General Relativity Fall 2019

Lecture 20: Geodesics of Schwarzschild

Yacine Ali-Haı̈moud
November 7, 2019

In this lecture we study geodesics in the vacuum Schwarzschild metric, at r > 2M .

Last lecture we derived the following equations for timelike geodesics in the equatorial plane (θ = π/2):
2
L2 M L2

dϕ L 1 dr M
= 2, + Veff (r) = E, Veff (r) ≡ − + 2− 3 , (1)
dτ r 2 dτ r 2r r

where E ≡ (E 2 − 1)/2 can be interpreted as a kinetic + potential energy per unit mass. The radial equation can also
be rewritten as
d2 r 0 M3 h 2 2 2
i
= −Veff (r) = − r̃ − L̃ r̃ + 3 L̃ , r̃ ≡ r/M, L̃ = L/M. (2)
dτ 2 r4

CIRCULAR ORBITS AND THE ISCO

We show the effective potential in Fig. 1. In contrast to the Newtonian effective potential for orbits around a central
mass (i.e. Veff ≡ −M/r + L2 /2r2 , without the last term −M L2 /r3 ), which always has √ a minimum at rNewt = L /M ,
2

the relativistic
√ effective potential has both a maximum
√ and a minimun for L > 12 M , an inflection point for
L = 12 M , and is strictly monotonic for L < 12 M .
0
Circular orbits (with √ r = constant) are such that Veff (r) = 0. Solving this equation, one finds that such
orbits exist only for L > 12 M . When this condition is satisfied, the radii of circular orbits are

L2  p 
rc± = 1 ± 1 − 12M 2 /L2 . (3)
2M
The Newtonian limit is obtained for L  M , in which case rc+ → L2 /M .

From classical mechanics, we know that circular orbits where d2 Veff /dr2 > 0 are stable, whereas those with
d Veff /dr2 < 0 are unstable. This can simply be understood graphically: for a given energy, an orbit is confined to
2

the interval in radii for which Veff (r) ≤ E. At a circular orbit, this is an exact equality. Perturb the orbit away by
giving it a bit of extra energy, and it will start oscillating between two radii if d2 Veff /dr2 > 0. If d2 Veff /dr2 ≤ 0, it
will either plunge towards r → 2M or diverge to infinity.
√
The smallest radius of a stable circular orbit is obtained for L = 12 M , and has value rc = 6M – strictly
speaking, only orbits with rc > 6M are actually stable. This orbit is called the innermost stable circular orbit
(ISCO).

The energy associated with a circular orbit at the ISCO is given by

1 p p
Eisco = Veff (6M ) = − , ⇒ Eisco = 1 + 2Eisco = 8/9 ≈ 0.94. (4)
18
This number is often invoked to quantify the efficiency of accretion around Schwarzschild black holes: suppose
material is orbiting around a central mass (typically a black hole) on a circular orbit, and slowly looses energy and
angular momentum due to dissipative forces, while remaining on a quasi-circular orbit. If it starts at large distances
with E ≈ 1, it will lose about 6% of its energy to dissipation before reaching the ISCO and plunging towards r → 2M .
Thus if the rate of mass-energy accretion is Ṁ , the luminosity radiated by dissipation is at most 0.06 Ṁ . This is
usually quoted as a maximal efficiency of ∼ 6% for accretion onto Schwarzschild black holes. Spinning black holes
(which we will explore in the last classes) can be much more efficient.
 2

���
L/M = Newtonian
effective potential
5
���
circular orbits
4.2
���� (�)
���
p
2 3
-���
3
-���
� �� �� ���
�/�
√
FIG. 1. Effective potential for orbits in the Schwarzschild metric. The numbers label the value of L/M . For L/M < √12, the
potential has no extremum, and there are no circular orbits – in fact, all orbits are plunging to r → 2M . For L/M > 12, the
effective potential has a local maximum, corresponding √ to an unstable circular orbit, and a local minimum, corresponding to a
stable circular orbit. The two merge together at L = 12M , for which the marginally stable circular orbit is at r = 6M . This
is the innermost stable circular obit (ISCO).

PRECESSION OF PERICENTER

In this section we consider bound orbits, i.e. for which E < 0, and the coordinate r is bound within a range
rmin ≤ r ≤ rmax , called the pericenter and apocenter, respectively. We adjust the polar angle ϕ such that, at some
arbitrary pericenter passage, r(ϕ = 0) = rmin , and dr/dτ (ϕ = 0) = 0. Note that the next pericenter passage need not
be at ϕ = 2π if the orbit does not close, as we will see now!

From the angular equation, we have d/dτ = (L/r2 )d/dϕ, so the radial equation can be rewritten as
 2
1 L2 dr
+ Veff (r) = E. (5)
2 r4 dϕ
Let us define the variable u ≡ L2 /(M r). It satisfies the following differential equation:
 2
1 du 1 1 2L2
− u + u2 − (M/L)2 u3 = − (1 − e2 ), 1 − e2 ≡ − 2 E (6)
2 dϕ 2 2 M
Differentiating again, we get
d2 u
+ u − 1 = 3(M/L)2 u2 . (7)
dϕ2
Recall that we have adjusted ϕ such that dr/dτ = 0 at ϕ = 0, hence du/dϕ = 0 at ϕ = 0.

We will solve this equation perturbatively in (M/L)2  1, corresponding to the Newtonian limit:
u(ϕ) = u0 (ϕ) + (M/L)2 u1 (ϕ) + ... (8)
2 2
Note that since E ∼ (L/M ) (consider, for instance, circular orbits), the quantity 1 − e is of order unity. The
lowest-order term satisfies
d2 u0
+ u0 = 1. (9)
dϕ2
 3

This has a simple solution: u0 (ϕ) = 1 + A cos ϕ, where we have adjusted the origin of ϕ such that du0 /dϕ = 0 at
ϕ = 0. To find the integration constant A, we plug into
 2
1 du0 1 1
− u0 + u20 = − (1 − e2 ), (10)
2 dϕ 2 2

implying A2 = e2 . For definiteness, we pick A = +e, corresponding to L > 0. The zero-the order solution is
therefore the well-known ellipse,
L2 /M
r0 (ϕ) = , (11)
1 + e cos ϕ
where e is the eccentricity. This orbit happens to be periodic, which is a unique feature of the 1/r potential.

Let us now find the next-order correction:

d2 u1 du1
+ u1 (ϕ) = 3u20 (ϕ), (ϕ = 0) = 0. (12)
dϕ2 dϕ
The solution is
u1 (ϕ) = (3 + e2 ) + (C − 3 − e2 ) cos ϕ + e2 sin2 ϕ + 3e ϕ sin ϕ, (13)
where C ≡ u1 (0) is an integration constant, which we won’t need to calculate explicitly for the problem of pericenter
precession, but is easy to find from Eq. (6): C = (1 + e)3 /e.
Obviously this solution cannot be correct for arbitrary ϕ, because of the growing term ϕ sin ϕ. This is because
deviations away from a Keplerian orbit accumulate over time. So we should only consider this perturbative solution
to be valid over one orbital period 0 ≤ ϕ ≤ 2π – after that, one can reset the origin of ϕ and reiterate the perturbative
expansion.

We want to know at what angle the next pericenter passage will be. In the Kepler problem, it is at ϕ = 2π, but,
because of the additional term in the effective potential, the next pericenter passage will be at ϕ = 2π + ∆ϕ, where
we expect ∆ϕ ∼ (M/L)2 . The pericenter passage is defined by dr/dϕ = 0, hence u0 (ϕ) = 0. Thus we must solve for
u00 (2π + ∆ϕ) + (M/L)2 u01 (2π + ∆ϕ) = 0. (14)
We Taylor-expand this around ∆ϕ = 0, and find, to lowest order,
u01 (2π)
∆ϕ = −(M/L)2 = 6π(M/L)2 . (15)
u000 (2π)

Thus we conclude that the pericenter (as well as apocenter) precesses by ∆ϕ = 6π(M/L)2 per orbit.

This is usually expressed in terms of the semi-major axis a and eccentricity. The semi-major axis is a = (rmax +
rmin )/2 = L2 /M/(1 − e2 ), i.e. L2 = M a(1 − e2 ), thus

6πM
∆ϕ = . (16)
a(1 − e2 )

We now apply this to the orbit of Mercury around the Sun, which was one of the earliest successfull tests of
GR. Here we have M = M , aM ≈ 6 × 1012 cm, e ≈ 0.2. This leads to a precession rate of ∆ϕ ≈ 0.1 arcsec per
orbital period, hence Ω = d∆ϕ/dt ≈ 43 arcsec/century.
It is interesting to compare this to the effect of Jupiter on the orbit. The tidal acceleration due to Jupiter is of order
MJ aM /a3J . The amplitude of this perturbed acceleration relative to the relativistic correction is therefore of order
3
MJ aM /a3J

δfJupiter aM MJ aM aM
∼ ∼ ∼ 10−6 ∼ 10, (17)
δfGR M 2 /a3M M M aJ M

where we used MJ ∼ 10−3 M and aJ ∼ 7 × 1013 cm ∼ 10 aM . Therefore we see that the relativistic precession is
a factor of ∼ 10 times weaker than the precession due to Jupiter’s tidal force. To be aware of this tiny “anomalous
precession” in the early 20-th century, physicisists therefore had to model the effect of Jupiter (and Solar-sytem
planets) to exquisite precision!
 4

RADIAL ORBITS

Consider a massive particle that is radially infalling, so that L = 0, and dr/dτ < 0. For L = 0 the radial geodesic
equation reduces to
dr p
(dr/dτ )2 − 2M/r = E 2 − 1 ⇒ = − E 2 + 2M/r − 1. (18)
dτ
The proper time elapsed between two radii ri > rf is
ri ri /2M ri /2M √
dr dx x dx
Z Z Z
∆τ = p = 2M p = 2M √ , (19)
rf E 2 + 2M/r − 1 rf /2M E 2 + 1/x − 1 rf /2M 1 − x + xE 2

For E 2 > 0, this integral is clearly convergent, for any rf ≥ 2M and even for E 2 →, it converges and remains finite
as rf → 2M . Therefore an observer falling towards r → 2M reaches r = 2M in a finite proper time.

Let us now compute the coordinate time that has elapsed. Recall that E = −ut = −gtt ut = (1 − 2M/r)dt/dτ .
Hence, we find
dr p
= −(1 − 2M/r) E 2 + 2M/r − 1
dt Z r0
dr
⇒ ∆t = E p . (20)
2M (1 − 2M/r) E + 2M/r − 1
2

This diverges logarithmically at r → 2M because of the term 1/(1 − 2M/r). Hence, it takes an infinite amount of
coordinate time to arrive at r = 2M , even though it only takes a finite amount of proper time to get there. This
is an indication that the coordinates are ill suited to describe the spacetime as r → 2M .