General Relativity: Homework 3 Solutions

1. Carroll Problem 3.1

Verify these consequences of metric compatability (∇σ gµν = 0):

∇σ g µν = 0

∇λ µνρσ = 0

Solution:
To verify the first claim, recall that g µν is the inverse of gµν so that we have:

g µν gµν = δµµ = n

where n is the dimension of the manifold. Now take the covariant derivative of both sides, noting that the
covariant derivative of a constant is zero,
∇σ (g µν gµν ) = 0
g µν (∇σ gµν ) + (∇σ g µν )gµν = 0
⇒ gµν (∇σ g µν ) = 0
And therefore ∇σ g µν = 0.

To verify the second claim note Carroll equations 2.69 and 2.70 which imply

µνρσ µνρσ = ˜µνρσ ˜µνρσ = −n!

where n is the dimension of the manifold we are working over, and we are assuming that the metric is
Lorentzian (this assumption is uneccessary, it’s just why the negative sign is there, all we really need is that
the right-hand side is a constant). Taking the covariant derivatives of both sides we obtain

∇λ (µνρσ )µνρσ + µνρσ ∇λ (µνρσ ) = 0

Since µνρσ is an honest to goodness tensor, we may relate it to µνρσ by factors of the g µν . Specifically we
have
µνρσ = g µα g νβ g ρδ g σγ αβδγ
By inserting this expression into the covariant derivative, using the Leibniz rule, and the fact that ∇σ g µν = 0,
we arrive at the expression

∇λ (µνρσ )µνρσ + µνρσ g µα g νβ g ρδ g σγ ∇λ (αβδγ ) = ∇λ (µνρσ )µνρσ + αβδγ ∇λ (αβδγ ) = 0

Replacing the dummy indices in the second expression we arrive at

2∇λ (µνρσ )µνρσ = 0

Multiplying both sides by µνρσ implies ∇λ µνρσ = 0.

1
2. Carroll Problem 3.5
Consider a 2-sphere with coordinates (θ, φ) and metric

ds2 = dθ2 + sin2 θdφ2

(a) Show that lines of constant longitude (φ = constant) are geodesics, and that the only line of constant
latitude (θ = constant) that is a geodesic is the equator (θ = π/2).
(b) Take a vector with components V µ = (1, 0) and parallel-transport it once around a circle of constant
latitude. What are the components of the resulting vector, as a function of θ?

Solution:
(a) To verify a curve is a geodesic we need to write the geodesic equation for this metric. First we need
the connection coefficients, to which we appeal to Carroll equation (3.154), although it is a worthwhile
exercise to derive these for yourself. The nonzero coefficients are:

Γθφφ = − sin θ cos θ

Γφθφ = Γφφθ = cot θ

Now we may write out the geodesic equation(s):

d2 xµ ρ
µ dx dx
σ
+ Γ ρσ =0
dλ2 dλ dλ
which give us the two equations:
d2 θ dφ
2
2
− sin θ cos θ =0
dλ dλ
d2 φ dθ dφ
2
+ 2 cot θ =0
dλ dλ dλ
d2 θ
In the case that φ is a constant the first equation reads dλ2 = 0, while the second equation is trivial.
Thus the curve xµ = (λ, φ) is a geodesic.
If instead we assume θ is a constant we get:
dφ
2
sin θ cos θ =0
dλ
d2 φ
=0
dλ2
Since we can’t set dφ
dλ = 0 (because then the ’curve’ would just be a point), if the first equation is
to be satisfied we must have θ = π/2 (note that θ = 0 is not a well-defined coordinate). Therefore
xµ = (π/2, λ) is a geodesic, and in fact it is the only geodesic (up to reparameterization) where θ is
constant.
(b) In order to parallel transport a vector we must choose a path (with parameterization) to transport
along. Any parameterization will work, however it is best to keep it simple, so choose xµ = (θ, λ). Now
we must solve the parallel transport equation:
d µ dxσ ρ
V + Γµσρ V =0
dλ dλ
dxµ
Note that dλ = (0, 1) so that we get the two equations:

dV θ
− sin θ cos θ V φ = 0 (1)
dλ

2
 dV φ
+ cot θ V θ = 0 (2)
dλ
To solve this system differentiate (2) with respect to λ (keeping in mind that θ is a constant), solve for
dV θ
dλ , plug it into (1) and do a little algebra to obtain:

d2 V φ
+ cos2 θ V φ = 0
dλ2
We can do the same procedure (differentiating (1) and plugging into (2)) to obtain a differential
equation for V θ :
d2 V θ
+ cos2 θ V θ = 0
dλ2
These equations are just those for a simple harmonic oscillator (with frequency cos θ), so the form of
the solutions are:
V θ (λ) = A cos(λ cos θ) + B sin(λ cos θ)
V φ (λ) = C cos(λ cos θ) + D sin(λ cos θ)
Now apply the boundary condition V (0) = (1, 0) which gives us that A = 1 and C = 0. To pin down
the last two constants, appeal to one of the original equations, say (1):

dV θ
= sin θ cos θ V φ
dλ
cos θ(− sin(λ cos θ) + B cos(λ cos θ)) = sin θ cos θ(D sin(λ cos θ))
− sin(λ cos θ) + B cos(λ cos θ) = sin θD sin(λ cos θ)
Comparing the coefficients of sin(λ cos θ) and cos(λ cos θ) on both sides we see that we must have B = 0
−1
and D = sin θ . Thus we have the solution to the parallel transport equation:
 
sin(λ cos θ)
V µ (λ) = cos(λ cos θ), −
sin θ

Plugging in λ = 2π we see that after parallel transporting V µ around a circle of constant latitude θ
the resulting vector is:  
sin(2π cos θ)
V µ (2π) = cos(2π cos θ), −
sin θ

As a check we can see that the vector is unchanged (i.e. V µ (2π) = (1, 0)) if we parallel transport
around the equator (θ = π/2), as expected.

3. Carroll Problem 3.6

A good approximation to the metric outside the surface of the Earth is provided by

ds2 = −(1 + 2Φ)dt2 + (1 − 2Φ)dr2 + r2 (dθ2 + sin2 θdφ2 )

where
GM
Φ=−
r
may be thought of as the familiar Newtonian gravitational potential. Here G is Newton’s constant and M
is the mass of the Earth. For this problem Φ may be assumed to be small.

(a) Imagine a clock on the surface of the Earth at a distance R1 from the Earth’s center, and another clock
on a tall building at distance R2 from the Earth’s center. Calculate the time elapsed on each clock as
a function of the coordinate time t. Which clock moves faster?

3
 (b) Solve for a geodesic corresponding to a circular orbit around the equator of the Earth (θ = π/2). What
is dφ/dt?
(c) How much proper time elapses while a satellite at radius R1 (skimming along the surface of the Earth,
neglecting air resistance) completes one orbit? You can work to first order in Φ if you like. Plug in
the actual numbers for the radius of the Earth and so on to get an answer in seconds. How does this
number compare to the proper time elapsed on the clock stationary on the surface?

Solution:
(a) For a clock on the surface of the Earth, the world-line (parameterized by t) is xµ = (t, R1 , θ, ωt), where
π rad
ω is the angular velocity of the Earth (ω = 12 hrs ). Now calculate the proper-time:

Z tr Z tq
dxµ dxν 0
∆τ = −gµν dt = (1 + 2Φ) − R12 sin2 θω 2 dt0
0 dt dt 0
r
2GM
∆τ = 1 − − R12 sin2 θω 2 ∆t
R1
Now let’s reinstate the factors of c. 2GM 2
R1 has units of velocity squared, so we should divide that by c .
The same goes for the second term under the square root. Also note that R1 ω << c, so we can drop
that term (we also have R12 ω 2 << 2GM 2GM
R1 , which justifies leaving in R1 ), leaving us with:
r
2GM
∆τ = 1− ∆t
c2 R1

For the clock on a tall building we replace R1 with R2 and see that the proper-time is decreased. Note
that this effect is purely due to the difference in gravitational potential (the effect due to time dilation
associated with the difference in speeds is negligible).
(b) To find a geodesic we must look at the geodesic equations. In fact, because we are looking for a specific
type of geodesics, namely one with constant radius, r and θ = π/2 we do not need to look at all of the
equations. First look at the geodesic equation for xµ = r which reads:

d2 r dxρ dxσ
2
+ Γrρσ =0
dλ dλ dλ
Keeping in mind that we are restricting ourselves to paths with constant r and constant θ, this expres-
sion simplifies to
dt
2 dφ
2 dφ dt
Γrtt + Γrφφ + 2Γrφt =0
dλ dλ dλ dλ
Now calculate just the necessary connection coefficients
1 rρ 1
Γrtt = g (∂t gtρ + ∂t gtρ − ∂ρ gtt ) = − g rr ∂r gtt
2 2
 
r 1 1 2GM GM
Γtt = ∂r (1 − )= 2
2 (1 − 2Φ) r r (1 − 2Φ)

1 rρ 1
Γrφφ = g (∂φ gφρ + ∂φ gφρ − ∂ρ gφφ ) = − g rr ∂r gφφ
2 2
r sin2 θ
 
1 1
Γrφφ =− ∂r (r2 sin2 θ) = −
2 (1 − 2Φ) (1 − 2Φ)

1 rρ
Γrφt = g (∂φ gtρ + ∂t gφρ − ∂ρ gφt ) = 0
2

4
 Now the geodesic equation reads:

GM dt
2 r sin2 θ dφ
2
− =0
r2 (1 − 2Φ) dλ (1 − 2Φ) dλ
r
dφ GM dt
⇒ = (3)
dλ r3 dλ
We still haven’t found the geodesic, to do so now consider the equation for xµ = t:

d2 t ρ
t dx dx
σ
+ Γ ρσ =0
dλ2 dλ dλ
Again, considering only paths such that r is constant and θ = π/2 this simplifies to:

d2 t dt
2 dφ
2 dt dφ
2
+ Γttt + Γtφφ + Γttφ =0
dλ dλ dλ dλ dλ
Now look at the equation for the necessary connection coefficients
1 tt
Γtµν = g (∂µ gνt + ∂ν gµt − ∂t gµν )
2
and we can see that for µ, ν ∈ {t, φ} the connection coefficients will all be zero. The geodesic equation
is now just
d2 t
=0
dλ2
the simplest non-trivial solution to which is t = λ. Taking this as our parameterization, (3) then
implies: r
dφ GM
=
dt r3
Putting it all together, the (up to reparameterization) geodesic corresponding to a circular orbit (of
radius r) around the equator (θ = π/2) is:
r
µ GM
x (t) = (t, r, π/2, t)
r3

(c) For the satellite skimming the surface of the Earth (at θ = π/2) we have that the elapsed proper-time
for a complete orbit is: s
2GM R2 ω 2 2π
∆τ = 1 − 2 − 12
c R1 c ω
q
where ω = GM R3
and we have included the term related to the motion of the satellite because it is on
1
the same order as the gravitational potential. Plugging in the expression for ω we have
s r r
2GM R12 GM R13 R13 3R2
∆τ = 2π 1 − 2 − 2 3 = 2π − 21
c R1 c R1 GM GM c
s
(6.371 × 106 m)3 3(6.371 × 106 m)2
∆τ = 2π − = 5061s
2
(6.674 × 10−11 Nkgm2 )(5.972 × 1024 kg) (2.998 × 108 ms )
2

Now to calculate the the elapsed proper-time for the clock stationary on the surface of the Earth,
borrow the result from part (a), setting θ = π/2.
r r r r
2GM 2π 2GM R13 R13 2R2
∆τ = 1 − 2 = 2π 1 − 2 = 2π − 21
c R1 ω c R1 GM GM c

5
 s
(6.371 × 106 m)3 2(6.371 × 106 m)2
∆τ = 2π − = 5061s
2
(6.674 × 10−11 Nkgm2 )(5.972 × 1024 kg) (2.998 × 108 ms )
2

For a clock stationary on the surface the elapsed proper-time (during the orbit of the satellite) is
exactly the same as that of the satellite! This is because both are the same distance from the center of
the Earth, and therefore they both have the same gravitational potential (and the effect due to time
dilation clearly has a negligible effect).
To fully convince ourselves that this effect is due purely to gravity, consider a stationary clock at the
top of Mount Everest (≈ 9000m high). Repeating the previous calculation we obtain ∆τ = 5072s.
Gravity slows the passage of proper-time! A true means for slowing the effects of aging.

4. Carroll Problem 3.13

Find explicit expressions for a complete set of Killing vector fields for the following spaces:

(a) Minkowski space, with metric ds2 = −dt2 + dx2 + dy 2 + dz 2 .

(b) A spacetime with coordinates (u, v, x, y) and metric

ds2 = −(dudv + dvdu) + a2 (u)dx2 + b2 (u)dy 2

where a and b are unspecified functions of u. This represents a gravitational wave spacetime. (Hints,
which you need not show: there are five Killing vectors in all, and all of them have a vanishing u
component K u .)

Be careful, in all of these cases, about the distinction between upper and lower indices.

Solution:
All answers are vectors with upper indices.
(a) First note that all of the metric components are independent of t, x, y, z and therefore the four trans-
lations of those coordinates correspond to symmetries of the metric. These symmetries are generated
by the four Killing vectors:
T µ = (1, 0, 0, 0)
X µ = (0, 1, 0, 0)
Y µ = (0, 0, 1, 0)
Z µ = (0, 0, 0, 1)
To see that each of these satisfy ∇(µ Kν) = 0 first note that all of the connection coefficients vanish for
the Minkowski metric, and therefore ∇(µ Kν) = ∂(µ Kν) . Since all of the components of these vectors are
constant, all of the partial derivatives will vanish as well and the Killing vector condition is satisfied.
Next consider the three generators of purely spatial rotations (see Carroll pages 138-9 for a more
detailed discussion of these). They are (with lowered indices):

Rxµ = (0, 0, −z, y)

Ryµ = (0, z, 0, −x)

Rzµ = (0, −y, x, 0)
Let’s explicitly verify that Rxµ satisfies the Killing vector condition. First check that ∇µ Rxµ = 0, but
this is clear since the µth component of Rxµ is independent of the µth coordinate (e.g. the y component
is independent of y). Next condsider terms where µ = 6 ν, in fact the only non-trivial combination is:

∇y Rxz + ∇z Rxy = ∂y (y) + ∂z (−z) = 1 + (−1) = 0

6
 Similar calculations show that all three of these vectors satisfy the Killing vector condition. All that’s
left to do is raise their indices, but that’s easy for this metric:

Rxµ = (0, 0, −z, y)

Ry µ = (0, z, 0, −x)
Rz µ = (0, −y, x, 0)

Finally, by looking at the calculations for the three rotation generators, let’s guess that the three
vectors corresponding to
Bxµ = (−x, t, 0, 0)
Byµ = (−y, 0, t, 0)
Bzµ = (−z, 0, 0, t)
are also Killing vectors (corresponding to boosts along each axis). Again, verify by following the same
argument as for the rotations. Let’s check the non-trivial case for Bxµ :

∇t Bxx + ∇x Bxt = ∂t (t) + ∂x (−x) = 1 + (−1) = 0

Now we need to again raise the indices, and in this case we need to be wary of the metric we are using:

Bxµ = (x, t, 0, 0)

By µ = (y, 0, t, 0)
Bz µ = (z, 0, 0, t)

In all we have 10 Killing vectors, which is in agreement with the fact that the Poincaré group (isometries
of the Minkowski metric) is 10 dimensional.
(b) In order to verify Killing’s equation we will need to calculate the connection coefficients, a quick
observation and use of the hints will let us get away without calculating all of the coefficients. Since,
as the hints indicate, the u-component of all of the Killing vectors is zero, by lowering the index,
K u gµν = −Kv , we see that this implies Kv = 0 for all Killing vectors. Therefore, by looking at the
covariant derivative of a 1-form ∇µ Kν = ∂µ Kν − Γρµν Kρ we conclude that we do not need to calculate
the Γvµν coefficients (upper index equals v). We do however need to calculate the rest of them.
We will proceed with the variational technique, that is consider variations of the integral (Carroll eqn.
3.49)
dxµ dxν
Z Z
1 1 du dv dx
2 dy
2
I= gµν dτ = (−2 + a2 (u) + b2 (u) )dτ
2 dτ dτ 2 dτ dτ dτ dτ
To get the coefficients with upper index u, we will actually need to consider variations of v (this is due
to the cross terms in the metric). Under variations v 7→ v + δv the only term in the integrand that
dv
changes is dτ dv
7→ dτ + d(δv)
dτ and therefore the variation in I is
Z Z 2
1 du d(δv) d u
δI = −2 = δvdτ
2 dτ dτ dτ 2
where the second equality comes from integrating by parts. Requiring that the variation vanishes we
are left with
d2 u
=0
dτ 2
which implies that all coefficients with upper index u vanish, Γuµν = 0.
Now we want the coefficients with upper index x, for which we consider variations x 7→ x + δx. Under
such variations the only change in the integrand is
dx
2 dx d(δx)
2 dx
2 dx d(δx)
7→ + = +2 + O(δx2 )
dτ dτ dτ dτ dτ dτ

7
Therefore the variation of the integral (up to first order in δx) is
Z
dx d(δx)
δI = a2 (u) dτ
dτ dτ
Integrating by parts we obtain

d2 x
Z
da du dx

δI = − a2 (u) 2
+ 2a(u) dτ
dτ du dτ dτ
Setting the variation equal to zero leads to the equation

d2 x da du dx
a2 (u) 2
+ 2a(u) =0
dτ du dτ dτ
d2 x 2 da du dx
⇒ 2
+ =0
dτ a(u) du dτ dτ
Now we may read off the connection coefficients
1 da
Γxux = Γxxu =
a(u) du

Similarly we obtain
1 db
Γyuy = Γyyu =
b(u) du

Now that we have all of the necessary connection coefficients we can now go about finding/guessing the
Killing vectors. We readily obtain three of them by noting that the metric has no explicit dependence
on the coordinates v, x, y, which implies that it is invariant under translations of those coordinates and
therefore we have that the vectors (in (u, v, x, y) coordinates)

V µ = (0, 1, 0, 0)

X µ = (0, 0, 1, 0)
Y µ = (0, 0, 0, 1)
are Killing vectors. Let’s verify that all three of these vectors satisfy Killing’s equation. First let’s
show it for Vµ . We need to be careful because lowering the index changes the components around,
specifically Vµ = (−1, 0, 0, 0). Now we have

∇ν Vµ = ∂ν Vµ − Γρνµ Vρ = 0 − Γuνµ (−1) = 0

where we’ve used that all of the components of Vµ are constants, hence all of the partial derivatives
vanish, and that all Γuµν = 0.
Next we should verify that Xµ satisfies Killing’s equation (and by similar calculations, that Yµ does
as well). This is a little trickier than in the previous case, because now lowering the index gives us
Xµ = (0, 0, a2 (u), 0).
∇ν Xµ = ∂ν Xµ − Γρνµ Xρ = ∂ν Xµ − a2 (u)Γxνµ
The most interesting cases we need to check are ν = u, µ = x and vice versa. In the former we get
1 da
da
∇u Xx = ∂u a2 (u) − a2 (u) = a(u)
a(u) du du

Switching indices we have

1 da
da
∇x Xu = ∂x (0) − a2 (u) = −a(u)
a(u) du du

8
 da da
⇒ ∇u Xx + ∇x Xu = a(u) − a(u) =0
du du
Unfortunately those are the only obvious symmetries of the metric, and as the hint suggests, we still
have two more Killing vectors to find. To find the last two vectors let’s explore a bit with the covariant
derivative. Look at
∇u Ku = ∂u Ku + Γρuu Kρ
Since all of the relevant connection coefficients with lower labels uu are zero, we see that in order
for Kµ to satisfy Killing’s equation we must have ∂u Ku = 0, that is, Ku must be independent of u.
Following the same logic with ∇x Kx we see that Kx must be independent of x. Now let’s look at the
implications of the cross terms.
2 da
∇x Ku + ∇u Kx = ∂x Ku − Γρxu Kρ + ∂u Kx − Γρux Kρ = ∂x Ku + ∂u Kx − Kx
a(u) du

In order for Kµ to satisfy Killing’s equation we must have

2 da
∂u Kx − Kx = −∂x Ku
a(u) du

Because Ku is independent of u, and Kx is independent of x, the right hand side must equal a constant
(i.e. Ku = λx for some constant λ, which we can set to -1 without loss of generality). Therefore we
have the linear first-order differential equation
2 da
∂u Kx − Kx = 1
a(u) du

Following the procedure for solving such equations (i.e. looking it up in your favorite differential
equations book) we solve for the integrating factor µ(u)
Z 
2 da
µ(u) = exp − du = exp{−2 ln |a(u)| + C} = Da−2 (u)
a(u) du

for some constant D. Now that we have the integrating factor the solution is given by
Z  1 2
Z Z
1
−1 −2 2
Kx = µ(u) µ(u)du = a (u) Da (u)du = a (u) du
D a2 (u)

Note that for a problem with boundary conditions there should be some constants hanging around
from the improper integrals, however since we do not have any such conditions we can just seek the
simplest such solution. Thus with lower indices our solution is
 Z 
du
Kµ = − x, 0, a2 (u) , 0
a2 (u)

You should check for yourself that this indeed solves the equation. Raising the indices we get the
Killing vector  Z 
du
K µ = 0, x, , 0
a2 (u)

Following the same procedure we obtain the final Killing vector

 Z 
du
K̃ µ = 0, y, 0,
b2 (u)

9
5.
In five-dimensional spacetime, xM = (xµ , y), the Randall-Sundrum metric is given by

ds2 = gM N dxM dxN = e−2ky ηµν dxµ dxν + dy 2

where k is a constant and ηµν =diag(−1, 1, 1, 1).

(a) Calculate the Christoffel connection coefficients. (Rather than direct calculation via Eq. (3.27), you
will probably find it easier to use the variational technique!)
(b) Calculate the Riemann tensor, Ricci tensor, and Ricci Scalar.
(c) Is this a maximally symmetric space? Verify by using Eq. (3.191).

Solution:
(a) Using the variational technique, we look for stationary points in the integral

dxµ dxν dx0
2 dx1
2 dx2
2 dx3
2

Z Z    
1 1 −2ky dy
2
I= gµν dτ = e − + + + + dτ
2 dτ dτ 2 dτ dτ dτ dτ dτ

First let’s consider variations x1 7→ x1 + δx1 . Under such variations the only change in the integrand is

dx1
2 dx1 d(δx1 )
2 dx1
2 dx1 d(δx1 )

+ O (δx1 )2


7→ + = +2
dτ dτ dτ dτ dτ dτ
Therefore, up to first-order in δx1 the variation of the integral is

1  −2ky dx1 d(δx1 ) 

Z
δI = 2e dτ
2 dτ dτ
1 d(δx1 )
Integrate by parts with u = e−2ky dx
dτ and dv = dτ giving us

dx1 1 d2 x1 −2ky 1
Z  
−2ky dx dy
δI = e−2ky δx |boundary − e − 2ke δx1 dτ
dτ dτ 2 dτ dτ

Demanding that the variation of x1 vanish at the boundary, and setting the variation of the integral equal
to zero we get the following equation
d2 x1 dx1 dy
− 2k =0
dτ 2 dτ dτ
which implies the non-zero Γ1ij are
Γ114 = Γ141 = −k
where we use the convention that y = x4 . Generalizing this calculation for other i’s we get

Γii4 = Γi4i = −k, i = 0, 1, 2, 3

0
2
Note that the minus sign in front of dxdτ does not affect the result of the calculation because there are no
off-diagonal terms in the metric and we are setting the variation equal to zero.
Now consider variations y 7→ y + δy. Every term in the integrand will change as a result of this variation.
One should look familiar by now

dy
2 dy d(δy)
2 dy
2 dy d(δy)
+ O (δy)2


7→ + = +2
dτ dτ dτ dτ dτ dτ
The other terms involve the exponential factor, whose variation is

e−2ky 7→ e−2k(y+δy) = e−2ky (1 − 2k δy + O(δy 2 ))

10
Therefore the variation of the integral is (dropping terms O(δy 2 ))
dx0
2 dx1
2 dx2
2 dx3
2
Z    
1 dy d(δy)
δI = − 2k e−2ky δy − + + + +2 dτ
2 dτ dτ dτ dτ dτ dτ
Integrating the last term by parts and dropping the boundary term we arrive at
dx0
2 dx1
2 dx2
2 dx3
2 d2 y
Z    
δI = − k e−2ky − + + + − 2 δy dτ
dτ dτ dτ dτ dτ
dx0
2 dx1
2 dx2
2 dx3
2 d2 y
 
δI = 0 ⇒ k e−2ky − + + + + 2 =0
dτ dτ dτ dτ dτ
Now we read off the following connection coefficients
Γ400 = −ke−2ky
Γ4ii = ke−2ky , i = 1, 2, 3
All other connection coefficients equal 0.

(b) Since the Riemann tensor has so many terms, even if you include symmetries, it is often helpful to
turn this calculation into a matrix calculation (matrices are the natural way of manipulating several equa-
tions at once). For this particular metric it isn’t actually all that difficult to just do this term by term, since
there are only so many non-zero connection coefficients, however for more complicated metrics this technique
comes in quite handy.
The definition of the Riemann tensor is:
Rρ σµν = ∂µ Γρνσ − ∂ν Γρµσ + Γρµλ Γλνσ − Γρνλ Γλµσ
Define the matrices
(Γµ )ρσ = Γρµσ
(Rµν )ρσ = Rρ σµν
where ρ is the row index and σ is the column index. Rµν should not be confused with the Ricci tensor Rµν .
What we have defined is merely a matrix that is useful for doing computations. With these definitions the
definition of the Riemann tensor can be recast as the matrix equation
Rµν = ∂µ Γν − ∂ν Γµ + Γµ Γν − Γν Γµ
The components of the Γ matrices can be written as
(Γ0 )ρσ = −k δ0ρ δ4σ − ke−2ky δ4ρ δ0σ
(Γi )ρσ = −k δiρ δ4σ + ke−2ky δ4ρ δiσ ; i = 1, 2, 3
3
X
(Γ4 )ρσ = −k δρi δσi
i=0
Also note that because of the antisymmetry of the Riemann tensor in the last two indices, it suffices to
consider just the case where µ < ν.
Let’s do some example calculations:

R01 = ∂0 Γ1 − ∂1 Γ0 + Γ0 Γ1 − Γ1 Γ0
Because none of the connection coefficients depend on x0 or x1 the derivative terms drop out and we have
     
0 0 0 0 −k 0 0 0 0 0 0 0 0 0 0 0 0 0 0 −k

 0 0 0 0 0  0
 0 0 0 −k  0
  0 0 0 −k 

 0 0 0 0 0 
R01 =  0 0 0 0 0  0 0 0 0 0 −0 0 0 0 0   0 0 0 0 0 
     
 0 0 0 0 0  0 0 0 0 0  0 0 0 0 0  0 0 0 0 0 
−ke−2ky 0 0 0 0 0 ke−2ky 0 0 0 0 ke−2ky 0 0 0 −ke−2ky 0 0 0 0

11
 −k 2 e−2ky
 
0 0 0 0
−k 2 e−2ky 0 0 0 0
 
=
 0 0 0 0 0

 0 0 0 0 0
0 0 0 0 0
Doing the same calculation for R02 and R03 we find the following non-zero components to the Riemann
tensor
0
Ri0i i
= R00i = −k 2 e−2ky ; i = 1, 2, 3
The calculation is also very similar (only differing by a minus sign) for Rij ; i < j = 1, 2, 3 resulting in the
components
j
i
Rjij = −k 2 e−2ky ; Riij = k 2 e−2ky ; i < j = 1, 2, 3
Now let’s try the calculation for R14

R14 = ∂1 Γ4 − ∂4 Γ1 + Γ1 Γ4 − Γ4 Γ1
    
0 0 0 0 0 0 0 0 0 0 −k 0 0 0 0
0
 0 0 0 −k 
 
0 0 0 0 −k   0 −k
 0 0 0

= −∂y 
0 0 0 0 0  + 0
 0 0 0 0   0
 0 −k 0 0

0 0 0 0 0  0 0 0 0 0  0  0 0 −k 0
0 ke−2ky 0 0 0 0 ke−2ky 0 0 0 0 0 0 0 0
  
−k 0 0 0 0 0 0 0 0 0
 0
 −k 0 0 0 
0 0 0 0 −k 

−
 0 0 −k 0 0 
0 0 0 0 0 
 0 0 0 −k 0 0 0 0 0 0 
0 0 0 0 0 0 ke−2ky 0 0 0
 
0 0 0 0 0
0
 0 0 0 −k 2 
= 0
 0 0 0 0 

0 0 0 0 0 
0 k 2 e−2ky 0 0 0
Repeating the previous calculation for the other components we find
0
R404 = −k 2 ; R004
4
= −k 2 e−2ky
i
R4i4 = −k 2 ; Rii4
4
= k 2 e−2ky ; i = 1, 2, 3
All other terms not related to one of these by a symmetry vanish.
Now we calculate the Ricci tensor by the definition, but this is aided by the observation that Rλµλν is
equal to zero unless µ = ν. Therefore all of the off-diagonal terms are zero and our work is almost done.
λ 1 2 3 4
R00 = R0λ0 = R010 + R020 + R030 + R040

Rewrite this in terms of the components that we have calculated:

1
R00 = −R001 2
− R002 3
− R003 4
− R004 = k 2 e−2ky + k 2 e−2ky + k 2 e−2ky + k 2 e−2ky = 4k 2 e−2ky

Proceeding in a similar manner we find

λ
Rii = Riλi = −4k 2 e−2ky ; i = 1, 2, 3
λ
R44 = R4λ4 = −4k 2
Finally calculate the Ricci scalar

R = Rµ µ = g µν Rνµ

12
where
−e2ky
 
0 0 0 0
 0 e2ky 0 0 0
µν
 
g  0
= 0 e2ky 0 0
 0 0 0 e2ky 0
0 0 0 0 1
So we get for the Ricci scalar

R = −e2ky (4k 2 e−2ky ) + e2ky (−4k 2 e−2ky − 4k 2 e−2ky − 4k 2 e−2ky ) + 1(−4k 2 ) = −20k 2

(c) A metric defines a maximally symmetric space if it satisfies Carroll eq. 3.191, which for a five-dimensional
manifold reads:
R
Rρσµν = (gρµ gσν − gρν gσµ )
20
Luckily the metric is diagonal and so we only need to worry about the terms Rµνµν and the symmetry of
both expressions in the second pair of indices will take care of the rest. For i, j = 1, 2, 3; i 6= j we have
k
Rijij = gik Rjij = e−2ky (−k 2 e−2ky ) = −k 2 e−4ky

Compare this with

R −20k 2 −2ky −2ky
(gii gjj − gij gji ) = e e = −k 2 e−4ky
20 20

Now consider i = 1, 2, 3
k
R0i0i = g0k Ri0i 0
= g00 Ri0i = −e−2ky (−k 2 e−2ky ) = k 2 e−4ky

Compare with
R −20k 2 −2ky
(gii g00 − gi0 g0i ) = e (−e−2ky ) = k 2 e−4ky
20 20

Now we just take care of the terms involving 4’s, first considering i = 1, 2, 3
k
Ri4i4 = gik R4i4 i
= gii R4i4 = (e−2ky )(−k 2 ) = −k 2 e−2ky

Compare with
R −20k 2 −2ky
(gii g44 − gi4 g4i ) = e (1) = −k 2 e−2ky
20 20

...Just one last case to check

k
R0404 = g0k R404 0
= g00 R404 = −e−2ky (−k 2 ) = k 2 e−2ky

Compare with
R −20k 2
(g00 g44 − g04 g40 ) = (−e−2ky )(1) = k 2 e−2ky
20 20
All cases check out, so this metric does indeed define a maximally symmetric space.

13