Solution for Chapter 23

(compiled by Xinkai Wu) A. Ex 23.2 Causality [A. Dvoretskii/April 2000] Consider two dierent reference frames - primed and unprimed. Assume without loss of generality that event P1 occurs at a point (0, 0) in spacetime in both frames and event P2 at a point (t, 0) in the unprimed frame (i.e. at the same spacial point) and at a point (t , x ) in the primed frame. Now using the invariance of the interval s2 = s 2 = t2 = t 2 + x 2 t = t2 + x 2

so

The transformations from one frame to another are continuous and in the limit of very small transformations t t, so we must use the + sign. Therefore, t = t2 + x 2 > 0

and so the temporal order of events is the same in all inertial frames. Of course, were this not true, causality would be violated. Its also obvious that t t and that apart from that there are no limits on the values of the spacial and temporal separation of the two events: t and x can be made arbitrarily large.

t P2 t
2

P2

t 2

t x

t
2

P1

x x P1 x
Figure 1: Causality

See Fig. 1 for the spacetime diagrams. As the velocity of the primed frame increases, the time of event P2 , t2 moves up the t axis. 1

Clearly t2 > t2 (c.f. dashed hyperbola of events all at same interval from P1 ). Similiar diagram can be drawn for velocity in the opposite direction. The diagram shows that for t2 , t2 t2 and t2 can be made arbitrarily large. B. Ex. 23.4 Index manipulation rules from duality [A. Dvoretskii/April 2000] (a) Lets expand a given dual basis vector in the original basis e = f e Now to nd the coecients f multiply both sides of the equation by e and use the duality relation to obtain
e e = f = f

But e e = g(e , e ) = g = f This proves the rst relation. e = g e The proof of the second relation is similar. (b) Using the result above F = F(e , e ) = F(g e , e ) Now use the linearity of the tensor to write F = g F(e , e ) = g F The proof of the second relation is similar. (c) Consider for example the rst identity F = F e e Now consider
F(e , e ) = F (e e )(e e ) = F = F

So all the components of the tensors on the left-hand side and the right hand side are equal which proves the identity. Similar proof can be given in the other cases. C. Ex. 23.6 Part (a) and (b) Connection coecients for circular coordinates [A. Dvoretskii/April 2000] First lets consider the Coordinate basis

= , e =

[e , e ] = [e , e ] = [e , e ] = 0 Hence c = 0 The metric tensor is given by g=( 1 0 0

2

The only non-zero Christoel symbols are then

= (1/2)g, =

and the connection coecients are

= ,

= 1/

(a) Orthornomal basis e = , e = 1/ In this basis the metric tensor is of course just g=( 1 0 ) 0 1

The commutation coecients are readily computed from [e , e ] = [ , (1/ ) ] = (1/

c = 1/ , 2

) = (1/ )e

Hence, the only non-zero commutation coecients are

, c = 1/

And the Christoel symbols and the connection coecients are

= = 1/ = = 1/

Ex.23.5 Part (b) Transformation matrices for circular polar bases [Xinkai Wu/02]

By chain rule, we have f f x f y f f f + = cos + sin x y x y f f = sin + cos x y =

for any function f . Combining this with 1 , e = ex = , ey = x y e = we get the transformation matrix
y Lx = cos, Ly = sin, Lx = sin, L = cos

inverting which gives

L x = cos, L x = sin, L y = sin, L y = cos

D. Ex. 23.9 Index gymnastics [Kip and Xinkai Wu/02] (a) First notice that P u = u + u u u = 0, using the fact that u2 = 1. Thus
P P = P (g + u u ) = P

(b) P A u = 0 because P u = 0, thus P A is orthogonal ot u. P A = A + u u A = A if u A = 0. (c) In the uids local rest frame, g = , and u = 0 . Thus P is diagonal in this frame, with P00 = 00 u0 u0 = 1 1 = 0, P11 = P22 = P33 = 1. (d) (u u) = u u; = a u2 = a where weve used eq. (23.53) and the fact that P , , and are all orthogonal to u. Thus we see u u = a. Also a u = a u = u u; u = 1 (u u ); u = 0 2

because u u = 1 is a constant. (e) Contracting eq. (23.53) with g , using a u = 0, tracelessness of and , and g P = g g + u2 = 4 1 = 3, we get u = . (f) Notice that the term a u in (23.53) is not orthogonal to u on its second slot. We can get rid of it by projecting with P . Thus u; P =1 3 P + + 4

 where in the second term we have used P P = P . Now is the 1 1 + u ; P ) 3 u ; P . symmetric, traceless part of this tensor: = 2 (u; P 1 Similarly is the antisymmetric part: = 2 (u; P u ; P ). (g) (i) The four velocity is given by u = (, v j ), where = 1/ 1 v j v j . So to rst order in v j , we have u0 = 1, and uj = v j . (ii) = u ; = u , , to rst order in v j , this becomes = uj ,j = v j ,j (since u0 = 1 to rst order of v j , u0 ,0 = 0). (iii) Using the expression of given in the previous part and noticing that ,to rst order in v j , Pjk = gjk , and in the brackets we can take P k = k (since the uj ; terms is already linear in v j ) ,we get 1 jk = 1 2 (vj,k + vk,j ) 3 gjk ,which is the uids nonrelativistic shear. (iv)Similar 1 to part (iii), we get jk = 2 (vj,k vk,j ), which is the nonrelativstic rotation.

E. Ex. 23.10 IntegrationGausss law [Xinkai Wu/02] E d = E j dj , where dj = (ej , d/, d/). By the antisymmetry of , only dr doesnt vanish, and is given by r dd = det||gjk ||dd = R2 sindd. On the r.h.s. of eq. (23.55), d was already worked out in the text: d = r2 sindrdd. Thus eq. (23.55) becomes E r R2 sindd =
r =R r<R

e
0

r2 sindrdd

F. Ex. 23.11 Stress energy tensor for a perfect uid [A. Dvoretskii/00 and Kip/02] (a) If the components of two tensors are equal in a given frame, then they will be equal in any frame, so we just need to verify that T = ( + P )u u + P g reduces to T 00 = , T ij = P jk

in the rest frame. Its a trivial exercise given the simple form of u in the rest frame u = (1, 0, 0, 0) (b) If the observer is moving with a speed v much smaller than the speed of light with respect to the rest frame of the uid, then the momentum density in this frame can be written as T 0j = ij v inertial i which is the denition of the tensorial inertial mass density. In the limit of small v the momentum density can be written as

T 0j = ( + P )u0 uj = ( + P )v j to rst order in v , and so ij = ( + P ) ij inertial (c) First notice that = 0 (1 + u) = N (1 v2 v2 )(1 + u) = N (1 + u ) 2 2

which well use frequently in the following derivation. T 00 1 = ( + P )u0 u0 P = ( + P )(1 + v 2 )2 P 2 v2 = (1 + v 2 ) = N (1 + u )(1 + v 2 ) 2 1 = N + N v 2 + N u 2

as desired. T j0 v2 v2 v2 ) + P (1 + )v j (1 + ) 2 2 2 2 2 v v = N (1 + u )(1 + v 2 )v j + P v j = N (1 + u + )v j + P v j 2 2 1 P = N v j + u + v 2 + N v j 2 N = ( + P )uj u0 = N (1 + u

since T 0j = T j 0 , this gives the desired expression for T 0j . And keeping only the term linear in v , this gives T j 0 = N v j as desired. T jk = ( + P )uj uk + P g jk = N (1 + u = N v j v k + P g jk as desired. (d) as in (b) T 0j = ij v = ( + P )vj inertial i v2 ) + P (1 + v 2 )v j v k + P g jk 2

G. Ex. 23.14 Stress-energy tensor for a point particle [A. Dvoretskii/00] 6

We want to prove that p (0 ) =

S

p ( )p ( ) (Q, P ( ))d d

The function vanishes everywhere except at the point Q in the 3-surface S at which the particles worldline pierces S . The value of at that point is 0 , so the only nonzero contribution to the integral comes from 0 . The right hand side reduces to RHS = p (0 )p (0 ) (Q, P ( ))d d

Let x be the coordinates of Q ( not necessarily Lorentzian ) and y ( ) - coordinates of P in the same coordinate system. Since the expression is frame invariant we can choose the coordinate system any way we like. To simplify the calculation we make it satisfy the following requirements: The surface is given by the eqn. x0 = 0. The world line of the particle intersects at xj = 0. The coordinate system is a local Lorentz one at the point of intersection x = 0. Then the surface element d has only its zero component non-vanishing: d0 = d3 x = dx1 dx2 dx3 , dj = 0, j = 1, 2, 3

and the -function can be written as a product (Q, P ( )) = (x0 y0 ( )) 3 (x y( )) = (y0 ( )) 3 (x y( )) The resulting integral can be easily calculated. RHS = p (0 )p0 (0 ) = p (0 )p0 (0 ) But
dy 0 d |0

(y 0 ( )) 3 (x y( ))d3 xd = 1 (y 0 ( ))d = p (0 )p0 (0 ) dy0 d |0

= p0 (0 ) by denition of momentum, so RHS = p (0 )

H. Ex. 23.15 Proper Reference Frame [A. Dvoretskii/April 2000] (a) Its fairly straightforward to obtain the transformation law in dierential form. The hats on the right-hand side are dropped to simplify notation, and to avoid confusion we use primes for the inertial coordinates(left hand side) 7

dx = dx + ax0 dx0 + ( dx)x0 + ( x)dx0 dx0 = dx0 (1 + a x) + x0 a dx Squaring and only keeping terms linear in x get dx 2 = dx2 + 2a dxx0 dx0 + 2( x) dxdx0 (dx0 )2 = (dx0 )2 (1 + 2a x) + 2a dxx0 dx0 Given this its easy to see that the new metric is indeed ds2 = (dx0 )2 + (dx )2 = (1 + 2a x)(dx0 )2 + 2( x) dxdx0 + dx2 (b) Recall that the components of the metric in the proper frame are (again dropping the hats for simplicity of notation) g00 = (1 + 2aj xj ), g0i =
ijk j k

x , gjk = jk

And to linear order in xj , the inverse metric g is given by taking xj xj in g . And to computate the connection coecients along the world line, we only need the inverse metric at xj = 0, which is just . We have 1 0 = (g 0, g0, ) 2 which gives 0 00 = 0, 0 j 0 = aj ; j
00

= a j , i

j0

ikj

Also its not hard to see that ij all vanish. The above results can be veried by, say, GRtensor, which is straightforward and we omit here. (c) Using the connection coecients we obtained in the previous part, we nd U e0 = 00 e = i and U e j = j 0 e = 0 j 0 e0 + i = aj e0 +
ikj k j 0 ei 00 ei

= ai ei = a

ei = (a ej )U + (U , , ej , ...)

(d) Now we are away from the world line, xj = 0. However, we see that at our order of approximation , i 00 and i j 0 are still given by the expressions worked out in part (b). Plugging them into eq. (23.95), we readily get d2 x = a 2 v (dx0 )2