General Relativity Fall 2019

Homework 5 solution

Exercise 1: Parallel transport on a sphere

The 2-sphere of radius r has line element ds2 = r2 dθ2 + r2 sin2 θdϕ2 .

(i) Compute the Christoffel symbols of this metric.

The inverse metric is diagonal, with components g θθ = 1/r2 , g ϕϕ = 1/(r sin θ)2 , g θϕ = 0.
There are only 6 independent Christoffel symbols to compute:
1
Γθθθ = gθθ,θ = 0, (1)
2r2
1
Γθθϕ = gθθ,ϕ = 0, (2)
2r2
1
Γθϕϕ = (2gθϕ,ϕ − gϕϕ,θ ) = − sin θ cos θ, (3)
2r2
1
Γϕ
θθ = (2gθϕ,θ − gθθ,ϕ ) = 0, (4)
2r sin2 θ
2
1 cos θ
Γϕ
θϕ = 2 2 gϕϕ,θ = sin θ , (5)
2r sin θ
1
Γϕ
ϕϕ = gϕϕ,ϕ = 0. (6)
2r sin2 θ
2

(ii) Write the equations of parallel transport of a tangent vector W = W θ ∂θ + W ϕ ∂ϕ along a small circle with
constant θ = θ0 – the first thing you will need is to do is find a vector V tangent to this curve.

Suppose we parametrize the curve by some parameter λ. The tangent vector d/dλ = V = V θ ∂θ + V ϕ ∂ϕ has
dθ
components V θ = dλ and V ϕ = dϕ/dλ. Since the curve has constant θ, we find that V θ = 0. Keeping the
parametrization of the curve general for now, thus we have
dϕ
V = ∂ϕ . (7)
dλ
The equations of parallel transport are
dW ν dW ν dϕ
V µ ∇µ W ν = 0 = + Γνµσ V µ W σ = + Γνϕσ W σ . (8)
dλ dλ dλ
We see that we can get rid of λ and write the parallel transport equation in terms of ϕ directly:
dW ν
= −Γνϕσ W σ . (9)
dϕ
We thus have two coupled equations:
dW θ
= −Γθϕσ W σ = sin θ0 cos θ0 W ϕ , (10)
dϕ
dW ϕ cos θ0
= −Γϕ σ
ϕσ W = − W θ, (11)
dϕ sin θ0
where I used the expressions for the Christoffel symbols derived above, evaluated on the curve of constant θ = θ0 .

(iii) Solve these equations, starting with some initial conditions W θ |0 and W ϕ |0 at ϕ = 0. How does the vector
compare to itself after being parallel-transported once around a small circle?

We can combine these equations into the second-order linear ODE:

d2 W θ
+ cos2 θ0 W θ = 0, (12)
dϕ2
 2

and similarly for W ϕ . These are just harmonic oscillators with frequency ω0 = cos θ0 , whose solution is
1 dW θ
W θ (ϕ) = W θ |0 cos[(cos θ0 )ϕ] + sin[(cos θ0 )ϕ], (13)
cos θ0 dϕ 0
1 dW ϕ
W ϕ (ϕ) = W ϕ |0 cos[(cos θ0 )ϕ] + sin[(cos θ0 )ϕ]. (14)
cos θ0 dϕ 0
Now using the original equations, we can related the first derivative of W θ to W ϕ and vice-versa. Thus we obtain,
finally,
W θ (ϕ) = W θ |0 cos[(cos θ0 )ϕ] + sin θ0 W ϕ |0 sin[(cos θ0 )ϕ], (15)
1
W ϕ (ϕ) = W ϕ |0 cos[(cos θ0 )ϕ] − W θ |0 sin[(cos θ0 )ϕ]. (16)
sin θ0
Let us moreover define W ϕ̂ = sin θ W ϕ . This is the component of W along eϕ̂ = sin1 θ ∂ϕ , which has norm r, like ∂θ .
Thus, after being transported once, we have
 θ    θ 
W (2π) cos(2π cos θ0 ) sin(2π cos θ0 ) W (0)
= (17)
W ϕ̂ (2π) − sin(2π cos θ0 ) cos(2π cos θ0 ) W ϕ̂ (0)
We see that the vector W is rotated by an angle 2π cos θ0 after being parallel-transported once around the small
circle.

Let us consider a small circle near the pole, so θ0  1. We then have cos θ0 ≈ 1 − θ02 /2, thus
cos(2π cos θ0 ) − 1 ≈ cos(πθ02 ) − 1 = O(θ04 ), (18)
sin(2π cos θ0 ) ≈ − sin(πθ02 ) ≈ −πθ02 . (19)
We then get
W θ (2π) − W θ (0) ≈ −πθ02 × W ϕ̂ |0 , (20)
W ϕ̂ (2π) − W ϕ̂ (0) ≈ πθ02 × W θ |0 . (21)
Therefore the difference if proportional to the area enclosed by the small circle.

Exercise 2: Killing vector fields and conservation laws

(i) Suppose that, in some coordinate system {xµ }, the metric components do not depend on a specific coordinate
σ∗
x , i.e. ∂σ∗ gµν = 0 for any µ, ν. Show that this implies that Pσ∗ (with the index downstairs!) is constant along
geodesics with tangent vector P α .

Suppose P = d/d/λ, i.e. that λ is the parameter along the geodesic (proper time or affine parameter for null
geodesics). Since P µ ∇µ P ν = 0, and the covarariant derivative is metric-compatible, we also automatically have
P µ ∇µ Pν = 0. Applying this to ν = σ∗, we have

dPσ∗
0 = P µ ∇µ Pσ∗ = − Γδµσ∗ P µ Pδ . (22)
dλ
Let’s compute the relevant Christoffel symbol:
1 δρ
Γδµσ∗ = g (gρµ,σ∗ + gσ∗ρ,µ − gσ∗µ,ρ ) = g δρ gσ∗[ρ,µ] , (23)
2
since gρµ,σ∗ = 0. Thus we have

dPσ∗
= g δρ gσ∗[ρ,µ] P µ Pδ = gσ∗[ρ,µ] P µ P ρ = 0, (24)
dλ
as the last term is the contraction of a pair of symmetric indices with a pair of antisymmetric indices. We thus
showed that Pσ∗ is constant along the geodesic. Important: P µ ∇µ Pσ∗ really means (∇P )µσ∗µ , and is not equal to
dPσ∗ /dλ, which is the quantity that must be zero for Pσ∗ to be constant along the geodesic.
 3

(ii) A Killing vector field K α is a vector field that satisfies Killings’ equation, ∇(α Kβ) = 0. Show that, if K α is a
Killing vector field, then Kα P α is constant along geodesics with tangent vector P α .

For an arbitrary vector field K α , P α Kα is a scalar field, thus

d
(Kα P α ) = P µ ∇µ (Kα P α ) = P µ (∇µ Kα )P α + P µ (∇µ P α )Kα = (∇µ Kα )P µ P α = (∇(µ Kα) )P µ P α (25)
dλ
since P µ ∇µ P α = 0, and the tesnor product P µ P α is symmetric. Thus, if K α is a Killing vector field, this vanishes,
thus K α Pα is constant along geodesics.

(iii) Show that, given a conserved stress-energy tensor T αβ , i.e. such that ∇α T αβ = 0, and a Killing vector field
K , the current J α ≡ Kβ T βα is conserved (i.e. divergence-free).
α

∇α J α = ∇α (Kβ T βα ) = (∇α Kβ )T βα + Kβ ∇α T βα = (∇(α Kβ) )T βα , (26)

since T αβ is divergence-free, and symmetric (so contractions with it only pick the symmetric part). Thus, if K α is a
Killing vector field, this vanishes.

(iv) In a coordinate system {xµ }, define K ≡ ∂σ∗ , for some specific coordinate xσ∗ . Show that K α satisfies Killing’s
equation if and only if ∂σ∗ gµν = 0 for all µ, ν.

Let us write explicitly the symmetric part of the gradient of Kα :

∇(α Kβ) = ∂(α Kβ) − Γδαβ Kδ , (27)

where the second term is automatically symmetric. Now since K = ∂σ∗ , we have

K µ = δσ∗
µ
, Kβ = gβµ K µ = gβσ∗ . (28)

So we found
1
∇(α Kβ) = ∂(α gβ)σ∗ − gδσ∗ Γδαβ = ∂(α gβ)σ∗ − (gασ∗,β + gβσ∗,α − gαβ,σ∗ ), (29)
2
where we used the expression for the Christoffel symbol and simplified the contraction with the inverse metric. The
first two terms in the parenthesis are exactly ∂(α gβ)σ∗ . Thus we have

1
∇(α Kβ) = gαβ,σ∗ . (30)
2
Hence we see that K = ∂σ∗ is a Killing vector field (i.e. the left-hand-side vanishes for any pair α, β, by definition) if
and only if the metric components do not depend on σ∗.

The concept of Killing vectors is important and you should remember it!

Exercise 3: Ideal fluid

Preliminaries: A projector is a tensor P αβ which satisfies P αβ P βγ = P αγ .
A projector parallel to a vector X α is a projector (P|| )αβ satisfying (P|| )αβ Y β ∝ X α for any vector Y α .
A projector perpendicular to a vector X α is a projector (P⊥ )αβ satisfying Xα P αβ Y β = 0, for any vector Y α .

(i) Suppose an ideal fluid has 4-velocity uα , normalized as usual such that uα uα = −1. Construct the pro-
jectors parallel to and perpendicular to the fluid’s 4-velocity. Rewrite the stress-energy tensor of the ideal fluid
T αβ = ρrf uα uβ + Prf (g αβ + uα uβ ) in terms of these projectors.

The desired projectors are

(P|| )αβ = −uα uβ , (P⊥ )αβ = δβα + uα uβ . (31)

 4

Let us first prove that they are indeed projectors:

(P|| )αβ (P|| )β γ = uα uβ uβ uγ = −uα uγ = (P|| )αγ , (32)
since uµ uµ = −1.
(P⊥ )αβ (P⊥ )β γ = (δβα + uα uβ )(δγβ + uβ uγ ) = δγα + 2uα uγ + uα uβ uβ uγ = δγα + uα uγ = (P⊥ )αγ . (33)
It should be clear that (P|| )αβ Y β = −(Y β uβ )uα is along uα , as it should. Moreover,
uα (P⊥ )αβ Y β = uα (δβα + uα uβ )Y β = uβ Y β (1 + uα uα ) = 0. (34)
Recalling that gβα = δβα , we have
T αβ = −ρrf (P|| )αβ + Prf (P⊥ )αβ . (35)
(ii) What physical quantity does the scalar field Q̇ ≡ −uα ∇β T αβ represent? Show that Q̇ = ∇α (ρrf uα ) + Prf ∇α uα .

In the rest-frame of the fluid, the components of the 4-velocity are, by definition uµ = (1, 0, 0, 0). Thus, taking
inertial coordinates in that rest-frame, we have uµ = (−1, 0, 0, 0) in the fluid’s rest frame. Thus, Q̇ = +∇β T 0β in the
fluid’s rest-frame. This is the zero-th component of the 4-force density (see lecture 10). A 4-force is a rate of change
of 4-momentum. The zero-th component of 4-momentum is energy. Thus, Q̇ is the rate of heating per unit
volume, in the fluid’s rest-frame – heating is just a rate of change of energy! In general, whenever we multiply a
quantity by −uα , it represents the value of the zero-th component of this quantity in the rest-frame of the “observer”
whose 4-velocity is uα .

Let us now compute this heating rate:

Q̇ = −uα ∇β T βα = −∇β (uα T βα ) + T βα ∇β uα . (36)
βα α
The first term in parenthesis only picks the component of T parallel to u . In addition,
1 1
uα ∇β uα = ∇β (uα uα ) = ∇β (−1) = 0. (37)
2 2
Thus the second term only picks the components of T αβ proportional to g αβ . Hence we find the desired expression.

(iii) Suppose that the ideal fluid is made of particles with conserved number. This means that there exists a number
4-current N α , such that N 0 is the number density and N ~ is the number flux, which satisfies ∇α N α = 0. Denoting by
nrf the particle number density in the fluid’s rest-frame, derive the expression for the particle number current N α in
an arbitrary frame.

By definition, in the fluid’s rest frame, N 0 = nrf and N i = 0, since this frame was defined as being isotropic, hence
there is no preferred direction for the number flux. Now in the fluid’s rest frame we have u0 = 1 and ui = 0. So in
that frame, N µ = nrf uµ . This is a vector equality, which should hold in any frame – note that nrf is a scalar field,
defined as the fluid number density in the fluid’s rest frame.

(iv) We define rf ≡ ρrf /nrf . What physical quantity does this represent? Combining particle number conservation
with question (ii), derive an equation involving Q̇ and rf , and explain why it represents the first law of thermodynamics.

rf is the ratio of energy density in the fluid’s rest frame to the number density in the fluid’s rest frame. It is thus
the mean energy per particle in the fluid’s rest frame.

Nα
   
1
Q̇ = ∇α (rf nrf uα ) + Prf ∇α nrf uα = ∇α (rf N α ) + Prf ∇α = nrf uα (∇α rf + Prf ∇α (1/nrf )) , (38)
nrf nrf
where we used ∇α N α = 0 and substituted N α = nrf uα . Let us now consider a small volume V enclosing a fixed
number of particles Npart ≡ nrf V , where we work in the fluid’s rest frame. We may thus replace 1/nrf = V /Npart .
We also define the total energy of these particles by U = Npart rf . Multiplying the above equation by V , we get
dU dV
V Q̇ = uα ∇α U + Prf uα ∇α V =
+ Prf (39)
dτ dτ
The left-hand side is the rate of heating of the volume V . This is exactly the first law of thermodynamics: δQ =
dU + P dV , written in differential form.