A First Course in General Relativity

Bernard F Schutz
(2nd Edition, Cambridge University Press, 2009 (reprinted with
corrections 2011))

Solutions to Selected Exercises

(Version 1.1, February 2011)

To the user of these solutions:

This document contains solutions to many of the Exercises in the second

edition of A First Course in General Relativity. The textbook offers an ex-
tensive collection of exercises, some of which prove results omitted from the
text, or form the basis for later exercises, or provide the foundation for re-
sults in later chapters. Doing exercises is integral to the process of learning a
subject as complex and conceptually challenging as general relativity. These
solutions, therefore, are meant to help users of this book master exercises
that they might have had difficulty with. It is assumed that an attempt has
already been made at solving the problem!

Solutions are provided to those problems which might present particular

conceptual challenges; those which require just routine algebra, and those
which require computer solutions, are generally not included.

I hope you find the solutions helpful!

Bernard Schutz
Potsdam, Germany

1
 Chapter 1
1 3.7×10−24 kg m−1 ; 3.5×10−43 kg m; 10−7 ; 10−4 kg; 1.1×10−12 kg m−3 ;
103 kg m−3 ; 3.7 × 10−16 kg m−3 .

2 3×106 m s−1 ; 9×1035 N m−2 ; 3.3×109 s; 9×1016 J m−3 ; 9×1017 m s−2 .

3 (a)–(k) See the top panel in Figure 1. Note that most of the items
continue into other quadrants (not shown). (l) See the bottom panel
in the same figure. Note that light always travels at speed 1 in this
frame even after reflecting off of a moving mirror.

(c) (g)
(k)
(b)(a) (i)
3
tO
tO
(m) (h)
2
(d) (c)
(e)
xO
(f )
1

(f )
(j)

1 2 x (m) 3

3
tO
tO
(l)
(m)
2

absorption
xO
1
mirror

reflection 1 2 xO (m) 3
detector

Figure 1: Solution to Ex. 3 of Chapter 1. See the solution text for explana-
tion.

2
 5 (a) See the top panel in Figure 2. The particles (short dashed world
lines) are emitted at event A and reach the detectors at events B and C.
The detectors send out their signals (more short dashed world lines)
at events D and E, which arrive back at the spatial origin at event
F. The line BC joining the two reception events (long dashed) at the
detectors is parallel to the x=axis in our diagram, which means that
they occur at the same time in this frame. Note that the lines DF
and EF are tilted over more than the lines AB and AC because the
returning signals go at speed 0.75 while the particles travel only at
speed 0.5.

(a) 6
F

4
tO
D (m) E
2 C
B

-4 -2 2 4
xO (m)

-2 A

tO
(c) 6 F

E
4 xO
C

tO (m)
2
D
B

-4 -2 2 4
xO (m)

-2
A

Figure 2: Solution to Ex. 5 of Chapter 1. See the solution text for explana-
tion.

3
 (b) The experimenter knows that the detectors are equidistant from
x = 0, and that the signals they send out travel at equal speeds.
Therefore they have equal travel times in this frame. Since they arrive
at the same time, they must have been sent out at the same time.
This conclusion depends on observer-dependent things, such as the
fact that the signals travel at equal speeds. Therefore the conclusion,
while valid in this particular frame, might not be valid in others.
(c) See the bottom panel in Figure 2. This is drawn using the axes of
frame Ō, which we draw in the usual horizontal and vertical directions,
since any observer would normally draw his/her axes this way. In this
frame, the frame O moves forwards at speed 0.75. These axes are
drawn in gray, and calibrated using invariant hyperbolae (not shown),
just as in the solution to Exercise 1.3. Then the events are located
according to their coordinate locations on the axes of O. Note that
to do this, lines of constant tO must be drawn parallel to the xO axis,
and lines of constant xO must be drawn parallel to the tO axis. One
such line is the long-dashed line BC. The detectors (heavy lines) must
pass through the points xO = ±2 m on the xO -axis. The two signal-
emission events D and E are clearly not simultaneous in this frame,
although they are simultaneous in O: D occurs much earlier than E.
Note also that the signal sent at the event E and received at F remains
at rest in Ō, since it was sent backwards at speed 0.75 in frame O,
exactly the same speed as the frame Ō.
(d) The interval is easily computed in frame O because the events D
and E have zero separation in time and are separated by 4 m in x.
So the squared interval is 16 m2 . To compute it in frame Ō, measure
as carefully as you can in the diagram the coordinates tŌ and xŌ for
both emission events. Given the thickness of the lines representing the
detectors, you will not get exactly 16, but you should come close.

6 Write out all the terms.

7 M00 = µ2 − α2 , M01 = µν − αβ, M11 = ν 2 − β 2 , M22 = a2 , M33 =

b2 , M02 = M03 = M12 = M13 = M23 = 0.

8 (c) Use various specific choices of ∆xi ; e.g. ∆x = 1, ∆y = 0, ∆z =

0 ⇒ M11 = −M00 .

10 Null; spacelike; timelike; null.

4
11 “Asymptotic” refers to the behavior of the curves for large values of x
and t. But when these variables are sufficiently large, one can neglect
a and b, and then one has approximately −t2 + x2 = 0, leading to
t = ±x. This approximation is better and better as t and x get larger
and larger.

13 The principle of relativity implies that if time dilation applies to one

clock (like one based on light travel times over known distances, which
is effectively the sort we use to calibrate our time coordinate), then it
applies to all (like the pion half-life). Algebra gives the result.

14 (d) For (a), 3.7 × 10−5 .

16 (a) In Fig. 1.14, we want the ratio t̄/t for event B (these are its time-
coordinate differences from A). The coordinates of B in O are (t, v t).
The first line of Eq. (1.12) implies t̄ = t(1 − v 2 )1/2 .

17 (a) 12 m.
(b) 1.25 × 10−8 s = 3.75 m; 211 m2 : spacelike.
(c) 9 m; 20 m. (d) No: spacelike separated events have no unique time
ordering. (e) If one observer saw the door close, all observers must
have seen it close. The finite speed of transmission (< c) of the shock
wave along the pole prevents it behaving rigidly. The front of the pole
may stop when it hits the wall, but the back keeps moving and can’t
be stopped until after the door has shut behind it. Many apparent
paradoxes in special relativity are resolved by allowing for the finite
speed of transmission of pressure waves: for a body to be perfectly
rigid would violate special relativity, since the communication among
its different parts would have to happen instantaneously.

18 (b) tanh[N tanh−1 (0.9)] ≈ 1 − 2(19)−N .

19 (c) The analog of the interval is the Euclidean distance x2 + y 2 . The

analog of the invariant hyperbola is the circle x2 + y 2 = a2 . The circle
allows one to transfer the measure of length along, say, the x-axis up
to the y-axis.

21 The easiest way to demonstrate these theorems is graphically: put two

arbitrary events onto a spacetime diagram, and join them. If they are
time-like separated, then the line joining them will be more vertical
than a light-line, and so there is a frame in which it is parallel to
the time-axis, in which frame it will represent a line of fixed spatial

5
position: the two events will be at the same spatial point in this new
frame. This proves (a). The proof of (b) is similar.

6
 Chapter 2
1 (a) −4; (b) 7, 1, 26, 17; (c) same as (b); (d) −15, 27, 30, −2 (different
from (b) and (c) because the sum is on the second index, not the first);
(e) A0 B0 = 0, A2 B3 = 0, A3 B1 = 12, etc.; (f) −4; (g) the subset of
(e) with indices drawn from (1, 2, 3) only.

2 (c) γ, α free; µ, λ dummy; 16 equations.

5 (b) No.

6 To get the basis vectors of the second frame, compose velocities to get
v = 0.882.

8 (b) The difference of the two vectors is the zero vector.

~ getting the four equations in
10 Choose each of the basis vectors for A,
Eq. (2.13) in turn.

12 (b) (35/6, −37/6, 3, 5).

14 (a) −(0.75/1.25) = −0.6 in the z direction.

15 (a) (γ, γv, 0, 0); (b) (γ, γv x , γv y , γv z ) with γ = (1 − v · v )−1/2 . (d)

v x = U x /U 0 = 0.5, etc.

18 (a) Given ~a · ~a > 0, ~b · ~b > 0, ~a · ~b = 0; then (~a + ~b) · (~a + ~b) =

~a · ~a + 2~a · ~b + ~b · ~b > 0.
(b) If ~a is timelike, use the frame in which ~a → (a, 0, 0, 0).

19 (b) v = αt(1 + α2 t2 )−1/2 , α = 1.1 × 10−16 m−1 ; t = 2.0 × 1017 m =

6.7 × 108 s = 22 yr.
(c) v = tanh (ατ ), x = α−1 [(1 + α2 t2 )1/2 − 1], 10 yr.

22 (a) 4 kg, 3.7 kg, 0.25 (e x +e y ); (b) (3, − 21 , 1, 0) kg, 3 kg, 2.8 kg, − 16 e x +
1
3 e y , 0.2 e y .

23 E = m + 12 m|v |2 + 83 m|v |4 + . . .; |v |2 = 2/3.

24 Work in the CM frame.

25 (a) Lorentz transformation of p~. (b) v = 2 cos θ/(1 + cos2 θ).

27 The cooler, because ratio of the rest masses is 1 + 1.1 × 10−16 .

7
31 2hν cos θ; hν cos θ.

33 Emax = 8 × 105 mp , above the γ-ray band.

8
 Chapter 3
3 (b) (i) −1; (ii) 2; (iii) −7; (iv) same.

4 (b) p̃ → (−1/4, −3/8, 15/8, −23/8); (b) −5/8; (d) yes.

5 The order of the multiplication of numbers does not matter, nor does
the order in which the sums are done; summation over ᾱ produces the
unit matrix because the two Λ matrices are inverse to one another;
summation over µ is just multiplication by the unit identity matrix.

6 (a) Consider p̃ = ω̃ 0 , an element of the basis dual to {~eα }; (b) (1, 0,

0, 1).

8 See Figure 3 for the diagram. Note that in the representation of a one-
form as a set of surfaces, the extent of the surfaces and the number
of them is not important. The critical property is their orientation
and spacing. The one-form d̃t has surfaces separated by one unit in
the t-direction so that, when the basis vector ~e0 is drawn on top of
it, it crosses exactly one surface, and when the other basis vectors are
placed on it they cross no surfaces (they lie in the surfaces).

}
3
tO
(m)
2 dt

1 2 x O (m) 3

3
tO
(m)
2

1
}

dx

1 2 x O (m) 3

Figure 3: Solution to Ex. 8 of Chapter 3. See the solution text for explana-
tion.

9
 9 d̃T (P) → (−15, −15); d̃T (L) → (0, 0).

10 (a) A partial derivative wrt xβ holds all xα fixed for α 6= β, so

∂xα /∂xβ = 0 if α 6= β. Of course, if α = β then ∂xα /∂xβ = 1.

12 (a) By definition, ñ(V~ ) = 0 if V

~ is tangent to S, so if for some vector
~ ~ ~
W , ñ(W ) 6= 0 then clear W is not tangent. But is the converse true,
~ that is not tangent but for
i.e. is it possible that there is a vector W
which ñ(W ~ ) = 0? To show that this cannot happen, we note that
the tangent space to the given plane x = 0 is two-dimensional, while
the whole space is three-dimensional. If W ~ is not tangent, then it is
linearly independent of any vector that is tangent. Therefore we can
form a basis of the entire three-dimensional spece using W ~ and any to
linearly-independent vectors that are tangent, say A ~ and B.~ But then,
~
by our assumption on W , it follows that the value of ñ on all three of
these basis vectors is zero, and therefore that its value on any vector
is zero: ñ can only be the null, or zero, one-form. But it was assumed
non-zero in the statement of the problem. So it follows that ñ(V ~)=0
if and only if V ~ is tangent. This proof would work for any number
of dimensions because a one-form can be “the” normal one-form to a
surface only if that surface has dimension one less than that of the
whole space.
~ and W
(b) If V ~ point to the same side of S then there exists a positive
number α such that W ~ = αV ~ + T~ , where T~ is tangent to S. Then
~ ~
ñ(W ) = αñ(V ) and both have the same sign.
(c) On the Cartesian basis, the components of ñ are (β, 0, 0) for some
β. Thus any ñ is a multiple of any other.

16 (e) 10 and 6 in four dimensions.

18 (b) ~q → (−1, −1, 1, 1).

20 (a) In matrix language, Λᾱ β Aβ is the product of the matrix Λᾱ β with
the column vector Aβ , while Λα β̄ pα is the product of the transpose of
Λα β̄ with the column vector pα . Since Λα β̄ is inverse to Λᾱ β , these are
the same transformation if Λα β̄ equals the transpose of its inverse.

21 (a) The associated vectors for t = 0 and t = 1 point inwards.

24 (b) No. The arguments are different geometrical objects (one vector,
one one-form) so they cannot be exchanged.

10
25 Use the inverse property of Λα β̄ and Λᾱ β .

26 (a) Aαβ Bαβ = −Aβα Bαβ = −Aβα Bβα = −Aµν Bµν = −Aαβ Bαβ .
Therefore 2Aαβ Bαβ = 0. The justification for each of the above steps:
antisymmetry, symmetry, relabel dummy indices, relabel dummy indices.
~.
28 Arbitrariness of U
~ ·D
30 (a) Since D ~ = −x2 + 25t2 x2 + 2t2 6= −1, D
~ is not a four-velocity
field.
(f) 5t.
(h) −5t because of (e).
(i) the vector gradient of ρ has components {ρ,α } = (−2t, 2x, −2y, 0).
~ → [t2 , 5t3 + 5x(1 + t2 ), √2(1 + t2 ), 0].
(j) ∇ ~ D
U

33 (d) Given any matrix (A) in 0(3), let (Λ) be the 4 × 4 matrix
 
1 0 0 0
0 
 .
0 (A) 
0

Show that this is in L(4) and that the product of any two such matrices
is one of the same type, so that they form a subgroup. These are
pure rotations of the spatial axes (relative velocities of the two frames
are zero). Transformations like Eq. (1.12) are pure boosts (spatial
axes aligned, relative velocity nonzero). The most general Lorentz
transformation involves both boost and rotation.

34 (c) guu = gvv = 0, guv = −1/2, guy = guz = gyz = 0, gyy = gzz = 1.
(e) d̃u = d̃t − d̃x, d̃v = d̃t + d̃x, g(~eu, ) = −d̃v/2, g(~ev, ) = −d̃u/2.
Notice that the basis dual to {~eu , ~ev }, which is {d̃u, d̃v}, is not the
same as the set of one-forms mapped by the metric from the basis
vectors.

11
 Chapter 4
1 (a) No. (b) Yes. (c) Dense traffic can indeed resemble a continuum,
and traffic congestion on highways often shows wave-like behavior,
with cars moving through a compression zone (slow speed) into a rar-
efaction zone (fast speed) and back into a compression zone. See, for
references, the Wikipedia article on “Traffic Flow”. (d) This is one
situation where cars are not a continuum: they move through the in-
tersection one by one. (e) If the plasma is dense enough then it is a
continuum, but very rarified plasmas, particularly in astronomy, can
be difficult to describe fully this way.

2 Particles contributing to this flux need not be moving exclusively in

the x direction. Moreover, consider a change to the non-orthogonal
coordinates (t, x, y 0 =: x + y, z). A surface of constant x is unchanged,
so the flux across it is unchanged, but the ‘x direction’, which is the
direction in which now t, y 0 , and z are constant, is in the old ~ex +
~ey direction. Is the unchanged flux now to be regarded as a flux in
this new direction as well? The loose language carries an implicit
assumption of orthogonality of the coordinates in it.

3 (a) In Galilean physics p changes when we change frames, but in

relativity p~ does not: only its components change.
(b) This is because the usual Galilean momentum is only a three-
vector. However, if in Galilean spacetime we define a four-vector
(m, p) then the Galilean transformation changes this to (m, p − mv),
where v is the relative velocity of the two frames. This is an approxi-
mation to the relativistic one (see Eq. (2.21)) in which terms of order
v 2 are neglected.
~ →
4 The required density is, by definition, N 0 in the frame in which U
~ ·U
(1, 0, 0, 0). In this frame N ~ = −N .
0

8 (a) Consider a two-dimensional space whose coordinates are, say, p

and T , each point of which represents an equilibrium state of the
fluid for that p and T . In such a space the dρ of Eq. (4.25) is just
~ where ∆
hd̃ρ, ∆i ~ is whatever vector points from the old state to the
new one, the change in state contemplated in Eq. (4.25). Since we
want Eq. (4.25) to hold for arbitrary ∆,~ it must hold in the one-
form version. See B.F. Schutz, Geometrical Methods of Mathematical
Physics (Cambridge University Press, Cambridge, 1980).

12
 ˜ = d̃q, then T ∂S/∂xi = ∂q/∂xi , where xi is either p or T . The
(b) If ∆q
identity ∂ 2 q/∂T ∂p = ∂ 2 q/∂p∂T implies (∂T /∂p)(∂S/∂T ) = ∂S/∂p,
which will almost never be true.

9 Follow the steps leading to Eq. 4.33 but changing the first index ‘0’
to, say, ’x’. Use the form of Newton’s law that says that the force is
the rate of change of the momentum. Interpret the terms analogous
to those on the left-hand-side of Eq. 4.30 as fluxes of momentum.

11 (a) By definition of ‘rotation’: see Exer. 20b, § 3.10.

(b) Suppose M has the property OT M O = M for any orthogonal
matrix O. Consider the special case of a rotation about x3 , where
O11 = cos θ, O12 = sin θ, O21 = − sin θ, O22 = cos θ, O33 = 1, all other
elements zero. Then OT M O = M for arbitrary θ implies M13 =
M23 = M31 = M32 = 0, M11 = M22 , and M21 = −M12 . By relabeling,
a rotation about x implies M12 = M21 = 0, M33 = M22 . Therefore M
is proportional to I.

13 Uα = ηαγ U γ ; ηαγ is constant, so ηαγ,β = 0; U α ,β U γ ηαγ = U γ ,β uα ηγα

(relabeling) = U γ ,β U α ηαγ (symmetry of η).

14 Since U~ →MCRF (1, 0, 0, 0), multiplying by Uα and summing on α

picks out the zero component in the MCRF.

16 There is no guarantee that the MCRF of one element is the same as

that of its neighbor.

20 (a) In Eq. 4.58, let V~ be N~ but now show that the expression does
not equal zero, but instead equals the integral of  over the four-
dimensional volume. Interpret the result as the difference between
the change in the number of particles and the number that have en-
tered over the boundaries. Show that this means that  is the rate of
creation of particles per unit volume per unit time.
(b) F 0 is the rate of generation of energy per unit volume, and F i is
the ith component of the force. F α is the only self-consistent general-
ization of the concept of force to relativity.
~ → γ(1, β, 0, 0), γ = (1 − β 2 )−1/2 .
21 (a) T αβ = ρ0 U α U β , U
(b) At any point on the ring, the particles have speed ωa. At position
(x, y) we have U α → γ(1, −ωy, ωx, 0), γ = (1 − ω 2 a2 )−1/2 . In the
inertial frame their number density is N [2π 2 a(δa)2 ]−1 , which equals

13
 nU 0 , where n is the number density in their rest frame. Therefore
n = N [2π 2 γa(δa)2 ]−1 and T αβ = mnU α U β .
(c) Add (b) to itself with ω → −ω. For example, T 0x = 0 and T xx =
2mnω 2 y 2 γ 2 .
22 No bias means that T ij is invariant under rotations. By Exer. 11
T ij = pδ ij for some p. Since T 0i = 0 in the MCRF, Eq. (4.36) holds.
Clearly ρ = γnm, where γ = (1 − v 2 )−1/2 . The contributions of each
particle to T zz , say, will be the momentum flux it represents. For a
particle with speed v in the direction (θ, φ), this is a z component of
momentum mγv cos θ carried across a z = const. surface at a speed
v cos θ. If there are n particles per unit volume, with random velocities,
then T zz = n(mγv)(v) times the average value of cos2 θ over the unit
sphere. This is 31 , so T zz = p = γnmv 2 /3. Thus, p/ρ = v 2 /3 → 13 as
v → 1. (In this limit mγ remains finite, the energy of each photon.)
23 This exercise prepares us for computing the quadrupole radiation of
gravitational waves in Ch. 9. Note that, although the stress-energy
tensor’s components vanish outside a bounded region of space, the
domain of integration in the spatial integrals in this exercise can still
be taken to be all of space: since the divergence vanishes outside the
system, integrating over the exterior does not add anything. This
seemingly small detail will be used in all parts of this exercise.
(a) The integral is over spatial variables, so the partial derivative with
respect to time may be brought inside, where it operates only on T 0α .
Then use the identity T µν ,ν = 0 to replace T 0α ,0 by −T jα ,j . This
is a spatial divergence, so its integral over d3 x converts to a surface
integral, by Gauss’ law in three dimensions. Since the region of in-
tegration in space is unbounded, this surface can be taken anywhere
outside the domain of the body, where the stress-energy components
all vanish. That means that the surface integral vanishes, and this
proves the result.
(b) The computation follows that in (a) closely. First bring one time-
derivative inside the integral, replace T 00 ,0 by −T 0k ,k , and this time
integrate by parts on xk . The integrated term is a full divergence,
and vanishes as above. But the integration by parts leaves another
term involving ∂(xi xj )/∂xk . Since partial derivatives of one coordinate
with respect to another are zero, the only terms that survive this
differentiation are where the indices i or j equal k. We can write
the result using Kronecker deltas: ∂(xi xj )/∂xk = δ i k xj + δ j k xi . The

14
 integral involves this expression multiplied by T 0k and summed on k.
This summation leaves a relatively simple integrand: −T 0i xj − T 0j xi .
But this integral still has one further time-derivative outside it. Bring
this one in now, let it operate on the components of T µν (it does
not affect the coordinates xi and xj ), and again replace these time-
derivatives by the spatial divergences. Again integrate by parts and
evaluate the full divergence on a surface outside the body. There will
be more Kronecker deltas, but when summed they will give the simple
result.
(c) This is a variation on the procedure in (b). Follow the same steps,
but there are now more factors of xk so the result will be the one given
in the problem.

24 (c) R/x = [(1 − v)/(1 + v)]1/2 .

25 (h) E x̄ = E x ; let E y e y +E z e z be called E⊥ , the part of E perpendicu-

lar to v . Similarly, let E 0⊥ = E ȳ e ν̄ +E z̄ e z̄ . Then E 0⊥ = γ(E ⊥ +v ×B).

15
 Chapter 5
3 (b) (i) Good except at origin x = y = 0: usual polar coordinates. (ii)
Undefined for x < 0, fails at x = 0, good for x > 0: maps the right-
hand plane of (x, y) onto the whole plane of (ξ, η). (iii) Good except
at origin and infinity: an inversion of the plane through the unit circle.

4 A vector has a slope which is the ratio of its y- and x-components. For
the given vector this is (dy/dλ)/(dx/dλ) = dy/dx, which is the slope
of the curve.

5 (a) and (b) have same path, the unit circle x2 + y 2 = 1. But their
tangent vectors are different even at the same point, because the
parametrization is different. Additionally, for this problem the points
λ = 0 in (a) and t = 0 in (b) are different.

6 Recall the solution to Exer. 3.8.

7
0
Λ1 1 = x/r = cos θ;
20
Λ 1 = −y/r2 = − 1r sin θ;
0
Λ1 2 = y/r = sin θ;
20
Λ 2 = x/r2 = 1
r cos θ;
1
Λ 10 = x/r = cos θ;
Λ 2 10 = y/r = sin θ;
1
Λ 20 = −y = −r sin θ;
Λ 2 20 = x = r cos θ.

8 (a)

f = r2 (1 + sin 2θ)
V r = r2 (cos3 θ + sin3 θ) + 6r(sin θ cos θ)
V θ = r sin θ cos θ(sin θ − cos θ) + 3(cos2 θ − sin2 θ)
W r = cos θ + sin θ
W θ = (cos θ − sin θ)/r

16
 (b)

(d̃f )x = 2x + 2y; (d̃f )y = 2x + 2y

(d̃f )r = 2r(1 + sin 2θ) = ∂f /∂r
(d̃f )θ = 2r2 cos 2θ = Λ1 20 (d̃f )x + Λ2 20 (d̃f )y

(c)(i)

(Ṽ )r = V r = r2 (cos3 θ + sin3 θ) + 6r(sin θ cos θ)

(Ṽ )θ = r2 V θ = r3 sin θ cos θ(sin θ − cos θ) + 3r2 (cos2 θ − sin2 θ)
(W̃ )r = W r = cos θ + sin θ
(W̃ )θ = r2 W θ = r(cos θ − sin θ)

(c)(ii) Same result by a different method, e.g.:

(W̃ )r = Λ1 10 (W̃ )x + Λ2 10 (W̃ )y = cos θ + sin θ

10 The key thing is to prove linearity. The lower index, associated with
the derivative, is linear in whatever vector we give it, just as is the
derivative of a scalar function. So if we double the vector argument,
we get a derivative twice as large, since it has to approximate the
change in the function when we go twice as far. Similarly, the one-
form argument is linear (associated with the upper index) because of
Eq. 5.52.

11 (a) The Christoffel symbols are zero in Cartesian coordinates so the

result is:

V x ,x = V x ;x = ∂V x /∂x = 2x; V x ,y = 3; V y ,x = 3; V y ,y = 2y.

(b) Although it is possible to do this using matrices, the straightfor-

ward expansion of the summations in the transformation equation is
less error prone. Thus, the r − r component of ∇~v is
0
V r ;r = Λ1 α Λβ 10 V α ,β
0 0 0 0
= Λ1 1 Λ1 10 V 1 ,1 + Λ1 2 Λ1 10 V 2 ,1 + Λ1 1 Λ2 10 V 1 ,2 + Λ1 2 Λ2 10 V 2 ,2
= 2r(cos3 θ + sin3 θ) + 6 sin θ cos θ.

17
 Other components are:

V r ;θ = 2r2 sin θ cos θ(sin θ − cos θ) + 3r(cos2 θ − sin2 θ)

V θ ;r = 2 sin θ cos θ(sin θ − cos θ) + 3(cos2 θ − sin2 θ)/r
V θ ;θ = 2r sin θ cos θ(sin θ + cos θ) − 6 sin θ cos θ

(c) This gives the same as (b).

(d) 2(x + y).
(e) 2r(sin θ + cos θ), same as (d).
(f) Same as (d).

12 (a) Same components as in Exer. 11a.

(b),(c) These components are the same for both (b) and (c) and are
related to the answers given for Exer. 11c as follows: pr;r = V r ;r , pr;θ =
V r ;θ , pθ;r = r2 V θ ;r , pθ;θ = r2 V θ ;θ . It happens that pθ;r = pr;θ for this
one-form field. This is not generally true, but happens in this case
because p̃ is the gradient of a function.

14 Two selected results: Arθ ;θ = r(1 + cos θ − tan θ); Arr ;r = 2r.

15 Of the first-derivative components, only V θ ;θ = 1/r is nonzero. Of

the second-derivative components, the only nonzero ones are V θ ;r;θ =
−1/r2 , V r ;θ;θ = −1, and V θ ;θ;r = −1/r2 . Note that this vector field is
just the unit radial vector of polar coordinates.

17
 
∂~eµ0 ∂ ∂~eα
ν 0 = β
(Λ µ0 ~eα ) Λβ ν 0 = Λα µ0 Λβ ν 0 β + Λα µ0 ,β Λβ ν 0 ~eα
α
∂x ∂x ∂x
0 0 0
⇒ Γλ µ0 ν 0 = Λα µ0 Λβ ν 0 Λλ γ Γγ αβ + Λα µ0 ,β Λβ ν 0 Λλ α

20 Γν αβ = 21 g νµ (gµα,β + gµβ,α − gαβ,µ + cαµβ + cβµα − cµαβ ).

21 (a) Compute the vectors (dt/dλ, dx/dλ) and (dt/da, dx/da) and show
they are orthogonal.
(b) For arbitrary a and λ, x and t obey the restriction |x| > |t|. The
lines {x > 0, t = ±x} are the limit of the λ = const. hyperbolae as
a → 0+ , but for any finite λ the limit a → 0+ takes an event to the
origin x = t = 0. To reach x = t = 1, for example, one can set
λ = ln(2/a), which sends λ → ∞ as a → 0.

18
 (c) gλλ = −a2 , gaa = 1, gaλ = 0, Γλaλ = 1/a, Γa λλ = a, all other
Christoffel symbols zero. Note the close analogy to Eqs. (5.3), (5.31),
and (5.45). Note also that gλλ (the time-time component of the metric)
vanishes on the null lines |x| = |t|, another property we will see again
when we study black holes.

22 Use Eq. (5.68).

19
 Chapter 6
1 (a) Yes, singular points depend on the system.
(b) Yes: if we define r = (x2 + y 2 )1/2 , the map {X = (x/r) tan(πr/2),
Y = (y/r) tan(πr/2)} shows that, as a manifold, the interior of the
unit circle (r < 1) is indistinguishable from the whole plane (X, Y
arbitrary). This map distorts distances, but the metric is not part of
the definition of the manifold.
(c) No, discrete.
(d) This consists of the unit circle and the coordinate axes. It has the
structure of a one-dimensional manifold everywhere except at the five
intersection points, such as (1, 0).

2 (a) Normally no metric is used on this manifold: since the axes repre-
sent physically different quantities, with different units, no combina-
tion like p2 + q 2 is normally useful or meaningful.
(b) The usual Euclidean metric is normally used here.
(d) On the different segments the one-dimensional Euclidean metric
(length) is used.

4 (a) Symmetry on (γ 0 , µ0 ) means there are 12 n(n + 1) = ten independent

pairs (γ 0 , µ0 ). Since α can assume four values independently, there are
40 coefficients.
(b) The number of symmetric combinations (λ0 , γ 0 , µ0 ) is 16 n(n + 1) ×
(n + 2) = 20, times four for α, gives 80.
(c) Two symmetric pairs of ten independent combinations each give 100.

5 Carefully repeat the flat-space argument in a local inertial frame.

6 g αβ gβµ,α = g βα gβµ,α (symmetry of metric) = g αβ gαµ,β (relabeling

dummy indices) = g αβ gµα,β (symmetry of metric). This cancels the
second term in brackets.

9 For polar coordinates in two-dimensional Euclidean space, g = r2 .

In three dimensions it is g = r4 sin2 θ. Notice in these cases g > 0,
so formulae like Eqs. (6.40) and (6.42), which are derived from Eq.
(6.39), should have −g replaced by g.

10 The vector maintains the angle it makes with the side of the triangle
as it moves along. On going around a corner the angle with the new

20
 side exceeds that with the old by an amount which depends only on
the interior angle at that corner. Summing these changes gives the
result.

15 Consider a spacelike curve {xα (λ)} parametrized by λ, and going from

Rb
xα (a) to xα (b). Its length is a [gαβ (dxα /dλ)(dxβ /dλ)]1/2 /dλ. Assume
that λ is chosen so that the integrand is constant along the given curve.
Now change the curve to {xα (λ) + δxα (λ)} with δxα (a) = δxα (b) = 0.
The
R b 1 first-order change in the length is, after an integration by parts,
[ g U α U β − d(g U α )/dλ]δxγ dλ, where U α ≡ dxα /dλ. For a
a 2 αβ,γ γα
geodesic, the term in square brackets vanishes.

18 (b) Each pair αβ or µν is antisymmetric, so has six independent com-

binations for which the component need not vanish. Each pair can
be chosen independently from among these six, but the component
is symmetric under the exchange of one pair with the other. (c) Eq.
(6.69) allows us to write Eq. (6.70) as Rα[βµν] = 0. There are only
n(n − 1) × (n − 2)/6 = four independent choices for the combination
βµν, by antisymmetry. In principle α is independent, but if α equals
any one of β, µ, and ν then Eq. (6.70) reduces to one of Eq. (6.69).
So Eq. (6.70) is at most four equations determined only by, say, the
value α. However, Eq. (6.69) allows us also to change Eq. (6.70) to
Rβ[αµν] = 0. Thus, if we had earlier taken, say, α = 1 and β = 2, then
that equation would have been equivalent to the one with α = 2 and
β = 1: all values of α give the same equation.

25 Use Eqs. (6.69) and (6.70).

28 (a) We desire the components gαβ in the new (spherical) coordinates,

so we need to compute derivatives of the Cartesian with respect to the
spherical, e.g.
∂x/∂r = sin θ cos φ,
and so on. Then we compute the new components from the transfor-
mation law, e.g.
∂x ∂x ∂x ∂y
grθ = gxx + gxy + . . . ,
∂r ∂θ ∂r ∂θ
summing over all components of the metric in Cartesian coordinates.
Since most of the components are zero the algebra is not very long.

21
 (b) If one moves on the surface of the sphere then one is at constant
r, so that dr = 0. With this in the line element, only the second and
third rows and columns of the matrix are relevant.
(c) g θθ = r−2 , g φφ = (r sin θ)−2 , g θφ = 0.

29 Rθφθφ = sin2 θ.

30 The most sensible coordinates to use are Euclidean: unwrap the cylin-
der so that it lies flat and put a flat coordinate system on it. To make
a cylinder you just have to remember that when you come to the edge
of the paper you jump to the other side, where the join was. But
locally, this makes no difference to the geometry. Since the metric is
Euclidean, its derivatives all vanish, and the Riemann tensor vanishes.

32 Compare with Exer. 34, § 3.10.

35 The nonvanishing algebraically independent Christoffel symbols are:

Γt tr = Φ0 , Γr tt = −Φ0 exp(2Φ − 2Λ), Γr rr = Λ0 , Γr θθ = −r exp(−2Λ),
Γr φφ = −r sin2 θ exp(−2Λ), Γθ rθ = Γφ r̄φ = r−1 , Γθ φφ = − sin θ cos θ,
Γφ θφ = cot θ. Here primes denote r derivatives. (Compare these with
the Christoffel symbols you calculated in Exer. 29.) As explained in
the solution to Exer. 18, in calculating Rαβµν we should concentrate
on the pairs (αβ) and (µν), choosing each from the six possibilities
(tr, tθ, tφ, rθ, rφ, θφ). Because Rαβµν = Rµναβ , we do not need to
calculate, say, Rθφrφ after having calculated Rrφθφ . This gives 21 in-
dependent components. Again following Exer. 18, one of the com-
ponents with four distinct indices, say Rtrθφ , can be calculated from
others. We catalog, therefore, the following 20 algebraically indepen-

22
 dent components:
00
Rtrtr = [Φ − (Φ0 )2 − Φ0 Λ0 ] exp(2Φ),
Rtrtθ = Rtrtφ = Rtrrθ = Rtrrφ = 0,
Rtθtθ = −r Φ0 exp(2Φ − 2Λ),
Rtθtφ = Rtθrθ = Rtθrφ = Rtθθφ = 0,
Rtφtφ = −r Φ0 sin2 θ exp(2Φ − 2Λ),
Rtφrθ = Rtφrφ = Rtφθφ = 0,
Rrθrθ = rΛ0 ,
Rrθrφ = Rrθθφ = Rrφθφ = 0,
Rrφrφ = r sin2 θΛ0 ,
Rθφθφ = r−2 sin2 θ[1 − exp(−2Λ)].

See if you can use the spherical symmetry and time independence of
the metric to explain why certain of these components vanish. Also
compare Rθφθφ with the answer to Exer.29 and see if you can explain
why they are different.

36 Since φ = 0 is already inertial, we can look for a coordinate trans-

0
formation of the form xα = (δ α β + Lα β )xβ , where Lα β is of order φ.
0
The solution to Exer. 17, § 5.9, gives Γλ µ0 ν 0 , which must vanish at P .
0
Since Λα β = δ α β + Lα β + Lα µ,β xµ , we find Lα (β,ν) = 21 Γα βν at P . The
antisymmetric part, Lα [β,ν] , is undetermined, and represents a Lorentz
transformation of order φ. Since we are only looking for an inertial
system, we can set Lα [β,ν] = 0. Calculating Γα βν at P (as in Exer. 3,
§ 7.6, below) gives the new coordinates. In particular, the equation
0
xi = 0 gives the motion of the origin of the new frame, whose acceler-
ation is d2 xi /dt2 = −Γi tt = −φ,i . We shall interpret this in the next
chapter, where we identify φ as the Newtonian gravitational potential
and see that this acceleration expresses the equivalence principle.

37 (b) The ranges of the coordinates must be deduced: 0 < χ < π, 0 <
θ < π, 0 < φ < 2π. Then the volume is 2π 2 r3 .

38 2πr sin θ.

23
 Chapter 7
1 Consider a fluid at rest, where U i = 0 and U 0 = 1 in a local inertial
frame. Then ∂n/∂t = R, so Eq. (7.3) implies creation (or destruction)
of particles by the curvature.

2 g 00 = −(1 − 2φ), g ij = δ ij (1 + 2φ).

3 Γ0 00 = φ̇, Γ0 0i = φ,i , Γ0 ij = −φ̇δij , Γi 00 = φ,i , Γi 0j = −φ̇δij ,

Γi jk = δjk φ,i − δij φ,k − δik φ,j .

7 (a)(i) In the Minkowski metric, all metric components are indepen-

dent of the given coordinates (t, x, y, z), so the momentum components
(pt , px , py , pz ) are conserved.
(a)(ii) The Schwarzschild metric components are independent of time
t, so the component pt is conserved. Also, the metric does not depend
on the coordinate φ, so that the associated momentum pφ is conserved.
In part (b)(ii) below we will see that this is an angular momentum.
(a)(iii) The Kerr metric is a somewhat more complicated generalization
of Schwarzschild. Like Schwarzschild, however, the components are all
independent of the two coordinates t and φ. Therefore there are two
conserved momentum components, pt and pφ .
(a)(iv) The Robertson-Walker metric does depend on time t, but like
the previous two it is independent of the coordinate φ. Therefore pφ
is conserved.
(b)(i) This coordinate transformation is to spherical coordinates, and
it produces a metric whose components are independent of the angle
φ. This means that pφ is conserved, the angular momentum about
the axis of our coordinates. But this axis could have been chosen
to point in any direction, so there are three independent conserved
components of angular momentum, usually referred to as (Jx , Jy , Jz ).
This exhausts the set of conserved quantities.
(b)(ii) Now we can see that there are more conserved quantities for
Schwarzschild. The metric is written in coordinates that look like the
spherical coordinates of Minkowski spacetime that we have just de-
rived, and indeed the way it depends on the angles θ and φ is identical
to Minkowski spacetime in spherical coordinates. This means that the
metric is in fact spherically symmetric. Now, since no metric compo-
nent depends on φ, then certainly pφ is conserved. But we can also
imagine performing exactly the kinds of coordinate transformations

24
 (rotations) on these coordinates as one would perform in Minkowski
spacetime if one wanted to re-orient the symmetry axis of the coordi-
nates in a different direction, and that would again produce a metric
independent of the new angle φ. So, just as in flat spacetime, there
are actually three conserved quantities associated with angles, three
angular momentum components. There are no other conservation laws
for Schwarzschild.
(b)(iii) There is no spherical symmetry for the Kerr metric, so we do
not find any further conserved quantities.
(b)(iv) The Robertson-Walker metric is spherically symmetric in the
same way as Schwarzschild and Minkowski are. Therefore there are
three conserved angular momentum components. But linear momen-
tum is not conserved.

9 (a)

R0i0j = φ,ij + δij φ,00 ,

R0ijk = φ,0j δik − φ,0k δij ,
Rijkl = δik φ,jl + δjl φ,ik − δjl φ,ik − δjk φ,il .

(c) The ‘acceleration’, in Newtonian language, is −φ,i . The differ-

ence between the accelerations of nearby particles separated by ξ j is
therefore −ξ j φ.ij .

10 (b) In terms of a Lorentz basis, the following vector fields are Killing
fields: ~et , ~ex , ~ey , ~ez , x~ey −y~ex , y~ez −z~ey , z~ex −x~ez , t~ex +x~et , t~ey +y~et , t~ez +
z~et . Any linear combination with constant coefficients of solutions to
Eq. (7.45) is a solution to Eq. (7.45), which means that the analogous
fields to these in another Lorentz frame are derivable from these.

25
 Chapter 8
2 (c) (i) 7.425×10−11 m−2 ; (ii) 8.261×10−12 m−2 ; (iii) 1.090×10−16 m−1 ;
(iv) 2.756 × 10−12 .
(d) m PL = 2.176×10−8 kg; tPL = 5.390×10−44 s. Typical elementary-
particle lifetimes are 10−24 s or greater. The heaviest known particles
are less than 10−23 kg.

3 (a)(i) −2.122 × 10−6 ; (ii) −9.873 × 10−9 ; (iii) −6.961 × 10−10 ; (iv)
9.936 × 10−5 .
(b) We and everything on Earth are in free-fall around the Sun, so the
value of the Sun’s potential does not matter, nor even the value of its
gradient (the gravitational acceleration the Sun produces on us). What
we do experience is the tidal force of the Sun, which is comparable in
size to that of the Moon, and which raises ocean tides and distorts
Earth’s shape. The tidal forces arise in the second derivatives of the
Sun’s potential, i.e. the differences in the Sun’s acceleration across
Earth. We experience Earth’s gravity, even though its potential is
much smaller than the Sun’s, because we are not in free-fall on Earth.
The forces we feel from our weight are actually the forces of the ground
or floor pushing up on us, stopping us from falling.

5 (a) This is an exercise in “index gymnastics”. Make sure that you

use all possible symmetries, in this case the freedom to exchange the
order of partial derivatives. (b) Since the expression is linear, a gauge
transformation is just the same as adding a pure-gauge value, and we
have just shown that this vanishes. Therefore the Riemann tensor at
linear order is gauge-invariant.

9 (b) No contradiction because this expression was derived by applying

the gauge condition Eq. 8.33, as is computed explicitly in the next
exercise, Ex. 10. One could use this condition to remove the time-
derivatives from the time-components of Eq. 8.42, at the expense of
making them more complicated to write down. The important thing
is that the gauge condition Eq. 8.33 creates a relationship among the
differential equations in Eq. 8.42, so that there are still only 6 inde-
pendent equations with second-time derivatives.

11 Gauge transformation: Aµ → Aµ + f,µ . Lorentz gauge condition:

Aµ ,µ = 0. These are very similar to Eq. 8.24 and Eq. 8.33. The dif-
ferences are all to do with the extra index needed in general relativity

26
 because our fundamental field is a tensor (the metric) rather than a
vector.
13 The ratio of momentum to mass is velocity, so the ratio of the corre-
sponding densities, T 0i /T 00 , will be of order v, where v is a typical
velocity. In the low-velocity limit this will be small. The argument for
the stresses is similar but more subtle. In a fluid the stress components
are just pressures, and when one uses statistical mechanics to derive
pressure from the random motions of particles of the gas, one finds
p ∼ ρv 2 . Therefore in this case T ij /T 0i  1. In more general mate-
rials, the stresses can be made of direct forces between particles, but
even here they must be of the same size: too great a stress would cre-
ate forces that would accelerate a body to relativistic speeds, violating
the low-velocity assumption.
17 (a) M = Rv 2 = (C 3 /2πP 2 ) ∼ 1.3 km ∼ 1 M . (b) 100 km ∼ 68 M .
18 (a)|Λ| . 4 × 10−35 m−2 . (b) This value of Λ represents a mass density
that one would get by spreading the mass of the Sun over the whole
Solar System. This is very much bigger than the average density of
the universe, since the nearest star is much further away than Pluto,
and on top of that there are vast empty spaces between galaxies. So
a cosmological constant of this size would have enormous and easily
observable consequences for the evolution of the universe. It follows
that a Λ of the same size as the cosmological mass density would not
be observable in the Solar System.
19 (a) T 00 = ρ, T 01 = −ρΩx2 , T 02 = ρΩx1 , T 03 = 0. The components
T ij are not fully determined by the given information, but they must
be of order ρv i v j , i.e. of order ρΩ2 R2 .
(b) Since ∇2 h̄00 = −16πρ, h̄00 is just minus four times the Newtonian
potential, h̄00 = 4M/r exactly. For h̄0i we have
Z
h̄ = −4ρΩ y 2 |x − y|−1 d3 y.
0i

Use the binomial expansion

|x − y|−1 = r−1 [1 + x · y /r2 + 0(R/r)2 ].
By symmetry,
Z Z
i 3
y d y = 0, y i y j d3 y = 0 if i 6= j, and

27
 Z Z Z
1 1 3 2 2 3
y y d y= y y d y= y 3 y 3 d3 y = (4π/15)R5 .

This implies h̄01 = −(16π/15)ρR5 Ωx2 /r3 . In terms of the angular

momentum J, we find

h̄01 = −2Jx2 /r3 , h̄02 = 2Jx1 /r3 , h̄03 = 0.

These fall off as r−2 and are correct to order r−3 . A more careful study
of the properties of the solutions of ∇2 f = g would show that these are
in fact exact: the higher-order terms all vanish in this simple situation.
The components h̄ij are small compared to h̄00 and h̄0i because T ij is
small. Therefore Eq. (8.31) gives hµν and the metric

g00 = −1 + 2M/r + 0(Ω2 R2 ),

g01 = 2Jx2 /r3 + 0(Ω3 R3 ), g02 = −2Jx1 /r3 + 0(Ω3 R3 ), g03 = 0,
2 2
gij = δij (1 + 2M/r) + 0(Ω R ).

Compare this with Eq. (7.8). In standard spherical coordinates, g00

is the same, g0r = g0θ = 0, g0φ = −2J sin2 θ/r, and the spatial line
element is

dl2 = [1 + 2M/r − 0(Ω2 R2 )](dr2 + r2 dθ2 + r2 sin2 θ dφ2 ).

(c) Such a particle obeys the geodesic equation with p0 := −E and

pφ := L constant, pr and pθ zero. To this order, the normaliza-
tion p~ · p~ = −m2 implies E = m(1 − M/r + L2 /2m2 r2 ), just as in
Newtonian theory (except for the rest mass). The r component of
the geodesic equation implies, again to lowest order, L = m(M r)1/2 ,
again as in Newtonian theory. One orbit, ∆φ = 2π, will take a time
∆t = (dt/dφ)∆φ. Now dt/dφ = (dt/dτ )/(dφ/dτ ) = U 0 /U φ = p0 /pφ ,
and this can be expressed in terms of E, L, and the metric; a straight-
forward calculation gives (∆t)prograde − (∆t)retrograde = −8πJ/M , in-
dependently of r. In principle, this allows measurement of a body’s
angular momentum by the study of particle orbits far from it.
(d) 0.16 ms.

28
 Chapter 9
7 A solution of Eq. (9.22) is uniquely determined by the initial position
and U α . The function U α = δ α 0 satisfies Eq. (9.22) for all time (by
virtue of Eq. (9.23)), and so must be the unique solution for initial
data in which U α = δ α 0 .

8 No: the equivalence principle.

9 For the light beam, ds2 = 0 ⇒ dt/dx = (gxx /|gtt |)1/2 ≈ 1 + 12 hTxxT (t).
Therefore, if the remote particle is at coordinate location x = , the
time elapsed for a round trip of light is ∆t ≈ (2+hhTT TT
xx i)ε, where hhxx i
TT
is some mean value of hxx during the time of flight of the photon.
Since hhTT
xx i changes with time while ε does not, free particles do see
accelerations relative to their neighbors.

10 (a) For example, Rtyxz = − 12 ω 2 hyz ; (d) ξt = − 21 B(x − t)2 , ξi = 0.

13 One way to do this is with rotation matrices. But consider a more

direct way. Write down the new basis vectors in terms of the old:
√ √
~ex0 = (~ex + ~ey )/ 2, ~ey0 = (~ey − ~ex )/ 2.

The components of the tensor h T T are its values on the basis vec-
tors, so its components in the new frame are, for example, hTx0Tx0 =
h T T (~ex0 , ~ex0 ). Putting the expressions for the new basis vectors into
this gives hTx0Tx0 = hTxyT + (hTxxT + hTyyT )/2. The vanishing trace gives the
desired result.

20 In the TT coordinates of the wave, choose the wave’s direction to be z,

the ellipse’s principal axes to be x and y, and the masses’ separation to
be along the unit vector s → (sin θ cos φ, sin θ sin φ, cos θ) in the usual
spherical coordinates of the frame. Then in Eq. (9.45) we replace hTT xx
2
by hTT α β TT TT
ss ≡ s s hαβ = sin θ(cos 2φ + ia sin 2φ)hxx . There will be no
driving term if θ = 0, i.e. if the masses lie along the direction of the
wave’s propagation (compare with Exer. 9.12).

26 (a) Does not violate relativity: this is just a coordinate speed, de-
pendent on the coordinate system. Relativity insists that the proper
distance divided by proper time along a light ray should always be 1,
but there is no constraint on the coordinate speed. (c) The statement
is correct as long as the distance is small enough that there is a local
inertial frame covering both ends of the light path. If the distance is

29
 larger, and the geometry is time-dependent, then there is not a unique
meaning to the distance to a remote object: it depends on time and
the path being measured. (e) This is a coordinate-independent result,
since it is a proper time as measured on a single clock. So it would be
the same in any coordinate system.
30 I lm = A m(A) xi(A) xi(A) .
P

31 (a) I ij = (4π/3)δ ij ρr4 dr; I–ij = 0.

R

(b) I ij =: Result of (a) + M ai aj ; –I ij = M (ai aj − 1

3 δ ij a2 ).
(c)
I xx = a2 M/5, I yy = b2 M/5, I zz = c2 M/5,
I xx = (2a2 − b2 − c2 )M/15,
–
I yy = (2b2 − a2 − c2 )M/15,
–
I zz = (2c2 − a2 − b2 )M/15,
–
all other components zero.
(d)
I xx = (a2 cos2 ωt + b2 sin2 ωt)M/5,
I yy = (b2 cos2 ωt + a2 sin2 ωt)M/5,
I zz = c2 M/5,
I xy = cos ωt sin ωt(a2 − b2 )M/5,
I xx = [a2 (3 cos2 ωt − 1) + b2 (3 sin2 ωt − 1) − c2 ]M/15,
–
I yy = [b2 (3 cos2 ωt − 1) + a2 (3 sin2 ωt − 1) − c2 ]M/15,
–
I zz = (2c2 − a2 − b2 )M/15,
–
I xy = I xy ,
–
others zero.
(e) I xx = 2ma2 = I yy , others zero;
I–xx = 2ma2 /3 = I–yy , –I zz = −4ma2 /3, others zero.
(f) Same as (e).
(g) I xx = 2m(A2 cos2 ωt + Al0 cos ωt + l02 /4), others zero;
I–xx = 2I xx /3, –I yy = −I xx /3 = I–zz , others zero.
(h) I xx = (m + M )(A2 cos2 ωt + Al0 cos ωt + l02 /4), other zero;
I–xx = 2I xx /3, –I yy = −I xx /3 = I–zz , and ω 2 = k/µ, where µ =
mM/(M + m) is the reduced mass.

30
35 There is no radiation in (a)–(c) and (e), and no quadrupole radiation
in (f).
For (d), first put the time-dependence of I–ij from Exer. 31 into complex
1
form, e.g., I–xx = 10 M (a2 − b2 ) exp(−2iωt). Then use Eqs. (9.84)–
(9.86) with Ω = 2ω and the correct permutations of the various indices.
Results:

Along x-axis: h̄TT TT 2 2 2 2

zz = −h̄yy = − 5 M ω (a − b ) exp[−2iω(t − r)]/r,
h̄TT
zy = 0.
Along y-axis: h̄TT TT 2 2 2 2
zz = −h̄xx = 5 M ω (a − b ) exp[−2iω(t − r)]/r,
h̄TT
xz = 0.
Along z-axis: h̄TT TT 4 2 2 2
xx = −h̄yy = − 5 M ω (a − b ) exp[−2iω(t − r)]/r,
h̄TT TT
xy = ih̄xx .

This means that the radiation is linearly polarized in the equatorial

plane, circularly polarized along the z-axis. Notice there is no radiation
if the body is axially symmetric, i.e. if a = b.
For (h), there is no radiation on the x-axis. On the y-axis:

h̄TT 2 2
zz = −(m + M )ω {2A exp[−2iω(t − r)] + Al0 exp[−iω(t − r)]}/r,
h̄TT TT
yy = −h̄zz ,
h̄TT
xy = 0.

36 Without loss of generality choose the wave to be moving in the x − y

plane. The unit vector along its direction has components nx =
cos θ, ny = sin θ.Then Pxx = sin2 θ, Pyy = cos2 θ, Pzz = 1, Pxy =
− sin θ cos θ, others zero. By matrix multiplication or by just writ-
2
ing out all the summation terms, we find h̄TT TT
xx = f sin θ, h̄yy =
2 TT TT
f cos θ, h̄xy = −f sin θ cos θ, h̄zz = −f , others zero, where f =
mω 2 sin2 θ{2A2 exp[−2iω(t − r)] + Al0 exp[−iω(t − r)]}/r. If we let
li be (− sin θ, cos θ, 0), then ¯l and ~ez are two orthogonal vectors per-
pendicular to the motion of the wave. We find h̄TT i j TT
ll := l l h̄ij = f .
Recalling h̄TT
zz = −f , we see that the wave is always 100% linearly
polarized with the ellipse’s axes in the x − y plane and parallel to ~ez :
the component rotated by 45◦ is absent.

38 Let the wave be traveling in the x − z plane at an angle θ with the

z axis. Let ~ey and ~l → (cos θ, 0, − sin θ) be orthogonal vectors in the

31
 1 2
plane of polarization. Then h̄TT TT i TT
yy = (1 − 2 sin θ)f, h̄ly := l h̄iy =
−i cos θf, h̄TT
ll := li lj h̄TT TT 2 2
ij = −h̄yy , where f = 2ml0 ω exp[−2iω(t −
r)]/r. From Exer. 9.15b we see that the wave is elliptically polarized
with principal axes ~l and ~ey . The percentage of circular polariza-
tion is |h̄TT TT 2
ly /h̄yy | , which ranges from zero at the equator to 100% at
the poles.

39 (a) Let the stars orbit in the x-y plane, with the center of mass at
the origin. If the polar coordinates of m are (r1 , θ), both functions of
time, and if those of M are (r2 , θ + π), then their total separation is
r = r1 + r2 , with r1 /r = M/(m + M ), r2 /r = m/(m + M ). In terms
of these variables the quadrupole tensor follows straightforwardly:

I xx = µr2 cos2 θ,
I yy = µr2 sin2 θ,
I xy = I yx = µr2 sin θ cos θ,

with µ = mM/(m+M ), which is called the reduced mass of the binary

system.
Let the reduced orbit have eccentricity e and semi-major axis a. The
Newtonian equations imply
a(1 − e2 )
r(t) = ,
1 + e cos[θ(t) − θ0 ]
where θ0 is the orientation of the major axis of the ellipse, the direc-
tion of their separation when they are closest. The distance of closest
approach is the minimum value of r,

l0 = a(1 − e).

Similarly, the angular velocity of the orbit is

dθ
Ω(t) = = r(t)−2 [(m + M )a(1 − e2 )]1/2 .
dt
The orbital period is
1/2
a3

P = 2π .
m+M
The total energy E and angular momentum L, being conserved, can
be computed at any point; the point of closest approach θ = θ0 is

32
 convenient. The results are E = −mM/2a, L2 = µ2 (m + M )a(1 − e2 ).
From these follow the desired relations:

P 2 = −π 2 (M + m)2 µ3 /2E 3 ,
e2 = 1 + 2EL2 /(M + m)2 µ3 ,
a = −mM/2E,
l0 = a(1 − e).

(Recall that E < 0.)

(b) The trace of I jk is µr2 , which is not a constant (unlike the case of
circular orbits). The components of I–jk are:

I xx = µr2 (cos2 θ − 31 ),
–
I yy = µr2 (sin2 θ − 31 ),
–
–I xy = I–yx = µr2 sin θ cos θ,
–I zz = − 13 µr2 .

(c) The radiation requires two time-derivatives of –I jk .

40 Bring out the R−1 as r−1 , as in Eq. (9.103), but expand t − R about
t−r. Eq. (9.104) and its consequences still follow. The first term is Eq.
(9.105) but higher terms depend on (d/dt)n ρyi yk . . . yl ym d3 y, where
R

there are n factors of yj . If ρ is spherical, then the integrations give in-

dices of the form δjk · · · δlm (see Exer. 42) and permutations. The TT
projection eliminates traces and so completely eliminates these terms.

33
 Chapter 10
1 This is essentially a repeat of Exer. 28, except with the time-dimension
present. However, the transformation affects only the spatial coordi-
nates, so the solution follows in simple way.

3 The factor e−Φ in Eq. 10.11 expands in the weak-field approximation

to 1 − Φ, and in this limit the relativistic function Φ just becomes the
Newtonian potential φ. When multiplied by the expression for E =
−p0 given in Eq. ch07:eqn7.34, this factor cancels the gravitational
potential energy, leaving just the rest-mass and kinetic energy.

4 First calculate the components of the Ricci tensor Rαβ = g µν Rαµβν .

This would give, for example,

Rtt = e−2Λ Rtrtr + r−2 Rtθtθ + (r sin θ)−2 Rtφtφ

= [Φ00 − (φ0 )2 − Φ0 Λ0 − 2Φ0 /r] exp(2Φ − 2Λ).

Then, from all the components of the Ricci tensor one forms the Ricci
scalar, R = g αβ Rαβ . Finally one forms the Einstein tensor, e.g.

Gtt = Rtt − 21 gtt R.

5 The definition of a static metric is that this transformation must leave

all quantities invariant, but it sends t → −t and hence dr/dt →
−dr/dt. Only if this vanishes can it be invariant.

7 The construction proceeds as described in Sec. 10.5, except that to

integrate equations with p one has to specify the function S in the
equation of state. It would be simple to give a function S(r), but
not very physical (unless S is constant through the star), because one
can’t know ahead of time how the star’s structure will come out of the
equations. It is more physical to give S(m(r)), so that the entropy is a
function of the mass interior to a radius r. This might approximately
describe a star formed by spherical collapse that conserves entropy, so
that the rest-mass inside a given shell would be constant during the
collapse and so would be the entropy, leading to a fixed relationship
between entropy and interior rest-mass. Now, m(r) is not exactly
proportional to the rest mass, so giving S(m(r)) only approximates
this situation. It would be possible, but more complicated, to compute
the rest-mass interior to r and take the entropy to be a function of

34
 that. Since there is no differential equation for S one does not change
the manner of integration; one just adds an auxiliary function or look-
up table to determine the entropy at any radius and hence allow the
pressure there to be computed.

8 (a) Use reasoning similar to that in Exer. 5. (b) Make sure to include
the Christoffel symbols in the computation of the covariant derivative!
These can be found in Exer. 35 of Sec. 6.9.

10 (b) exp(Φ) = 0.999 997, 0.760; z = 3 × 10−6 , 0.315.

15 Putting the power series into the differential equations and matching
up powers of r shows that ρ1 = 0 and

ρ2 = −2πρc (ρc + pc )(ρc + 3pc )/(3Γc pc ).

The power series solution for the three functions then becomes

ρ = ρc + ρ2 r2 ,
p = pc (1 + ρ2 Γc /ρc r2 ),
m = 4πr3 ρc /3 + 4πr5 ρ2 /5.

Estimate the error in neglecting (uncalculated) the next term of each

equation to be about equal to the square of the last calculated
p term,
i.e., choose r such that the calculated terms are smaller than (0.01) ≈
3%. This means that
r2 6 |ρc /30ρ1 Γc |.

19 (a) 1.12 × 1042 kg m2 s−1 ; (b) 1.4 × 104 s−1 ; yes, by 50%; (c) 2.7 × 10−4 ;
(d) 5 × 109 Gauss.

35
 Chapter 11
dr
1 In Eq. 11.9, find the minimum approach distance by setting dτ = 0,
and calling this radial distance b. If M = 0 solving for b gives the
result. Note that if m = 0 the result is the same as if one had started
with Eq. 11.10. The impact parameter is defined for any orbit (even
if M is not zero) by the equation
b = L/[E 2 − m2 ]1/2 .
It can be thought of as the offset of the “aim” of the trajectory from
the center of the metric, when the trajectory is far away. If M is non-
zero then the trajectory will approach the center more closely than
b. For a photon orbit with m = 0, again even when M is not zero,
one can replace in Eq. 11.12 L by bE. Then if one re-scales the affine
parameter λ to a new one λ0 = Eλ, the resulting equation depends
only on one parameter, b. Re-scaling the affine parameter is always
allowed and does not change the path of the geodesic through the
spacetime. So we learn that photons follow geodesics that depend
only on their “aim”, or offset, from the center. Massive particles do
not: the degree to which they approach more closely than b depends
on their velocity far away, which (unlike that of a photon) depends on
the initial conditions.
4 The key radii are 3M , the location of the unstable photon circular
orbit, and 6M , the location of the last stable circular orbit for massive
particles. So if a star has a radius of 2.5M , as in (a), then all possible
orbits exist outside it, even photon circular orbits. If a star, as in (b),
has a radius of 4M , then it will not have any photon orbits, but the
last stable massive-particle orbit will still be outside its surface, so all
stable massive-particle orbits will exist. Finally, in case (c), if the star
has a radius of 10M then there will be stable circular orbits down
to its surface, and of course any quasi-elliptical and quasi-hyperbolic
orbits that approach no closer than the surface radius.
5 (a) The question is badly worded, since the value 1−2M/R is the rela-
tivistic value of −g00 , not its Newtonian approximation.pThe function
that is to be compared with the Newtonian potential is (−g00 )−1 =
−M/r −M 2 /r2 −. . .. This differs from the Newtonian potential −M/r
by the term M 2 /r2 , which makes a 1% change when M/r = 0.01.
(b) For a 106 M black hole, this is at a distance of about 1.5 × 1011 m,
just 15 time larger than the radius of a normal star. So nor more than

36
 of order 1000 normal stars can be entirely inside the highly relativistic
zone around a black hole of that mass. But if the mass is 109 M , then
the radius is 1000 times larger, and the volume is 109 times larger, so
that billions of normal stars could be in the highly relativistic region
near such a mega-massive black hole.
√
7 (a) Use Eq. (11.24) in ds2 = g00 dt2 + gφφ dφ2 to get τ = 20π 7 M .
√
(b) Same as coordinate time interval, Eq. (11.25): 20π 10 M .
√
(c) Integrate ds2 = g00 dt2 over the time in (b): 40π 2 M .
(d) 6.4 ms: innermost stable orbit.
√
(e) 40/ 2 yr, independent of M .

9 (a) Turning points are at r = 12.5 M and 5.47 M . (Notice that the
high L̃ enables the particle to turn around inside 6M . This is not a
contradiction with is radius being the last stable orbit, because that
applies only to circular orbits.) The orbit changes by ∆φ = 7.4 rad
between these points. A full orbit has ∆φ = 14.8 rad, or a perihelion
shift of 8.5 rad. The approximation Eq.11.37) gives only 2.3 rad. This
shows that highly noncircular orbits can have much greater shifts.

10 (a) (Ẽ/L̃)2 = [1 + 36M 2 /L̃2 + (1 − 12M 2 /L̃2 )3/2 ]/(54 M 2 ).

14 In arc sec per orbit and per year: Venus (0.052, 0.085), Earth (0.038,
0.038), Mars (0.025, 0.013).

20 (a) The Christoffel symbols may be computed as a special case of those

in the solution of Ex. 35 of Sec. 6.9.
(d) The quantity computed here is called the Riemann scalar. Note
that it is perfectly well-behaved at the horizon and only singular at
the center r = 0.

22 Approximately 105 M .

24 The point u = v = 0 has r = 2M , so is located on the horizon.

We expect it to be locally flat, but this can also be easily shown by
expanding the metric in a Taylor series in u and v. It is clear from
Eq. 11.68 that the expansion does not contain terms linear in u and
v, and so with some rescaling of the coordinates by constant factors it
can be brought into the form of a Minkowski metric with corrections
of order u2 and v 2 . Of course, the metric is locally flat everywhere
except at the singularity at r = 0, but we have chosen to look at the

37
 origin because it belongs to the extended Kruskal-Szekeres metric and
is not easily examined in Schwarzschild coordinates, and also because
it is relatively easy to do the computation there!

25 Estimate the total breakup force to be M/R2 , giving RT ∼ (MH /ρ)1/3 .

For the given density we find that the disruption mass is about 106 M .
This is an interesting mass considering that there is strong evidence
that a black hole of about this mass resides in the center of our Galaxy.

31 The ZAMO has angular velocity ω so that U φ /U 0 = ω. Denoting U 0

by the variable A, one can compute A from the normalization condition
U α U β gαβ = −1 in the Kerr metric.
√ √
34 (a) (2 − 2)M , (b) E 6 m1 + m2 − (m21 + m22 ) ≈ m1 (1 − 12 m1 /m2 ).

37 (a) The photon follows a null line along which only the coordinates t
and φ change. So the relevant interval is

ds2 = 0 = gtt dt2 + 2gtφ dtdφ + gφφ dφ2 .

This can be solved to give the photon’s angular velocity:

dφ p
= ω ± Ω2 + ω 2 ,
dt
where ω is the frame-dragging angular velocity defined in Eq. 11.90,
whose value in the equatorial plane (θ = π/2) is
−gtφ 2M a/r
ω= = 2 ,
gφφ r + a+ 2M a2 /r
and where Ω is the basic photon circular angular velocity and is defined
(again in the equatorial plane) by
−gtt 1 − 2M/r
Ω2 = = 2 .
gφφ r + a2 + 2M a2 /r

The forward-going photon’s angular velocity is (Ω2 + ω 2 )1/2 + ω, while

the backward-going angular velocity is −[(Ω2 + ω 2 )1/2 − ω]. The
backward-going photon travels more slowly than the forward-going
one. This is an illustration of the action of the dragging of inertial
frames, which drags in the forward direction.
Notice that, as r gets large, Ω approaches 1/r while ω falls off as 1/r3 .
So Ω dominates the angular velocity and the frame-dragging ω makes

38
a correction. The limiting value of Ω = 1/r is just what one needs in
order for the photon to travel around a distant circle of radius r in a
time equal to the circumference of the circle, 2πr.
The coordinate time that elapses is 2π divided by the absolute value
of the angular velocity. In the forward direction this is
r !
2π ω2 ω
∆t+ = 1+ 2 − .
Ω Ω Ω

In the backward direction this is

r !
2π ω2 ω
∆t− = 1+ 2 + .
Ω Ω Ω

The difference in the round-trip times is

4πω
∆t− − ∆t+ = .
Ω2
There is no redshift or blueshift of the photons: the components of
the photon’s 4-momentum are constant in time as it moves around
the circle, since the geometry is the same everywhere along the path.
(b) The ZAMO observer orbits with angular velocity ω. It follows that
the angular velocity of a photons relative to him is dφ/dt − ω, which
evaluates simply to ±(Ω2 + ω 2 )1/2 : for this observer the two photons
have the same (absolute value) angular velocity, so they take the same
time to go once around and meet him again! If he launches the two
photons simultaneously, they arrive back and meet him at the same
time. From the point of view of the first observer at rest, the forward-
going photon is going faster but has to travel further in order to meet
him again, since he moves around the black hole in the meantime,
while the backward-going photon is going slower but has less far to go
since the moving observer’s motion takes it towards this photon.
Notice that all the speeds of the photons we are talking about are
coordinate angular velocities. The photons always travel at speed 1
relative to any local inertial frame, but the curvature of space and time
makes them take different amounts of time to traverse their circular
paths.

39
 Chapter 12
8 (a) Use dχ2 = dr2 /(1 + r2 ). Integrating leads to r = sinh χ.
(c) A Lorentz transformation leaves the metric and hence the hyper-
bola unchanged, but changes the origin of spatial coordinates. Any
point on the hyperbola can be made into this origin by choosing the
correct transformation.
9 (a) In the text it is pointed out that pχ is conserved due to the homo-
geneity of the universe. A radially propagating photon has just two
non-zero four-momentum components, p0 and pχ , so we have
0 = g00 (p0 )2 + gχχ (pχ )2 ,
which by the diagonality of the metric can also be written as
0 = g00 (p0 )2 + g χχ (pχ )2 .
For each kind of universe gχχ = R2 (t), so that g χχ = 1/R2 (t). Since
g00 = −1, we finally get that
p0 = pχ /R(t).
The constancy of pχ establishes the fact that the locally measured
energy of the photon, p0 , decreases with time as 1/R(t). Notice that,
since the location of the origin of coordinates in a homogeneous space
is arbitrary, the radially-moving photon really can be any photon, so
this law is quite general.
(b) The redshift relation follows from the fact that the wavelength of
a photon is inversely proportional to its energy.
10 In Eq. 12.67, which is the result of the previous Exercise, expand R(tr )
about R(te ):
R(tr ) = R(te ) + Ṙ(te )(tr − te ) + . . . .
Now, for a photon traveling at the speed of light, tr − te = d0 , the
distance to the emitter as defined in Eq. 12.24. So we get
z = (Ṙ/R)(te )d0 .
Now, to the accuracy we are working (first order in time intervals or
d0 ) we can replace te in this expression with tr , which is the same as
today’s time t0 as used in Eq. 12.24. This makes z from Eq. 12.67
identical to v in Eq. 12.24 at this order.

40
13 Make sure you use the definition of the luminosity distance dL in
Eq. 12.34.

15 The result follows from the fact that the energy density of black-body
radiation is proportional to T 4 .

18 Pressure forces depend on pressure gradients, and in a homogeneous

universe the gradient is zero. So a cosmological tension does not lead
to a pulling inward. Rather, the entire effect of the negative pressure
is in the “active gravitational mass per unit volume” term ρ + 3p in
Eq. 12.55, where the large negative pressure overwhelms the density
and causes gravity itself to become repulsive.

22 (a) The energy density of black-body radiation is ρr = 4σT 4 /c, where

σ is the Stefan-Boltzmann constant, whose value in SI units is 5.67 ×
10−8 Wm−2 K−4 . Divide by another factor of c2 to get this in mass-
density units. When divided by ρm we get
ρr
ε= ∼ 4m−1 × 10−4 .
ρm

Then we get the number of photons per baryon to be about 2m−1 ×109 .
In the early universe, if there were no baryon-antibaryon asymmetry,
then one would have expected of order one baryon and one antibaryon
per photon, because the baryons would have been in thermal equilib-
rium with the photons at temperatures where the photons had enough
energy to create baryon-antibaryon pairs. What this large number tells
us is that the excess of baryons over antibaryons is of order one part
in 109 .

23 z = 3 × 103 .

24 The density must exceed ρ = 3H02 /8π = 10−26 kg m−3 . This is called
the closure density or critical density and is denoted by ρc .

41