You are on page 1of 56

Theoretical Astrophysics

Course Notes 2016

Martin Pessah
mpessah@nbi.ku.dk
2
3

Preface

This is the second version of the set of notes for the course Theoretical Astrophysics taught at the Niels Bohr
Institute in the Fall of 2016.
This course provides an overview of some of the most important astrophysical processes that shape the evolution,
and observational properties, of astrophysical systems, from planets to stars, and from supermassive black holes
to entire galaxies. The course is strongly recommended for all students starting at the M.Sc. and Ph.D. levels in
preparation for their further study and research in any area of astrophysics, including planetary sciences and cosmol-
ogy. This course will provide students with a wide range of interests in observational, theoretical, or computational
astrophysics with a valuable toolkit to become more competent researchers. We will cover the basic equations, learn
how to solve them, and understand their implications. The competences acquired in this course are a valuable
complement to those obtained in observational and phenomenological astrophysics courses and are an indispensable
asset for students wishing to pursue studies in any branch of astrophysics.
The content of this course draws from a broad variety of sources, including the courses "Topics on Theoretical
Astrophysics" and "High Energy Astrophysics" that I took as a student in the University of Arizona in the years
2001 and 2002 with Prof. Adam Burrows and Prof. Fulvio Melia, as well as select sections from several books as
mentioned throughout the notes and referenced in the bibliography. These notes are not intended to be a substitute
to any of these books but are merely a guide to introduce you to some of the key physical processes that are relevant
in astrophysics. Having said this, in defining the topics for the course, I have done my best to select the core set of
concepts that I believe any student interested in astrophysics should be acquainted with. In defining the material in
these notes, I have done my best to provide a relatively self-contained set of lectures. It is remarkably difficult to be
consistent with notation when drawing from a large set of resources which use different conventions. The notation
should be consistent at least within each class and hopefully the context will be clear enough to minimize any possible
ambiguity.
I am grateful to Oliver Gressel, Tobias Heinemann, Christian Brinch, and Neil Vaytet, who gave guest lectures
and contributed with the respective set of notes. I am also thankful to Thomas Berlok, Gopakumar Mohandas, and
Michael Kffmeier, who were teaching assistants for this course, and were instrumental in making this set of notes a
reality. I am committed to improving this set of notes, so please let me know if you have any comments or suggestions
as they will be most welcome.

Martin Pessah
Niels Bohr Institute
4
5

CONTENTS 5.4.1 Gaseous self-gravitating system . . 22


5.4.2 Ideal monoatomic gas . . . . . . . 22
5.5 Gravitational collisions and relaxation . . 23
1 Order of Magnitude Astrophysics 7
5.6 Dynamical friction . . . . . . . . . . . . . 23
1.1 Physical constants, Astronomical units . . 7
1.2 Dimensional Analysis . . . . . . . . . . . . 7 5.7 Self-gravitating barotropic fluids . . . . . 24
1.2.1 Period of a Pendulum . . . . . . . 7 5.8 Lane-Emden equation polytropes . . . . 24
1.2.2 Units of the Gravitational Constant 7 5.8.1 Boundary conditions for the Lane-
1.2.3 Characteristic Scales in Black Holes 7 Emden equation . . . . . . . . . . 24
1.2.4 Coulomb Force . . . . . . . . . . . 8 5.9 Dynamical stability of self-gravitating disk 25
1.2.5 Alfvn Speed . . . . . . . . . . . . 8 5.9.1 Steady state background . . . . . . 25
1.2.6 Free fall time . . . . . . . . . . . . 8 5.9.2 Linear Mode Analysis . . . . . . . 25
1.3 Characteristic Scales - Harmonic Oscillator 8
1.4 Back-of-the-Envelope Estimates . . . . . . 8 6 Basics of Magnetohydrodynamics 27
1.4.1 Speed of Earth at Equator . . . . 8
6.1 Magnetohydrodynamics via Hydromagnetics 27
1.4.2 Galactic year . . . . . . . . . . . . 9
1.4.3 Keplerian Orbits . . . . . . . . . . 9 6.2 Ideal MHD . . . . . . . . . . . . . . . . . 28
6.3 Energetics in MHD . . . . . . . . . . . . . 28
2 Basics of Fluid Dynamics 10
2.1 Lagrangian derivative . . . . . . . . . . . 10 7 Magnetohydrodynamic Waves 30
2.2 Equations for fluid dynamics . . . . . . . 10 7.1 Flux freezing . . . . . . . . . . . . . . . . 30
2.3 Enthalpy . . . . . . . . . . . . . . . . . . 11 7.2 The plasma parameter . . . . . . . . . . . 31
2.4 Mass Conservation . . . . . . . . . . . . . 11
7.3 Magnetic field-line tension . . . . . . . . . 31
2.5 Energy Conservation . . . . . . . . . . . . 11
7.4 Basic MHD waves . . . . . . . . . . . . . 32
2.6 Plane-Parallel Atmospheres . . . . . . . . 11
2.7 Sound Waves . . . . . . . . . . . . . . . . 12 7.4.1 Alfvn waves . . . . . . . . . . . . 33
7.4.2 Slow and fast magnetosonic waves 33
3 Hydro Instabilities 13
3.1 Linear Mode Analysis for Sound Waves . 13 8 Radiation 35
3.2 Bernoulli Equation and Barotropic Fluids 14 8.1 Solid Angle . . . . . . . . . . . . . . . . . 35
3.3 Incompressible and Irrotational Flows . . 14 8.2 Attenuation of Radiation . . . . . . . . . 35
3.4 Viscous Dissipation and Reynolds Number 14
8.3 Diffusion of Particles and Photons . . . . 36
3.5 Hydrodynamic Instabilities . . . . . . . . 15
8.4 Specific Intensity . . . . . . . . . . . . . . 36
3.5.1 Kelvin-Helmholtz Instability . . . 15
3.5.2 Rayleigh-Taylor instability . . . . . 15 8.5 Black Body Radiation . . . . . . . . . . . 37

4 Shocks 17 9 Radiative Transfer 38


4.1 Shock waves in astrophysics . . . . . . . . 17 9.1 The Equation of Radiative Transfer . . . 38
4.2 Conservative form of fluid equations . . . 17 9.2 Diffusion Approximation for Radiative
4.3 Shock frame vs. lab frame . . . . . . . . 18 Transfer . . . . . . . . . . . . . . . . . . . 38
4.4 Rankine-Hugoniot jump conditions . . . 18 9.3 Thermalization Time . . . . . . . . . . . . 39
4.5 Properties of shocks . . . . . . . . . . . . 18
9.4 Thermal Structure of a Grey Atmosphere 39
4.6 Phases of supernova remnant evolution . 19
4.6.1 Free expansion/constant velocity
10 Radiative Processes 40
phase . . . . . . . . . . . . . . . . 19
4.6.2 Adiabatic, energy-conserving/ Se- 10.1 Larmor Formula . . . . . . . . . . . . . . 40
dov phase . . . . . . . . . . . . . . 20 10.2 Cyclotron Radiation . . . . . . . . . . . . 40
4.6.3 Radiative/momentum conserv- 10.3 Synchrotron Radiation . . . . . . . . . . . 41
ing/snowplow phase . . . . . . . . 20 10.4 Scattering of Electromagnetic Waves by
non-relativistic Particles . . . . . . . . . . 42
5 Gravitational Dynamics 21 10.4.1 Thomson Scattering . . . . . . . . 42
5.1 The gravitational potential . . . . . . . . 21
10.4.2 Rayleigh Scattering . . . . . . . . 43
5.2 Gravitational binding energy . . . . . . . 21
5.3 Virial theorem for N-body system . . . . . 21 10.4.3 Resonant Scattering of Line Radia-
5.3.1 Approximately steady state I ' 0 22 tion . . . . . . . . . . . . . . . . . 43
5.3.2 Time-averaged virial theorem for a 10.5 Compton Scattering . . . . . . . . . . . . 43
bound system . . . . . . . . . . . . 22 10.6 Inverse Compton Scattering and SZ-effect 44
5.4 Virial theorem for self-gravitating gases . 22 10.7 Thermal Bremsstrahlung Radiation . . . . 44
6

11 Spherical accretion 45 List of Figures


11.1 Steady isentropic flow of gas . . . . . . . . 45
11.2 Maximum Flow Speed . . . . . . . . . . . 45 3.1 Kelvin-Helmholtz Instability. . . . . . . . 15
11.3 Sonic Point . . . . . . . . . . . . . . . . . 45 3.2 Kelvin-Helmholtz Instability dissipation. . 15
11.4 de Laval Nozzle . . . . . . . . . . . . . . . 46 3.3 Rayleigh-Taylor instability. . . . . . . . . 16
11.5 Subsonic and Supersonic Flows . . . . . . 47 4.1 Steepening of wave into shock. . . . . . . 17
11.6 Spherical Accretion of a Polytropic Gas . 48 4.2 Receding shock wave. . . . . . . . . . . . 18
11.7 Bondi Accretion . . . . . . . . . . . . . . 49 4.3 Expanding blast wave. . . . . . . . . . . . 19
11.8 Asymptotic Properties of Transonic Flows 49 4.4 Phases of supernova blast wave. . . . . . . 20
11.9 Accretion Luminosity . . . . . . . . . . . . 50 5.1 Solutions to Lane-Emden equation. . . . . 24
11.10The Eddington Limit . . . . . . . . . . . . 50 5.2 Stability of self-gravitating disk. . . . . . 26
7.1 Magnetic flux. . . . . . . . . . . . . . . . . 30
12 Disk accretion 51 7.2 Flux freezing. . . . . . . . . . . . . . . . . 30
12.1 Introduction . . . . . . . . . . . . . . . . . 51 7.3 Magnetic tension. . . . . . . . . . . . . . . 32
12.2 On Angular Momentum Transport . . . . 51 7.4 MHD wave speeds. . . . . . . . . . . . . . 34
12.3 Energetics . . . . . . . . . . . . . . . . . . 52 8.1 Differential of solid angle. . . . . . . . . . 35
12.4 Temperature at the Disk Midplane . . . . 52 8.2 Finite solid angle. . . . . . . . . . . . . . . 35
12.5 Nature of Disk Viscosity . . . . . . . . . . 52 8.3 Particle diffusion. . . . . . . . . . . . . . . 36
12.6 Viscous force per unit length and Stress . 53 8.4 Intensity in radiative transfer. . . . . . . . 37
12.7 Time Scales in Accretion Disks . . . . . . 53 9.1 Slab geometry. . . . . . . . . . . . . . . . 38
12.8 The structure of the Standard Disk Model 54 10.1 Electromagnetic fields in the radiation zone 40
12.9 The magnetorotational instability . . . . . 54 10.2 Cyclotron radiation . . . . . . . . . . . . . 40
12.9.1 The shearing box approximation . 54 10.3 Synchrotron radiation . . . . . . . . . . . 41
12.9.2 MRI dispersion relation . . . . . . 55 10.4 Synchrotron spectrum for one particle. . . 41
10.5 Non-thermal synchrotron spectrum. . . . 42
10.6 Compton scattering . . . . . . . . . . . . 43
10.7 Compton cross section. . . . . . . . . . . . 43
10.8 Thermal Bremsstrahlung. . . . . . . . . . 44
11.1 Isentropic flow of gas. . . . . . . . . . . . 45
11.2 Sketch of the de Laval nozzle. . . . . . . . 46
11.3 Subsonic and supersonic flows. . . . . . . 47
11.4 Nozzle coupled to a diffuser. . . . . . . . . 47
11.5 Maximum mass flux in isentropic flow. . . 48
11.6 Isentropic, spherical winds and accretion. 49
12.1 Thin accretion disk. . . . . . . . . . . . . 51
12.2 Viscous angular momentum transport. . . 51
12.3 Thin accretion disk radiation. . . . . . . . 52
12.4 Shearing box approximation. . . . . . . . 55
12.5 Magnetorotational Instability. . . . . . . . 55
7

1 ORDER OF MAGNITUDE ASTROPHYSICS and thus


+ = 0 = , (1.2)
1.1 Physical constants, Astronomical units
1 + 2 = 0 = 1/2 . (1.3)
When faced with a complex problem it is always useful to This implies that s
try to get a sense of what to expect so you will be able to L
assess whether the result makes any sense. Astrophysical P = . (1.4)
g
systems span a huge range of physical conditions, so get-
ting an answer within factors of a few is sometimes pretty
good! 1.2.2 Units of the Gravitational Constant
Order of magnitude calculations are not a substitute
for a detailed answer and some times the nature of the We can find the units of G from Newtons law
problem at hand does demand complex calculations or
GmM
numerical studies. However, before embarking in what ma = , (1.5)
could be a time consuming task it is always useful to get R2
a sense of what kind of result to expect. where m and M are the masses separated by a distance
When making quick estimates to have a rough idea it is r and a is the acceleration of m. In this case we find
very useful to know the most relevant physical constants.
In order to avoid confusion, it is a good idea to stick to L L2 L3
[G] = 2
= . (1.6)
a given unit system. A lot of people in astrophysics use T M MT2
cgs units, which is the system we will generally use in this
course. If you are consistent with the units you use, you 1.2.3 Characteristic Scales in Black Holes
will only need to remember the numbers characterizing
the physical constants in that unit system and there is no The Schwarzschild radius of a black hole is given by
need to learn them better than the first significant digit!
For example, the speed of light is c ' 3 1010 cm s1 , RBH = 2
GM
, (1.7)
the Planck constant is h ' 6 1027 erg s, etc. This c2
is a handy link: http://www.astro.wisc.edu/~dolan/
where G is the Gravitational constant, M is the black
constants.html
hole mass and c is the speed of light. Let us see if we can
You should also get comfortable with the most used "guess" this result using dimensional analysis.
units in astrophysics, e.g., solar mass, Gyr, astronomical We are seeking to derive a scale, with dimensions of
unit, light-yr, pc, solar luminosity, etc. and how they re- length. Let us assume that this quantity can be build as
late to cgs units. Knowing this will help you calculate a combination of G, M and c. In order to do this we need
quickly, for example, what does a velocity of pc/yr corre- to find constants , and such that
sponds to in km/s.
With some practice you should be able to carry out L = G M c , (1.8)
simple calculations in your head. For example, try to
calculate the Schwarzschild radius, i.e., R = 2GM/c2 of a such that the units are the same on the LHS and the RHS
neutron star in this way. This should give you something of the previous equation.
like R 4km. Now, using that R scales linearly with We find that
mass, you can estimate the Schwarzschild radius for a  3   
L L
solar mass black hole or the supermassive black hole at L= 2
M
= L3 M T 2 M L T ,
the galactic center. MT T
(1.9)
or
1.2 Dimensional Analysis L = L3+ M T (2+) . (1.10)
From this expression we can see that
1.2.1 Period of a Pendulum
1 = 3 + , (1.11)
We can estimate the period of a pendulum using dimen-
sional analysis. We expect the period to be shorter if the = 0, (1.12)
gravitational force acting on it is stronger. We also ex- 2 + = 0 . (1.13)
pect the period to be longer if the length of the pendulum
is larger. So lets start with P L g . Now, [g] = LT 2 Solving these equations we find = 1, = 1 and = 2
so we must have and so we conclude that
GM
L L= . (1.14)
L0 T 1 L (1.1) c2
T 2
8

1.2.4 Coulomb Force 1.3 Characteristic Scales - Harmonic Oscillator

You are probably mostly familiar with electromagnetism If the force acting on a point mass is proportional to the
in SI units. In these units the force between two charged displacement so that F = kx, then the equation of
particles is given by Coulombs law motion for a particle moving in one dimension is
1 q1 q2
F = , (1.15) d2 x
40 r2 m = kx , (1.23)
dt2
where q1 and q2 are the charges, r is the separation and
 is the vacuum permittivity. d2 x k
+ x = 0, (1.24)
In CGS units, the corresponding equation is simply dt2 m
q1 q2 d2 x
F = 2 . (1.16) + 02 x = 0 . (1.25)
r dt2
What is the unit of charge in CGS units? Answer: If we p
where 0 = k/m has dimensions of a frequency. If we
use the fact that force has units of [F ] = M L/T 2 , we can define as the characteristic length scale in this problem
show that L0 and the characteristic time scale 01 , we can define a
1/2
L3 set of dimensionless variables according to

2 1/2

[e] = M L/T L= M 2 . (1.17)
T
x0 = x/L0 , (1.26)
This result can be simplified slightly by writing it in
terms of the units of energy (think of kinetic energy) and
E = M L2 /T 2 . We then find t0 = t0 . (1.27)
1/2
[e] = (EL) . (1.18) The dimensionless equation of motion found by dividing
the original equation by L0 02
1.2.5 Alfvn Speed 1
 2
d x

2
+ 0 x = 0 , (1.28)
L0 02 dt2
The Alfvn speed, which is the speed at which Alfvn
waves propagate1 , is given by d2 (x/L0 ) 02 x
+ 2 = 0. (1.29)
B B d(t0 )2 0 L0
vA = . (1.19)
4 In the new variables, the equation is therefore simply
From this relation we can find the units of the magnetic
field. We can isolate B 2 d2 x0
02
+ x0 = 0 . (1.30)
2 2
dt
B = v , (1.20)
This dimensionless equation is the kind of equation that
and immediately see that you can solve in a computer! We will later encounter
 2 M L 2
E more complex examples in which the advantages of work-
B = 3 2 = 3, (1.21) ing with dimensionless variables are even greater. The
L T L
underlying principle, nevertheless, is always the same. If
where we recognized the E = M L2 /T 2 in the last step. you choose your scales wisely, the equations will look sim-
The intermediate step could have been skipped by seeing pler!
that v 2 /2 must have units of energy per volume.

1.4 Back-of-the-Envelope Estimates


1.2.6 Free fall time
1.4.1 Speed of Earth at Equator
We can use dimensional analysis to estimate the free-fall
time characterizing the gravitationally collapse of a homo- The radius of Earth is
geneous medium of density . It is not hard to argue that
the result should depend only on G and . Given that the R = 6 108 cm = 6000 km , (1.31)
units of are [] = M L3 and that [G] = L3 T 2 M 1 , it
is easy to see that [G] = T 2 . Therefore and the rotation period is t = 24 hr. Using this infor-
1 mation, let us estimate the speed of Earth at the equator
tff . (1.22)
G 2R 6000 km km
1 Alfvn waves will be properly introduced later in the course
v= = 2 1670 . (1.32)
t 24 hr hr
9

1.4.2 Galactic year

The speed of the Sun around the Galactic center is


roughly
km
v 230 , (1.33)
s
whereas its distance to it is about

R 8 103 pc , (1.34)

This implies that a "galactic year" is about


2R
t 225 106 yr . (1.35)
v

1.4.3 Keplerian Orbits

Let us estimate the Period of Jupiter around the Sun.


The gravitational force due to a point source of mass
M acting on a test particle of mass m is

v2 GmM
m = . (1.36)
r r2
We can write v = r, where is the angular frequency,
and rewrite
GM
= 2 r2 , (1.37)
r
or, solving for the angular frequency,
GM
2 = . (1.38)
r3
The orbital period is associated with this angular fre-
quency is
2
P = , (1.39)

and thus
4 2 3
P2 = r . (1.40)
GM
Therefore the ratio of periods for two planets orbiting the
same star is given by
 2  3
P2 r2
= (1.41)
P1 r1

If we use the fact that the Earth orbits at r1 = 1AU in


a period P1 = 1yr, then the period of Jupiter, which is
roughly at r2 = 5.2AU is roughly P2 11.8 yr.
10

2 BASICS OF FLUID DYNAMICS for the energy per unit volume, e , and the result
reads
2.1 Lagrangian derivative e
+ (ev) = P (v). (2.10)
t
The total change of a quantity as we move along with the The 5 equations of fluid dynamics listed so far contain 6
fluid is given by the Lagrangian derivative variables, namely , v, P and either s or . We therefore
need an equation of state, either P P (, s) or P
d
= + v (2.1) P (, ), that relates the pressure, P , to the density, ,
dt t and the entropy, s, or the internal energy, .
Think about the number density of particles: n(x, t) For an ideal fluid the pressure is given by

n n P = ( 1) (2.11)
dn(xi , t) = dxi + dt (2.2)
xi t where
If we think about how dn changes along the trajectory of = CP /CV (2.12)
the moving fluid element xi and vi are related via dxi = is the ratio of specific heats. Therefore, we can write the
vi t. Thus internal energy per unit mass as
 
n dxi n n n
 
1 P
dn = dt + dt = vi + dt (2.3) = . (2.13)
xi dt t xi t 1
and we find therefore that The pressure for an ideal gas is related to the density,
dn


 temperature and the mean molecular weight, , by
= + v n (2.4)
dt t kB T
P = (2.14)
mH
where the terms inside the square brackets are the rate of
change. where mH the mass of the Hydrogen atom.
Note that the mean molecular weight depends on the
composition and ionization state of the gas. If the gas is
2.2 Equations for fluid dynamics pure, neutral Hydrogen then = 1, if it is pure, fully ion-
ized Hydrogen then = 1/2, and if it is a mixture of fully
The equations of fluid dynamics for an ideal fluid are the ionized Hydrogen and Helium with primordial composi-
continuity equation tion then ' 0.6. If you merely need a rough estimate
of the pressure, then it is OK to use ' 1 for practical
+ (v) = 0 , (2.5) purposes.
t
Recall that for a monoatomic gas with three degrees of
the momentum equations freedom = 5/3 so that
v P 3 kB T
+ (v)v = , (2.6) = , (2.15)
t 2 mH
and the equation for the entropy per particle s S/N and each degree of freedom contributes with kB T /2.
The entropy per unit mass for fluid is
ds s
= + (v)s = 0 . (2.7) kB
dt t s= ln P (2.16)
1
The RHS of the entropy equation is only zero for an ideal
fluid where there is no dissipation. For an adiabatic process entropy is constant along stream-
The entropy equation can also be written as an equa- lines
tion for the energy per unit mass, , using the fundamen- ds
=0 (2.17)
tal thermodynamic relation dt
  which is equivalent to
1
T ds = d + P d , (2.8) P = constant (2.18)

After some algebra, we obtain Perturbations in the pressure and density propagate
with the speed of sound, c, is given by
d
+ P v = 0 , (2.9) P

P
dt 2
c = = . (2.19)
s
where T is the temperature. Alternatively, we can write
this equation for the energy per unit mass as an equation We will talk more about this later.
11

2.3 Enthalpy 2.5 Energy Conservation

The enthalpy is a thermodynamic potential which is use- We consider the change in kinetic energy v 2 /2
ful to consider for adiabatic processes. It is defined so  
that 1 2 1 v
P v = v 2 + v , (2.30)
h=+ (2.20) t 2 2 t t

and internal energy
or, substituting ,
 s
P =h + T . (2.31)
h= . (2.21) t t t
1
We add the two equations and obtain an equation for the
In terms of the sound speed, c, we have
change in the total energy v 2 /2 + 
2
c     
h= (2.22) 1 2 1 2
1 v +  = v v +h (2.32)
t 2 2
An incremental change in enthalpy can be written as
  Integrating over volume, as in the case of mass conserva-
1 dP
dh =  + P d + (2.23) tion, we find that

Z   Z  
Using the thermodynamic relation 1 2 1 2
v +  dV = v v + h dS .
  t V 2 S 2
1
d = T ds P d (2.24) (2.33)
The quantity  
we have 1 2
v v +h (2.34)
dP 2
dh = T ds + . (2.25)
is the energy flux.
If the change in entropy is zero (ds = 0), then the change
in enthalpy is
dP 2.6 Plane-Parallel Atmospheres
dh . (2.26)

Let us consider a plane parallel atmosphere where all
So in an adiabatic process, the change in enthalpy is just
quantities are constant along x and y but can vary with
the change in pressure divided by the density.
z. The gravitational potential can be taken to be

2.4 Mass Conservation = gz (2.35)

and the gas is standing still (v = 0). In this case the


In the absence of sources or sinks of matter, mass is con- momentum equation reduces to
served. This can be shown explicitly by integrating the
continuity equation P
= g . (2.36)
z
+ (v) = 0 , (2.27)
t When this equation is fulfilled the atmopshere is in hydro-
static equilibrium. The differential equation has a simple
over the volume, V . We have that
Z Z solution when the temperature and composition of the
gas is constant with z. In this case, since
dV = (v) dV . (2.28)
t V V
kB T
P = , (2.37)
Converting the volume integral on the RHS to a surface mH
integral and defining the mass enclosed in the volume to
be M , we find that the density profile has the solution

(z) = 0 emH gz/kB T = 0 ez/H .


Z
M (2.38)
= vdS . (2.29)
t S
Here, 0 is the density at z = 0 and we have defined the
Here v is the flux of mass per unit time per unit area scaleheight of the isothermal atmosphere to be
and the equation states that the change in mass contained
inside the volume, V , is due to a flux of mass through the kB T
H . (2.39)
surface, S. mH g
12

2.7 Sound Waves This is a wave equation for the density perturbation. Its
solutions are sound waves that propagate with speed c.
Starting from the continuity equation Let us take a perturbation of the form

= A sin(kx t) , (2.53)
+ (v) = 0 , (2.40)
t
where A is the amplitude. The derivatives with respect
and the momentum equation to time and space are
dv
= P (2.41) = A cos(kx t) (2.54)
dt t
we will derive the wave equation for sound waves. 2
We consider an equilibrium with v 0 = 0, 0 and P0 = A 2 sin(kx t) (2.55)
t2
constant in space and time with small perturbations ,
P and v. The quantities are then
= Ak cos(kx t) (2.56)
x
= 0 + , (2.42) 2
= Ak 2 sin(kx t) (2.57)
x2
P = P0 + P , (2.43)
So that the one dimensional wave equation
v = v . (2.44)
2 2
We plug these into the equations and find 2
= c2s , (2.58)
t x2
0
 + + (0 v + 
v)
 =0,
 (2.45) gives
t t
A 2 sin(kx t) = Ak 2 c2s sin(kx t) . (2.59)
 
v Simplifying we find the dispersion relation for sound
P
 P

(0 + ) +
(vv)
 =  (2.46)
t waves
2 = c2s k 2 . (2.60)
The perturbations are small P/P  1, /  1 which
means that we can ignore all terms that are more than The dispersion relation is a relation between the scale of
linear in the perturbations. We have crossed out all the the perturbation and the frequency.
terms that can be ignored due to being small or zero. The
linearized equations are then


+ 0 v = 0 . (2.47)
t
v
0 = P . (2.48)
t
Let us now differentiate the equation for with respect
to time. We find
2 v
+ 0 =0. (2.49)
t2 t
We can then use the linearized momentum equation to
substitute for tv , yielding

2
= 2 P . (2.50)
t2
Now, we can use the relation between pressure and den-
sity
P = c2 , (2.51)
where c is the soundspeed, to find

2
= c2 2 . (2.52)
t2
13

3 HYDRO INSTABILITIES where v k,j (t) are the eigenvectors associated with the ma-
trix M and j are the roots of the characteristic polyno-
3.1 Linear Mode Analysis for Sound Waves mial associated with the matrix M and are in general
complex scalars.
Let us start with the linearized equations, Equations The advantage of working in this base is that the equa-
(2.47) and (2.48), but specializing to one spatial dimen- tions are decoupled, i.e.,
sion, x, and setting P = c2s In this case we have
k (t)
vx = j k (t) , (3.13)
+ 0 =0, (3.1) t
t x
vk (t)
vx = j vk (t) , (3.14)
0 = c2s . (3.2) t
t x and the solutions are exponentials
Before proceeding we make these equations dimensionless
by using the background value, 0 , the characteristic scale k (t) = k (0) ej t , (3.15)
L, the soundspeed cs and the characteristic time t0 =
L/cs . In terms of the dimensionless parameters vk (t) = vk (0) ej t . (3.16)
x vx t Note that here we have omitted the index j in the eigen-
0 = x0 = vx0 = t0 =
(3.3) vectors for the sake of brevity.
0 L csL/cs
Now how do we find the eigenvalues j ? Given our
the equations are then previous equations we need to find j such that
0 vx0 
0 ik

k (t)
 
j 0

k (t)

+ = 0 , (3.4) = ,
t0 x0 ik 0 vk (t) 0 j vk (t)
vx0 0 (3.17)
+ = 0 . (3.5) or, moving everything to the LHS,
t0 x0
We drop the primes on the dimensionless variables in the 
j ik

k (t)

following. The main idea of a linear mode analysis is to =0. (3.18)
ik j vk (t)
now assume that the perturbations depend on space and
time as X You should remember from linear algebra that such a lin-
(x, t) = k (t)eikx (3.6) ear, homogeneous system only has non trivial solutions
k (that is a solution that is not just vk = k = 0) if the
(3.7) determinant of the matrix is zero. Taking the determi-
X
ikx
vx (x, t) = vk (t)e
k
nant, we find what is called the characteristic polynomial
where k is the wavenumber. Substituting this into our j2 + k 2 = 0 , (3.19)
linearized equations we obtain
X  k (t)  which has solutions
+ ikvk (t) eikx = 0 (3.8)
t j = ik . (3.20)
k

X  vk (t) 
If only one mode is excited, and we restore the physical
+ ikk (t) eikx = 0 (3.9)
t dimensions, e.g., = kc then the perturbations in real
k
space become
This can be written in matrix form as


k (t)
 
0 ik

k (t)
 (x, t) = k (0)e(ikxit) + k (0)e(ikxit) , (3.21)
= . (3.10)
t vk (t) ik 0 vk (t)
and then since
This matrix equation can be written as ei + ei
cos , (3.22)
2
v k (t) = M v k (t) (3.11)
t we find
In order to solve this matrix equation, it is convenient to (x, t) cos(kx t) . (3.23)
work in the base of eigenvectors in which the matrix is
diagonal. In this base, the action of M over the set of
eigenvectors is equivalent to a scalar multiplication, i.e.,

Mdiag v k,j (t) = j v k,j (t) , for j = 1, 2, (3.12)


14

3.2 Bernoulli Equation and Barotropic Fluids 3.3 Incompressible and Irrotational Flows

The momentum equation for an ideal fluid in a gravita- For incompressible flows the density is constant. Using
tional potential is the momentum equation, this implies that

v P v = 0 (3.35)
+ (v)v = . (3.24)
t and thus the velocity field is divergence-free.
Let us know assume that the flow is curl-free, or irro-
We can revrite the term (v)v as tational, i.e.,
v = = 0 . (3.36)
1
(v)v v 2 v(v) (3.25) This implies that v can be written as the gradient of a
2
scalar function , i.e.,
and use this relation to define the vorticity, , as
v = . (3.37)
v . (3.26) This should be familiar from electrostatics where a curl-
free electric field can be written as the gradient of a po-
In terms of the , the momentum equation is tential.
If we now have a flow which is incompressible and irro-
v 1 P
v = v 2 , (3.27) tatinal, i.e., both divergence-free and curl-free, then
t 2
v = = 2 = 0 . (3.38)
Let us consider an ideal fluid in steady state by setting
and the potential satisfies Laplaces equation.
v
=0 (3.28)
t 3.4 Viscous Dissipation and Reynolds Number
in the momentum equation we find
So far we have only discussed ideal fluids where there is
1 P no dissipation. Let us now add a dissipation term, 2 v,
v 2 v = , (3.29) on the RHS of the momentum equation to find what is
2
refereed to as the Navier-Stokes equation
If we take the scalar product of this equation with the v P
unit vectos i = v/|v|, we can obtain and equation for the + (v)v = + 2 v . (3.39)
t
total change in the energy as
One of the important dimensionless parameters of flu-
1 2 ids dynamics is the Reynolds number, Re. We can de-
dv + dh + d 0 , (3.30)
2 termine the Reynolds number by considering the dimen-
sionless version of the momentum equation. We make the
or
v 2 equation dimensionless by introducing the dimensionless
+ h + constant . (3.31) variables
2
v P d L d
If the entropy is constant along streamlines ds = 0 then v0 = P0 = = 0 = L 0 =
v0 0 v02 dt0 v0 dt 0
(3.40)
dP
dh = (3.32) where L, v0 , L/v0 and 0 are respectively the character-

istic length, speed, time and density. We find
and v 0 0 0 0 0 P 0 02 0
P 0
+ (v )v = 0
+ v . (3.41)
h=+ (3.33) t Lv

Or by introducing the Reynolds number
The Bernoulli equation is
Lv
Re = , (3.42)
v2
P
+ + + constant (3.34) The momentum equation is
2
v 0 0 0 0 0 P 0 1 02 0
Note that if s(, P ) constant globally then the flow is 0
+ (v )v = 0
+ v . (3.43)
referred to as isentropic. For isentropic flows the pressure t Re
is only a function of the density and thus P = P (). and the flow is characterized only by the Reynolds num-
Fluids for which P = P () are called barotropic. ber.
15

Figure 3.2: Kelvin-Helmholtz Instability dissipation.

form
Figure 3.1: Kelvin-Helmholtz Instability. P = f (z)ei(kxt) , (3.46)
Laplaces equation leads to
3.5 Hydrodynamic Instabilities
d2 f
= k2 f , (3.47)
dz 2
Fluid instabilities can arise in a wide range of physical
conditions and they play an important role in nature. which has solutions f exp[kz].
Here we will outline the derivation of some important With a little bit of algebra using the previous equa-
instabilities in astrophysics without going into all the de- tions, it can be seen that demanding that the pressure is
tails. You can find some of the intermediate steps in the continuous at the interface, i.e., P1 P2 , leads to
book by Padmanhaban (Section 8.13).
The general strategy is always the same. We consider a 1 (kv )2 2 2 , (3.48)
system which is in equilibrium and we perturb all the fluid
This is no other than the dispersion relation associated
variables by a small amount. When the perturbations are
with the Kelvin-Helmholtz Instability, which has solu-
small, we can linearize the equations of motion for the
tions
perturbations by neglecting terms that are quadratic in 
1 i 1 2

the small perturbations. We can attempt to find solutions = kv ., (3.49)
1 + 2
of the form exp[i(kx t)]. Substituting this expression
in the equations for the perturbations will lead, in general If the two fluids have the same density, i.e., 1 2 , then
after some algebra, to a polynomial in as a function of the imaginary part of reads
the wavenumber k. If any of the roots of this polynomial kv
has an imaginary part, then we have an instability. =() = (k) = (3.50)
2
In Figure 3.2, you can see how the growth rate de-
3.5.1 Kelvin-Helmholtz Instability pends linearly on the wavenumber k. Obviously, this
growth can not be infinite for large k, or small physi-
The Kelvin-Helmholtz Instability can arise when there is cal scales, and in reality viscous dissipation prevents the
a velocity gradient along the direction that separates two small scales fro going unstable.
fluids.
For the sake of simplicity we will assume that the fluid
is incompressible and we will work in the frame of refer- 3.5.2 Rayleigh-Taylor instability
ence in which the fluid velocity is zero on one side of the
interface and v on the other. We are going to look for We have two fluids with an interface, which is slightly
solutions that are of the form exp[i(kx t)]. If the dis- distorted. The setup is unstable in case of a heavier fluid
persion relation between and the wavenumber k leads on top of a lighter fluid due to the gravitational field.
to an with an imaginary part, then the small amplitude For the sake of simplicity, let us assume that we have
perturbations will go unstable. an incompressible and irrotational flow, i.e.,
The linearized equations for an incompressible ideal
v , (3.51)
fluid are v = 0 and
v P and thus
+ (v )v = (3.44) 2 0 . (3.52)
t
Using the Ansatz for the potential function
Taking the divergence of this equation we find that the
pressure perturbations must satisfy Laplaces equation, = f (z) cos(kx t), (3.53)
i.e.,
2 P 0 . (3.45) where again f (z) is
Thus assuming that the pressure perturbations are of the f (z) ekz (3.54)
16

Figure 3.3: Rayleigh-Taylor instability.

Requesting that the velocity perturbations vanish at


the top and at the bottom

v(h0 ) = v(h) 0 (3.55)

it can be seen that the solutions for 1 and 2 are:

1 = A cosh[k(z + h)] cos(kx t), (3.56)

2 = B cosh[k(z h0 )] cos(kx t). (3.57)


After some algebra involving the equations of motion for
, it can be seen that must satisfy

kg( 0 )
2 = (3.58)
coth(kh) + 0 coth(kh0 )

This dispersion relation for (k) governs the stability


properties for the small amplitude perturbations. In par-
ticular when < 0 , then

2 < 0 (3.59)

and the system is Rayleigh-Taylor (RT) unstable. In gen-


eral RT instabilities occur in a stratified medium if
d
< 0. (3.60)
dz
Note that for > 0 there are various types of waves,
which are very interesting.
17

4 SHOCKS time we will used the conserved quantity, the total energy,
which is just the sum of internal and kinetic energy.
4.1 Shock waves in astrophysics
1 2
E= v + e (4.3)
It seems to be a good principle that the prediction of a 2
singularity by a physical theory indicates that the theory  
has broken down, i.e. it no longer provides a correct E 1 3
+ v + (e + P )v =0 (4.4)
description of observations - S. Hawking t 2
Notice that the moment equations have successively
Astrophysical examples of shock waves include super- higher orders of v. Can you think of a reason for this?
nova remnants like the Crab Nebula, SN1006, SN1604 The internal energy is given by
and SN1987A. Jets from active galactic nuclei also form
shocks which are observed as hot spots. Herbig-Haro e =  (4.5)
objects are shock waves which appear periodically along
the beams of bipolar jets from young stars. On Earth, we For an adiabatic equation of state
have shock waves produced in thunderstorms, which are
produced when air is heated up to 50,000 K by light- p
e= . (4.6)
ning. The air expands outward rapidly and creates a 1
shock wave. Schlieren plots show shock waves from whips
cracking travelling at twice the sound speed. The angular The adiabatic index/ratio of specific heats is defined as
velocity imparted to the handle of the whip can produced follows:  
Q
a tangential velocity which is supersonic. Other examples Cp m
are blasts from nuclear explosions in air, e.g. the atmo- =  P (4.7)
Cv Q
spheric tests from 1945-1980. Shock waves affect engi- m
V
neering applications, supersonic jets often have pointed
Typical values are: = 5/3 ideal, monatomic gas = 7/5
noses and delta-shaped wings which subtend the narrow-
diatomic gas (air)
est possible conical bow shocks. Blunt objects, e.g. space
Consider a velocity oscillation, propagating in a gas.
shuttle noses or re-entry capsules, create wide, curved,
The crests have higher velocity than the troughs, and
bow shocks, which can be desirable to increase drag and
eventually overtake them. The system becomes multival-
slow down on re-entry from orbit.
ued and is therefore no longer monotonic and no longer
To read a little more, see Lautrups Physics of Contin-
differentiable. What has happened? As the gradient be-
uous Matter , chap 26 and Landau & Lifshitz, Fluid Me-
comes steeper, the non-ideal terms (2 v) in the con-
chanics For more details, see Zeldovich & Raizer (1967).
servation of momentum dominate over the ideal terms.
Even for very large Reynolds numbers/ very small values
4.2 Conservative form of fluid equations of , the viscous term eventually dominates and smears
out the discontinuity over a distance of one mean free
Recall the equations of hydrodynamics, which can be de- path (for collisional shocks) or one Debye length (for col-
rived by taking moments of the Boltzmann equation. lisionless plasmas). We treat this thin non-ideal layer as a
We write them in conservative form: discontinuous jump and apply weak solutions which are
piecewise continuous up to the thin discontinuous jumps

+ (v) = 0 , (4.1) or shocks.
t
the vector momentum equation
v
+ (v 2 + P ) = 0, (4.2)
t
Note the pressure2 term can also be written
q as: P = c2s
p
where cs is the speed of sound cs = . You should
confirm the dimensions make sense! The pressure is a Figure 4.1: Sinusoidal velocity oscillation, propagating
really a tensor quantity, P I3 , where I3 is the identity in a gas. The crests have higher velocity than the
matrix. The quantity v 2 is the tensor product of velocity troughs, and outrun (b) and eventually overtake them
with itself. (c). The system does not become multivalued (c, dashed
Finally we write the energy equation, which before you line). Instead non-ideal effects smooth out the solution
have seen in terms of entropy and internal energy. This over a thin layer, which we call a shock.
2 For this class we will use P and p interchangeably for the pres-

sure and v and u interchangeably for the velocity.


18


v1 v2
1 2
p1 p2

4.3 Shock frame vs. lab frame

For many physical applications it is natural to think of


a shock propagating through a medium at rest. We can
convert a shock from a moving/observer/lab frame to the
shocks own rest frame by subtracting the shock speed VS .
This simplifies the treatment somewhat. If we consider
a jet with a density, j and velocity vj , driving a shock
with vs into an ambient medium at rest with density a ,
we can estimate the advance speed of the shock. We do
this by neglecting the pressure on either side of the shock
front. We transform into the shock frame by subtracting
vs from both sides of the Figure 4.2: Space-time diagram of receding shock wave
2 forming. Density increases across the shock wave, while
j (vj vs ) = a (vs )2 (4.8)
velocity decreases by an equal amount, in accordance
p with the first Rankine-Hugoniot jump condition. From
vs j /a
= p (4.9) Courant & Friedrichs Supersonic flow and shock waves
vj 1 + j /a
We notice that if the jet is 10 times more dense than
the ambient medium, vs /vj .76. What happens to the 1
hxi (x1 + x2 ) (4.15)
speed of the shock if the jet is ten times less dense? Does 2
this make sense? [xy] = [x]hyi + hxi[y] (4.16)
Lets rewrite the jump conditions using the more compact
4.4 Rankine-Hugoniot jump conditions notation.
[v] = 0 (4.17)
Consider the shock in its own rest frame. [v 2 + p] = 0 (4.18)
Density, velocity and pressure change rapidly across an  3 
infinitely thin shock front. v
+ v(e + p) = 0 (4.19)
Can use the the conservation laws, in their steady state 2
form, d/dt = 0, to determine the relationships of each Let
quantity across the front. A = 1 v1 = 2 v2 (4.20)
1 v1 = 2 v2 (4.10) [Av + p] = A[v] + [p] (4.21)
 
1 2
1 v12 + p1 = 2 v22 + p2 (4.11) Av + v(e + p) = A[v]hvi + [v]he + pi + hvi[e + p]
2
1 1 (4.22)
1 v13 + u1 (e1 + p1 ) = 2 v23 + v2 (e2 + p2 ) (4.12)
2 2 [v]he + pi + hvi[e] = 0 (4.23)
1 v2 [v] = []hvi + hi[v] = 0 (4.24)
= (4.13)
2 v1
[] [v] [e] 1 [p]
The first Rankine-Hugoniot jump condition shows that = = = (4.25)
hi hvi he + pi hpi
density increases across the shock front, while velocity
decreases by an equal amount. The jump in each quantity, divided by the mean value, is
equal across a shock wave.

4.5 Properties of shocks hpi hvi Ahvi Ahvi


= = = (4.26)
[p] [v] A[v] [p]
We introduce some helpful notation here: Ahvi = hAvi = hpi (4.27)
2
[x] x1 x2 (4.14) Av = v = the ram pressure of the flow (4.28)
19

If the ram pressure decreases across the shock, thermal 4.6 Phases of supernova remnant evolution
pressure must increase! We might expect this from Li-
ouvilles theorem. Since the total volume in phase space Supernovae (SNe) occur when massive stars implode and
is conserved, a convergence in real/configuration space eject their outer layers. 99% of the energy released is car-
implies a divergence in velocity space. ried away by neutrinos. The remainder drives a super-
On one side of the shock, kinetic energy is converted to nova remnant into the ISM. The SN ejecta expand into
pressure by dissipative, non-ideal processes. the interstellar medium, and shock it.
Can divide the process into 3 phases.
v 2 < p (4.29)
1. free expansion/constant velocity phase.
On the other side kinetic energy dominates over thermal 2. adiabatic, energy-conserving/ Sedov phase.
pressure.
3. radiative/momentum conserving/snowplow phase.
v 2 > p (4.30)

Can define adiabatic speed of sound as 4.6.1 Free expansion/constant velocity phase
r
p
cs (4.31) Simple model: Mass Mej of ejecta expands at the max-

imum velocity, v0 Assume distributed in a spherical vol-
ume, and uniform density ej .
Flow on one side is supersonic
Radius R0 = v0 t
v 2 > c2s (4.32) 3Mej
ej = (4.39)
4(v0 t)3
Flow on other side is subsonic
V (r) = r/t, 0 < r < V0 t (4.40)
2
v < c2s (4.33) We can now make an expression for the supernova energy.
Z v0 t  r 2
Kinetic energy is irreversibly converted to thermal energy. 1 3
ESN = 4r2 ej dr = Mej V02 (4.41)
Shock waves increase entropy. 0 2 t 10
Mach number definition
Ejecta shocks ISM, creates a forward shock. ISM gas
v shocks ejecta, creates a reverse shock. Ejecta expand to
M= (4.34)
cs

Lets assume the shock is strong , with a M  1. We


can then neglect pressure on supersonic side.
Simplify conservation of momentum and energy:

A(u1 u2 ) = p2 (4.35)

1
A(u1 u2 )(u1 + u2 ) = u2 (e2 + p2 ) (4.36)
2
Substituting:
e2
u1 + u2 = 2u2 (1 + ) (4.37)
p2
or
u1 2 e2 +1
= =1+2 = (4.38)
u2 1 p2 1
We have the interesting result that the jump in density
and velocity and pressure depends only on the properties
of the medium, the adiabatic index. For ideal monatomic Figure 4.3: Expanding blast wave showing forward and
gas, compression ratio 21 is 4. For diatomic gas, compres- reverse shock and contact discontinuity.
sion ratio is 6. For gases with strong radiative cooling,
compression can be much higher as the effective adiabatic
index approaches unity. radius RCD (See Figure 4.3)
20

Forward shock as @ radius RF . All ISM gas is initially along under its own inertia. Momentum is conserved, en-
inside RF , has been compressed to a thin shell, between ergy is lost. The swept-up mass now gradually slows the
RCD and RF . front to subsonic speeds. At this point it begins to merge
with the ISM.
4 3 4 3 3
R 0 ' (RF RCD )40 (4.42) In order to calculate the expansion radius RS (t) and
3 F 3 speed VS (t) in the snowplow phase, we can set the mo-
  43 mentum equal to that at the end of the Sedov phase.
4
RF RCD ' 1.1006RCD (4.43)
3 4 3 4 3
RS 0 VS = constant = R 0 V 0 (4.49)
Assume each element does not expand or contract after 3 3 0
being compressed in shock. RS
Phase ends @ tSW , sweep-up time, where swept-up VS = (4.50)
t
mass = mass of ejecta.
Integrate to get the expression for the radius in the
4 3 3 snowplow regime.
V t 0 = Mej (4.44)
3 0 SW Z Z
3 3
RS dRs = R0 V0 dt (4.51)
4.6.2 Adiabatic, energy-conserving/ Sedov phase
1
RS t 4 (4.52)
The Sedov phase begins after about 102 years, when the
sweep-up time is over, and enough mass has been accu- You can differentiate to get the velocity, or just use the
mulated to decelerate the remnant from a constant veloc- momentum conservation expression above.
ity. During 1941-1945, in unpublished work, J. von Neu- RS 3
mann, L. Sedov and G. I. Taylor independently studied VS = t 4 (4.53)
t
the instantaneous input of fixed amounts of energy E0 E FRIDER
IA

into systems with uniform density, 0 . They found the M


Putting the pieces together, we can plot the life cycle of
IC
M ACADE

O
ALEXA

only free parameters were the initial energy (at the end a supernova remnant, from constant velocity expansion,
LV

ND

441
RI IL
AE SIG
N

of the free expansion phase), the initial ambient density to constant energy, toSummary
constant momentum.
and the radius of the blast wave. See homework exercise.

R = (constant) ESN 0 t (4.45) Blast wave
v=const.
Sedov
E=const.
Snow plough Merging
p=const. r=const.

T108 K T106 K T104 K


Another way to derive the Sedov relations approximate 10 2
v104 km s-1 v200 km s-1 v10 km s-1

the pressure, p by the energy density in the volume t1/4 t0 10 4

ESN / 4R3 /3 . Omitting the constants we get and or-


t2/5
der of magnitude estimate for the sound speed: 1
10

velocity [km s-1]


Radius [pc]

s
10 3
r
p ESN
' (4.46)
R3
10 0

Assuming the shock speed is comparable to the internal


10 2
sound speed, t
-1
s 10

R ESN 3/2
Vs = ' R (4.47)
t 0
10 1
10-2
100 102 104 106
Integrate to get the radius. Time [years]

 1/5 after Padmanabhan (2002, Fig. 4.6)


ESN
R' t2/5 (4.48) Figure 4.4: The radius and velocity of the supernova shell
as a function of time during the different phases
Supernova Remnants 24

4.6.3 Radiative/momentum conserving/snowplow phase

The Sedov phase ends after a few 103 years, when


temperature has dropped to T 106 K. C,N,O ions now
recombine and cool the remnant efficiently. The energy
driving the explosion is now lost. The remnant now coasts
21

5 GRAVITATIONAL DYNAMICS 5.2 Gravitational binding energy

5.1 The gravitational potential The gravitational binding energy of a system is the
amount of energy that needs to be removed from it in
The gravitational force on a particle of mass mi due to a order to build the system doing work against the grav-
particle of mj is: itational field. Using dimensional analysis, it is easy
to argue that the gravitational binding energy of a sys-
Gmi mj (r i r j )
F ij = . (5.1) tem of mass M and size R, should be of the order of
3
|r i r j | U (R) = GM 2 /R, where the minus sign indicates that
the system is gravitationally bound.
We can write the acceleration due to gravitational force
Let us analyze in detail the case of a uniform sphere of
exerted by a continuous distribution of matter as:
radius R, mass M and density = 3M/(4R3 ). Let us
G(t, x0 )(x x0 ) 3 0 consider that we are at the stage where we have already
Z
g(t, x) = 3 d x . (5.2) built a spherical object of mass m(r). The incremental
|x x0 |
change in the potential energy of the system associated
The gravitational acceleration can be written as the gra- with the addition of a new shell of mass m(r) that it is
dient of the gravitational potential , i.e., brought from infinity to the distance r is given by

g = , (5.3) m(r)m(r)
dU = G , (5.10)
r
with
G(t, x0 ) 3 0
Z
2 3
(t, x) d x . (5.4) where m(r) = 4r dr and m(r) = 4r /3 Therefore,
|x x0 | the total energy associated with the build up of the entire
This implies that the gravitational field is conservative, sphere can be obtained as
i.e., the work done by the gravitational field between two Z
points in space is independent of the path used to go U (R) = dU (r) (5.11)
from one point to the other one. You can see that the 0
4r2 34 r3
Z R
gravitational field tends to 0, as |x| , assuming that
3 = G dr (5.12)
drops faster than |x| . 0 r
Using that the Laplacian of 1/|x| is related to the Dirac 16
delta function = 2 G2 R5 , (5.13)
15
 
1
2 = 4(x) , (5.5) which can be written in terms of the total mass of the
|x| sphere as
where 3 GM 2
U (R) = . (5.14)
Z x2  5 R
f (x0 ) if x0 [x1 , x2 ]
(xx0 )f (x)dx = (5.6)
x1 0 if x0
/ [x1 , x2 ] In general, the density can be a complicated function of
position and we will have:
we can show that the gravitational potential satisfies the
Poisson equation
GM 2
U (R) = , (5.15)
R
2 = 4G . (5.7)
with a constant of order unity.
The previous set of equations apply to situations in which
relativistic effects can be neglected. Newtonian gravity is
a good approximation provided that ||  c2 , where c is 5.3 Virial theorem for N-body system
the speed of light. For a test mass orbiting a mass M in
circular orbit we have The equation of motion for a particle of mass mi subject
GM to gravitational interactions with a set of particles of mass
= v2 . (5.8) m is
r j

Thus we have for the potential mi mj (r i r j )


F i = mi r i = j6=i G 3 (5.16)
|| GM  v 2 |r i r j |
2
= =  1. (5.9)
c rc2 c
We can calculate the "virial" associated with this force
As expected, we can use Newtonian gravity provided the
speed involved is small compared to the speed of light c. virial i F i r i = i mi r i r i (5.17)
22

by dotting the equation for the force acting on mi with 5.4 Virial theorem for self-gravitating gases
r i and summing over all indices i. This leads to
r i (r i r j ) When we consider a fluid, there will be a force contribu-
i mi r i r i = i j6=i Gmi mj 3 (5.18) tion arising from the pressure gradient
|r i r j |
using the identity ri F i = r i P (5.27)
= (r i P ) + P r i (5.28)
1 1
ri = (r i r j ) + (r i + r j ) (5.19) = (r i P ) + 3P . (5.29)
2 2
we find that the virial in no other thing than the gravi- Recall that for ideal fluids the pressure and the internal
tational potential energy: energy are related via an equation of state
1 P = ( 1) . (5.30)
mi r i r i = j6=i Gmi mj U . (5.20)
|r i r j |
We now consider the virial that we introduced before and
On the other hand, the left-hand side (LHS) of this write it in the continuous limit as
equation is related to the second time derivative of the
Z Z
moment of inertia of the system of particles, i.e., I i ri Fi (ri P )dV + 3 P dV (5.31)
V
2
mi r i since Z
= 3P s V + 3 P dV (5.32)
1 d2
I = i mi r i r i + i mi r i r i . (5.21) Z
2 dt2 = 3Ps V + ( 1) dV (5.33)
Realizing that the second term on the RHS of this equa-
tion is twice the kinetic energy of the system of particles, = 3Ps V + 3( 1)Uint (5.34)

1 The virial theorem for a fluid is thus


i mi r 2i K , (5.22)
2 1
I = 2K + Ugr + 3( 1)Uint 3Ps V
(5.35)
we can write the virial theorem for a system of gravitating 2
particles as
Here, K is the kinetic energy due to the bulk motion and
1 d2 U int is the kinetic energy due to random motions.
I U + 2K . (5.23)
2 dt2
Lets look at two important cases where this is useful.
5.4.1 Gaseous self-gravitating system

5.3.1 Approximately steady state I ' 0 Lets assume the system is not moving as a whole, K = 0,
and that we can neglect the pressure acting on its surface,
In steady state, the virial theorem implies i.e., Ps = 0. The virial theorem becomes
1 1
2K + U = 0 K = U (5.24) I = Ugr + 3( 1)Uint (5.36)
2 2
Thus, we can write the total energy of the system as = Ugr + 3( 1)[E Ugr ] (5.37)
1 = 3( 1)E (3 4)Ugr . (5.38)
E = K + U E = K = U (5.25)
2 Therefore, in steady state
The total energy of a system in virial equilibrium is half of
3 4
its gravitational potential energy, which must be negative E= Ugr . (5.39)
for bound systems. 3( 1)

This implies that gravitationally bound systems, i.e., E <


5.3.2 Time-averaged virial theorem for a bound system 0, require that > 34 . Note that a gas made out of
photons has = 4/3!
Taking the time average in the virial system we get
2K + U 0. (5.26) 5.4.2 Ideal monoatomic gas

This form of the virial theorem has been applied to glob- Let us consider a ideal monoatomic gas in steady state
ular clusters and galaxy clusters. In the former case it and for which there is no bulk motion, i.e., K = 0, U =
int
can be used to estimate the mass of the cluster whereas 3k T /2, = 5/3. The virial theorem reduces then to
B
in the latter it has been used to infer the presence of dark
matter in galaxy clusters. 3kB T + Ugr = 3Ps V . (5.40)
23

We can solve for the pressure acting over the surface of If the system is in virial equilibrium we have v 2 GM/R
the volume under consideration as and thus
R
GM 2 tcr N , (5.50)
 
1 3M 2
Ps = v 4 . (5.41) v
4 R 3 R
which is N times the crossing time.
Recalling that the binding gravitational energy is A more refined treatment leads to:

GM 2 v3
Ugr (5.42) tcr = (5.51)
R 2G2 m2 n ln N

and that the characteristic velocity associated with ther- and subsequently to:
mal motions is
kB T 1 N R
v2 = (5.43) tcr ' . (5.52)
m 8 ln N v
we can derive the critical value for the mass that would
lead to Ps = 0. This critical mass for gravitational col- For a galaxy, we get
lapse is called the Jeans mass:
tgal ' 1018 yr  tuniverse , (5.53)
2
 
3v R 3 kB T R
Mcrit = = . (5.44) whereas for a globular cluster we find
G Gm
tgc ' 0.4 Gyr . tuniverse . (5.54)
We can estimate a Jeans length associated with this mass
using that M R3 to obtain Beware of other processes such as resonant relaxation
(where the presence of a massive body can enforce Kep-
s   lerian orbits that interact resonantly) and violent relax-
3 kB T
crit = . (5.45) ation (where the gravitational potential changes signifi-
Gm cantly on the timescales of interest) that operate faster.

5.5 Gravitational collisions and relaxation 5.6 Dynamical friction

We can estimate the cross section for gravitational inter- Let us now consider a particle of mass M moving on a
action in an ensemble of particles interacting gravitation- sea of lighter particles of mass m. As the particle moves
ally by using dimensional analysis. We could in principle through the medium it will create a wake of lighter parti-
ask, what is the combination of G, M , and v, that would cles that will exert an enhanced gravitational pull. This
give as a length-scale squared, i.e., "dynamical friction" will slow down the mass M on a time
scale that can be estimated by replacing Gm2 GM m
2
L =G M v . (5.46) in the equations that we derived for the collisional relax-
ation.
We could take a short cut and realize that the ratio of v3
tdf ' (5.55)
the potential energy to the kinetic energy is dimensionless 2G2 M ln N
and thus Note that this time is shorter than the timescale associ-
GM
L
# = 2 L GM v 2
(5.47) ated with the relaxation timescale by a factor of m/M ,
v i.e.,
Therefore the cross section for interaction scales like m
tdf = tcr (5.56)
M
2 2
G M
grav = (5.48) This dynamical friction is thought to provide a mech-
v4 anism for the condensation of heavier bodies at the cen-
ters of clusters of lighter objects. This "slow down" has
Lets us now consider N identical particles with mass m.
associated a loss of angular momentum that makes the
The total mass of the system is M = N m. We will denote
embedded heavy object spiral toward the center of the
the characteristic number density and velocity as n and
medium of lighter bodies. This could be applied to heavy
v, respectively. We can estimate the timescale associated
stars and stellar mass black holes in globular clusters, or
with with collisional relaxation as
even to massive black holes and globular clusters in the
1 R3 v 3 N R3 v 3 R R2 v 4 galaxy.
tcr = ' 2 2
' 2 2 =N .
nvgrav N G m G M v G2 M 2
(5.49)
24

5.7 Self-gravitating barotropic fluids

We now consider the static equilibrium of fluid configu-


rations with spherical symmetry in which self-gavity is
balanced by the pressure gradient.
Setting v 0 in the equations for fluid dynamics we
have
P = , (5.57)

where the gravitational potential satisfies Poissons equa-


tion
2 = 4G. (5.58)

We can combine these two equations to obtain Figure 5.1: Sketch of the solutions to the Lane-Emden
equation as a function of the politropic index n.
r2 dP
 
1 d
= 4G (5.59)
r2 dr dr 5.8.1 Boundary conditions for the Lane-Emden equation

In case of a barotropic fluid, P P () and we can inte- In order to integrate the Lane-Emden equation we need
grate this equation to get P (r) and (r), and thus solve two boundary conditions on (). We must have
for the radial structure of the self-gravitating object.
( = 0) 0 (5.67)

5.8 Lane-Emden equation polytropes so the density matches the correct value at the origin, and

d
A particularly simple form of the barotropic fluid is given 0 (5.68)
d =0
by a polytropic equation of state
if we require that the pressure does not exhibit a cusp at
P K , (5.60) the origin, i.e.,
dP
0 (5.69)
where both K and are constants. In this case, we dr r=0
can rewrite equation (5.59) in dimensionless form, the In general, the Lane-Emden equation only has a few
so called Lane-Emden equation, analytic solutions:

1 d
  = , n = 0 (sphere of constant density)
2 d
= n , (5.61)
2 d d 2
0 = 1 (5.70)
6
where we have defined
10 = 6 (5.71)
n
= c , (5.62) = 2, n = 1 (planetary interiors and brown dwarfs)

with sin()
1 = (5.72)
c (0) , (5.63)
11 = (5.73)
and
r = , (5.64) = 56 , n = 5 (galaxies)
 12
with 3

  12 5 = 1+ (5.74)
n+1 1n 3
Kc n , (5.65)
4G
Here, we have denoted the first zero of the polytrope of
and the polytropic index n is related to via index n by 1n the. Only the solutions with n < 5 have
a finite 1n such that n (1 ) = 0. In general, if K and
1
= 1. (5.66) n are given, we can integrate the Lane-Emden equations
n from = 0 up to 1 , and state the results in terms of the
central density c or the total mass M .
25

5.9 Dynamical stability of self-gravitating disk or, in terms of the angular frequency, 0 = v0 (r)/r

The derivation in this section follows closely the one in 1 dP0 GM


2 r = + 2 . (5.88)
pg. 134 in Armitage (2013). 0 dr r
We consider the dynamical stabiltiy of a razor-thin,
Because the radial pressure gradient in the disk is in gen-
self-gravitating disk against axi-symmetric perturbations.
eral negative, the angular frequency is in general lower
If we are not concerned with the vertical structure of the
than the Keplerian value
disk, we can write the equations in terms of the surface
density The equations of interest are thus GM
2K . (5.89)
r3
+ (v) = 0 , (5.75)
t
v P 5.9.2 Linear Mode Analysis
+ (v )v = , (5.76)
t
2 = 4G , (5.77) Let us consider small amplitude, axisymmetric perturba-
tions such that
where the density is given by
= 0 + 1 (t, r) , (5.90)
(x) = M (x) + (z) , (5.78)
P = P0 + P1 (t, r) , (5.91)
and the sound speed is
= 0 + 1 (t, r) , (5.92)
dP
c2s = . (5.79)
d and
v = v 0 + vr1 (t, r)r + v1 (t, r) (5.93)
5.9.1 Steady state background In order to perform the linear mode analysis, we assume

We first need to find steady state solutions over which 1 (t, r) P1 (t, r) 1 (t, r) e(krt) (5.94)
we will perform the linear mode analysis. Denoting this
background solution with a subscript "0", we obtain for with kr  1 (locally). Recall that in cylindrical coordi-
the continuity, momentum, and Poisson equations nates the velocity is v = vr (r)r + v (r) and thus

(0 v 0 ) = 0 , (5.80)
! 
v2

vr v vr v
(v )v vr r vr + .
P r r r r
(v 0 )v 0 = 0 , (5.81)
0 (5.95)
2 Substituting for , P , and v, we obtain
0 = 4G [M (x, y, z) + 0 (z)] , (5.82)
Note that if the speed at which the gas rotates in the i1 + vr 0 ik = 0 , (5.96)
disk depends only on radius, i.e.,
1 dP1 d1
v 0 = v0 (r) , (5.83) ivr 2v = , (5.97)
0 dr dr
then the continuity equation is trivially satisfied because 2
iv +
vr = 0 . (5.98)
2
0 ( v 0 ) + v 0 0 = 0 . (5.84)
Here, we have defined the epicyclic frequency as
The momentum equations reads
dln
v2 P0 1 dP0 0 2 = 42 + 22 (5.99)
0 = 0 = , (5.85) dln r
r 0 0 dr dr
One way to derive the dispersion relation associated
where the gravitational potential due to the central star with this linear, homogeneous system is to recast all the
and disk is quantities that depend on the perturbations in terms of,
GM
0 = + 2G0 |z| . (5.86) for instance, 1 . This is easy to do for the radial and
r
azimuthal velocity fluctuations
Using this expression in the previous equation, we obtain

v02 1 dP0 GM 1
= 2 , (5.87) vr = , (5.100)
r 0 dr r k 0
26

and thus
2 i 1
v = . (5.101)
2 k 0
We now need to relate dP1 /dr and d1 /dr with 1 .
From the equation for the sound speed,
dP
c2 = , (5.102)
d
we obtain
1 dP1 ik 2
= c 1 . (5.103)
0 dr 0 s
From the Poisson equation for the perturbations Figure 5.2: Sketch to find the roots of the dispersion re-
lation in equation (5.112).
2 1 = 4G1 (z) , (5.104)

we obtain The wavenumber at which the parabola in Fig. 5.2


d2 1 shows a minimum, which corresponds to the wavenumber
= 2 1 + 4G1 (z) . (5.105) exhibiting fastest growth, can be obtained from
dz 2
The only solution that remains finite for large |z| is d 2
= 0, (5.113)
dk
1 = ce|kz| , (5.106)
corresponding to
where the constant c needs to be determined. In order to G0
do this, we integrate the Poisson equation between z =  kmin = , (5.114)
c2
and z = +
Z + Z  Note that the disk will be marginally stable if 2 (kmin )
2
1 dz = 4G1 (z)dz , (5.107) 0 or when the Toomre parameter Q is less than unity, i.e.,
 
cs
+
d1 Q <1. (5.115)
= 2|k|1 = 4G1 , (5.108) G0
dz 
Within factors of order unity this criterion also applies
This implies that in the limit of  0, 1 () = (0) = c
to a self-gravitating stellar disk by replacing the sound
2G1 speed cs with the velocity dispersion of the stars, usually
1 (z = 0) = , (5.109) called .
|k|

which gives the relationship between the potential and


surface density fluctuations in the z = 0 plane. Taking
the derivative with respect to radius, we find

d1 2ikG1
= . (5.110)
dr |k|

Substituting equations (5.100), (5.101), (5.103), and


(5.109) into (5.97) we obtain

1 2 1 1 2ik
i +i = ikc2s + G1 , (5.111)
k 0 k 0 0 |k|

which, after some simplifications, leads to

2 = 2 + c2s k 2 2G0 |k| , (5.112)

where the epicyclic frequency defined in equation (5.99),


is = for a Keplerian disk.
27

6 BASICS OF MAGNETOHYDRODYNAMICS and then define

6.1 Magnetohydrodynamics via Hydromagnetics = m+ n+ + m n (6.14)


number
Go back to to Maxwells Equations, which describe how = mass per particle (6.15)
volume
magnetic and electric fields behave written here in CGS:
and arrive back at
B = 0 (6.1)
E = 4q (6.2) + (v) = v (6.16)
t
4 1 E
B = j+ (6.3) Now, what about forces on the charged fluids? Write
c c t the momentum equations, for the forces on each of the +
1 B
E = (6.4) and in a volume element, we include the Lorentz force
c t on the charged fluids
where q is electric charge density, E is the electric field  + 
and B is the magnetic field. Watch out - this form + + v +
m n + (v)v =
depends on the choice of CGS Gaussian v.s. SI v.s. t
e+ n+ E + c1 v + B f + P + m+ n+ (6.17)

Heaviside-Lorentz.
Where q = 0 we can combine these to get a wave- 
v


equation: m n + (v)v =
t
1 2B e n E + c1 v B f P +m n

2 B = (6.5)
c2 t2
| {z } | {z }
Lorentz Force: spiralling Fraction of pressure per
motion about B fluid f + + f 1. Will
You have seen a similar thing for sound waves.
go away.
The fields interact with charges through the Lorentz
force, given by (6.18)

1
 add them together. Define
F =q E+ vB . (6.6)
c q n+ e+ + n e net charge density (6.19)
where if q is a charge density, F is a force density. + + +
j n e v +n e v current density (6.20)
Recall the equation of mass conservation (aka continu-
ity), and momentum equation from fluid dynamics Note

v = m+ n+ v + + m n v (6.21)
+ (v) = v (6.7)
 t  so the two charged fluid momentum equation is
v
+ (v)v = P (6.8)
t 
v

1
+ (v)v = qE + j B P (6.22)
Take a fluid made up of two particle species of oppo- t | c
{z }
site charges. The conservation laws for the two types of Need to figure
particles are out what this is.

n+ Ohms law in the familiar form is


+ (n+ v + ) = 0 (6.9)
t V
n I= (6.23)
+ (n v ) = 0 (6.10) R
t
more generally we can write
(6.11)
 
1
with j = E+ vB (6.24)
c

n = (6.12) is the conductivity in 1/(resistivity)
m
the particle number density. Deeper physics at work in (collisions, ...)
If we define the center-of-mass velocity as Watch out for the unit system here!
+ + +
m n v +m n v v really should be v + and v but charge seperation
v= (6.13)
m+ n+ + m n is damped by the electric field so v v + v
28

That specified j, but what about E and B? and introduce the definiton of the magnetic diffusivity in
The critical next step is to impose charge neutrality CGS Gaussian

q=0 (6.25) c2
(6.37)
4
and then given
then we get
1 B
E = (6.26) B 0
c t (v 0 B 0 ) = 02 B 0 (6.38)
t0 v0 L
cE ` (flow lengthscale)
v (6.27)
B (flow timescale) This form immediately suggests an analogous definition
to the Reynolds number introduced with a viscous fluid.
This is hence the magnetic Reynolds number:
1 E 1 E
 2
c t 1 ` v2
B
2
2 1 (6.28) v0 L
B c ` c c

ReM (6.39)

so
4 6.2 Ideal MHD
B = j (6.29)
c
Take perfect conduction (and perfect Lenzs law).
which is a charge-neutral, non-relativistic approximation. So the Ideal MHD induction equation is
The momentum equation is then
B
v 1 P = (v B) (6.40)
+ (v)v = 0 + BB t
t 4
(6.30) We have here the same situation as Kelvins Circulation
Therom (regarding vorticity). Note the (v)B is hidden
or inside the (v B). Magnetic field lines are frozen
v 1jB P in (pinned, stuck) to the fluid.
+ (v)v = (6.31)
t c
6.3 Energetics in MHD
Now back to
4 In a fluid, we have
B = j (6.32)
c
Internal thermal energy density e. For an ideal gas
Take the curl, substitute in Ohms law with = constant e = P/( 1)

4 Kinetic energy density EK = 21 v 2


( B) = [ E +c1 (v B)]
c | {z }
Magnetic energy density EB = B 2 /(8)
| {z }
as B =0 then =2 B B
= 1c t
(6.33) Fixed gravitational potential EG =
So this becomes the MHD induction equation So evolution equations can be written like
B c2 2 P dV work,
(v B) = B (6.34) EK to e
t 4
e z }| { 1 2
This has no explicit E field! Thats the magic of MHD. + (v)e = ev P (v) + j
|t {z
| {z }
Dimensionally this is
} contiinuity |{z}
Lagrangian deriv. Ohmic heating
P = I2R
L2 B
 
B 1 L (6.41)
B = (6.35)
T L T T L2
Sometimes such a form is used, but conservative form
The boxed term is a diffusion (like viscosity). Now, if we is clearer here
non-dimensionalize the MHD induction equation E
+ F = Q (6.42)
B v d L d t
0 0 0
B = v = 0
= = L (6.36)
B0 v0 dt v0 dt where
29

E is conserved if there are no sources or sinks For magnetic energy, go back to

F is the flux associated with E B


= c (E) (6.53)
t
Q is the source or sink of the quantity E
EB B 2 /(8) B B B
Theres a special case you have seen a few times before, = = = (E)
where the source term Q is zero. A conservation law has t t 4 t 4/c
the special form (6.54)

E 
B
 
EB

+ F = 0 (6.43) = E (6.55)
t 4/c 4/c
| {z } | {z }
For example, if you take E = and F = v then this is j = B Poynting
4/c
conservation of mass, or the continuity equation. Flux
So, we can write for the thermal energy S = E
B/(4/c)
e 1 2
+ ((e + P )v ) = P
| {zv} + j (6.44) With Ohms law E = 1c v B.
1
t | {z } j
flux pressure work |{z}
Ohmic heating  
1 2 1 EB
= j + (v B)j (6.56)
This is for a fluid which does not lose or transfer energy | c {z } 4/c
by radiating it away, there is no sink term. Its adiabatic. triple product
Now, for the others = 1c (j B)v

1 v 2 So
EK 1 v
= 2 = v2 +v (6.45) Poynting Flux (incl. Ohmic)
t t | {zt}
2 t z }|
|{z}   {
continuity momentum EB c 1 1
+ vB+ j B
t 4 c
 
1 2

1jB P
 1 2 jB
= v (v) + v (v)v + = j v (6.57)
2 c }
| {z c
| {z }
(6.46) Ohmic Lorentz
 
1 jB And the gravitational energy
= v 2 (v) v(v)v + v
2 c
EG = (6.58)
vP v (6.47)

Now, note EG
= = (v) = (v) + v
t t
(6.59)
 
1 2 1 1
v v = v 2 (v) + v(vv) (6.48)
2 2 2 so
EG
with the identity + (v) = v (6.60)
t | {z } | {z }
1 Grav. Pot. Flux work against gravity
(AA) = (A)A + A ( A) (6.49)
2 Now if we add these all together
(Note we used the same sort of identity when deriving the (e + EK + EB + EG )
Bernoulli Equation.)
t
1
= v 2 (v) + v(v)v + v(v ( v)) (6.50) 1 EB
2 + (e + P )v + v 2 v + + v = 0 (6.61)

1 | {z } |2 {z } 4/c |{z}
= v 2 (v) + v(v)v + ( v)
(vv)
 (6.51) Thermal | {z } Grav.
Kinetic
2 Poynting

so Thats how energy conservation works out. Obviously,


EK 1 jB
  due to the lack of a source term on the right hand side,
+ ( v 2 v) = v P v v this describes a system with no loss of energy to anything
t 2 c | {z } | {z }
| {z } Pressure work Gravity outside of the fluid, resulting in it being adiabatic. Note
Lorentz work however that at no point did the equation of state appear
(6.52) in this section.
30

processes (stemming from mutual collisions of charged


species) become important. If such effects are negligi-
ble, flux freezing implies that magnetic field lines are at-
tached to the flow. By virtue of Kelvins circulation
theorem, we have
Z
B d
= (vB) = B dS = 0 , (7.2)
t dt S
R R
Figure 7.1: Magnetic flux B S B dS. which implies that the magnetic flux, B S B dS, is
conserved through any surface, S, that is moving with
7 MAGNETOHYDRODYNAMIC WAVES
the fluid (a so-called Lagrangian surface) see Fig. 7.1.
To prove the theorem in Eqn. (7.2), we have to account
We begin by recapitulating the ideal MHD equations,
for two different possible ways of changing the flux, B
where ideal refers to the limit of vanishing diffusion of
in time:
magnetic fields, corresponding to 1/(0 0 ) 0, or
1. due to B B
R
t , that is: S dS t
alternatively speaking, to the limit of infinite conductiv-
ity, 0 . The equations are (in simple form)
d
R
2. due to the motion of the surface: S B dt (dS)
%
+ (v) % = % v continuity eqn.,
t
v
%( + (v) v ) = p + j B momentum cons.,
t
B
= (vB) induction eqn. ,
t
where j B/0 is the electric current density. Al-
ternatively, we can write the two fluid-related equations
above (that is, the continuity and momentum equations)
in conservation form (and introducing the abbreviation
t for the partial derivative /t).
t % + (%v) = 0,
t (%v) + ( %vv + p1 B B/0 ) = 0.
| {z } | {z } Figure 7.2: Magnetic flux lost due to motion of a surface.
1 2
To visualise the latter effect, we imagine a cylindrical vol-
where (v v)ij vi vj indicates the outer (or tensorial) ume in space-time, as illustrated in Fig. 7.2. The mag-
product of the velocity vector with itself, and 1 means the netic flux B lost through the wall of the cylinder dur-
identity matrix. The terms marked 1 , and 2 , denote ing a finite time, t, is equivalent of the normal component
the Reynolds-, and Maxwell stress, respectively, whose of B integrated over the shaded area (t vl), which can
divergences, (. . . ), represent the forces exerted within be expressed as
the fluid itself, and via the Lorentz force, j B. These I
are important quantities, that we will encounter again in B = t B (vl) . (7.3)
the context of accretion disks later. C
Note that, in the conservative form, the equations have Cyclic permutation of the scalar triple product yields
the general form I
t a + Fa = 0 , (7.1) = t l(B v) , (7.4)
C

with a being a conserved fluid variable (such as the den- where we additionally obtain a minus sign from swapping
sity, %, the momentum density %v, the total energy den- the terms in the cross product, that is B v vB,
sity, etot v 2 /2 + . . . ), and Fa representing the hydro- I
dynamic flux associated with that variable (for instance, = t l(vB) . (7.5)
v in the case of a = %). C

According
H to Stokes theorem, we can replace
R a line inte-
7.1 Flux freezing gral CS
lA with a surface integral S
dS A for
any vector field A. Thus the last expression becomes
The so-called flux freezing is an important property of
Z
ideal MHD fluids, and as such is violated once diffusive = t dS (vB) , (7.6)
S
31

which we recognise as the induction term in Faradays the upper right-hand corner, with x pointing to the right,
law. Taking the limit t 0, where B/t t B, we y going into the page, and z pointing up. The field ini-
obtain the corresponding infinitesimal change, which we tially consists of a constant vertical field B = B0 z. Such
a field does not have a curl, and hence j = 0, implying
can insert into the expression for the total time derivative
of the magnetic flux, yielding that no Lorentz force is exerted.
Z By imposing a shear flow v = v0 sin(kz)x, with a si-
d d nusoidal variation of wavenumber k along z, we will cre-
B B dS
dt dt S ate a distortion in the field lines. As a result, after some
Z  
d B finite time, the field lines will become corrugated, as is
= dS (vB) , (7.7)
dt S t shown in the right hand side of Fig. 7.3. To compute the
change in the magnetic field, we have to solve the induc-
where the integrand vanishes according to the ideal MHD tion equation. For this, we first evaluate the vB term,
equation, completing the proof that B = 0 for any co- employing the useful property that the vector product
moving surface, S. can conveniently be written as a determinant (with the
unit coordinate vectors in the first row).
7.2 The plasma parameter
x

y z 0
vB = vx vy vz = v0 B0 sin(kz) (7.11)
An important dimensionless number in plasma physics Bx By Bz 0
is the so-called plasma parameter, or plasma beta, P .
It signifies the relative importance of the p and the which means that vB only has a y component, and hence
Lorentz force j B. points out-of (into) the page on the top (bottom) half
of the distorted pattern (note the location of the z = 0
| j B | | p | . (7.8) location in Fig. 7.3). The induction equation yields

The magnitude of the Lorentz force can be expressed by B


the so-called magnetic pressure, pmag , which is defined as = (vB)
t
x y z
B2
pmag . (7.9) = x y z (7.12)
20
0 v0 B0 sin(kz) 0
It can easily be shown that the longitudinal part of the = z [v0 B0 sin(kz)] x
Lorentz force has a contribution owing to a gradient, = v0 B0 k cos(kz) x , (7.13)
pmag , in magnetic pressure. Assuming that the mag-
netic field changes on similar scales as the gas pressure, which we can simply time-integrate using the initial con-
pgas , one typically defines dition B(0) = B0 z as the integration constant.
pgas 20 p B(t) = B0 z v0 B0 k cos(kz) t x , (7.14)
p = . (7.10)
pmag B2
that is a field distortion that grows linear in time.3 Note
Note that this convention implies a large value of p in that, since the sine differentiates into a cosine, the growth
the case of a weak influence of magnetic fields, whereas it of the transverse field (that is, Bx ) is fastest where the
implies a small value for strong fields. distortion velocity is zero. This however corresponds to
the location where the velocity shear (that is, z vx ) is the
highest.
7.3 Magnetic field-line tension We will now evaluate the Lorentz force of the distorted
field lines. For this, we first need to compute the electric
Before we can proceed with deriving linear MHD waves, current. Using the fact that B0 z is curl free, and Bx =
we have to introduce the important concept of magnetic Bx (z), we obtain
field-line tension. We have learned earlier that the gas
pressure can act as a restoring force when we compress or 1
j B = (z Bx ) y
rarefy a parcel of gas. Together with the inherent inertia 0
of a massive fluid, this lead to the propagation of sound v0 B 0 2
= k sin(kz) t y . (7.15)
waves. In the following, we will see that magnetic field 0
lines have a similar property, and that we can think of
them as a kind of rubber bands threading the fluid. We can see that j also grows linear in time, implying that
We will illustrate this tension effect by explicitly com- j B will grow quadratic in time (at least initially). We
puting the Lorentz force for the simple field configuration 3 This is of course only true, as long as the initial velocity field

depicted in Fig. 7.3. The coordinate axes are shown in is not impeded by the Lorentz force in the momentum equation.
32

Figure 7.3: Simple example illustrating the concept of field-line tension.

can now compute the Lorentz force, which is of state. We now insert an ansatz for wave-like perturba-
tions, that is,
2 1
B %1 = %1 exp [i(kr t)] ,
j B = v0 sin(kz) k 2 t 0 0 , (7.16)
0
v0 k t cos(kz) and accordingly for the other perturbed variables. With
this ansatz, spatial derivatives (curl, gradient, divergence)
which scales with the magnetic energy density, emag
2 will morph into factors of the (vector-like) wavenumber
B0 /0 , of the background field. The jB is always per-
k, preserving the respective vector operation associated,
pendicular to the magnetic fields as depicted on the right
for instance ik . . . ; similarly, the partial time
hand side of Fig. 7.3. We summarise:
derivative will translate into a factor i. In the follow-
the magnetic field distortion grows linear in time ing we will moreover drop the subscript 1, for easier
growth is fastest where the velocity shear is biggest notation. Dividing by common factors of the imaginary
(but: the field line does not actually get displaced) unit, i, we obtain the following algebraic system of equa-
tions:
distorted field lines exert perpendicular stresses
field lines oppose bending via fluid motions % + %0 kv = 0 (7.21)
magnitude of the force scales with magnetic energy 1
%0 v + kp (kB) B 0 = 0 (7.22)
0
B + k (v B 0 ) = 0 (7.23)
 
7.4 Basic MHD waves p %
= 0 . (7.24)
p0 %0
As shown above, the field line tension provides a restoring
Assuming 6= 0, we can solve the resulting system of
force (in the transverse direction) introducing an elastic
equations for three dependent variables
property of the field-line fabric. Together with the flux
freezing property, and the plasmas inertia, this allows % = %0 kv/ (7.25)
to propagate wave-like disturbances along field lines. We p = p0 kv/ (7.26)
can derive the dispersion relation for such MHD waves
B = [(kv) B 0 (kB 0 ) v] / (7.27)
by following the procedure that we have already seen for
acoustic waves. We begin by writing down the linearised already, leaving only the velocity v undetermined. By
MHD equations, where subscript 0 refers to background substituting these relations into the momentum equa-
state, and subscript 1 indicates fluctuations, for in- tion 7.22, we obtain the following vector equation for v,
stance, % = %0 + %1 . As previously, we ignore any terms 
(kB 0 )2

(kB 0 ) (vB 0 )
2
quadratic in fluctuations, and thus obtain: v = k
%0 0 %0 0
B02
  
t %1 + %0 v 1 = 0 (7.17) p0 kB 0
+ + k B 0 (kv) ,
%0 t v 1 + p 10 (B 1 ) B 0 = 0 (7.18)
%0 %0 0
|{z} | {z }
%0 0
t B 1 + (v 1 B 0 ) = 0 (7.19) =c2s =c2A
 
p1 %1 (7.28)
t = 0 , (7.20)
p0 %0 where we identify the (adiabatic sound speed)
with Cp /CV the ratio of specific heats, and where the
r
p0
last equation is derived from the thermodynamic equation cs ,
%0
33

which is a property of the background pressure and den- 7.4.2 Slow and fast magnetosonic waves
sity. The corresponding magnetic term that adds to this
signal speed, is called the Alfvn speed, cA , given by The two remaining roots of (7.31) follow from setting
s the term in square brackets to zero. Solving this simple
B02 quadratic equation in 2 , we obtain
cA ,
%0 0
= k c+ and = k c ,
which is itself a property of the background magnetic field
and the gas density. We can see that this speed is fast for with the fast (c ) and slow (c ) magnetosonic velocity
+
strong magnetic fields (high tension), or low density (low defined by
gas inertia). We will identify this speed as the propaga-
s 
tion velocity of transverse perturbations along the field 1 2
q 
c c + c2 (c 2 + c2 )2 4c2 c2 cos2 .
lines. s s A s
2 A A
Without loss of generality, we can assume that B 0 k z,
and that k y = 0, where the latter implies k x-z- (7.32)
plane. With these restrictions, and indroducing the angle
B 0 , k, we can rewrite equation 7.28 as a simple Note that c cs for the unmagnetised case B0 0
linear eigenvalue problem (EVP) (i.e., for cA 0), illustrating the relation of these waves
! ! with regular sound waves. Because all terms appearing
2k2 (c2A+c2s sin2 ) 0 k2 c2s sin cos vx
0 2 2 2
k cA cos 2 0 vy = 0 .
in (7.32) are positive definite, we easily see that c+ > c ,
k2 c2s sin cos 0 2k2 c2s cos2 vz explaining the names of the two wave branches.
In contrast to the Alfvn wave, the EV for the magne-
(7.29)
tosonic waves is
To guarantee that the linear system of equations has a vx
solution, we evaluate the determinant of the matrix v MS 0 ,
2 2 2 2 2 v z
k (c +c sin )
A s 0 k 2 c2s sin cos
, implying compressional waves (kv 6= 0) with longitudinal
2 2 2 2

0 k cA cos 0
k 2 c2 sin cos
s 0 2
k 2 2
c s cos 2
polarisation (vB 0 6= 0). The different propagation speed
of the two wave branches can be understood in terms of
(7.30)
the pressure perturbations being in phase or out of phase
which yields the dispersion relation with the (linear) magnetic pressure fluctuation B 0B/0 .
2 2 2 2
 More specifically
k cA cos
 4
2 k 2 (c2A + c2s ) + k 4 c2A c2s cos2 = 0 , kv B02 (kB 0 ) (vB 0 )

(7.31) B 0 B
= = ...
which is a third-order polynomial in 2 . The three inde- 0 0
c2 k 2 c2s cos2
 
pendent roots correspond to 3 distinct wave types. = A2 1 p. (7.33)
cs 2
7.4.1 Alfvn waves From this equation, we can see that both perturbations
have the same sign if c > cs cos , and have opposite
The obvious pair of roots in the dispersion relation (7.31) signs if c < cs cos . As a simple exercise, check that this
for ideal-MHD waves can be obtained by setting ( 2 is indeed the case for c = c+ and c = c , as is claimed
k 2 c2A cos2 ) = 0, which yields above.
The angle dependence of the three wave families is il-
= kcA cos ,
lustrated in Fig. 7.4 for the case cA > cs (left panel)
where the plus and minus sign corresponds to the left- and and the case cA < cs (right panel). We can see that
right-travelling wave, respectively. The group speed, c the pure Alfvn wave as well as the slow magnetosonic
/k, of the perturbations is simply the projected Alfvn branch vanish for /2. For 0, the slow speed,
speed cA cos , that is, the maximum propagation speed c min (cA , cs ), that is, it approaches whichever of the
is obtained for k k B 0 , and waves that are perpendicular two speeds is lower. In the same limit, the fast speed
to the field lines do not propagate at all. The associated c max (cA , cs ), that is, it approaches the faster of the
eigenvector (EV) of this particular solution of (7.31) is two wave velocities. As an exercise, verify this behaviour
by setting cos = 1 in (7.32), and applying the binomial
0
formula to simplify terms. The absolute speed limit for
v A vy ,
MHD waves is reached for the fast magnetosonic wave in
0
the case /2. Setting to zero the pcorresponding term
satisfying both kv = 0 ( incompressible perturbation), cos in (7.32), we arrive at cmax
+ = c2A + c2s , which con-
and vB 0 = 0 ( transverse polarisation of the wave). cludes our brief introduction of linear waves in ideal MHD
34

Figure 7.4: MHD wave speeds, c /k, as a function of the angle between the background magnetic field and
the wave vector, k. Left: magnetically dominated case, cA > cs . Right: gas-pressure dominated case, cA < cs .

fluids. In the case of non-ideal MHD, where resistive ef-


fects come into play, the waves will become dissipative,
resulting in wave damping.
35

8 RADIATION 8.2 Attenuation of Radiation

The vast majority of the information that we obtain from Imagine a beam of light going through a medium. The
astrophysical systems is carried by electromagnetic waves. optical depth, , tells how the initial intensity of radia-
In many cases, the photons that reach our eyes or detec- tion, I0 , is attenuated as radiation propagates throughout
tors have interacted with matter after they were emitted. he medium according to
The propagation of radiation through a medium is af-
fected by absorption, emission, and scattering processes. I = I0 e . (8.4)
The equation of radiative transfer describes these inter-
actions mathematically. Before we can write down this There are two different reasons for attenuation: absorp-
equations, we need to introduce several concepts. tion and scattering. In, for example a star, there will
also be emission which counteracts the attenuation. For
a medium that is semi-opaque the optical depth can be
8.1 Solid Angle calculated as Z L
= d` , (8.5)
In order to characterize the energy contained in a beam 0
of light, it is useful to introduce the concept of a solid
where d = d`. Here, is the extinction coefficient,
angle. You can think of it as the two-dimensional angle
is the density and d` is an incremental distance through
in three-dimensional space that an object subtends at a
the medium. The coefficient has units of
point. It is a measure of how large the object appears
to an observer looking from that point. A solid angle is L2
expressed in a dimensionless unit called a steradian, or sr [] = cross section per unit mass . (8.6)
M
for short.
If  1 the medium is said to be optically thin and
Mathematically, the differential of solid angle is defined
as thus most radiation goes through without attenuation. If
dA dA  1 the medium is said to be optically thick and blocks
d 2 = 2 , (8.1)
r r most of the incident radiation. The transition between
Note that what matters is the projection of the element these two regimes takes place around optical depths of
order unity, i.e., ' 1.
of area into the direction given by the unit vector , see
Figure 8.1, The opacity is given by and it can be related to the
number density of the medium, n, and the cross section
In general, for a finite more complex surface, as illus-
of collisions between photons and the components of the
trated in Figure 8.2, the total solid angle is calculated as
an integral medium, , as
1
= n . (8.7)
Z
dA
= . (8.2) `
A r2
Here ` is the mean distance, or mean-free path, between
As an example, the solid angle subtended by a sphere by collisions. Intuitively, high opacity corresponds to a very
an observer inside of it is short distance between each collision experienced by a
2 photon. On the other hand, low opacity corresponds to
4R
= 4 sr , (8.3) essentially freely streaming photons.
R2
and for a semi-sphere, the solid angle is = 2 sr.

Figure 8.2: Finite solid angle.

Figure 8.1: Differential of solid angle. As an example of these concepts, consider photons
propagating through a gas made out of free electrons.
36

The process of a photon scattering off an electron, i.e.,


+ e + e is governed by the Thomson cross section
 2 2
8 e 2
T = 1024 cm2 . (8.8)
3 me c2 3
We can estimate the mean free time between collisions
experienced by the photons as
` 1 Figure 8.3: Particle diffusion.
t = . (8.9)
c cne T
In our case, this would be
8.3 Diffusion of Particles and Photons
12 = 2Dt . (8.17)
There are several problems in astrophysics in which the In our three dimensional case, the dispersions in each in-
propagation of radiation is modeled as a photons diffusing dependent direction add up quadratically so we have
through a medium. Photons do not behave like material
2
particles but we can learn a few things by thinking about r = 312 = 6Dt . (8.18)
the diffusion of material particles and make connections
We can use some of this insights to get a feeling for
with photons at the end.
how a burst of photons would diffuse through a medium.
If particles cannot be created or destroyed, a diffusion
Figure 8.3 illustrates how the number density depends on
equation for the number density of particles n has the
time at a fixed distance R from the origin. The maximum
general form
n of n(R, t) happens when tmax = R2 /4D. If we take the
= D2 n , (8.10) diffusion coefficient for photons to be D = lc/3, we can
t
write
where D is the diffusion coefficient given by   
3 R R 3
1 tmax = = tlight , (8.19)
D hvi ` . (8.11) 4 c l 4
3
where tlight = R/c is the light travel time and = R/l is
where hvi is the mean particle velocity and l is the mean
the optical depth since = R = nR = R/l.
free path.
This equation has as a general solution
8.4 Specific Intensity
n(r 0 , 0) (r r 0 )2 /4Dt 0
Z
n(r, t) = e dr . (8.12)
(4Dt)3/2
The specific intensity of the radiation field I(r, t, , )
If the initial condition is given by a burst or particles at is a scalar quantity that describes the rate of radiative
the origin, i.e., n(r 0 , 0) = N (r 0 ), where N is the total transfer of energy at the point r at time t and depends
number of particles, then at a later time we have in general on the direction given by the unit vector
and the frequency , see Figure 9.1. I(r, t, , ) is de-
N r 2 /4Dt
n(r, t) = e . (8.13) fined to be such that a virtual source area, dA, contain-
(4Dt)3/2 ing the point r, is an apparent emitter of a small but
which we can write as finite amount of energy dE transported by radiation of
frequencies (, + d) in a small time duration dt, i.e.,
N 2 2
n(r, t) = er /2( 2Dt) . (8.14)
( 2)3 (2Dt)3/2 dE I (r, t, )n dA d dt d . (8.20)

Note that if r2  4Dt n t3/2 . Here, the frequency dependence is indicated with a sub-
In order to understand the dispersion of particles at script, n = (0, 0, 1) is the unit vector perpendicular to
the differential area and


time t, we would like to know the mean value r2 taken
over the number density n particle distribution (Note that
= (sin cos , sin sin , cos ) , (8.21)
the mean value hri = 0!). This is simplified if we recall
that for a one-dimensional distribution is the unit vector characterizing the solid angle under con-
1 2 2 sideration expressed in spherical coordinates, with respect
g(x, t) = e(xhxi) /21 . (8.15) to n. Note that n no energy is transported in the di-
21
rection perpendicular to the surface dA. This dot product
The dispersion is given by appears often so we define
2
2 x2 hxi = 12 .


(8.16) n = cos . (8.22)
37

Therefore, the energy density of the radiation field (pho-


ton gas) is
Z Z
2 3 2
E= 3 (p)f (p) d p = 3 (p)f (p)4p2 dp (8.29)
h h
where (p) = cp = h is the energy of a photon with
momentum p and f (p) is the distribution function for the
photons. We can carry out the integral involved as follows
c2 p 2
Z
4 2
E = cp c dp , (8.30)
c3 h3 ecp/kB T 1
h2 2
Z
4 2
E = h h d , (8.31)
c3 h3 eh/kB T 1
3
Z
8h
E = d , (8.32)
c3 eh/kB T 1
Figure 8.4: Intensity in radiative transfer.
This implies that the energy per unit frequency is
8h 3
We can take the moments of the intensity I with re- E = 3 h/k T . (8.33)
spect to in order to find the energy density, the energy c e B 1
flux and the pressure of the radiation as follows In thermal equilibrium, the radiation field must be
isotropic. i.e.,
Energy density 4
E = I . (8.34)
Z c
1 E 1
E = I d [E ] = 3 (8.23) The intensity associated with blackbody radiation is thus
c L Hz
2h 3 1
I = 2 h/k T . (8.35)
Energy Flux c e B 1
Z
E 1 We can obtain the total energy density contained in the
F = I d [F ] = 2 (8.24) black body radiation field by defining the dimensionless
L T Hz
variable x = h/kB T and integrating over all frequencies
4 Z
Pressure x3

8h kB T
E BB = 3 x
dx . (8.36)
1
Z
E 1 c h e 1
P = I 2 d [P ] = 3 (8.25)
c L Hz The integral can be shown to give 4 /15 so
4
The differential of solid angle d in spherical coordi- 2 (kB T )
EBB = , (8.37)
nates is d = sin dd and can be written in terms of 15 (~c) 3
as d = dd. If the problem has axial (azimuthal
or
symmetry), we can write d = 2d.
EBB = aT 4 , (8.38)
Note that for isotropic radiation the energy flux is
Z where we have defined the radiation constant
F = I d = 0 , (8.26) 2 kB 4
4 a= , (8.39)
15 (~c)3
whereas the radiation pressure satisfies
which is related to the Stefan-Boltzmanns constant via
Z
I 4 I E 4
P = 2 d = = . (8.27) a= . (8.40)
c 4 3 c 3 c
This implies that the total intensity of black body radia-
8.5 Black Body Radiation tion can be written as
IBB = T 4 . (8.41)
We now seek to determine the intensity associated with
black body radiation. Recall that in thermodynamic equi- Note that the total luminosity emitted by a spherical
librium the number density of photons is blackbody of radius R is given by the Stefan-Boltzmanns
law as
1 L = 4R2 T 4 . (8.42)
n() = h/k T . (8.28)
e B 1
38

9 RADIATIVE TRANSFER

9.1 The Equation of Radiative Transfer

The equation for radiative transfer describes the evolution


of the intensity when matter in the medium can absorb,
emit, or scatter photons and is given by

1 I  
+ I = I + j (9.1)
c t Z
s
+ I0 (0 )(, 0 ) d0 .
4
Here, the total extinction coefficient, , is the sum of the
extinction coefficient due to absorption, a , and extinc- Figure 9.1: Slab geometry.
tion coefficient due scattering, s , i.e.,

= a + s . (9.2) where we have defined the single scattering albedo

Both of these effects act to attenuate the amount of ra- s


= . (9.8)
diation along the direction given by and can depend a + s
explicitly on the frequency . The emissivity j , on the
Defining the source function S as the quantitive in
other hand, acts to increase the amount of radiation in
square brackets, the equation for radiative transfer can
the direction . The last term on the right hand side de-
be written as
scribes how the scattering of photons from all directions dI
changes the intensity in the direction . = I S . (9.9)
d
In general, the equation for radiative transfer is an
integro-differential equation with the integral making This means that the intensity tends to the source func-
analysis difficult. It can sometimes be simplified by argu- tion.
ing that I will be isotropic, thereby making it possible to Kirchhoffs law of thermal radiation states that in local
take I out of the integral which then only depends on the thermodynamic equilibrium (LTE) the emissivity satisfies
scattering function (, 0 ). If the scattering process is j
isotropic then (, 0 ) = 1 and the problems simplifies IBB , (9.10)
a
significantly.
Note that the units of I and j are Note, however, that I IBB only if  1.
erg erg
[I ] = [j ] = . (9.3)
cm2 s st Hz gr s st Hz 9.2 Diffusion Approximation for Radiative Transfer
If we consider a plane-parallel atmosphere, we have
Let us take moments of the equation of radiative transfer
d d with respect to . The zeroeth moment yields
= , (9.4)
dz ds
E F
where = cos and is the angle subtended by the ray = cE + a cEBB + s cE , (9.11)
t z
under consideration and the vertical direction, and z is = a c [E EBB ] , (9.12)
the variable measuring depth taken to increase downward,
= c(1 ) [E EBB ] . (9.13)
see Figure 9.1. Note that dz = ds cos = ds.
If we define the differential optical depth measured in The first moment is
the vertical direction as
1 F P
c = F . (9.14)
d = dz = ds , (9.5) c t z

then Note that the equation for the moment n involves moment
d
ds = , (9.6) n + 1, i.e., the equation for the energy involves the flux,
and the equation for the flux involves the pressure. In
and the equation of radiative transfer becomes order to have a complete set of equations we need closure
relations to relate E, F and P . One way to do this is to
 
use the Eddington approximation that assumes that
Z
dI j 0
= I + I d , (9.7) the intensity depends linearly on , i.e., I = I + I .
d 4 0 1
39

Using the definitions for E, F and P , we obtain information about the physical state of the gas. In order
Z to simplify this problem we will assume a plane-parallel
1 4
E = I d = I0 , (9.15) geometry and that all the opacities involved do not de-
c c pend on frequency and thus the atmosphere is said to be
Z
4 grey. We will also ignore scattering.
F = I d = I1 , (9.16) In radiative equilibrium, i.e., F/z = 0 and thus
3
Z F/ = 0 with
1 4 E
P = I 2 d = I0 = . (9.17) c 1 E c E
c 3c 3 F = = . (9.24)
3 z 3
This implies that the intensity is
Integrating this equation with respect to , we obtain
cE 3F
I= + . (9.18)
4 4 3F
E( ) = + E(0) . (9.25)
In steady state, equation 9.14 leads to an expression c
for the flux In order to determine the constant of integration, we make
c 1 E
F = , (9.19) use of an approximate boundary condition by assuming
3 z
that the emergent flux is given by
which provides a closure relation between F and E. Note
that because of our convention, a positive value for F c
F = E(0) . (9.26)
means that the flux is travelling upwards (in the negative 2
z direction). This means that about half of the photons at the stellar
If we consider a slab geometry, substituting this ex- surface stream freely out of the star. This leads to
pression for the flux in equation 9.13 provides a diffusion
equation for the energy in one dimension 3F 2F
E( ) = + , (9.27)
c c
2
E c E
= c(1 ) [E EBB ] . (9.20) thus the energy density as a function of optical depth is
t 3 z 2
 
The fact that the right hand side of this equation does 3F 2
E( ) = + . (9.28)
not vanish in general reflects the fact that the number c 3
density of photons is not conserved.
If we now define the effective temperature Teff as the
temperature that a black body must have in order for the
9.3 Thermalization Time emitted flux to be
4
If we consider radiative equilibrium, i.e., F/z = 0 then F = Teff . (9.29)

E In order to relate this effective temperature at the surface


= a c [E EBB ] . (9.21) to the temperature at the optical depth we need to
t
realize that because of radiative equilibrium the energy
which can be integrated to obtain density of the radiation field E( ) must correspond to
h i the energy density of the black body radiation at that
E(t) = EBB 1 e(a c)t , (9.22) optical depth, i.e.,

When t = 0 the energy density is zero and as time goes 4 4 4


E( ) = EBB ( ) = IBB ( ) = T ( ) , (9.30)
to infinity the energy density approaches the black body c c
energy density. The characteristic time for this to happen
which implies that the temperature of the atmosphere as
is given by
a function of optical depth is
1
tc = . (9.23)
 
4 3 4 2
a c T ( ) = Teff + . (9.31)
4 3

9.4 Thermal Structure of a Grey Atmosphere

We now would like to determine the thermal structure of


a model stellar atmosphere. This is the site where the
radiation produced at the stellar interior escapes and it
and thus the photons that we observe carry important
40

10 RADIATIVE PROCESSES 10.2 Cyclotron Radiation

10.1 Larmor Formula A charged particle moving in the presence of a magnetic


field is subject to the Lorentz force that acts perpendicu-
The electric and magnetic fields far away from an accel- lar to the direction of motion and the magnetic field lines.
erated, non-relativistic charged particle are given by This force accelerates the particle causing it to emit cy-
e clotron radiation as it spirals around the magnetic field.
E rad = n(nv) , (10.1)
Rc2 This cyclotron radiation from plasma provides an impor-
B rad = nE rad . (10.2) tant source of information about the magnetized plasma
in the interstellar medium or around black holes and other
Here, e is the charge of the particle, c is the speed of light,
astronomical environments. When the electrons are mov-
R is the distance from the charge to the point where the
ing at relativistic speeds, cyclotron radiation is known as
electromagnetic fields are measured along the direction
synchrotron radiation (see below).
given by the unit vector n and v is the acceleration, with
The equation of motion for a particle with charge e
v  c. Note that the instantaneous direction of E rad is
moving with speed v in the presence of a magnetic field
determined by v and n. If the charge accelerates along a
B is given by
straight line, the radiation will be 100% linearly polarized
in the plane containing both vectors. e e
mx = (vB) = vB sin p , (10.7)
c c
where p is the pitch angle between v and B, see Fig-
ure 10.2. Note that this force will only alter the motion
of the particle in the direction perpendicular to the mag-
netic field, v = v sin p , while leaving unchanged the
component parallel to it, , vk = v cos p . The centripetal
acceleration induced by the Lorentz force
2
v eB
Figure 10.1: Electromagnetic fields in the radiation zone. v = = v , (10.8)
rg mc
Recall from electromagnetism that the Poynting vector can be written as
c
S= (E rad B rad ) , (10.3) v = rg g , (10.9)
4
represents the directional energy flux density (the rate of where we have defined the particle gyro-radius
energy transfer per unit area) associated with the electro-  mc 
rg = v , (10.10)
magnetic field. Note that the Poynting vector is directed eB
along n and its modulus is given by and gyro-frequency
c 2 c e2 v 2 2
   
S= Erad = sin , (10.4) eB 7 B 1
4 4 R2 c4 g = 2 10 s . (10.11)
mc G
where is the angle between the acceleration vector v
and n. The power, energy per unit time, emitted in a We can use Larmors equation to calculate the total
given solid angle is then power emitted via cyclotron radiation
dP 1 e2 v 2 2 e2 2
= R2 S = sin2 . (10.5) P = v . (10.12)
d 4 c3 3 c3
This corresponds to the characteristic dipole pattern
sin2 in which no radiation is emitted along the direction
of acceleration, and the maximum is emitted perpendic-
ular to this direction, see Figure 10.1.
The total power emitted by the accelerated charge is
obtained by integrating over all solid angles
2 e2 v 2
Z
dP
P = d = , (10.6)
4 d 3 c3
where we have used that d = 2 sin d because of the
azimuthal symmetry with respect to v. Thus, the power
emitted is proportional to the square of the charge and the
square of the acceleration. This is referred to as Larmors
equation for a single accelerated charge. Figure 10.2: Cyclotron radiation.
41

Figure 10.3: Synchrotron radiation.


Figure 10.4: Synchrotron spectrum for one particle.
2
Using that the acceleration is v = v /rg = g2 rg = g v ,
the total power is thus
power emitted by the particle in the frame in which it
2 e2 2 2 2 is at rest, P 0 = dE 0 /dt0 , is the same as the power ob-
Pcycl = v sin p , (10.13) served in the lab frame where the particle is moving with
3 c3 g
speed v, P = dE/dt, therefore
Averaging over all angles p , we obtain
4
4 Psync = cT uB 2 2 . (10.17)
Pcycl = cT uB 2 , (10.14) 3
3
The fact that synchrotron radiation is beamed means
where = v/c, uB = B 2 /8 is the energy density asso-
that an observer sees an intensity that varies in time, this
ciated with the magnetic field and we have defined the
implies that the radiated energy is no longer emitted in
Thomson cross section
a unique frequency but in a spectrum. The synchrotron
 2 2
8 e spectrum for a single particle increases with 1/3 for 
T = . (10.15) c and decays like exp(/c ) for  c , reaching the
3 mc2
peak at max ' 0.3c with c ' (3/2) 2 g , with g =
g /2, see Figure 10.4.
Assuming that the particles loose most of their energy
10.3 Synchrotron Radiation
at this frequency, in the absence of a process that would
replenish this energy, the lifetime of a source that emits
When a charged particle moves with relativistic speed via synchrotron can be estimated as
in the presence of a magnetic field, it emits synchrotron
radiation. At high speeds the radiation pattern is no  1/2  B 3/2
c
longer an isotropic dipole pattern and it is beamed into tcool 0.5 Gyr . (10.18)
106 Hz 106 G
forward-pointing cone of radiation, see Figure 10.3. The
openingpangle of this cone is proportional to 1/, with So far we have been interested in the synchrotron radia-
2
= 1/ 1 (v/c) . This planar acceleration geometry tion emitted by a single particle. We can derive the spec-
produces radiation that is linearly polarized when ob- trum emitted by a collection of particles if we know the
served in the orbital plane, and circularly polarized when number density of the emitting particles as a function of
observed from a direction perpendicular to that plane. their energy, n(). While understanding the synchrotron
We will not provide a formal derivation of the total radiation emitted by a plasma in thermal equilibrium if
power emitted by a charged particle emitting synchrotron of interest, it turns out that most synchrotron radiation
radiation but will limit ourselves to use what we have observed in astrophysical sources, including relativistic
learned from cyclotron radiation and argue how some of jets in AGN and supernova remnants, is of non-thermal
the equations need to be modified to account for relativis- origin.
tic effects. If the spectrum of emitting electrons is a power-law,
In the electrons frame, the acceleration induced by the n() = Cp , then the radiation spectrum is given ap-
Lorentz force, and thus the relativistic angular frequency, proximately by
increases by a factor . Therefore, according to Larmors (p1)/2
18

equation the power emitted by the electron in its own rest j 2 1022 CB (p+1)/2 6.26 10 Hz erg
,

frame is cm3 sHz
2 e2 2 2 2 (10.19)
0
Psync = 3
g v sin p , (10.16)
3c Note that this emissivity increases without bound at lower
It can be shown that because the way that energy and frequencies, potentially becoming larger than the emis-
time transform when subject to a Lorentz boost, the sivity corresponding to black body radiation, which is
42

is proportional to the velocity and the damping frequency

2e2 2
d = . (10.22)
3me c3

The solution to this equation of motion is

e E 0 eit
x(t) = . (10.23)
me 02 + id
2

Therefore, the power emitted by the electron over one


Figure 10.5: Non-thermal synchrotron spectrum. cycle of the incident electromagnetic wave, i.e.,

1 t+ /2
Z
proportional to 2 at low frequencies. In reality, these hP i = P dt , (10.24)
low frequency photons will be self-absorbed by the emit- t /2
ting plasma and the emissivity at low frequencies behaves,
quite universally, as 5/2 below a break frequency b that where = 2/, is
depends on the source, see Figure 10.5. This behavior at  2 
(e2 /m2e ) E02

low frequencies is a signature of non-thermal synchrotron e
hP i = . (10.25)
radiation and the frequency b provides a measure of the 3c3 ( 2 02 )2 + d2 2
optical depth through the emitting plasma. Measuring
the spectral index of the spectrum at high frequencies Note that the time-averaged incident flux is
provides insight on the power-law index of the emitting c 2
particles and the observed power can be used to estimate hSi = E . (10.26)
8 0
the magnetic field.
Defining the total cross section as the ratio of the scat-
tered power to the incident flux, i.e., = hP i/hSi, we
10.4 Scattering of Electromagnetic Waves by
obtain
non-relativistic Particles
4e2 d 2
  
Note that the power depends on the product ev which can = . (10.27)
me c ( 2 02 )2 + d2 2
be thought to be associated with the changes induced in
the dipole moment characterizing two charges of opposite Note that re = e2 /me c2 ' 3 1013 cm is the classi-
sign separated by a distance x, i.e., d = ex, since d = cal electron radius. For wavelengths much greater that
ex = ev and thus the radius of the electron the damping frequency is much
2

2 |d| smaller than the frequency of the incident wave, i.e.,
P = . (10.20) d  . The general scattering cross section that we
3 c3
derived enable us to study a few limiting cases of inter-
We have managed to link the power emitted in terms
est.
of the second time derivative of the dipole moment char-
acterizing the system. We now briefly address what hap-
pens when the changes in the dipole moment are induced 10.4.1 Thomson Scattering
by the action of an electromagnetic wave on the system
characterized by a dipole moment d. Consider the Thom- If the electrons are free, i.e., not bound to an atom, then
son model for the atom in which electrons are assumed we can take 0 = 0 and obtain the Thomson scattering
to be in an equilibrium configuration such as to neutral- cross section
ize the positive charges. When the electrons are displaced
from this configuration by a distance x, they will oscillate 8 2
T = r , (10.28)
with a characteristic frequency 0 . The dipole moment 3 e
associated with this displacement is d = ex.
If we now consider an incident electromagnetic wave, which is independent of the frequency of the incident ra-
diation, . This result is valid only if the energy of the in-
E = E 0 ei(kxt) , acting on this atom we can write the
cident photons is much smaller than the rest mass energy
equation of motion for the displaced electron as
of the electrons, i.e., me c2 0.5MeV. When the energy
e
x = 02 x d x E 0 eit , of the incident radiation is not negligible compared to
(10.21)
me the rest mass energy the non-relativistic approach breaks
where we have included the driving force due to the elec- down and the relevant cross section is the Compton cross
tromagnetic wave and the radiative damping force, which section (see below).
43

Figure 10.7: Compton cross section.

Figure 10.6: Compton scattering


of the particles relativistic and quantum effects cannot
be ignored. If a photon of initial wavelength interacts
10.4.2 Rayleigh Scattering with an electron originally at rest, see Figure 10.6, the
photon wavelength after the interaction 0 is given by
If the frequency of the incident radiation in  0 , we
obtain 0 = + c (1 cos ) , (10.32)
 4
8 2
R = r . (10.29) where we have defined the Compton wavelength as
3 e 0

Note the strong dependence of the cross section on 4 . h


c = = 0.0243 . (10.33)
This strong wavelength dependence of the scattering me c
means that shorter (blue) wavelengths are scattered more
strongly than longer (red) wavelengths. This results in Note that when the incoming photon has an energy much
the indirect blue light coming from all regions of the sky. smaller than the rest mass of the electron, h  me c2 ,
Rayleigh scattering applies to particles that are small we obtain the classical result 0 = which is described
with respect to wavelengths of light. The scattering by by Thomson scattering by free electrons. In the ultra-
relativistic limit, h  me c2 and the scattered photon
particles similar to, or larger than, the wavelength of light
is typically treated by the Mie theory. ends up roughly with the electron rest-mass energy. The
principal effect is to reduce the cross section from its clas-
sical value as the photon energy increases, thus Compton
10.4.3 Resonant Scattering of Line Radiation scattering becomes less efficient as the energy of the pho-
ton increases, see Figure 10.7.
If the frequency of the incident radiation is ' 0 , then
We can heuristically derive the power emitted via
the cross section becomes
Compton processes as follows. A relativistic electron
 2
e moving through a bath of photons is subject to a radia-
line = L, (10.30) tive energy flux given by S 2
me c rad = curad in the reference
frame where the electron is at rest. Thus, estimating the
where L is the Lorentz profile radiated power via P 0 = Srad , and recalling that the
  power is a Lorentz invariant, i.e., P 0 = P , we have
1 (d /4)
L= , (10.31)
( 2 02 )2 + (d /4)2 4
PComp = cT urad 2 2 , (10.34)
with = /2 and 0 = 0 /2. This functional de- 3
pendence with frequency is intimately related to the line
profile that you will see in many astrophysical contexts. Note that this result is very similar to what we obtained
Realize that this result has been obtained using classical when we calculated synchrotron radiation where the en-
theory and quantum corrections, introduced via "oscilla- ergy density corresponds to the magnetic field. In astro-
tor strengths" coefficients need to be considered. physical sources which emit via synchrotron and Comp-
ton radiation, say in radio waves and gamma-rays, the
ratio Psync /PComp can be used to estimate either the mag-
10.5 Compton Scattering netic field if we know the local radiation field, or to gain
insight into the ambient radiation field is the magnetic
When the energy of the incident radiation cannot be field can be estimated by other means, e.g., measuring
strictly neglected with respect to the rest mass energy Faraday rotation.
44

10.6 Inverse Compton Scattering and SZ-effect the scattering protons, the emissivity associated with
Bremsstrahlung radiation is given by
When a photon of initial energy  interacts with a fast  
moving electron with a Lorentz factor , the energy of the j 6.8 1038 ne ni Z 2 T 1/2 exp h gff erg
,
0 2 kT cm 3 sHz
photon gets boosted to  = . This is how a cosmic ray
electron with energy 109 eV, i.e., 103 , can boost up a (10.35)
CMB photon ( 103 eV) to X-ray (with 0 104 keV).
This process is intimately related to the so-called where ne and ni are the numbers densities of the electrons
Sunyaev-Zeldovich effect (or SZ effect). This is the result and ions with charge number Z and the Gaunt factor gff
of high energy electrons distorting the cosmic microwave is a number of order unity for practical purposes. The
background (CMB) radiation through inverse Compton total energy radiated can be obtained by integrating over
scattering, in which the low-energy CMB photons receive all frequencies
an average energy boost during collision with the high en- erg
ergy electrons in clusters of galaxies. The magnitude of jbrem 1.4 1027 n2e T 1/2 3 , (10.36)
cm s
this temperature change has a very unique frequency de-
pendence, below 220 Ghz (1.4mm) the SZ effect produces where we have assumed a plasma made of Hydrogen.
a decrement, or a "cold spot" relative to the mean CMB The thermal bremsstrahlung energy spectrum is flat
temperature. Above 220 Ghz, it produces a temperature for photon energies h  kT , see Figure 10.8. This is
increase ("hot spot"). The distortions observed in the a rather distinctive feature that provides a unique ob-
CMB can be used to detect dense clusters of galaxies. servational signature. Because of the exponential cut-off
the spectrum turns over at energies h kT , which pro-
vides a way to measure the temperature of the emitting
10.7 Thermal Bremsstrahlung Radiation plama. Thermal bremsstrahlung is the dominant source
of X-rays in the hot (107 108 K!) intracluster medium
An unbound electron moving in the presence of an ion permeating cluster of galaxies. This radiation is easily ob-
will be deflected from a straight-line trajectory and emit served with space-based telescopes such as Chandra X-ray
bremsstrahlung radiation. This radiation has a contin- Observatory, XMM-Newton, ROSAT, ASCA, EXOSAT,
uos spectrum which becomes more intense and whose Suzaku, RHESSI and future missions like IXO and Astro-
peak intensity shifts toward higher frequencies as the H. Bremsstrahlung is also the dominant emission mecha-
change of the energy of the accelerated particles increases. nism for H II regions at radio wavelengths.
Bremsstrahlung emitted from plasma is sometimes re- If the plasma is optically thin, the bremsstrahlung
ferred to as free/free radiation when the emitting charged radiation leaves the plasma, carrying part of the in-
particles are free both before and after the deflection (ac- ternal plasma energy. This effect is known as the
celeration) that caused the emission. A formal deriva- bremsstrahlung cooling and can be quite effective at high
tion of the equations involved is beyond the scope of temperatures. The characteristic cooling time can be es-
these notes and the details can be found, for example, timated by taking the ratio between the thermal energy
in Rybicki & Lightman (1986); Shu (1991); Padmanab- density ne kT and the energy lost per second per unit vol-
han (2000); Melia (2009). Here, we will limit ourselves ume jbrem as
to analyze the result and understand some of its conse-   1
quences. T ne
tcool 3 Gyr . (10.37)
108 K 0.01cm3

If the plasma is optically thick then the emitted pho-


tons will interact with the emitting matter before es-
caping the system. The particles with which the pho-
tons interact have energies of the order of kT , therefore
the photons with lower energy will be upscattered and a
bump in the energy spectrum will develop close to kT ,
at the expense of a decrement at lower energies. If the
optical depth is high enough, the radiation and plasma
reach local thermodynamic equilibrium and the radiation
emerges with the intensity characterizing blackbody ra-
diation with temperature T .

Figure 10.8: Thermal Bremsstrahlung.

If the electrons are in thermal equilibrium with


45

11 SPHERICAL ACCRETION

11.1 Steady isentropic flow of gas

For a steady isentropic flow, the Lagrangian derivative of


the entropy vanishes, i.e., ds = 0. For an ideal gas, this
implies that
P . (11.1)
Recall that for an ideal gas we had
kB T
P = , (11.2)
m
and thus the pressure P is related to the internal energy Figure 11.1: Sketch for isentropic flow of gas considering
via gas that is initially at rest and allowed to escape through
P = ( 1) , (11.3) a small cavity.
and the sound speed cs is such that
The maximum possible value of v is attained when the
P
c2s = . (11.4) pressure vanishes, i.e.,

vmax = v(P = T = 0) . (11.12)
The Bernoulli equation is satisfied along each stream-
line Therefore Bernoullis equation reads
P 1 2
 + + v = const. , (11.5)
2 1 1 2
h0 + v02 = vmax . (11.13)
2 2
or using the expression for the enthalpy (remember: its
the amount of heat content used or released in a system which leads to p
at constant pressure) vmax = 2h0 , (11.14)
P P or expressed in terms of the sound-speed c0
h=+ = , (11.6)
1 r
2
the equation becomes vmax = c0 . (11.15)
1
1
h + v 2 = const. , (11.7)
2
11.3 Sonic Point
we obtain the following three relations:
In general, we will have
c2s
h= , (11.8)
1 P 1 c20
+ + v2 = , (11.16)
  2 1
P
= , (11.9) c2s 1 c20
P0 0 + v2 = . (11.17)
1 2 1
 1
 1
T Taking the ratio of
= . (11.10)
0 T0 P
c2s = , (11.18)

and
11.2 Maximum Flow Speed P0
c20 = , (11.19)
0
Using Bernoullis equation, we can find the maximum yields
speed at which gas can flow in terms of the speed of sound c2s

s
1
when the gas is at rest. When there is no gas flow, v0 = 0 = , (11.20)
c20 0
and the enthalpy is
and thus
P0 c20 2
h0 = = . (11.11) v2 = [c2 c2s ] , (11.21)
1 0 1 1 0
46

becomes
"  1 #
2 2c20
v = 1 , (11.22)
1 0
" #
2c20
  1
T
= 1 , (11.23)
1 T0
"   1 #
2c20 P
= 1 . (11.24)
1 P0

There is a critical point, that we denote with "?", at


which the flow speed v coincides with the sound speed cs ,
i.e., v? c? .
At this sonic point
Figure 11.2: Sketch of the de Laval nozzle.
c2? 1 c20
+ c2? = , (11.25)
1 2 1
11.4 de Laval Nozzle
and thus
2
c2? = c2 = v?2 , (11.26) We are now interested in understanding the properties of
+1 0 a steady isentropic flow which is confined to move along
and recall that the maximum speed for the flow is a pipe with a varying area A, as shown in Figure 11.2.
Recall the equations describing fluid dynamics
2 2 2
vmax = c0 . (11.27)
1
+ (v) = 0 , (11.34)
We can then write the temperature, pressure, and den- t
sity as a function of the flow speed either in terms of the v 1
+ (v )v = P , (11.35)
sound speed when the fluid is at rest, c0 , or the sound t
speed at the critical point, c? as
" In steady state t 0, equation (11.34) leads to
 2 # "  2 #
1 v 1 v
T = T0 1 = T0 1 (v) = 0 , (11.36)
2 c0 + 1 c?
(11.28) which we can integrate to obtain
Z Z
" #
 2 1 " #
 2 1 vdv = vdA = vA = const . (11.37)
1 v 1 v
P = P0 1 = P0 1
2 c0 + 1 c? This states that the amount of mass that flows through
(11.29) the area A per unit time is constant. We can write this
as
"  2 # 11 "  2 # 11 M = vA = jA , (11.38)
1 v 1 v
= 0 1 = 0 1 where we have defined the mass current density
2 c0 + 1 c?
(11.30) j v . (11.39)
Using these expressions we can relate the temperature,
pressure, and density at the sonic point with the tem- We now seek an expression connecting dA with dv.
perature, pressure, and density when the gas is at rest.
d(vA) = vdA + d(v)A , (11.40)
Setting v = c? , we find
 
2 d
T? = T0 (11.31) d(v) = dv + vd = + v dv , (11.41)
+1 dv
Lets do the same for equation 11.35.
 1
 1
2 dP d dP c2s
? = 0 (11.32) vdv = = = d , (11.42)
+1 d


2
 1 which yields
P? = P0 (11.33) d v
+1 = 2 , (11.43)
dv cs
47

Figure 11.3: Sketch of subsonic and supersonic flows in de Laval nozzles and diffusers.

Substituting this in (11.41) and introducing the Mach


number M = v/cs , we obtain

dj = d(v) = (1 M2 )dv . (11.44)

Using this expression in (11.40) we obtain


d(vA) = vdA + 1 M2 Adv = 0 ,
 
(11.45)
or
dA  dv
= 1 M2

. (11.46)
A v
Note that if v < cs , v will decrease if the cross-sectional
area A increases. However, the opposite is true in the
regime where v > cs ; where v increases as A increases.
We can also related the fractional chance in the area
with the change in density, pressure, etc. in terms of the
Mach number M = v/cs
dA d
= M2 [1 M2 ] , (11.47)
A
dA P dP
= 2
[1 M2 ] . (11.48)
A v P

11.5 Subsonic and Supersonic Flows


Figure 11.4: Sketch showing a nozzle coupled to a diffuser
In the previous section we have seen that the character- such that the gas speed reaches the sound speed at the lo-
istics of the flow are very different depending on whether cation where the devise has the minimum cross-sectional
the fluid is in the subsonic M < 1 or supersonic M > 1 area Amin .
regime. Figure 11.3 summarizes the different situations
we can have depending on whether we start with a sub-
sonic or supersonic flow and whether the area of the cross
section increases or decreases in the direction of the flow.
What happens if we couple a nozzle to a diffuser and
we require that the flow transitions from subsonic to su-
personic at the location where the cross-sectional area A
reaches its minimum value Amin , as shown in Figure 11.4?
48

This implies that the flux of matter through a spherical


surface is constant

r2 v = const . (11.56)

It makes sense to define the accretion rate as

M = 4r2 v , (11.57)

where v is the modulus of the radial speed.


The momentum equation can be written in spherical
Figure 11.5: Sketch showing the dependence of the matter coordinates as
current density as a function of flow speed in the de Laval v 1 P GM
nozzle coupled to a diffuser shown in Figure 11.4. v + + 2 = 0. (11.58)
r r r

Recall from equation (11.44) we can write the rate of We can re write the term involving the pressure gradi-
change of the matter current density with respect to the ent in terms of the sound speed by recalling that
speed of the flow as P P
= = c2s , (11.59)
dj dv  2
 r r r
= = 1M . (11.49)
dv dv
and realizing that, from equation (11.55), we have
This implies that the matter current density reaches an
extreme value when M = 1, see Figure 11.5. Moreover,
= 2 (vr2 ) . (11.60)
by construction, this is the sonic point so that r vr r

jmax = j? = ? c? . (11.50) Using this, the momentum equation can be written as

Using equations (11.26) and (11.32) v c2 GM


v s2 (vr2 ) + 2 = 0 . (11.61)
r vr r r
 +1
 2(1)
2
jmax = 0 c0 . (11.51) This can be rearranged to obtain
+1
c2s v 2 2c2s r
   
The maximum value for the rate at which matter can flow 1 GM
1 2 = 2 1 . (11.62)
is 2 v r r GM
Mmax = ? c? Amin . (11.52)
This equation allows us to learn a few interesting things.
which is independent of the external pressure Pe provided
At very large distances, r , cs c s and
Pe < P? .
2c2s
 
1 r < 0, (11.63)
11.6 Spherical Accretion of a Polytropic Gas GM

and thus the right hand side is positive. On the left hand
Let us consider the accretion of matter onto a central
side, we expect
object of mass M . This is a fairly complicated process
v 2
but we can get a handle on the basic physics invoking < 0, (11.64)
r
some assumptions. We will assume the process to be time

independent, spherically symmetric, and isentropic. The since v ' 0. This implies that at large distances the
equations describing the gas dynamics are then flow is subsonic, i.e.,

v 2 < c2s .
+ (v) = 0 , (11.53)
t
As r decreases the quantity
v P
+ (v )v = , (11.54)
t 2c2s
1 r
where t = 0, P , = GM/r, and all the fluid GM
variables are only a function of the spherical radius r. will tend to zero and vanish at the critical radius rc
The continuity equation can be written in spherical co-
ordinates as 1 GM
1 2 rc = , (11.65)
(r v) = 0 . (11.55) 2 c2s (rc )
r2 r
49

we denote by v? = c? .
c2
 
2 1 1
c? + 2 = . (11.71)
2 1 1
This implies that the speed of sound at the critical radius
is related to the speed of sound at infinity via
r
2
c? = c (11.72)
5 3

whereas the densities at these locations are related via


 2
 1
c?
? = (11.73)
c
Figure 11.6: Sketch of the solutions to Equation (11.62).
Using these results, we can write for the (constant) ac-
or in terms of the "?" variables we defined before cretion rate

1 GM M = 4r2 v = 4r?2 ? c? , (11.74)


r? = . (11.66)
2 c2? or, in terms of the values at infinity,4
We can use typical values to get a sense of the scales  2
involved GM
M = 4() c (11.75)
c2
104 K
  
M
rc ' 7 1013 cm 1AU . (11.67) with
T M  53
 2(1)
1 2
() = (11.76).
For r < rc we must have v 2 > c2s . At r rc the flow 4 5 3
transitions from subsonic to supersonic! Furthermore at In order to complete the solution, we find v(r) in terms
r rc , we must have that either of cs (r) from the constancy of M as
 2
 1
v 2 c2s (11.68) M c
v(r) = . (11.77)
4r2 cs
or
v 2 Substituting this relation into the Bernoulli equation
0 (11.69) (11.70) we can solve for cs (r) and using equation (11.77)
r
obtain v(r), and thus for fixed M = 4r2 (r)v(r) we can
Equation (11.62) can be integrated to provide v/cs as find (r).
a function of radius, or r/rc . The solutions are shown in
Figure 11.6. Note that not all of the curves give physically
11.8 Asymptotic Properties of Transonic Flows
viable solutions. The curves labeled with 3 and 4 can-
not describe sonic transitions, whereas the ones labeled
with 5 and 6 are multi-valued and are thus unphysical. For r > r? : weak gravity, pressure dominates, subsonic
The solutions labeled with 1 and 2 correspond to an
accretion flow and a wind, respectively. M = 4vr2 , (11.78)
 2
GM
M = 4 c , (11.79)
11.7 Bondi Accretion c2
and v r2 and const.
Integrating equation (11.58), we obtain the Bernoulli For r < r? : gravity strong, weak pressure, supersonic
equation for a spherical flow (free-fall) r
2GM 1
v 2
c 2
GM c 2 v v r 2 , (11.80)
+ s = const = , (11.70) r
2 1 r 1
M 3

where we have assumed that the flow speed vanishes at 2


r 2 . (11.81)
4r v
infinity. 4 Note that if the accreting body were moving with respect to
On the other hand, at the critical radius r = rc , the the interstellar medium with velocity v, we can obtain the accretion
flow speed matches the speed of sound vc = cs (rc ), which rate onto it by substituting c2 in the denominator by c2 + v2 .
50

11.9 Accretion Luminosity This means that there is an upper limit to the accre-
tion luminosity, such that radiation forces do not prevent
Matter falling onto a central object at a rate M will re- the accretion of matter. This maximum luminosity is
lease gravitational energy at a rate given by achieved when frad = fgrav and is called the Eddington
luminosity
GM M 4GcM
Lacc (r) = . (11.82) LEdd = . (11.88)
r
If all the kinetic energy of the infalling matter is given In summary, the Eddington luminosity is the maximum
up in the form of radiation at the stellar surface, the luminosity a body can achieve via steady-state spherical
accretion luminosity is given by accretion when there is balance between the force of ra-
diation acting outward and the gravitational force acting
GM M inward. For pure ionized hydrogen = T /mp
Lacc = . (11.83)
R?
4GcM mp
LEdd = , (11.89)
Typical values for accreting white dwarfs and neutron T
stars in binary systems are M erg
= 1038 , (11.90)
WD 33 erg NS 36 erg
M s
Lacc = 10 , Lacc = 10 , (11.84) M
s s = 3 104 L . (11.91)
M
where we have assumed that RWD 109 cm and RNS
105 cm and M 1016 gr s1 which is roughly 1.5 Note that this is significantly larger than the typical val-
1010 M yr1 . Note that this values are much larger ues we have quoted for accreting stars in binary sys-
than what we would expect when these objects are ac- tems. However, supermassive black holes are thought to
creting from the interstellar medium! go through phases where they accrete close to this limit!
In the case of a black hole, which lacks a hard surface,
the accretion luminosity can be written as

GM M
Lacc = 2 = M c2 , (11.85)
RBH
where we have used the definition of the Schwarzschild
radius RBH = 2(GM/c2 ). The parameter is a measure
of how efficiently the rest mass energy of the accreted ma-
terial is converted into radiation. Realistic estimates sug-
gest that 0.1 for accretion onto compact objects. This
is very large compared to the efficiency of 0.007, at
which rest mass energy is converted into radiation when
hydrogen is burned into helium in stellar interiors.

11.10 The Eddington Limit

A radiation field characterized by a flux F exerts a force


per unit volume given by
F F
frad = n = , (11.86)
c c
where n is the number density of the particles whose in-
teractions with the photons is characterized by the cross
section , and we have used that n = , where is the
absorption coefficient. If the flux of photons arises due to
the release of gravitational energy due to accretion onto
a central object then F = L/4r2 , where L is the lumi-
nosity arising from accretion. The accreting matter will
be pushed away if the force exerted by the radiation field
can overcome the gravitational attraction given by
GM
fgrav = . (11.87)
r2
51

12 DISK ACCRETION 12.2 On Angular Momentum Transport

12.1 Introduction The transport of angular momentum can be modeled by


assuming that there is viscous force per unit length acting
If the angular momentum of infalling material is not zero, on each ring in the disk given by
a planar, disk like structure will form (Fig. 12.1). This  
happens because the gravitational force exerted on the d
F = R , (12.4)
gas is mostly radial (when the central body is responsible dR
for most of the gravitational attraction, i.e., M?  Mdisk )
where is an effective viscosity. Then the viscous torque
and the angular momentum of the gas is, to a large extent,
around the whole circumference 2R is:
conserved. This leads to gas rotating faster at smaller
radii. When the timescale for the gas to cool down is G(R) = (2RF)R , (12.5)
short compared to the dynamical timescale, a flattened,
differentially rotating disk forms. or
d
There is plenty of evidence suggesting that matter in G(R) = 2R3 . (12.6)
these disks is able to lose angular momentum and spi- dR
ral inwards onto the central object releasing gravitational We can write an equation that relates the loss of an-
energy in the form of radiation that we observe. These ac- gular momentum with the torque exerted on a given disk
cretion disks can be found around young stars, compact annulus as follows. The specific angular momentum of a
stars in binary systems, and supermassive black holes. disk annulus at a distance R from the origin is given by
Understanding how accretion disks work has been a ma-
l(R) = R2 (R) , (12.7)
jor challenge in astrophysics for the last several decades.
In what follows we will consider thin accretion disks for whereas for an annulus at a distance R + R we obtain
which the vertical height is small compared to the disk
radius, i.e.. H  R. We will neglect the detailed vertical l(R + R) = (R + R)2 (R + R) . (12.8)
structure by working with vertically integrated variables,
such as (R), the disk surface density as a function of Thus the net angular momentum loss in for a given an-
radius. The accretion rate M = 2R(R)vr (R), where nulus in a disk where the accretion rate is given by M is
the radial infall speed is much smaller that the rotation (see Fig. 12.2):
speed vr  v .
M l + G = 0 , (12.9)
In thin disks, H  R, and the gas rotates in orbits
which are very close to Keplerian. This means that M [l(R + ) l(R)] + G = 0 , (12.10)
M [R2 + 2RR] + G = 0 . (12.11)
v (R) ' vK (R) = RK (R) . (12.1)
The change in the angular frequency can be found from
where the Keplerian orbital frequency is the expansion
r d
GM (R + R) ' (R) + R , (12.12)
K (R) = . (12.2) dR
R3
so that
and thus the specific angular momentum of the gas in the
disk decreases with decreasing radius. d
= (R + R) (R) R . (12.13)
1
dR
2
l v R = R K R 2 (12.3)

This implies that in order for matter in the disk to


accrete through successive Keplerian orbits the gas needs
to constantly lose angular momentum.

Figure 12.1: Thin accretion disk. Figure 12.2: Viscous angular momentum transport.
52

If we assume that each disk annulus emits like black-


body with a surface temperature Ts (R), we can relate the
dissipation rate to the temperature as

2Ts (R)4 D(R) , (12.20)

where is the Stefan-Boltzmann constant and the factor


of 2 accounts for the fact that the disk has two emitting
areas. This implies that
9 2
Ts (R)4 = , (12.21)
8
or
"   12 #! 14
3GM M R
Ts (R) = 1 . (12.22)
8R3 R
Figure 12.3: Thin accretion disk radiation.

12.4 Temperature at the Disk Midplane


Substituting this back into equation (12.11), we obtain
Recall that for a plane-parallel atmosphere we had
M (R2 ) = G , (12.14)
 
4 3 2
or T ( ) = + F (12.23)
d d

d
 4 3
M [R2 ] = v2R3 . (12.15)
dR dR dR where the optical depth is = /2 and F is the emer-
This expression can be integrated to give gent flux F = Ts4 .
If  1 we can approximate the temperature at the
1 M 1 disk midplane by
R 2 = R 2 + const . (12.16) 3
3 Tc4 = F . (12.24)
4
If we require that the torque vanishes at the stellar radius, Therefore the temperature at the disk midplane is related
i.e., G 0 for R = R we can fix the constant and obtain to the temperature at the disk surface via
"   12 # 3
M R Tc4 = Ts4 (12.25)
= 1 . (12.17) 4
3 R
and thus
which relates the accretion rate with the viscosity and the
   
4 3 9 2
disk surface density. Tc = (12.26)
4 2 8
which leads to
12.3 Energetics 27 2 2
Tc4 = . (12.27)
64
The viscous dissipation rate per unit area is Note that these estimates assume that all of the disk en-
 2    ergy is dissipated locally and within the disk.
d GM M R
D(R) = R =3 1 ,
dR 4R3 R
(12.18) 12.5 Nature of Disk Viscosity
where we have used equation (12.17). The corresponding
luminosity is Until now, we have not specified the nature of the viscos-
ity that enables disk accretion.
Z
GM M Let us examine the kinematic viscosity that arises due
L= D(R)2RdR . (12.19) to molecular scattering between the particles in the gas.
R 2R
A characteristic value for the viscosity can be obtained
This luminosity corresponds to half of the total potential by considering
gravitational energy lost from to R . In BH accretion, mol = cs (12.28)
the remaining half goes into the BH as kinetic energy. In
where
compact stars, the remaining half is released in a bound- 1 1
ary layer. = (12.29)
n n4b2c
53

is the mean free path for particles to interact, and the Note that the ratio of timescales associated with molec-
cross section characterizing Coulomb interactions, bc , can ular viscosity and its turbulent counterpart is very large
be obtained from
Ze2 3 tmol turb H
kT (12.30) 1010 (12.41)
bc 2 tturb mol mol
Considering typical values for the mean free path and This is in agreement with observations of binary systems
the sound speed where the systems are seen to undergo dramatic changes
 2
T
 in brightness in timescales of a week or so as opposed
4
6 10 cm (12.31) to millions of years as it would be implied if transport
n
processes where govern by molecular viscosity rather than
1 turbulent processes.
cs 104 T 2 cm s1 (12.32)
The model for the disk viscosity such that
the molecular viscosity can be written as
turb = cs H (12.42)
5
8T
2
2 1
mol 6 10 cm s
(12.33) is referred to as the "alpha" model and leads to what is
n
known as the standard model for accretion disks that has
Let us now estimate the Reynolds number characteriz- been widely adopted since it was introduce by Shakura
ing an accretion disk. Recall that and Sunyaev in the 70s.
LV Rv
Re . (12.34) 12.6 Viscous force per unit length and Stress
mol
Using that can be estimated as d
Z
F = R Tr dz HTr (12.43)
  12   12 dR
M R
Rv 1018 10
cm2 s1 (12.35) 1
Z
M 10 cm Tr = Tr dz (12.44)
H
we obtain for the Reynolds number
E
 12   12 [Tr ] = (12.45)
L3

Rv M R 25
Re 109 nT (12.36)
mol M 10
10 cm F E
[F] = = 2 (12.46)
L L
Using as characteristic values for the particle density,
temperature, and rarius
d
15 3
HTr = HR (12.47)
n 10 cm (12.37) dR
 
d ln
T 104 K (12.38) = H (12.48)
d ln R
we can see that the Reynolds number characterizing an d ln
= Hcs H (12.49)
astrophysical disk are huge d ln r
= Hc2s q (12.50)
Re 1014 (12.39)
= HP q (12.51)
In laboratory, flows become turbulent if Re > 103 .
When this happens the transport properties in the flow Tr = P q (12.52)
change dramatically. In particular, the effective viscos-
ity increases significantly. Turbulent astrophysical disks
can be modeled with an effective viscosity that has an 12.7 Time Scales in Accretion Disks
enhanced value due to turbulence
There are several relevant time scales in accretion disks
turb vturb lturb cs H (12.40)
H 1
thydrostatic (12.53)
where we have assumed that the turbulent velocity is of cs
the order of the sounds speed and the turbulent eddies
have a characteristic scale which is a fraction of the scale 2
R2 R2 R2

height of the disk. The parameter is a number that R 1 1
tviscous 2 
needs to be provided and characterizes the model for the cs H cs H
disk structure and its evolution. (12.54)
54

3
cs R 2
H= 1 (12.67)
E c2s c2s c2s cs 1 (GM ) 2
tthermal 2
2
2
, Pc
E D(R) H c2s = (12.68)
(12.55) c
where we have used that "   12 #
M R
 2   cs H = 1 (12.69)
d 2 d ln 9 3 R
D(R) R 2 .
dR d ln R 4 " 1 #
4T 4

(12.56) 3GM M R 2
= 1 (12.70)
Note that these timescales are related via 3 8R3 R
2 kT 4 4
Pc = c + T (12.71)
  
R R hyd
tv th (12.57) m p 3c
H H
Note that the dominant pressure support as well as
and thus tv  th hyd . the dominant source of opacity vary in the different disk
regions as follows
outer: Pg  Prot , f f dominates.
12.8 The structure of the Standard Disk Model
middle: Pg  Prot , c dominates.
For a thin disk, hydrostatic equilibrium in the vertical
inner: Pg  Prot , e dominates.
direction implies
If we know M , M , , we can calculate c , , H, cs , Pc
P GM z
= = 2 = 2K z (12.58) and T at position R.
z z r r
Thus the density gradient satisfies
12.9 The magnetorotational instability
 
K
= (12.59) Recall the equations for MHD (ideal). Continuity equa-
z c2s
tion
and thus

2K z 2
 + (v) = 0, (12.72)
(z) = c exp (12.60) t
2cs2 where we consider = const and thus v = 0 Momentum
It make sense to define the scale height of the disk H as equation
 
cs cs R v
+ (v )v = P + J B, (12.73)
H 2 = 2 (12.61) t
K vK
where we use = GM r . Induction equation
which implies that the disk thickness is related to the
ratio between the sound speed and the Keplerian speed B
= (v B), (12.74)
t
H cs
(12.62) where we use J = 1 B.
R vK 0

and thus cs  vK if H  R.
12.9.1 The shearing box approximation
We can estimate the pressure support in the radial co-
ordinate as
If we only care about a small piece of the disk, we could
1 P Pc c2 GM H 2 expand the equations of motion around r0 in a frame
s 2 2 (12.63)
R c R R R R rotating with V0 = /r0 . Provided that the departures
with respect to r0 are small; i.e. rx0 , ry0 , rz0  1.
and therefore
r P  r (12.64)
v B 2 (B )B
which implies that the rotation speed is ver close to Ke- + (v )v = 2 v + 2q2 x +
t 8 4
plerian (12.75)
v vK (12.65)
B
We can now collect all the relationships we have derived + (v )B = (B )v (12.76)
t
in a concise way as
These equations admit as a solution v0 (x) = qxy,
c = (12.66) B0 = Bz z.
H
55

This has 4 solutions, two waves, one exponentially de-


caying and one exponentially growing. Thus, MRI pro-
vides a mechanism to turn Keplerian disks unstable. The
induced turbulence leads to angular momentum transport
in accretion disks.

Figure 12.4: Shearing box approximation.

Figure 12.5: Magnetorotational Instability.

12.9.2 MRI dispersion relation

These solutions, however, are not stable. Lets consider


perturbations of the form

v = (vx , vy , 0)eikz+t (12.77)

B = (Bx , By , 0)eikz+t (12.78)


Substituting v = v0 (x) + v and B = B0 (x) + B, we
obtain

vx vA Bx
= 2vy + (12.79)
t 4 z

vy vA By
= (2 q)vx + (12.80)
t 4 z

Bx vx
= B0 (12.81)
t z
By vy
= qBx + B0 (12.82)
t z
This can be written in matrix form.


0 2 ikvA 0 vx vx
(q 2) 0 0 ikvA vy
= vy


ikvA 0 0 0 Bx Bx
0 ikvA q 0 By By
(12.83)
Av = v, det(A 1) 0.
This leads to a fourth degree dispersion relation for
(k):

[ 2 + (kvA )2 ]2 + 2(2 q)2 [ 2 + (kvA )2 ] 42 (kvA )2 = 0


(12.84)
56

References

Armitage, P. J. 2013, Astrophysics of Planet Formation


Melia, F. 2009, High-Energy Astrophysics

Padmanabhan, T. 2000, Theoretical Astrophysics - Vol-


ume 1, Astrophysical Processes
Rybicki, G. B., & Lightman, A. P. 1986, Radiative Pro-
cesses in Astrophysics

Shu, F. H. 1991, The physics of astrophysics. Volume 1:


Radiation.
Zeldovich, Y. B., & Raizer, Y. P. 1967, Physics of shock
waves and high-temperature hydrodynamic phenomena

You might also like