physically independent components. The Gaucurvature of a surface perpendicular to the x3 axis is given by K= P1212 . 2 11 22 12 (1.4.76)

In four dimensions there are 6 independent pairs, 01, 02, 03, 12, 23, and 31, so 5 = 15 with dierent pairs. The there are 6 components with identical pairs and 62 cyclic identity reduces the number of independent components by 1, so the Riemann tensor in four dimensions has generally 20 independent components. Choosing the Cartesian coordinates at a given point and applying 6 rotations brings Pijkl to the canonical form with 20 6 = 14 physically independent components. 29

1.4.5

Weyl tensor

In four dimensions the Weyl tensor is dened as 1 1 Wiklm = Piklm (Pil gkm + Pkm gil Pim gkl Pkl gim ) + P (gil gkm gim gkl ). (1.4.77) 2 6 This tensor has all the symmetry and antisymmetry properties of the Riemann tensor, and is also traceless (any contraction of the Weyl tensor vanishes). 1.4.6 Metric geodesics

Consider two points in spacetime, P and Q. Among curves that connect these points, one curve has the minimal value of the interval s = ds, and is referred to as a metric geodesic. The equation of a metric geodesic is given by the condition that ds be an extremum with the endpoints of the curve xed: + 1 2 ds = (gik dxi dxk )1/2 = dxi gij dxj 1 + ds 2 duixi + gij dxi dxj = ds gij,k xk uiuj ds (1.4.78) gij uj dxi

gij,k xk uiuj ds = dui i 1 x ds + ds 2

d(uixi )

1 2

gjk,ixi uj uk ds = 0,

where we omit the total dierential term Since xi is arbitrary, we obtain

d(uixi ) because xi = 0 at the endpoints. gjk,iuj uk ds (1.4.79)

duj d 1 1 gjk,iuj uk ds = gij (gij uj ) + uk gij,k uj ds 2 ds 2 duj j k + {jm = gij k }gim u u = 0 ds or, after multiplying (1.4.79) by g il : D {} ul dul = + {jlk }uj uk = 0. ds ds The metric geodesic equation (1.4.80) can be written as

(1.4.80)

k l d2 xi i dx dx + { } = 0. (1.4.81) kl ds2 ds ds Using (1.4.28) and (1.4.29), the ane geodesic equation (1.2.44) can be written as k l l k d2 xi i dx dx i dx dx + { } + 2 S = 0. (1.4.82) kl kl ds2 ds ds ds ds If the torsion tensor is completely antisymmetric then the last term in (1.4.82) vanishes and the ane geodesic equation coincides with the metric geodesic equation. The equation of geodesic deviation with respect to the Levi-Civita connection is, analogously to (1.3.44), D {}2 v i = P ijkl uj uk v l . (1.4.83) 2 ds

30

1.4.7

Galilean frame of reference and Minkowski tensor

At a given point, the nondegenerate ( = 0) metric tensor can be brought to a diagonal (canonical) form gik = diag(1, 1, 1, 1). Physical systems are described by the metric tensor with < 0. Without loss of generality, we assume that the canonical form of the metric tensor is gik = ik = diag(1, 1, 1, 1), g ik = ik = diag(1, 1, 1, 1). (1.4.84)

A frame of reference in which gik has the canonical form is referred to as Galilean. The transformation (1.2.34) with (1.2.37) brings a symmetric ane connection, thus the Christoel symbols, to zero at a given point without changing the components of the metric tensor because of (1.2.35). Therefore a frame of reference can be both geodesic and Galilean. In such locally inertial frame rst derivatives of the metric tensor vanish because of (1.4.32). The corresponding metric tensor (1.4.84) is referred to as the Minkowski tensor. In a locally inertial frame the coordinates xi , not only the dierentials dxi , are components of a contravariant vector. In the absence of torsion, spacetime with a vanishing Riemann tensor P iklm = 0 is at. In the new coordinates ya (1.3.17), (1.3.18) gives g ab (y ) = gik (x) xi xk = gik (x)hia hkb = ab . ya yb (1.4.85)

Therefore in a at spacetime without torsion one can always nd a system of coordinates which is Galilean everywhere. 1.4.8 Intervals, proper time and distances

The form of the Minkowski tensor distinguishes the coordinate x0 from the rest of the coordinates x , where the index can be 1,2,3. The temporal coordinate x0 = ct, where t is referred to as time and c is referred to as the velocity of propagation of interaction. The coordinates x are spatial and span space. The set of 4 coordinates xi describe an event and span spacetime. The curve xi (), where is a parameter, is referred to as a world line of a given point. The quantitites v = dx dt (1.4.86)

where dxi are innitesimal coordinate dierences between the two points. The interval between two nitely separated points is s2 = ik xi xk = c2 t2 x x , 31

are the components of a three-dimensional vector, the velocity of this point. An innitesimal interval ds is timelike if ds2 > 0, spacelike if ds2 < 0, and null if ds2 = 0. In the Galilean frame, the interval between two innitesimally separated points (events) is ds2 = ik dxi dxk = c2 dt2 dx dx , (1.4.87)

(1.4.88)

where xi are nite coordinate dierences between the two points. If s is timelike, one can always nd a frame of reference in which the two events occur at the same place, x = 0. A frame of reference in which dx = 0 describes a point at rest and is referred to as the rest frame. In this frame t = , ds2 = c2 d 2 , (1.4.89)

where is the proper time. If dx = 0 along a world line then the point moves. The proper time for a moving point is equal to the time measured by a clock moving with this point. If s is spacelike, one can always nd a frame of reference in which the two events occur at the same time (are synchronous), x0 = 0. If ds = 0 along a world line, this world line describes the propagation of a signal (interaction), with v = (v v )1/2 = c. Equations (1.4.87) and (1.4.89) give 1 d 2 = dt2 dx dx , c (1.4.90)

so the proper time goes more slowly than the coordinate time t. In the rest frame dx = 0 gives u = 0. At each point in space, the condition dx = 0 gives the relation between the proper time and the coordinate time: d = which requires g00 0. The relation (1.4.19) gives u0 = (g00 )1/2 . (1.4.93) The distance between two innitesimally separated points cannot be obtained by imposing dx0 because x0 transforms dierently at these points. Instead, consider a 0 signal that leaves point B (x + dx ) at x0 + dx0 , reaching point A(x ) at x and coming back to point B at x0 + dx0 + . Therefore ds2 = g00 (dx0 )2 + 2g0 dx0 dx + g dx dx = 0 gives dx0 = 1 (g0 dx g00 (1.4.94) (1.4.92) 1 g00 dx0 , c (1.4.91)

(g0 g0 g00 g )dx dx ).

(1.4.95)

The dierence in the time coordinate between emitting and receiving the signal at 0 g00 /c, and the distance point B is equal to the dierence between dx0 + and dx times dl between points A and B is equal to this dierence times c/2: dl2 = dx dx , where = g + 32 g0 g0 g00 (1.4.96) (1.4.97)

is the symmetric spatial metric tensor of spacetime. This tensor is used to raise and lower spatial indices of quantities written in a three-dimensional (spatial) form: A = A , B = B , where is the inverse of :
= ,

(1.4.98) (1.4.99) (1.4.100)

One can show that the following formulae hold: = g , = g00 , g = g 0 , 1 g g , g 00 = g00 where (1.4.101) (1.4.102) (1.4.103) (1.4.104)

= det ,
g =

(1.4.105) (1.4.106)

g0 . g00

The components g form a three-dimensional vector g. The event at point A at x0 is synchronized with the event at point B at the arithmetic mean of the time coordinates of emitting and receiving the signal, i.e. at 1 0 0 x0 + (dx0 + dx+ ) = x + g dx . 2 Therefore x0 = g x , (1.4.108) which is equivalent to x0 = 0, is the dierence in x0 between two synchronized innitesimally separated points. 1.4.9 Spatial vectors (1.4.107)

The spatial components of a contravariant vector Ai form a three-dimensional vector A: Ai = (A0 , A ) = (A0 , A). (1.4.109) The spatial components of a covariant-vector operator i form a spatial gradient operator grad = : = i = , , . (1.4.110) ct cx ct The scalar product of two spatial vectors is A B = A B . 33 (1.4.111)

The square of a spatial vector A is A2 = A A. In three-dimensional space, the permutation symbol is dened as = 0 . The three-dimensional equivalent of (1.4.15) is e = 1 (1.4.113) (1.4.112)

, e = .

(1.4.114)

The cross product of two three-dimensional vectors A and B, C = A B is C = e A B , C = e A B . (1.4.115)

The three-dimensional divergence of a spatial vector A is, in analogy to (1.4.37, 1 divA = A = ( A ), .

(1.4.116)

The three-dimensional curl of a spatial vector is (curlA) = ( A) = e A, . The Laplacian operator is the divergence of the gradient, = 2 = . (1.4.118) (1.4.117)

In a locally galilean frame of reference, the covariant and contravariant three-dimensional components of a vector are identical, because = , where is the Cartesian metric tensor, = diag(1, 1, 1), = diag(1, 1, 1). (1.4.120) (1.4.119)

In this frame we refer to the coordinates x1 , x2 , x3 , which are Cartesian, as x, y, z . The permutation symbol (1.4.113) satises = ,

(1.4.121) (1.4.122) (1.4.123)

= 2 , = 6.

34

One can show that the following formulae hold: A B = B A, curl grad = 0, div curl A = 0, grad( ) = grad + grad, grad(A B) = (A )B + (B )A + A curl B +B curl A, div(A) = grad A + div A, curl(A) = grad A + curl A, div(A B) = B curl A A curl B, curl(A B) = (B )A (A )B + A div B B div A, curl curl A = grad div A A, where (A )B = A B. References: [1, 2, 3, 4]. (1.4.134) (1.4.124) (1.4.125) (1.4.126) (1.4.127) (1.4.128) (1.4.129) (1.4.130) (1.4.131) (1.4.132) (1.4.133)

1.5
1.5.1

Tetrad and spin connection

Tetrad

In addition to the coordinate systems, at each spacetime point we set up four linearly independent vectors ei a such that ei a eib = ab , (1.5.1)

where a, b = 0, 1, 2, 3 are Lorentz indices and ab = diag(1, 1, 1, 1) is the coordinateinvariant Minkowski metric tensor in a locally geodesic frame of reference at this point. This set of four vectors is referred to as a tetrad. The inverse tetrad eai satises
b b ei a ei = a , i a ei a ek = k .

(1.5.2) (1.5.3)

The coordinate metric tensor gik is related to the Minkowski metric tensor through the tetrad: b gik = ea (1.5.4) i ek ab . Accordingly, the determinant of the tetrad = |ea i | by of the metric tensor gik is related to the determinant | |= . (1.5.5)

35

Any vector V can be specied by its components V i with respect to the coordinate system or by the coordinate-invariant projections V a of the vector onto the tetrad eld:
i i V a = ea i V , Va = ea Vi , a a V i = ei a V , Vi = ei Va ,

(1.5.6) (1.5.7)

and similarly for tensors and densities with more indices. We can use ab and its inverse ab to lower and raise Lorentz indices, as we use gik and its inverse g ik to lower and raise coordinate indices. 1.5.2 Lorentz transformation

The relation (1.5.4) imposes 10 constraints on the 16 components of the tetrad, leaving 6 components arbitrary. If we change from one tetrad ei i a to another, e b , then the vectors of the new tetrad are linear combinations of the vectors of the old tetrad:
b i e i a = a eb .

(1.5.8)

The relation (1.5.4) applied to the tetrad eld e i b, gik = e a b ie k ab , imposes on the matrix ba the orthogonality condition: ca db cd = ab . (1.5.10) (1.5.9)

We refer to ba as a Lorentz matrix, and to a transformation of form (1.5.8) as the Lorentz transformation. 1.5.3 Tetrad transport

A natural choice for the zeroth component of a tetrad at a given point is

i ei 0 = u .

(1.5.11)

Along a world line this tetrad should be transported such that the zeroth component always coincides with the four-velocity. The Fermi-Walker transport of a tetrad is dened as Duj Dui j ei a = ui ej + e uj . (1.5.12) a ds ds ds a Putting a = 0 in (1.5.12) gives u i Dui = , (1.5.13) ds ds so the Fermi-Walker transport of the four-velocity is equivalent to its covariant change and thus (1.5.11) is valid at all points. This transport does not change the orthogonality relation for tetrads (1.5.1) since (1.5.12) gives i (e eib ) = 0. ds a 36 (1.5.14)

1.5.4 Dene

Spin connection
i i j iak = ei a;k = ea,k + j k ea .

(1.5.15) (1.5.16)

The quantities

j abi = ea j bi

transform like vectors under coordinate transformations. We can extend the notion of covariant dierentiation to quantities with Lorentz coordinate-invariant indices by regarding abi as a connection, referred to as Lorentz or spin connection. For a contravariant Lorentz vector
a a b Va |i = V ,i + bi V ,

(1.5.17)

where |i is a covariant derivative of such a quantity with respect to xi . The covariant derivative of a scalar V a Wa coincides with its ordinary derivative: (V a Wa )|i = (V a Wa ),i, which gives a covariant derivative of a covariant Lorentz vector: Wa|i = Wa,i bai Wb . (1.5.19) (1.5.18)

The chain rule implies that a covariant derivative of a Lorentz tensor is equal to the sum of the corresponding ordinary derivative of this tensor and terms with spin connection corresponding to each Lorentz index:
ab... a eb... b ae... e ab... e ab... T ab... cd...|i = T cd...,i + ei T cd... + ei T cd... + . . . ci T ed... di T ce... . . . . (1.5.20)

We assume that the covariant derivative |i is total, that is, also recognizes coordinate indices, acting on them like ;i . For a tensor with both coordinate and Lorentz indices
aj... aj... aj... ej... j al... e l a T aj... bk...|i = T bk...,i + ei T bk... +l i T bk... + . . . bi T ek... k i T bl... . . . . (1.5.21)

A total covariant derivative of a tetrad is

i i j b i ei a|k = ea,k + j k ea ak eb = 0,

(1.5.22)

due to (1.5.15). Therefore total covariant dierentiation commutes with converting between coordinate and Lorentz indices. Equation (1.5.22) also determines the spin connection abi in terms of the ane connection, tetrad and its ordinary derivatives:
k k j abi = ea k (eb,i + j i eb ).

(1.5.23)

Conversely, the ane connection is determined by the spin connection, tetrad and its derivatives: j ijk = jik + ea (1.5.24) i,k ea . The torsion tensor is then
j S jik = j[ik] + ea [i,k ] ea ,

(1.5.25)

37

and the torsion vector is

k Si = k[ik] + ea [i,k ] ea .

(1.5.26)

Metric compatibility of the ane connection leads to

c c a b b gik;j = gik|j = ea i ek ab|j = ei ek ( aj cb + bj ac ) = (kij + ikj ) = 0,

(1.5.27)

so the spin connection is antisymmetric in its rst two indices: abi = b ai . (1.5.28)

Accordingly, the spin connection has 24 independent components. The contortion tensor is Cijk = ijk + ijk , (1.5.29) where are the Ricci rotation coecients. The rst term on the right-hand side in (1.5.29) is expected because both the contortion tensor and spin connection are antisymmetric in their rst two indices. The quantities
i i j iak = ei a:k = ea,k + {j k }ea a a ijk = eia ea [j,k ] eja e[i,k ] eka e[i,j ]

(1.5.30)

(1.5.31)

form the Levi-Civita spin connection and are related to the Ricci rotation coecients by (1.5.29) with Cijk = 0, ijk = ijk , (1.5.32) so Cijk = ijk ijk . 1.5.5 Tetrad representation of curvature tensor (1.5.33)

The commutator of the covariant derivatives of a tetrad with respect to the ane connection is k l k 2ek (1.5.34) a;[ji] = R lij ea + 2S ij ea;l . This commutator can also be expressed in terms of the spin connection:
k k b kb b k ek a;[ji] = a[j ;i] = (eb a[j );i] = ba[j i] + a[j ;i] eb l k = ba[j kbi] + ba[j,i] ek b + S ij al .

(1.5.35)

Consequently, the curvature tensor with two Lorentz and two coordinate indices depends only on the spin connection and its ordinary derivatives: Rabij = abj,i abi,j + aci cbj acj cbi . (1.5.36)

Because the spin connection is antisymmetric in its rst two indices, the tensor (1.5.36) is antisymmetric in its rst two (Lorentz) indices, like the Riemann tensor. The contraction of the curvature tensor (1.5.36) with a tetrad gives the Ricci tensor with one Lorentz and one coordinate index: Rbj = Rabij ei a. 38 (1.5.37)

The contraction of the tensor Rai with a tetrad gives the Ricci scalar,
ab i j R = Rai ei a = R ij ea eb .

(1.5.38)

The Riemann tensor with two Lorentz and two coordinate indices depends on the Levi-Civita connection (1.5.31) the same way the curvature tensor depends on the ane connection: P abij = abj,i abi,j + aci cbj acj cbi . (1.5.39)

The contraction of (1.5.39) with a tetrad gives the Riemannian Ricci tensor with one Lorentz and one coordinate index: Pbj = P abij ei a. The contraction of the tensor P ai with a tetrad gives the Riemann scalar,
ab i j P = P ai ei a = P ij ea eb .

(1.5.40)

(1.5.41)

References: [3, 5, 6, 7, 8].

1.6
1.6.1

Lorentz group
Subgroups of Lorentz group and principle of relativity

A composition of two Lorentz transformations 1 and 2 ,

c a b = a (1)c (2)b ,

(1.6.1)

a also satsatises (1.5.10), so it is a Lorentz transformation. The Kronecker symbol b ises (1.5.10), so it can be regarded as the identity Lorentz transformation. Therefore Lorentz transformations form a group, referred to as the Lorentz group. Taking the determinant of the relation (1.5.10) gives

|ab | = 1.

(1.6.2)

A Lorentz transformation with |ab | = 1 is proper and with |ab | = 1 is improper. Proper Lorentz transformations form a group because the determinant of the product of two proper Lorentz transformations is 1. Improper Lorentz transformations include the parity transformation P ab (P ) = diag(1, 1, 1, 1), and the time reversal T The relation (1.5.10) gives 00 00 0 0 = 1, so |00 | 1. 39 ab (T ) = diag(1, 1, 1, 1). (1.6.4) (1.6.3)

(1.6.5)

Lorentz transformations with 00 1 are orthochronous and form a group. If xi is a timelike vector, xi xi > 0, then for an orthochronous transformation x0 = 00 x0 + 0 x , (1.6.6) |0 x | 0 0 x x < (00 )2 (x0 )2 = |00 x0 |. Thus the time component of a timelike vector does not change the sign under orthochronous transformations. Einsteins principle of relativity states that all physical laws are invariant under transformations within the orthochronous proper subgroup of the Lorentz group. Under the parity transformation, the spatial components of contravariant and covariant vectors (three-dimensional vectors) change the sign, while the spatial components of dual vectors (cross products) do not change the sign. Similarly, the scalar contraction of the Levi-Civita symbol and a tensor changes the sign, while a scalar does not. Quantities that transform under proper Lorentz transformations like vectors and do not change the sign in their spatial components under parity are referred to as axial vectors or pseudovectors. Quantities that transform under proper Lorentz transformations like scalars and change the sign under parity are referred to as pseudoscalars. 1.6.2 Innitesimal Lorentz transformations

Consider an innitesimal Lorentz transformation

= + ,

(1.6.7)

where are innitesimal quantities. The relation (1.5.10) gives = , (1.6.8)

where the indices are raised and lowered using the Minkowski metric tensor. Therefore Lorentz transformations are given by 6 independent antisymmetric parameters . The corresponding transformation of a contravariant vector A is 1 1 A = A + A = A + ( )A = A + J A , 2 2 where
J = .

(1.6.9)

(1.6.10) (1.6.11)

Dene matrices J such that

(J ) = J .

Therefore, in the matrix notation (with A treated as a column), 1 A = 1 + J A. 2 (1.6.12)

The 6 matrices J are the innitesimal generators of the vector representation of the Lorentz group. The explicit form of the generators of the Lorentz group in the vector 40

representation is

J01 =
J03 =

0 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0

0 0 0 0

J23 = 1.6.3

0 0 0 0 0 1 1 0

0 0 1 0 0 0 , J12 = 0 0 0 0 0 0
,

J02 =

0 0 1 0

0 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0

0 0 0 0

J31 =

0 0 0 0 0 0 0 1

(1.6.13)

Generators and Lie algebra of Lorentz group

The commutator of the generators of the Lorentz group in the vector representation is, using (1.6.10) and (1.6.11),
[J , J ] = (J ) (J ) (J ) (J ) = (J J + J + J ) , (1.6.14) so [J , J ] = J J + J + J . (1.6.15)

The relation (1.6.15) constitutes the Lie algebra of the Lorentz group. If a set of qantitites transforms under a Lorentz transformation with a matrix D () D (), then D is a representation of the Lorentz group if D (I ) = I, D(1 2 ) = D (1 )D (2),

(1.6.16)

(1.6.17)

where I denotes the identity transformation, and 1 and 2 are two Lorentz transformations. Therefore D (1 ) = D 1 (), (1.6.18) where 1 is the Lorentz transformation to : 1 = I . For an innitesimal Lorentz transformation in any representation, 1 D () = I + J , 2 according to (1.6.12). The relation
1 1 D (1 2 1 ) = D (1 )D (2 )D (1 )

(1.6.19)

(1.6.20)

gives (1.6.15), valid for any representation of the Lorentz group. 41

1 If 1 and 2 are two group transformations then 3 = 1 2 1 is a group transformation. If 2 = I + 2 G2 is an innitesimal group transformation with generator 1 G2 then 3 = I + 2 1 G2 1 is an innitesimal group transformation with genera1 tor G3 = 1 G2 1 . If 1 = I + 1 G1 is an innitesimal group transformation with generator G1 then, neglecting terms in 1 of higher order, G3 = G2 + 1 [G1 , G2 ], so [G1 , G2 ] is a generator. For a nite number N of linearly independent generators, a general innitesimal group transformation is = I + N a=1 a Ga . Because [Ga , Gb ] is a generator, it is a linear combination of the N generators: [Ga , Gb ] = N c=1 fabc Gc , where fabc are structure constants of the Lie algebra of the given group. For the Lorentz group, a Ga = D () I , where D () is given by (1.6.19).

1.6.4

Rotations and boosts

Rotations are proper orthochronous Lorentz transformations with 0 = 0 = 0, 00 = 1. (1.6.21)

Rotations act only on the spatial coordinates x and form a group, referred to as the rotation group. Boosts are proper orthochronous Lorentz transformations with = 0. Dene 1 J = e J , 2 K = J0 , and 1 = e , 2 = 0 (1.6.25) (1.6.26) (1.6.23) (1.6.24) (1.6.22)

(for the Lorentz group = 1, so the tensors e and densities are numerically identical). The explicit form of the generators of the rotation group J in the vector representation is 0 1 0 0 0 1 0 0 0 J1 = 0 0 1 , J2 = 0 0 0 , J3 = 1 0 0 . 0 0 0 1 0 0 0 1 0 For an innitesimal Lorentz transformation (1.6.19) D = 1 + J + K. (1.6.28)

(1.6.27)

A nite Lorentz transformation can be regarded as a composition of successive identical innitesimal Lorentz transformations: D = limn (1 + J/n + K/n)n = eJ+K . 42 (1.6.29)

The nite parameters , are the canonical parameters for a given Lorentz transformations. For a nite Lorentz transformation, (1.6.19) gives D () = e 2 so
1 J

(1.6.30)

D () . (1.6.31) =I The explicit form of a nite Lorentz transformation in the vector representation is J =

R1 = eJ1 =
R3 = eJ3 =

1 0 0 0

0 0 0 1 0 0 0 cos sin 0 sin cos 0 0 0 1

1 0 0 0 cos sin 0 sin cos 0 0 0 cosh 0 sinh 0

R2 = eJ2 =
B1 = eK1 =

1 0 0 cos 0 0 0 sin

B2 = eK2 =

0 sinh 0 1 0 0 0 cosh 0 0 0 1

cosh sinh 0 0 sinh cosh 0 0 0 0 1 0 0 0 0 1 cosh 0 0 sinh 0 1 0 0

0 0 0 sin , 1 0 0 cos

B3 = eK3 =

0 sinh 0 0 , 1 0 0 cosh (1.6.32)

where R denotes a rotation about the x -axis and B denotes a boost along this axis. The canonical parameters and are referred to as the angle of rotation and rapidity, respectively. The explicit form of a nite rotation in the three-dimensional vector representation is cos 0 sin 1 0 0 1 0 R1 () = 0 cos sin , R2 () = 0 , sin 0 cos 0 sin cos cos sin 0 R3 () = sin cos 0 . 0 0 1

(1.6.33)

For instance, Vx Vx cos Vy sin Vx Vx Vy Vy = R3 Vy = Vx sin + Vy cos . Vz Vz Vz Vz

(1.6.34)

The relation (1.6.31) gives J = R () 43

=0

(1.6.35)

The commutation relation (1.6.15) gives [J , J ] = e J , [J , K ] = e K , [K , K ] = e J . (1.6.36) (1.6.37) (1.6.38)

Therefore rotations do not commute and form a nonabelian group, rotations and boosts do not commute, and boosts do not commute - changing the order of two nonparallel boosts is equivalent to applying a rotation, referred to as the ThomasWigner rotation. The structure constants of the Lie algebra of the rotation group are fabc = eabc . Moreover, the square of the generators of rotation, J 2 = J J , commutes with J : [J 2 , J ] = [J , J ]J + J [J , J ] = e (J J + J J ) = 0. Denining 1 L = (J + iK), 2 1 Q = (J iK), 2 gives [L , L ] = e L , [Q , Q ] = e Q , [L , Q ] = 0, (1.6.43) (1.6.44) (1.6.45) (1.6.41) (1.6.42) (1.6.40) (1.6.39)

so the Lorentz group is isomorphic with the product of two complex rotation groups. Accordingly, the Lorentz group can be regarded as the group of four-dimensional rotations in the Minkowski space, or the group of tetrad rotations. 1.6.5 Poincar e group

Thus the tensor ik is invariant under (1.2.54) (isometric) if i is a Killing vector, (i,k) = 0, which has the solution i = ik xk + i , 44

Under the innitesimal coordinate transformation (1.2.54) in a locally at spacetime, (1.4.41) gives ik ik i,k k,i. (1.6.46) (1.6.47) (1.6.48)

where ik and i are constant. The rst term on the right-hand side of (1.6.48) corresponds to a Lorentz rotation described by 6 parameters ik . The second term on the right-hand side of (1.6.48) corresponds to a translation. A combination of two translations does not change if their order is reversed, so translations commute: [T , T ] = 0, (1.6.49)

where T is the generator of translation. The relations (1.6.36) and (1.6.37) mean that J and K are spatial vectors under rotations. Spatial translations are spatial vectors under rotations, while a time translation is a scalar: [J , T ] = e T , [J , T0 ] = 0. (1.6.50) (1.6.51)

The last relation indicates that the generators of rotations, like generators of spatial translation, correspond to conserved quantities, which are quantities that do not change in time. The covariant generalization of (1.6.50) and (1.6.51) is [J , T ] = T T . (1.6.52)

The relations (1.6.15), (1.6.49) and (1.6.52) constitute the Lie algebra of the inhomogeneous Lorentz or Poincar e group. In particular, [K , T ] = T0 , [K , T0 ] = T . (1.6.53) (1.6.54)

The last relation indicates that the generators of boosts do not correspond to conserved quantities. For an innitesimal rotation about the z -axis, (1 + Jz )f (ct, x) = D (Rz ())f (ct, x) = f (ct, Rz ()x) f (ct, x y, x + y, z ) f f + x , (1.6.55) = f (ct, x) y x y y , y x which gives the dierential representation of rotations: Jz = x J = e x . For an innitesimal boost along the z -axis, (1 + Kz )f (ct, x) = D (Bz ( ))f (ct, x) = f (Bz ( )(ct, x)) f (ct + z, y, z + ct) f f + ct , (1.6.58) = f (ct, x) + z ct z or Kz = z + ct , ct z 45 (1.6.59) or (1.6.56)

(1.6.57)

which gives the dierential representation of boosts: K = x + ct . ct x (1.6.60)

The relation for an innitesimal translation, analogous to (1.6.19), is D (t) = I + T , so a nite translation is given by D (t) = e
T

(1.6.61)

(1.6.62)

Translation in (1.6.48) can also be written as

t ()x = x + .

(1.6.63)

The relation analogous to (1.6.35) is T = t () . (1.6.64)

=0

The dierential representation of a translation is thus T = 1.6.6 . x (1.6.65)

Casimir operators of Lorentz and Poincar e group

Analogously to (1.6.40), [L2 , L ] = 0, [Q2 , Q ] = 0, (1.6.66) (1.6.67)

so L2 and Q2 commute with all 6 generators of the Lorentz group. Consequently, J 2 + K 2 and J K commute with all generators of the Lorentz group, that is, are the invariants or Casimir operators of the Lorentz group. The Casimir operators of Lorentz group do not commute with the generators of translation T , so they are not the invariants of the Poincar e group. Instead, the mass operator m2 = T T and W 2 = W W , where W is the Pauli-Luba nski pseudovector 1 W = e J T , 2 46 (1.6.70) (1.6.69) (1.6.68)

commute with all generators of the Poincar e group, so they are the Casimir operators of the Poincar e group. The Pauli-Luba nski pseudovector obeys the commutation relations [T , W ] = 0, [J , W ] = W W , [W , W ] = e W T . (1.6.71) (1.6.72) (1.6.73)

The relation (1.6.72) is analogous to (1.6.52) because W behaves like a vector under proper Lorentz transformations. Dene the four-momentum operator P = iT , (1.6.74)

whose time component is the energy operator P0 = iT0 and spatial components form the momentum operator P = iT . Dene the angular four-momentum operator M = iJ , whose spatial components form the angular momentum operator M = iJ . Therefore the following relations are satised: [M , M ] = i(M + M M M ), [P , P ] = 0, [M , P] = i(P P ), m2 = P P , 1 W = M P , 2 [P , W ] = 0, [M , W ] = i(W W ), [W , W ] = ie W P , [M , M ] = ie M . 1.6.7 Relativistic kinematics (1.6.77) (1.6.78) (1.6.79) (1.6.80) (1.6.81) (1.6.82) (1.6.83) (1.6.84) (1.6.85) (1.6.76) (1.6.75)

Consider a boost in the direction of the z -axis xi = eK3 xi , (1.6.86)

where xi and xi have a form of a column (41 matrix), and eK3 is given by (1.6.32). Therefore the coordinates in an inertial K -system (unprimed) are related to the coordinates in an inertial K -system (primed) by ct = ct cosh + z sinh, x = x , y = y , z = z cosh + ct sinh. 47

(1.6.87)

Consider the origin of the K -system, x = y = z = 0, in the K -system. Therefore ct = ct cosh, z = ct sinh, which gives the relation between the rapidity and velocity V = K: tanh = , V . c Accordingly, cosh = and sinh = , where = = 1 The relations (1.6.87) become V z , c2 x = x , y = y , z = (z + V t ), t = t + V2 c2
1/2 dz dt

(1.6.88) of K relative to (1.6.89) (1.6.90)

where

(1.6.91)

(1.6.92)

and are referred to as a special Lorentz transformation in the z -direction. The reverse transformation is V z , c2 x = x, y = y, z = (z V t). t = t

(1.6.93)

For a boost along an arbitrary direction, the spatial vector x = (x, y, z ) transforms such that its component parallel to the velocity V = c of K relative to K , x = (x V)V/V 2 (similarly for primed), behaves like z in (1.6.92) and its component perpendicular to V, x = x x , behaves like x in (1.6.92): t = t + V x , c2 (1.6.94)

x = x , x = (x + Vt ), so x = (x + Vt ) + x = Vt + x +

( 1)(V x )V . (1.6.95) V2 Therefore the transformation law for the coordinates in two inertial frames of reference is ct ct , (1.6.96) = ( 1) x x 1 + 2 48

or equivalently ct x = 1) 1 + ( 2 ct x . (1.6.97)

The matrix in (1.6.97) is called a boost matrix. In the local Minkowski spacetime, contravariant vectors transform like xi , according to (1.6.96), W0 W = ( 1) 1 + 2 V 0 W , (1.6.98)

covariant vectors transform such that they remain related to contravariant vectors by the Minkowski metric tensor, and tensors transform like products of vectors. For example, if V = c z is parallel to the z -axis, a tensor of rank (0,2) transforms according to T00 = (T00 + T03 ) = 2 (T0 0 + T3 0 + T0 3 + 2 T3 3 ), T0 = (T0 + T3 ), T03 = (T03 + T00 ) = 2 (T0 3 + T3 3 + T0 0 + 2 T3 0 ), T = T , T3 = (T3 + T0 ), T33 = (T33 + T30 ) = 2 (T3 3 + T0 3 + T3 0 + 2 T0 0 ),

(1.6.99)

T where the index denotes either 1 or 2, and the transposed components Tik = Tki transform like the transpositions of the right-hand sides in (1.6.99). If Tik is antisymmetric then T03 = T0 3 . The relations (1.6.92) can be written as

V dz , c2 dx = dx , dy = dy , dz = (dz + V dt ), dt = dt + which gives

vx vx = , /c2 ) (1 + V vz vy , vy = /c2 ) (1 + V vz vz +V vz = , /c2 1 + V vz

(1.6.100)

(1.6.101)

where

dx dx , v = . (1.6.102) dt dt Two special Lorentz transformations in the same direction commute because of (1.6.38). If a Lorentz transformation from K to K has parameters 1 and 1 , and a Lorentz v= 49

transformation from K to K has parameters 2 and 2 , then a Lorentz transformation from K to K has parameters 3 and 3 such that 3 = 1 + 2 , 3 = 1 2 (1 + 1 2 ). 1 + 1 2 (1.6.103)

If v = |V | = c then v = |V| = c, in agreement with the constancy of the velocity of propagation of interaction. Consider two points at rest in the inertial frame of reference K with positions z1 and z2 , so the distance between them is z = z2 z1 . In the inertial frame K , moving relative to K in the z -direction with velocity V , z1 = (z1 + V t1 ) and z2 = (z2 + V t2 ), so if t1 = t2 is the time at which we measure (simultaneously) the positions of the two points then z = (z2 z1 ) = z . Therefore the length of an object in K , whose length in the rest frame K is l (proper length), is l = l < l,

For a boost along an arbitrary direction, (1.6.96) gives the Lorentz transformation of velocities: v + V + ( 1)(v V)V/V 2 . (1.6.104) v= (1 + v V/c2)

(1.6.105)

which is referred to as the Lorentz-FitzGerald contraction. The volume of an object in K , whose volume in the rest frame K is V (proper volume), is V = V . (1.6.106)

Suppose that there are two rods of equal lengths, moving parallel relative to each other. From the point of view of an observer moving with the rst rod, the second one is shorter, and from the point of view of an observer moving with the second rod, the rst one is shorter. There is no contradiction in this statement because the positions of both ends of a rod must be measured simultaneously and the simultaneity is not invariant: from the transformation law (1.6.92) it follows that if t = 0 then t = 0 and if t = 0 then t = 0. Consider a clock (any mechanism with a periodic or evolutionary behavior) at rest in K with position z ; the time dierence between two events with t1 and t2 , as measured by this clock, is t = t2 t1 . In the frame K , t1 = (t1 + V z /c2 ) and t2 = (t2 + V z /c2 ), so t = t2 t1 = t > t . (1.6.107) Thus the rate of time is slower for moving clocks than those at rest (time dilation), in agreement with (1.4.90) and (1.4.96), from which c2 d 2 = c2 dt2 dl2 and d = 1 dt.

(1.6.108)

Suppose that there are two clocks linked to the inertial frames K and K , and that when the clock in K passes by the clock in K the readings of the two clocks coincide. 50

From the point of view of an observer in K clocks in K go more slowly, and from the point of view of an observer in K clocks in K go more slowly. There is no contradiction in this statement because to compare the rates of the two clocks in K and K we must compare the readings of the same moving clock in K with dierent clocks in K ; we require several clocks in one frame and one in the other, thus the measurement process is not symmetric with respect to the two frames of reference. The clock that goes more slowly is the one which is being compared with dierent clocks in the other frame. The time interval measured by a clock is equal to the integral 1 t = ds (1.6.109) c along its world line. Since the world line is a straight line for a clock at rest and a curved line for a clock moving such that it returns to the starting point, the integral ds taken between two world points has its maximum value if it is taken along the straight line connecting these two points. For a Lorentz transformation with velocity V = |V|, (1.6.104) gives tan = v sin , (v cos + V ) (1.6.110)

where is the angle between v and V, and is the angle between v and V. If v = v = c then cos + V c , (1.6.111) cos = cos 1+ V c which is referred to as the aberration of a signal. Suppose an observer in frame K 1 c measures a periodic signal with period T , frequency = T and wavelength = , propagating in the z direction; the number of pulses in time dt is n = dt. A second observer in frame K , moving in the z direction with velocity V relative to the rst dt one, travels a distance V dt and measures V more pulses: n = (1 + V )dt. Because c dt the time interval dt with respect to K is dt = , the frequency of the signal in K is = (1 + V ) or c = e . (1.6.112) This dependence of the frequency of a signal on a frame of reference is referred to as the Doppler eect. When c (at which 1) the above formulae, referring to relativistic kinematics, reduce to their nonrelativistic limit. The Lorentz transformation (1.6.96) reduces to the Galileo transformation, t = t , x = x + Vt, (1.6.113)

so the time is an absolute (invariant) quantity in nonrelativistic (Newtonian) physics. Any two Galileo transformations commute. The transformation law for velocities (1.6.104) reduces to the simple addition of vectors, v = v + V. 51 (1.6.114)

1.6.8

Four-acceleration

In a locally inertial frame of reference, the four-velocity is ui = , where v is the velocity and = (1 wi = v v , ui = , , c c
v2 1/2 ) . c2

(1.6.115)

Dene the four-acceleration (1.6.116)

D 2 xi Dui = , ds ds2

which is orthogonal to ui because of (1.4.19): w i u i = 0, (1.6.117)

thus having 3 independent components. In a locally inertial frame of reference, the four-acceleration is wi = dui d2 xi 4v a 2 2 4 (v a)v = = c , a + , ds ds2 c c2 (1.6.118)

where a is the three-dimensional acceleration vector a= d2 x dv = 2. dt dt (1.6.119)

The invariant square of the four-acceleration is thus w i wi = 4 2 2 a + 2 (v a)2 . c4 c (1.6.120)

If v = 0 at a given instant of time, the corresponding frame of reference is referred to as the instantaneous rest frame. In this frame w i wi = so a0 = c2 w i wi is the absolute value of the acceleration in the instantaneous rest frame. References: [2, 3]. (1.6.122) a2 , c4 (1.6.121)

1.7
1.7.1

Spinors
Spinor representation of Lorentz group

Let a be the coordinate-invariant 44 Dirac matrices dened as a b + b a = 2 ab I, 52 (1.7.1)

where I is the unit 44 matrix (4 is the lowest dimension for which (1.7.1) has a solutions). Accordingly, the spacetime-dependent Dirac matrices, i = ei a , satisfy i j + j i = 2g ij I. Under a tetrad rotation, (1.5.8) gives a = ab b . Let L be a 44 matrix such that a = ab L b L1 = L a L1 , (1.7.4) (1.7.3) (1.7.2)

where L1 is the matrix inverse to L: LL1 = L1 L = I . The condition (1.7.4) represents the constancy of the Dirac matrices a under the combined tetrad rotation and transformation LL1 . We refer to L as the spinor representation of the Lorentz group. The relation (1.7.4) gives the matrix L as a function of the Lorentz matrix ab . For an innitesimal Lorentz transformation (1.6.7), the solution for L is 1 1 L = I + ab Gab , L1 = I ab Gab , 2 2 (1.7.5)

where Gab are the generators of the spinor representation of the Lorentz group: 1 Gab = ( a b b a ). 4 (1.7.6)

A spinor is dened as a quantity that, under tetrad rotations, transforms according to = L. (1.7.7) is dened as a quantity that transforms according to An adjoint spinor = L 1 , is a scalar: so the product = . (1.7.8)

(1.7.9)

The indices of the a and L that are implicit in the 44 matrix multiplication in (1.7.1), (1.7.2) and (1.7.4) are spinor indices. The relation (1.7.4) implies that the Dirac matrices a can be regarded as quantities that have, in addition to the invariant transforms index a, one spinor index and one adjoint-spinor index. The product like the Dirac matrices: = L L 1 . (1.7.10) can be used to construct tensors. For example, a transforms The spinors and like a contravariant Lorentz vector: a L 1 a L b L1 L = a b . b b 53 (1.7.11)

1.7.2

Spinor connection

The derivative of a spinor does not transform like a spinor: ,i = L,i + L,i . If we introduce the spinor connection i that transforms according to i = Li L1 + L,i L1 , then a covariant derivative of a spinor, ;i = ,i i , is a spinor: ;i = ,i = L,i + L,i (Li L1 + L,i L1 )L = L;i . i is a scalar, Because );i = ( ),i, ( (1.7.15) (1.7.16) (1.7.14) (1.7.13) (1.7.12)

the chain rule for covariant dierentiation gives a covariant derivative of an adjoint spinor ;i = ,i + i . (1.7.17) Also |i = ;i . |i = ;i , , whose covariant derivative is The Dirac matrices a transform like );i = ;i + ;i = ( ),i i + i = ( ),i [i , ]. ( Therefore a covariant derivative of the Dirac matrices is a;i = a,i [i , a ] = [i , a ], so Accordingly
j k j j;i = j|i = j,i + k, i [i , ].

(1.7.18)

(1.7.19)

(1.7.20) (1.7.21) (1.7.22)

i |i transforms under Lorentz rotations like a scalar: The quantity i |i L 1 L i L1 L|i = i |i . The relation ab|i = 0 implies that a|i = 0, (1.7.24) because the Dirac matrices a only depend on ab . Multiplying both sides of (1.7.22) by a from the left gives abi a b a i a + 4i = 0. 54 (1.7.25) (1.7.23)

a|i = abi b [i , a ].

We seek the solution of (1.7.25) in the form 1 i = abi a b Ai , 4 (1.7.26)

where Ai is a spinor-tensor quantity with one vector index. Substituting (1.7.26) to (1.7.25), together with the identity c a b c = 4 ab , gives a Ai a + 4Ai = 0, (1.7.27)

so Ai is an arbitrary vector multiple of I . Therefore the spinor connection i is given, up to the addition of an arbitrary vector multiple of I , by the Fock-Ivanenko coecients: 1 1 (1.7.28) i = abi a b = abi Gab . 4 2 Using the denition (1.5.15), we can also write (1.7.28) as 1 1 j c i = ej c;i [j , ] = [ ;i , j ]. 8 8 1.7.3 Curvature spinor (1.7.29)

The commutator of total covariant derivatives of a spinor is |ji |ij = (|j ),i i |j jki |k (|i ),j + j |i + ik j |k where Kij = Kji is dened as Kij = i,j j,i + [i , j ]. Substituting (1.7.13) to (1.7.31) gives ij = i,j j,i + [ i, j ] = L(i,j j,i + [i , j ])L1 = LKij L1 , K (1.7.32) (1.7.31)

= j,i + i j + i,j j i + 2S kij |k = Kij + 2S kij |k , (1.7.30)

so Kij transforms under tetrad rotations like the Dirac matrices a , that is, Kij is a spinor with one spinor index and one adjoint-spinor index. We refer to Kij as the curvature spinor. The relation (1.7.24) leads to k|i = 0. (1.7.33) Thus the commutator of covariant derivatives of the spacetime-dependent Dirac matrices vanishes: 2 k|[ji] = Rklij l + 2S lij k|l + [Kij , k ] = Rklij l + [Kij , k ] = 0. Multiplying both sides of (1.7.34) by k from the left gives Rklij k l + k Kij k 4Kij = 0. 55 (1.7.35) (1.7.34)

We seek the solution of (1.7.35) in the form 1 Kij = Rklij k l + Bij , 4 (1.7.36)

where Bij is a spinor-tensor quantity with two vector indices. Substituting (1.7.36) to (1.7.35) gives k Bij k 4Bij = 0, (1.7.37) so Bij is an antisymmetric-tensor multiple of I . The tensor Bij is related to the vector Ai in (1.7.26) by Bij = Aj,i Ai,j + [Ai , Aj ]. (1.7.38) Because has no indices other than spinor indices, Ai is a vector and [Ai , Aj ] = 0. The invariance of (1.7.35) under the addition of an antisymmetric-tensor multiple Bij of the unit matrix to the curvature spinor is related to the invariance of (1.7.25) under the addition of a vector multiple Ai of the unit matrix to the spinor connection. Setting Ai = 0, which corresponds to the Fock-Ivanenko spinor connection, gives Bij = 0. Therefore the curvature spinor Kij is given, up to the addition of an arbitrary antisymmetric-tensor multiple of I , by 1 1 Kij = Rklij k l = Rklij Gkl . 4 2 References: [3, 4]. (1.7.39)

56

2
2.1

Fields
Principle of least action

The most general formulation of the law that governs the dynamics of classical systems is Hamiltons principle of least action, according to which every classical system is characterized by a denite scalar-density function , and the dynamics of the system is such that a certain condition is satised. Let A (xi ) be a set of physical elds, being dierentiable functions of the coordinates, and let be a Lorentz covariant quantity constructed from the A and their derivatives. Consider a scalar quantity 1 d, (2.1.1) c where the integration is over some region in locally Minkowski spacetime. Let A be arbitrary small changes in A (regarded as a dynamical variable) over the region of integration, which vanish on the boundary. Then the change in S can be written as 1 F A A d. (2.1.2) S = c The principle of least action states that the dynamics of a physical system is given by the condition the scalar S be a local minimum. Therefore any innitesimal change in the dynamics of the system does not alter the value of S : S= S = 0 (2.1.3) (S is a local extremum). If is covariant and A transform covariantly under the Lorentz group, the variational condition (2.1.3) gives the Lorentz covariant equations A = 0. (2.1.4)

These equations are also invariant for any other transformations (internal symmetries) for which is invariant. is referred to as the Lagrangian density, S is the action functional, S = 0 is the principle of least action, and F A = 0 are the eld equations. The eld equations of a physical system are the result of the action being a local extremum. The condition that the action be a local minimum imposes additional restrictions on possible choices for S . The number of independent eld equations for a given system is referred to as the number of the degrees of freedom representing this system. In most cases contains only A and their rst derivatives (the Lagrangian density for the gravitational eld contains second derivatives). A Lagrangian density containing higher derivatives can always be written in terms of rst derivatives by increasing the number of the components of A . Consider a physical system in the galilean frame of reference. If depends only on A and i A , = (, ,i), then S = = 1 c 1 c 1 + (,i) d = + (),i d (,i ) c (,i ) + i i d. (2.1.5) (,i ) (,i ) 57

The last term in the second line of (2.1.5) is a divergence, which, after integration, can be transformed into a hypersurface integral over the boundary of integration region, where = 0 on the boundary, so this term does not contribute to the action variation: 1 1 d + d. i dSi = i S = c (,i ) (,i ) c (,i ) (2.1.6) If S = 0 for arbitrary variations that vanish on the boundary then = 0, i (,i ) or = 0, (2.1.7)

(2.1.8)

(2.1.9) = i (,i ) is a variational derivative of with respect to . This set of equations, for each component A , is referred to as the Lagrange equations. Generalizing the Lagrange equations to an arbitrary coordinate frame gives {} = 0. i (,i ) (2.1.10)

where

There is some arbitrariness in the choice of ; adding to it the divergence of an arbitrary vector density or multiplying it by a constant produces the same eld equations. If a system consists of two noninteracting parts A and B , with corresponding Lagrangian densitites A (A , A ) and B (B , B ), then the Lagrangian for this system is the sum A + B . This additivity of the Lagrangian density express the fact that the eld equations for either of the two parts do not involve quantities pertaining to the other part. If A also depends on B and/or B , and/or B depends on A and/or A , then the subsystems A and B interact. References: [1, 2, 3].

2.2

Action for gravitational eld

Consider a Lagrangian density that depends on the ane (or spin) connection and its rst derivatives. Such Lagrangian density can be decomposed into the covariant part that contains derivatives of the ane/spin connection, which is referred to as the Lagrangian density for the gravitational eld, and the covariant part that does not contain these derivatives, which is referred to as the Lagrangian density for matter. The simplest covariant scalar that can be constructed from the ane/spin connection and its rst derivatives is the Ricci scalar R. The corresponding Lagrangian density for the gravitational eld is proportional to the product of R and the scalar density : 1 R, (2.2.1) g = 2 58

where is Einsteins gravitational constant. There exist two variational principles in the theory of the gravitational eld. The metric variational principle regards the metric tensor or tetrad as a dynamical variable and assumes the ane connection to be the Levi-Civita connection. The metric-ane variational principle regards both the metric tensor (or tetrad) and the metric-compatible ane connection (or spin connection) as dynamical variables. In the metric variational formulation, the Lagrangian density for the gravitational eld is proportional to the Riemann scalar P :

g =

1 P. 2

(2.2.2)

Because P is linear in derivatives of {kil }: l m l P = g ik ({ilk },l {ill },k + {im k }{m l } {i l }{m k }) = ( g ik {ilk }),l {ilk }( g ik ),l ( g ik {ill }),k + {ill }( g ik ),k l m l + g ik ({im (2.2.3) k }{m l } {i l }{m k }), we can subtract from P total derivatives without altering the eld equations, replacing P by a noncovariant quantity G: l m l G = {ill }( g ik ),k {ilk }( g ik ),l + g ik ({im k }{m l } {i l }{m k }) = {ill }(( g ik ):k + {jjk } g ik {jik }g jk {jkk }g ij ) {ilk }(( g ik ):l + {jjl } g ik {jil }g jk {jkl }g ij ) j l m l l ik + g ik ({im {jik }g jk k }{m l } {i l }{m k }) = {i l }({j k } g {jkk }g ij ) {ilk }({jjl } g ik {jil }g jk {jkl }g ij ) l m l l m l + g ik ({im g ik ({im (2.2.4) k }{m l } {i l }{m k }) = l }{m k } {i k }{m l }). Therefore and the Lagrangian density for the gravitational eld is
l m l G = g ik ({im l }{m k } {i k }{m l }),

(2.2.5)

g =
.

1 G. 2

(2.2.6)

Any coordinate transformation results in variations of g ik , so S is not necessarily a minimum with respect to these variations (only an extremum) because not all g ik correspond to actual variations of the gravitational eld. In order to exclude the variations g ik resulting from changing the coordinates, we must impose on the metric tensor 4 arbitrary constraints. If we choose g0 = 0, |g | = const, then G becomes 1 G = g 00 g g g,0 g,0 . 4 59 (2.2.7)

(2.2.8)

In the locally galilean frame g = , so 1 G = g 00 (g,0 )2 . 4 (2.2.9)

For physical systems g 00 > 0. Therefore in order for S to have a minimum, must be positive, otherwise an arbitrarily rapid change of g in time would result in an arbitrarily low value of S and there would be no minimum. References: [2, 3].

2.3
2.3.1

Matter
Metric dynamical energy-momentum density

The variation of the matter action Sm = Sm = 1 2c

m d with respect to the metric tensor,

1 2c T ij gij d, (2.3.1)

Tij g ij d =

denes the metric dynamical energy-momentum density Tij , which is symmetric: Tij = Tji . Equivalently Tij = 2 m m m =2 k ( ij . ij ij g g g Tij Tij = . 2.3.2 Tetrad dynamical energy-momentum density (2.3.2)

(2.3.3)

The metric dynamical energy-momentum tensor Tij is dened as (2.3.4)

The variation of the matter action Sm with respect to the tetrad, Sm = 1 c

iaeia d, ia . Equivalently

(2.3.5)

denes the tetrad dynamical energy-momentum density m = ia ei a or

(2.3.6)

m . (2.3.7) ei a If m depends only on tensor matter elds expressed in terms of the coordinate indices and it depends on neither derivatives of the metric tensor nor derivatives of the tetrad then the tetrad enters m only through the metric tensor, in a combination j g ij = ab ei a eb . Thus 1 ij (2.3.8) ei a = eaj g . 2

ia =

60

Substituting (2.3.8) to (2.3.5) gives Sm = where 1 2c

ij g ij d,

(2.3.9)

ij = eaj ia.

(2.3.10)

The tensor ij is generally not symmetric. Comparing (2.3.9) with (2.3.1) gives the relation between the tetrad dynamical energy-momentum density and the metric dynamical energy-momentum density for tensor matter elds:

(ij ) = Tij .
2.3.3 Canonical energy-momentum density

(2.3.11)

If we express the matter Lagrangian density m , depending on matter elds and their rst derivatives ,i, only in terms of Lorentz and spinor indices, then the tetrad appears in m only through a derivative of , in a covariant combination ei a |i . Since m = L, where L is a scalar, we obtain
i m = L ea i Lea =

m i |i m ea i ea . ,a

L m a i i |i ei |i m ea a m ei ea = i ea |a |a (2.3.12) m |i ea i m , ,a m i m . |j j ,i

The last term in (2.3.12), ia =

(2.3.13)

is referred to as the canonical energy-momentum density. Accordingly j i = (2.3.14)

Comparing (2.3.12) with (2.3.6) shows that the canonical energy-momentum density is identical with the dynamical tetrad energy-momentum density: ia = ia . 2.3.4 Spin density (2.3.15)

The variation of the matter action Sm with respect to the spin connection, Sm = denes the dynamical spin density 1 i abi d, 2c ab m , abi (2.3.16)

abi : abi = 2
61 (2.3.17)

which is antisymmetric in the Lorentz indices:

abi = bai .

(2.3.18)
{}ab

In the metric variational formulation of gravity, the variations abi = i are functions of the variations ei a and their derivatives, so the spin density is a function of the energy-momentum density. In the metric-ane variational formulation of gravity, the variations abi are independent of ei a and their derivatives. The relation (1.5.29) indicates that the spin density is generated by the contortion tensor:

ij k = 2
Accordingly, the variation of

m . C ijk

(2.3.19)

m with respect to the torsion tensor,

i jk = 2 m , S ijk (2.3.20)

is a homogeneous linear function of the spin connection because of (1.4.29): m m C lmn l m n m n l n m l [j k] + i [j k] + i [j k] ) = 2 = lmn (i S ijk C lmn S ijk = ijk jki + kij , (2.3.21) ijk = [ij ]k , (2.3.22) ijk = 2 antisymmetric in the last two indices: ijk = ikj . (2.3.23)

The variation of m with respect to the metric-compatible ane connection in the metric-ane variational formulation of gravity is equivalent to the variation with respect to the torsion (or contortion) tensor. The spin connection abi enters m only through derivatives of , in a combination i , where i is the covariant ,i derivative acting on : 1 i = abi Gab . (2.3.24) 2 Consequently, the dynamical spin density abi =

abi is identical with

(2.3.25)

m Gab , ,i

referred to as the canonical spin density. Spin tensor is dened as sijk =

ijk

(2.3.26)

62

2.3.5

Belinfante-Rosenfeld relation 1 c 1 2c 1 2c 1 2c

The total variation of the matter action with respect to geometrical variables is either Sm = or Sm = Equation (1.5.25) gives 1 2 dj ik S jik = 1 2
ab a j a j dj ik (ej a eib k ) + ei,k ea + ei,k ea

dia ei a+ dTik g ik +

dabi abi dj ik S jik .

(2.3.27) (2.3.28)

1 ik j ik j ik a ab i ab a a j d j li (ej a elb ) i + ab i + (j ea ei ),k (j ea ),k ei + j ei,k ea 2 1 li ab j ik j lm b i ab a d j lk cbk elb ej el,m ej = c + j i ea elb + ab i (j ea ),k ei + j b 2 1 1 il b j a a j a i ab dSk j ik ej d j lk cbk elb ei + a ei = c ea ei + j al eb ei + ab i 2 2 lm b j a (a ik|k S ijk a jk 2Sj a ij + bak b ik )ea el,m ei i j b ea ei 1 a ij a = d ab i abi j ik;k ej (2.3.29) a ei + 2Sj a ei , 2 so comparing of (2.3.27) with (2.3.28) leads to = dia ei a+ 1 2 1 2 1 ka i dTik e ea + 2 dabi abi = dTik g ik + 1 2
a d ab i abi j ik;k ej a ei i di jk;k ea j ea

+2Sj a ij ea i =

dab i abi +

1 2

b i dSj b kj ea k ei ea .

(2.3.30)

The terms with abi give (2.3.22), while the terms with ei a give
aj ia = Tik eka + i jk;k ea j Sj i

1 2

(2.3.31)

1 Tik = ik j (ikj kj i + j ik ) + Sj (ikj kj i + j ik ). (2.3.32) 2 Equation (2.3.32) is referred to as the Belinfante-Rosenfeld relation between the dynamical metric and dynamical tetrad (canonical) energy-momentum densites. In the absence of torsion, (2.3.32) is consistent with (2.3.11). The Belinfante-Rosenfeld relation can be written as 1 1 j j j (2.3.33) Tik = ik j (sik sk i + s ik ), 2 where i = i 2 S i (2.3.34)

or

is the modied covariant derivative. References: [2, 3, 5, 6, 7].

63

2.4
2.4.1

Symmetries and conservation laws

Noether theorem

Consider a physical system in the galilean frame of reference, described by the Lagrangian density that depends on matter elds A , their rst derivatives ,i, and the coordinates xi . The change of the Lagrangian density under an innitesimal coordinate transformation (1.2.54) is thus = + (,i ) + i i , ,i x (2.4.1)

where the changes and (,i ) are brought by the transformation (1.2.54) and i denotes partial dierentiation with respect to x at constant and ,i . The variation under this transformation is also given by (1.2.63): = i,i . Using the Lagrange equations (2.1.7) and the identities (2.4.2)

,i =

,ji, + ,i + i x ,j

(2.4.3) (2.4.4)

(,i) = (),i j,i ,j , we bring (2.4.1) to = i ,i + ( j ,j ) . ,i ,i

(2.4.5)

Combining (2.4.2) and (2.4.5) gives the conservation law,

i,i = 0,
for the current

(2.4.6)

i = i +

( j ,j ) = i + . ,i ,i

(2.4.7)

Equations (2.4.6) and (2.4.7) represent the Noether theorem, which states that to each continuous symmetry of a Lagrangian density there corresponds a conservation law. Generalizing (2.4.6) to an arbitrary coordinate frame gives

i:i = 0.
2.4.2 Conservation of spin

(2.4.8)

The Lorentz group is the group of tetrad rotations. Since a physical matter Lagrangian density m (, ,i) is invariant under local, proper Lorentz transformations, it is invariant under tetrad rotations: m = m m 1 i ab + (,i) + ia ei a + ab i = 0, ,i 2 64 (2.4.9)

where the changes correspond to a tetrad rotation. Under integration of (2.4.9) over spacetime, the rst two terms vanish because of the Lagrange equations for (2.1.7):

ia eia + abi abi d4 x = 0.

For an innitesimal Lorentz transformation (1.6.7), the tetrad ea i changes by
a a b a a ea a i = e i ei = b ei ei = i , a j and the tetrad ei a , because of the identity (ei ea ) = 0, according to i ei a = a .

1 2

(2.4.10)

(2.4.11)

(2.4.12)

The spin connection changes by

jb jb a bc ab a a jb a cb a jb abi = (ea j i ) = j i ej ;i = c i ej |i + c i = |i .

(2.4.13)

Substituting (2.4.12) and (2.4.13) to (2.4.10), together with partial integration (1.2.33), gives =

ia ia + abi ab|i d4 x =
[ij ] Sk ij k +

1 2

1 k ij d4 x. 2 ij ;k

ij ij + ijk ij |k d4 x
(2.4.14)

1 2

Since the innitesimal Lorentz rotation ij is arbitrary, we obtain the covariant conservation law for the spin density:

ij k;k = ij ji + 2Sk ij k
or 1 k (ij ji). k sij =

(2.4.15)

(2.4.16)

The conservation law for the spin density (2.4.16) also results from antisymmetrizing the Belinfante-Rosenfeld relation (2.3.33) with respect to the indices i, k . If we use the metric-compatible ane connection ik j , which is invariant under tetrad rotations, instead of the spin connection abi as a variable in m then we must replace the term with abi in (2.4.9) by a term with (ei a,j ). 2.4.3 Conservation of metric energy-momentum

Consider the metric variational formulation of gravity. Under an innitesimal coordinate transformation (1.2.54), the matter Lagrangian density m (, ,i) changes according to m = m m m m + (,i) + ik g ik + ik (g ik,l ). ,i g g ,l 65 (2.4.17)

The matter action Sm = transformation: Sm = 1 c

1 c

m(, ,i)d is a scalar, so it does not change under this

(2.4.18)

m m m m + (,i) + ik g ik + ik (g ik,l ) d = 0. ,i g g ,l

The rst two terms in (2.4.18) vanish because of the Lagrange equations for (2.1.7), so Sm = 1 c 1 m m l ik g ik d = ik g g ,l c 1 m ik g d = ik g 2c Tij g ij d = 0.

(2.4.19) If the components of the metric tensor change because of an innitesimal coordinate transformation (1.2.54) then the corresponding variation of the metric tensor is given by (1.4.41): ij = 2(i:j ) , gij = g (2.4.20) so ij d = 1 T ij (i:j ) d = 1 T ij i:j d m = 1 T ij g Sm = S 2c c c 1 1 1 1 = (T ij i ):j d + T ij (T ij i ),j d + T ij : j i d = : j i d c c c c 1 1 T ij i dSj + T ij (2.4.21) = : j i d = 0. c c If the variation of the coordinates i vanishes on the boundary of the region of integration then T ij (2.4.22) : j i d = 0, which, for arbitrary variations i gives the covariant conservation of the metric energymomentum density (4 equations): T ij : j = 0. Equivalently T ij:j = 0. (2.4.23) (2.4.24)

ij d in (2.4.21) does not imply T ij = 0, because 10 Note that vanishing of T ij g ij are functions of 4 variations i and thus not independent. variations g 2.4.4 Conservation of tetrad energy-momentum

The matter Lagrangian density m is invariant under innitesimal translations of the coordinate system (1.2.54). The corresponding changes of the tetrad and spin connection are given by Lie derivatives i = L ei = i ej j ei , e a a ,j a a,j j ab ab ab i = L i = ,i j j abi,j . 66 (2.4.25) (2.4.26)

Equation (2.4.10) becomes now i + i ab d4 x = 0. ia e a ab i Substituting (2.4.25) and (2.4.26) into (2.4.27) gives
a j i i j i j ab 4 ab ia i,j ej a i ea,j ab ,i j ab i,j d x

1 2

(2.4.27)

1 1 j ab j ab i 4 ij ,j j a ej a,i + (ab i ),j ab j,i d x = 0. (2.4.28) 2 2

1 2

1 2

This equation holds for an arbitrary vector i , so we obtain

abj,j abi + abj ( abi,j abj,i ) 2ij,j 2ja ej a,i j j c k j c ab = (ab |j 2Sk ab + cb aj + ac bj ) i 2ij ,j 2j a ej a,i (2.4.29) +abj (Rabij + aci cbj acj cbi ) = 0,
which reduces to = (jlk;k 2Sk jlk ) jli Rklij klj 2ij ;j + 4Sj ij 2jk jki (abj |j 2Sk abk ) abi Rabij abj 2ij ;j + 4Sj ij 2jk jki + 4S jki jk (2.4.30)

+4S jki jk = 0.

The conservation law for the spin density (2.4.15) brings (2.4.30) to the covariant conservation law for the energy-momentum density:

ij;j = 2Sj ij + 2S jkijk + klj Rklji

or

1 2

(2.4.31)

ij:j = Cjk i jk + klj Rklji.

Conservation laws for Lorentz group

1 2

(2.4.32)

2.4.5

Consider a matter Lagrangian m for a physical system in the galilean and geodesic frame of reference, that depends on the coordinates only through a eld and its rst derivatives ,i. Therefore i m =

,i +

m m m m ,i + ,ji = j ,ji = j ,i , ,j ,j ,j ,j

(2.4.33)

where we use the Lagrange equations (2.1.7), from which we obtain the conservation law, ij,j = 0, (2.4.34) for ij = m j m . ,i i ,j 67 (2.4.35)

The conservation law (2.4.34) is a special case of (2.4.31) in the absence of torsion and spin, expressed in the Galilean and geodesic frame. The quantity (2.4.35) is a special case of the canonical energy-momentum density (2.3.14) in the absence of torsion and spin, expressed in the galilean and geodesic frame. If xi are Cartesian coordinates then for translations, i = i =const and = 0, the current (2.4.7) is i = i m m j ,j . (2.4.36) ,i The conservation law (2.4.6) gives j j i,i = 0, (2.4.37)

which gives (2.4.34) because i are arbitrary. For Lorentz rotations, i = ij xj and 1 ij Gij , where Gij are the generators of the Lorentz group, the current (2.4.7) = 2 is

i = ij xj m +

m 1 kl 1 m m ,l xk li m + Gkl . Gkl jk xk j,j = kl xk ,i 2 ,i 2 ,i (2.4.38) The conservation law (2.4.6) gives kl 1 m m ,[l xk] [il xk] m + Gkl ,i 2 ,i ,
,i

(2.4.39)

which, because kl are arbitrary, gives

kli,i = 0,
where

(2.4.40) m Gkl . ,i (2.4.41)

kli = xk l i xl k i +
The quantity

kli is referred to as the angular momentum density, and is the sum, kli = kli + kli ,
(2.4.42)

of two densities: the orbital angular momentum density, kli = xk l i xl k i , (2.4.43)

and the canonical spin density (2.3.25). The conservation law (2.4.40) for the angular momentum density is equivalent to kl lk kli,i = 0, (2.4.44)

which is a special case of the conservation law for the spin density (2.4.15) in the absence of torsion, expressed in the galilean and geodesic frame. The canonical energymomentum density ik is not symmetric. However, the quantity 1 ik = ik j (ikj kj i + j ik ), 2 68 (2.4.45)

is symmetric, which follows from (2.4.44), and conserved: ik = ki , k i,i = 0. (2.4.46) (2.4.47)

The symmetric energy-momentum density ik corresponds to the metric dynamical energy-momentum density (2.3.3), expressed in the galilean and geodesic frame. Equation (2.4.45) is a special case of the Belinfante-Rosenfeld relation (2.3.32) in the absence of torsion, expressed in the Galilean and geodesic frame. The second term on the right-hand side of (2.4.45) has the form j ikj , where ikj = ijk . Adding such term to ik preserves the conservation law (2.4.34) and brings ik to a symmetric form. 2.4.6 Components of energy-momentum tensor

Integrating the conservation law (2.4.34), valid in the galilean and geodesic frame of reference, over a hypersurface enclosing matter represented by ik and using the Gau-Stokes theorem gives ik dSk = 0, (2.4.48) which gives the conservation of the four-momentum vector Pi = 1 c ik dSk = const. (2.4.49)

Choosing the volume hypersurface dV = dS0 gives Pi = 1 c i0 dV, (2.4.50)

so the components 1 i0 form the four-momentum density. The component 00 , rec ferred to as the energy density, m m , W = 00 = (2.4.51)

integrated over the volume gives the time component of the four-momentum, the energy L L, E = cP0 , cP 0 = 00 dV = (2.4.52) where L=

m dV

(2.4.53)

is the Lagrange function or Lagrangian. Hereinafter, a dot above any quantity = d , and two dots above denotes the partial derivative of with respect to time, dt = d22 denote the second derivative of with respect to time, . Consequently, the dt action of a physical system is the time integral of the Lagrangian, S= 69 Ldt. (2.4.54)

 0 , referred to as the momentum density, integrated over the The components 1 c volume give the spatial components of the four-momentum, the momentum vector P : P = 1 c 0 dV. (2.4.55)

Adding a total divergence j ikj to ik does not alter the denition of the fourmomentum vector (2.4.49). The symmetry of ik can be written as l (xi kl xk il ) = 0, (2.4.56)

which upon the integration over a hypersurface enclosing matter represented by ik and using the Gau-Stokes theorem gives (xi kl xk il )dSl = 0, which gives the conservation of the angular momentum tensor M ik = (xi dP k xk dP i) = 1 c 1 c (xi kl xk il )dSl = const. (2.4.58) (2.4.57)

Choosing the volume hypersurface dV = dS0 gives M ik = The conservation of M 0 , M 0 = 1 c x 00 dV x0 0 dV = 1 c x 00 dV ctP = const, (2.4.60) (xi k0 xk i0 )dV. (2.4.59)

divided by the conservation of P 0 (2.4.50), P 0 = const, gives a uniform motion, X = V t + const, cP , P0 of the center of inertia with the coordinates X , V = X = x 00 dV . 00 dV with velocity (2.4.61)

(2.4.62)

(2.4.63)

The coordinates of the center of inertia (2.4.63) are not the spatial components of a four-dimensional vector. The conservation law (2.4.34) can be written as 1 00 0 + = 0, c t x 1 0 + = 0. c t x 70 (2.4.64) (2.4.65)

Integrating these equations over the volume hypersurface and using the Gau-Stokes theorem gives t t where
df = df0

00 dV = c 1 0 dV = c

0 df , df ,

(2.4.66) (2.4.67)

(2.4.68)

is the spatial surface element (1.1.30). The integral of a three-dimensional vector V over the two-dimensional surface element df , V df , is referred to as the ux of this vector. Therefore the components S = c 0 (2.4.69)

of the energy current S form, upon integrating over df , the energy ux. The components represent the momentum current and give, upon integrating over df , the momentum ux. The stress tensor is dened as = . The components of the energy-momentum tensor form the matrix ik = Dene the spatial surface force vector, F = df . (2.4.72) W
S c

(2.4.70)

S c

(2.4.71)

The relations (2.4.55), (2.4.67), (2.4.70) and (2.4.72) equal the time derivative of the momentum P to the surface force F , = F . P (2.4.73)

In an arbitrary frame of reference, the metric dynamical energy-momentum tensor Tik describing isotropic matter (without a preferred direction in its rest frame) can be decomposed into the part proportional to ui uk , the part proportional to the projection tensor, hik = gik ui uk , (2.4.74) which is orthogonal to ui , hik uk = 0, (2.4.75)

and parts containing covariant derivatives of ui . The projection tensor satises hij hj k = hik . 71 (2.4.76)

Assume that Tik does not depend on derivatives of ui. Therefore Tik = ui uk phik , (2.4.77)

where a scalar is equal to the energy density W in the locally Galilean rest frame and a scalar p is the pressure. In this frame T ik = diag(, p, p, p) and the stress tensor = p , giving F = p df = p n df, which states that the force per unit surface df acting on a surface is parallel, with the opposite sign, to the outward normal vector of this surface n , dF = pn , and df which is referred to as Pascals law. Matter described by the tensor (2.4.77) represents an ideal uid. The relation between and p is referred to as the equation of state. In the Galilean frame of reference, combining (1.6.115), (2.4.71) and (2.4.77) gives + pv 2 /c2 , 1 v 2 /c2 ( + p )v , S= 1 v 2 /c2 ( + p)v v = 2 p . c v2 W = The relation (2.4.77) gives T = T ii = 3p.

(2.4.78)

(2.4.79) (2.4.80) (2.4.81)

(2.4.82)
0

g00 dx +g0 dx 2 The component T00 = u2 , (1.4.96) and 0 + p(u0 g00 ) is, using u0 = ds (1.4.97), equal to dl 2 , (2.4.83) T00 = u2 0 + pg00 ds so it is positive under physical conditions > 0, p > 0 and g00 > 0. If Tik depends also on derivatives of ui then matter described by the tensor (2.4.77) with the corresponding additional terms represents a viscous uid.

2.4.7

Mass and Papapetrou equations of motion

Consider matter which is distributed over a small region in space and consists of points with the coordinates xi , forming an extended body whose motion is represented by a world tube in spacetime. The motion of the body as a whole is represented by an arbitrary timelike world line inside the world tube, which consists of points with the coordinates X i ( ), where is the proper time on . Dene xi = xi X i , x0 = 0, ui = Also dene the following integrals: M ik = u0 dX i . ds (2.4.84)

ik dV,
72

(2.4.85)

M ijk = u0 N ijk = u0 J ik =

xi jk dV,

(2.4.86) (2.4.87) 1 (M ik0 + M ki0 + N ik0 ).(2.4.88) 0 u

ijk dV,

(xi k0 xk i0 + ik0 )dV =

The quantity J ik is equal to (xi kl xk il + ikl )dSl taken for the volume hypersurface, so it is a tensor, which we call the total spin tensor. The quantity N ijk is also a tensor. The relation x0 = 0 gives M 0jk = 0. (2.4.89)

Assume that the dimensions of the body are small, so integrals with two or more factors xi multiplying jk and integrals with one or more factors xi multiplying jkl can be neglected. The conservation law for the tetrad energy-momentum density (2.4.32) is

ji,i + {ijk }ik Cik j ik Riklj ikl = 0.

1 2

(2.4.90)

Integrating (2.4.90) over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals gives

j 0,0 dV + {ijk }ik dV Cik j ik dV

Expanding
i(0) i(0)

1 2

Riklj ikl dV = 0.

(2.4.91)

jik = j k + j k,l xl ,

(2.4.92)

where the superscripts 0 denote the values at X i , and substituting these expressions into (2.4.91) gives (omitting the superscripts)

j 0 dV
Cik j,l

,0

+ {ijk }

xl ik dV

1 2

ik dV + {ijk },l xl ik dV Cik j

Riklj ikl dV = 0

ik dV
(2.4.93)

or, using the denitions (2.4.85), (2.4.86) and (2.4.87), d M j0 1 +{ijk }M (ik) {ijk },l M l(ik) Cik j M [ik] +Cik j,l M l[ik] Riklj N ikl = 0. (2.4.94) 0 ds u 2 The conservation law (2.4.90) gives 1 (xl ji),i = jl xl {ijk }ik + xl Cik j ik + xl Rikmj ikm , 2 (xl xm ji),i = xm jl + xl jm xl xm {ijk }ik + xl xm Cik j ik 1 + xl xm Riknj ikn . 2 73 (2.4.95)

(2.4.96)

Integrating (2.4.95) over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals gives (xl j 0 ),0 dV =

jl dV xl {ijk }ik dV + xl Cik j ik dV +

1 2

xl Rikmj ikm dV. (2.4.97)

Substituting (2.4.84) into (2.4.97) gives ul u0

j 0 dV + X l j 0,0 dV + (xl j 0 ),0 dV =

xl {ijk }ik dV + X l Rikmj ikm dV, Cik j ik dV +

jl dV X l {ijk }ik dV
(2.4.98)

xl Cik j ik dV

1 + Xl 2

which reduces, due to (2.4.91), to ul u0

j 0dV +

(xl j 0 dV

,0

(2.4.99) Substituting (2.4.92) into (2.4.99), omitting the superscripts and using the denitions (2.4.85), (2.4.86) and (2.4.87), turns (2.4.99) into ul j 0 d M lj 0 M u0 ds u0 = M jl + {ijk }M lik Cik j M lik . (2.4.100)

jl dV xl {ijk }ik dV + xl Cik j ik dV.

Putting l = 0 in (2.4.100) gives the identity because of (2.4.89). Integrating (2.4.96) over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals gives (xl xm j 0 ),0 dV = + xm jl dV + 1 2 xl jm dV xl xm {ijk }ik dV (2.4.101)

xl xm Cik j ik dV +

xl xm Riknj ikn dV.

Substituting (2.4.84) into (2.4.97) gives X lX m + um u0

j 0,0 dV +

ul m X u0

j 0 dV +

ul u0

xm j 0 dV + xl j 0,0 dV

um l X u0

j 0 dV

xl j 0 dV + X l

xm j 0,0 dV + X m Cik j ik dV 1 2

= X l X m +X l +X m +

{ijk }ik dV

Riklj ikl dV

jmdV xm {ijk }ik dV + xm Cik j ik dV jl dV xl {ijk }ik dV + xl Cik j ik dV

xl jm dV, 74 (2.4.102) xm jl dV +

which reduces, due to (2.4.91) and (2.4.99), to ul u0 or xm j 0 dV + um u0 xl j 0dV = xm jl dV + xl jm dV (2.4.103)

ul mi0 um li0 M + 0 M = M mil + M lim . (2.4.104) u0 u The expressions analogous to (2.4.95) and (2.4.96) with higher multiples of xi do not introduce new relations. The conservation law for the angular momentum density (2.4.15) is

ijk,k lik jlk + ljk ilk 2[ij ] = 0.

(2.4.105)

Integrating (2.4.105) over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals gives

ij 0,0 dV lik jlk dV + ljk ilk dV 2 [ij ]dV = 0.

(2.4.106)

Substituting (2.4.92) into (2.4.106), omitting the superscripts and using the denitions (2.4.85) and (2.4.87), turns (2.4.106) into d N ij 0 lik N jlk + ljk N ilk 2M [ij ] = 0. ds u0 The conservation law (2.4.105) gives (xl ijk ),k = ijl + xl lik jlk xl ljk ilk + 2xl [ij ]. (2.4.108) (2.4.107)

Integrating (2.4.108) over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals gives (xl ij 0 ),0 dV =
j i jmk imk ijl dV + xl m dV xl m dV + 2 xl [ij ] dV. k k

(2.4.109)

Substituting (2.4.84) into (2.4.109) gives ul u0 +2

ij 0 dV + X l ij 0,0 dV = [ij ]dV + 2 xl [ij ]dV,

ijl dV + X l

jmk i m dV k

j imk m dV k

(2.4.110)

which reduces, due to (2.4.106), to ul u0 or M l[ij ] =

ij 0dV =

ijl dV + 2 xl [ij ]dV

1 ul ij 0 N N ijl . 2 u0 75

(2.4.111)

(2.4.112)

Putting l = 0 in (2.4.112) gives the identity because of (2.4.89). The expressions analogous to (2.4.108) with higher multiples of xi do not introduce new relations. Taking the cyclic permutations of the indices i, l, m in (2.4.104), adding the rst and second of these relations, and subtracting the third, gives ul [mi]0 ui [ml]0 um (li)0 M + 0M + 0M = M l[im] + M i[lm] + M m(il) . 0 u u u Substituting (2.4.88) and (2.4.112) into (2.4.113) gives M m(il) = u(i J l)m + um (il)0 M + N m(il) . 0 u (2.4.114) (2.4.113)

Putting m = 0 in (2.4.114), substituting it into (2.4.114) and using (2.4.89) gives M m(il) = u(i J l)m um (i l)0 (u J + N 0(il) ) + N m(il) . u0 (2.4.115)

Combining (2.4.112) and (2.4.115) gives M mil = u(i J l)m Therefore Combining the antisymmetric part of (2.4.100) and (2.4.107) gives ul j 0 uj l 0 M M = 0 M 0 M ({ijk } Cik j )M lik + ({ilk } Cik l )M jik u u d M lj 0 M jl0 . (2.4.118) ds u0
jl lj

um (i l)0 1 um il0 0(il) m(il) ( u J + N ) + N N N ilm . u0 2 u0 M (ik)0 = (u(i J k)0 + N 0(ik) ).

(2.4.116)

(2.4.117)

Using (2.4.88), (2.4.112) and (2.4.115) brings (2.4.118) to ijk N lik + ilk N jik = 1 Cik j 2 1 + Cik l 2 which, using ul j 0 uj l 0 ul i k 0 j i kl lik M M { } u J + N (u J + N 0ik ) i k 0 0 0 u u u j u ul ik0 N N ikl + {ilk } ui J kj + N jik 0 (uiJ k0 + N 0ik ) u0 u j d u ik0 N N ikj + J lj , (2.4.119) 0 u ds
D {} lj J ds

d lj J ds

+ uk {ilk }J ij + uk {ijk }J li , turns into

D {} lj uj 1 uj uj J = 0 M l0 + 0 {ilk }(uiJ k0 + N 0ik ) Cik l 0 N ik0 N ikj + C lik N jik l j ds u u 2 u (2.4.120) or, using the four-momentum Pl = 1 c 76

l0 dV,

(2.4.121)

into uj 1 uj D {} lj J = cuj P l + 0 {ilk }(uiJ k0 + N 0ik ) Cik l 0 N ik0 N ikj + C lik N jik l j . ds u 2 u (2.4.122) Therefore 1 1 D {} J li 1 ui = cP l + 0 {ilk }(uiJ k0 + N 0ik ) Cik l 0 N ik0 N ikj uj + C lik N jik uj ds u 2 u l l u uj 1 u cul uj P j 0 {ijk }(uiJ k0 + N 0ik ) + Cik j 0 N ik0 N ikl uj u 2 u j lik C ik N uj , (2.4.123) which gives with (2.4.122) D {} J lj D {} J ji D {} J li 1 + ul ui uj ui = C lik N jik + Cik l N ikj C jik N lik ds ds ds 2 1 1 1 j ikl Cik N uj um C lik N mik + Cik l N ikm C mik N lik Cik m N ikl 2 2 2 1 1 +ul um C jik N mik + Cik j N ikm C mik N jik Cik m N ikj 2 2 1 j j j l l n mik (2.4.124) = 2([ln m + Cik n N ikm . ] u u[m n] + u u[m n] ) C ik N 2 Multiplying (2.4.124) by uj gives the identity, so only 3 equations in (2.4.124) are independent. Thus 3 components of J ik are arbitrary and we can impose 3 constraints on J ik . A simple choice is J ik uk = 0, (2.4.125) which means that in the local rest frame J 0 = 0, so the three independent components of J ik are the spatial J . Analogously to the Pauli-Luba nski pseudovector (1.6.70), dene the four-spin pseudovector 1 J i = eijkl uj Jkl , 2 which is orthogonal to ui , J i u i = 0. The condition (2.4.125) gives the relation inverse to (2.4.126): J ik = eikjl uj Jl . (2.4.128) (2.4.127) (2.4.126)

Dierentiating (2.4.126) covariantly with respect to {jik } and using (2.4.124) gives 1 1 D {} uj D {} J i n mpr l + Cpr n N prm = eijkl eklmn um J n + eijkl uj [k n m] C pr N ds 2 ds 2 {} k D u 1 = ui Jk + eij nm uj C npr N mpr + Cpr n N prm ds 2 {} k {} i D u 1 D u = ui Jk + uk Jk + eij nm uj C npr N mpr + Cpr n N prm . (2.4.129) ds ds 2 77

Thus the covariant (with respect to the Levi-Civita connection) change of the spin pseudovector along the world line is the sum of the corresponding Fermi-Walker transport (with respect to the Levi-Civita connection) and a term which depends on the torsion and spin density. In (2.4.123), the four-momentum P l depends on terms proportional to the fourvelocity ul and terms in which the index l appears in other quantities. Dene the mass of the system described by the energy-momentum density ik as the coecient l , m of ul in the expansion for P c Pl = mul + . . . , c so m= where uj uj j uj uj P + 2 0 {ijk }(uiJ k0 + N 0ik ) 2 0 Cik j N ik0 = j , c cu 2c u c j = P j + (2.4.131) (2.4.130)

1 j 1 i k0 0ik { } ( u J + N ) Cik j N ik0 (2.4.132) i k 0 0 cu 2cu is the modied four-momentum. Substituting (2.4.131) and (2.4.132) into (2.4.123) gives D {} li 1 1 J ui = cl mc2 ul + uj C lik N jik + Cik l N ikj uj C jik N lik + Cik j N ikl , ds 2 2 (2.4.133) l l j l l so mcu = (j u uj ) is a vector. Thus the modied four-momentum i is a vector and the mass m is a scalar. Substituting (2.4.132) into (2.4.122) gives the Papapetrou equation of motion for the spin: 1 1 D {} lj J = cuj l cul j + C lik N jik + Cik l N ikj C jik N lik Cik j N ikl . (2.4.134) ds 2 2 Putting (2.4.100), (2.4.107), (2.4.112), (2.4.115), (2.4.117) and (2.4.132) into (2.4.94) gives 1 d 1 1 cj 0 {ijk }(uiJ k0 + N 0ik ) + 0 Cik j N ik0 + {ijk }uk ci 0 {ljm }(ul J m0 ds u 2u u (ik )0 1 d M +N 0lm ) + 0 Clm j N lm0 {ijk }({lim } Clm i )M klm {ijk } 2u ds u0 ul 1 d N ik0 {ijk },l u(i J k)l 0 (u(i J k)0 + N 0(ik) ) + N l(ik) Cik j u 2 ds u0 1 1 ul 1 ilm ) Cik j,l 0 N ik0 N ikl Riklj Nikl Cik j (lim N klm + lk mN 2 2 u 2 D {} j =c {ijk }{lim }(ul J mk + N klm ) {ijk },m (uiJ km + N mik ) ds 1 1 1 j (2.4.135) + {i k }Clm i N lmk + Cik j lim N klm + Cik j,m N ikm Riklj Nikl = 0. 2 2 2 78

Using (1.4.51) turns (2.4.135) into the Papapetrou equation of motion for the momentum: 1 1 D {} j = P jimk ui J mk Nikl C ikl:j . (2.4.136) ds 2c 2c If the spin density vanishes then the Einstein-Cartan gravitational eld equations reduce to the Einstein-Hilbert gravitational eld equations. The conservation law for the spin density (2.4.15) with the condition ijk = 0 gives the symmetry of the energy-momentum density, ik = ki . The relations (2.4.85), (2.4.86), (2.4.87) and (2.4.88) give then M ik = M ki , M ijk = M ikj , N ijk = 0, J ik = cLik = (xi k0 xk i0 )dV = (2.4.137) (2.4.138) (2.4.139) 1 (M ik0 + M ki0 ), (2.4.140) 0 u

where Lik is the angular momentum tensor, analogous to (2.4.58). The modied four-momentum (2.4.132) reduces to j = P j + and (2.4.133) gives l = mcul + The relation (2.4.129) reduces to D {} J i D {} uk D {} ui k = ui Jk + u Jk , ds ds ds (2.4.143) D {} Lli ui . ds (2.4.142) 1 j i k0 { }u L u0 i k (2.4.141)

so the covariant (with respect to the Levi-Civita connection) change of the spin pseudovector along the world line is equal to the corresponding Fermi-Walker transport. Multiplying (2.4.143) by Ji and using (2.4.127) gives J i Ji = const, (2.4.144)

so the change of the spin pseudovector along a world line is a rotation, called spin precession. The Papapetrou equation of motion for the spin (2.4.134) reduces to D {} Llj = uj l ul j , ds (2.4.145)

while the Papapetrou equation of motion for the momentum (2.4.136) reduces to 1 D {} j = P jimk uiLmk . ds 2 (2.4.146)

79

The change of the mass m along the world line is, using (1.4.48), (2.4.125), (2.4.131), (2.4.142) and (2.4.146), D {} m 1 D {} j 1 j D {} uj 1 D {} uj 1 D {} Lji D {} uj dm = = uj + = j = ui ds ds c ds c ds c ds c ds ds {} {} 1 D ui D uj = Lji = 0, (2.4.147) c ds ds so m = const. (2.4.148) In the absence of the external gravitational eld and neglecting the gravitational eld of the body, the relation (2.4.141) gives j = P j , so (2.4.146) reduces to dP j = 0, (2.4.150) ds whose integration gives the conservation of the four-momentum along a world line: P i = const. The equation of motion for the spin (2.4.145) becomes dLlj = uj P l ul P j , ds (2.4.152) (2.4.151) (2.4.149)

whose integration gives the conservation of the angular momentum along a world line: Lik + X i P k X k P i = const. (2.4.153)

The tensor Lik is the intrinsic angular momentum of the body, while the tensor (in the absence of the gravitational eld) X i P k X k P i is the orbital angular momentum associated with the motion of the body as a whole. If Lik = 0 then (2.4.153) gives P i ui , so (2.4.151) is equivalent to ui = const and thus X i is a linear function of the proper time . If Lik = 0 then X i can be given by 3 arbitrary functions of (since ui ui = 1). In the momentum rest frame, in which P = 0, u = 0, so the body has an arbitrary internal motion. The 3 constraints (2.4.125) eliminate this arbitrariness, so the equations of motion entirely determine the motion of the body. 2.4.8 Energy-momentum tensor for particles

If the body is not spatially extended then it is referred to as a particle. The corresponding condition x = 0 gives M ijk = 0, Lik = 0. (2.4.154)

80

Therefore (2.4.112) reduces to

ul N ij 0 u0

N ijl = 0, which with (2.4.88) gives (2.4.155) (2.4.156)

N ijk = uiJ jk , so J ij = S ij = N ijk uk ,

where S ij is the intrinsic spin tensor. If the body is spatially extended then the dierence Rik = J ik S ik (2.4.157) is the rotational spin tensor. The dierence between the rotational spin tensor and angular momentum tensor is, due to (2.4.88) and (2.4.112), N ik0 R cL = 0 N ikl ul = 2M l[ik] ul . u
ik ik

(2.4.158)

This expression vanishes, because of (2.4.89), in the velocity rest frame, in which u = 0, which is also locally Galilean, so u = 0. If a particle is spinless then its four-momentum is proportional to its four-velocity due to (2.4.141) and (2.4.142): P l = mcul , (2.4.159) which gives P 2 = m2 c2 , (2.4.160) in agreement with (1.6.80). Equations (2.4.100), (2.4.121), (2.4.137), (2.4.139) and (2.4.159) give ui ui uk M ik = 0 M k0 = 0 2 M 00 = mc2 uiuk , (2.4.161) u (u ) so ui uk (2.4.162) ik dV = mc2 0 u or ui uk (2.4.163) ik (x) = mc2 (x x0 ) 0 , u where (x x0 ) is the spatial Dirac delta representing a point mass located at x0 . Dene the mass density such that dV = dm, (2.4.164) where is given by (1.4.105). The mass density for a particle located at xa is m (x) = (x xa ),

(2.4.165)

so (2.4.163) turns into

ik = c2
81

ui uk . u0

(2.4.166)

Thus the energy-momentum tensor for a spinless particle is given by ui uk c dxi dxk T ik = c2 = g00 u0 g00 ds dt or, for a system of particles, T ik (x) =
a

(2.4.167)

ui uk . ma c2 (x xa ) u0

(2.4.168)

The Papapetrou equation of motion (2.4.146) for a spinless particle reduces to the metric geodesic equation (1.4.80), D {} ui = 0. ds (2.4.169)

In the absence of torsion and in the locally Galilean frame of reference, the conservation law for the energy-momentum tensor is given by (2.4.34), so T i,i = 0. (2.4.170)

Consider a closed system of particles which carry out a nite motion, in which all quantities vary over nite ranges. Dene the average over a certain time interval = 1 f dt. The average of the derivative of a of a function f of these quantities as f 0 = 1 (f ( ) f (0)) 0 as . Thus averaging (2.4.170) over bounded quantity f the time gives = 0. T (2.4.171) , Multiplying (2.4.171) by x and integrating over the volume gives, omitting surface integrals, dV = T dV = 0. x T (2.4.172) , The average energy of the system (2.4.52) is thus = E 0 0 dV = T i i dV. T (2.4.173)

Substituting (1.6.115) into (2.4.168) gives Ti i (x) = ma c2 (x xa ) 1 v2 c2

1/2

(2.4.174)

so Ti i 0. Putting (2.4.174) into (2.4.173) gives = E

a

ma c2 1

v2 c2

1/2

(2.4.175)

which is referred to as the virial theorem. 82

Comparing (2.4.82) with (2.4.174) gives 3p =

a

ma c2 1

v2 c2

1/2

(2.4.176)

where the summation extends over all particles in unit volume, so p /3. In the nonrelativistic limit p 0, while in the ultrarelativistic limit (v c) p /3. Consider a system of noninteracting identical particles of mass m, which we call an ideal gas, with the number of particles in unit volume (concentration) n, so = nm. (2.4.177) Comparing (2.4.77) in the locally Galilean rest frame with (2.4.167) gives the kinetic formulae for ideal gases: = nmc2 , nm 2 v . p= 3 In a locally inertial frame of reference, (1.6.115) and (2.4.159) give v P i = mc 1, , c so the energy and momentum of the particle are E = mc2 , P = m v. Thus (2.4.160) gives E 2 = (P c)2 + (mc2 )2 . We also obtain v dP = = cu cu cu d(mcu ) = d(mcu ) = d(mcu )g (2.4.184) (2.4.183) (2.4.178) (2.4.179)

(2.4.180)

(2.4.181) (2.4.182)

cu0 d(mcu0 )g 00 = cdP0 = dE.

In the rest frame of the particle, P = 0, (2.4.183) reduces to Einsteins formula for the rest energy, E = mc2 . (2.4.185) The formulae (2.4.181) and (2.4.182) give c2 p. E If a particle is massless, m = 0, then (2.4.183) gives v= E = P c, which is consistent with (2.4.186) only if v = c. References: [2, 3, 5, 6, 7, 9]. 83 (2.4.188) (2.4.186)

(2.4.187)

2.5
2.5.1

Gravitational eld equations

Einstein-Hilbert action and Einstein equations

The Einstein-Hilbert action for the gravitational eld and matter is, due to (2.2.2), S= 1 2c P d + Sm , (2.5.1)

where the metric tensor is regarded as a variational variable and the ane connection is the Levi-Civita connection. Varying (2.5.1) with respect to the metric tensor gik g ik (which results from gives, using (2.3.1) and the identity = 1 2 = g ik gik = gik g ik ), S = (2.5.2) Partial integration of the rst term on the right-hand side of (2.5.2), using (1.4.57), brings this term to zero: Pik gik d = where gik = ( {ilk }):l ( {ill }):k gik d = (gik:l {ilk } gik:k {ill })d = 0, g ik 1 2c 1 1 Pik g ik + Pik g ik P gik g ik d+ 2 2c Tik g ik d.

(2.5.3)

(2.5.4)

is the contravariant metric density, whose covariant derivative with respect to the Christoel symbols vanishes, gik:l = 0. Equaling S = 0 in (2.5.2) gives the Einstein equations of the general theory of relativity: Gik = Tik or (2.5.5)

1 Pik = Tik T gik . (2.5.6) 2 Because P d = G d, where the noncovariant quantity G is given by (2.2.5), the left-hand side of the Einstein equations is 1 ( G) . (2.5.7) Gik = g ik

The covariant conservation of the Einstein tensor (1.4.67) imposes the conservation of the metric dynamical energy-momentum tensor (2.4.23). Therefore the gravitational eld equations contain the equations of motion of matter. In vacuum, where Tik = 0, the Ricci tensor in (2.5.6) vanishes: Pik = 0. (2.5.8)

Thus vanishing of Pik at a given point in spacetime is a covariant criterion of whether matter is present or absent at this point. 84

The Einstein equations (2.5.5) are 10 second-order partial dierential equations for: 10 4 = 6 independent components of the metric tensor gik (the factor 4 is the number of the coordinates which can be chosen arbitrarily), 3 independent components of the four-velocity ui , and either or p (which are related to each other by the equation of state). The contracted Bianchi identity (1.4.67) gives the equations of motion of matter. In vacuum, the Einstein equations are 10 4 = 6 independent equations (the factor 4 is the number of constraints from the contracted Bianchi identity) for 6 independent components of the metric tensor gik . In the Einstein equations, the only second time-derivatives of gik are the derivatives of the spatial components of the metric tensor, g , and they appear only in the components of the eld equations (2.5.5). Therefore the initial values (at t = 0) for g and g can be chosen arbitrarily. The rst time-derivatives g 0 and g 00 appear only in the components of the eld equations (2.5.5). The 0 and 00 components of the eld equations (2.5.5) give the initial values for g0 and g00 . The undetermined initial values for g 0 and g 00 correspond to 4 degrees of freedom for a free gravitational eld. A general gravitational eld has 8 degrees of freedom: 4 degrees of freedom for a free gravitational eld, 3 related to the four-velocity, and 1 related to (or p). 2.5.2 Dene G= Einstein pseudotensor and principle of equivalence gG, (2.5.9)

where G is the noncovariant quantity (2.2.5). The action for the gravitational eld and matter, 1 1 1 (2.5.10) S= G + m d, Gd + Sm = 2c c 2 produces the Einstein eld equations by varying the metric tensor, because G diers from gP by a total divergence: 1 G + m = 0. ik g 2 (2.5.11)

Construct a canonical energy-momentum density (2.4.35) corresponding to the gravitational eld, treating 21 G (which depends only on g ij and its rst derivatives g ij,k ) like m and g ik like a matter eld :

ki =

1 jl i jl g ,k k 2 g ,i

(2.5.12)

This quantity is not a tensor density since is not a scalar density and its division ki , is referred to as the Einstein energy-momentum pseudotensor for the by , gravitational eld. The four-momentum corresponding to the total energy-momentum density for the gravitational eld and matter (which is not a vector) is then Pi = 1 c (ik + Ti k )dSk , 85

(2.5.13)

where the sum ik + Ti k is called the Einstein energy-momentum complex. The denition (2.5.12) gives jl jl jl jl g jl,k jl g ,k jl g ,ki + ,k = i jl g ,k jl g ,ki + jl g g ,i g ,i g ,i g ,i i jl g jl,k = jl g jl,k , (2.5.14) + jl g jl,ik = jl g g g ,i g ,i 2ki,i = i which, using (2.5.11), gives 1 m jl g = Tjl g jl,k . ,k g jl 2

ki,i =

(2.5.15)

The covariant conservation (2.4.23) gives 1 1 i Tk ,i = {kl i }Tl i = g lm gim,k Tl i = g lm,k Tlm , 2 2 (2.5.16)

so the total energy-momentum density for the gravitational eld and matter is ordinarily conserved: (2.5.17) (ki + Tk i ),i = 0. Integrating (2.5.17) over the four-dimensional volume and using the Gau-Stokes theorem gives (ki + Tk i )dSi = 0, (2.5.18)

so the four-momentum (2.5.13) is conserved, Pi = const. Because the quantity ik is not symmetric in the indices i, k , the total angular momentum constructed from P i as in (2.4.58), M ik = (xi dP k xk dP i) = 1 c (xi (kl + T kl ) xk (il + T il ))dSl ,

(2.5.19)

is not conserved. The conservation law (2.5.17) gives kl + Tk l = k li,i , where kli = kil , so kl lk = (k li lk i ),i . Analogously to (2.4.44) and (2.4.45), we could bring kl + Tk l to a symmetric form. However, using (k li lk i ) instead of kli in (2.4.45), where ik is replaced by ik + Ti k , gives i k = 0, so this symmetrization procedure does not work for the Einstein pseudotensor. The Einstein pseudotensor (2.5.12) can be explicitly written as

k i =

1 i l mi i ({l m }glm,k {m l }g ,k + k ), 2

(2.5.20)

so it is a homogeneous quadratic function of the Christoel symbols. Thus it vanishes in the local Galilean frame of reference. It can also dier from zero in the Minkowski spacetime (in the absence of the gravitational eld) if we choose the coordinates such that the Christoel symbols do not vanish. Therefore the energy of the gravitational eld is not absolutely localized in spacetime; it depends on the choice of the coordinates. The gravitational eld can be always eliminated locally by transforming 86

the coordinate system to the local Galilean frame of reference in which the Einstein pseudotensor vanishes. This property of the gravitational eld is referred to as the principle of equivalence. The construction of a conserved four-momentum for the gravitational eld and matter is possible because the Lagrangian density for the gravitational eld g is linear in the second derivatives of the metric tensor. The Lagrangian density (2.2.2) can be generalized to 1 g = (P + 2), (2.5.21) 2 where is referred to as the cosmological constant, without altering the Einstein energy-momentum pseudotensor (2.5.12). Another scalar density which is linear in curvature is ijkl Pijkl , but this parity-violating expression vanishes due to the cyclic identity (1.4.64). Therefore the simplest choice for a gravitational Lagrangian density, linear in P , is the only one that admits ordinary conservation laws for the gravitational eld and matter, and thus physical. 2.5.3 Landau-Lifshitz energy-momentum pseudotensor

The covariant conservation (2.4.23) in the local Galilean frame of reference is

i Tk ,i = 0,

(2.5.22)

so T ik can be expressed as T ik = ikl,l , where ikl = ilk . The Einstein equations (2.5.5) in the Galilean frame are ( )T ik = hikl,l , where hikl = iklm,m = hilk , 1 iklm = ( )(g ik g lm g il g km). 2 In an arbitrary frame of reference, (2.5.23) is not valid. Dene tik such that ( )(tik + T ik ) = hikl,l . Therefore (( )(tik + T ik )),k = 0, Pi = 1 c ( )(tik + T ik )dSk . (2.5.27) so there is a conservation of the four-momentum of the gravitational eld and matter, (2.5.28) (2.5.26) (2.5.24) (2.5.25) (2.5.23)

The quantity tik is not a tensor density, so the conserved four-momentum P i (2.5.28) is not a vector. The four-momentum P i is not a vector even for Lorentz transfor mations, because of the factor instead of the correct (weight 1) density in 87

 (2.5.28). Dividing P i by at some xed point (a natural choice is innity) turns it into a vector under Lorentz transformations. Using (2.5.26) turns (2.5.28), for the hypersurface dS0 = dV , into 1 1 1 hikl,l dSk = hikl dfkl = hi0 df . (2.5.29) Pi = c 2c c The quantity tik is referred to as the Landau-Lifshitz energy-momentum pseudotensor for the gravitational eld, and the sum ( )(tik + T ik ) is called the Landau-Lifshitz complex. The explicit expression for the Landau-Lifshitz pseudotensor is 1 p n p n p tik = (g il g km g ik g lm )(2{ln m }{n p } {l p }{m n } {l n }{m p }) 2 p p p k k k p +g il g mn ({lk p }{m n } + {m n }{l p } {n p }{l m } {l m }{n p })
p p p i i i p +g kl g mn ({lip }{m n } + {m n }{l p } {n p }{l m } {l m }{n p }) k i k +g lm g np({lin }{m p } {l m }{n p })

(2.5.30)

or

1 1 ik lm g ,l g ,m gil,l gkm,m + g ik glm gln,p gpm,n 2 2 mp mp il kn kl in (g gmn g ,p g ,l + g gmn g ,p g ,l ) + glm g np gil,n gkm,p 1 + (2g il g km g ik g lm)(2gnp gqr gpq gnr )gnr,l gpq,m . (2.5.31) 8 This pseudotensor is symmetric in the indices i, k , so there is a conservation of the total angular momentum constructed from P i as in (2.4.58), 1 (xi (tkl + T kl ) xk (til + T il ))( )dSl . (2.5.32) M ik = (xi dP k xk dP i ) = c Dividing M ik by at innity turns it into an antisymmetric tensor under Lorentz transformations. Using (2.5.24) and (2.5.26) turns (2.5.28), for the hypersurface dS0 = dV , into 1 1 M ik = (xi klmn,nm xk ilmn,nm )dSl = (xi klmn,n xk ilmn,n )dflm c 2c 1 1 klin ilkn i k 0 k i0 ( ),n dSl = (x h x h + i0k )df . (2.5.33) c c Choosing the volume hypersurface dV = dS0 gives 1 M ik = (xi (tk0 + T k0 ) xk (ti0 + T i0 ))( )dV. (2.5.34) c The conservation of M 0 in (2.5.34) divided by the conservation of P 0 in (2.5.28) gives a uniform motion (2.4.61) with velocity (2.4.62) of the center of inertia for the gravitational eld and matter, with the coordinates ( )tik = X = x (t00 + T 00 )( )dV . (t00 + T 00 )( )dV (2.5.35)

The coordinates of the center of inertia (2.5.35), like (2.4.63), are not the spatial components of a four-dimensional vector. 88

2.5.4

Utiyama action

The Utiyama action for the gravitational eld and matter is equal to (2.5.1), where the tetrad is regarded as a variational variable and the spin connection is the Levi-Civita spin connection (1.5.31). Thus (2.3.5) gives S = 1 2c ( P )d + 1 c

ia eia d.

(2.5.36)

The Lagrangian density for the gravitational eld is given by (2.2.2), with the Riemann scalar P given by (1.5.39) and (1.5.41):
jb a a a c a c P = ei a e ( bj,i bi,j + ci bj cj bi ) = 2 ij ab ab ( j,i

+ aci cbj ), (2.5.37) (2.5.38)

where

ij ab

= e[ai eb .

j]

Varying P and omitting total derivatives gives in the absence of torsion, using = ij ij ij c c ij a ei a ei and ab|j = ab,j aj cb bj ac = 0 (which results from (1.5.22)), +2(
i ( P ) = (2P ai P ea i ) ea + 2 ij ab,j

caj

ij cb

ij ab a cb ab ( j,i + ci j ) = ab a a i cbj ij ac ) i = (2P i P ei ) ea .

i (2P ai P ea i ) ea

(2.5.39)

Equaling S = 0 gives the tetrad Einstein equations: 1 a P ai P ea i , i = 2 (2.5.40)

equivalent to the metric Einstein equations (2.5.5) because of (2.3.4) and (2.3.32) (in the absence of torsion). 2.5.5 Mller pseudotensor

The Riemann scalar P is linear in derivatives of abi :

j ac b i j ab i j ab i j ac b ei a eb j c i = 2( ea eb j ),i 2( ea eb ),i j + ea eb i c j j ac b ei a eb j c i . j ab i j ab i j ab i j ab i j ac b P = ( ei a eb j ),i ( ea eb ),i j ( ea eb i ),j + ( ea eb ),j i + ea eb i c j

(2.5.41)

Thus we can subtract from P total derivatives without altering the eld equations, replacing P by a noncovariant quantity M : = 2 ({kki } ijj + iai ajj {kii } kjj + jbi ibj {kji } ikj ) + ( ici c jj icj c ji ) = ( iai jaj iaj jai ), M = iai jaj iaj jai . 89 using (1.4.35) and (1.5.31). Therefore (2.5.43)
j ab i j ac b i j ac b M = 2( ei a eb ),i j + ea eb i c j ea eb j c i

(2.5.42)

Because of (1.5.15), the quantity (2.5.43) depends on the tetrad ei a and its rst derivatives ei . Therefore, analogously to (2.5.12), we can construct a canonical energya,j M: momentum density corresponding to the gravitational eld, treating 21

k i =

1 ( M ) j i M . ea,k k 2 ej a,i

(2.5.44)

This quantity is not a tensor density since M is not a scalar density and its division by i , k , is referred to as the Mller energy-momentum pseudotensor for the gravitational eld. One can show, analogously to the steps leading to (2.5.17, that the total energymomentum density for the gravitational eld and matter is ordinarily conserved: (ki + Tk i ),i = 0. Thus the corresponding total four-momentum is conserved: Pi = 1 c (ik + Ti k )dSk = const, (2.5.46) (2.5.45)

where the sum ik + Ti k is called the Mller energy-momentum complex. The Mller pseudotensor depends on the choice of both the coordinates and the tetrad. To x the tetrad, one can impose on it 6 constraints which are covariant under constant Lorentz transformations but not under general Lorentz transformations (otherwise these constraints would not x the tetrad since Lorentz transformations are tetrad rotations). A natural choice is to constrain the 6 components of the spin connection ijk in which the last index is contracted with a covariant derivative or the trace of the spin connection. 2.5.6 Einstein-Cartan action

If we regard the torsion tensor as a variational variable (in addition to the metric tensor) then the action for the gravitational eld and matter is, due to (2.2.1), S= 1 2c R d + Sm , (2.5.47)

and it is referred to as the Einstein-Cartan action. Using (1.4.54) gives S= 1 2c P g ik (2C lil:k + C jij C lkl C lim C mkl ) d + Sm . (2.5.48)

Partial integration of the terms with covariant derivatives : and omitting total derivatives (which do not contribute to the eld equations) reduces (2.5.48) to S= 1 2c P g ik (C jij C lkl C lim C mkl ) d + Sm . (2.5.49)

Varying (2.5.49) with respect to the metric tensor and contortion tensor (which is equivalent to varying with respect to the torsion tensor) gives, using (2.3.19), (2.3.26) 90

and (2.5.3), 1 1 Pik P gik C jij C lkl + C lim C mkl + gik (C jmj C lml C ljm C mjl ) 2 2 1 1 kj lj k (C i C l i ) C ijk d + Tik g ik d g ik d c 2c 1 j ik C ik d. (2.5.50) + sj 2c S = 1 2c For variations g ik , S = 0 gives the rst Einstein-Cartan equation Gik = (Tik + Uik ), where Uik = or Uik = 1 1 l j (S lij + 2S(ij ) )(S jkl + 2S(kl) ) + 4Si Sk + gik (S mjl + 2S (jl)m ) 2 (2.5.53) 1 1 C jij C lkl C lij C jkl gik (C jmj C lml C mjl Cljm ) 2 (2.5.52) (2.5.51)

(Sljm + 2S(jm)l ) 2gik S j Sj . For variations C jik , S = 0 gives the second Einstein-Cartan equation
l C k[ji] [k i C j ]l =

k s 2 ij

(2.5.54) (2.5.55) (2.5.56)

or where

T jik = sikj , 2
j j T jik = S jik Si k + Sk i

is the modied torsion tensor. The relation (2.5.55) is equivalent to l S kij = (sij k + [k (2.5.57) i s j ] l ). 2 This relation between the torsion and spin tensors is algebraic: torsion at a given point in spacetime does not vanish only if there is matter at this point, represented in the Lagrangian density by a function which depends on torsion. Unlike the metric, which is related to matter through a dierential eld equation, torsion does not propagate. Combining (2.5.52) and (2.5.57) gives 1 1 1 Uik = sij [l sklj ] sijl skjl + sjli sjlk + gik (4slj [msjml] + sjlmsjlm ) . (2.5.58) 2 4 8 The tensor (2.5.58) represents a correction to the dynamical energy-momentum tensor from the spin contributions to the geometry of spacetime, quadratic in the spin density (so the sign of the spin density does not aect this correction) and corresponding to a spin-spin contact interaction. If matter elds do not depend on torsion then Uik = 0 and the rst Einstein-Cartan equation (2.5.51) reduces to the Einstein equations (2.5.5). 91

2.5.7

Kibble-Sciama action

The Kibble-Sciama action for the gravitational eld and matter is equal to (2.5.47), where both the tetrad and spin connection are regarded as variational variables. Thus (2.3.27) gives S = 1 2c ( R )d + 1 c

iaeia d +

1 2c

abi abi d.

(2.5.59)

The Lagrangian density for the gravitational eld is given by (2.2.1), with the curvature scalar R given by (1.5.36) and (1.5.38):
jb a a a c a c R = ei a e ( bj,i bi,j + ci bj cj bi ) = 2 ij ab ab ( j,i

+ aci cbj ).
ij ab,j

(2.5.60) cbj
ij ac

Varying R and omitting total derivatives gives, using ij ab|j = j ik i kj k ij k j ab + k j ab k j ab = 0 (which results from (1.5.22)),
i ( R) = (2Rai Rea i ) ea + 2 i = (2Rai P ea i ) ea i = (2Rai Rea i ) ea

caj

ij cb

ij ab a cb ab ( j,i + ci j ) ij c c ij ab + 2( ij ab,j aj cb bj ac ) i ij ab 2(S ikj kj ab + 2Sj ab ) i .

(2.5.61)

For variations abi , S = 0 gives

i S iab Sa ei b + Sb ea =

i, 2 ab

(2.5.62)

equivalent to the second Einstein-Cartan equation (2.5.55). For variations ei a , S = 0 gives 1 a Rai Rea i (2.5.63) i = 2 or 1 (2.5.64) Rki Rgik = ik . 2 Substituting (2.5.55) and (2.5.64) into the conservation law for the spin density (2.4.16) gives
k k 2(S kij ;k Si;j + Sj ;i) = Rji Rij 4Sk (S kij Si j + Sj i ),

(2.5.65)

which is equivalent to the contracted cyclic identity (1.4.61). Thus the contracted cyclic identity imposes the conservation law for the spin density in the EinsteinCartan gravity. Substituting (2.5.55) and (2.5.64) into the conservation law for the energy-momentum density (2.4.31) gives 1 1 k 1 j j + 2S jki Rkj Rj Rj i;j R;i = 2Sj Rj i Ri (S jkl Sk lj + Sl k )Rklji, 2 2 2 (2.5.66) which is equivalent to the contracted Bianchi identity (1.4.62). Thus the contracted Bianchi identity imposes the conservation law for the energy-momentum density in 92

the Einstein-Cartan gravity. Substituting (2.5.55) and (2.5.64) into the BelinfanteRosenfeld relation (2.3.33) gives 1 1 j j j j Tik = Rki Rgik + j (S ik + 2Sk i 2S(ik ) 2S gik ) = Rki Rgik 2 2 j j lj l +j (C ki + C kl i C l gik ). (2.5.67) Combining (1.4.52), (1.4.54) and (2.5.67) gives 1 1 Tik = Pik P gik + C lki:l C lkl:i + C jkiC ljl C jkl C lji gik (2C ljl:j 2 2 C ljl C mjm + C mjl Cljm ) C jki:j C jlj C lki + C lkj C jli + C lij C jkl + C jkj :i
j C lki C jlj gik (C ljl:j + C jlj C mlm ) Cj (C jki + C lkl i C ljl gik ),

(2.5.68)

which is equivalent to the rst Einstein-Cartan equation (2.5.51). Thus the relation between the Ricci tensor and the Riemannian Ricci tensor is equivalent to the Belinfante-Rosenfeld relation in the Einstein-Cartan gravity, and (2.5.64) is another form of the rst Einstein-Cartan equation. 2.5.8 Einstein-Cartan pseudotensor

Replacing the action for the gravitational eld and matter (2.5.49) by S= 1 2c G g ik (C jij C lkl C lim C mkl ) d + Sm (2.5.69)

produces the rst Einstein-Cartan equation by varying the metric tensor, because gG diers from gP by a total divergence: 1 G gnp (C jnj C lpl C lnm C mpl ) + m = 0. ik g 2 (2.5.70)

The canonical energy-momentum density for the gravitational eld is also given by (2.5.12). The relations (2.5.14) and (2.5.70) give

i k ,i

(m + 2gnp (C jnj C lpl C lnm C mpl )) jl 1 = g = ( T + Ujl )g jl,k . jl ,k jl g 2

(2.5.71)

The covariant conservation (2.4.23) gives (Tk i + Uk i ),i = {kl i }(Tl i + 1 Ul i ) = g lm gim,k (Tl i + Ul i ) 2 (2.5.72)

1 = g lm,k (Tlm + Ulm ), 2

so the total energy-momentum density for the gravitational eld and matter is ordinarily conserved: (ki + Tk i + Uk i ),i = 0. (2.5.73) 93

Thus the corresponding four-momentum is conserved: Pi = 1 c (ik + Ti k + Ui k )dSk = const. (2.5.74)

i The quantity

+ Ui k is referred to as the Einstein-Cartan energy-momentum pseudotensor for the gravitational eld, and the sum ik + Ti k + Ui k is called the Einstein-Cartan energy-momentum complex.

2.5.9

Palatini variation

If the matter action Sm does not depend on the ane connection, its variation with respect to the metric and connection ( ijk is a tensor) is referred to as the Palatini variation. Varying (2.5.47) with respect to ik j gives, due to (1.3.39), S = 1 2c Rik gik d = 1 2c (( ilk );l ( ill );k 2S jlk ilj )g ik d.

(2.5.75) Partial integration and omitting total derivatives in (2.5.75) gives, using (1.2.33), S = 1 2c ( ilk gik;l 2Sl ilk gik ill gik;k +2Sk ill gik +2S jlk ilj gik )d. (2.5.76)

Since the ane connection is metric-compatible, gij ;k = 0, S = 0 turns the torsion tensor into zero, so the connection is formed by the Christoel symbols and the eld equations are the Einstein equations (2.5.5). Thus varying the action for matter elds, which do not depend on the ane connection, with respect to the connection is equivalent to varying it with respect to the torsion tensor. However, if the matter action Sm depends on the ane connection then (2.5.76) becomes S = + 1 2c 1 2c ( ilk gik;l 2Sl ilk gik ill gik;k + 2Sk ill gik + 2S jlk ilj gik )d (2.5.77)

ij k ijk d,

where the hypermomentum density is dened as ij k = 2 m . ijk (2.5.78)

Since the connection is metric-compatible, S = 0 gives k i g ik Sj j S S kij = ij k . 2 Contracting the indices i, j gives iik = 0, (2.5.80) (2.5.79)

94

which also results from the invariance of the Lagrangian density under a projective transformation (1.2.49) (the symmetric part of the Ricci tensor is invariant under this transformation): 1 1 j Ak = 0. (2.5.81) = m = ij k ijk = ij k i 2 2 The relation (2.5.80) constrains possible forms of matter Lagrangians algebraically, so it is not a conservation law. Therefore varying the action with respect to the ane connection, unlike that with respect to the torsion (or spin connection), does not constitute a physical variational principle. Only the antisymmetric part of the connection (torsion) can be regarded as a dynamical variable; its symmetric part can always be brought locally to zero by a suitable transformation of the coordinates. 2.5.10 Gravitational potential

If the metric tensor gij is approximately equal to the Minkowski metric tensor ij then the corresponding gravitational eld is weak. We can write 2 g00 1 + 2 , (2.5.82) c where is referred to as the gravitational potential. Thus nonrelativistic gravitational elds, corresponding to the limit c , are weak. Also u0 1 and u 0. In this limit, the leading component of the Levi-Civita connection is 1 g00 1 {0 = 2 , (2.5.83) 0} g 2 x c x so the metric geodesic equation (1.4.80) reduces to dv = . (2.5.84) dt The quantity G in (2.2.5) reduces to 2 G = 4 ( ) 2 . (2.5.85) c The leading component of the Riemannian Ricci tensor is 1 2 1 {0 0} = 2 2 = 2 . (2.5.86) P00 x c x c The leading component of the energy-momentum tensor (2.4.167) is T00 = c2 . (2.5.87) Therefore the Einstein equations in the nonrelativistic limit reduce to the Poisson equation: = 4G, (2.5.88) c4 (2.5.89) 8 is Newtons gravitational constant. In vacuum, where = 0, the Poisson equation reduces to the Laplace equation: = 0. (2.5.90) G= 95 where

2.5.11

Hydrodynamics

The covariant conservation (2.4.24) of the metric energy-momentum tensor (2.4.77) gives (( + p)uk ):k ui + ( + p)uk ui:k = p,k g ik . (2.5.91) Multiplying (2.5.91) by ui gives (( + p)uk ):k = p,k uk , which, upon substituting into (2.5.91) yields the Euler equation: ( + p ) D {} ui = p,k hik . ds (2.5.93) (2.5.92)

If p,i ui (which includes the case p = const) then (2.5.93) reduces to the metric geodesic equation (1.4.80). Dening a quantity w such that d dw = w +p brings (2.5.92) to (wui):i = 0. In the nonrelativistic limit, c , u0 1, u (2.5.92) reduces to the equation of continuity: + div s = 0, t where s = v is referred to as the mass current. Integrating (2.5.96) over the volume gives dV + s df = 0, (2.5.98) t which means that the change in time of the total mass inside a volume, m = dV , is balanced by the mass ux through the surface bounding this volume, representing the conservation of the total mass of a uid. The Euler equation (2.5.93) reduces in this limit to v (2.5.99) + v , v = , + p, t or dv v = + (v )v = p. (2.5.100) dt t Integrating (2.5.100) over the volume gives, using P = vdV , the change in time of the total momentum of a uid: dP = dm pdf . (2.5.101) dt Without pressure gradients, (2.5.100) reduces to (2.5.84). References: [1, 2, 3, 4, 5, 6, 7] 96 (2.5.97)
v c

(2.5.94)

(2.5.95) , c2 and p , so (2.5.96)

2.6
2.6.1

Spinor elds
Dirac matrices

The Dirac matrices dened by (1.7.1) are complex. A particular solution of (1.7.1) is given by the Dirac representation: 0 = I 0 0 I , = 0 0 , (2.6.1)

where I is the unit 22 matrix and 1 = 0 1 1 0 , 2 = 0 i i 0 , 3 = 1 0 0 1 (2.6.2)

are the Pauli matrices (all indices are coordinate invariant). The Pauli matrices are traceless tr( ) = 0 and Hermitian = (the Hermitian conjugation of a matrix A is the combination of the complex conjugation and transposition, A = AT ), satisfy = + i , and their square is I . The identity (2.6.3) gives the anticommutation relation = i , , 2 2 2 (2.6.4) (2.6.3)

so 2 form the lowest, two-dimensional representation of the angular momentum operator M (1.6.76). The properties of imply that the Dirac matrices are traceless tr( i ) = 0 and satisfy

0 = 0 , = , i = 0 i 0 . i eijkl i j k l = i 0 1 2 3 , 24 which is traceless tr( 5 ) = 0 and Hermitian 5 = 5 , and satises 5 = { i , 5 } = 0 , ( 5 )2 = 1 . In the Dirac representation 5 = 0 I I 0 . Dene

(2.6.5)

(2.6.6)

(2.6.7)

(2.6.8)

The anticommutation relation (1.7.1) gives i i = 4, i j i = 2 j , i j k i = 4 jk I, i j k l i = 2 l k j , i j k = ij k + jk i ik j + iijkl l 5 . 97 (2.6.9) (2.6.10) (2.6.11) (2.6.12) (2.6.13)

The Dirac representation is not unique; the relation (1.7.1) is invariant under a similarity transformation i S i S 1 , where S is a nondegenerate (det S = 0) matrix. I I 1 S 1 . Taking S = turns the Dirac Accordingly, S and 2 I I repreentation into the Weyl representation, in which 0 = 0 I I 0 , = 0 0 , 5 = I 0 0 I . (2.6.14)

For an innitesimal Lorentz transformation (1.6.7), the relations (1.7.5) and (1.7.6) 1 ab ( a b b a ), so give L = I + 8 1 L = I + ab ( b a a b ) 8 (2.6.15)

is equal to L1 (so L is unitary) for rotations and equal to L for boosts. The relation (2.6.5) gives then 1 1 L 0 = 0 + ab ( b a a b ) 0 = 0 ab 0 ( a b b a ) = 0 L1 . (2.6.16) 8 8 Thus the quantity 0 transforms under (1.7.7) like an adjoint spinor: 0 L 0 = 0 L1 . (2.6.17)

The spinors and 0 can be used to construct tensors, as in (1.7.11): 0 transforms like a scalar, 0 i is a vector, 0 5 is a pseudoscalar, 0 i 5 is a pseudovector, and 0 [i j ] is an antisymmetric tensor. Higher-rank tensors constructed from and 0 reduce to the above 5 kinds of tensors because of (2.6.13). to denote 0 . Hereinafter, we will use Dene the chirality projection operators P = I 5 2 , P+ + P = I, P = I, P+ P = P P+ = 0. 2 (2.6.18)

They project a spinor into the right-handed spinor R and left-handed spinor L , R = P+ , L = P , = R + L . 2.6.2 Dirac equation (2.6.19)

A Lagrangian density for dynamical spinor elds must contain rst derivatives of i ;i , spinors. The simplest scalar containing derivatives of spinors is quadratic in , where ;i is the covariant derivative of (1.7.14). This quantity is complex. In the locally inertial frame of reference, its complex conjugate is
i i ,i ) = ( i ,i ) = ,i ,i 0 i 0 = ,i i , ( =

(2.6.20)

i ,i + ,i i and i( i ,i ,i i ) are real. The former is, however, equal so both i ),i , so a Lagrangian density proportional to such term to a total divergence ( 98

does not contribute to eld equations. Thus the simplest dynamical part of a spinor i ,i ,i i ). Another scalar that can be Lagrangian density is proportional to i( . Therefore the simplest Lagrangian used in a spinor Lagrangian is proportional to density for spinor elds, in the locally Galilean frame of reference, has the form i ,i ,i i ) m, = ( i 2 (2.6.21)

where m is a real scalar constant called the spinor mass, and it is referred to as the Dirac Lagrangian density. For any frame of reference, = ei ( i ;i ;i i ) m a ;i ;i a ) m . = ( a i 2 i 2 (2.6.22)

Consider the metric formulation of gravity with the Einstein-Hilbert action (2.5.1). Therefore spacetime has the Riemannian geometry, so ;i = :i . Varying (2.6.22) with and omitting total derivatives gives respect to (i i :i m ), = gives the Dirac equation: so the stationarity of the action S = 0 under i i :i = m. (2.6.24) (2.6.23)

Varying (2.6.22) with respect to and omitting total derivatives gives the adjoint conjugate of (2.6.24): :i i = m. i (2.6.25) The Dirac equation is linear in , so can be multiplied by an arbitrary constant without altering (2.6.24). Varying (2.6.22) with respect to ei a gives the tetrad energymomentum density for the spinor eld,
a a :i :i a ea j :j + ea j ia = ( i i :j ) + m ei ,

i 2

(2.6.26)

so

i j j Tik = ( (2.6.27) (k :i) :(i k ) gik :j + gik :j ) + mgik . 2 The conservation law (2.4.24) applied to the energy-momentum tensor (2.6.27) gives the Dirac equations (2.6.24) and (2.6.25). gives, using Subtracting (2.6.25) multiplied by from (2.6.24) multiplied by (1.7.33) and |i = :i , i )|i = ( i ): i = 0 , ( (2.6.28) so the vector density i ji V = , (2.6.29)

called the vector Dirac current, is conserved: ji V,i = 0 or + j = 0, t 99 (2.6.30)

where

. c = , j =

(2.6.31)

The spinor density is real and positive. The conservation law (2.6.30) is referred to as the equation of continuity, like (2.5.96). The Dirac equation (2.6.24) gives j ( i :i )|j = im j :j or, due to (1.7.33), j i |ij = m2 . 1 i |i + m2 = Rklij k l i j . 8 (2.6.33) Using (1.7.30) and (1.7.39) turns (2.6.33) into the Klein-Gordon-Fock equation: (2.6.34) (2.6.32)

If a spinor is equal to its either left- or right-handed projection, = L or = R , then it is called a Weyl spinor. Multiplying (2.6.24) by P gives iP i :i = i i P :i = mP or i i :i
L(R)

(2.6.35) (2.6.36)

= m R(L) .

Thus if is a Weyl spinor then m = 0. 2.6.3 Spinors in Einstein-Cartan gravity

Consider the metric-ane formulation of gravity with the Einstein-Cartan action (2.5.47), in which spacetime has the Riemann-Cartan geometry. Varying (2.6.22) with respect to ei a gives the tetrad energy-momentum density for the spinor eld,
a a ;i ;i a ea j ;j + ea j ia = ( i i ;j ) + m ei .

i 2

(2.6.37)

Putting the denition of the covariant derivative of a spinor (1.7.14) into (2.6.22) gives i i ,i i ) i { i , i} m . = ( ,i (2.6.38) 2 2 Using the Fock-Ivanenko coecients (1.7.28) as the spinor connection i turns (2.6.38) into i i ,i i ) + i abi { i , a b } m . ,i (2.6.39) = ( 2 8 The spin density (2.3.17) corresponding to the Lagrangian density (2.6.39) is, due to the identity { i , j k } = 2 [i j k] , [i j k ] ijk = i 2 (2.6.40)

100

or

i [i j k ] . sijk = 2 The spin density (2.6.40) is independent of m and totally antisymmetric,

(2.6.41)

ijk = [ijk].

(2.6.42)

The second Einstein-Cartan equation (2.5.57) for the spin tensor (2.6.41) gives a totally antisymmetric torsion tensor, Sijk = i [ij k] , 4 (2.6.43)

so Si = 0. Thus the contortion tensor is, using (2.6.13), Cijk = The pseudovector density l 5 . ijkl 4 (2.6.44)

i 5 ji A =

(2.6.45)

is called the axial Dirac current. gives, after omitting total divergences, Varying (2.6.39) with respect to i k ( ,k + ( k ),k {k , k } ) m = 0. 2 Substituting ( k ),k = k ,k + k;k 2 Sk k = k ,k + [k , k ] into (2.6.46) gives i k ,k i k k m = i k ;k m = 0. The relation (1.5.33) gives 1 ;k = :k + Cijk i j , 4 from which we obtain, upon substituting (2.6.44), l 5 ) i j k = k :k + i ijkl ( l 5 )ijkm m 5 ijkl ( 16 16 3i l 5 k 5 = :k + ( )l . (2.6.50) 8 k ;k = k :k + Therefore (2.6.48) becomes the Heisenberg-Ivanenko equation: i k :k 3 5 (k ) k 5 = m. 8 101 (2.6.51) (2.6.49) (2.6.48) (2.6.47) (2.6.46)

Varying (2.6.39) with respect to gives the adjoint conjugate of (2.6.51), k 5 ) k 5 = m. :k k 3 ( i 8 (2.6.52)

The Heisenberg-Ivanenko equation (2.6.51) diers from the Dirac equation (2.6.24) by a nonlinear term, cubic in the spinor eld and representing a spinor self-interaction, corresponding to a spin-spin interaction in the tensor (2.5.58). The conservation law (2.4.32) applied to the energy-momentum density (2.6.37) gives the HeisenbergIvanenko equations (2.6.51) and (2.6.52). Subtracting (2.6.52) multiplied by from gives the conservation of the vector Dirac current (2.6.29). (2.6.51) multiplied by The total antisymmetry of the spin density implies N ijk = N [ijk] , where N ijk is given by (2.4.87). Also N ijk = 3S [ij uk] , (2.6.54) (2.6.53)

where S ij is the intrinsic spin tensor (2.4.156). The covariant (with respect to the Levi-Civita connection) change (2.4.129) of the spin pseudovector along a world line becomes D {} J i D {} ui k 3 D {} uk = ui Jk + u Jk + eij nm uj S nik N ikm = 3S ijk uj N k , ds ds ds 2 where (2.6.55)

1 (2.6.56) N i = eijkl Njkl . 6 If N i J i then (2.6.55) gives J i Ji = const. For a point particle, M ijk given by (2.4.86) vanishes. Thus (2.4.112) reduces to N ijl = ul ij 0 N , u0
l

(2.6.57)

u i00 which for a spinor particle gives N il0 = u and thus N ijk = 0 or 0N

= 0.

(2.6.58)

Therefore a spinor eld in the Einstein-Cartan gravity cannot be approximated as a point particle. References: [3, 4, 5, 6, 7].

2.7
2.7.1

Electromagnetic eld
Gauge invariance and electromagnetic potential

The Lagrangian density (2.6.22) is a real combination of the complex Dirac matrices . It is invariant under a gauge transformation of the rst type of i and spinors , the spinor elds, = eie , = eie , (2.7.1) 102

if e is a real scalar constant, but it is not invariant for e(xi ), because ; = eie (; + ie, ). (2.7.2)

Introduce a compensating vector eld A , called the electromagnetic potential, such that the Weyl or electromagnetic covariant derivative D = ieA of a spinor , D = ; ieA , transforms under (2.7.1) like : D = eie D . This requirement gives ; ieA = eie (; ieA ), (2.7.6) (2.7.5) (2.7.4) (2.7.3)

which, with (2.7.1) and (2.7.2), yields the transformation law for the electromagnetic potential, (2.7.7) A = A + , , called a gauge transformation of the second type. The real scalar constant e is called the spinor electric charge. The adjoint conjugation of (2.7.4) is = ; + ieA . D is invariant under (2.7.1), so The scalar ) = ( ), D ( which constraints the electromagnetic potential to be real: A = A . (2.7.10) (2.7.9) (2.7.8)

The time component of A , = A0 , is called the electric potential and the spatial components A form the magnetic potential A: A = (, A). The gauge transformation (2.7.7) reads = + , A = A . ct (2.7.12) (2.7.11)

In the local Minkowski spacetime, A transforms according to (1.6.98), A = ( 1) 1 + 2 103 A . (2.7.13)

The gauge-invariant modication of the Dirac Lagrangian density (2.6.22) is a Di Di a ) m . = eia ( i 2 (2.7.14)

The electromagnetic potential corresponds, up to the multiplication by an arbitrary constant, to the vector multiple of I in the formula for the spinor connection (1.7.26). The electromagnetic potential is analogous to the ane connection: it modies a derivative of a spinor so such derivative transforms like a spinor under unitary gauge transformations of the rst type, while the connection modies a derivative of a tensor so such derivative transforms like a tensor under coordinate transformations. 2.7.2 Electromagnetic eld tensor

The commutator of total covariant derivatives of a spinor is given by (1.7.30) with the curvature spinor Kij given by (1.7.36), where the tensor Bij is related to the vector Ai in (1.7.26) by (1.7.38). Therefore the commutator of the electromagnetic covariant derivatives of a spinor, [Di , Dj ] , is given by (1.7.30) with the curvature spinor 1 Kij = Rklij k l + ieFij I, 4 where the antisymmetric tensor Fij = Aj,i Ai,j = Aj :i Ai:j (2.7.16) (2.7.15)

is referred to as the electromagnetic eld tensor. The electromagnetic eld tensor is analogous to the curvature tensor: it appears in the expression for the commutator of electromagnetic covariant derivatives of a spinor, while the curvature tensor appears in the expression for the commutator of coordinate-covariant derivatives of a tensor. Substituting (2.7.7) into (2.7.16) gives
Fij = Fij ,

(2.7.17)

so the electromagnetic eld tensor is gauge invariant. The denition (2.7.16) is equivalent to the rst Maxwell-Minkowski equation Fij,k + Fjk,i + Fki,j = Fij :k + Fjk:i + Fki:j = 0 or ijkl Fjk,l = ijkl Fjk:l = 0. Dene E = F0 , 1 B = F , B = B , B 2 (2.7.20) = B , (2.7.21) (2.7.19) (2.7.18)

where is given by (1.4.105). The component of (2.7.18) with all spatial indices, B, + B, + B, = 0, gives, using (1.4.116), div B = 0. 104 (2.7.22)

The components of (2.7.18) with one temporal index, B,0 + Ealpha, E, = 0, gives, using (1.4.117), 1 ( B ) curl E = . (2.7.23) c t The spatial vector E is called the electric eld and the spatial pseudovector B is the magnetic eld. In the locally geodesic and Galilean frame of reference, these elds depend on the components of the electromagnetic potential (2.7.11) according to (2.7.16): A , ct B = A, E= and they are invariant under (2.7.12). The tensor Fij is given by

(2.7.24) (2.7.25)

Fij =

0 Ex Ey Ez Ex 0 Bz By Ey Bz 0 Bx Ez By Bx 0

and transforms according to (1.6.99). Thus the electric and magnetic elds transform according to E = (E + B ) + 1 ( E ) , 2 1 ( B ) . B = (B E ) + 2 (2.7.27) (2.7.28)

(2.7.26)

In this frame, (2.7.22) and (2.7.23) become the rst pair of the Maxwell equations: div B = 0, curl E = B . ct (2.7.29) (2.7.30)

Applying the div operator to (2.7.25) gives (2.7.29) and applying the curl operator to (2.7.24) gives (2.7.30). Applying the div operator to (2.7.30) gives (2.7.29). Integrating the rst pair of the Maxwell equations over the volume and surface area, respectively, gives B d f = 0, E dl = (2.7.31) B df . (2.7.32)

ct

The integral A df is the ux of a vector A through the surface f and the integral A dl is called the circulation of A along the contour l. Thus the ux of the magnetic eld through a closed surface vanishes and the circulation of the electric eld along a contour, which is called the electromotive force, is equal to the minus time derivative of the ux of the magnetic eld through the surface enclosed by this contour (Faradays law). 105

2.7.3

Lagrangian density for electromagnetic eld

The simplest gauge-invariant Lagrangian density representing the electromagnetic eld is a linear combination of terms quadratic in Fij : Fij F ij and ijkl Fij Fkl . The second term is a total divergence because of (2.7.19): ijkl Fij Fkl = 2(ijkl Fij Al ),k , (2.7.33)

so it does not contribute to the eld equations. Thus the Lagrangian density for the electromagnetic eld is given by

EM =

1 Fij F ij , 16

(2.7.34)

1 sets the units of Ai . In the locally geodesic and Galilean where the Gauian factor 16 frame of reference, (2.7.34) becomes

EM =

1 (E 2 B 2 ). 8

(2.7.35)

Therefore in order for the action S to have a minimum, there must be the minus sign in front of the right-hand side of (2.7.34). Otherwise an arbitrarily rapid change of A in time would result in an arbitrarily large value of E, according to (2.7.24), and thus an arbitrarily low value of S , so the action would have no minimum. A generalization of the tensor (2.7.16) to a covariant derivative with respect to the ane connection k ik j , Aj ;i Ai;j = Fij + 2S ij Ak , is not gauge invariant, so the torsion tensor cannot appear in a gauge-invariant Lagrangian density which is quadratic in Fij . Thus the electromagnetic eld, unlike spinor elds, does not couple to torsion. 2.7.4 Electromagnetic current c m , Ai

Dene the electromagnetic current density ji = and the electromagnetic current vector ji ji = . (2.7.37) (2.7.36)

The invariance of the action under an arbitrary innitesimal gauge transformation Ai = Ai Ai = ,i gives, upon partial integration and omitting a total divergence, S = 1 c2 ji Aj d = 1 c2 ji ,i d = 1 c2 ji,i d = 0, (2.7.38)

so the electromagnetic current is conserved, ji,i = 0, j i:i = 0. 106 (2.7.39)

The gauge-invariant Lagrangian density (2.6.22) for spinor matter is = ei ( eAi i , a i i a ) m a ;i ;i a ) m = eia ( a so the electromagnetic current for the spinor eld is i , ji = ec (2.7.41) i 2 i 2

(2.7.40)

which is proportional to the conserved vector Dirac current (2.6.29). The electromagnetic current density (2.7.36) corresponds to the current (2.4.7) with i =0 and = ie (which is equal to the innitesimal due to (2.7.1)). Consider matter which is distributed over a small region in space, as in section (2.4.7). Integrating (2.7.39) over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals gives j0,0 dV = 0. The conservation law (2.7.39) also gives
k i (xk ji ),i = xk,i ji + xk ji,i = i j = jk ,

(2.7.42)

(2.7.43)

which, upon integrating over the volume hypersurface and using Gau-Stokes theorem to eliminate surface integrals, gives xk j0 dV
,0

jk dV.

(2.7.44)

Using (2.4.84) turns (2.7.44) into uk u0 For a particle located at xa , so j0 dV + xk j0 dV

,0

jk dV.

(2.7.45)

xk j0 dV = 0 and ji (x) is thus proportional to (x xa ), jk = (2.7.46)

uk 0 j. u0 Dene the electric charge density such that c j0 = . g00

(2.7.47)

The electric charge density is not a tensor density. Dene the electric charge e such that dV = de. (2.7.48) The electric charge density for particles with charges ea located at xa is (x) =
a

ea (x xa ),

(2.7.49)

107

and

j0 dV (which is equal to j0 dV =
a

ji dSi for a volume hypersurface, so it is a scalar) is cea (x xa )dV = c g00 ea ,

a

(2.7.50)

so the electric charge is a scalar. Thus the electromagnetic current vector for a system of charged particles is cuk ea (x xa ), (2.7.51) j k (x) = 0 a u analogously to (2.4.168). The relation (2.7.42) represents the conservation of the total electric charge of a physical system. ui In the locally geodesic and Galilean frame of reference, u 0 = (1, v /c), so j i = (c, j), where j is the spatial current vector, j = v. (2.7.53)

(2.7.52)

The conservation law (2.7.39) in this frame, j i,i = 0, has the form of the equation of continuity (2.6.30). For one particle located at x0 (t), (x) = e (x x0 ), (2.6.30) is explicitly satised since (x x0 ) = ev = e (x x0 ) = ev (x x0 ) t t x0 x = ev (x x0 ) = j, (2.7.54) x where v =
dx0 . dt

For a system of charged particles, we also have jdV =

a

ea va .

(2.7.55)

The equation of continuity (2.6.30) represents, upon integrating over the volume, the conservation of the total electric charge: t 2.7.5 Maxwell equations dV + j d f = 0. (2.7.56)

The total Lagrangian density for the electromagnetic eld and matter is the sum of (2.7.34) and the term Ai j i due to (2.7.36):

EM =

1 1 Fik F ik Ak j k , 16 c

(2.7.57)

108

where we omit the terms corresponding to the gravitational eld and matter which does not depend on Ak . Varying (2.7.57) with respect to Ak , integrating partially and omitting total divergences gives 1 1 jk jk F ik Fik Ak = F ik (Ak,i Ai,k ) Ak 8 c 8 c k 1 1 j 1 = ( F ik ),i Ak F ik Ak,i Ak = j k Ak , (2.7.58) 4 c 4 c EM = so the principle of least action S = 0 for arbitrary variations Ak yields the second Maxwell-Minkowski equation 4 k j ( F ik ),i = c or F ik:i = (2.7.59)

4 k j . (2.7.60) c The electromagnetic eld equation (2.7.59) implies that ji is conserved, ji,i = 0, which corresponds to the conservation of the total electric charge, but does not constrain the motion of particles. Therefore a conguration of charged particles producing the electromagnetic eld can be arbitrary, subject only to the condition that the total charge be conserved, unlike a conguration of particles producing the gravitational eld which is not arbitrary but constrained by the gravitational eld equations. Dene D = g00 F 0 , (2.7.61) 1 1 H , H = H . (2.7.62) H = g00 F , H = 2 The relations F0 = g0i gj F ij and F = g i g j Fij give then E + g H , D = g00 H B = gE + g E , g00 or, in the spatial-vector notation, E g H, D= g00 H + g E. B= g00 Using (1.4.102) brings the temporal component of (2.7.59) to 1 ( D ), = 4 (2.7.65) (2.7.66) (2.7.63) (2.7.64)

(2.7.67)

109

or div D = 4. The spatial components of (2.7.59) read 1 dx 1 ( H ), + ( D ),0 = 4 0 dx or 1 ( D) 4 curl H = + j. c c t (2.7.69) (2.7.68)

(2.7.70)

The conservation law (2.7.39) reads 1 ( to D = E, B = H. (2.7.72) (2.7.73)

+ div j = 0.

(2.7.71)

In the locally geodesic and Galilean frame of reference, (2.7.65) and (2.7.66) reduce

In this frame, (2.7.68) and (2.7.70) become the second pair of the Maxwell equations: div E = 4, ) E 4 + j. curl B = ct c (2.7.74) (2.7.75)

Applying the div operator to (2.7.75) and using (2.7.74) gives (2.6.30). Integrating the second pair of the Maxwell equations over the volume and surface area, respectively, gives E df = 4q, B dl = ct E df + (2.7.76) 4 c j df . (2.7.77)

Thus the ux of the electric eld through a closed surface is proportional to the total charge inside the volume enclosed by the surface f (Gau law) and the circulation of the magnetic eld along a contour is equal to the time derivative of the ux of the electric eld through the surface enclosed by this contour, called the displacement current, plus the surface integral of the current vector (Amp` eres law). The two pairs of the Maxwell equations are linear in the elds E and B. The sum of any two solutions of the Maxwell equations is also a solution of these equations. Thus the electromagnetic eld of a system of sources (particles) is the sum of the elds from each source. The additivity of the electromagnetic eld is referred to as the principle of superposition.

110

2.7.6

Energy-momentum tensor for electromagnetic eld

EM The metric energy-momentum tensor (2.3.3) for the electromagnetic eld Tik is given by the Lagrangian density (2.7.34):

1 1 glm Fik F ik g lm Fik Flm g il g km 32 8 1 1 gik Flm F lm Fi j Fkj g ik , = 8 4 EM = so

EM Tik =

(2.7.78)

1 1 gik Flm F lm Fi j Fkj . (2.7.79) 4 4 The corresponding energy density W , energy current S called the Poynting vector, and stress tensor called the Maxwell stress tensor, are given in the locally geodesic and Galilean frame of reference, due to (2.4.71), by 1 (E 2 + B 2 ), 8 c S= E B, 4 1 1 E E + B B (E 2 + B 2 ) . = 4 2 W = (2.7.80) (2.7.81) (2.7.82)

Multiplying (2.7.30) by B and (2.7.75) by E and adding these scalar products gives 1 E 1 B 4 E + B = j E (B curl E E curl B), c t c t c from which we obtain 1 4 (E 2 + B 2 ) = j E div(E B) 2c t c or W + j E + div S = 0. t The energy-momentum tensor for the electromagnetic eld is traceless,
EM ik Tik g = 0,

(2.7.83)

(2.7.84)

(2.7.85)

(2.7.86)

so (2.4.174) and the virial theorem (2.4.175) remain unchanged if the particles interact electromagnetically. The condition (2.7.86) also gives, using (2.4.82), EM = 3pEM , so (2.4.176) implies that the free electromagnetic eld is ultrarelativistic. (2.7.87)

111

2.7.7

Lorentz force

Consider a charge particle interacting with the electromagnetic eld. The total energy-momentum tensor for the particle and electromagnetic eld is covariantly conserved, which gives the motion of the particle. The electromagnetic part yields, using (2.7.18) and (2.7.60), Ti k:k = 1 1 1 1 1 Fmi:l F lm Fil:m F lm Flm:i F lm Fil:k F kl Fil F kl:k = 4 2 4 2 2 1 1 kl l kl kl Fil F :k = Fil j . (2.7.88) Fil:k F Fil F :k = 4 c

The particle part gives, using (2.4.167), ui uk Ti k:k = c2 g00 u0 so we obtain ui uk c2 g00 u0 ,
:k

(2.7.89)

1 Fil j l = 0. c :k uk g00 u0

(2.7.90)

Multiplying (2.7.90) by ui and using (2.7.46) gives c2 which turns (2.7.90) into c2 or 1 ul uk u = F i:k il g00 u0 c g00 u0 (2.7.92) ,
:k

(2.7.91)

D {} ui e = F ij uj , (2.7.93) ds c which is the equation of motion of a particle of mass m and charge e in the electromagnetic eld Fij . Multiplying (2.7.93) by ui gives the identity, so (2.7.93) has 3 independent components. The right-hand side of (2.7.93) is referred to as the Lorentz force. 0 d {} d = ds = uc dt and ui = In the locally geodesic and Galilean frame of reference, D ds (, v/c), so (2.7.93) reads (we choose the spatial components as the 3 independent ones) du e mc = eF 0 + F v (2.7.94) dt c or, using (2.4.159), e dP = eE + v B. (2.7.95) dt c The temporal component of (2.7.93) is mc mc e du0 = F0 v dt c 112 (2.7.96)

or

dE = ev E, (2.7.97) dt which also results from multiplying (2.7.95) by v and using (2.4.184). Integrating (2.7.85) over the volume gives t W dV + j EdV + S d f = 0, (2.7.98)

which, with (2.7.55) and (2.7.97), yields the conservation of the total energy (2.4.66) of the electromagnetic eld and particles: t References: [2, 3]. W dV +
a

Ea +

S d f = 0.

(2.7.99)

113

References
[1] E. Schr odinger, Space-Time Structure (Cambridge Univ. Press, 1950). [2] L. D. Landau and E. M. Lifshitz, The Classical Theory of Fields (Pergamon, 1975). [3] E. A. Lord, Tensors, Relativity and Cosmology (McGraw-Hill, 1976). [4] V. de Sabbata and M. Gasperini, Introduction to Gravitation (World Scientic, 1986). [5] F. W. Hehl, P. von der Heyde, G. D. Kerlick, and J. M. Nester, Rev. Mod. Phys. 48, 393 (1976). [6] F. W. Hehl, J. D. McCrea, E. W. Mielke, and Y. Neeman, Phys. Rep. 258, 1 (1995). [7] R. T. Hammond, Rep. Prog. Phys. 65, 599 (2002). [8] C. W. Misner, K. S. Thorne, and J. A. Wheeler, Gravitation (Freeman, 1973). [9] K. Nomura, T. Shirafuji, and K. Hayashi, Prog. Theor. Phys. 86, 1239 (1991).

114