Lecture XXVI: Kerr black holes: I.
Metric structure and regularity of particle orbits
Christopher M. Hirata
Caltech M/C 350-17, Pasadena CA 91125, USA∗
(Dated: February 22, 2012)
I. OVERVIEW
We now wish to consider the case of a neutral black hole with angular momentum. This is an astrophysically
relevant case – collapsing stars may be rotating, and accreting black holes (such as the ones we see!) collect material
with angular momentum. We will see that the problem is very similar to that for the charged black hole, with the
corrections to the radial equation from angular momentum leading to a similar structure – two event horizons, etc.
However the deviation from spherical symmetry leads to some interesting additional properties outside the outer
horizon, i.e. in the part of the hole we can actually see.
Reading:
• MTW Ch. 33.
• If you’re interested in learning more about the global structure, see Hawking & Ellis, The Large-Scale Structure
of Spacetime.
• The great details of perturbations of the Reissner-Nordstrøm metric, and what goes wrong at the inner horizon,
can be found in Chandrasekhar, The Mathematical Theory of Black Holes.
II. THE KERR METRIC
The Kerr metric is the only stationary, axisymmetric vacuum black hole solution. (For a long proof, see Chan-
drasekhar §§52–55.) Its form is
(r2 + a2 )2 − a2 ∆ sin2 θ
2M r 4aM r Σ
ds2 = − 1 − dt2 − sin2 θ dt dφ + sin2 θ dφ2 + dr2 + Σ dθ2 , (1)
Σ Σ Σ ∆
where M and a are constants, and
∆ = r2 − 2M r + a2 and Σ = r2 + a2 cos2 θ. (2)
There are many algebraically equivalent forms in which this can be written with the same set of coordinates {t, r, θ, φ}
called Boyer-Lindquist coordinates.
The Kerr metric tensor components are independent of t and φ and therefore there are two Killing fields ξ and ζ
with ξ α = (1, 0, 0, 0) and ζ = (0, 0, 0, 1). They commute, [ξ, ζ] = 0, and correspond respectively to the conservation
of energy E = −p · ξ and z-angular momentum L = p · ζ.
It is clear that for a = 0, the Kerr metric becomes the Schwarzschild metric with mass M . The significance of a
can be obtained by taking the large-r limit – i.e. expanding to order M/r or a/r2 :
2M 4aM 2 2M
2
ds = − 1 − 2
dt − 2 sin θ dt dφ + 1 + dr2 + r2 (dθ2 + sin2 θ dφ2 ), (3)
r r r
or – switching to Cartesian-like coordinates where x1 = r sin θ cos φ, x2 = r sin θ sin φ, and x3 = r cos θ –
2M 4aM 2M 2
2
ds = − 1 − dt2 − 3 ǫ3ij xi dt dxj + [(dx1 )2 + (dx2 )2 + (dx3 )2 ] + dr . (4)
r r r
2
The a term thus corresponds to an angular momentum of the black hole (as measured by gyroscope experiments from
the outside) of magnitude J ≡ aM and directed in the +3 direction. If we flip the sign of a, this is equivalent to a
black hole rotating in the opposite direction (φ → −φ): thus in astrophysical applications involving accretion disks,
one often refers to a retrograde system (disk orbiting opposite to the black hole spin) as a < 0.
The angular momentum has units of mass squared, so in addition to the parameter a it is common to define a
dimensionless angular momentum for a black hole:
a J
χ ≡ a⋆ ≡ = 2. (5)
M M
III. THE GLOBAL STRUCTURE
We wish to understand the global behavior of the Kerr metric, just as we did for Schwarzschild. It turns out the
most convenient way to do this is to complete the square in the metric. In general, we have for the tφ sector of the
metric:
2
AC − B 2 2
2 2 B
A dt + 2B dt dφ + C dφ = C dφ − dt + dt . (6)
C C
For example:
(r2 + a2 )2 − a2 ∆ sin2 θ
C= sin2 θ. (7)
Σ
Then:
AC − B 2 −(Σ − 2M r)[(r2 + a2 )2 − a2 ∆ sin2 θ] − 4a2 M 2 r2 sin2 θ
=
C [(r2 + a2 )2 − a2 ∆ sin2 θ]Σ
−Σ[(r2 + a2 )2 − a2 ∆ sin2 θ] + 2M r[(r2 + a2 )2 − a2 ∆ sin2 θ − 2a2 M r sin2 θ]
=
[(r2 + a2 )2 − a2 ∆ sin2 θ]Σ
−Σ[(r2 + a2 )2 − a2 ∆ sin2 θ] + 2M r(r2 + a2 )Σ
=
[(r2 + a2 )2 − a2 ∆ sin2 θ]Σ
−(r2 + a2 )(r2 + a2 − 2M r) + a2 ∆ sin2 θ
=
(r2 + a2 )2 − a2 ∆ sin2 θ
−∆Σ
= . (8)
(r + a )2 − a2 ∆ sin2 θ
2 2
We then find:
2
(r2 + a2 )2 − a2 ∆ sin2 θ
2 ∆Σ 2 2 2aM r Σ
ds = − 2 dt + sin θ dφ − 2 dt + dr2 +Σ dθ2 .
(r + a2 )2 − a2 ∆ sin2 θ Σ (r + a2 )2 − a2 ∆ sin2 θ ∆
(9)
This is an algebraically equivalent form to Eq. (1).
What is interesting about Eq. (9) is that, like the usual form of Schwarzschild, it tells us about the range of allowed
trajectories in the (t, r)-plane. This is because the normalization condition for the 4-velocity is now
2
∆Σ(ut )2 (r2 + a2 )2 − a2 ∆ sin2 θ
φ 2aM r t Σ
−1 = − 2 2 +
2
sin θ u − 2 2 u + (ur )2 +Σ(uθ )2 .
2 2 2
(r + a ) − a ∆ sin θ Σ 2 2 2
(r + a ) − a ∆ sin θ ∆
(10)
Thus we have the usual situation – the allowed regions in the (ut , ur )-plane are hyperbolae that allow one to move
either inward or outward – as long as ∆ > 0. If ∆ < 0, then particles moving inward (ur < 0) are forced to maintain
this motion. The situation is exactly analogous to the Reissner-Nordstrøm case, with a replacing the charge: we have
∆ = 0 at
p
r = r± = M ± M 2 − a2 , (11)
a “normal region” at r > r+ , and an outer horizon at r = r+ . Particles falling through the outer horizon are doomed
to have their r continue to decrease until they reach the inner horizon, r = r− , at which point they see the end of the
universe and are destroyed.
3
If particles could pass through r = r− without being destroyed, then the analytic continuation of Kerr, like
Reissner-Nordstrøm, goes all the way down to the singularity at r = 0. In fact, this singularity is not a single point
– if cos θ 6= π/2, then the metric, including spherical-like coordinates, are perfectly well-behaved at r = 0! There
is thus a ring singularity in the analytic continuation of Kerr, and on the other side (r < 0) a whole universe with
asymptotically Minkowski-like geometry. However, it seems that in the actual collapse of a rotating star, this structure
does not form for the aforementioned reasons.
IV. EXTERIOR STRUCTURE OF KERR SPACETIME
Outside the outer horizon, at r > r+ , the Kerr spacetime allows particles to move either inward or outward in
radius. However, their motion in longitude may be restricted. In particular, using Eq. (10), and that dφ/dt = uφ /ut ,
we see that
√
dφ 2aM r Σ ∆
− < . (12)
dt (r2 + a2 )2 − a2 ∆ sin2 θ [(r2 + a2 )2 − a2 ∆ sin2 θ] sin θ
At large radii, this is not of particular note – it becomes |dφ/dt| < 1/r. But near the hole, strange things start
to happen. In particular, there is a region called the ergosphere where dφ/dt must have the same sign as a – any
particle is forced to revolve in the direction of the hole’s spin! This is where
√
2|a|M r sin θ > Σ ∆. (13)
Squaring this gives
4a2 M 2 r2 sin2 θ > Σ2 ∆. (14)
Our work on Eq. (8) gave an alternative form for Σ2 ∆:
Σ2 ∆ = (Σ − 2M r)[(r2 + a2 )2 − a2 ∆ sin2 θ] + 4a2 M 2 r2 sin2 θ, (15)
so the ergosphere is the region outside the horizon (r > r+ ) and with Σ < 2M r, or
r2 − 2M r + a2 cos2 θ < 0. (16)
Therefore the region is
p p
r+ = M + M 2 − a2 < r < rsl (θ) = M + M 2 − a2 cos2 θ. (17)
The outer boundary of the ergosphere is the static limit. At the static limit, a photon can instantaneously have
dφ/dt = 0, but all matter particles are inexorably dragged along with the rotation of the hole. The static limit
touches the horizon at the poles, and extends outward to r = 2M at the equator. For the Schwarzschild case, it
merges with the horizon (the ergosphere vanishes).
The behavior as one approaches the horizon itself is weirder. Recall that as we approach the horizon, ∆ → 0+ .
Then all objects near the horizon must orbit the hole at the rate
dφ 2aM r+ a
= ΩH = 2 = 2 . (18)
dt (r+ + a2 )2 r+ + a2
Thus (classically) the objects that fall onto the hole have their exponentially fading images appear to rotate at a
constant velocity. The black hole thus appears to rotate as a solid body.
It is worth summarizing these results both in the case of small spin parameter, χ ≪ 1, and large spin parameter,
χ = 1 − ǫ.
• For small spins χ ≪ 1, the outer horizon is located at r+ = (2 − 21 χ2 )M , and the rotation rate of the horizon is
ΩH = χ/(4M ).
√
• For large spins
√ χ = 1 − ǫ, the outer horizon is located at r+ = (1 + 2ǫ)M , and the rotation rate of the horizon
is ΩH = (1 − 2ǫ)/(2M ).
4
V. PARTICLE ORBITS
We are now interested in test particle orbits in the Kerr spacetime. In order to study such orbits, we need to
understand the conserved quantities. For a test particle of mass µ, the energy per unit mass E = −ut and the z-
angular momentum per unit mass L = uφ will be conserved because of the existence of the appropriate Killing fields.
In order to understand the allowed motions, then, we must use the 4-velocity normalization,
g αβ uα uβ = −1. (19)
This requires us to derive the inverse-metric components by inversion of gαβ . This is tedious but straightforward: the
result is
(r2 + a2 )2 − a2 ∆ sin2 θ
g tt = − ,
∆Σ
2aM r
g tφ = − ,
∆Σ
∆ − a2 sin2 θ
g φφ = ,
∆Σ sin2 θ
∆
g rr = , and
Σ
1
g θθ = . (20)
Σ
The normalization condition is then
−(r2 + a2 )2 E 2 + a2 ∆ sin2 θ E 2 + 4aM rLE + ∆L2 csc2 θ − a2 L2 ∆(ur )2 + (uθ )2
−1 = + . (21)
∆Σ Σ
In the Schwarzschild case, we found the allowed region by rotating to a coordinate system that made the orbit
equatorial. No such luck for Kerr: the spacetime has a preferred axis. We could find an allowed region in the
(r, θ)-plane by taking the last term in Eq. (21) to be non-negative. But we can do better.
A. The Carter constant
It turns out that the Kerr spacetime has a peculiar additional property that is non-obvious and whose reason for
existence remains shrouded in mystery. We may write
duθ d
= (Σuθ )
dτ dτ
duθ dΣ
= Σ + uθ
dτ dτ
= −ΣΓµν u u + uθ Σ,µ uµ
θ µ ν
1
= gµν,θ uµ uν − gµθ,ν uµ uν + uθ Σ,µ uµ . (22)
2
In the gµθ,ν uµ uν , only the µ = 0 terms can give a nonzero result; since gθθ = Σ, this cancels the last term. Therefore,
duθ 1 1 1
= gµν,θ uµ uν = gµν,θ g αµ g βν uα uβ = − g αβ ,θ uα uβ , (23)
dτ 2 2 2
where in the last equality we have used the derivative of a matrix inverse. Given the prevalence of Σ in the denominator
of the inverse metric, it is easier to write this as
duθ 1 1
Σ = − (Σg αβ ),θ uα uβ − Σ,θ , (24)
dτ 2 2
where we recall the normalization condition g αβ uα uβ = −1. This relation simplifies if we recall that
Σ,θ = −2a2 sin θ cos θ (25)
5
and
duθ uθ duθ 1 d(uθ )2
Σ = = , (26)
dτ dθ/dτ dτ 2 dθ
so that
d(uθ )2
= −(Σg αβ ),θ uα uβ + 2a2 sin θ cos θ. (27)
dθ
Now for the miracle: of the metric coefficients (rescaled by Σ), only Σg tt and Σg φφ depend on θ, and the derivatives
do not depend on r:
d(uθ )2 d d
= −a2 E 2 sin2 θ − L2 csc2 θ + 2a2 sin θ cos θ. (28)
dθ dθ dθ
Therefore, one may integrate to get
(uθ )2 = −a2 E 2 sin2 θ − L2 csc2 θ + a2 sin2 θ + constant. (29)
Using the Pythagorean identities sin2 θ = 1 − cos2 θ and csc2 θ = 1 + cot2 θ, and noting that E and L are constant, we
may write this as
(uθ )2 = −a2 (1 − E 2 ) cos2 θ − L2 cot2 θ + Q, (30)
where Q is the Carter constant. The Carter constant is an accidental conserved quantity of trajectories in the Kerr
spacetime. It is obvious that for orbits with |E| ≤ 1 (which will include all the bound orbits), we must have Q ≥ 0.
In Eq. (30), for the bound orbits (|E| < 1) the right-hand side achieves a maximum value of Q at θ = π/2, and
decreases to −∞ at θ = 0 or θ = π. It crosses zero at θ = π/2 ± I, where I is the coordinate-inclination of the orbit.
(Note: there are several other definitions of “inclination” in use in the Kerr spacetime.) Thus the particle is confined
to the region within I of the equator, and it bounces back and forth between θ/2 − I and θ/2 + I. The Carter constant
is related to the inclination via
Q = a2 (1 − E 2 ) sin2 I + L2 tan2 I. (31)
In the Newtonian limit, 1 − E 2 is small and L is large, so we have tan I ≈ Q1/2 /L. Note that for retrograde orbits,
L < 0 and tan I < 0. The only orbits that can reach the poles (θ = 0, π) have L = 0 and finite Q.
We may now attempt to solve for the θ coordinate along the particle’s trajectory. We see that
dθ
q
= uθ = Σ−1 −a2 (1 − E 2 ) cos2 θ − L2 cot2 θ + Q. (32)
dτ
We unfortunately cannot integrate this to give θ(τ ) because Σ depends on r as well as θ. We can however define a
reparameterization of the particle trajectory from the proper time τ to the rescaled coordinate
dτ
Z
λ= . (33)
Σ
[Warning: This is not an affine parameter!] Then we have
θ
dθ
Z
p = λ − cθ , (34)
π/2−I −a (1 − E ) cos2 θ − L2 cot2 θ + Q
2 2
where cθ is a constant. The integral on the left-hand side is an elliptic function, but we will not evaluate it here. In
the Newtonian (or the Schwarzschild) limit where a can be dropped, it becomes L−1 arccos(cos θ/ sin I) so that
cos θ ≈ sin I cos[L(λ − cθ )] (a ≈ 0); (35)
the more general case is qualitatively similar. In all cases, θ is a periodic function of λ with some period Pθ (which is
2πL−1 in the Schwarzschild limit).
6
B. The radial motion
We are next interested in the radial motion of a particle. To find this, we take the normalization condition Eq. (21)
and substitute Eq. (30):
−(r2 + a2 )2 E 2 + a2 ∆E 2 + 4aM rLE + ∆L2 − a2 L2 − a2 ∆ cos2 θ + ∆Q ∆(ur )2
−1 = + . (36)
∆Σ Σ
Multiplying through by ∆Σ gives −∆Σ = −∆(r2 + a2 cos2 θ) on the left-hand side; moving this to the right then gives
0 = ∆r2 − (r2 + a2 )2 E 2 + a2 ∆E 2 + 4aM rLE + ∆L2 − a2 L2 + ∆Q + ∆2 (ur )2 . (37)
Expanding the ∆s and algebraically simplifying gives
−∆2 (ur )2 = (1 − E)2 r4 − 2M r3 + [a2 (1 − E 2 ) + L2 + Q]r2 − 2M [(aE − L)2 + Q]r + a2 Q ≡ V (r). (38)
Here V (r) is a fourth-order polynomial in r; the particle is trapped in the regions where it is negative. Clearly V (r)
becomes large for positive r, at least for bound (|E| < 1) orbits, and it is positive for r = 0 (or zero for a nonspinning
hole a = 0 or equatorial orbit Q = 0). Descartes’s rule of signs tells us that there can be no negative zeroes of V (r)
since all signs alternate, and finally by inspection of the intermediate step Eq. (37) we have V (r+ ) ≤ 0. Thus there
are two possibilities: there is either 1 zero of V (r) between r+ and ∞, or there are 3. In the former case, the allowed
region in r spans from below the horizon to some rmax , and all particles on such orbits fall into the hole. There are no
stable orbits for these values of (E, L, Q). In the latter case, there is a bounded trapped region where V (r) is negative
between two roots rmin and rmax . It is in the latter case that stable orbits are possible.
We can construct the particle’s trajectory by noting that
dr dr dτ p
= = Σur = ∆ur = −V (r), (39)
dλ dτ dλ
and so
Z r
dr
p = λ − cr , (40)
rmin −V (r)
R rmax p
where cr is a constant. Again the function r(λ) is periodic, with period Pr = 2 rmin dr/ −V (r). This is also an
elliptic integral.
C. The longitude and time motions
The remaining two motions, φ(λ) and t(λ), can be obtained from the inverse-metric:
2
(r + a2 )2
dt t 2 2 2aM rL
= Σu = − a sin θ E − (41)
dλ ∆ ∆
and
a2
dφ t 2 2aM rE
= Σu = csc θ − L+ . (42)
dλ ∆ ∆
These functions, while complicated, are the sums of functions of θ and r (and constants of the motion). Therefore
there are average values (over λ), bt = hdt/dλiλ and bφ = hdφ/dλiλ . The coordinate time is then
t = ct + bt λ + Dtr (λ) + Dtθ (λ), (43)
where Dtr is periodic with period Pr and Dtθ is periodic with period Pθ . Similarly one may write
φ = cφ + bφ λ + Dφr (λ) + Dφθ (λ), (44)
where Dφr is periodic with period Pr and Dφθ is periodic with period Pθ .
We conclude that the motion is regular when measured with the parameter λ, and that the particle proceeds to
exhibit oscillatory motions superposed on the mean march forward in time (at rate bt ) and longitude (at rate bφ ).
The orbits are confined to a certain range in θ and r, and are integrable (in the sense of ordinary classical mechanics).
In the next lecture, we will examine this motion in greater detail, and evaluate such quantities as precession rates,
the energies and angular momenta of circular orbits, and stability.
