Part C Major Option Astrophysics

High-Energy Astrophysics
Garret Cotter
garret@[Link]
Office 756 DWB
Michaelmas 2012 Lecture 7
Today’s lecture: Accretion Discs Part I

• The Eddington luminosity and accretion rate.

• Accretion discs.

• Properties of the thin accretion disc.

• (Evidence for accretion onto black holes - will start if time

permits).
The Eddington Luminosity

Today we will consider the accepted mechanism for the production

of the extreme luminosities of active galaxies: accretion of material
onto supermassive black holes. First we will consider the
Eddington accretion limit.

The Eddington luminosity was introduced in the context of massive

stars. The notion is very simple: for any object in the depths of
space, there is a maximum luminosity beyond which radiation
pressure will overcome gravity, and material outside the object will
be forced away from it rather than falling inwards.
Sir Arthur Stanley Eddington (1882-1944)

Journalist: “Sir Arthur, it is said that only

three people in the world understand rela-
tivity!”

Eddington: “Yes I’ve heard that. I am trying

to work out who the third person is...”
The ingredients we need to derive the Eddington luminosity are:

• The mass of the central object, M

• The total luminosity, L

• A suitable opacity for radiation pressure against any

surrounding material. We will see that this must have the
dimensions of area per unit mass.
 Small cloud of "stuff"
Distance R from the radiating object
Cloud has mass m

Luminous object with mass M and luminosity L

We wish to find the luminosity at which the gravitational force
inwards balances with the radiation force outwards. The
gravitational force is given simply via:
GM m
Fgrav =
R2
To calculate the radiation force first we need to get the radiation
pressure at R:
L 1
Prad =
c 4πR2
Then to calculate the radiation force on the cloud, we need its
opacity, κ. Radiation pressure is force per unit area; opacity is the
cross-sectional area per unit mass for radiation scattering.

Frad = Pradκm
Balancing the two forces gives:

GM m Lκm
=
R2 4πR2c

And solving for this luminosity we get:

4πGM c
L=
κ
Some important things to note here are:

• The Eddington luminosity depends only on the mass of the

radiating object.

• We have assumed spherical symmetry.

Now, we have yet not specified a particular value for κ. In

high-energy accretion scenarios we make a useful approximation
on the basis that the accreting material is mostly ionized hydrogen
and the opacity is provided by Thomson scattering. The
cross-section will then come almost exclusively from radiation
pressure on the electrons, but the mass lies almost exclusively in
the protons.
There will still be electrostatic forces between the e−s and the p+s;
if we exert a radiation force on the cloud that is felt mostly by the
electrons, they will drag the protons along with them. Thus the
approximation is to set κ = σT/mp, and we get the following
approximation for the Eddington luminosity:

4πGM cmp
LEdd = σT

Note that this approximation for κ is not valid in all situations,

especially in stars of all but the highest masses. E.g. in low mass
stars the opacity follows Kramer’s Law, κ ∝ ρ/T 3.5.
Eddington accretion limit

This becomes interesting when the luminosity of the central object

is derived from matter falling into it. In accretion onto compact
objects, infalling matter travels deep into the gravitational potential
well of the central object. If it is possible to turn the GPE of the
infalling material into heat, huge luminosities can result.

First, however, let us consider the consequences for the limiting

rate at which such accretion can occur. Suppose a compact object
is accreting mass from its surroundings at a rate Ṁ . Next assume
that some fraction of the GPE can be radiated away. If we express
this as a fraction  of the rest-mass energy, then the luminosity
radiated away becomes
 L = Ṁ c2

This has a profound implication. If our accreting object radiates at

more than the Eddington luminosity, even a glut of “fuel” will be
blown away by radiation pressure: we get a natural feedback
process with a limiting accretion rate. We derive this by setting the
accretion luminosity equal to the Eddington luminosity:

4πGM cmp
Ṁ c2 =
σT

From which the limiting Eddington accretion rate is:

ṀEdd = 4πGM
cσ
mp
T
Accreting objects in practice

The rather unsubtle catch with this is that we have made no

attempt to estimate . In principle it could be as low as zero: if we
simply drop a brick radially into a black hole, it will disappear over
the event horizon taking all its energy with it.

However in a realistic astrophysical situation, accreting matter will

have angular momentum, forming an accretion disc. We will later
calculate the canonical value of  which is used for black hole
accretion, and estimate the associated temperature and luminosity
of the accreting material just before it disappears over the horizon.

First we will recap on the essential properties of black holes, then

consider the conditions in the accretion disc.
Properties of black holes

In GR any point mass is described by the Schwarzschild (static) or

Kerr (rotating) metrics. We use the term black hole to describe an
object sufficiently compact (for its mass) that its event horizon has
noticeable effects on spacetime and matter around the object. For
a point mass M , the Schwarzschild radius is

RS = 2GM
c2

Note that this scales only with mass; RS for a one-solar-mass

object is 3 km. Formally this radius is called the event horizon
because it is the furthest distance that a photon starting inside
RScan reach; and once photons or matter from outside RS pass
beyond it, they cannot escape.
The event horizon is not a solid surface; matter falling inwards
passes straight through it. An external observer sees any light
emitted by the infalling object becoming infinitely redshifted as the
object passes over the horizon.

However, due to spacetime curvature near the horizon it is not

possible to have a stable circular orbit near the horizon. The last
stable orbit is at r = 3RS. For this part of the course we will take
this result as given.

Rotating black holes have a more complicated horizon and the last
stable orbit is closer in. This may be very important for accretion
onto real black holes.
Accretion discs

The key to the extraction of energy by material falling into a black

hole is to remember that in real astrophysical situation is will have
significant angular momentum. Gas falling onto a black hole in a
binary star system will start with the tangential speed of the binary
orbit; gas falling onto the central black hole of an active galaxy may
well have an initial tangential velocity of hundreds of km s−1 if it
begins its descent from the outer regions of the galaxy. A
broad-brush picture is:
• An initially large cloud of gas extending well beyond the compact object will, if it
has net angular momentum, tend to flatten into a disc. This is because collisions
between particles in a direction parallel to the angular momentum L vector will
tend to sum to zero, whereas collisions perpendicular to the L direction will tend
to maintain the average circular velocity.

• As the disc becomes sufficiently dense, viscosity inside it both transfers

angular momentum outwards and heats the disc. This is how the GPE of the
infalling material is radiated away.

• In the most well-studied model, the disc is assumed to be physically thin and
optically thick. This allows the maximum amount of heat to radiate away from the
surface of the disk before matter falls into the black hole.

• Eventually we arrive at a reasonably stable state where matter spirals in

through the disc, losing angular momentum via friction on its way in, becoming
hotter and hotter, until it falls off the inside edge (at the last stable orbit) and
crosses over the horizon.
Properties of the thin accretion disc
First of all, the disc must be in hydrostatic equilibrium. In the z direction
perpendicular to the plane of the disc we must satisfy
dP
= −ρgz
dz
where gz is the vertical component of the gravitational acceleration due to the
central object,
GM z
gz ≈
R2 R
(using small-angle approximation). We can relate P and ρ via the sound speed
in the gas, dP = c2s dρ, and then integrate to find that the density in the disc falls
off exponentially with height
2
 
z
ρ(z) = ρ0 exp − 2
2h
with a height scale factor given by h2 = c2s R2 /GM .
Next we consider the speed of rotation. The particles in the disc
will have orbits very close to Keplerian, so

2 = GM
vrot .
R
The scale height can be re-written in terms of the rotational velocity
via
2 c2
sR
2
h = 2
vrot
so if we have R  h we must have vrot  c2 s and so the rotation
of the disc is highly supersonic. The same condition applies to
keep cold gas in the plane of the Milky Way with a rotational
velocity of ≈ 200 km s−1.
Viscosity in the disc

For the accreting material to fall into tighter orbits in the disc there must be an
outwards flow of angular momentum—a torque acting on the disc. Take the disc
viscosity to be η and consider a radius r in the disc with thickness t and angular
velocity ω. The tangential force per unit area exerted by the disc inside r on the
disc outside r is given by
dω
F = ηr .
dr
This force acts over an area 2πrt so the total torque between adjacent pieces of
the disc is
dω
Γ = 2πr3 tη .
dr
Remember torque is rate of change of angular momentum!
 q
GM
Again taking the orbits to be Keplerian we have ω = r2
, so subbing into the
previous equation for the torque we have

dL
= −3πηt(GM r)1/2
dt

which is the rate of change of angular momentum of the inner piece of the disc.
This must equal the change of angular momentum duw to inflow of disc material,
i.e.

dL
= ṁr2 ω = ṁ(GM r)1/2
dt

and so we have a relationship between the accretion rate and the disc viscosity,
ṁ = 3πηt
The problem with viscosity... The problem arises when we consider the
Reynolds number of the material in the disc—a measure of how turbulent it is.
VL
R∼
ν
Here V and L are characteristic speed and length scales and ν is the kinematic
viscosity, η/ρ. We find (see e.g. Longair pp 145–146) that R ∼ 1012 . The flow is
highly turbulent, and so standard kinetic theory dynamic visosity η = 13 ρcλ will
make a negligible contribution.

We do not yet understand the precise mechanism of viscosity in accretion discs.

Highly turbulent flow helps, but precise calculations are difficult. Magnetic fields
will be present and will certainly contribute. Much of the progress to date has
come from a neat side-step developed by Shakura and Sunyaev (1972). They
invented a parameter
ν
α=
hcs
which allows detailed models to be made without knowing the exact mechanism
for the viscosity.
Luminosity of a thin accretion disc

Neglecting the energy transport due to viscosity, we can calculate the rate at
which accreting material in the disc must lose gravitational potential energy if it is
to fall closer to the accreting object. For an annulus between r and r + dr, the
energy which must be dissipated will be
 
dE GṀ M
L(r) = − = dr
dt 2r2
where Ṁ is the accretion rate and M is the mass of the central object. Including
viscous energy transport we gain a total luminosity three times this value
(non-examinable—see Longair pp 149-150).
Temperature structure of a physically thin, optically thick disc

If the disc is optically thick, each annulus between r and r + dr will radiate as a
blackbody with the luminosity derived above. Hence via Stefan’s Law
(remembering the disc has two surfaces), the annulus at at r will radiate with
2σT 4 × 2πrdr. Thus

3GṀ M
σT 4 =
8πr3

and so
 1/4
T (r) = 3G Ṁ M
8πr3σ
Spectrum of the thin disc

We are now in a position to describe the form of the overall spectrum of the disc,
i.e. the sum of all the black-body contributions at different radii.
Z router
Iν ∝ 2πrBν {T (r)} dr
rinner

where from lecture 1 we have

 −1
3 hν/kT
Bν ∝ ν e −1
From before we have T ∝ r−3/4 so dr ∝ (1/T )1/3 d(1/T ) and we can integrate
dT instead of dr. Including Bν explicitly we therefore have

Z router  4/3  −1  1 1/3  1 

1
Iν ∝ ν 3 ehν/kT − 1 d
rinner T T T

We can proceed by changing variable x = (hν/kT )—recall from the

second-year thermo problem set where you used the same substitution to derive
the functional form u(T ) ∝ T 4 . This yields

xouter
ν3
Z
Iν ∝ 8/3 x4/3 (ex − 1)−1 x1/3 dx
ν xinner
The integral dx is just a numerical constant so we now have the shape of the
spectrum over most of its range:

Iν ∝ ν 1/3 .

Note, though, that the low- and high-frequency ends will have a different form:

• From the outer edge of the disc we will see the Rayleigh-Jeans tail of
Touter , Iν ∝ ν 2 .

• From the inner edge, an exponential cut-off Iν ∝ e−hν/kTinner .

Theoretical spectrum of thin accretion disc.
Total luminosity of the thin disc

Let’s now estimate . If we approximate with a Newtonian potential, take the

particle to have started its trip at r = ∞ and total energy zero, and calculate the
total energy it has on the last stable orbit. The amount of GPE which must be
lost (by radiation) is equivalent to 1/12 of the rest-mass energy of the particle.

For the best possible case—the closest orbit around a rapidly-rotating black
hole—the efficiency rises to a whopping 0.42. Compare this to nuclear fusion in
stars, which has an efficiency of only 0.7 percent!

Hence in practice, astronomers usually adopt an approximate value of  = 0.1

for accretion onto black holes.

Example: estimating a quasar accretion rate

Suppose we observe a quasar to have a total power output of 1040 W. We are

now in a position to estimate the mass of the central black hole and the rate at
which its mass is increasing.

First let us assume that the accretion is Eddington limited. From our equation for
the Eddington luminosity we have
 4πGM cmp
LEdd =
σT

from which
M = 7 × 108 M .

And from our Eddington accretion rate, using  = 0.1 we have

4πGM mp
ṀEdd = ≈ 3M yr−1
cσT
Getting round the Eddington limit

The accretion may not always be Eddington-limited. It is, for example, possible to
achieve Ṁ much greater than would be inferred by using the Eddington
luminosity with  = 0.1, by making the disc physically thick, and very low density,
so that it is optically thin and matter doesn’t have time to radiate away so much
energy before it falls over the horizon. This has the advantages of allowing black
holes to grow at a very high rate in the early Universe, and also of providing
“funnels” which could be a mechanism for collimating outflows from accreting
objects.

Unfortunately simple analytical models of these discs are unstable, but the
advantages of thick discs are so great that much effort is put into modelling them
numerically... including the effects of strong magnetic fields.
It is also possible for an object to have a luminosity significantly greater than the
Eddington luminosity:

• In supernovae (somewhat trivially!)

• Where spherical symmetry is broken, with extremely collimated radiation in

a direction different to the accretion direction

• Where accretion is not steady, e.g. bursts or radiation emitted when

discrete clouds of matter fall onto a neutron star or white dwarf.
Evidence for black holes in AGN
There are several canonical pieces of evidence that supermassive black holes
really are there at the heart of AGN. Among these are:

• Variability (in combination with Eddington luminosity).

• Stellar velocity dispersions.

• Rotation speeds inferred from emission lines.

• The controversial history of X-ray line profiles.