General Relativity Fall 2019

Homework 9 solution

Exercise 1: Cosmological constant

Suppose that instead of Einstein’s equations Gµν = 8πTµν , gravity was described by the modified Einstein field
equation (EFE) with a cosmological constant Λ:

Gµν + Λgµν = 8πTµν . (1)

Λ
(i) We can move the term Λgµν to the right-hand side and see it as the stress-energy tensor Tµν ≡ −Λgµν /8π.
Show that it can be seen as the stress-energy tensor of an ideal fluid, and give explicitly its rest-frame density and
pressure ρΛ , PΛ in terms of Λ.

Recall that the stress-energy tensor of an ideal fluid is Tµν = ρuµ uν + P (gµν + uµ uν ) = P gµν + (ρ + P )uµ uν , where
uµ is the fluid’s 4 velocity. We want this to be equal to −Λgµν /8π. Dotting this into the fluid’s velocity, we find

Λ Λ
uµ Tµν = − uν = −ρuν ⇒ ρΛ = . (2)
8π 8π

Taking the trace we find

Λ µ Λ Λ
T µµ = − g =− = 3P − ρ ⇒ PΛ = − = −ρΛ . (3)
8π µ 2π 8π

Plugging back, you see that the fluid’s velocity indeed drops out when P = −ρ, and that we do recover the correct
stress-energy tensor.

(ii) What dimensions must the constant Λ have? Current cosmological observations indicate that a cosmological
constant is indeed likely present, with a density corresponding to ρΛ ∼ 1011 M /Mpc3 . What lengthscale does this
density correspond to? Do you expect any effect on solar-system orbits, and if not, why?

From the modified EFE, knowning that Gµν has dimensions of inverse length squared, we find that Λ has dimensions
of inverse length squared. Anther way to see it is that we found that ρΛ ∼ Λ, so Λ has dimensions of energy density,
i.e. again, inverse length squared.

Numerically, the lengthscale corresponding to ρΛ is

1
−1/2 −1/2
√ ∼ 1011 M /Mpc3 ∼ 3 × 10−6 (km/Mpc) Mpc ∼ 15 Gpc. (4)
ρΛ

We expect this to be completely negligible on solar-system orbits. Indeed, the radius of curvature of spacetime
in the solar system is at most of order ∼ (r3 /M )1/2 ∼ 100 pc, where I took r ∼ 100 AU as the semi major
axis of the outermost planets. Note that larger radii of curvature mean less gravity, so saying that the radius of
curvature is at most 100 pc is equivalent to setting a minimum bound to the strength of gravity. This means that
Gµν /ρΛ & (100pc)−2 /(15Gpc)−2 & 108 , thus the cosmological constant term is completely negligible on solar-system
scales.

(iii) Consider a spherically-symmetric spacetime with line element ds2 = gtt (t, r)dt2 + grr (t, r)dr2 + r2 dΩ2 – note
that the general form of the metric does not depend on Einstein’s field equations. Using the expressions given in the
lecture notes for the components of the Ricci tensor, solve for the metric coefficients in vacuum with the modified
EFE. Rescale your variables appropriately so the solution reduces to the Schwarzschild solution when Λ = 0. Hint:
start by finding a simple expression for the Ricci scalar from the trace of the vacuum modified EFE.

Lets start by finding R = Rµµ = −Gµµ . Taking the trace of the vacuum EFE, we find

0 = Gµµ + Λg µµ = −R + 4Λ ⇒ R = 4Λ . (5)
 2

Therefore, the vacuum EFEs are

1
0 = Rµν − Rgµν + Λgµν = Rµν − Λgµν ⇒ Rµν = Λgµν . (6)
2
The tr equation is identical as the one we derived in class, and gives ∂t grr = 0, i.e. grr is a function of r only.
Setting Rθθ = Λgθθ = Λr2 , we find again that ∂r ln |gtt | is a function or r only, hence, here again we can write
gtt = ±eA(t) eB(r) . Upon recaling the time, we can get rid of the time dependence alltogether. So the metric is now
ds2 = gtt (r)dt2 + grr (r)dr2 + r2 dΩ2 . Combining the tt and rr equations, we find

1 grr grr
∂r ln |gtt grr | = Rrr − Rtt = Λgrr − Λgtt = 0. (7)
r gtt gtt

Thus, here again, gtt grr = constant, and we may chose this constant to be −1 by appropriately rescaing the time.
Let us finally solve for gtt . The θθ equation is now

Rθθ = 1 + ∂r (rgtt ) = Λgθθ = Λr2 . (8)

Integrating, we find
 
2M Λ
gtt = − 1 − + r2 , (9)
r 3

where M is once again an integration constant. Thus we found the following solution:
   −1
2 2M Λ 2 2 2M Λ 2
ds = − 1 − − r dt + 1 − − r dt2 + r2 dΩ2 . (10)
r 3 r 3

We see that the cosmological constant will become relevant only at sufficiently large separations,

r & (M/Λ)1/3 . (11)

(iv) Derive the effective potential for radial motion Veff (r) corresponding to this metric.

If we define
Λ 3 4π 3
M (r) ≡ M + r =M+ r , (12)
6 3
we see that the line element takes eaxctly the Schwarzschild form with M → M (r). The effective potential is thus
identical to that of the Schwarzschild spacetime, provided we substitute M → M (r).

Exercise 2: Classifying timelike orbits in Schwarzschild

In Newtonian theory, the radial equation of motion with specific energy E and angular momentum L is simply
2
L2

1 dr M
+ Veff (r) = E, Veff (r) ≡ − + 2. (13)
2 dt r 2r

Newonian orbits can be classified as follows:

• if E < 0, orbits are bound, i.e. the radius remains in a finite range rmin ≤ r ≤ rmax , where the bounds are
such that Veff (rmin ) = Veff (rmax ) = E. These equations can in fact be solved explicitly for rmin and rmax , and give
rmin = a(1 − e), rmax = a(1 + e), where a = −M/(2E) and 1 − e2 = −2EL2 /M 2 .
• if E ≥ 0, orbits are unbound at large r: the radius is in the interval rmin ≤ r < +∞, where again rmin can be found
explicitly from Veff (rmin ) = E.

Do a similar classification of timelike geodesics in the Schwarzschild metric, and express relevant parameters ana-
lytically when possible (and if not, define them by an implicit equation). The classification is a bit more subtle than
Newtonian orbits, as the types of orbits depend on the ratio L/M .
 3

����
����
����

���� (�)
����
-����
-����
-����
� �� �� ���
�/�
√
FIG. 1. Effective potential for ` = L/M = 12, 4, 5 from bottom to top.
.

Let’s recall the effective potential, in terms of the dimensionless variables x ≡ r/M and ` ≡ L/M :

`2 `2 `2
 
1 1
Veff (x) = − + 2 − 3 = − 3 x2 − x + `2 . (14)
x 2x x x 2
√
• For ` > 12, we saw in class that the effective potential has a minimum at

` p 
x+ = ` + `2 − 12 (15)
2
and a maximum at
` p 
x− = ` − `2 − 12 . (16)
2
Plugging in, one can see that Veff (x+ ) is always negative, and that Veff (x− ) > 0 if ` > 4 and negative otherwise, see
Fig. 1. Therefore, we have the followning sub-classification, depending on whether ` is greater or less than 4:

I For ` > 4, Veff (x− ) > 0,

? If E < 0, orbits are bound between two radii, xmin ≤ x ≤ xmax , which are the solutions of Veff (xmin ) = Veff (xmax ) = E.
They are qualitatively similar to the elliptical orbits in the Kepler problem, except they do not close.
? If 0 < E < Veff (x− ), orbits are bound from below, but not above: xmin < 0 < ∞. These are qualitatively similar to
the hyperblic orbits in the Kepler problem.
? If E > Veff (x− ), orbits can plunge to r → 2M (unless they start with dr/dτ > 0, in which case they escape to
infinity). These plunging orbits have no equivalent in the Kepler problem.

I For ` < 4, Veff (x− ) < 0,

? If E < Veff (x− ) < 0, orbits are bound xmin < x < xmax , those are qualitatively similar to elliptical orbits in the
Kepler problem, albeit with a large “precession” since ` ≤ 4.

? If E > Veff (x− ), orbits can plunge to r → 2M . In this case there are no hyperbolic-like orbits that can reach
infinity and are bound by a minimum radius
√
• For ` ≤ 12, orbits are not bound from below, and eventually plunge to r → 2M (unless a particle is thrown in
the dr/dτ > 0 direction with E > 0, in which case it escapes to infinity).
 4

Exercise 3: Null geodesics of Schwarzschild

In class we considered timelike geodesics. Now consider null geodesics with affine parameter λ.

(i) Using the symmetries of the Schwarzschild metric, derive the geodesic equations without computing Christoffel
symbols. By combining the equations for r and ϕ, derive an equation for dr/dϕ for equatorial orbits (θ = π/2)

Let us define p = d/dλ to be the tangent vector. Spherical symmetry implies that the orbit is planar, and we will
set them in the plane θ = π/2 without loss of generality. We moreover have L = pϕ = constant. Stationarity implies
that E ≡ −pt is constant. Let us moreover recall that gµν pµ pν = 0, and use the fact that the metric is diagonal, to
arrive at
dϕ
L = pϕ = gϕϕ pϕ = r2 , (17)
dλ
 2
dr 1 2
0 = pµ pµ = g tt p2t + grr (pr )2 + g ϕϕ L2 = −(1 − 2M/r)−1 E 2 + (1 − 2M/r)−1 + L . (18)
dλ r2
The second equation can be rewritten in the form
 2
dr L2
+ 2 (1 − 2M/r) = E 2 . (19)
dλ r
Combining it with the first equation, we find
 2
L2 dr L2
4
+ 2 (1 − 2M/r) = E 2 . (20)
r dϕ r

We see that the equation in fact only depends on the ratio E/L:
2
E2 4

dr
+ r2 (1 − 2M/r) = r . (21)
dϕ L2

(ii) Consider a photon moving towards the “origin”, starting at r → ∞ and ϕ → 0 with impact parameter b. Relate
b to the constants of motion of the geodesic. For what values of b does the photon stay at r > 2M and escape back to
infinity? When this condition is satisfied, compute the deflection angle ∆ϕ, i.e. the value of ϕ when r → ∞ again.
The result should be an integral, depending on b/M . Evaluate this integral numerically if needed, and plot ∆ϕ as a
function of b/M . Explain in words (with a cartoon if needed) how photon orbits look like for different values of b/M .

At large distance, we see that |dr/dϕ| → (E/L)r2 . On the other hand, we have b/r = sin ϕ ≈ ϕ, thus
|dr/dϕ| → |b/ϕ2 | = |r2 /b|. So we conclude that b = L/E .

We can rewrite the radial equation as

 2
dr
= r(r3 /b2 − r + 2M ) ≡ rF (r). (22)
dϕ

Orbits can only exist in regions where F (r) > 0. The cubic
√ polynomial F (r), seen as a function of r ∈ R, starts at
b
−∞ at r → −∞, reaches a local maximum at r1 = −b/ 3 < 0, with value F (r1 ) = 2(M + 3√ 3
) > 0, is still positive
√ b
at r = 0, then reaches a local minimum at r2 = +b/ 3, with value F (r2 ) = 2(M − 3√3 ), and then increases again to
√
+∞ as r → +∞. If F (r2 ) > 0, i.e. b < 27M √ , nothing prevents the geodesics from reaching arbitrarily small radii.
On the other hand, if F (r2 ) < 0, i.e. b > 27M , then orbits cannot reach radii below rmin > r2 > 3M , such that
√
F (rmin ) = 0. Thus, the minimum impact parameter for the orbit to escape back is bmin = 27 M . This is related
to the “black hole shadow” – and would be an interesting final project.

Eq. (22) can be rewritten as

dϕ 1
= ±p , (23)
dr 3 2
r(r /b − r + 2M )
 5

��
�

�
Δϕ-π ����

����
����
� �� �� ���
�/�
FIG. 2. Defelection angle of null geodesics in the Schwarzschild metric as a function of impact parameter (solid line). The
dashed line shows the asymptotic result ∆ϕ − π ≈ 4M/b, valid for b  M , which we had already derived in the weak-field
limit.

where the sign is, say, + as r decreases and − as r increases – recall that r2 dϕ/dλ = L is constant, so dϕ/dλ has a
constant sign.
The deflection angle is obtained by integrating dϕ/dr between r = ∞ and rmin , and again from rmin to ∞, i.e.
Z ∞
dr 3
∆ϕ = 2 p , rmin /b2 − rmin + 2M = 0. (24)
r(r 3 /b2 − r + 2M )
rmin

It is always better to write things in a dimensionless fashion. We rescale x ≡ r/b so that

Z ∞ √
dx
∆ϕ = 2 p , x3min − xmin + 2M/b = 0, xmin > 1/ 3. (25)
3
x(x − x + 2M/b)
xmin

First, when M/b → 0, xmin = 1, and the integral can be computed analytically and give ∆ϕ = π, i.e. no deflection:
the photon just moves on a straight line.

For general b/M , we must compute the integral numerically, and show the result in Fig. 2. For b  M , we recover
√ analytic limit we had computed in the far-field of a point mass, ∆ϕ − π ≈ 4M/b. Second, as b/M approaches
the
27, the deflection increases, and diverges. This means that for small enough impact parameters,
√ photon orbits circle
the central
√ mass before going back to infinity, and they circle it more and more as b → 27M . In the limit that
b →√ 27M , photons get nearly trapped on an unstable√circular orbit at r = 3M , before heading back to infinity if
b > 27M , or plunging towards the central mass if b < 27M .