Massachusetts Institute of Technology

Department of Physics
8.962 Spring 2006

Problem Set 1

1. (a) [5 pts] Show that the sum of any two orthogonal spacelike vectors is spacelike.
(b) [5 pts] Show that a timelike vector and a null vector cannot be orthogonal.

2. In some reference frame, the vector fields U� and D � have the components
. √
U α = (1 + t2 , t2 , 2t, 0)
. √
Dα = (x, 5tx, 2t, 0) .

The scalar ρ has the value

ρ = x2 + t2 − y 2 .
.
(The relationship “LHS = RHS” means “the object on the left-hand side is represented
by the object on the right-hand side in the specified reference frame.”)
� is suitable as a 4-velocity. Is D
(a) [3 pts] Show that U �?

� , for arbitrary
(b) [3 pts] Find the spatial velocity v of a particle whose 4-velocity is U
t. Describe the motion in the limits t = 0 and t → ∞.

(c) [3 pts] Find ∂β U α for all α, β. Show that Uα ∂β U α = 0. (There’s a clever way to
do this; do it the brute force way instead.)

(d) [3 pts] Find ∂α Dα .

(e) [3 pts] Find ∂β (U α Dβ ) for all α.

(f) [3 pts] Find Uα ∂β (U α Dβ ). Why is the answer so similar to that for (d)?

(g) [3 pts] Calculate ∂α ρ for all α. Calculate ∂ α ρ.

(h) [3 pts] Find �U� ρ and �D� ρ.

1
 � and the tensor
3. Consider a timelike unit 4-vector U

Pαβ = ηαβ + Uα Uβ .

Show that this tensor is a projection operator that projects an arbitrary vector V� into
� . In other words, show that the vector V�⊥ whose components are
one orthogonal to U

V⊥α = P α β V β

is

�
(a) [5 pts] orthogonal to U

(b) [5 pts] unaffected by further projections:

α
V⊥⊥ ≡ P α β V⊥β = V⊥α .

�:
(c) [5 pts] Show that Pαβ is the metric for the space of vectors orthogonal to U

Pαβ V⊥α W⊥β = V�⊥ · W

�⊥ .

(d) [5 pts] Show that for an arbitrary nonnull vector �q, the projection tensor is given
by
qα q β
Pαβ (q α ) = ηαβ − .
q γ qγ

Do we need a projection tensor for null vectors?

4. [15 pts] Let ΛB (v) be a Lorentz boost associated with 3-velocity v. Consider

Λ ≡ ΛB (v1 ) · ΛB (v2 ) · ΛB (−v1 ) · ΛB (−v2 )

where v1 · v2 = 0. Assume v1 � 1, v2 � 1.

Show that Λ is a rotation. What is the axis of rotation? What is the angle of rotation?

2
5. “Superluminal” motion
The quasar 3C 273 emits relativistic blobs of plasma from near the massive black hole
at its center. The blobs travel at speed v along a jet making an angle θ with respect
to the line of sight of the observer. Projected onto the sky, the blobs appear to travel
perpendicular to the line of sight with angular speed vapp /r where r is the distance to
3C 273 as and vapp is the apparent speed.

(a) [7 pts] Show that

v sin θ
vapp = .
1 − v cos θ

(b) [5 pts] For a given value of v, what value of θ maximizes vapp ? What is the corre­
sponding maximal value of vapp ? Can this be greater than the speed of light? If so, is
special relativity violated?

(c) [3 pts] For 3C 273, vapp � 10c. What is the largest possible value of θ (in degrees)?

6. GZK cutoff in the cosmic ray spectrum

(a) [8 pts] Calculate the threshold energy of a nucleon N for it to undergo the reaction
γ + N → N + π 0 , where γ represents a microwave background photon of energy kT
with T = 2.73 K. Assume the collision is head-on and take the nucleon and pion masses
to be 938 MeV and 135 MeV, respectively.

(b) [5 pts] Explain why one might expect to observe very few cosmic rays of energy
above ∼ 1011 GeV.

(c) [3 pts] This expectation is called the Griesen-Zatsepin-Kuzmin (GZK) cutoff. Mod­
ern observations show no sharp cutoff; there may even be evidence for an upturn in
cosmic ray flux at these energies. Can you suggest a mechanism by which the GZK
cutoff can be avoided?

3
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

4
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 2

�
1. Show that the number density of dust measured by an observer whose 4-velocity is U
� � �
is given by n = −N · U , where N is the matter current 4-vector.

2. Take the limit of the continuity equation for |v| � 1 to get ∂n/∂t + ∂(nv i )/∂xi = 0.

3. In an inertial frame O, calculate the components of the stress-energy tensors of the

following systems:
(a) A group of particles all moving with the same 3-velocity v = β�ex as seen in O.
Let the rest-mass density of these particles be ρ0 , as measured in their own rest frame.
Assume a sufficiently high density of particles to enable treating them as a continuum.
(b) A ring of N similar particles of rest mass m rotating counter-clockwise in the x − y
plane about the origin of O, at a radius a from this point, with an angular velocity
ω. The ring is a torus of circular cross-section δa � a, within which the particles are
uniformly distributed with a high enough density for the continuum approximation to
apply. Do not include the stress-energy of whatever forces keep them in orbit. Part of
this calculation will relate ρ0 of part (a) to N , a, δa, and ω.
(c) Two such rings of particles, one rotating clockwise and the other counter-clockwise,
at the same radius a. The particles do not collide or otherwise interact in any way.

4. Use the identity ∂ν T µν = 0 to prove the following results for a bounded system (i.e., a
system for which T µν = 0 beyond some bounded region of space):
�
(a) ∂t T 0α d3 x = 0. This expresses conservation of energy and momentum.
� �
(b) ∂t2 T 00 xi xj d3 x = 2 T ij d3 x. This result is a version of the virial theorem; it will
come in quite handy when we derive the quadrupole formula for gravitational radiation.
� � �
(c) ∂t2 T 00 (xi xi )2 d3 x = 4 T i i xj xj d3 x + 8 T ij xi xj d3 x. No pithy wisdom for this one.

1
 �=.
5. The vector potential A (A0 , A) generates the electromagnetic field tensor via

Fµν = ∂µ Aν − ∂ν Aµ .

(a) Show that the electric and magnetic fields in a specific Lorentz frame are given by

B = �×A
∂A
E = − − �A0 .
∂t
Here, � is taken to be the normal gradient operator in Euclidean space.
(b) Show that Maxwell’s equations hold if and only if

∂µ ∂ µ Aα − ∂ α ∂µ Aµ = −4πJ α .

(c) Show that a gauge transformation of the form

Anew
µ = Aµold + ∂µ φ

leaves the field tensor unchanged.

(d) Show that one can adjust the gauge so that

∂µ Aµ = 0 .

Show that Maxwell’s equations take on a particularly simple form with this gauge
choice. Use the operator � ≡ ∂µ ∂ µ to simplify your result.

2
6. An astronaut has
√ acceleration g in the x direction (in other words, the magnitude of his
¯ x,
4-acceleration, �a · �a, is g). This astronaut assigns coordinates (t, ¯ y,
¯ z)
¯ to spacetime
as follows:
First, the astronaut defines spatial coordinates to be (x̄, y,
¯ z̄), and sets the time coor­
dinate t̄ to be his own proper time.
Second, at t̄ = 0, the astronaut assigns (¯ x, y,
¯ z)
¯ to coincide with the Euclidean co­
ordinates (x, y, z) of the inertial reference frame that momentarily coincides with his
motion. (In other words, though the astronaut is not inertial — he is accelerating —
there is an inertial frame that, at t̄ = 0, is momentarily at rest with respect to him.
This is the frame used to assign (x̄, y, ¯ at t¯ = 0.) Observers who remain at fixed
¯ z)
values of the spatial coordinates (x̄, y,¯ z̄) are called coordinate-stationary observers
(CSOs). Note that proper time for these observers is not necessarily t̄! — we cannot
assume that the CSOs’ clocks remain synchronized with the clocks of the astronaut.
Assume that some function A converts between coordinate time t̄ and proper time at
the location of a CSO:
dt̄
A=
dτ
The function A is evaluated at a CSO’s location and thus can in principle depend on
all four coordinates t̄, x̄, ȳ, z̄.
Finally, the astronaut requires that the worldlines of CSOs must be orthogonal to
the hypersurfaces t̄ = constant, and that for each t̄ there exists an inertial frame,
momentarily at rest with respect to the astronaut, in which all events with t̄ = constant
are simultaneous.
It is easy to see that ȳ = y and z̄ = z; henceforth we drop this coordinates from the
problem.
(a) What is the 4-velocity of the astronaut, as a function of t̄, in the initial inertial
frame [the frame that uses coordinates (t, x, y, z)]? (Hint: by considering the conditions
on �u · �u, �u · �a, and �a · �a, you should be able to find simple forms for ut and ux .)
(b) Imagine that each coordinate-stationary observer carries a clock. What is the
4-velocity of each clock in the initial inertial frame?
(c) Explain why A(x̄, t̄) cannot depend on time. In other words, why can we put
A(x̄, t̄) = A(x̄)? (Hint: consider the coordinate system that a different CSO may set
up.)
¯ x̄) and t(t,
(d) Find an explicit solution for the coordinate transformation x(t, ¯ x̄).
(e) Show that the line element ds2 = d�x · d�x in the new coordinates takes the form

ds2 = −dt2 + dx2 = −(1 + gx̄)2 dt̄2 + dx̄2 .

3
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

4
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 3

1. Accelerated observer revisited:

(a) On a spacetime diagram, show the trajectory (t, x) exhibited by the uniformly
accelerated astronaut from Problem 6 of pset 2.
(b) Show that there is a region of spacetime which is causally disconnected from this
astronaut. In other words, show that there is a region of spacetime in which events can­
not effect the astronaut without violating the fundamental postulate that information
cannot propagate faster than the speed of light.
(c) Find the boundary between the region that is causally connected and causally
disconnected from the astronaut. Such a boundary is called a particle horizon; it shares
some features of the event horizon that separates the interior and exterior spacetimes
of black holes.
Throughout this problem, only consider t ≥ 0.

2. Perfect fluids:
In class, I listed one of the defining characteristics of a perfect fluid that it have no
viscosity — i.e., no force parallel to the interface between fluid elements. This implied
that the stress-energy tensor must be diagonal — any component T ij for i �= j would
violate this assumption. I then claimed (without too much justification) that the
stress-energy tensor could be written
.
T αβ = diag[ρ, P, P, P ]

in Cartesian coordinates (t, x, y, z).

Suppose that the form were instead
.
T αβ = diag[ρ, P (1 + �), P, P ) .
� �
Show that if one performs a rotation around the z axis by an angle φ that T α β has
off-diagonal components of order �P . Hence we must have � = 0 in order for the tensor
to be diagonal in all Cartesian coordinate systems.

1
3. “3+1” split of the electromagnetic field:
An observer with 4-velocity U � interacting with an electromagnetic field F measures
�
electric and magnetic fields EU� and B� � in their instantaneous local inertial reference
U
� ). These fields are 4-vectors with
frame (that is, in an orthonormal basis with �e0̂ = U
components
1
EUα� = F αβ Uβ , BUα� = − �αβγδ Uβ Fγδ .
2

(a) Show that E � � lie orthogonal to observer’s worldline. Thus, they are spatial
� � and B
U U
vectors according to the observer, living entirely in that observer’s hypersurface of
simultaneity. (Hint: recall the projection tensor defined in Pset 1.)
(b) Show that the field tensor can be reconstructed from the observer’s 4-velocity and
the electric and magnetic fields they measure via the following tensor equation (valid
for any basis):

F αβ = U α EUβ� − EUα� U β + �αβ γδ U γ BUδ� .

� �
The identity �αβρσ �µνρσ = 2 δ α ν δ β µ − δ α µ δ β ν may prove useful.
(c) The wedge product between two vectors is defined as

�∧B
A � =A
�⊗B
� −B
� ⊗A
�.

The Hodge dual of a (0, 2) tensor is defined as

1
∗
Cµν = �αβ µν Cαβ .
2
Show that the field tensor may be written
� �
� ∧E
F = aU �� + b∗ U
� ∧B
�� .
U U

What are the values of the real constants a and b?

4. Transformation of Christoffel symbols:

(a) Show that, under a coordinate transformation, the components of the Christoffel
symbol transform as follows:
� �
� ∂xα ∂xβ ∂xγ α ∂ 2 xα ∂xβ ∂xγ
Γα β � γ � = Γ βγ −
∂xα ∂xβ � ∂xγ � ∂xβ ∂xγ ∂xβ � ∂xγ �
Do this by considering the form of the Christoffel symbol in terms of derivatives of the
metric.
(b) Show that, using this rule, the components of the covariant derivative of a vector
transform as tensors should:
�
� ∂xα ∂xβ
�α� Aβ = α � β
�α Aβ .
∂x ∂x

2
5. Carroll: Chapter 3, Problem 2. In this problem, define the curl via

(curl V� )i = �i jk �j Vk .

6. Carroll: Chapter 3, Problem 3.

7. Prove the following connection identities:

(a) ∂λ gµν = Γµνλ + Γνµλ .
(b) gµκ ∂λ g κν = −g κν ∂λ gµκ .
(c) ∂λ g µν = −Γµ λκ g κν − Γν λκ g κµ
The next three parts rely on an identity I will prove on either Thursday March 2nd or
on Tuesday March 7th. The quantity g is the determinant of the metric gµν .
� �
(d) �ν Aµ ν = |g|−1/2 ∂ν |g|1/2 Aµ ν − Γλ νµ Aλ ν in a coordinate basis.
� �
(e) �ν F µν = |g|−1/2 ∂ν |g|1/2 F µν in a coordinate basis, if F µν is antisymmetric.
� �
(f) �S ≡ g µν �µ �ν S = |g|−1/2 ∂µ |g|1/2 g µν ∂ν S in a coordinate basis. (S is a scalar
function.)

3
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

4
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 4

1. Connection in Rindler spacetime

The spacetime for an accelerated observer that we derived on Pset 2,

ds2 = −(1 + gx̄)2 dt̄2 + dx̄2 + dȳ 2 + dz̄ 2 (1)

is known as “Rindler spacetime”. Compute all non-zero Christoffel symbols for this
spacetime. (Carroll problem 3.3 will help you quite a bit here.)

2. Relativistic Euler equation

� ⊗U
(a) Starting from the stress-energy tensor for a perfect fluid, T = ρU � + P h, where
� �
h = g + U ⊗ U , using local energy momentum conservation, � · T = 0, derive the
−1

relativistic Euler equation,

� = −h · �P .
(ρ + P )�U� U (2)

(Note: Because both T and h are symmetric tensors, there is no ambiguity in the dot
products that appear in this problem.)
(b) For a nonrelativistic fluid (ρ � P , v t � v i ) and a cartesian basis, show that this
equation reduces to the Euler equation,
∂vi 1
+ v k ∂ k v i = − ∂i P . (3)
∂t ρ
(i, k are spatial indices running from 1 to 3.) What extra terms are present if the
connection is non-zero (e.g., spherical coordinates)?
(c) Apply the relativistic Euler equation to Rindler spacetime for hydrostatic equilib­
rium. Hydrostatic equilibrium means that the fluid is at rest in the x̄ coordinates, i.e.
U x̄ = 0. Suppose that the equation of state (relation between pressure and density) is
P = wρ where w is a positive constant. Find the general solution ρ(x̄) with ρ(0) = ρ0 .
(d) Suppose now instead that w = w0 /(1 + gx̄) where w0 is a constant. Show that the
solution is ρ(x̄) = ρ0 exp(−¯
x/L. Find L, the density scale height, in terms of g and
w0 . Convert to “normal” units by inserting appropriate factors of c — L should be a
length.
(e) Compare your solution to the density profile of a nonrelativistic, plane-parallel,
isothermal atmosphere (for which P = ρkT /µ, where T is temperature and µ is the
mean molecular weight) in a constant gravitational field. [Use the nonrelativistic Euler
equation with gravity: add a term −∂i Φ = gi , where Φ is Newtonian gravitational
potential and gi is Newtonian gravitational acceleration, to the right hand side of Eq.
(3).] Why does hydrostatic equilibrium in Rindler spacetime — where there is no
gravity — give such similar results to hydrostatic equilibrium in a gravitational field?
1
3. Spherical hydrostatic equilibrium
As we shall derive later in the course, the line element for a spherically symmetric
static spacetime may be written
� �−1
2 2Φ(r) 2 2GM (r)
ds = −e dt + 1 − dr2 + r2 (dθ2 + sin2 θ dφ2 ) ,
r

where Φ(r) and M (r) are some given functions. In hydrostatic equilibrium, U i = 0 for
i ∈ [r, θ, φ]. Using the relativistic Euler equation, show that in hydrostatic equilibrium
p = p(r) with

∂p ∂Φ
= −(ρ + P ) .
∂r ∂r

4. Converting from non-affine to affine parameterization

Suppose v α = dxα /dλ∗ obeys the geodesic equation in the form

Dv α
= κ(λ∗ )v α .
dλ∗
Clearly λ∗ is not an affine parameter.
Show that uα = dxα /dλ obeys the geodesic equation in the form

Duα
=0
dλ
provided that
dλ
�� �
= exp κ(λ ) dλ .
∗ ∗
dλ∗

5. Conserved quantities with charge

A particle with electric charge e moves with 4-velocity uα in a spacetime with metric
gαβ in the presence of a vector potential Aµ . The equation describing this particle’s
motion can be written

uβ �β uα = eFαβ uβ ,

where

Fαβ = �α Aβ − �β Aα .

The spacetime admits a Killing vector field ξ α such that

Lξ�gαβ = 0 ,
Lξ�Aα = 0 .

Show that the quantity (uα + eAα )ξ α is constant along the worldline of the particle.

2
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

3
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 5

1. Space garbage
In a convenient coordinate system, the spacetime of the earth is approximately
2GM 2GM � 2
� � � � � ��
ds2 = − 1 − dt2 + 1 + dr + r2 dθ2 + sin2 θ dφ2
r r
2GM 2GM
� � � �
2
= − 1− dt + 1 + (dx2 + dy 2 + dz 2 ) ,
r r
where M is the earth’s mass. In the second version, we’ve remapped the spherical
coordinates to Cartesian coordinates in the usual way:

x = sin θ cos φ , y = sin θ sin φ , z = cos θ .

Note that the Cartesian form of the spacetime metric is conveniently written gαβ =
ηαβ + 2Φdiag(1, 1, 1, 1), with Φ ≡ GM/r. You may assume Φ � 1 throughout this
problem.
The space shuttle orbits the earth in a circular (ur = 0), equatorial (θ = π/2, uθ = 0)
orbit of radius R.
(a) Using the geodesic equation, show that an orbit which begins equatorial remains
equatorial: duθ /dt = 0 if uθ = 0 and θ = π/2 at t = 0. (Hint: begin by computing the
non-zero connection coefficients. Use the results of Carroll 3.3.)
We now require that the orbit must remain circular: dur /dτ = 0.
(b) By enforcing this condition with the geodesic equation, derive an expression for
the orbital frequency
dφ/dτ
Ω≡ .
dt/dτ
Does the result look familiar?
The next part is most conveniently described in Cartesian coordinates; you may de­
scribe the shuttle’s orbit as x = R cos Ωt, y = R sin Ωt.
An astronaut releases a bag of garbage into space, spatially displaced from the shuttle
by ξ i = xigarbage − xishuttle .
(c) Using the equation of geodesic deviation, work out differential equations for the
evolution of ξ x , ξ y , and ξ z as a function of time. You may neglect terms in (GM/r)2 ,
and you may treat all orbital velocities as non-relativistic. You will need the Cartesian
connection coefficients for this.
(d) Suppose the initial displacement is ξ x = ξ y = 0, ξ z = L, dξ i /dτ = 0. Has the
astronaut succeeded in getting rid of the garbage?
1
2. Light bending
The spacetime of the sun can be written using the same line element as that of Problem
1 (substituting the sun’s mass for M ). Consider a light ray which initially is moving
purely along the x axis. Suppose that it passes near the sun with impact parameter
(the distance between the axis passing through the sun’s center and the axis along
which the light initially moves) b, as shown in this figure:

y
pinitial

Δφ

b pfinal

Using the geodesic equation, compute the angle by which the sun bends the light ray,
Δφ = (py /px )final , expressing your answer using GM and b. You may assume that
GM/r � 1 and that the bending angle is very small (so that pxfinal � pxinitial to leading
order in Δφ).
Historical note: One can crudely calculate this effect in Newtonian theory by imagining
2
that a photon has an effective mass meff = hω/c ¯ and asking what is the transverse
momentum deflection imparted by Newtonian gravity. The answer one finds is exactly
half the angle computed using the geodesic equation applied to the weak-field space-
time. The factor of two arises because of “spatial curvature” — the factor (1 + 2Φ)
that multiplies the spatial sector of the spacetime metric. An expedition to observe an
eclipse in Africa led by the British astrophysicist Arthur S. Eddington confirmed the
relativistic prediction by carefully measuring stellar positions near the limb of the sun
and comparing to their positions when the sun is in a different part of the sky. These
results were instrumental in convincing the scientific world that general relativity was
a valid description of gravity, and were largely responsible for Einstein becoming a
household name.
It’s worth noting that the expedition’s data were not of particularly good quality; it’s
something of a miracle that they were able to get such good agreement with the rel­
ativistic prediction. Modern measurements using radar pulse propagation in the solar
system measure this bending angle with very high precision; the relativistic prediction
is typically confirmed with a fractional precision ∼ 10−5 .

2
3. Parallel transport on a sphere
� �
On the surface of a 2-sphere of radius a, ds2 = a2 dθ2 + sin2 θ dφ2 . Consider the
vector A0 = �eθ at θ = θ0 , φ = 0. The vector is parallel transported all the way
around the latitude circle θ = θ0 (i.e, over the range 0 ≤ φ ≤ 2π at θ = θ0 ). What
� ? What is its magnitude (A
is the resulting vector A �·A� )1/2 ? (Hint: derive differential
θ φ
equations for A and A as a function of φ.)

4. Curvature of a sphere
(a) Compute all the nonvanishing components of the Riemann tensor Rijkl [(i, j, k, l) ∈
(θ, φ)] for the surface of a 2-sphere.
� = Aθ �eθ + Aφ �eφ on the sphere
(b) Consider the parallel transport of a tangent vector A
around an infinitesimal parallelogram of sides �eθ dθ and �eφ dφ. Using the results of part
(a), show that to first order in dΩ ≡ sin θ dθ dφ, the length of A� is unchanged, but its
direction rotates through an angle equal to dΩ.
(c) Show that, if A � is parallel transported around the boundary of any simply con­
nected solid angle Ω, its direction rotates through an angle Ω. (“Simply connected”
is a topological term meaning that the boundary of the region could be shrunk to
a point; it tells us that there are no holes in the manifold or other pathologies. See
http://mathworld.wolfram.com/SimplyConnected.html for illustrations.) Compare with
the result of Problem 3.

5. Killing vectors and curvature

Prove that the relations

�ν �µ ξ α = Rα µνβ ξ β
�ξ α ≡ �µ �µ ξ α = −Rα β ξ β

are satisfied by any Killing vector ξ α . This result is not too difficult to derive using
the commutator [�α , �β ] = �α �β − �β �α and the identities

[�α , �β ] V µ = Rµ ναβ V ν
[�α , �β ] Vµ = −Rν µαβ Vν

(If you are reading Schutz, these identities are incorrectly given on p. 171 — the minus
sign is missing in the second identity.)

6. Riemann tensor for 1+1 static spacetimes

(a) Compute all the nonvanishing components of the Riemann tensor for the spacetime
with line element

ds2 = −e2φ(x) dt2 + e−2ψ(x) dx2 .

(b) For the case φ = ψ = 12 ln |g(x − x0 )| where g and x0 are constants, show that the
¯ x̄)
spacetime is flat and find a coordinate transformation to globally flat coordinates (t,
such that ds2 = −dt̄2 + dx̄2 .

3
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

4
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 6

1. Constraint and evolution equations

(Note: A long setup to a fairly short problem.)
Maxwell’s equations, written in terms of electric and magnetic fields and using cgs
units with the speed of light c = 1, take the form

∂i E i = 4πρ , ∂i B i = 0 ;
∂t E i = �ij k ∂j B k − 4πJ i , ∂t B i = −�ij k ∂j E k .

The first two equations — those relating the divergence of the fields — are called
“constraint equations”. This is because they involve no time derivatives — they are
differential relations among the fields (and sources) at a single moment in time.
The second two equations are called “evolution equations”. Because they involve the
time derivative of fields, they relate the fields and sources from one moment to the
next. We can thus use the second pair of Maxwell equations to evolve “data” (which
must satisfy the constraints) from some initial moment to an arbitrary late time.
We will now prove that the Einstein field equations have a similar structure. Because
the Einstein field equations are second order, we expect our “initial data” to consist
of fields (the metric) and first time derivatives. Constraint equations should thus be
those components of the Einstein equation which contain no more than a single time
derivative; evolution equations are those components which contain two derivatives.
(This is analogous to the kinematics of a particle: the “initial data” is the starting
position and velocity; we use the acceleration as our “evolution equation” to find the
particle’s motion from then on.)
Suppose we have chosen a timelike direction, so that x0 = t; we do not specify the
spatial coordinates xi other than to say that they are coordinates covering the x0 =
constant hypersurface.
Show that the Einstein tensor components G00 and G0i contain no more than one time
derivative. Thus, the equations G00 = 8πGT00 and G0i = 8πGT0i can be considered
constraints which relate the metric, its first time derivative, and sources at a single
moment of time; the equations Gij = 8πGTij are evolution equations.
(Hint: A brute force construction of the curvature tensor Gαβ in terms of the metric
and its derivatives will give you the correct solution to this problem. This is not a rec­
ommended procedure, though. A much quicker solution can be deduced by considering
the Bianchi identity applied to the Einstein tensor.)

1
2. Action for a cosmological constant
Show that varying the action
�
√
S= d4 x −g (R + a) ,

(where R is the Ricci scalar and a is a constant) yields the Einstein equation with a
cosmological constant. How does a relate to the cosmological constant Λ?

3. Nordstrøm’s gravity theory

A metric theory devised by G. Nordstrøm in 1913 (before general relativity was final­
ized) relates gµν to Tµν by the equations

Cαβγδ = 0 , R = κgµν T µν

where Cαβγδ = 0 is the Weyl curvature tensor.

The vanishing of Weyl tells us that the metric is conformally flat; this follows from
the fact that vanishing Riemann implies that spacetime is truly flat (not just confor­
mally flat), and that the Weyl tensor is invariant under conformal transformations.
Conformal flatness means that

gµν = e2φ ηµν ,

where φ = φ(xµ ) is a function of the spacetime coordinates. (To relate this to the
notation used in lecture, eφ = Ω; the exponential form is convenient for the calculations
here.)
(a) Show that in the limits φ2 � 1 and |∂t φ| � |∂i φ|, the geodesic equation for a test
body moving slowly (ui � 1) in this spacetime reproduces the kinematics of Newtonian
gravity. We’ll call this the “Newtonian limit” from now on.
(b) Show that in the Newtonian limit the Ricci scalar R is just a second order differ­
ential operator acting on φ. Compute that operator.
(c) Show that Nordstrøm’s field equation reduces in the Newtonian limit to the gravi­
tational field equations, and determine the value of κ.
(d) Is this theory consistent with the Pound-Rebka gravitational redshift experiment?
(This is the experiment which established that light in fact redshifts as the equivalence
principle predicts.)
(e) Show that there is no deflection of light by the sun in this theory of gravity.

2
4. Weighing a relativistic body
An object of mass m is at rest on a bathroom scale in a weak, uniform, static gravita­
tional field. That is, the object has fixed spatial coordinates (x, y, z) and the spacetime
metric has the standard weak-field form gµν = ηµν + 2φdiag(1, 1, 1, 1), with φ the nor­
mal Newtonian potential. We take φ2 � 1, ∂z φ = constant = −g, and ∂µ φ = 0 for
µ �= z. Neglect terms of order φ2 and φg in what follows.
In this problem, we will see that if one wants to interpret gravity as a force rather than
as the effect of spacetime curvature, then it must be a velocity-dependent force. This
is not a fundamental insight; the main purpose of the problem is to practice relating
the metric to measurable quantities in curved spacetime.
(a) What force does the bathroom scale apply on the body? Compute both the com­
ponents and the scalar magnitude of the 4-force. The principle to apply here is that
the body does not follow a geodesic: The equation of motion for the body is

D 2 xµ
m = muβ �β uα = F µ .
dτ 2

This relation may be taken to define the 4-force F µ .

(b) Now suppose that the object moves with constant, relativistic coordinate 3-velocity
v = dx/dt = (dx/dτ )(dt/dτ )−1 in the x-direction:

V x = vV t ; Vy =Vz =0.

What is V t ? (Don’t just use a special relativity formula!) While the mass is on the
bathroom scale, what force (components and magnitude) does the scale apply to the
mass?
(c) Now transform coordinates by applying a naive Lorentz transformation: t̄ = γ(t −
vx), x̄ = γ(x − vt), ȳ = y, z̄ = z. Evaluate the components of the metric in the new
coordinate system, gµ̄ν̄ . To first order in φ, what are the force components in this new
coordinate basis?
(d) Show that the barred coordinate basis can be transformed to an orthonormal basis,
�eµ̂ = E µ̄ µ̂�eµ̄ with a tetrad matrix

E µ̄ µ̂ = δ µ̄ µ̂ + φAµ̄ µ̂ .

Find the matrix Aµ̄ µ̂ . To first order in φ, do the force components F µ̂ differ from F µ̄ ?

3
5. Converting between geometrized and “normal” units
Especially as we discuss astrophysical applications, we will find it useful to work in
“geometrized units”, in which the gravitational constant and the speed of light are
both set to unity. When this is done, mass, length and time are measured in the same
units.
We convert among different units by multiplying by powers of G and c; since such a
factor is just 1 in geometrized units, we can include as many such factors as is necessary.
For example, to express the solar mass as a time, we write M�geom = GM� /c3 . Using

M� = 1.99 × 1033 gm ,
G = 6.67 × 10−8 cm3 gm−1 sec−2 ,
c = 3.00 × 1010 cm/sec ,

we find M�geom = 4.93 × 10−6 seconds.

Do the following conversions:
(a) Mass of the earth (M⊕ = 5.98 × 1027 gm) in centimeters.
(b) Characteristic mean density of neutron stars (ρ̄ = 1015 gm/cm3 ) in inverse square
centimeters.
(c) Characteristic mean pressure at core of a neutron star (P̄ = 1034 gm sec−2 cm−1 )
in inverse square centimeters.
(d) Acceleration of gravity at the surface of the earth (g = 9.8 m/s2 ) in inverse seconds
and inverse years.
(e) The typical (isotropic) luminosity of a gamma ray burst: L = 1053 erg/sec. (Re­
minder: 1 erg = 1 gm cm2 /sec2 .)
(f) Planck’s constant (h̄ = 1.05 × 10−27 erg sec) in square centimeters.
The square root of this quantity is called the “Planck length”, and is denoted lp .
Since it involves the constants which set quantum effects (h̄), gravitational effects (G),
and relativistic effects (c), it is thought that quantum gravitational effects (i.e., the
“quantization of spacetime”, whatever that actually means) must be important at
lengthscales L ∼ lp .
(g) Convert lp to a mass (gm); this is known as the “Planck mass”. Convert that to an
energy; express this energy in electron volts. Comment on the likelihood of observing
Planck mass scale effects at a particle collider.

4
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

5
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 7

1. Gravitomagnetism
In lecture and working in Lorentz gauge, we examined the linearized Einstein field
equations for a static source,

�h̄µν = −16πGTµν → �2 h̄µν = −16πGTµν ,

where �2 is the ordinary Euclidean 3-space Laplacian operator. For a static, non­
relativistic source, the only non-zero stress-energy component is (to sufficient accuracy
for our purposes)

T00 = ρ .

Using this, we found

h̄00 = −4Φ → hµν = −2Φ diag(1, 1, 1, 1) ,

where Φ = −GM/r is the Newtonian gravitational potential.

We will now modify this slightly by imagining that the source rotates, and thus is
characterized by a spin angular momentum with spatial components S i as well as a
mass M .
(a) Consider the source to be spherically symmetric, with uniform density ρ and radius
R. Take it to be rotating rigidly about the x3 ≡ z axis with constant angular velocity
Ω. Working in a Lorentz frame that is at rest with respect to the center of mass of
the source, work out all components of the stress energy tensor Tµν to first order in Ω.
(Assume ρ, R, and Ω are constant.) Indicate which components would change if you
included terms to second order in Ω, but don’t calculate those second order corrections.
(You may neglect pressure terms throughout your calculation.)
¯ 0i
(b) Solve for the Cartesian off-diagonal components h0x , h0y , h0z . (Note that h0i = h
since trace reversal has no effect on off-diagonal components.)
This is a moderately challenging calculation. The following tips should help:

• Recall that the formal solution to the Poisson-type equation for h0i is
�
T0i (x� ) 3 �
h0i (x) = 4G dx
|x − x� |

where x is the “field point”, the location of the point at which h0i is to be evalu­
ated, and x� is the “source point”, a coordinate within the source over which the
.
integral is taken. [Boldface quantities denote 3-vectors: x = (x, y, z).]

1
 • The following expansion for the factor 1/|x − x� | is very useful:
�
1 1 xj xj
= + 3 + ...
|x − x� | r r

You may assume this identity in your solution. Note also that a sum over j is
implied here; we are allowed to be sloppy about the placement of indices since
the spatial metric is δij to leading order. [This identity is more often seen as an
expansion in spherical harmonics; see, for example, J. D. Jackson, Sec. 3.6 (2nd
edition). This form in terms of Cartesian coordinates is equivalent.]
• After you have set up your integral, convert the primed integration variable to
spherical coordinates to do the integration:
�
x1 = x� → r� sin θ� cos φ�
�
x2 = y � → r� sin θ� sin φ�
�
x3 = z � → r� cos θ�

Your final metric components should be proportional to ρR5 /r3 .

(c) Using the identity S i = IΩi where I is moment of inertia and Ωi is the ith com­
ponent of the angular velocity vector, rewrite your answer in terms of the angular
momentum S i .
Although we derived this result for a special situation (uniform density, spherical body,
rigid rotation), the result we obtain in terms of S i is completely general; see, for
example, MTW Sec. 19.1.
(d) Converting to spherical coordinates, find h0r , h0θ , h0φ .
Hint: Only one of these components is non-zero. After changing coordinates, you
should find that this non-zero component is ∝ S z sin2 θ/r.

2. Comparison of linearized GR and Maxwell’s theory

Consider the line element

ds2 = −(1 + 2Φ)dt2 + (1 − 2Φ)(dx2 + dy 2 + dz 2 ) − 2β i dxi dt ;

in other words, the usual weak field line element on the diagonal with h0i = −β i .
(a) Show that the geodesic equation for a particle moving in this spacetime gives the
following equation of motion to first order in the particle’s velocity v:

d2 x
m = mg + m(v × H) .
dt2
Here, x is a 3-vector representing the position of the particle, and

g = −�Φ ,
H = �×β ,

where � represents the ordinary gradient operator in Euclidean 3-space.

2
 (b) Show that for stationary sources (i.e., no component of the stress energy tensor
shows time variation) the Einstein field equations may be written

�·g = −4πGρ ,
�×H = −16πGJ
�·H = 0,
�×g = 0.

The current J = ρv, where v is the velocity of fluid flow in the source. (Note that the
second two equations follow from the definitions of g and H, so the only labor is in
working out the first two.)
(c) These equations clearly bear a strong resemblance to Maxwell’s equations in the
limit ∂t E = ∂t B = 0; the main differences are the reversed sign in both equations,
and the extra factor of 4 (compared to Maxwell) in the curl equation. Can you give a
simple explanation for these differences?

3. Carroll: Chapter 7, Problem 1.

4. Carroll: Chapter 7, Problem 3.

5. Carroll: Chapter 7, Problem 4.

3
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

4
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 8

1. In lecture, we derived the following formula for the leading gravitational radiation
generated by a source:
2 1
� �
hTT
ij = I¨kl Pik Pjl − Pij Pkl .
r 2
Here, overdot denotes d/dt, Ikl = dV ρ xk xl , we are ignoring the distinction between
�

upstairs and downstairs indices since we are in a nearly flat region and can work in
nearly inertial coordinates, and the combination of projection tensors guarantees that
the resulting tensor is transverse and traceless.
Show that the same result is obtained if one uses the traceless “quadrupole moment
tensor” Ikl = Ikl − 13 δkl I instead of Ijk , where I = δkl Ikl .
Comment 1: This is less trivial than it may seem since there are really two different
trace operations defined here: A trace with respect to the spatial metric ηij = δij ,
and a trace with respect to the metric of the subspace orthogonal to the propagation
direction, Pij .
Comment 2: In general, a radiative l-pole has 2l+1 separate components — a scalar has
one component, a dipole has three, a quadrupole has five. The symmetric spatial tensor
Iij has 6 components — there must be “extra” information in that tensor unrelated to
radiation. This exercise proves that this extra information is bound up in the trace.
For this reason, you will often see the formula for hTT
ij written in terms of Iij rather
than Iij .
2. Binary system
Consider a binary consisting of two masses m1 and m2 in a circular orbit of radius R
about one another. Consider the orbit to be adequately described using Newtonian
gravity; in this problem, we will use this description to compute the leading effects due
to gravitational-wave emission.
[Hint: Don’t forget that orbits in a problem of this type are most easily described using
the “reduced system”: a body of mass µ = m1 m2 /(m1 + m2 ) in circular orbit around
a body of mass M = m1 + m2 . If you need a refresher, this result is derived in all
junior-level mechanics textbooks; see, for example, Goldstein Sec. 3.1.]
(a) Compute the gravitational-wave tensor hTT
ij as measured by an observer looking
down the angular momentum axis of the system (i.e., the z-axis if you define the orbital
plane as the x − y plane).
(b) Compute the rate at which energy is carried away from the system by gravitational
waves.
Due to this loss of energy, the radius of the orbit will gradually shrink, and the frequency
of the binary will “chirp” to higher frequencies as time passes.
1
 (c) By asserting global conservation of energy in the following form,

d
(Ekinetic + Epotential + EGW ) = 0 , (1)
dt
derive an equation for dr/dt, the rate at which the orbital radius shrinks.
[Hint: Don’t forget that for circular, Newtonian orbits, there is a simple relationship
expressing Ekinetic + Epotential as a function of r, as well as a simple result for the orbital
frequency as a function of r (Kepler’s 3rd law).]
(d) Derive the rate of change of the orbital angular frequency Ω due to gravitational-
wave emission. You should find that the masses only appear in the combination M ≡
µ3/5 M 2/5 , perhaps raised to some power. This combination of masses is known as the
“chirp mass”, since it sets the rate at which the frequency “chirps”.

3. Wave equation for the Riemann tensor in linearized theory

As we have emphasized from time to time, there is a nice analogy between the metric
of GR and the electromagnetic potential, and likewise between curvature tensors and
the electromagnetic field. This suggests that it should be possible to build a wave
equation for the curvature tensor.
As background to this problem and the next one, recall that the Einstein field equations
can be written in the trace-reversed form

Rµ αµβ = Rαβ = 8πT̄αβ

where T̄αβ = Tαβ − 12 gαβ T γ γ .

To keep things simple, we begin with linearized theory, working in nearly inertial
coordinates: gµν = ηµν + hµν with ||hµν || � 1.
(a) To linear order in h, the Bianchi identity can be written

∂α Rβγµν + ∂β Rγαµν + ∂γ Rαβµν = 0 .

Using this equation, show that the divergence of the Riemann tensor is related to the
gradient of the trace-reversed stress energy tensor:
� �
∂α Rα βγδ = 8πG source involving gradient of T̄µν .

(b) Now use the Bianchi identity and the solution to part (a) to develop a wave equation
for the Riemann tensor of the form
� �
�Rαβµν = 8πG source involving double gradients of T̄µν .

Solve this equation (formally) using the radiative Green’s function introduced in lec­
ture.
(c) Now specialize to a plane gravitational wave propagating in the z-direction through
vacuum. The corresponding solution to the above wave equation takes the form
Rαβµν = Rαβµν (t − z). Using the Bianchi identity and the symmetries of Riemann,
show that the only non-zero components of the Riemann tensor are of the form Ri0j0
2
 (plus components that are trivially related by symmetries; recall that indices i, j only
refer to spatial indices).
(d) Show that the only non-zero Ri0j0 are Rx0x0 (t−z) = −Ry0y0 (t−z) and Rx0y0 (t−z) =
Ry0x0 (t−z). The first non-zero components correspond to the + polarization discussed
in lecture; the second corresponds to the × polarization.
(e) Define fields h+ (t − z) and h× (t − z) in terms of these components of the Riemann
tensor by
1 1
Rx0x0 = − ∂t2 h+ , Rx0y0 = − ∂t2 h× .
2 2
Also recall the expression for the Riemann tensor in linearized theory:
1
Rαβµν = (∂α ∂ν hβµ + ∂β ∂µ hαν − ∂α ∂µ hβν − ∂β ∂ν hαµ )
2
Comparing these two forms, show that h+ = hTT TT TT TT
xx = −hyy , h× = hxy = hyx .

(f) Show that when one rotates the coordinate system about the waves’ propagation
direction (i.e., about the z-axis) by an angle θ [so that x� + iy � = (x + iy)e−iθ ], the
gravitational-wave fields h+ and h× transform such that

h�+ + ih�× = (h+ + ih× )e−i2θ

This equation is equivalent to the statement that the graviton is spin 2 — the quantity
h+ + ih× has “spin-weight” 2.

4. Nonlinear wave equation for the Riemann tensor

Begin now with the full Bianchi identity:

�α Rβγµν + �β Rγαµν + �γ Rαβµν = 0 .

(a) Develop the fully covariant analog to your answer to part (a) of problem (3):
� �
�α Rα βγδ = 8πG source involving covariant gradient of T̄µν .

(b) Using the Bianchi identity and the solution to part (a), develop a nonlinear wave
equation for the Riemann tensor of the form

�Rα βγδ = 8πG [source] .

In this case, the source should involve three kinds of terms: double covariant gradients
of T̄µν , coupling of T̄µν to the Riemann tensor, and coupling of Riemann to Riemann.
Here, the operator � = �µ �µ .
Comment: This non-linear wave equation was first developed by Roger Penrose, and
is sometimes called the Penrose wave equation. A variant of this equation based on
an expansion around the Riemann tensor for a black hole spacetime was developed
by Teukolsky and has played an extremely important role in astrophysical relativity
research.

3
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

4
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 9

1. Time scale for variation in microlensing

Consider the microlensing scenario described in the April 20 lecture. (“Microlensing”
is the limit in which the images cannot be separated; one just examines the total
intensity produced by the sum of the two images.) Imagine that the source is a star
in the Large Magellanic Cloud (about 100 kiloparsecs away; 1 pc = 3.26 light years =
3.1 × 1013 km); imagine that the lens is an unseen object of about one solar mass in
the galaxy, roughly 10 kiloparsecs away. If the lens is orbiting in the galactic potential
with a typical galactic orbit speed vorb ∼ 200 km/sec, what is the timescale over which
you expect the light curve to vary? In other words, roughly how long will it take for
the light curve to brighten and then dim?
For students who could not attend that lecture, see Hartle’s textbook, Chapter 11.

2. Proper motion distance

Using the definition of proper motion distance given in Carroll [Eq. (8.124)], compute
dM (z). Your final result should be similar in form to Eq. (8.123) of Carroll and should
confirm the rule that

dL (z) = (1 + z)dM (z) = (1 + z)2 dA (z) ,

where dL (z) is luminosity distance and dA (z) is angular diameter distance.

3. Particle horizons in the early universe

[Note: Carroll Sec. 8.8 provides important background to this and the next two prob­
lems. A very readable and even more comprehensive discussion is given in Chapter 3
of Cosmological Inflation and Large Scale Structure, by Andrew R. Liddle and David
H. Lyth (Cambridge University Press, 2000), and in Chapter 13 of An Introduction to
Modern Cosmology, by Andrew R. Liddle (Wiley, 2003).]
The early universe was so hot and dense that all matter existed in the form of a plasma.
As such, photons did not stream freely, but rather scattered repeatedly off the ionized
charges in the plasma. “Recombination” is the name given to the point at which the
universe cooled sufficiently that electrons and protons could bind into atoms; at this
point, photons can stream freely. The cosmic microwave background (CMB) dates
from the recombination epoch, which is now known to have occurred at a redshift
zr � 1200.

1
 Consider a k = 0 FRW cosmology which is radiation dominated prior to recombination
and is matter dominated afterwards1 . Let rH (tr ) be the maximum coordinate distance
that a photon can travel from the big bang to the epoch of recombination; this is
the “comoving horizon radius” at recombination. (More properly, it is the comoving
particle horizon radius, to be contrasted with the event horizon we will define in about
two weeks.) Only points that are within each other’s particle horizons are in causal
contact.
Let robs (tr ) be the coordinate distance from recombination to an observer on earth;
thus 2 × robs (tr ) is the coordinate separation between patches of the CMB on opposite
parts of the sky.
(a) Compute rH (tr ) and robs (tr ).
The CMB is extremely isotropic; its temperature varies by roughly a part in 105 over
the sky.
(b) Given the isotropy of the CMB, why is the smallness of the ratio rH /robs so puz­
zling? (If you find that this ratio is not small, you might want to revise your answer
to part a!) This is known as the “horizon problem”.
(c) Compute the angular scale θc such that sources separated by θ < θc as seen today
were in causal contact prior to t = tr , but sources separated by θ > θc were not.

Some background: For the remaining two problems, we repeatedly need to refer to
temperature and energy scales commonly used in early universe cosmology. Since
this subject grew out of particle physics, the units of particle physics describe its
relevant quantities — e.g., temperatures are usually given in MeV or GeV. This means
that temperatures are really being reported in the combination kB T , where kB is
Boltzmann’s constant.
The following conversions and definitions will prove useful in problems 4 and 5:
T0 ≡ Mean temperature of universe today (CMB temperature) = 2.73 K
kB ≡ Boltzmann’s constant = 8.6 × 10−11 MeV/K
h̄ = 6.6 × 10−22 MeV sec
π 2 kB4
σ ≡ Stefan-Boltzmann constant = .
60h̄3 c2
The last constant comes in handy when converting between temperature and energy
density for radiation: A gas of photons2 at temperature T has an energy density
ρ = 4(σ/c)T 4 .
Finally, don’t forget that the temperature of radiation redshifts: Radiation whose
temperature is T today was at temperature a(t)T /a(0) = (1 + z)T at an earlier time t.

1
In fact, the universe became matter dominated somewhat prior to recombination: The redshift at which
matter and radiation contribute equal to the mean density of the universe is zeq � 3000. The error we
make in approximating zeq � zr is small enough for the purposes of this and the next two problems that we
may safely consider the universe to be radiation dominated prior to recombination, and matter dominated
afterwards.
2
If you read the texts by Liddle, you will discover that we are neglecting neutrinos and other relativistic
particles in this analysis. Our conclusions will thus be off by a factor of order unity; fine for our purposes.
2
4. The horizon problem resolved: The inflationary universe
Suppose that at some very early epoch, t1 ≤ t ≤ t2 (where t1 � t2 � tr ), the universe
resides in a “false vacuum” state: A scalar field φ fills spacetime and provides an
effective vacuum energy ρvac � V (φ). (For more details about how a scalar field of
this type is characterized and evolves, see Carroll pp. 369 – 370.) This vacuum energy
acts like a cosmological constant Λ = 8πρvac . Up to factors of order unity, we may
4
write ρvac � (4σ/c)Tvac , where kB Tvac ∼ 1015 GeV. (Note: Don’t take this formula for
density as a function of temperature too seriously! The vacuum doesn’t actually have
a temperature, and it is certainly not a gas of photons. This formula is designed to
give this scalar field an energy density that is comparable to the scale of grand unified
theories. We are recycling the photon energy density formula because it gives us a nice
way of understanding the energy scales at which this field operates.)
As the universe expands, the potential slowly evolves. Around time t2 , the scalar
field decays into standard model particles, and the stress energy tensor is no longer
dominated by V (φ). This particles provide the matter and radiation content for our
universe; it is then radiation dominated until recombination, and matter dominated
thereafter.
We will take “slowly evolving” to mean that we can regard V (φ) as a constant from
t1 ≤ t ≤ t2 , and assume that V (φ) � 0 for t > t2 . (See Carroll Sec. 8.8, Liddle and
Lyth Chapter 3, or Liddle Chapter 13 for further discussion.)
(a) By inserting appropriate factors of G and c, and using the fact that 1 MeV =
1.6 × 10−12 gm cm2 /sec2 , compute the value of the effective cosmological constant Λ
in inverse seconds squared.
(b) Repeat the calculation in Problem 3, part (a). Find the value of Ne ≡ Δt(Λ/3)1/2
(where Δt = t2 − t1 ) such that robs (tr )/rH (tr ) = 1. (Ne is often referred to as the
number of “e-foldings” of expansion for reasons that are hopefully obvious.) Estimate
t2 . (You may approximate t1 � 0 for this estimate).
(c) What is the spatial expansion factor, a(t2 )/a(t1 ), during this “inflationary” epoch?
(d) Recalculate the angular scale θc in this inflationary universe.

5. The flatness problem

(a) Use the Friedmann equations to derive a general expression relating values of Ω
at different times — i.e., Ω1 at t1 (corresponding to scale factor a1 ) and Ω2 at t2
(corresponding to a2 ). Do this calculation for a matter dominated universe, and for
a radiation dominated universe. Express your answer in terms of Ω − 1 and the scale
factor. Do not take into account the inflationary physics of Problem 4.
(b) Observations today tell us that the universe is flat with about 1% accuracy; i.e.,
Ω = 1 ± 0.01. Continuing to neglect inflation, estimate the value of Ω − 1 at the epoch
of nucleosynthesis (TN ∼ 1 MeV), and at the presumed epoch of grand unification
(TGUT ∼ 1015 GeV). Given that the universe is flat with at least 1% accuracy today,
how close to flat must it have been in these earlier epochs?
The incredible precision with which the universe must have been flat is called the
“flatness” problem. The early universe’s initial conditions must either have been tuned
3
with incredible precision to give the universe we see today, or some mechanism must
exist to drive the universe into a nearly flat state.
(c) Repeat part (a) using the inflationary universe of Problem 4.
(d) How does inflation resolve the flatness problem? If the universe passes through
enough “e-foldings” to cure the horizon problem of Problem 4, is it likely to leave the
universe flat enough to account for the observed spatial curvature today?

4
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

5
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 10

1. Isotropic representation of the Schwarzschild metric

(a) The Schwarzschild line element, written in Schwarzschild coordinates, takes the
form
2GM dr2
� �
ds2 = − 1 − dt2 + + r2 dΩ2 .
r (1 − 2GM/r)

Show that changing to the radial coordinate r̄, defined by

�2
GM
�
r = r̄ 1 + ,
2r̄
puts the Schwarzschild line element into the form

ds2 = gtt (r̄)dt2 + g(r̄)(dr̄2 + r̄2 dΩ2 ) .

This new coordinate system is called “isotropic coordinates”, since it emphasizes the
fundamental local isotropy of the three spatial directions.
Compute the metric functions gtt (r̄) and g(r̄).
(b) Take the limit r̄ � GM . Show that the line element then reduces to a form
appropriate for describing the exterior of a spherical body in the linearized limit of
general relativity.

2. Numerical construction of neutron star models in GR

A moderately accurate approximation to the equation of state of the material which
makes up a neutron star is given by the polytropic form

P = KρΓ0

where P is the pressure, ρ0 is the rest matter density, and the constants are given by

Γ = 5/3 ,
¯2
32/3 π 4/3 h
K = 8/3
5 mn
= 5.38 × 109 (dyne/cm2 )(gm/cm3 )−5/3 .
= 5.38 × 109 gm−2/3 cm4 sec−2 .

In this problem, you will numerically integrate the TOV equations of stellar structure
to build models of neutron stars in general relativity. Your goal will be to examine the
total mass M∗ and radius R∗ as a function of central density ρ0,c . As such, you must
1
familiarize yourself with some system for numerically solving a system of ordinary
differential equations. My personal experience has shown that the routine NDSolve
built into the package Mathematica is well suited to this problem. (An example of
using NDSolve will be posted to the 8.962 website, to illustrate how it works.) More
ambitious students with sufficient programming and numerical methods expertise may
prefer to write their own code in C or C++, or whatever you are most comfortable with.
During your calculations, don’t forget to take into account the difference between rest
mass density ρ0 and relativistic energy density ρ:

P KρΓ0
ρ = ρ0 + = ρ0 + .
Γ−1 Γ−1
The equations you will integrate take the form
dm
= 4πρr2
dr
dP (ρ + P )(m + 4πr3 P )
= − .
dr r(r − 2m)

We have here written these equations in units in which both G and c are set equal to
1. The initial conditions you need to apply are

m(r = 0) = 0 ,
P (r = 0) = Pc = P (ρ0,c ) = KρΓ0,c .

One integrates these equations until the pressure drops to zero: P (R∗ ) = 0 defines
the star’s surface. (In fact, you will probably find in your numerical integration that
Mathematica or your code attempts to take an infinite number of infinitesimally
small steps in the vicinity of R∗ . The radius at which this numerical inaccuracy sets
in will still define the surface of the star to very high accuracy.) The total mass of the
star is then defined as m(R∗ ) ≡ M∗ .
(a) Experience has shown that numerical calculations of this kind behave best when
the units are chosen such that many key quantities are within an order of magnitude
or so of unity. A good choice is to put G = 1 and c = 1; we then pick the units of
all dimensionful quantities to be powers of kilometers. With this choice, the order of
magnitude of the total mass will be roughly 1 kilometer, and the order of magnitude
of the stellar radius will be roughly 10 kilometers.
Using this unit system, convert
(i) ρ = 1 gm/cm3 to km−2 .
(ii) P = 1 dyne/cm2 = 1 gm cm−1 sec−2 to km−2 .
(iii) K = 1 gm−2/3 cm4 sec−2 to km4/3 . Use this conversion factor to convert the K that
appears in the polytropic equation of state to km4/3 .
(b) Pick a central density ρ0,c = 1015 gm/cm3 . Inserting appropriate conversion factors
such that all quantities are in units of km (or powers of km), integrate the TOV
equations to compute the radius R∗ and mass M∗ of the star that has this central
density.
2
(c) If a photon is emitted radially with energy Eem from the surface of this star, what
is the energy Eob with which this photon is observed by distant (r → ∞) observers?
Using these energies, compute the surface redshift
Eob − Eem
zsurf = .
Eob

(d) As described in lecture, the mass M∗ is not what one would get by integrating all
fluid density elements over the proper volume of the star’s interior. Let us define Mp as
the mass that would be obtained by this integration. Re-integrate the TOV equations
for the central density 1015 gm/cm3 , but add the equation
� �−1/2
dmp √ 2m(r)
= 4πρr2 grr = 4πρr2 1 − .
dr r

Compute the mass MP , which is mp (R∗ ).

(e) The difference between Mp and M∗ reflects the fact that the self-gravity of the star
contributes to the star’s mass. These differences can be regarded as the “gravitational
binding energy” of the star. Compute
Mp − M∗
Δ≡ .
M∗
Δ is the fraction of the star’s mass that is due to this binding energy. What is this
fraction for this choice of central density?

3
3. Stability of a TOV star
By computing a range of TOV models, we can assess whether a star is stable against
radial (i.e., purely spherical) perturbations. Detailed analysis shows that a stable star
satisfies
dM
>0;
dρc
an unstable model satisfies
dM
<0.
dρc

(Notice it is ρc that appears in this criterion, not ρ0,c .) A detailed explanation and
derivation of this stability criterion can be found in Black holes, white dwarfs, and
neutron stars: The physics of compact objects, by S. L. Shapiro and S. A. Teukolsky,
§6.8 - 6.9. An intuitive explanation goes as follows:
Suppose we construct a TOV model, and then squeeze it, decreasing its radius by
some small amount δR∗ . Clearly, the mean density, and thus the central density, will
be augmented, since we are decreasing the volume of the object. If the star resists
the squeezing, its mass will also increase: It does work to resist the squeezing, work is
energy, and mass and energy are equivalent. On the other hand, if the star does not
resist the squeezing, its mass will decrease: A small squeeze puts it into an energetically
more favorable smaller radius. This clearly leads to a runaway, causing the star to
collapse into a black hole.
(a) Compute a range of TOV models with central densities ρ0,c = 1014 , 1015 , 1016 , 1017
and 1018 gm/cm3 . Make a plot of M∗ vs ρc for these models. Do you see any evidence
for a change from stable to unstable behavior?
(b) Zoom in on any unstable region you find and locate, as accurately as possible, the
marginally stable star (dM/dρc = 0). This star has the maximum possible mass for
this equation of state in general relativity. What is this maximum mass? Convert your
result from km to solar masses.

4
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

5
 Massachusetts Institute of Technology
Department of Physics
8.962 Spring 2006

Problem Set 11

1. Formation of a black hole

An objection to the Schwarzschild black hole solution described in class is that it is
“eternal”. In this problem, we will see how a Schwarzschild black hole can be formed
from the collapse of a simple, non-singular physical object.
We will examine the collapse of a “star” composed of pressureless dust. We take the
star to be spherical, of initial radius R∗ , of mass M , and composed of an isotropic
distribution of pressureless dust. By Birkhoff’s theorem, the exterior of this star is
simply described by the Schwarzschild metric:
2GM dr2
� �
ds2r>R∗ =− 1− dt2 + + r2 dΩ2 .
r 1 − 2GM/r
Since the interior is spatially isotropic, it is perfectly described by the Robertson-
Walker metric. Since the star clearly must collapse under its own self gravity, we use
the closed RW metric to describe the interior:
� �
ds2r<R∗ = −dτ 2 + a2 (τ )R02 dχ2 + sin2 χdΩ2 .

(We use different notation for time on the interior and the exterior: The exterior time
t is Schwarzschild time, a convenient label very far away, but perhaps confusing in the
strong field; the interior time τ denotes the proper time experienced by an element of
dust inside the star.)
(a) The evolution of the scale factor for a closed RW line element turns out to have a
simple, closed form solution. Show that the parametric solution
amax
a = (1 + cos η) ,
2
amax R0
τ = (η + sin η) ,
2
with 0 ≤ η ≤ π, solves the Friedmann equations for k = 1 assuming ρ is given by
pressureless dust matter. Since a always appears multiplied by the lengthscale R0 in
the line element, you may set amax = 1. Find a relationship between the initial density
ρ0 and the lengthscale R0 .
(b) This solution for the interior time coordinate τ is only good up to τ = πR0 /2.
What happens to the interior solution after that?
We now need to examine the star’s surface from the point of view of the external
coordinate system.
1
(c) Consider a purely radial “orbit” (i.e., a trajectory with no angular momentum:
L = 0). For a given energy per unit mass E, find the radius R at which the radial
velocity goes to zero.
We will use this solution to define the “orbital energy” of a dust element at the surface
of the star as it begins to collapse.
(d) Using the radial geodesic equation for the Schwarzschild geometry and the rela­
tionship between E and R you found in (b), write down an integral for the proper time
τ it takes for a fluid element at the star’s surface to fall from its initial radius R∗ to r.
You should find the answer
� r
dr�
τ =− � .
R∗ 2GM/r� − 2GM/R∗
(The minus sign is due to the infalling motion of the surface.) By introducing the
parameterization
R∗
r= (1 + cos η) , 0≤η≤π,
2
show that this integral can be evaluated to give
�
R∗3
τ = (η + sin η) .
8GM
We now match the inner and the outer coordinate systems: We require that the star’s
circumference be the same in both the inner and the outer coordinate systems for all
η, and we require that the two expressions for the proper time τ experienced by a fluid
element on the star’s surface be the same for all η.
(e) By enforcing these two conditions, determine the lengthscale R0 and the Robertson-
Walker radius of the star χ.
For the next part of the problem, assume that the star’s initial radius is R∗ = 5GM .
A Schwarzschild black hole’s event horizon is a null surface: It is “generated” by null
geodesics whose coordinate locations are r = 2GM for all time. The event horizon of
a black hole that forms in collapse is “generated” by the null geodesic that begins at
the star’s center and reaches the surface just as the surface passes through r = 2GM ;
at that point, by Birkhoff’s theorem this horizon “generator” will remain at r = 2GM
for all time.
(f) Determine the time τ at which the horizon generator leaves the center of the star.
Hint: It is easiest to solve for this geodesic by noting that the parametric solution for
the closed RW spacetime allows us to write
� �
ds2 = a(η)2 R02 −dη 2 + dχ2 + sin2 χ dΩ2 .
An outward propagating null geodesic thus obeys dχ/dη = 1. Using the value of χ you
found for the star’s surface, it should thus be easy to integrate backwards from the
moment that the star’s surface crosses r = 2GM to determine the value of η at which
the horizon generator leaves the star’s center. You then just need to convert to τ .
(g) On a spacetime diagram, sketch the evolution of the star’s surface and of the event
horizon.
2
2. Consider a static, spherical star cluster in which all stars move in circular orbits. Ignore
collisions between stars (i.e., approximate the stars as non-interacting dust). Adopt
Schwarzschild-type coordinates with r = 0 at the center of the cluster and write the
metric in the form

ds2 = −e2Φ dt2 + e2Λ dr2 + r2 dΩ2 ,

Φ = Φ(r), Λ = Λ(r) .

(a) Find e2Λ and dΦ/dr in terms of m = 0r 4πρr2 dr, where ρ = ρ(r) is the stars’
�

mass density in the cluster. (We assume that there are enough stars that a continuum
treatment is accurate.) Hint: Your final equations should be similar to the TOV
equations but with one rather crucial difference.
(b) Define an appropriate effective potential Veff (r). Use it to determine the energy
per unit mass Ê and angular momentum per unit mass L̂ of a star in the cluster.
Your answer should be expressed in terms of r, m(r), and Φ(r). Determine the orbital
frequency Ω ≡ dφ/dt = (dφ/dτ )/(dt/dτ ).
(c) Use Veff to analyze the stability of orbits of individual stars in the cluster. What
local condition must Gm(r)/r satisfy if all orbits at r are to be stable?
(d) Apply the above results to homogeneous cluster of total mass M and radius R.
[Homogeneous means ρ(r) = const, so m(r) = M (r/R)3 for r ≤ R; you will need to
use this to solve for Φ(r) to complete this part of the problem.] Find the maximum
value of GM/R if all orbits are to be stable.
(e) Find the cluster with maximal GM/R, compute the redshift of photons emitted
from the cluster’s surface, and from its center. When quasars were first discovered,
their typical redshift was on the order of z ∼ 0.3. Could a cluster of this type explain
this redshift? Today, quasars are measured with redshifts as high as z � 6.5. How well
does the relativisic cluster hypothesis explain these quasars?

3
3. Numerical studies of black hole orbits
In lecture, we showed that the following equations govern the motion of a test body
moving around a Schwarzschild black hole:
� �2
dr
= Ê 2 − Veff (r) , where
dτ
L̂2
�� �
2GM
�
Veff (r) = 1− 1+ 2 ;
r r
dφ L̂
= 2 ;
dτ r
dt Ê
= .
dτ 1 − 2GM/r

In this exercise, you will numerically integrate these equations to study some inter­
esting black hole trajectories. Note, the equation for r involves a slightly troublesome
singularity at turning points that can make numerical integration tricky: At a turning
point, dr/dτ changes sign. Naively, you might imagine that this means you have to
carefully change which root to take as you pass through these turning points. Instead,
try taking an additional derivative of both sides and rearranging.
It is useful to define units in which GM = 1; this amounts to redefining r → r/GM ,
t → t/GM , τ → τ /GM , L ˆ → L/GM
ˆ .
(a) With this choice, what are the basic units of time and length if M = 10 solar
masses?
Now, construct examples of the following types of orbits:
(b) An unbound orbit — one which comes in from infinity, passes near the black hole,
and zooms back out to infinity. How close can you get the orbit to come to the black
hole without being captured?
(c) An “elliptical” bound orbit. You will rapidly discover that these orbits are not truly
ellipses (as in Newtonian theory), but are at best precessing ellipses. Try to make an
orbit whose innermost radius and outermost radius are as widely separated as possible.
(d) A stable circular orbit.
(e) An unstable circular orbit. How long can your integrator hold the unstable circular
orbit at fixed r? It’s entirely possible that your numerical integrator will hold this
orbit for quite a long time; if that’s the case, consider slightly perturbing the energy or
angular momentum to see how it evolves. Can you drive the system into both outcomes
of instability (plunge into the hole and escape to infinity)?

4
MIT OpenCourseWare
https://ocw.mit.edu

8.962 General Relativity

Spring 2020

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.