General Relativity Fall 2019

Homework 2 solutions

Exercise 1: Christoffel symbols

(i) In class we derived an expression for the Christoffel symbols in terms of the metric components and their
derivatives,
1 ασ
Γα
βγ = g (gσβ,γ + gσγ,β − gβγ,σ ) . (1)
2
0
We also showed that, starting from a ICS {xµ }, and transforming to a general coordinate system {xµ } (note that
here I chose to have primes for the ICS, and no primes for the general coordinates), we arrive at the geodesic equation
with
0
∂xα ∂ 2 xα
Γα
βγ = . (2)
∂xα0 ∂xβ ∂xγ
0
Show that this is indeed identical to the Christoffel symbol – don’t forget that {xµ } is defined to be an ICS!
0
By definition, gµ0 ν 0 = ηµ0 ν 0 since {xµ } is an ICS. Changing coordinates, we have
0 0
∂xµ ∂xν
gσβ = ηµ0 ν 0 . (3)
∂xσ ∂xβ
Taking the derivative with respect to xγ we get
0 0 0 0
!
∂ 2 xµ ∂xν ∂xµ ∂ 2 xν
gσβ,γ = σ γ β
+ ηµ0 ν 0 . (4)
∂x ∂x ∂x ∂xσ ∂xβ ∂xγ

Thus we obtain
gσβ,γ + gσγ,β − gβγ,σ
0 0 0 0 0 0 0 0 0 0 0 0
!
∂ 2 xµ ∂xν ∂xµ ∂ 2 xν ∂ 2 xµ ∂xν ∂xµ ∂ 2 xν ∂ 2 xµ ∂xν ∂xµ ∂ 2 xν
= + + + − − ηµ0 ν 0 . (5)
∂xσ ∂xγ ∂xβ ∂xσ ∂xβ ∂xγ ∂xσ ∂xβ ∂xγ ∂xσ ∂xβ ∂xγ ∂xβ ∂xσ ∂xγ ∂xβ ∂xσ ∂xγ

The third and fifth terms cancel out. Since ηµ0 ν 0 is symmetric, the first and last terms also cancel out. Moreover, the
second and fourth terms are identical. So we are left with
0 0
∂xµ ∂ 2 xν
gσβ,γ + gσγ,β − gβγ,σ = 2 ηµ0 ν 0 . (6)
∂xσ ∂xβ ∂xγ
We also know how the inverse metric transforms:
∂xα ∂xσ α0 σ0
g ασ = η , (7)
∂xα0 ∂xσ0
0
where we recall that {xµ } is an ICS. Thus we get
0 0
1 ασ ∂xα ∂xσ α0 σ0 ∂xµ ∂ 2 xν
g (gσβ,γ + gσγ,β − gβγ,σ ) = η ηµ0 ν 0 . (8)
2 ∂xα0 ∂xσ0 ∂xσ ∂xβ ∂xγ
I highlighted the summed-over index σ. Upon summation, we get
0 0 0 0
1 ασ ∂xα α0 σ0 µ0 ∂ 2 xν ∂xα α0 µ0 ∂ 2 xν ∂xα ∂ 2 xν α0 ∂xα ∂ 2 xα
g (gσβ,γ + gσγ,β − gβγ,σ ) = 0 η δ σ 0 η µ0ν0 = 0 η ηµ0ν0 = 0 δ ν 0 = ,
2 ∂xα ∂xβ ∂xγ ∂xα ∂xβ ∂xγ ∂xα ∂xβ ∂xγ ∂xα0 ∂xβ ∂xγ
0
which proves the equality. Note: it was key to use the fact that {xµ } is an ICS.
 2

(ii) Using either expression, compute the transformation law of Christoffel symbols between general coordinate
systems.
00 0
I will use the coordinate definition. Suppose now that {xµ } is an ICS, and that {xµ } and {xµ } are general
coordinates. We use our very farvorite rule (the chain rule) to obtain
" # " #
α0 2 α00 α0 α γ α00 α0 α 2 γ α00 γ β 2 α00
0 ∂x ∂ x ∂x ∂x ∂ ∂x ∂x ∂x ∂x ∂ x ∂x ∂x ∂x ∂ x
Γα
β0 γ0 = = × = × + .
∂xα00 ∂xβ 0 ∂xγ 0 ∂xα ∂xα00 ∂xβ 0 ∂xγ 0 ∂xγ ∂xα ∂xα00 ∂xγ 0 ∂xβ 0 ∂xγ ∂xγ 0 ∂xβ 0 ∂xγ ∂xβ
00
The second term can be rewritten in terms of Γα
βγ , and the first term can be simplified when summing over α :

0 0
0 ∂xα ∂ 2 xα ∂xα ∂xβ ∂xγ α
Γα
β0 γ0 = + Γ . (9)
0
∂xα ∂xβ ∂xγ 0
∂xα ∂xβ 0 ∂xγ 0 βγ
We will see later on that this is NOT the transformation rule for a tensor (it would be if the first piece was not there),
i.e. the Christoffel symbol is NOT a tensor.

(iii) Show that the following quantity transforms as a vector under general coordinate transformations:
d2 xµ ρ
µ dx dx
σ
aµ ≡ + Γ ρσ . (10)
dτ 2 dτ dτ
These are the components of the 4-acceleration vector in a general coordinate system.
0 0
We already know that dxµ /dτ = (∂xµ /∂xµ ) dxµ /dτ – this is just the chain rule. The next step is to compute how
2 µ0
d x /dτ 2 transforms, (we basically already did this calculation in lecture 3):
0 0
! 0 0
! 0 0
d2 xµ d ∂xµ dxµ ∂xµ d2 xµ dxµ d ∂xµ ∂xµ d2 xµ dxµ dxν ∂ 2 xµ
= = + = + . (11)
dτ 2 dτ ∂xµ dτ ∂xµ dτ 2 dτ dτ ∂xµ ∂xµ dτ 2 dτ dτ ∂xµ ∂xν

We see that this does not transform as a vector on its own. Combining with what we foud earlier, we now get
0
!
ρ0 σ0 0 0 0 0 0 0
µ0 d2 xµ µ0 dx dx ∂xµ d2 xµ dxµ dxν ∂ 2 xµ ∂xµ ∂ 2 xµ ∂xµ ∂xρ ∂xσ µ ∂xρ ∂xσ dxλ dxδ
a = +Γ ρ0 σ 0 = + + + Γ .
dτ 2 dτ dτ ∂xµ dτ 2 dτ dτ ∂xµ ∂xν ∂xµ ∂xρ0 ∂xσ0 ∂xµ ∂xρ0 ∂xσ0 ρσ ∂xλ ∂xδ dτ dτ

This looks pretty awful, but can be simplified a lot. Combining the first term and the term multiplying Γµρσ just gives
0
∂xµ
∂xµ aµ . So we wan to show that the other two terms actually cancel out. Upon renaming of dummy indices, we want
to show that
0 0 0 0
∂ 2 xµ ∂xµ ∂ 2 xλ ∂xρ ∂xσ
+ = 0. (12)
∂xµ ∂xν ∂xλ ∂xρ0 ∂xσ0 ∂xµ ∂xν
Using the chain rule, we rewrite
0 0 0 0 0 0
" 0
! 0
#
∂xµ ∂ 2 xλ ∂xρ ∂xσ ∂xµ ∂xσ ∂ ∂xλ ∂xµ ∂xσ ∂xλ ∂xλ ∂ 2 xσ
 
∂
0 0 = = − . (13)
λ ρ σ
∂x ∂x ∂x ∂x ∂xµ ν ∂xλ ∂xν ∂xµ ∂xσ0 ∂xλ ∂xµ ∂xν ∂xσ0 ∂xσ0 ∂xµ ∂xν

Note that we had to be careful: while partial derivatives with respect to unprimed coordinates commute, mixed partial
derivatives do not:
   
∂ ∂f ∂ ∂f
6= . (14)
∂xν 0 ∂xµ ∂xµ ∂xν 0

This is because the variables that are kept constant are different in one set of coordinates and the other. Let us
compute the difference explicitly:
∂xν ∂ 2 f
     ν   ν
∂ ∂f ∂ ∂f ∂ ∂x ∂f ∂ ∂x ∂f
− = − =− µ ×
∂xν 0 ∂xµ ∂xµ ∂xν 0 ∂xν 0 ∂xν ∂xµ ∂xµ ∂xν 0 ∂xν ∂x ∂xν 0 ∂xν
0
∂xµ ∂ 2 xν ∂f
=− . (15)
∂xµ ∂xν 0 ∂xµ0 ∂xν
 3

Now back to our calculation, the term in parentheses is just δνλ , which is a constant, and the last term simplifies to
0 0 0 0 0
∂xµ ∂ 2 xλ ∂xρ ∂xσ 2 σ
µ0 ∂ x ∂ 2 xµ
0 0 = −δσ 0 = . (16)
∂xλ ∂xρ ∂xσ ∂xµ ∂xν ∂xµ ∂xν ∂xµ ∂xν
So we have indeed proved Eq. (12), thus found that
0
µ0 ∂xµ µ
a = a , (17)
∂xµ
i.e. the 4-acceleration (including the Christoffel symbol part) transforms as a vector under a change of coordinates.

Exercise 2: geodesics in different metrics

Consider the following two line elements:

ds2 = −dt2 + dr2 + r2 dθ2 + r2 sin2 θdϕ2 , (18)
ds2 = a(t)2 −dt2 + dx2 + dy 2 + dz 2 .
 
(19)
For each line element:

(i) Derive the geodesic equation by extremizing

dxµ dxν
Z
I≡ dτ gµν . (20)
dτ dτ

Let us first consider the first line element:

Z   Z
2 2 2 2 2 2 2
I = dτ −ṫ + ṙ + r θ̇ + r sin θϕ̇ ≡ dτ L(t, r, θ, ϕ, ṫ, ṙ, θ̇, ϕ̇), (21)

where an overdot means d/dτ . We use the Euler-Lagrange equations to find the 4 equations of motion:
d
(−2ṫ) = 0, (22)
dτ
d  
(2ṙ) = 2r θ̇2 + sin2 θϕ̇2 , (23)
dτ
d
(2r2 θ̇) = 2r2 sin θ cos θϕ̇2 , (24)
dτ
d
(2r2 sin2 θϕ̇) = 0. (25)
dτ
These equations simplify to

ẗ = 0,
 
r̈ − r θ̇2 + sin2 θϕ̇2 = 0,
ṙ
θ̈ + 2 θ̇ − sin θ cos θϕ̇2 = 0,
r
ṙϕ̇ cos θ
ϕ̈ + 2 +2 θ̇ϕ̇ = 0.
r sin θ

Let’s do the same thing for the other metric. Denote the spatial coordinates {x1 , x2 , x3 } ≡ {x, y, z}, so the metric is
ds2 = a(t)2 [−dt2 + δij dxi dxj ]. The Lagrangian is then

L(t, xi , ṫ, ẋi ) = a(t)2 −ṫ2 + δij ẋi ẋj

 
(26)
Euler-Lagrange equations then give
d da
(−2a(t)2 ṫ) = 2a(t) −ṫ2 + δij ẋi ẋj ,


(27)
dτ dt
d 2 i
(2a(t) ẋ ) = 0. (28)
dτ
 4

This implies

ẗ + H(t)ṫ2 + H(t)δij ẋi ẋj = 0,

ẍi + 2H(t)ṫẋi = 0,

where
1 da
H(t) ≡ . (29)
a(t) dt

(ii) From this equation, and the general expression for the geodesic equation, infer the Christoffel symbols (i.e. do not
use the explicit expression of the Christoffel symbols in this question).

We can just read off the Christoffel symbols from the geodesic equations. The only non-zero symbols are, for the
first metric:

Γrθθ = −r, Γrϕϕ = −r sin2 θ,

1
Γθrθ = , Γθϕϕ = − sin θ cos θ,
r
1 cos θ
Γrϕ = , Γϕ
ϕ
θϕ = .
r sin θ

For the second metric, we find the following only non-zero Christoffel symbols:

Γttt = H(t),
Γtij = H(t)δij ,
Γitj = H(t)δji .

(iii) Now compute the Christoffel symbols explicitly from their expression in terms of the metric components, and
check that you got the right answer in (ii).

There are a priori 4 × 6 = 24 different Chirstoffel symbols (remember that they are symmetric in the lower two
indices). So before starting to compute them all, let’s first try to eliminate as many as possible.
Both metrics are diagonal, so we get
1 αα
Γα
βγ = g (gβα,γ + gγα,β − gβγ,α ) [NOT summed over α] (30)
2
Since the metrics are diagonal, it means that at least two indices must be the same for non-zero Christoffel
symbols (otherwise all the terms in the parenthesis involve metric coefficients with unequal indices). Let us now look
at each metric.

The coefficients of the first line element only explicitly depend on r and θ. This means that non-zero Christoffel
symbols have at least one r index or one θ index (so there can be a derivative with repsect to either r or θ). With
the result from above, it means that the only possibly non-zero coefficients are a priori
Γrtt , Γrrr , Γrθθ , Γrϕϕ , Γtrt , Γθrθ , Γϕ r r r
rϕ , Γrt , Γrθ , Γrϕ (31)
Γθtt , Γθrr , Γθθθ , Γθϕϕ , Γtθt , Γrθr , Γϕ θ θ
θϕ , Γθt , Γθϕ (32)
These are still 19 coefficients to compute explicitly. I will not write the calculations, as it merely amounts to applying
the formula, but you can see that it is a lot more painful than extracting them from the geodesic equation!

For the second line element, the metric coefficients only depend on t. This means that non-zero Chirstoffel symbol
must include at least one t. With the requirement that two indices are the same, we thus find that the non-zero
coefficients are Γttt , Γtii , Γiti [the i are NOT repeated here], which is indeed the ones we had found. Computing them
gives the same answer.
 5

Exercise 3: Uniformly accelerated motion

(i) Starting from gµν uµ uν = −1, show that gµν uµ aν = 0, where aµ is the 4-acceleration defined in the first exercise.

In a ICS,

dxµ dxν
gµν uµ uν = ηµν , (33)
dτ dτ
thus, in an ICS,

d dxµ d2 xν
(gµν uµ uν ) = 2ηµν . (34)
dτ dτ dτ 2
But, in an ICS, we also have aν = d2 xν /dτ 2 , hence

d
(gµν uµ uν ) = 2ηµν uµ aν = 2gµν uµ aν . (35)
dτ
But the right-hand-side is now coordinate independent, as you can check explicitly. On the other hand, the left-hand-
side is d/dτ (−1) = 0. Thus, regardless of the coordinates in which this expression is computed, gµν uµ aν = 0.

(ii) For the remainder of this problem, suppose that {xµ } is a globally inertial coordinate system. Consider a
particle moving along the 1-direction (i.e. the x-direction), starting at rest from the origin at t = 0. Suppose that
in the particle’s instantaneous rest-frame, the components of its 4-acceleration are constant. Assuming the particle
carries with it a cartesian coordinate system, whose axes are aligned with the spatial axes of the glocal ICS, write
explicitly the components of the 4-acceleration in the particle’s instantaneous rest-frame. This is a long question, with
a very short answer!

Assuming that the particle uses a ICS in its rest-frame,

0
the metric in its reste-frame 0is the
0
Minkowski metric. The
components of the 4-velocity in its rest-frame are {uµ } = (1, 0, 0, 0). Imposing ηµ0 ν 0 aµ uν = 0 in the particle’s rest-
0
frame, this implies a0 = 0. If we further assume that motion is along the 1-direction, we thus get {aµ } = (0, 1, 0, 0)
in the particle’s instantaneous rest frame, where a is constant.

(iii) In the globally inertial frame of reference, in which the particle is initially at rest at the origin, describe the
particle’s motion, i.e. compute its 3-velocity, position, and proper time as a function of coordinate time t. Discuss
how these results match with the non-relativistic limit.

In the global ICS, the particle has 3-velocity ~v = (v, 0, 0), and 4-velocity
p
{uµ } = γ(1, v, 0, 0), γ ≡ 1/ 1 − v 2 . (36)

Doing a Lorentz transformation from the instantaneous rest-frame, to the global ICS, which has velocity −~v with
respect to the particle’s rest-frame, we get the following components of the 4-acceleration in the global ICS:
0
a0 = γ(a0 + va) = γva, (37)
1 00
a = γ(a + va ) = γa, (38)
a2 = a3 = 0. (39)

Now the 4-acceleration is just aµ = duµ /dτ , since we are in an ICS (otherwise it contains another piece, see first
exercise). Thus we have the following equations:

dγ d(γv)
= γva, = γa. (40)
dτ dτ
Note that these two equations are redundant (hence self-consistent) because, using γ 2 v 2 = γ 2 − 1, we have

d(γv) 1 d(γv)2 1 dγ 2 1 dγ
= = = . (41)
dτ 2γv dτ 2γv dτ v dτ
 6

Differentiating the first equation with respect to τ again, we find

d2 γ d(γv)
2
=a = a2 γ. (42)
dτ dτ
Thus we find

γ(τ ) = γ(0) cosh(aτ ) + γ̇(0) sinh(aτ ). (43)

Now, since γ̇ = aγv and the particle is initially at rest, we find that γ(0) = 1, γ̇(0) = 0, thus

γ(τ ) = cosh(aτ ). (44)

Now, remember that γ = dt/dτ is the first component of the 4-velocity.

dt 1 1
= cosh(aτ ) ⇒t= sinh(aτ ) ⇒τ = sinh−1 (at) . (45)
dτ a a

Now we want to express the velocity as a function of coordinate time:

p p
v = 1 − 1/γ 2 = γ 2 − 1/γ = tanh(aτ ). (46)

Using hyperbolic trigonometry, we get

at
v(t) = p . (47)
1 + (at)2

Integrating this, we find

1 p 
x= 1 + (at)2 − 1 . (48)
a

We see that the velocity asymptotes to 1 (in units of the speed of light), and becomes relativistic on a timescale
t ∼ 1/a. As long as t  1/a, we have τ ≈ t, v ≈ at and x ≈ 21 at2 , as we expect in the Newtonian limit.