Solutions to General Relativity by Wald

Daniel J. Cross
November 17, 2008
Chapter 2 Solutions
1. a) Show that the overlap functions f

i
(f

j
)
1
are C

, thus completing the demonstration given in

section 2.1 that S
2
is a manifold.
b) Show by explicit construction that two coordinate systems (as opposed to the six used in the text) suce to
cover S
2
. (It is impossible to cover S
2
with a single chart, as follows from the fact that S
2
is compact, but
very open subset of R
2
is noncompact.)
a) The maps are dened as acting on the standard embedding of the sphere into R
3
. f

i
is a projection of
upper (lower) hemisphere into the plane, for example
f

2
(x
1
, x
2
, x
3
) = (x
1
, x
3
),
and the others are similar. The inverse map stretches the plane up or down into the corresponding hemi-
sphere,
(f

2
)
1
(y
1
, y
2
) = (y
1
,
_
1 ((y
1
)
2
+ (y
2
)
2
), y
2
),
and so on. The overlap map then projects out two of these coordinates, so up to exchanging coordinates it
is either of the two maps
(y
1
, y
2
) (y
1
, y
2
)
(y
1
, y
2
) (y
i
,
_
1 ((y
1
)
2
+ (y
2
)
2
)),
both of which are C

in their domain of denition, (y

1
)
2
+ (y
2
)
2
< 1.
b) Construct a chart by drawing a line from the north pole N through any given point p = N on the sphere.
The point where this line hits the plane x
3
= 0 is the image of p under the chart. This map is dened for all
points except N and is called stereographic projection from N. A similar chart dened at S (or any other
chart about S) then together with the rst form an atlas for S
2
. An explicit form of the map can be found
geometrically. The line through p and N (thought of as vector in R
3
) is (with N = (0, 0, 1))
p + t(p N) = (t(1 + p
1
), t(1 + p
2
), t(1 + p
3
) t).
Solving for when x
3
= 0 we get
t =
p
3
1 p
3
,
which then gives
x
i
=
p
i
1 p
3
.
The inverse maps are found similarly by starting with a point (y
1
, y
2
) in the plane, constructing the line
N + t(N y) through N and y and nding where this hits the sphere:
1 = |x(t)|
2
= (y
1
)
2
t
2
+ (y
2
)
2
t
2
+ (t + 1)
2
,
which has solution
t =
2
1 +|y|
2
,
and gives the mapping
(y
1
, y
2
)
1
1 +|y|
2
(2y
1
, 2y
2
, |y|
2
1).
1
2. Prove that any smooth function F : R
n
R can be written in the form equation (2.2.2).
Start with the fundamental theorem of calculus
F(x) F(a) =
_
x
a
F

(s)ds,
and make the substitution s = t(x a) + a, which linearly rescales the interval [a, x] to [0, 1]. Then
ds = dt(x a) and we get
F(x) F(a) = (x a)
_
1
0
F

(t(x a) + a)dt,
which is the result for n = 1. If n > 1 then can write F : R
n
R as F = (F
1
, . . . , F
n
), where each
F
i
: R R. Then apply the above to each F
i
. To check the derivative condition compute
F
x

a
= (x a)

x
_
1
0
F

(t(x a) + a)dt +
_
1
0
F

(t(x a) + a)dt

a
=
_
1
0
F

(a)dt
= F

(a),
which is the result.
3. a) Verify that the commutator, dened by equation (2.2.14), satises the linearity and Leibnitz properties,
and hence denes a vector eld.
b) Let X, Y, Z be smooth vector elds on a manifold M. Verify that their commutator satises the Jacobi
identity:
[[X, Y ], Z] + [[Y, Z], X] + [[Z, X], Y ] = 0.
c) Let Y
(1)
, . . . , Y
(n)
be smooth vector elds on an n-dimensional manifold M such that at each point p M
they form a basis of the tangent space T
p
M. Then, at each point, we may expand each commutator [Y
()
, Y
()
]
in this basis, thereby dening the functions C

= C

by
[Y
()
, Y
()
] = C

Y
()
.
Use the Jacobi identity to derive an equation satised by C

. (This equation is a useful algebraic relation

if the C

are constants, which will be the case if Y

(1)
, . . . , Y
(n)
are left (or right) invariant vector elds on
a Lie group.)
a) First, linearity:
v(w(f + g)) = v(w(f) + w(g)) = vw(f) + vw(g),
similarly,
w(v(f + g)) = wv(f) + wv(g),
so that
[v, w](f + g) = (vw wv)(f) + (vw wv)(g) = [v, w](f) + [v, w](g).
And for Leibnitz we have
v(w(fg)) = v(fw(g) + gw(f))
= v(f)w(f) + fv(w(g)) + v(g)w(f) + gv(w(f)),
likewise
w(v(fg)) = w(fv(g) + gv(f))
= w(f)v(f) + fw(v(g)) + w(g)v(f) + gw(v(f)).
when put together the v(f)w(f) and like terms cancel, giving
[v, w](fg) = fv(w(g)) fw(v(g)) + gv(w(f)) gw(v(f))
= f[v, w](g) + g[v, w](f).
2
b) Consider the rst term
[[x, y], z] = [xy yx, z]
= (xy yx)z z(xy yx)
= xyz yxz (zxy zyx).
Its then easy to see that writing all the terms out they will cancel pairwise.
c) That the commutator can be written as such just says that since the commutator is vector, it can be written
out as a linear combination of basis vectors and the combination you get depends on the basis vectors youre
commuting. Taking a third basis vector Y

we get
[[Y
()
, Y
()
], Y
()
] = [C

Y
()
, Y
()
]
= C

[Y
()
, Y
()
]
= C

Y
()
.
The cyclic sum over (, , ) then gives
C

+ C

+ C

= 0.
4. a) Show that in any coordinate basis, the components of the commutator of two vector elds v, w are
given by
[v, w]

= v

.
b) Let Y
(1)
, . . . , Y
(n)
be as in problem 3(c). Let Y
(1)
, . . . , Y
(n)
be the dual basis. Show that the components
Y
()

of Y
()
in any coordinate basis satisfy
Y
()

Y
()

= C

Y
()

Y
()

.
a) We have
[v, w](f) = v

_
w

f
x

_
w

_
v

f
x

_
= v

2
f
x

+ v

f
x

+
+ w

2
f
x

+ w

f
x

=
_
v

_
f
x

= [v, w]

(f)
b) Since the commutator is a vector eld, let it act on a dual basis vector Y
()
. On the one hand we get
C

Y
()
(Y
()
) = C

by denition of the dual. On the other hand we get

[Y
()
, Y
()
]Y
()
=
_
Y

()
Y

()
x

Y

()
Y

()
x

_

x

(Y
()

dx

)
= Y

()
Y

()
x

Y
()

Y

()
Y

()
x

Y
()

= Y

()
Y

()
Y
()

Y

()
Y

()
Y
()

,
where in line three we use

(Y

()
Y
()

) =

= 0. The result follows by multiplying both sides by

Y
()

Y
()

and contracting.
3
5. Let Y
(1)
, . . . , Y
(n)
be smooth vector elds on an n-dimensional manifold M which form a basis of T
p
M
for each p M. Suppose that [Y
()
, Y
()
] = 0 for all , . Prove that in a neighborhood of each p M there
exists coordinate y
1
, . . . , y
n
such that Y
(1)
, . . . , Y
(n)
are the coordinate vector elds, Y
()
=

y

.
First select an arbitrary chart about p with coordinates x

. We wish to construct coordinates y

with
the indicated properties. We can create a dierential relation between y and x by taking the dierential of
the functions y
j
, which is then
dy
j
=
y
j
x
i
dx
i
.
Now, if Y
()
=

y

, then the dual satises Y

()
= dy

, hence the Jacobian matrix in the above equation

expresses the components of the dual basis vectors in the dx
i
basis, i.e.
dy
j
= Y
(j)
= Y
(j)
i
dx
i
,
or, comparing the two equations,
Y
(j)
i
=
y
j
x
i
Now, for each j we have an equation of the form
f
x

= F

,
with the necessary integrability condition (regarding f and F as dierential forms)
d
2
f = dF =
F

= 0,
which is also sucient when the rst cohomology group H
1
vanishes. Whatever the domain of the chart we
may restrict to an open star-shaped subset where this condition holds, hence these equations have solutions
when that condition is met. By substituting in Y
(j)
for F we obtain the dierential equation from the
previous problem. Since the Y s commute, C 0, and the integrability condition is met.
6. a) Verify that the dual vectors {v

} dened by equation (2.3.1) constitute a basis of V

.
b) Let e
1
, . . . , e
n
be a basis of V and e
1
, . . . e
n
be its dual basis. Let w V and V

. Show that
w = e

(w)e

= (e

)e

.
c) Prove that the operation of contraction, equation (2.3.2), is independent of the choice of basis.
a) Let v V and let e
i
be a basis of V and e
j
be dened by e
j
(e
i
) =
j
i
. Let f = f
j
e
j
be a dual vector, then
f
j
e
j
(v
i
e
i
) = f
j
v
i
e
j
(e
i
) = f
j
v
j
,
which is zero for arbitrary v only if f
j
= 0, hence the e
j
are linearly independent. On the other hand, for an
arbitrary dual f we have
f(v) = f(v
i
e
i
) = v
i
f(e
i
) = f
i
so, it is determined by its action on the n basis vectors of V . Then we can write f = f
j
e
j
, since f
j
e
j
(e
i
) =
f
j

j
i
= f
i
, and the e
j
span V

.
b) We have w = w
i
e
i
and e
j
(w) = e
j
(w
i
e
j
) = w
j
. The second equation comes out similarly.
c) It is sucient to look at a (1,1) tensor. Then we have for any isomorphism M
CT = T(e
j
, e
j
)
= T(M
j
a
e
a
, e
b
(M
1
)
b
j
)
= M
j
a
(M
1
)
b
j
T(e
a
, e
b
)
=
b
a
T(e
a
, e
b
)
= T(e
a
, e
a
)
= CT
4
7. Let V be an n-dimensional vector space and let g be a metric on V .
a) Show that one always can nd an orthonormal basis v
(1)
, . . . , v
(n)
of V , i.e, a basis such that g(v
()
, v
()
) =

.
b) Show that the signature of g is independent of the choice of orthonormal basis.
a) Suppose n = 1, then by non-degeneracy there is a vector v such that g(v, v) = l = 0, then the vector
v

= v/
_
|l| satises g(v

, v

) = 1. Now suppose the condition holds for n 1. Choose a vector v V .

It may happen that v is null, g(v, v) = 0. We cannot choose the orthogonal complement of v because the
induced metric would be degenerate. By non-degeneracyof V , there must be at least one other vector w such
that g(v, w) = 0. If w is not null use w, otherwise use v = v + w, then
g(v

, v

) = g(v + w, v + w)
= g(v, v) + 2g(v, w) + g(w, w)
= 2g(v, w).
Normalize v

and then consider the orthogonal complement v

of v

. We claim the induced metric g

= g|
v

is non-degenerate. If v v

satises g

(v, w) = 0 for all w v

, then g(v, v

) = 0 by denition of
orthogonal complement, but then g is degenerate. Let the dimension of v

be m. By induction we have
an orthonormal basis {e
i
}. We need to show that {e
i
, v

} form a basis of V . First we need to show

that v

is linearly independent of the {e

i
}, but this follows from the denition of orthogonal complement.
Finally, take a vector v and remove its projection onto v

, that is let w = v ||v

||g(v, v

)v

. Then we have
g(w, v

) = g(v, v

) ||v

||
2
g(v, v

) = 0, and w is in v

, hence these elements span V .

b) Let {e
a
} be a basis as in a) so that g(e
a
, e
b
) =
ab
and let M be an orthonormal transformation. Then
g(M
r
a
e
r
, M
s
b
e
s
) = M
r
a
M
s
b
g(e
r
, e
s
)
=
rs
M
r
a
M
s
b
=
ab
,
where the last line follows by denition of orthonormal transformation (i.e. the matrix is orthogonal, hence
the norm of the row vectors is one).
8. a) The metric of at, three-dimensional Euclidean space is
ds
2
= dx
2
+ dy
2
+ dz
2
.
Show that the metric components g

in spherical polar coordinates r, , dened by

r = (x
2
+ y
2
+ z
2
)
1/2
cos = z/r
tan = y/x
is given by
ds
2
= dr
2
+ r
2
d
2
+ r
2
sin
2
d
2
.
b) The spacetime metric of special relativity is
ds
2
= dt
2
+ dx
2
+ dy
2
+ dz
2
.
Find the components, g

and g

, of the metric and inverse metric in rotating coordinates dened by

t

= t
x

= (x
2
+ y
2
)
1/2
cos( t)
y

= (x
2
+ y
2
)
1/2
sin( t)
z

= z,
where tan = y/x
5
a) The metric is determined by the transformation rule
g

=
x

.
The derivatives determine the inverse Jacobian matrix:
x

= (J
1
)

(since as a matrix the unprimed variables label rows and primed columns). So we have
g

= (J
1
)

(J
1
)

.
and can write the matrix equation
g

= (J
1
)
t
g(J
1
).
The inverse transformation is easy to obtain. We know z = r cos . We then have
y = xtan
y
2
= x
2
(sec
2
1)
y
2
+ x
2
= x
2
sec
2

r
2
z
2
= x
2
sec
2

r
2
(1 cos
2
) = x
2
sec
2

r
2
sin
2
= x
2
sec
2
,
which gives x = r cos sin and y = r sin sin .
It is straightforward to calculate the partial derivatives to obtain
J
1
=
_
_
cos sin r cos sin r sin sin
sin sin r sin cos r cos sin
cos r sin 0
_
_
.
Since g = I,we have
(J
1
)(J
1
)
t
=
_
_
1 0 0
0 r
2
0
0 0 r
2
sin
2

_
_
,
or
ds
2
= dr
2
+ r
2
d
2
+ r
2
sin
2
d
2
.
b) We will rst change the metric to stationary cylindrical coordinates. This is similar to the transformation
in part a) and gives the metric
ds
2
= dt
2
+ dr
2
+ r
2
d
2
+ dz
2
,
with inverse

_

t
_
2
+
_

r
_
2
+
1
r
2
_


_
2
+
_

z
_
2
The Jacobian is
J =
_
_
_
_
1 0 0 0
y x/r y 0
x y/r x 0
0 0 0 1
_
_
_
_
.
The inverse metric transforms as
g

=
x

= J

,
or
(g
1
)

= J(g
1
)J
t
,
which gives
(g
1
)

=
_
_
_
_
1 y x 0
y 1
2
y
2

2
xy 0
x
2
xy 1
2
x
2
0
0 0 0 1
_
_
_
_
.
6
The metric itself can be obtain from inverting the 3 3 submatrix. The Jacobian determinant is r, and
det(g
1
) = r
2
, so det(g
1
)

= 1 and we get (computing minors)

g =
_
_
_
_
1 +
2
r
2
y x 0
y 1 0 0
x 0 1 0
0 0 0 1
_
_
_
_
.
Chapter 3 Solutions
1. Let property (5) (the torsion free condition) be dropped from the denition of derivative operator
in section 3.1
a) Show that there exists a tensor T
c
ab
(called the torsion tensor) such that for all smooth functions f, we
have (
a

b

b

a
)f = T
c
ab

c
f.
b) Show that for any smooth vector elds X, Y we have
T
c
ab
X
a
Y
b
=
X
Y
c

Y
X
c
[X, Y ]
c
.
c) Given a metric, g, show that there exists a unique derivative operator with torsion T such that g = 0.
Derive the analog of equation 3.1.29, expressing this derivative operator in terms of an ordinary derivative
and T.
a) We note that (3.1.7):
a

b
=

b
C
c
ab

c
is still valid with torsion, so setting
b
=
b
f =

b
f, we get

b
=

b
f C
c
ab

c
f. Since

is torsion free we get
(
a

b

b

a
)f = (

a
)f (C
c
ab
C
c
ba
)
c
f
= 2C
c
[ab]

c
f,
Thus T is essentially the anti-symmetric part of C.
b) We compute
[v, w]f = v
a

a
(w
b

b
f) w
a

a
(v
b

b
f)
= v
a
w
b
(
a

b

b

a
f) + (v
a

a
w
c
w
a

a
v
c
)f
= T
c
ab
v
a
w
b

c
f + (v
a

a
w
c
w
a

a
v
c
)
c
f,
or, dropping f and basis vector
c
=
c
, [v, w]
c
= T
c
ab
v
a
w
b
+ (v
a

a
w
c
w
a

a
v
c
).
c) The condition g = 0 becomes, again

a
g
bc
=

a
g
bc
C
d
ab
g
dc
C
d
ac
g
bd
=

a
g
bc
C
cab
C
bac
.
If we add the permutation (ab) and subtract the permutation (cab) as before we get

a
g
bc
+

b
g
ac

c
g
ab
= T
bac
+ T
abc
+ 2C
c(ab)
,
so that we nay solve for the symmetric part of C given the antisymmetric part, the torsion T. Thus we have
C
c(ab)
=
1
2
(
a
g
bc
+
b
g
ac

c
g
ab
)
1
2
(T
bac
+ T
abc
) ,
or since C
abc
= C
c(ab)
+ C
c[ab]
= C
c(ab)
+ T
cab
/2, we get
C
abc
=
1
2
(
a
g
bc
+
b
g
ac

c
g
ab
)
1
2
(T
bac
+ T
abc
T
cab
) .
2. Let M be a manifold with metric g and associated derivative operator . A solution of the equation

a
= 0 is called a harmonic function. In the case where M is 2-dimensional, let be harmonic and let

ab
be an antisymmetric tensor eld satisfying
ab

ab
= 2(1)
s
, where s is the number of minuses occurring
in the signature of the metric. Consider the equation
b
=
ab

b
.
a) Show that the integrability conditions for this equation are satised, and thus, locally, there exists a solution,
. Show that also is harmonic.
7
b) By choosing and as coordinates, show that the metric takes the form
ds
2
= (, )
_
d
2
+ (1)
s
d
2

.
a) Since
ab

ab
is a constant, we have
c

ab

ab
= 0, but we can also write

ab

ab
=
ab

ab
+
ab

ab
= 2
ab

ab
,
where we use g = 0. This then requires that
e

ac
= 0. Since = d, the integrability condition is again
d
2
= d = 0, which is (
b

a

a

b
) = 0. We have
(
b

a

a

b
) =
a
(
bc

c
)
b
(
ac

c
)
= (
bc

ac

c
).
Now we can write
bc

c
as the covector
_
0
0
__

1

2
_
=
_

2
,
1
_
,
and then consider (
bc

c
)
a
as an outer product

2
,
1
_

_

1

2
_
=
_

1

1
_
= M,
and the equation becomes
(M M
t
) =
_
0
1

2
+
1

1
0
_
,
and both non-trivial terms vanish since is harmonic and thus the integrability condition is satised.
Now we want to compute
g
ab

b
=
a

a
=
a

ab

b
=
ab

b
,
where

ab

b
=
_

2
,
1
_
_

1

2
_
= (
2

2
),
which vanishes acting on (no torsion), and so is harmonic.
b) Given an arbitrary system of coordinates x

, the inverse metric transforms as

g

=
x

.
with x
1
= and x
2
= , we have
g

= g

) (

)
_
=
_
g

_
.
Analogously to part a) the o-diagonal terms become

= (
2

1

1

2
) = 0.
To evaluate the 22-component we will write it as g

and consider the tensor contracted against

the derivatives. This can be written as the matrix equation

t
g
1
,
with =

= g

. So we have

= g

=
_
0 1
1 0
__
g
11
g
12
g
21
g
22
_
=
_
g
21
g
22
g
11
g
12
_
,
8
and thus

t
g
1
=
2
_
g
21
g
22
g
11
g
12
__
g
11
g
12
g
21
g
22
__
g
21
g
11
g
22
g
12
_
=
2
det g
_
g
21
g
22
g
11
g
12
__
0 1
1 0
_
=
2
det g
_
g
11
g
12
g
21
g
22
_
=
2
det g(g
1
),
so that the 22-element simplies to (
2
det g)g

, and the inverse metric becomes

_

s
_
2
= (g

)
_
_


_
2
+
2
det g
_


_
2
_
.
Finally, since
t
g
1
=

, we have

= g

= (
2
det g)g

=
2
det g tr(

)
= 2
2
det g,
since n = 2. Now this must equal 2(1)
s
, so
2
det g = (1)
s
, and the metric becomes
_

s
_
2
=
1
(, )
_
_


_
2
+ (1)
s
_


_
2
_
,
(where we set
1
= ||

||
2
), or
ds
2
= (, )
_
d
2
+ (1)
s
d
2

.
3. a) Show that R
abcd
= R
cdab
.
b) In n dimensions, the Riemann tensor has n
4
components. However, on account of the symmetries (3.2.13),
(3.2.14), and (3.2.15), not all of these components are independent. Show that the number of independent
components is n
2
(n
2
1)/12.
a) The rst two identities imply that cyclic sum on the rst three indices vanishes, R
abcd
+ R
bcad
+ R
cabd
= 0.
If we then add four copies of this equation using the four cyclic permutations on all four indices then all
terms pairwise cancel except for four, leaving 2(R
acdb
+ R
bdac
) = 0, or R
acbd
= R
bdac
.
b) Start with R
abcd
with n
4
unconstrained components, and consider the various identities as imposing con-
straints on these components. The rst symmetry is ab = ba, which imposes n constraints when a = b and
_
n
2
_
constraints when a = b, which gives a total of
_
n
2
_
+ n =
n!
2(n 2)!
+ n =
n(n 1)
2
+ n =
n(n + 1)
2
constraints on the rst two indices. At the same time it leaves n
2
n(n+1)/2 = n(n 1)/2 terms in a, b to
be freely specied. Now, since there are choices of a, b for all c, d, that actually makes n
2
n(n + 1)/2 total
constraints.
Now the condition cd = dc imposes n(n + 1)/2 constraints on c, d for all choices of a, b. But the rst
two indices are already constraints and only n(n 1)/2 are independent, so the total number of imposed
constraints is
n(n + 1)
2
n(n 1)
2
=
1
4
n
2
(n
2
1).
The nal symmetry is abc + bca + cab = 0. This gives
_
n
3
_
constraints for a = b = c for each choice of d.
No new constraints are introduced if any two or all three of a, b, c are equal, since these cases reduce to the
antisymmetry relations considered already. Thus the number of introduced constraints is
n
_
n
3
_
=
nn!
3!(n 3)!
=
1
6
n
2
(n 1)(n 2).
9
The total number of constraints is then
1
2
n
2
(n
2
+ n) +
1
4
n
2
(n
2
1) +
1
6
n
2
(n 1)(n 2)
=
1
12
n
2
(6n
2
+ 6n + 3n
2
3 + 2n
2
6n + 4)
=
1
12
n
2
(11n
2
+ 1).
Thus there remains
n
4

1
12
n
2
(11n
2
+ 1) =
1
12
n
2
(n
2
1)
independent components. So for n = 1 . . . 5 we get
n tot indep
1 1 0
2 16 1
3 81 6
4 256 20
5 625 50
so the savings by using the symmetries is tremendous. We note that when n = 1 there are no free curvature
terms: no 1-dimensional manifolds have curvature. When n = 2 there is only one free components, which is
essentially the Gaussian curvature of the surface.
4. a) Show that in two dimensions, the Riemann tensor takes the from R
abcd
= Rg
a[c
g
d]b
.
b) By similar arguments, show that in three dimensions the Weyl tensor vanishes identically; i.e., for n = 3,
equation (2.2.28) holds with C
abcd
= 0.
a) Consider the tensor
abcd
= g
a[c
g
d]b
= (g
ac
g
bd
g
ad
g
cb
)/2. We have
2
bacd
= g
bc
g
da
g
bd
g
ca
= (g
ac
g
db
g
ad
g
cb
)
= 2
abcd
,
so that has the rst Riemann symmetry. In the same way
abdc
=
abcd
and
cdab
=
abcd
, so that has
all the symmetries of the Riemann tensor. Thus from 3b) both tensors have one free component, and thus
must be proportional: R
abcd
=
abcd
. We can establish by contracting:
g
bd
g
ac
R
abcd
= g
bd
g
ac

abcd
R = g
bd
g
ac
(g
ac
g
bd
g
ad
g
cb
)/2
R = ((tr g)
2
tr g)/2
R = ,
since tr g =
a
a
= 2. We note that 2K = R,where K is the Gaussian curvature of the surface. Further, by
taking only one contraction in the above we get
g
bd
R
abcd
= g
bd

abcd
R
ac
= Rg
bd
(g
ac
g
bd
g
ad
g
cb
)/2
R
ac
= R(g
ac
tr g g
ac
)/2
R
ac
= Rg
ac
/2
R
ac
= Kg
ac
.
b) Write R
abcd
= C
abcd
+
abcd
. has the same symmetries as R, and so we can write
R
abcd
= (a, b, c, d)
abcd
,
where (a, b, c, d) is a collection of proportionality coecients. Since C is traceless, we have
R
ac
= g
bd

abcd
= g
bd
(a, b, c, d)
abcd
=

b
(a, b, c, b)
b
abc
,
10
which can only hold if in fact (a, b, c, b) = 1. This determines all the coecients unless all four indices are
dierent, but this is impossible when n = 3, so all normalization constants are +1, so that R
abcd
=
abcd
and
C
abcd
= 0. This also demonstrates why when n > 3 no reductions are possible and we must consider the full
Riemann tensor.
5. a) Show that any curve whose tangent satises equation (3.3.2) can be reparametrized so that equation
(3.3.1) is satised.
b) Let t be an ane parameter of a geodesic . Show that all other ane parameters of take the form
at + b, where a and b are constants.
a) In coordinates the generalized geodesic equation is
d
2
x

dt
2
+

dx

dt
dx

dt
=
dx

dt
.
Introduce a new parameter s = s(t) so that d/dt = (ds/dt)d/ds. Then
d
dt
dx

dt
=
d
dt
_
ds
dt
dx

ds
_
=
_
ds
dt
_
2
d
2
x

ds
2
+
dx

ds
d
dt
_
ds
dt
_
.
Now we rewrite the geodesic equation as
_
ds
dt
_
2
d
2
x

ds
2
+
_
ds
dt
_
2

dx

ds
dx

ds
=
dx

ds
_

ds
dt

d
dt
_
ds
dt
__
,
so that to make the r.h.s zero we need to solve the equation(setting ds/dt = )
=
d
dt
dt =
d

which integrates to
=
0
e
(tt
0
)
.
b) Let s
1
and s
2
be ane parameters with
i
= ds
i
/dt in terms of some arbitrary parameter t. From part a)
we have
ds
1
ds
2
=

1

2
=
(
1
)
0
(
2
)
0
= const.
Thus ds
1
/ds
2
= const and s
1
= as
2
+ b, for constants a, b.
6. The metric of Euclidean R
3
in spherical coordinates is ds
2
= dr
2
+ r
2
(d
2
+ sin
2
d
2
).
a) Calculate the Christoel components

in this coordinate system.

b) Write down the components of the geodesic equation in this coordinate system and verify that the solutions
correspond to straight lines in Cartesian coordinates.
a) There are only three non-trivial derivatives of the metric tensor:
g

r
= 2r,
g

r
= 2r sin
2
,
g

= 2r
2
sin cos .
Since the metric is diagonal the components of the inverse are the inverse of the components. We will
consider the six dierent combinations for the bottom indices, which gives
2

r
= g

(
r
g

g
r

g
r
) = g

r
g

rr
= g

(2
r
g
r

g
rr
) = 0
2

= g

(2

) = g
r

r
g

= g

) = g

= g

(2

) = g
r

r
g

r
= g

(
r
g

g
r

g
r
) = g

r
g

,
11
and the non-zero components come out to be

r
= g

r
g

/2 =
1
r

= g
rr

r
g

/2 = r

= g

/2 = cot

= g
rr

r
g

/2 = r sin
2

= g

/2 = sin cos

r
= g

r
g

/2 =
1
r
b) If we write dx

/dt = x

= (v
r
, v

, v

) we have
0 = v
r
+
r

+
r

= v
r
rv

r sin
2
v

0 = v

+ 2

r
v
r
v

= v

+
2
r
v
r
v

sin cos v

0 = v

+ 2

+ 2

r
v
r
v

= v

+ 2 cot v

+
2
r
v
r
v

Consider x = r cos sin . We have

x = r cos sin r

sin sin + r

cos cos ,
and similarly,
x = r cos sin 2 r

sin sin + 2 r

cos cos
r

sin sin r

2
cos sin 2r

sin cos
+r

cos cos r

2
cos sin .
If we then substitute in for r = v
r
, etc. all the terms cancel and x = 0. The other axis are worked out
similarly. (Is there a better way to do this?)
7. As shown in problem 2, an arbitrary Lorentz metric on a two-dimensional manifold locally always can
be put in the form ds
2
=
2
(x, t)[dt
2
+dx
2
]. Calculate the Riemann curvature tensor of this metric (a) by
the coordinate basis methods of section 3.4a and (b) by the tetrad methods of section 3.4b.
a) We know that when n = 2 there is only one free component of R which we take to be R
1212
. To calculate
this we need only R
2
121
g
22
, since g is diagonal. Thus we need
R
2
121
=
2

2
11

1

2
21
+
e
11

2
e2

e
21

2
e1
.
Now we note that g
ab
=
2

ab
, so that

c
g
ab
= 2
ab

c

=
2

g
ab

c

= 2g
ab

c
ln .
Likewise g
ab
=
2

ab
and
c
g
ab
= 2g
ab

c
ln . Since the metric is diagonal we have g
11
g
11
= g
22
g
22
= 1.
Moreover, g
22
g
11
= 1. We compute the Christoel symbols using these properties to get
2
1
11
= g
1r
(2
1
g
1r

r
g
11
) = g
11

1
g
11
= 2
1
ln
2
1
12
= g
1r
(
1
g
2r
+
2
g
1r

r
g
12
) = g
11

2
g
11
= 2
2
ln
2
1
22
= g
1r
(2
2
g
2r

r
g
22
) = g
11

1
g
22
= 2
1
ln
2
2
11
= g
2r
(2
1
g
1r

r
g
11
) = g
22

2
g
11
= 2
2
ln
2
2
12
= g
2r
(
1
g
2r
+
2
g
1r

r
g
12
) = g
22

1
g
22
= 2
1
ln
2
2
22
= g
2r
(2
2
g
2r

r
g
22
) = g
22

2
g
22
= 2
2
ln .
Summarizing, we have

1
11
=
1
22
=
2
12
=
1
ln

1
12
=
2
11
=
2
22
=
2
ln .
The rst term in R
2
121
becomes

2

2
11
= (
2
)
2
ln .
12
The second term becomes

1

2
21
= (
1
)
2
ln .
The third term becomes

1
11

2
12
+
2
11

2
22
= (
1
ln )
2
+ (
2
ln )
2
.
The nal term is

1
21

2
11

2
21

2
21
= (
2
ln )
2
(
1
ln )
2
,
which will cancel the third term. All together, we get
R
2
121
=
_
(
1
)
2
+ (
2
)
2

ln
=

ln
=
2
ln .
Finally, lowering the index with g
22
=
2
gives
R
1212
=
2

2
ln .
b) Next we calculate using the tetrad method. The only non-trivial connection one-forms are

121
=
112

221
=
212
.
We can construct an orthonormal basis by
e
1
=
1

t
e
2
=
1

x
,
since g(e
i
, e
j
) =
2

ab
(e
i
)
a
(e
j
)
b
=
2

ab
(
1

a
i
)(
1

b
j
) =
ij
. The components in the coordinate basis are
(e
1
)
i
=
1
(1, 0)
(e
2
)
i
=
1
(0, 1)
(e
1
)
i
= (1, 0)
(e
2
)
i
= (0, 1)
This simplies equation (3.4.21) for Riemann to
R
(1212)
=
e
1

212

e
2

112
+
2
112

2
212
,
where the s here are

= (e

)
a

a
, and these latter components are determined by equation (3.4.25)

a
(e

)
b

b
(e

)
a
=

((e

)
a

b
(e

)
b

a
).
Due to the anti-symmetry we need only calculate for a, b = 1, 2 and = 1, 2:

1
(e
1
)
2

2
(e
1
)
1
=

((e

)
1

21
(e

)
2

11
)

2
=
112

112
=
2
ln ,
and

1
(e
2
)
2

2
(e
2
)
1
=

((e

)
1

22
(e

)
2

12
)

1
=
221

212
=
1
ln .
The s in Riemann are then related to the above by

212
= (e
2
)
2

212
=
1

1
ln

112
= (e
1
)
1

112
=
1

2
ln .
13
Now we can calculate the derivative terms in Riemann:

e
1

212
= (e
1
)
a

a
(
1

1
ln )
=
1

1
(
1

1
ln )
=
1
_

1

1
ln +
1
(
1
)
2
ln
_
=
2
((
1
ln )
2
(
1
)
2
ln )
and similarly

e
2

112
= (e
2
)
a

a
(
1

2
ln )
=
1

2
(
1

2
ln )
=
1
_

2

2
ln +
1
(
2
)
2
ln
_
=
2
((
2
ln )
2
(
2
)
2
ln ).
The nal two terms are

2
112
=
2
(
2
ln )
2

2
212
=
2
(
1
ln )
2
.
These two terms will cancel the corresponding terms in the derivative terms, leaving
R
(1212)
=
2
_
(
1
)
2
+ (
2
)
2

ln
=
2
ln .
Finally, as a sanity check, the frame and coordinate versions are related through the tetrad by
R
(abcd)
= R

(e
a
)

(e
b
)

(e
c
)

(e
d
)

,
or
R
(1212)
= R
1212
(e
1
)
1
(e
2
)
2
(e
1
)
1
(e
2
)
2
=
2
ln (
4
)
=
2
ln .
8. Using the antisymmetry of
a
in and , equation (3.4.15), show that

= 3
[]
2
[]
.
Use this formula together with equation (3.4.23) to solve for

in terms of commutators (or antisym-

metrized derivatives) of the orthonormal basis vectors.
Due to the anti-symmetry in the last two indices, the full anti-symmetrization reduces to a double cyclic
sum:
3
[]
=
1
2
(

)
=

.
The second term is
2
[]
=

,
which cancels the last two terms, leaving the rst, which is the identity. Let

(e

)
a
[e

, e

]
a
= 2
[||]
from equation (3.4.23). By anti-symmetrizing over all indices we get
1
2

[]
=
[]
.
Now, we can write
3
[]
=
[]
+
[]
+
[||]
=

+
[]
+
[||]
.
Comparing this against the equation from the rst part of the problem we see that
3
[]
=
||]
=
1
2

,
so that

= 3
[]
2
[]
=
3
2

[]

1
3

.
14