Astro 201 – Radiative Processes – Solution Set 1

by Roger O’Brient and Eugene Chiang

Readings: Rybicki & Lightman Chapter 1 except the last section 1.8.

Problem 1. Blackbody Flux

Derive the blackbody flux formula

F = σT 4 , (1)

from the Planck function for specific intensity,

2hν 3
Bν = . (2)
c2 (ehν/kT − 1)

dx
= π 4 /15.
R∞
You may use the integral 0 x5 (e1/x −1)

Those who know what they have to do can stop reading here. Those who need a bit of
help should read on.

Imagine a perfectly flat, blackbody patch at the center of a sphere of radius R. The patch
has tiny surface area dA and is lying flat at r = 0 in the θ = π/2 plane (in spherical
coordinates, where r, θ, φ are the radius, polar angle, and azimuth). Only one side of the
patch is warm; the other side is at absolute zero. Assume that the warm side radiates
as a perfect thermal blackbody.

A thermal blackbody emits radiation isotropically—i.e., with no preference for direction.

In other words, imagine each point on the patch beaming out a tiny, uniform, hemispher-
ical dome of rays (like an umbrella). It is a hemispherical dome and not a full sphere
because only one side of the patch is warm. Note that the Planck blackbody function has
no variable inside it that specifies direction of emission; the direction of emission doesn’t
matter for purely thermal emission. The only variables that need to be specified for the
Planck function are temperature and frequency (wavelength).

Despite the isotropic nature of the emission, because the patch is geometrically flat, has
definite area, and emits only on one side of itself, a detector glued to the inside of the
sphere at r = R detects different amounts of radiation depending on where it is placed.
For example, if the detector is glued on the half of the sphere that can’t see the bright side

1
of the patch, the detector detects nothing. If the detector is glued to the sphere directly
above the patch, the detector picks up the most number of photons, because it sees the
full face-on area of the patch (namely, dA). If the detector is glued at an angle to the
pole, then it sees less than area dA. If it is glued in the plane of the patch, it sees nothing
but a infinitesimally thin line segment.

This problem asks you to place detectors everywhere on the inside of the sphere and sum
up the radiation collected; this operation is equivalent to “integrating the Planck function
over all solid angles into which the radiation is beamed.”

Calculate the total luminosity (in units of energy/time) emitted by the patch of area dA.
You must integrate the Planck function over the entire emitting area (which is maximally
dA but in general will be less, depending on the viewing geometry), over all solid angles
into which the radiation is beamed, and over all frequencies. Think about placing tiny
detectors all over the inside of the sphere and asking how much energy/time each of the
detectors receives.

Then divide the luminosity by dA to calculate the total flux (in units of energy/time/area)
emitted by the patch. You should recover the usual blackbody flux formula, σT 4 . By
definition, σT 4 is the total amount of energy radiated per time per unit area of a blackbody
surface, radiated into all solid angles and over all frequencies.

To derive the flux (r&l 1.43) from the Planck Distribution (r&l 1.51), I noted that flux
has units of erg·s−1 ·cm−2 while the distributions has units of erg·s−1 ·cm−2 ·Hz −1 ·ster −1 .
So clearly I need to integrate out the frequency and angular dependence. Only catch is
that I also need to account for the detectors’ seeing the full blackbody patch directly
above the radiating patch (at θ = 0) but seeing an effective vanishing area (dA = 0) off
to the side (θ = π2 ). See fig.1 to understand this better. Hence I need to toss in an extra
factor of cos(θ) to get this right. And since one side is cold, I only integrate θ up to
(θ = π2 ).

Let Φ be the frequency integral of the Planck Distribution at a certain temperature.

Then:

R∞
Φ ≡ B (T )dν
 0 4 νR 3
2h kT ∞ x
(3)
= c2 h 0 ex −1 dx
4
2π k 4
= 4
15 c2 h3 T

erg
in units of s·cm2 ·ster . So the total Flux is:

R
F = Bν (T )dνcos(θ)dΩ
R 2π π
(4)
R
= Φ 0 dφ 2
0 dθcos(θ)sin(θ)
= πΦ

2
 erg
in units of s·cm2 . Plugging in Eqn 1:

2π 5 k4
F = σT 4 where σ ≡ (5)
15 c2 h3

Note that the above result (Eqn 3) is r&l 1.58.

Problem 2. Flat Disks

This problem forms the foundation for understanding the spectral energy distributions of
circumstellar disks, e.g., those surrounding pre-main-sequence stars and AGN.1

Consider a perfectly flat, blackbody disk encircling a blackbody star. The star has radius
R∗ and effective temperature T∗ . The disk begins at a stellocentric radius far from the
star, i.e. ri = R∗ /, where   1 is a small but finite constant. The disk extends to
infinity.

Calculate the temperature of the disk, T (r), as a function of radius, r. Make whatever
approximations you deem necessary in light of the fact that   1. Be sure you get the
scaling of T with r correctly. It is less important that you get the numerical coefficient
correctly.

This disk is PERFECTLY flat. Do not consider individual particles in the disk.

Also, neglect radial transport of energy. Each annulus is independent of neighboring

annuli. This is a fine approximation for thin disks.

Starlight strikes the flat disk not at normal incidence (as Rybicki & Lightman implic-
itly assume in their derivation leading up to r&l 1.13), but rather at grazing incidence.
A quick but crude way to solve this problem proceeds as follows. Imagine that you are
a patch of unit area lying in the flat disk. From your perspective, you see the upper
half face of the star. Idealize the upper half face of the star as a single point source
lying above the disk plane at height ∼R∗ . We should model this point source as having
a luminosity of order 1/2 the luminosity of the total star, because we can only see the
upper half face of the star (if the bottom of the disk were not there, we could see the
entire half face of the star).

Naively—and incorrectly—we would say that the flux of light from this source, eval-
uated at distance r, would be (L∗ /2)/4πr 2 , where L∗ is the total luminosity of the star.
But this would be incorrect, because what we really want is the flux of light passing
through the flat patch of the disk. This is the light that the disk actually absorbs. The
amount of energy crossing the flat patch per time equals (L∗ /2)/4πr 2 TIMES the sine
of the angle at which rays from the point source strike the patch. This angle is not 90
1
The problem has the further distinction that for some reason, whenever I teach this course, it is the
one problem that drives some students crazy and that they never forget.

3
degrees (normal incidence), but is rather equal to ∼R∗ /r  1. Therefore the true flux
passing through the patch equals

(L∗ /2) R∗
. (6)
4πr 2 r

Set this absorbed flux equal to the flux emitted by the patch:

(L∗ /2) R∗
= σT 4 . (7)
4πr 2 r

Since L∗ = 4πR∗2 σT∗4 , we have

 1/4 3/4
1 R∗

T = T∗ . (8)
2 r

Note that the index for the radial scaling is 3/4, not 1/2. The temperature decays
more steeply with distance for a flat disk than for, say, a system of planets because rays
strike the disk at grazing incidence, not at normal incidence.

We can re-do this problem more carefully by integrating over the upper half face
of the star rather than by idealizing the upper half face as a single point source. The
result of such an integration is to replace the (1/2)1/4 numerical coefficient above with
[2/(3π)]1/4 .

Problem 3. Albedos

Optical astronomers detect rocky bodies in the solar system by virtue of their reflected
sunlight. The received flux is proportional to aA, where a is the albedo (reflectivity) of
the object in some optical passband, and A is the area of the object. Typically, albedos
of minor bodies in planetary systems range from ∼0.01 to ∼0.7.

The proportionality of the flux to aA presumes that all of the optical light received from
an object is from reflected sunlight. Of course, that is not quite true; there is some
contamination at optical frequencies from the object’s thermal emission (the fact that it
is warm).

Give a quantitative explanation as to why this contamination is not worth troubling over.
If you wish, you can adopt as your “case study” a Kuiper belt object of optical albedo
0.07 and heliocentric distance 40 AU, observed in the visual (V) passband. Flesh out
the problem as you deem appropriate (“make it up as you go along.”) You may find
tables of the integrated Planck function, found in the old edition of Allen’s Astrophysical

4
Quantities, useful. Such tables might also be found in the new Allen, but I have not
checked.

Assume that the efficiency of emission at the (infrared) wavelengths at which most of the
object’s thermal power is radiated is one. Of course, its efficiency of emitting at optical
wavelengths is assumed to be less than one (by Kirchoff ’s law).

The KBO emits as much flux as it absorbs. Flux balance reads

L
(1 − a) = σT 4 (9)
4πr 2

where (1 − a) is the fraction of incident light absorbed by the KBO. (Here, unlike in
problem 2, we are taking sunlight to strike the KBO at normal incidence.) Plugging in
the numbers from the case study, we find that the temperature of the KBO is about

T ≈ 62 K . (10)

This little calculation ignores all real-world complications such as rotation of the body,
latitudinal variation of incident flux, finite thermal conductivity, etc., that professional
planetary scientists actually do trouble themselves over (sometimes for good reason, but
many times for no good reason, in my opinion). The thermal blackbody spectrum of the
KBO peaks at a wavelength of about λpeak ∼ hc/3kT ∼ 77 microns, or the far-infrared.
There is no need to worry about optical thermal emission from this cold blackbody; at
a visible wavelength of λ = 0.5 micron, the specific intensity from the KBO is

2hν 2hν
Iν,KBO−emission = Bν (T = 62K) = = , (11)
λ2 (ehν/kT − 1) λ2 (e465
− 1)

well on the Wien exponential tail of the blackbody. We would like to compare this ther-
mal specific intensity with the reflected specific intensity. The reflected light spectrum
of the KBO matches in shape, but not in magnitude, the spectrum of the sun, which
is a blackbody that peaks in the optical. In other words, the reflected spectrum of the
blackbody is a dilute blackbody; Iν,KBO−ref lect = f Bν (T ), where f  1. Use energy
conservation to find f . We know that the total reflected luminosity of the KBO must
equal

2
πRKBO
aL , (12)
4πr 2

in other words, the albedo-diluted power intercepted from the sun by the KBO, where
RKBO is the radius of the KBO. But we also know that the reflected light luminosity

5
can be obtained by integrating Iν,KBO−ref lect over all frequencies (to get f σT 4 ) and
then multiplying by the surface area of the KBO. This gives f σT 4 4πRKBO2 . Set this
4 2
expression equal to the one above, and use L = σT 4πR to find

R2
f =a ≈ 2 × 10−10 . (13)
4r 2

Therefore, the reflected specific intensity from the KBO equals

R2 2hν
Iν,KBO−ref lect = a (14)
4r λ (e4.8 − 1)
2 2

which is greater than Iν,KBO−emission by a factor of e460 f >> 1 despite the smallness of
f . So at λ = 0.5µm, there is no worry of contamination. Now the visible passband is
actual a broadband filter, so technically we need to integrate the power from ∼0.4µm to
∼0.6µm, but there seems to be little point because e460 is a mighty big number that’s
hard to drive down, even if we go to λ = 0.6µm. The moral of the story: the exponential
Wien tail really kills the emission at frequencies higher than the peak frequency.

Problem 4. Make like a tree and ...

If all the leaves of a tree fall to the ground, how thick is the layer of leaves on the ground?
Give your answer in leaf thicknesses, to order-of-magnitude.

What is the order-of-magnitude of the optical depth presented by the leafy tree to someone
lying below the tree? Why does this answer make sense from the perspective of the
vegetation? (Be the tree.)

The above two paragraphs are flip sides of the same question. You can try the second
paragraph before the first, or vice versa.

For simplicity, I’ll consider trees with leaves that bunch into vertical cylinders (see
attached fig. 3). Let’s say these trees have a height s = 10m. Each leaf should have
roughly the surface area of my driver’s license (σ ≈ 3in × 2in ≈ 5in2 ≈ 30cm2 ). Looking
outside, I guess that trees have between 10 and 100 leavesm3
. So I’ll split the difference
−5 leaves
and say that there are 5 × 10 cm3 .

This means that the optical depth (assuming constant density) is: τ = nσs =
(5 × 10−5 )(30)(1000) = 1.5.

This makes sense. A tree ought to absorb most of the light falling on it (i.e. τ ≈ 1).
If it absorbed any less, it could grow leaves near the base to absorb the extra light. And
if it absorbed more, then leaves near the base would die due to lack of light exposure.
In fact τ probably determines the density of leaves.

6
 As for the part about the depth of leaves on the ground, consider a 10 m tall square
column of leaves with a cross sectional area of 5 in2 that all fall off the tree and collect
directly on top of each other on the ground. With the same density as guessed above,
there should be: nAh = (50 leaves
m3
)(3 × 10−3 m2 )(10m) = 1.5 leaves.

So the pile is between 1 and 2 leaves thick. Gee-wiz! That’s the same as the first
part! The questions are mathematically identical!

7
 Astro 201 – Radiative Processes – Solution Set 2

by Jared Mehl and Eugene Chiang

Problem 1. Flat Disk SED

We return to the perfectly flat, blackbody disk encircling a blackbody star of problem set
1.

(a) Explain why νFν is a measure of the “flux radiated by an object per logarithmic
interval in frequency.” Here Fν is the flux density [units of energy per time per area per
frequency], and ν is the frequency of radiation. (Some refer more loosely to νFν as the
flux radiated per octave in frequency. An octave, in either the acoustic or electromagnetic
spectrum, represents a factor of 2 in frequency.)

Can you understand why νFν , and not Fν , is a quantity of interest to those who wish
to understand the overall energetics of an object? It is called the “broadband spectral
energy distribution,” or “broadband SED,” or “SED” for short.

Hint: Most macroscopic objects in the universe are broadband emitters; that is, they
radiate continuum radiation at all wavelengths. Plotting their spectrum would give a
curve that varies smoothly with frequency. For any object, you can ask whether it is
putting out its energy predominantly in X-rays, gamma rays, radio waves, infrared waves,
etc (is it an “X-ray object?” “a gamma-ray object?” “an infrared object?”) Now ask
yourself why a person asking such questions should plot νFν and not Fν .

An infinitesimal flux dF radiated into a logarithmic frequency interval d(ln ν) is given

by:
dF dF d ∞
Z 
SED ≡ = 1 =ν Fν dν = νFν
d(ln ν) ν dν dν 0
R∞
where we have used d(ln x) = dx/x and F ≡ 0 Fν dν.
If we are interested in the
energetics of some radiating object, the SED would be the quantity of interest, because
the specific flux Fν has units:
erg
[Fν ] ∼
Hz cm2 s
which, since it is divided by frequency (proportional to energy for photons), gives a
number density of photons. To recover the energy within a given frequency band, we
need to multiply by frequency (which is proportional to energy).

In other words, νFν = dF/d ln ν; it is a measure of the total amount of energy per time
per area (dF ) over a logarithmic interval of frequency (d ln ν). It is a crude integral of
Fν over a logarithmic interval in frequency (from ν to e × ν). Whatever frequency νFν
peaks for a broadband emitter, that is the frequency where most of the energy is being
emitted. If νFν peaks in the infrared, then we say the object is emitting most of its
energy at infrared wavelengths.

1
(b) Is νFν equal to λFλ , where λ is the wavelength of the radiation, and Fλ is per
wavelength rather than per frequency?

Comparing νFν to λFλ , using λ = c/ν:

dF dF c dF c dF dF
λFλ ≡ =λ = = = −ν = −νFν
d(ln λ) dλ ν d(c/ν) ν (− νc2 dν) dν

so formally they differ by a minus sign. The minus sign arises because going up in
wavelength means going down in frequency. We forget about this minus sign because
we know where we’re going, and energy in radiation is always positive.

(c) Write down an expression for νFν for the blackbody disk. That is, write down the
formula for the spectrum of the disk as measured by an observer for whom the disk is a
point source.

Recall that the disk has a temperature T (r) at every stellocentric radius, r. The inner
radius of the disk is ri and the outer radius is ro . The disk is a distance D away from
the observer, and is inclined by an angle i (i = 0 corresponds to a face-on disk). Your
expression should take the form of an integral.

Recall that the disk has temperature T (r) for each stellocentric radius r given by:
1 3
2 R∗
 
4 4
T (r) = T∗
3π r
where R∗ and T∗ are the radius and temperature of the blackbody star at the center of
the disk. To get the SED of the disk, we only have to add up the blackbody flux seen
by a distant observer for each infinitesimal annulus of the disk at temperature T (r) and
multiply by the frequency ν. Now the flux seen by a distant observer (to whom the disk
appears to be a point source) will be given by flux ∼ specific intensity × solid angle. As
seen by the distant observer at distance D, each annulus, which is oriented at an angle
i with respect to the observer’s line of sight, presents a solid angle of:
dA 2πrdr cos i
dΩ = = .
D2 D2
Applying equation (1.3b) of Rybicki and Lightman, we get:

Z
νFν = ν Bν (T (r)) cos θ dΩ (1)
Z ro 2πr
= ν Bν (T (r)) cos i dr (2)
ri D2
4πhν 4 cos i ro rdr
Z
= (3)
c2 D2
 1/4  3/4 
ri 3π hν r
exp 2 kT∗ R∗ −1

2
Here we have assumed cos θ ∼ 1 because the disk is a point source to the observer.

(d) Sketch (no heroics necessary) νFν vs. ν. If the spectrum exhibits power-law behavior,
give the slope of the power law (i.e., d ln(νFν )/d ln ν). Overlay on your sketch the SED
of the central stellar blackbody. Log-log space is best.

The accompanying .ps or .pdf figure shows the disk’s contribution to the total SED.
To make this plot, I assumed i = 0, T∗ = 4000 K, R∗ = 2.5R , ri = 6R∗ , and ro =
2.3 × 104 R∗ . These are typical T Tauri star parameters.

Indeed the disk SED does fall off like a power-law, but much less steeply than the
Rayleigh-Jeans tail of the stellar blackbody. We say that the disk is responsible for an
“infrared excess” compared to the stellar photosphere. We can derive the index to the
power law of the disk SED by making a change of variable to our integral in part (c).
Change to the new variable x ≡ (hν/kT∗ )(3π/2)1/4 (r/R∗ )3/4 . Then the denominator in
the integral is just ex − 1. The numerator rdr = constant × ν −8/3 x5/3 dx. The constant
R xo 5/3
×ν −8/3 can be taken outside of the integral. The integral becomes xi [x /(exp(x) −
1)]dx. Now xi and xo both depend on ν, but provided we are interested in frequencies
far from those characterizing the inner and outer radius, we can take xi ≈ 0 and xo ≈ ∞.
Then the integral is just some dimensionless number that we could look up in an integral
table if we wanted to. Therefore, νFν ∝ ν 4 × ν −8/3 ∝ ν 4/3 . The index is therefore 4/3—
plenty less steep than the Rayleigh-Jeans tail of the stellar blackbody, which gives an
index of 3.

The lecture in class gives an alternative, more physical way of deriving the index of
4/3.

Finally, at the very lowest frequencies, the disk will exhibit Rayleigh-Jeans behavior
like any good blackbody: νFν ∝ ν 3 .

Problem 2. Practice with jν , αν , Sν , Bν , Iν

(a) A plane-parallel slab of uniformly dense gas is known to be in LTE (local thermody-
namic equilibrium) at a uniform temperature T. Its thickness normal to its surface is s.
Its absorption coefficient is αν,gas . Write down the specific intensity, Iν , viewed normal
to the slab, in terms of the variables given.

A slab having uniform source function Sν and total normal optical depth τν has a
specific intensity normal to its surface of

Iν = Sν (1 − e−τν ) (4)

This is a result worth remembering. Now τν = αν,gas s. Furthermore, LTE tells us

that Sν = Bν . Then

3
 Iν = Bν (T )(1 − e−αν,gas s ) (5)

(b) The same slab is now filled uniformly with non-emissive dust having absorption
coefficient αν,dust . The dust is non-emissive, so its emissivity jν,dust = 0. Write down
Iν viewed normal to the slab, in terms of all variables given so far.

The source function now changes. The source function is the TOTAL emissivity of
the gas divided by its TOTAL absorption coefficient. The total emissivity is still given
by the gas in part (a): jν = αν,gas Bν . But the total absorption coefficient is now greater
because of the extra dust: αν = αν,gas + αν,dust . Therefore Sν = αν,gas Bν /(αν,gas +
αν,dust ). Returning to the equation (4) worth remembering, we find

αν,gas Bν
Iν = (1 − e−(αν,gas +αν,dust )s ) (6)
(αν,gas + αν,dust )

(c) The slab of gas and dust is further mixed with a third component: an emissive, non-
absorptive uniform medium having emissivity jν,med and absorption coefficient αν,med =
0. Write down Iν viewed normal to the slab, in terms of all variables given.1

This is the same as part (b) except that now the total emissivity is now greater:
jν = αν,gas Bν + jν,med . Then

(αν,gas Bν + jν,med )
Iν = (1 − e−(αν,gas +αν,dust )s ) (7)
(αν,gas + αν,dust )

Problem 3. Photon Pinball in Clouds

This problem shows how the usual e−τ attenuation factor for flux passing through an
absorbing medium can be incorrect for flux passing through a purely scattering medium.
(And thank goodness, or else it would be really dark on cloudy days.) Clouds can be
viewed as a purely scattering medium.

Idealize the cloud as a uniform, 1-dimensional slab comprising particles that can only
scatter light. A uniform flux of photons irradiates the top of the cloud deck (that is, just
above the cloud deck, there exists a uniformly bright sheet). A photon passing through
the cloud gets bounced like a pinball from cloud droplet to cloud droplet, preserving its
frequency and never getting absorbed by any droplet. A few photons are lucky enough to
1
A physical realization of this problem might be an HII region surrounding an ionizing O star. The
material in LTE would be the fully ionized plasma, emitting thermal bremsstrahlung radiation. The
dust would be dust. The emissive, non-absorptive medium would be the same ionized plasma emitting
recombination (line) radiation. For the assumptions stated in the problem to be valid, we would have
to evaluate ν at, say, an optical recombination line like Hα.

4
make it through the cloud, while most get pinballed back out the way they came. We will
calculate the fraction that make it through.

Take the cloud to have a droplet density [droplets per cubic volume] η, the droplet radius
to be R, and the vertical thickness of the cloud to be zmax . Measure vertical distance
through the cloud by z, where the top of the cloud is located at z = 0 and the base of the
cloud is located at z = zmax .

(a) Write down the optical depth of the cloud, τ .

Optical depth is given by the formula τ = (density) × (cross-section) × (thickness),

which, assuming the droplets have a geometric cross-section, gives:
τ (z) = ηπR2 z.

(b) Incident photons from the sun strike the top of the cloud. The photons have a
number flux, Fi [number per time per area]. What is the number density of incident
photons at the top of the cloud? Call this photon number density ni . These incident
photons have NOT been scattered yet by any droplet. Remember that flux is a number
density multiplied by a speed.

In a time ∆t the photons will have moved a distance c∆t, so the volume they occupy
during that time is the area A they are passing through times this distance. The incident
flux is given by Fi = #/A∆t. The incident number density is therefore given by:
# # Fi
ni = = = .
vol Ac∆t c

(c) These photons pinball/random walk through the scattering droplets. Random walks
are described by the diffusion equation,

∂n ∂2n
=D 2 (8)
∂t ∂z

where D is the diffusion coefficient and n is the photon number density.

Express D in terms of symbols defined above and whatever fundamental constants you
deem appropriate. Hint: dimensional analysis may prove useful.
2
By dimensional analysis, the diffusion coefficient must have units [D] = lengthtime . The
relevant quantities for diffusion that give these units are the particle velocity [c] = length
time
and the mean free path [λ] = length. Digging through Rybicki and Lightman (1.31) we
find that the mean free path is λ = 1/nσ, so using what we know about the scattering
cross-section from (a) we find:
c
D = cλ = .
ηπR2

5
 This diffusivity is microscopic quantity. Therefore we cannot the macroscopic length-
scale zmax ; we must use the microscopic lengthscale λ.

(d) In steady-state, ∂n/∂t = 0 (the number density of photons everywhere in the cloud
does not change with time). Write down the solution to the diffusion equation for n(z).
You should have two, as yet unknown, constants of integration.

The general solution to ∂ 2 n/∂z 2 = 0 is n = Az + B , where A and B are constants.

(e) To solve for the two constants of integration, you need two boundary conditions. The
first condition is that nt = n(z = zmax ). Here nt is the number density of photons at
the base of the cloud. These photons comprise the transmitted flux.

The second condition is that the (net, number) flux, F , of photons at z = 0 equals the
incident flux, Fi (directed down into the cloud) MINUS the outgoing, reflected flux, Fr
(directed up, away from the cloud into space). Recall Fick’s law, which is just another
way of writing the diffusion equation, that the (net) flux F = −D∂n/∂z. So we have
F (z = 0) = Fi − Fr .

Use the above, and the fact that the incident flux, Fi , must equal the reflected flux, Fr ,
PLUS the transmitted flux, Ft , to calculate T , the ratio of the transmitted flux to the
incident flux, in terms of τ . Are you glad that clouds scatter but do not absorb light?

At z = 0, there are TWO contributions to the local photon number density. The
first contribution is from incident photons, ni . The second contribution is from photons
reflected out of the cloud, nr . Therefore, at z = 0,

n = ni + nr = B (9)

Now Fi = Fr + Ft means that ni = nr + nt (just divide by c). Use the latter equation
to write nr = ni − nt and substitute into (9) to solve for

B = 2ni − nt (10)

At z = zmax , n = nt = Azmax + B. Use this equation and equation (10) to solve for

2(nt − ni )
A= <0 (11)
zmax

Now, the net flux F = −D∂n/∂z = −DA. Notice the net flux is constant with height, as
it must be lest the photon density increase or decrease somewhere in the cloud, violating
our steady-state assumption. At z = zmax , F = Ft = −DA. Then nt = −DA/c. Note

6
that since A < 0, nt > 0 which is correct (you cannot have a negative number density).
Thus,

T ≡ nt /ni (12)
DA
= − (13)
cni
2D(nt − ni )
= − (14)
zmax cni
2(T − 1)
= − (15)
τ

Finally, solve for T in terms of τ :

2
T = 2+τ (16)

which is certaintly not exponentially sensitive to optical depth. And thank goodness, or
else it would be plenty darker on the Earth on cloudy days. Note our expression has the
right limits: the transmission goes to 1 if τ = 0, and goes to zero (linearly) as τ goes to
infinity.

This pure scattering transmission coefficient T gives much more light passing through
the cloud than a purely absorbing cloud (attenuation e−τ ) of the same optical depth.
For example, with τ = 1, we have

T = 0.67 (17)
−τ
e = 0.37 (18)

For τ = 2 we have

T = 0.50 (19)
−τ
e = 0.14 (20)

Absorbing clouds would make the sky much darker than scattering clouds with the
same droplet parameters.

Problem 4. Galaxies

7
In an idealized model of a galaxy, stars are uniformly distributed within a cylinder of
radius R and height H. The number density of stars is n [in units of stars per volume].
Model each star as a spherical blackbody of temperature T∗ and radius R∗ . The density of
stars is so low that of all possible lines of sight running through the galaxy, the fraction
that intersect a star is  1.

An observer resolves the galaxy in a face-on viewing geometry, but does not resolve the
individual stars making up the galaxy. In other words, the observer detects the galaxy as
a circular disk that is very nearly uniformly bright from center to edge.

Write down the specific intensity the observer measures. Neglect terms of order H/d,
where d is the distance to the galaxy.

The answer is Iν = Bν (T∗ )nπR∗2 H. An easy way to see this is note that the specific
intensity of a single star is Bν , but that we have to dilute this intensity by the fraction
of area covered by stars, which is nπR∗2 H (which is τ ).

But we can also answer this question in a formal way, which is worthwhile for those
who like to test their grasp of formalisms. The galaxy is an example of our proverbial
uniform slab. We know that for uniform slabs, Iν = Sν × (1 − e−τν ). For optically thin
slabs (which galaxies are), τν  1 and so Iν ≈ Sν τν . But we also know τν = nσH =
nπR∗2 H, so Iν ≈ Sν nπR∗2 H.

It remains to show that Sν = Bν . Well by definition, Sν = jν /αν . The extinction

coefficient αν = nσ = nπR∗2 . The emissivity jν = nLν /4π, where the 4π accounts for
the per solid angle in jν , and Lν is the “luminosity density of a star”, defined such that
Lν dν = L∗ = 4πR∗2 σSB T∗4 . Now it must be that Lν = CBν where C is some constant,
R

since all the frequency dependence of Lν must be captured by Bν . So we solve for the
normalization constant:

σSB T∗4
Z Z
Lν dν = C Bν dν = C = 4πR∗2 T∗4
π

which implies that C = 4π 2 R∗2 . So jν = nBν πR∗2 , which means that yes, Sν = Bν .

8
 Astro 201 – Radiative Processes – Solution Set 3

by Katie Peek and Eugene Chiang

Problem 1. The Eddington Limit (Rybicki & Lightman 1.4)

(a) Show that the condition that an optically thin cloud of material can be ejected by
radiation pressure from a nearby luminous object is that the mass to luminosity ratio
(M/L) for the object be less than κ/(4πGc), where G = gravitational constant, c =
speed of light, κ = mass absorption coefficient of the cloud material (assumed indepen-
dent of frequency).

Force of radiation must exceed the force of gravity:

frad > fG , (1)

where

κ L
frad = (2)
c 4πR2

and

GM
fG = . (3)
R2

Note that L/c is a “momentum luminosity” (units of momentum per time). The factor
of κ is a cross-section for absorbing this momentum (energy), per unit mass. Thus, the
condition for the cloud’s ejection is given by

κ L GM
2
> 2
c 4πR R
κ M
> . (4)
4πGc L

(b) Calculate the terminal velocity v attained by such a cloud under radiation and grav-
itational forces alone, if it starts from rest a distance R from the object. Show that
2GM κL
 
v2 = −1 .
R 4πGM c

1
 Initially at rest at position R, the cloud is simultaneously pulled inward under grav-
itation and pushed outward by radiation. If the cloud begins moving outward, the force
due to radiation exceeds the force due to graviation. Hence, the gravitation serves to
“soften” the radiation force outward.

κL GM
fnet = frad − fG = 2
− 2 .
4πcR R

Let r = the non-initial position (R at some time t). Then the total, potential, and
kinetic energies are given by the expressions
κL GM
Etot = − (5)
4πcR R
κL GM
U= − (6)
4πcr r
1
T = v2 . (7)
2

At terminal velocity, fnet = 0, so

κL GM κL
− 2 = 0 ⇒ GM = ,
4πcr 2 r 4πc

therefore (softened) U = 0 for terminal velocity vT .

Since Etot = T − U ,

T = Etot − U (8)
1 2 κL GM
 
vT = − −0
2 4πcR R
1 2 κL GM
v = −
2 T 4πcR R
2GM κL
 
2
vT = −1 . (9)
R 4πcGM

(c) The mimimum value for κ may be estimated for pure hydrogen as that due to Thom-
son scattering off free electrons, when the hydrogen is completely ionized. The Thomson
cross section is σT = 6.65 × 10−25 cm2 . The mass scattering coefficient is therefore
> σT /mH , where mH = mass of hydrogen atom. Show that the maximum luminosity

2
that a central mass M can have and still not spontaneously eject hydrogen by radiation
pressure is
LEDD = 4πGM cmH /σT
= 1.25 × 1038 ergs−1 (M/M ),
where
M ≡ mass of sun = 2 × 1033 g.
This is called the Eddington limit.

Regardless of whether the radiation is being scattered or absorbed, the limiting case
for ejection of material occurs when
M σT /mH
= , (10)
L 4πcG
as can be seen by setting the term in parentheses in (b) equal to zero. In this limit,
L → LEDD . Thus,
4πcGM mH
LEDD = . (11)
σT

Plugging in numbers for everything,

4π(3 × 1010 cms−1 )(6.673 × 10−8 cm3 g−1 s−2 )(1.674 × 10−24 g)M
LEDD =
6.65 × 10−25 cm2
LEDD = 6.33 × 104 M s−3 cm2 ,
or, in terms of solar masses,

LEDD = 1.27 × 1038 ergs−1 · M/M .

Most objects in the universe, even the high-energy ones like accreting X-ray binaries and
AGN, are radiating at sub-Eddington rates. I believe there are a few cases where compact
objects are puzzlingly radiating above the Eddington limit; it has been conjectured that
such objects are accreting through a thin disk, and radiating above and below the disk.
That is, the inflow of matter and the outflow of radiation occur at different locations,
so there is less danger of the radiation preventing the inflow.

Problem 2. Pulsar Dispersion Measure

Don Backer notes that pulses from a certain pulsar observed at a radio frequency of
ν = 2 GHz arrive slightly ahead of the pulse train observed at ν = 1 GHz. The lead time
is 1 s.

(a) Use the dispersion relation derived in class for a cold, ionized plasma, and the in-
formation above, to derive the column density of electrons along the line of sight to this
pulsar. Express in standard pulsar-community units of cm−3 pc.

3
 The dispersion relation (as derived in class) is given by

2
ck 4πne2

=1+ .
ω me (ωo2 − ω 2 )

For a plasma, ωo2 = 0, so the dispersion relation becomes

ωp2
2
ck

=1− 2 (12)
ω ω

where

4πne2
ωp2 = .
me

Thus, the dispersion equation can be rewritten as

q
ω= ωp2 + k2 c2 . (13)

dω
Since the group velocity vgroup = dk ,

dq 2
vgroup = ωp + k2 c2
dk
c2 k
vgroup = . (14)
ω
q
Since the wave number k = c−1 ω 2 − ωp2 , vgroup can be rewritten as

s
ωp2
vgroup = c 1 − (15)
ω2

Enter the pulsar. The group velocity of the 2 GHz signal is such that it arrives 1 second
before the 1 GHz signal. Henceforth, I shall use ω1 for the 2-GHz wave and ω2 for the
1-GHz wave.

ωp
2 −1/2
 
If it can be assumed that ωp  ω, then the expression 1 − ω can be approxi-
ωp
2
mated as 1 + 12 ω . If s is the distance between Don and the pulsar, the time between
signals can be expressed as

4
 s s
∆t = t2 − t1 = − ,
vgroup,2 vgroup,1

which becomes

" 2 # " 2 #
s 1 ωp s 1 ωp
 
∆t = 1+ − 1+
c 2 ω2 c 2 ω1
2 2
2c ωp ωp
   
∆t = − . (16)
s ω2 ω1

Since ω2 = 21 ω1 ,

2
2c ωp
  
∆t = . (17)
3s ω1

Replacing the expression for ωp2 ,

!
4πne2 8π 2 2
s= cν1 ∆t. (18)
me 3

Cancelling convenient things like factors of π and plugging in ∆t = 1 s (with standard

values for me , etc.), one ends up with a column density of

n · s = 3 × 102 cm−3 pc. (19)

(b) For an assumed density of electrons in the interstellar medium of 0.03 cm−3 , calculate
the distance to this pulsar. Does this seem reasonable?

The distance to the pulsar if n = 0.03 cm−3 , is

 
0.03 cm−3 s = 3 × 102 cm−3 pc

s = 1 × 104 pc = 10 kpc
which is reasonable because it resides inside our Galaxy. I don’t believe our Earth-bound
detectors are sufficiently sensitive to detect extragalactic pulsars.

(c) For such an assumed electron density, is Don safely observing above the plasma
cut-off frequency?

5
 The plasma cutoff frequency is given by

4πne2 4π(0.03cm −3 )(4.8 × 10−10 esu)2

ωp2 = =
me 9.11 × 10−28 g

ωp = 104 s−1 .

And ω1 is

ω1 = 2πν1 = 4π GHz = 1010 s−1 .

Thus

ωp
= 10−6 ,
ω

plenty small enough for Don to have observed the pulsar in the first place.

(d) Calculate the optical depth to Thomson scattering along the line-of-sight to this
pulsar.

The optical depth due to Thomson scattering is given by

τ = σT (n · s) (20)
τT = (6 × 10−25 cm2 )(3 × 102 cm−3 pc).

Taking 1 pc = 3 × 1018 cm,

τT = 0.0005.

Thomson scattering attenuates the signal very little!

Problem 3. Hyperfine 3 He+

Observations of the hyperfine transition in 3 He+ are used to probe the 3 He/H abundance
in the galaxy. This abundance reflects the primordial yield from big bang nucleosynthesis
and galactic chemical evolution.

6
(a) Estimate, using the scaling relations presented in class and whatever facts you re-
member, the wavelength of the ground-state hyperfine transition in 3 He+ . Compare to
the true answer of 3.46 cm.

This is a magnetic dipole transition. Energies of magnetic dipole transitions scale as

E ∼ µe B, where µe is the magnetic moment of the electron and B is the magnetic field
strength experienced by the electron.

Let’s scale from the more well-known 21-cm transition in hydrogen. Now µe , the
moment intrinsic to the electron, does not change in going from H to He. But B does
change. For a dipole nuclear field, B ∼ µnucleus/r 3 , where µnucleus is the magnetic
moment of the nucleus, and r is the distance between the nucleus and the electron.
Both these quantities change in going from H to He.

Now the magnetic moment of a single proton is eh̄/(2mp c). We infer that the mag-
netic moment of a nucleus of charge Ze and mass Amp is µnucleus = Zeh̄/(2Amp c). The
Bohr model tells us that r ∼ 1/Z. Therefore B ∝ (Z/A) × Z 3 . For our 3 He nucleus,
Z = 2 and A = 3. Therefore the wavelength of our transition is 21 cm × A/Z 4 ∼ 3.9 cm,
which is pretty close to the laboratory-measured answer of 3.46 cm.

(b) Estimate the Einstein A coefficient (transition probability) of this line. Compare to
the true answer of 1.95 × 10−12 s−1 .

Scale again from the 21-cm transition in H, for which we know A21 ≈ 2.9 × 10−15 s−1 .
We know that Einstein A’s scale as some moment squared and frequency cubed. Now
the moment appropriate to this problem is the magnetic moment of the electron which
does not change in going from H to He. The frequency certainly does change by a factor
of (21/3.9). Then the Einstein A for this transition is about 2.9×10−15 ×(21/3.9)3 s−1 ∼
5 × 10−13 s−1 , which is a factor of 4 too low compared to the true answer. Had we used
the true wavelength of 3.46 rather than our above estimate of 3.9 cm, we would be off
by a factor of 3.

Here’s an alternative scaling from the Ly α electric dipole transition in H that just
happens to get the answer almost exactly correct:

!3
1216Å
A21 = A21 α2 ·
∆Ehf ,3 He+ Lyα 3.46 cm

A21 = 1.3 × 10−12 s−1 .

∆Ehf ,3 He+

Problem 4. Shades of the Sun

(a) When the sun sits 5 degrees above the horizon and is about to set, it looks red. Give
a quantitative explanation why. Neglect dust.

7
 Consider the Sun overhead, where the distance it passes through the atmosphere is
given by H. The distance it passes through the atmosphere at the horizon (assuming
that the horizon is an flat infinite slab) is H/ sin θ, where θ is the angle above the horizon.

Sunlight experiences Rayleigh scattering as it tries to propagate through the atmo-

sphere. The cross-section for Rayleigh scattering is

ω4
σscattering = σT homson .
ωo4

Compare the two extremes of the visible spectrum (taking λred = 7000Å and λblue =
4000Å):

4
σscat (red) σ · λ4 /λ4 λb

= T 4o 4r = = 0.11.
σscat (blue) σT · λo /λb λr

What is the absolute optical depth, say at blue wavelengths? At zenith, the column
of (mostly nitrogen) molecules is N ∼ 5 × 1019 cm−3 × 10 km ∼ 5 × 1025 cm−2 . The
cross-section for Rayleigh scattering at blue wavelengths is σT homson ω 4 /ω04 , where ω0 ≈
10 eV/h̄. The 10 eV is the typical energy for an electronic transition in molecular
nitrogen. Putting it all together, we get τb ∼ 0.2. Correct this by 1/ sin 5◦ to get
τb ∼ 2.3 towards the sun at sunset. Thus, e−2.3 ∼ 10% of the blue light reaches us.

Some of you wished to use the scattering law proved in the last problem set, but
the geometry of this problem is different; for this problem, light that is scattered out
of the incident beam remains out of the beam (3-D geometry), whereas in the previous
problem set on clouds, light has no choice but to be scattered parallel or anti-parallel
to the incident beam (1-D geometry). Our assumption here that light that is scattered
out of the direction of a particular beam remains out of the beam is a good one because
otherwise the disc of the sun would appear larger than just the geometric value, and we
know this is not the case.1 In other words, multiple scattering is not important for this
problem; the extinction in this problem is still exponential; energy losses suffered by a
given beam due to scattering are not regained by scattering from other directions.

By the ratio of cross-sections, τr ∼ 0.02 towards the sun at sunset. Thus, about 98%
of the red light reaches us.

And that’s why sunsets are red: red and blue light are in the rough ratio of 10:1.
Actually they’d be more orange, if weren’t for all the particulates spewed by humanity
that reddens the light even further.
1
That the sun and moon appear larger near the horizon than at zenith is, I believe, an optical illusion
caused by the proximity of these celestial objects to terrestrial ones near the horizon.

8
(b) During a total solar eclipse, the sun’s corona appears white to the naked eye. The
corona is composed of plasma. State why the corona looks white (one sentence will
suffice).

In a plasma, ωo = 0. So σscattering = σT homson . Since σscat no longer has any

wavelength dependence, all the light from the sun’s photosphere is scattered evenly by
the corona into our line-of-sight. Hence it appears “white.”2

Problem 5. Back-of-the-envelope CO 1-0

Most astronomers know that the Einstein A coefficient for the Lyman alpha (n = 2 to
n = 1) transition in atomic hydrogen is of order 109 s−1 (actually, 5× 108 s−1 , but what’s
a factor of 2 between friends?). We derived this result in class to order-of-magnitude
by considering an (accelerating) electron on a spring that displaces a Bohr radius, a0 ,
and has natural angular frequency ω. We took the inverse lifetime of the excited state
as A ∼ P/h̄ω, where P is the power radiated by an accelerating charge.

The ubiquitous carbon monoxide molecule, 12 CO, is used by astronomers to trace the
presence and temperature of molecular gas in everything from galaxies to circumstellar
disks. We would rather try to detect H2 , but sadly H2 has no permanent dipole moment
because of its symmetry. Carbon monoxide does have a permanent dipole moment.

(a) Estimate the wavelength of the lowest energy, rotational transition in CO (J =

1 to J = 0, where J is the rotational quantum number). Do this by considering a
barbell spinning about its axis of greatest moment of inertia and recognizing that angular
momentum comes quantized in units of h̄.

Compare your estimate to the true answer of 2.6 mm.

We know from quantum mechanics that angular momentum is quantized as L2 =

h̄2 J(J + 1), and from classical mechanics we know the rotational kinetic energy of a
barbell to be E = L2 /2I. Here, the moment of inertia I can be approximated as
I = i mi ri2 ∼ 30mp a20 , where mp is the mass of a proton and we assume that the
P

width of the barbell is roughly two times the Bohr radius. For transition J = 1 to J = 0
we find:
h̄2 hc
∆E1→0 ∼ 2 ∼
30mp a0 λCO
which upon plugging in the appropriate values of the constants gives:
λCO ∼ 2.6mm
which just happens to be exactly the same as the real value. Hooray for us!

(b) Use the scalings of our semi-classical spring model to estimate the Einstein A coef-
ficient of this transition, i.e., the inverse lifetime of the excited J = 1 state.
2
Actually, by this reasoning, it should appear slightly yellow, since yellow is the color of the sun’s
photosphere. Whether the corona is yellow or white I have yet to see with my own eyes.

9
Bring to bear one fact that is difficult to guess from first principles: the dipole moment of
the CO molecule is 0.1 Debyes (1 Debye = 10−18 cgs. Note that ea0 = 2.5 Debyes, where
e is the electron charge), not ∼1 Debye, as one might have guessed naively. The smaller-
than-usual dipole moment of CO is a consequence of the strong double bond connecting C
to O. Most other molecules—e.g., H2 O, CS, SiS, SiO, HCN, OCS, HC3 N—have dipole
moments that are all of order 1 Debye.

Compare your estimate to the true answer of 6 × 10−8 s−1 .

We can estimate the Einstein A coefficient using the approximate equation derived
in class:
2d2 ω03
A21 ∼ .
3h̄c3
Note we can just plug in our value of λCO = 2πc/ω0 and standard constants for every-
thing except for the dipole moment, which has a difficult to guess value of 0.1 Debye
∼ 10−19 [cgs]. Plugging in these values gives:
A21 ∼ 8.7 × 10−8 s−1
which is pretty close to the actual value of 6 × 10−8 s−1 .

(c) Estimate σ, the cross-section of the transition at line center, in cm2 . Assume the
line to be thermally broadened at a typical molecular cloud temperature of 20 K.

To estimate the cross-section at line center of this transition, we apply the equation
(again from lectures):
λ2 A21
σlc ∼ .
8π ∆ν
Here ∆ν is the broadening of the line. For our particular case, the broadening mechanism
is thermal broadening, which gives ∆ν ∼ ν vthermalc . The thermal velocity is given by
q
vthermal = 3kT
m . Using T = 20K and m ∼ 30mp we find that the frequency broadening
is ∆ν ∼ 5 × 104 s−1 . Plugging this into the cross-section equation gives
σlc ∼ 5 × 10−15 cm2

(d) If the number abundance of CO molecules to H2 molecules is nCO /nH2 ∼ 10−4 (i.e.,
close to the solar abundance ratio of nC /nH ∼ 10−4 ), estimate the column of hydrogen
molecules required to produce optical depth unity at the center of the CO line. Is this
column likely to be exceeded in a molecular cloud? Take a typical molecular cloud H2
density of 104 cm−3 and a cloud dimension of 100 pc.3

Assuming nH2 ∼ 104 cm−3 in molecular clouds (where we expect to see CO), and
zero everywhere else, we use nCO /nH2 ∼ 10−4 to find a rough number density of CO
3
In Spitzer’s book on the interstellar medium, molecular cloud densities of molecular hydrogen are
said to range from 103 to 106 cm−3 . The density can be orders of magnitude higher than even 106 cm−3
as you approach star forming regions. Moreover, cloud dimensions can range from 1 pc to 100 pc.
Molecular clouds are clumpy structures.

10
in molecular clouds of 1 CO molecule per cubic centimeter. Using this number density
and the cross-section from part (c), we expect to see an optical depth of unity when the
column of CO molecules has a height s given by:

τ = 1 = nCO σlc s which gives s ∼ 2 × 1014 cm ∼ 7 × 10−5 pc.

Compared with the rough size of a molecular cloud of 100 pc, we expect this column to
be exceeded in our galaxy whenever the line of sight points at a molecular cloud.

Moral: Molecular clouds in the lowest-order (J = 1 to J = 0) rotational transition

of the most common isotopomer of carbon monoxide, 12 CO, are likely to be extremely
optically thick. How do astronomers get around this problem to measure the mass of
a molecular cloud? Some try using less common isotopomers of CO, such as 13 CO
(the terrestrial abundance of 13 C to 12 C is about 1/89) or C18 O (the terrestrial isotopic
ratio is 18 O/16 O ∼ 1/500). This helps, but not for the most massive of clouds that are
considered here.

Others try looking at higher-order (higher J) transitions in CO for which level pop-
ulations will be smaller because molecular clouds are cold. And others give up on CO
altogether and try to detect emission from optically thin dust (and then multiply by
a factor of 100 to account for the gas-to-dust ratio) or other other, even less abun-
dant molecules (whose abundance relative to hydrogen must be inferred from chemical
modelling).

(e) Based on your answer to (d), would you conclude that this transition is a good way
to measure the mass of molecular gas in a galaxy? How about the temperature?

If the line were merely thermally broadened, we could simply measure the width of
the line and thereby infer the temperature. Real-world difficulties crop up, however, from
non-thermal, bulk streaming motions (turbulence) in molecular clouds that complicate
interpretation of a measured line width.

Because the clouds tend to be so optically thick in this transition, we are only
sensitive to their outer skins, so an emission line measurement does not provide a good
way to measure the mass of the most massive of clouds. In other words, we can’t tell
how much material lies beneath the surface of optical depth unity. See part (d) for
alternative ways of measuring the masses of molecular clouds.

11
 Ay 201 – Radiative Processes – Problem Set 5 Solutions
Linda Strubbe and Eugene Chiang
October 2, 2003

Problem 1. Energy Density of Starlight and Grain Temperature

Interstellar space is filled with radiation. The bulk of the radiation arises from light
emitted by the most massive (O-type) stars, each of mass 102 M . The number of such
stars in our Galaxy is about 5 × 104 , distributed over a cylindrical disc of radius 50 kpc
and height 200 pc.

(a) Estimate the energy density of starlight in the Galaxy. Express your answer in
eV /cm3 . To estimate the luminosity of the most massive stars, use the fact that they
are radiating at the Eddington limit; i.e., recall problem set 3.

Energy density (u) is related to flux (F ) by the following equation:

F
u= , (1)
c

and the flux from an object with luminosity L is

L
F = , (2)
4πr 2

where r is the distance to the object. Our strategy in this problem is thus to find the
total flux at an “average” point in the Galaxy by adding up the flux received at that
point from each O-star in the Galaxy.

To do this, we need a model to describe the locations of all the O-stars in the Galaxy.
Let’s approximate the Galaxy as a cylinder, with all the O-stars uniformly distributed
within the Galaxy’s volume. To have an idea of how dense the Galaxy is in O-stars,
let’s find the average distance (rnear ) between an observer in the Galaxy and the nearest
O-star. Let N ≡ the number of Galactic O-stars, rGal ≡ the radius of the Galaxy, H ≡
the height of the Galaxy, and VGal ≡ the volume of the Galaxy.

1/3 !1/3

VGal 2 H
πrGal
rnear ∼ = = 320 pc > H. (3)
N N

Since the distance to the nearest O-star is greater than the height of the Galaxy, we
can make the approximation that all O-stars lie in a plane. (The radial component
of the distance to any star is greater (and usually much greater) than the component
of the distance normal to the disk.) Let’s consider an observer at the center of the

1
Galaxy. Beyond a distance of about rnear , the fact that the stars are point sources is
unimportant. Let’s therefore model the Galaxy as a 2-dimensional annulus with inner
radius of rnear and outer radius of rGal , which has a continuous uniform distribution of
luminosity sources (the individual stars are smeared out over the whole area). Then we
can define a luminosity surface density σ :

LO−star N
σ≡ 2 (4)
πrGal

We can calculate LO−star using the relationship between mass and luminosity we found
in problem set 3, question 1 (p47 in Rybicki & Lightman) :

M
 
LEDD = 1.25 × 1038 erg s−1 (5)
M

with MO−star = 102 M , so

LO−star ∼ 1040 erg s−1 . (6)

Now we can calculate the total flux received at the center of the Galaxy:

σdA
dF = (7)
4πr 2
Z rGal σ σ rGal
 
F = 2
2πr dr = ln (8)
rnear 4πr 2 rnear

So we can finally obtain the energy density at the Galactic center:

F 1 LO−star N rGal
 
u= = 2 ln (9)
c 2 cπrGal rnear
 
50
(1040 erg s−1 )(5 × 104 )ln .32
u= = 0.4eV cm−3 (10)
2(3 × 1010 cm s−1 )(1.5 × 1023 cm)2 (1.6 × 10−12 ergeV−1 )

In truth, the logarithmic correction factor over-estimates the true correction factor be-
cause stars from the other side of the Galaxy are obscured from view by dust; recall
the slide show of the first class, during which I said that most of the visible starlight
originates from within ∼1 kpc of Earth. Also see the next problem where we show that
at visible wavelengths, we suffer about 1 mag extinction for every kpc travelled. But all
logarithms are of order unity anyway, so the error accrued in including every O-star is
negligible; compare the flux from the single nearest O-star and the flux from all O-stars
and you will see that they differ by a factor of 5.

2
(b) The interstellar radiation field (ISRF) heats dust grains in the interstellar medium
(ISM). Estimate the temperature, T , of the largest grains in the ISM. These grains have
radii of a ∼ 0.1 µm. Take their emissivity (Qemis ) = absorptivity (Qabs ) to be unity
for wavelengths shorter than 2πa and to fall off as 2πa/λ for longer wavelengths. Is
the grain hotter, cooler, or equal to the temperature of an ideal blackbody placed in the
ISRF?

Most of the interstellar radiation field (ISRF) is produced by O-stars, which have
temperature T ∼ 40000K. From the Wien Peak Law, we find the peak wavelength of
the ISRF:

0.29 cm K
λpeak = = 0.07µm (11)
T

In class we learned that about 95% of the intensity from a blackbody comes from
the range of 13 λpeak to 3λpeak , so most Bλ (ISRF ) comes from .02µm < λ < 2µm. Thus
almost all of the energy absorbed by our grain from the ISRF is in the wavelength
range λ < 2πa where Qabs = 1, so the grain absorbs energy from the ISRF pretty much
perfectly:

Fabs = Ffrom O−stars = cu = 1.2 × 1010 eV s−1 cm−2 = 0.02 erg s−1 cm−2 . (12)

The grain is likely to be emitting most of its power at infrared wavelengths that are
much greater than 2πa. Then we can approximate the emitted flux as

∞ 2πa
Z
Femis ≈ π Bλ (Tgrain ) dλ (13)
0 λ

Here we have assumed that most of the area (power) underneath the Planck function
is at λ  2πa, so the error accrued at λ ≤ 2πa—where our expression for Qemis is
incorrect, because it should be equal to 1—is negligible. Guessing different temperatures
in Mathematica gives T ≈ 19 K.

We can arrive at this result without having to resort to numerics. Write Femis =
σT 4 Qemis . Now Qemis = 2πa/λ for λ > 2πa. The wavelengths at which most of
the power from our grain will be emitted are near λ ≈ hc/5kT from the Wien peak
law. Now use of the Wien peak law is not rigorously justified because the Wien peak
law is for blackbodies and our grain is not a blackbody. Still, the grain will be at
some temperature T and it will be emitting most of its power at wavelengths which are
inversely proportional to T ; the numerical value of “5” in the Wien peak law should shift
to a slightly lower number because our grain’s power at long wavelengths is diluted, and
therefore most of the power will get shifted to shorter wavelengths. Let’s stick with “5”
and see how our answer depends on this guess. Then Femis = σT 4 2πa/(hc/5kT ), which

3
we set equal to Fabs above. This yields T ≈ 17K, and it scales as “5”−1/5 —not that
sensitively.

The temperature of an ideal blackbody placed in the ISRF is

4
FfromO−stars = σTideal ⇒ Tideal = 4K (14)

Our grain is thus hotter than a blackbody at thermal equilibrium in the ISRF, as it
must be since it absorbs with perfect efficiency but emits with imperfect efficiency; to
maintain global energy balance, the grain must boost its temperature above that of a
blackbody.

(c) At what wavelength, λpeak , does νFν (the SED; recall problem set 3) of such grains
peak? Give a number in microns, and also an expression in terms of T , fundamental
constants, and dimensionless numbers.

We can continue our approximation that all energy is radiated from the grain in the
Qemis = 2πa
λ ∝ ν regime. To find λpeak of νFν , we find νpeak by setting the derivative of
νFν to zero, and then converting from νpeak to λpeak . Since we’ll be setting the derivative
to zero, we can drop constant coefficients in the expression for the SED.

ν5
νFν ∝ νBν Qemis (ν) ∝ ν 2 Bν ∝ hν (15)
kTgrain
e −1

hν
Define x ≡ kTgrain , take the derivative of equation 15, and set it equal to 0:

d(νFν ) d ν5 d x5 5x4 (ex − 1) − x5 ex

∝ hν ∝ = = 0. (16)
dν dν kTgrain dx ex − 1 (ex − 1)2
e −1

Using Mathematica’s NSolve on the numerator, we find that xpeak = 4.97.

kTgrain ch
νpeak = xpeak ⇒ λpeak = = 140 µm (17)
h kTgrain 4.97

(d) These largest grains also carry the lion’s share of the mass in the interstellar grain
distribution. Given a dust-to-gas mass density ratio of ρdust /ρgas ∼ 10−2 (metallicity),
a rough average density of gas in the Galaxy of 0.1 H atom cm−3 , and a radial extent of
the Galaxy of 50 kpc (kiloparsecs), calculate the specific intensity of the infrared Milky
Way at a wavelength of λpeak . Express in mJy arcsec−2 , where mJy = milliJansky
= 10−3 × 10−23 erg s−1 cm−2 Hz−1 is a radio astronomer’s unit of “flux density” (the
“density” here refers to spectral density; i.e., it refers to the per Hertz) and arcsec = 1
arcsecond = 1/206265 (a phone number worth remembering) radians.

4
 Let’s again assume we’re in the center of the Galaxy, so our line-of-sight distance is
rGal . We can calculate the number density of grains in the Galaxy, using the information
given in the problem, and an estimate of the density of a single dust grain (ρgrain ∼
2 g cm−3 ):

ρdust 4
ndust = ngas mH ( πa3 ρgrain )−1 = 2 × 10−13 cm−3 (18)
ρgas 3

Now we multiply ndust by rGal to get a column density, and then multiply by the
effective cross-sectional area of a grain, Qemis πa2 , to obtain the optical depth along the
line of sight:

τlos = (ndust )(rGal )Qemis (πa2 ) = 0.04 (19)

These grains comprise a homogeneous, thermally emitting (and absorbing) slab of source
function Sν = Bν . The observed specific intensity for such a slab equals

Iνpeak = Sνpeak (1 − e−τlos ) (20)

−0.04
= Bνpeak (Tgrain )(1 − e ) (21)
≈ Bνpeak (Tgrain ) × 0.04 (22)
−14 −1 −2 −1 −1
≈ 3 × 10 erg s cm Hz sr (23)
−2
≈ 80mJy arcsec (24)

using 1 sr = (206265 arcsec)2 . We bravely compare this answer to the truth as measured
by the DIRBE sky map ([Link] images/[Link]):
in the plane of the Galaxy, DIRBE measured ∼8000 MJy/sr at λ = 100 microns. Our
answer of 80 mJy/square arcsecond converts to ∼3000 MJy/sr at λ = 140 microns. Not
bad for an order-of-magnitude estimate!

Problem 2. Dust Opacity

A rough model for the dust in the ISM tells us dn/da ∝ a−3.5 , where dn is the differential
number of dust grains having radii between a and a + da. The largest radius in the
distribution is amax = 0.1µm, and the smallest radius is amin = 0.001µm.

(a) Plot the opacity, κ(λ), contributed by all dust grains as a function of wavelength
from λ = 0.1µm to λ = 10µm. Express the opacity in units of cm2 g−1 , where the g−1
equals “per gram of gas.” Use whatever parameters you need as given by problem 1 of
this set.

Indicate over every decade in wavelength which grain sizes dominate the opacity. (For
example, at wavelengths between λ = 0.1 and 1 µm, do the a ∼ 0.1µm grains dominate

5
the opacity? If not, do the a ∼ 0.01µm grains dominate? And if not them, what about
the a ∼ 0.001µm grains?)

By definition,

Z amax dn
ρdust κdust (λ) = Qabs (a, λ)πa2 da (25)
amin da

where dn/da is the differential number density of grains having radii between a and
a + da, and ρdust is the volumetric (bulk) density of grains in space (not the internal
grain density, ρgrain , which is like that of water). Note also that κdust is the cross-section
for absorption by dust per gram of DUST. At the very end of the problem, we only have
to multiply by the dust-to-gas mass ratio, 10−2 , to get the desired cross-section per gram
of GAS.

Begin by finding an expression for dn/da. All we are told is that dn/da ∝ a−3.5 , so we
need to find the constant of proportionality. We can find it by using the fact that the
mass of all grains, integrated over the entire size range, must yield a volumetric mass
density of ρdust . In other words, defining dn/da = Ca−3.5 ,

Z amax dn 4 3
ρdust = πa ρgrain da (26)
amin da 3
4π
≈ Cρgrain a1/2
max (27)
3
ρdust 3 −1/2
→C = a (28)
ρgrain 8π max

since the integral is dominated by the upper limit. Plug C into our beginning expression
for κdust :

3
Z amax
κdust (λ) = √ a−1.5 Qabs (λ, a)da (29)
8ρgrain amax amin

Now we attack the integral. First we note that for wavelengths longer than 2πamax ≈
0.6 µm, we are always on the falling power-law section of Qabs = 2πa/λ. To wit,

Z amax 2π amax −0.5

Z
a−1.5 Qabs (λ > 0.6 µm)da = a da (30)
amin λ amin
4π √
≈ amax (31)
λ

So for λ > 0.6 µm, plugging our answer for the integral into (29),

6
 3π
κdust (λ > 0.6 µm) = λ−1 (32)
2ρgrain

Finally convert to the desired κ by multiplying by ρdust /ρgas ∼ 10−2 :

3π
κ(λ > 0.6 µm) = λ−1 (33)
200ρgrain

At these wavelengths, the BIGGEST grains, near amax , dominate the opacity. When we
evaluated the integral in (31), amax dominated the integral.

Now we attack λ < 0.6 µm. The integral in (29) breaks into two pieces,

Z λ/2π 2πa −1.5

Z amax
a da + a−1.5 da (34)
amin λ λ/2π

which evaluates to

√
4 2π 4π √ 2
√ − amin − √ (35)
λ λ amax

This cumbersome expression numerically evaluates to 2141 (cgs) at λ = 0.1 µm, where
the major contribution is from the first (and only positive) term. Therefore at λ =
0.1 µm, the grains that contribute most to the opacity are those whose radii a ≈ λ/2π ≈
0.01 µm. At λ = 0.6 µm, the cumbersome expression evaluates to 630 (cgs); the major
contribution is from a ≈ 0.1 µm.

Putting it all together, the opacity (per gram of GAS) at λ = 0.1 µm equals

κ(λ = 0.1 µm) = 2500 cm2 g−1 (36)

(dominated by smallish grains, a ≈ 0.01 µm), falling to

κ(λ = 0.6 µm) = 700 cm2 g−1 (37)

(dominated by the biggest grains, a ≈ 0.1 µm), and falling from thereon as λ−1 , so that
at λ = 10 µm,

κ(λ = 10 µm) = 42 cm2 g−1 (38)

7
(and dominated throughout this long wavelength regime by the biggest grains, a ≈
0.1 µm).

(b) Convert your plot to read magnitudes of absorption per kiloparsec travelled in the
Galaxy (mag/kpc) as a function of λ. Again, use whatever parameters you need from
problem 1 above.

A magnitude is a logarithmic unit measured in base 2.5; each magnitude of absorption

reduces the flux from a star by a factor of 2.5; a rule of thumb is that 5 magnitudes is
a factor of 100. (Editorial note: Astronomer’s magnitudes were invented by Hipparcos
to measure star brightnesses; a magnitude 0 star is 2.5 times brighter than a magnitude
1 star, and so on. For some reason, optical and near-infrared astronomers refuse to
abandon this system, even though it runs backwards, crosses zero, and is normalized to
different values depending on the wavelength band. As far as I can see, its only redeeming
quality is that base 2.5 is close to the natural base e = 2.712...)

By definition,

τλ (s) = ρgas κλ s (39)

As light passes through a distance s in the ISM, the flux is reduced by eτλ (s) . We want
to plot this quantity on the magnitude scale, which is base 1001/5 , and which increases
as flux is reduced. Therefore,

1
magλ (s) = −log1001/5 e−τlambda (s) = τλ (s)log1001/5 e = τλ (s) 2 (40)
5 ln10

so

mag 3 × 1021 cm
 
= 1.086 × ρgas κλ × (41)
kpc λ kpc

where ρgas = ngas mH from 1d. Plugging in numbers from 2a and from 1d, we get

mag mag
 
= 1.4 (42)
kpc λ=0.1 µm kpc
mag mag
 
= 0.4 (43)
kpc λ=0.6 µm kpc
mag mag
 
= 0.02 (44)
kpc λ=10 µm kpc

8
The plot of mag/kpc (λ) has identical shape to the plot for κ(λ). Moral: the extinction
at infrared wavelengths is much reduced below that in the optical.

9
 Ay 201 – Radiative Processes – Problem Set 4 Solutions
Linda Strubbe and Eugene Chiang
October 2, 2003

Problem 1. Hyperfine Emission from Neutral Hydrogen

This problem is an exercise in learning more astronomy jargon and in practicing some
of the formalism in Rybicki & Lightman Chapter 1.

Neutral hydrogen in the electronic ground state can be in one of two hyperfine states.
Denote the number density of atoms in the ground hyperfine level (singlet state) as n0 ,
and the number density of atoms in the excited hyperfine level (triplet state) as n1 .
DEFINE the excitation temperature, Tex , of the transition as

n1 g1
= e−hν/kTex . (1)
n0 g0

Here hν = hc/λ is the mean energy difference between the levels, and g0 = 1 and g1 = 3
are the statistical weights of the levels. The excitation temperature is merely another
way of expressing the ratio of ground state to excited state populations. By definition,
if a gas is in local thermodynamic equilibrium (LTE) at some gas kinetic temperature
T , then Tex = T ; the level populations are distributed in Boltzmann fashion at the
local temperature T . Some people refer to the excitation temperature for the λ = 21 cm
transition as the spin temperature. But use of the term “excitation temperature” is
general to any line transition; it is simply a measure of how excited an atom is.

(a) Define T∗ = hν/k and compute its value.

It is likely that Tex  T∗ . For the remainder of this problem, work in the T∗ /Tex  1
limit.

For the hyperfine transition in hydrogen (λ = 21cm),

hν hc
T∗ ≡ = = 0.067 K (2)
k λk

The excitation temperature is defined by the following:

n1 g1 hν
= e− kTex (3)
n0 g0

with g0 = 1 and g1 = 3. We assume that Tex  T∗ . The next few parts take equations
from Rybicki & Lightman pages 30 and 31.

1
(b) Write down the absorption coefficient, αν (units of per length), for this transition.
Express your answer in terms of φ(ν) (the line profile function), A21 (Einstein A coef-
ficient), λ, whatever densities you need, and T∗ /Tex . Do not forget the correction for
stimulated emission.

Including the correction for stimulated emission,

hν hν g0 n1
αν = φ(ν)(n0 B01 − n1 B10 ) = n0 B01 (1 − )φ(ν). (4)
4π 4π g1 n0

2hν 3 g0
With A10 = c2 g1 B01 , equation 4 becomes

hν g1 c2 hν
n0 3
A10 (1 − e− kTex )φ(ν) (5)
4π g0 2hν
1 c2 T∗
= n0 3 2 A10 (1 − e− Tex )φ(ν) (6)
4π 2ν
3 T∗
≈ n0 λ2 A10 (1 − (1 − ))φ(ν) (7)
8π Tex

so
3 T∗
αν ≈ n0 λ2 A10 ( )φ(ν) (8)
8π Tex

(c) Write down the volume emissivity, jν (units of erg s−1 cm−3 Hz−1 sr−1 ), for this
transition. Use whatever quantities defined above that you need.

hν0
jν = n1 A10 φ(ν) (9)
4π
c
with ν0 = 21cm .

(d) Write down the source function, Sν , for this transition.

jν hν0 3 T∗
 −1
Sν ≡ = n1 A10 φ(ν) n0 λ2 A10 ( )φ(ν) (10)
αν 4π 8π Tex
2 n1 Tex 2 Tex Tex
   
hν
= hν0 λ−2 = hν0 3e− kTex λ−2 ≈ 2hν0 λ−2 (11)
3 n0 T∗ 3 T∗ T∗

(e) Write down the specific intensity, Iν , of a cloud of HI that is optically thin along the
line-of-sight (l-o-s). Take the l-o-s dimension of the cloud to be L, and give the answer
only to leading order in τ  1, where τ is the optical depth at an arbitrary wavelength.

Does your answer depend on Tex ?

2
If someone gives you a spectrum of the 21 cm line that appears in emission and tells you
that the line was emitted from an optically thin cloud, what physical quantities can you
infer from the spectrum?

The source function Sν is constant throughout the uniform HI cloud, so Sν is inde-

pendent of τν .

Iν (τν ) = Iν (0)e−τν + Sν (1 − e−τν ) (12)

Since we want to know the specific intensity of the cloud itself, we can say that
Iν (0) = 0 (there is no light entering the cloud). We can also apply our assumption of an
optically thin cloud (τν  1 for all ν).

Iν (τν ) = Sν (1 − e−τν ) ≈ Sν (1 − (1 − τν )) = Sν τν (13)

Now we use the definition of τν and the fact that we want the specific intensity of
light emerging from the cloud, meaning an optical depth corresponding to a path length
L.

Iν (L) ≈ Sν nν σν L = Sν αν L = jν L (14)
hν0
= n1 A10 φ(ν)L (15)
4π

The specific intensity thus has no dependence on Tex .

Now ν0 and A10 are known quantities from the lab/quantum mechanics. A spectrum
would tell us Iν and φ(ν), so we could find n1 L, the column density of excited hydrogen
atoms. Assuming Tex  T∗ , n0 ≈ gg10 n1 = 13 n1 , so we can find n0 L as well (the column
density of ground state hydrogen atoms). We can add these column densities to get the
total column density of hydrogen atoms:

4 16π
nL = (n1 + n0 )L = n1 L = (hν0 )−1 A−1
10 (φ(ν)) Iν
−1
(16)
3 3

Knowing the column density gives us the mass if the cloud is roughly spherical and we
know its length.

Finally, some idea of the temperature and/or velocity field of the gas could be gained
by looking at the line shape (line profile).

(f ) Write down the optical depth of the cloud. Does your answer depend on Tex ?

Well,

3
 3 T∗
 
τν = nν σν L = αν L = n0 λ2 A10 φ(ν)L (17)
8π Tex

The optical depth does depend on Tex .

(g) How large would L have to be for the cloud to be marginally optically thick? Use our
canonical gas density of n = 1 cm−3 , a gas temperature of T = 100 K, and an excitation
temperature Tex = T . Assume the line is only thermally broadened.

The condition for marginal optical thickness is τ ∼ 1.

1
n = n1 + n0 ≈ 4n0 ⇒ n0 ≈ n (18)
4
3 T
 
n0 λ2 A10
∗
1= φ(ν)L (19)
8π Tex
32π −2 −1 Tex
 
⇒L∼ λ A10 φ(ν)−1 (20)
3n T∗

To find L, we evaluate equation 20 at λ = 21cm. The Einstein A coefficient ∼

3 × 10−15 s−1 at 21 cm for this transition, we use T∗ from part (a), and take φ(ν) ∼
1/∆ν ∼ c/(vν0 ) ∼ λ/v at line center, where v ∼ 1 km/ s for T = 100 K gas. Then

L ∼ 50 pc (21)

Problem 2. Multiple Multipoles

(a) Write down the electric field a distance r away from a monopole of charge q.

q/r 2

(b) Someone moves another monopole of charge −q next to the original monopole. The
two charges are separated by a distance b. Derive, to order-of-magnitude, the factor by
which this dipole electric field is reduced from the monopole field.

A first-order Taylor expansion gives the reduction factor b/r.

(c) Someone moves two more charges, −q and q, into position to form a square. The
edge of the square has length b. Going around the square, the charges are −q, q, −q,
and q. Derive, to order-of-magnitude, the factor by which this quadrupole electric field
is reduced from the dipole field.

A second-order Taylor expansion gives the reduction factor b/r.

4
(d) Draw a picture of a pure electric octopole, that is, a charge distribution for which the
electric field decreases as 1/r 5 (and no less gradually). For bonus points, draw a picture
of an electric hexadecapole (electric field dies no less gradually than 1/r 6 ).

You can draw the octopole by placing charges on the vertices of a cube, with alter-
nating signs.

Here’s another way to draw an octopole: draw a circle, and arrange 8 charges sym-
metrically on the circle with alternating signs. I spent part of Thanksgiving holiday
verifying that this last configuration indeed gives a field that dies as b3 /r 5 along an
axis that overlays a diameter of this circle intersecting two of the charges. So one can
construct an octopole without having to go into the 3rd dimension!

Presumably, one can construct an electric hexadecapole by drawing a circle, and

arranging 16 charges symmetrically on the circle with alternating signs. (But I haven’t
checked explicitly that such a configuration gives a field that dies as b4 /r 6 .)

(e) Now imagine the dipole and quadrupole configurations in (b) and (c) rotating about
their centers-of-charge with frequency ν. We have a rotating barbell and a rotating
square, respectively. (Think CO and H2 ).

To order-of-magnitude, what is the maximum distance from each object inside of which
the electric fields are nearly perfectly in phase with the rotation?

This is the boundary of the near zone, inside of which the electric field geometry rotates
with frequency ν as if it were a rigid body.

The information that the charge distribution has rotated propagates at the speed of
light away from the distribution. Therefore a distance ∼c/ν marks the boundary of the
near zone.

(f ) Electromagnetic waves are emitted at the boundary of the near zone, into the far
(radiation) zone.

Derive (one line of argument suffices) the factor by which the power carried by waves
emitted by the rotating quadrupole is smaller than the power carried by waves emitted by
the rotating dipole.

The waves are launched at the boundary of the near zone.

Since the power carried by electromagnetic waves goes as the square of the field, we
must square the reduction factor for the field. In other words, the power reduction factor
is (b/r)2 , where r ∼ c/ν ∼ λ is the size of the near zone. Then the power reduction
factor is (b/λ)2 , as used throughout this class.

Problem 3. Rotating Magnetic Dipole

5
Rybicki & Lightman Problem 3.1

See solution 3.1 at back of the book.

6
 Astronomy 201 – Radiative Processes – Solution Set 5

Sam Williams and Eugene Chiang

Readings: Hand-outs from Osterbrock; Rybicki & Lightman 9.5; however much you like
of Mihalas 108-114, 119-127, 128-137 (even skimming Mihalas can prove enlightening);
however much you like of the hand-out of Purcell & Field (1957).

Problem 1: Maxwellian is not quite LTE

A certain atom suffers collisions with another species (perhaps merely its own). These
collisions populate and de-populate a certain level in the atom that lies above another
level in that atom by energy E. Prove that if the relative velocity distribution, f(v),
between the atom and surrounding colliders obeys a Maxwellian at temperature T
( ∫ f(v)dv = 1 ), then the excitation rate coefficient,

q12 = ∫ σ 12 (v ) f (v )vdv

is related to the de-excitation rate coefficient,

q 21 = ∫ σ 21 (v ) f (v )vdv

by
g 2 − E / kT
q12 = q 21 e
g1

where gi is the statistical weight of level i, σij is the velocity-dependent collisional cross-
section for making a transition from level i to level j, and v is the relative velocity
between the atom and the colliding species. You should find this problem easy given the
Einstein analogue presented in lecture.

NOTE: Nowhere in this discussion have we assumed that the level populations are
distributed in a Boltzmann fashion at temperature T. That is, none of the relations above
assume LTE (local thermodynamic equilibrium); LTE is a more restrictive condition than
merely assuming that the relative velocity distribution is Maxwellian.

1
 g 2 − E / kT
show : q12 = e q 21
g1
font note : v = velocity , υ = frequency
start with the Einstein analogue : g1σ 12 (v)v 2 = g 2σ 21 (v ' )v '2
3
 m 
2
 − mv 2 
Multiply both sides by: 4π   exp  
 2πkT   2kT 
3 3
 m 
2
 − mv 2   m 
2
 − mv 2 
g1σ 12 (v)v 4π 
2
 exp  = g 2σ 21 (v ' )v '2 4π   exp 
 2πkT   2kT   2πkT   2kT 
remember : 1 2 mv 2 = 1 2 mv '2 + E
−1 −1 − mv '2
so K
kT
( 1
2 mv )
2
=
kT
( 1
2 mv ' + E =
2
) 2kT
−
E
kT
3 3
 m 
2
 − mv 2   m 
2
 − mv '2  −E
g1σ 12 (v )v 4π 
2
 exp 
 
 = g σ ( v ' ) v ' 2
4π   exp   exp 
 2πkT   2πkT 
2 21
 2kT   2kT   kT 
3
 m 
2
 − mv 2 
isn' t it lucky that : f(v) = v 4π  2
 exp   so K
 2πkT   2 kT 
g1σ 12 (v ) f (v ) = g 2σ 21 (v ' ) f (v' )e − E / kT

now multiply by vdv and integrate both sides

∫gσ 1 12 (v ) f (v )vdv = ∫ g 2σ 21 (v' ) f (v ' )e − E / kT vdv

remember: vdv = v'dv'

∫gσ 1 12 (v ) f (v )vdv = ∫ g 2σ 21 (v' ) f (v ' )e − E / kT v' dv '

g 2 − E / kT
∫σ 12 (v ) f (v )vdv =
g1
e ∫ σ 21 (v' ) f (v' )v' dv
remember : q12 = ∫ σ 12 (v ) f (v)vdv and q 21 = ∫ σ 21 (v ) f (v )vdv

g 2 − E / kT
so K q12 = e q 21
g1

2
Problem 2: 21 (flavors of) Temperatures for 21cm radiation

This problem continues our examination of the famous hyperfine transition in neutral
hydrogen, begun in problem set 4. Here we try to understand what sets the excitation
(spin) temperature, Tex. (In the last problem set, we merely contented ourselves with the
statement that Tex>>T* ) While we're at it, we learn more astronomer's jargon.

In general, the excitation temperature of a transition is influenced by two factors: (1) the
radiation field, and (2) collisions with surrounding species.

The radiation field can either excite the atom through photon absorption, or de-excite
through stimulated emission. Measure the strength of the ambient radiation field at 21 cm
by J , the mean (i.e., angle-averaged) intensity integrated over the hyperfine line profile
(recall Rybicki & Lightman chapter 1).

As for collisions, consider here exciting and de-exciting collisions with fellow neutral
hydrogen atoms; Purcell and Field (1957, hereafter PF) conclude that collisions between
a given electronic-ground-state H atom and other electronic-ground-state H atoms are
most important in the predominantly neutral HI clouds of the ISM. (Electrons are 42
times faster and tend to dominate the excitation dynamics, but assume there are too few
of them in these cold clouds.)
Denote the collisional excitation rate coefficient by q12, and the collision de-excitation
rate coefficient by q21 (see problem 1). Assume for this problem that both hyperfine-
excited and hyperfine-ground atoms can excite or de-excite the hyperfine level in an
atom.

(a) Write down the equation of global (not detailed!) balance for this transition. That is,
write down the statement that the rate of excitations (from all possible channels) per
volume per time equals the rate of de-excitations (from all possible channels) per volume
per time.

Use only the following variables: n1 and n2 are the number densities of atoms in the
ground and excited states, respectively, n = n1+ n2, TK is the kinetic temperature of the
atoms that move according to a Maxwellian, q12, any Einstein coefficients you want, J ,
and the statistical weights g1 and g2 of the ground and excited states, respectively.

What you have written down is an equation for the excitation temperature (Tex↔n1/n2) in
terms of the radiation field and the rate of collisions. Regard the latter two as given
throughout this problem.

3
 Rabsorption + Rcollisional_excitation = Rspontaneou s_emission + Rstimulated_emission + Rcollisional_de −excitation
n1 B12 J + n1nq12 = n2 A21 + n2 B21 J + n2 nq 21
n1 (B12 J + nq12 ) = n2 ( A21 + B21 J + nq 21 )
n1 ( A21 + B21 J + nq 21 )
=
n2 (B12 J + nq12 )
g1 T* / TK
from #1 : q21 = e q12
g2
 g 
 A21 + B21 J + nq12 1 eT* / TK 
n1  g2 
=
n2 (B12 J + nq12 )

(b) DEFINE a "radiation temperature," TR, from J as

J = Bυ (TR )

Note that we are NOT saying the ambient radiation field is Planckian. We are merely
DEFINING a number TR by using Planck's function, Bυ, where υ = 1420MHz, the
frequency of the 21 cm line.

Re-write your equation in (a) to solve for Tex in terms of the following variables: TK, TR,
T* ≡ hν/k (recall last problem set), and the dimensionless variable

g1 nq12 T*
z=
g 2 A21 TK

where A21 = 2.85x10-15 s-1 is the Einstein decay coefficient. Use the very likely condition
that TK, TR >> T* to rid your equation of all exponentials.

Verify that if z >> 1, Tex ≈ TK (collisions beat radiation; the transition is in LTE at TK),
but that if z << 1, Tex ≈ TR (radiation beats collisions; the transition is not in LTE at TK).

4
 note : T* << TR and T* << TK
c2
g2 c 2
2 hυ 3 2hυ 3 TR
B21 ≡ A21 B12 ≡ A21 J = ≈
2 hυ 3 g1 2 hυ 3 c 2 [exp(T* TR ) − 1] c 2 T*

we now see why the form of J was chosen

TR g T 
A21 + A21 + nq12 1  * + 1
n1
=
T* g 2  TK  = g1 exp(T T ) ≈ g1 1 + T* 
g 2  Tex 
* ex
n2  g 2 TR  g2
 A21 + nq12 
 g1 T* 

 T g nq  T  T g nq12 T* 
A21 1 + R + 1 12  * + 1 1 + R + 1  + 1
T
1+ * =  T* g 2 A21  TK 
=
T* g 2 A21  TK 
Tex g g T nq   TR g1 nq12 
A21 1  2 R + 12   + 
g 2  g1 T* A21   T* g 2 A21 

TR T 
1+ + z + z  K 
g1 nq12 T* T* T*  T* 
let z = so K 1 + =
g 2 A21 TK Tex 1
(TR + zTK )
T*
 T  T 
T* 1 + R + z + z  K   − (TR + zTK )
1 T*  T*   T + TR + zT* + zTK − (TR + zTK )
=  = *
Tex T* (TR + zTK ) T* (TR + zTK )

1 1+ z TR + zTK
= so K Tex =
Tex TR + zTK 1+ z

If z >> 1, Tex ≈ zTK / z = TK

If z << 1, Tex ≈ TR

5
(c) To order of magnitude (actually much better than that), what fraction of HI is in the
excited hyperfine state? Recall that g1 = 1 and g2 = 3 and use the very likely condition
that TK, TR >> T*.

n2 g 2
= exp (− T* Tex ) if we assume : T* << Tex
n1 g1
n2 g 2
= = 3 so K n 2 = 3n1
n1 g1
n2 n2 3n1
= = = 0.75
n n1 + n 2 n1 + 3n1

(d) Estimate the value for z for an HI cloud at TK = 100K, n = 1 cm-3. Use Table 1 and
equation (9) of PF; note that PF's collision frequency ν = n<σν> is not the same as our
line frequency ν; call PF's ν = νPF; then nq12 = 3 νPF / 8. (For those interested, the 3/8
can be understood easily; skim the first 3 pages of PF and use your answer for part (c).)

Based on your answer, would you expect collisions or radiation to be more important in
determining the degree of excitation?
1
 8kTK 
2

2 nσ  
g1 nq12 T* g 3 8 υ pf hυ 3 g1  πM  hυ
z= = 1 =
g 2 A21 TK g 2 A21 kTK 8 g 2 A21 kTK
(1) ( )( )( )  8(1.38 × 10 )100 
1

2 5.65 × 10 −15 6.6 × 10 − 27 1.42 × 10 9 −16 2

=
( )( )
8 2.85 × 10 −15 1.38 × 10 −16 (100 )  π (1.67 × 10 ) 

− 24

= 34.43

z>>1, so the system is collisional

(e) "Critical densities," ncrit, for exciting the line by collisions are defined by setting the
rate of radiative de-excitations equal to the rate of collisional de-excitations. Show that
such a procedure gives
A21
nc =
q21

and solve for its value for this line at TK = 100K.

One can define ncrit for any line transition at any temperature; it is a crude gauge of the
density of colliders required for collisions to be important in exciting the line.

6
 ignore stimulated emission
n 2 A21 = n 2 ncrit q 21
A21
ncrit =
q 21
A21 g2 A21 g2 8 A21
ncrit = = = =
g 1 T* / TK 3 υ e * KT / T 1

e q12 g1 8 n pf g1 3  8kT K  hυ / kTK

2

g2 2σ   e
 πM 

=
(
8 2.85 × 10 −15 )
) (( ) ( )( )
1
 8 1 .38 × 10 −16 100   6 .6 × 10 − 27 1.42 × 10 9
(
2

2 5.65 × 10 −15   exp 

 π 1.67 × 10
− 24 
 )  1.38 × 10 −16 100 ( ) 

= 2 × 10 −5 cm −3

(f) Suppose radio observations are made that spatially resolve emission from a uniform
HI cloud that is optically thick to its own 21 cm line radiation. Prove that the observed
specific intensity, Iν equals

2kTex
Iυ =
λ2

optically thick, so use S υ

−1 −1
jυ 2 hυ 3  g 2 n1  2 hυ 3  g 2 g1 T* Tex 
I υ ≈ Sυ = = {see RL eq 1.79} = 2  − 1 = 2  e − 1
αυ c  g1 n2  c  g1 g 2 
−1
2 hυ 3  T  2hυ 3 Tex 2hυ 3 kTex 2υ 2 kTex 2kTex
≈ 2 1 + * − 1 = 2 = 2 = = 2
c  Tex  c T* c hυ c2 λ

7
Problem 3: Photoionized Quasar winds

Quasars are luminous X-ray sources sitting in the cores of ancient galaxies. They are
supermassive black holes that accrete surrounding gas; the gravitational potential energy
of gas spiralling down the potential well of the black hole is converted into radiation.
The luminosity of a typical quasar is L≈1046 erg s-1, mostly in the Lyman continuum,
with a substantial fraction in a power-law X-ray tail. The flux density in the X-ray tail
obeys Fν ∝ ν -β, with 1 < β < 2. (Fν has units of energy per time per frequency per
area).

This radiation streams out from regions closest to the black hole and may illuminate more
distant but still circumnuclear gas. In so doing, it photo-ionizes the more distant gas and
threatens to turn it into a complete and utter plasma, with all electrons stripped from
parent nuclei. This problem estimates the varying degrees to which this threat is made
good. It is inspired by work done by Norm Murray.

(a) Define the "ionization parameter," ξ, as the number density of Lyman continuum
photons, η, divided by the total number density of hydrogen, n H = n H + + n H 0 , where n H +
is the number density of ionized hydrogen, and n H 0 is the number density of neutral
hydrogen. Show that in photo-ionization equilibrium (rate of photo-ionizations per
volume per time equals the rate of radiative recombinations per volume per time):

N 0 10 −6
=
N+ ξ
valid for n H 0 / n H + << 1 . This is a rule of thumb worth remembering. Just consider
photons near the ionization edge! Furthermore, assume a 1-dimensional geometry for the
problem; consider only a semi-infinite slab, upon which is incident a radiation flux. In
your derivation, you will see that the dimensionless factor of 10-6 can be expressed in
terms of fundamental constants.

u η
note : η = ξ=
hυ N

σ bf σ bf
N 0 4π ∫ J υ dυ → 4 N 0 F = αN + N e
hυ hυ

8
 σ bf c σ bf 4 c u u
4N0 F = 4 N0 F = 4 N 0σ bf = N 0σ bf c = N 0σ bf cη = αN + N e
hυ 4 hυ c 4 hυ hυ

N0 αN e
=
N + cσ bf η
now look up α and σ bf in tables
N0 6.8 × 10 −13 Ne −6 N 10 −6
= ≈ 4 × 10 ≈
N+ ( )(
3 × 1010 6.3 × 10 −18 η ) η ξ

(b) Prove for hydrogenic ions of species X and nuclear charge Z (atoms that are holding
on desperately to their last electron) that

N x0 10 −6 2 β + 4
≈ z
N x+ ξ

where n X 0 is the number density of hydrogenic ions of species X that each still retain
their last electron, and n X + is the number density of fully stripped ions of species X.
Assume that all of the hydrogenic metal ions are recombining with electrons provided by
hydrogen, and that nearly all of the hydrogen is ionized. We check this last statement in
part (d).

Steps you can take:

Use the Bohr model to understand how ionization energy scales with Z.

Use our dirty derivation of the photo-ionization cross-section,

λ2 A21
σ ≈ ,
8π υ

to argue that σ scales as the radius of the hydrogenic ion squared. Then use the Bohr
model to see how this radius scales with Z.

Finally, once you have determined how σ scales with Z, use the Milne relation to see
how the radiative recombination coefficient scales with Z.

9
 (1) Bohr atomic radius : a ∝ 1 / z
(2) Bound Free crosssecti on:
σ bf ∝ A21λ3
d2
A21 ∝ υ 3 d 2 ∝
λ3
d ( ≡ dipole moment ) ∝ a
1
σ bf ∝ a 2 ∝ 2
z
(3) Ionization Energy: hυ = ze 2 /a → υ ∝ z 2
1
( )
2
(4) Milne: σ fb ∝ σ bf υ 2 ∝ 2 z 2 ∝ z 2
z
 υF 
(5) N x0  υ σ bf ≈ N x+ N eσ fb f (v)v
 hυ 
 υFυ 
( )
−β
−β
  ≡ Number Flux of ionizing photons ∝ Fυ ∝ υ ∝ z
2
∝ z −2 β
 h υ 
 υF 
∴ N x0  υ σ bf ∝ N x0 z −2 β ηz −2 & N x+ N eσ fb ∝ N x+ nH z 2
 hυ 
N x0 z −2 β ηz −2 ∝ N x+ n H z 2
All the constants are the same as in (a) so...
N x0 10 −6 2 β + 4
≈ z
N x+ ξ

(c) Notice that no matter how large ξ is, there are always a few electrons bound to nuclei
at any given moment (if there were none, there would be nothing for photons to ionize,
and photo-ionization equilibrium would be violated). This tiny neutral/hydrogenic
population attenuates the UV-to-X-ray radiation as it tries to propagate through gas. The
gas will have an ionization gradient: at its unshielded face, naked before the radiation,
nearly all the ions will be stripped, but as we move further away from the face, the
ionizing radiation weakens due to increasing absorption by the tiny neutral/hydrogenic
population, and the neutral/hydrogenic fraction grows.

Show that an element of fractional number abundance fx = nX / nH relative to hydrogen

( n X = n X + + n X 0 ) goes from being fully stripped to predominantly hydrogenic over a
column density of hydrogen of order

N H ( Z ) ≈ 10 23 ξz −2 β − 2 f x−1cm −2

10
Assume that such columns are achieved over regions sufficiently geometrically thin that
we can neglect the dilution of the radiation flux by the inverse square law. Again,
consider only ionizations near the ionization edges of various species.

note : N H ( z ) = n H L (number density * depth )

1
τ = n x0σ bf , x L ≈ 1 → n x0 L =
σ bf , x
n x0 nH 1 nH z 2 n H n x+
N H ( Z ) = nH L = nH L = n 0
L ≈ = =
n x0 σ bf , x n x0 σ bf , H n x n x0
x
n x0
z2
= −18
f x−110 6 ξz −2 β −4 ≈ 10 23 ξz −2 β −2 f x−1cm −2
6.3 × 10 cm 2

(d) If β > 1.5, show that the layers of fully stripped carbon, nitrogen, oxygen, and neon
are each thinner than the layer of fully stripped hydrogen. In other words, soft X-rays are
stopped by the metals before the Lyman continuum photons are stopped by hydrogen.
This is a fact of relevance in understanding how gas can be radiatively accelerated to
form quasar winds.

N H ( Z ) ∝ z −2 β −2 f x−1 ∝ Thickness ≡ P
if β = 1.5 P = z −5 f x−1
show that P ( z ) < P (1) = 1
look up f x in tables

carbon z=6 f 6 = 3.3 × 10 − 4 P ( 6 ) = 0 .4

−5
nitrogen z=7 f 7 = 8.3 × 10 P ( 7 ) = 0 .7
oxygen z =8 f 8 = 6.8 × 10 −4 P (8) = 0.04
neon z = 10 f10 = 1.2 × 10 −4 P (10) = 0.08
if β > 1.5 P ( z ) is even smaller

11
 Astro 201 – Radiative Processes – Solution Set 6
Yuki Takahashi and Eugene Chiang

Problem 1. Saha and Redshift of Recombination

Rybicki & Lightman problem 9.4, parts (a), (b), and (c)

Assuming that the partition functions are of order unity, and with λ ≡ h/(2πmkT )1/2 , Saha equation
connecting two successive ionization stages is:

nj+1 e−χj /kT

∼ , or (1)
nj ne λ3

   
nj+1 χj 1
ln ∼− + ln . (2)
nj kT ne λ3

(a) The ionization stage passes from j to j + 1 when nj+1 ∼ nj , so:

 
χj 1 χj
∼ ln ≡ γ =⇒ kT ∼ ≪ χj . (3)
kT ne λ3 γ

(b) The population ratio changes with respect to temperature like:

d ln(nj+1 /nj ) χj
∼ + 3/2 since ne λ3 ∝ T −3/2 . (4)
d ln T kT

[Note: d/d(ln T ) = T d/dT .] The transition temperature is kT ∼ χj /γ (and γ ≫ 3/2), so the ionization
stage changes over a temperature range

 −1
d ln(nj+1 /nj )
∆T ∼ T ∼ T γ −1 ≪ T. (5)
d ln T

(c) For an atom/ion in state j, the ratio of excited to ground state populations is:

ni,j gi,j −χi,j /kT

= e , where χi,j is the excitation potential. (6)
n0,j g0,j

Before being ionized, kT <

∼ χj /γ, so:

ni,j gi,j −γχi,j /χj

<
∼ e . (7)
n0,j g0,j

1
The excitation potential χi,j is of the same order as the ionization potential χj (except for very low-lying
states), so when the electrons are very nondegenerate (γ large), the atom/ion stays mostly in its ground
state (ni,j ≪ n0,j ).

(d) Calculate the redshift, zrec , at which recombination occurred in the early universe. Use the temperature-
redshift relation

Tphoton = T0 (1 + z) (8)

and the relation between baryonic number density and redshift

n = n0 (1 + z)3 . (9)

Here T0 = 2.73 K is the temperature of the cosmic microwave photon gas today, and n0 = 10−7 cm−3 is
the number density of baryons (read: hydrogen atoms) today, grossly averaged over the entire universe.
You may define the epoch of recombination to be when 0.5 of the protons have recombined to form neutral
hydrogen.

(Redshift is a cosmologist’s measure of time. A redshift z = 0 corresponds to today. As one goes back
in time, the redshift increases. Photons that have travelled from an epoch corresponding to a redshift z
have their original wavelengths, λ, stretched to a longer wavelength, λ′ , by the expansion of the universe.
By definition, λ′ /λ = 1 + z.)

(d) Defining the recombination epoch to be when np = nH ,

np 2Up (T ) e−χH /kT e−χH /kT

1= = = (in cgs units). (10)
nH UH (T ) ne λ3 4.1 × 10−16 ne T −3/2

For hydrogen in the ground state, UH (T ) = 2Up (T ). With T = T0 (1 + z) and ne = ne,0 (1 + z)3 =
n0 (1 + z)3 ,

e−χH /kT0 (1+zrec )

1= −3/2
(in cgs units). (11)
4.1 × 10−16 n0 T0 (1 + zrec )3/2

With χH = 13.6 eV and T0 = 2.73 K (kT0 = 0.235 meV), using n0 = 10−7 / cm3 gives a solution
zrec ∼ 1400 (actually less significant figures because a gross average was used for n0 ).

Problem 2. The Orion Nebula

(a) Sketch the flux density, Fν , for the Orion Nebula due to its free-free emission from a wavelength of
λ = 100 µm to λ = 100 cm. Express Fν in Janskys (10−23 in cgs units; see PS 4 if in doubt). Plot the
spectrum on a log-log plot, and indicate power-law indices where appropriate.

Take the density of electrons to be ne = 2 × 103 cm−3 , the electron and proton temperatures to be
T = Te = Tp = 8 × 103 K, the dimension of the ionized cloud to be R = 1 pc, and the distance to the
Orion Nebula to be d = 500 pc. Take whatever geometry (cube, sphere) for the cloud is most convenient.
Please do not forget free-free self-absorption.

For a Maxwellian (thermal) distribution of particle speeds,

 m 3/2 −mv 2
f (v) = 4π v 2 e 2kT
2πkT

free-free radiation is emitted at a rate per volume per frequency of:

1/2
25 πe6

dW 2π hν
= Z 2 T −1/2 e− kT g f f = Eνf f = 4πjνf f
dV dtdν 3mc3 3km

Also assume thermal free-free absorption, so

25 πe6 2π 1/2 hν hν hν
Z 2 T −1/2 ne ni e− kT g f f 22 e6 (2π)1/2 Z 2 ne ni e− kT g f f (1 − e− kT )


jff 3mc3
ανf f = ν = 3km
hν =
Bν (T ) 2hν 3 /[c2 (e kT − 1)] 3mc(3km)1/2 T 1/2 hν 3

Assume pure ionized Hydrogen, Z = 1 and ne = ni =n.

4e6 2π 1/2 −1/2 2

ανf f = ( ) T n (1 − e−hν/kT )g f f ν −3
3mhc 3km

Now, to compare hν ( = hc/λ) to kT for the spectral range of interest:

hc (7 × 10−27 erg · s)(3 × 108 m/s)

≃ ≃ 2 × 10−14 erg
λmin 10−4 m
hc (7 × 10−27 erg · s)(3 × 108 m/s)
≃ ≃ 2 × 10−18 erg
λmax 1m
kT ≃ (10−16 erg/K)(8 × 103K) ≃ 10−12 erg

So, hν << kT for the whole range. This is the Rayleigh-Jeans region of the spectrum. So we can
hν hν
approximate (1-e−hν/kT ) as ≃ (1 − (1 − kT )) = kT . Also assume g f f ∼ 1.

1/2
4e6

2π
ανf f ≃ T −3/2 n2 ν −2
3mkc 3km

αfν f 2kT ν 2
So now we have expressions for both ανf f and jνf f , which gives Sfν f = jνf f
= c2
. With this, we can
use the solution to the radiative transfer equation:

Z τν
Iν (τν ) = Sν (tν )e−(τν −tν ) dtν + Iν (tν = 0)e−τν
0

3
For constant Sν (assume uniform cloud composition) and Iν (tν = 0) = 0 (no background lighting), this
equation simplifies to

Z τν  τν
Iν (τν ) = Sν e −τν
etν dtν = Sν e−τν etν 0 = Sν e−τν (eτν − e0 ) = Sν (1 − e−τν )
0

Rs
For a uniform medium, τν = 0
αν (s′ )ds′ = αν · s, where s is the path length.

2kT ν 2 −4e6 1/2 −3/2 2

3mkc ( 3km )
2π
T n g f f ν −2 ·s
Iνf f (τν ) = Sν (1 − e−τν ) = Sν (1 − e−αν ·s ) = [1 − e ]
c2

But what is “s”? Well, to be careful I would need to calculate it for every line of sight through this
spherical cloud (see below, left).

t=0 s t = tau
observer
The mostly-correct and infinitely easier way.

R The correct way.

s
observer

But I am far too lazy for that, so I will assume that the Orion nebula is one of an exotic new class of
objects discovered by graduate students. Behold the mysterious cubical nebula (above, right). Hey, at
least I conserved volume! s = ( 4π
3
)1/3 R ...

Finally, to turn this specific intensity into a flux, we must multiply by the solid angle subtended by the
2
object from our position. Fν = Iν dΩ. At distance d, dΩ = ds2 . So

νs 2 −4e6 2π 1/2 −3/2 2 −2

Fν (τν ) = 2kT ( ) (1 − e 3mkc ( 3km ) T n ν ·s
),
cd

where d = 500 pc, s = ( 4π 3

)1/3 R, R = 1 pc, n = 2000 cm−3 , T = 8000 K. ν is the only variable.
Everything else is a constant.
  h
erg K pc2 s2 erg
i
The units are = cm2 ·s·Hz . To express in Janskys, divide by 10−23 . See the plot a few
K s2 cm2 pc2
pages from now. The spectral slope is nearly zero on high frequency portion of the spectrum, and is 2
on the low frequency portion.

(b) Overlay on your plot Fν from pure electron-electron scatterings. Here aim only for order-of-
magnitude accuracy. Remember that electron-electron collisions involve time-varying quadrupole mo-
ments, not time-varying dipole moments, and recall the quadrupole vs. dipole scalings from our discus-
sion of Einstein A’s.

For this part, consider the emissivity (jν ) purely from electron-electron scatterings. Drop the emissivity
from electron-proton scatterings. However, for the absorptivity (αν ), consider the contributions from both
electron-proton and electron-electron scatterings. If you think one absorption process is more important
than the other, justify why you think that is the case and proceed by including just the one.

4
Recall from lecture our discussion of the Einstein coefficients: A21,electric quadrupole ∼ A21,electric dipole ×
 s 2
λ
, where λ is the wavelength of the photon created in the interaction, and s is the “characteristic
length scale” of the problem, which I will consider as the average impact parameter and call “b”. The
ratio of Einstein A’s is just the ratio of power emitted by a dipole vs. power radiated by a quadrupole.
Since e-/ion is a dipolar and e-/e- is a quadrupolar interaction, the emission coefficients are related like
the Einstein A’s above:

 2
b
jνee ∼ jνf f ×
λ

And, because this will still be thermal e-/e- absorption as well as emission (just like in the e-/ion case
before), the source function is the Planck function, Bν (Te ) = Bν (T ). Rearranging the definition of the
source function as the ratio of the emission and absorption coefficients gives a relationship between the
e-/e- absorption coefficients that closely resembles that of the e-/ion absorption coefficients:

 2  2
jνee j f f ( b )2 jff b b
ανee = = ν λ = ν ff
= αν ×
Bν (T ) Bν (T ) Bν (T ) λ λ

But what is the average impact parameter, b? Recall our derivation of electron-proton bremmstrahlung
in which we identified each encounter having impact parameter b as generating radiation having a typical
frequency ν ∼ v/b, where v was the velocity of encounter. Then

 2  
b v 2 kT
∼ ∼ ∼ 10−6 (12)
λ c mc2

Does one absorption process—either e-/ion free-free or e-/e- scattering—dominate over the other? Since
(b/λ)2 ∼ 10−6 is very small, e-/ion free-free dominates the absorption. Then we can write

jνee ee +αf f )·s jνf f (b/λ)2 ff

Iνee = Sνee (1 − e−τν ) = ff
(1 − e−(αν ν
)≃ (1 − e−αν ·s
) = 10−6 × Iνf f
ανee + αν ανf f

As before, we multiply by s2 /d2 to convert to Flux. See the second plot below the first one on the next
page.

5
No wonder nobody ever includes it! The entire e-e spectrum is down by 6 orders of magnitude w.r.t.
the usual e-proton spectrum.

Problem 3. Blowing Stromgren Bubbles

Consider a lone O star emitting η Lyman limit photons per second. It sits inside hydrogen gas of infinite
extent and of number density n. The star ionizes an HII region—a.k.a. a “Stromgren sphere,” after
Bengt Stromgren, who understood that such spheres have sharp boundaries—in which hydrogen is nearly
completely ionized.

Every Lyman limit photon goes towards ionizing a neutral hydrogen atom. That is, every photon emitted
by the star goes towards maintaining the Stromgren bubble. Put yet another way, no Lyman limit photon
emitted by the star travels past the radius of the Stromgren sphere.

The rate at which Lyman limit photons are emitted by the central star equals the rate of radiative
recombinations in the ionized gas. The sphere is nearly completely ionized.1 These facts of photo-
ionization equilibrium determine the approximate radius of the Stromgren sphere.

The temperature inside the sphere is about 10000 K. (A class on the interstellar medium can show you
why.)
1
The sphere cannot be 100% ionized because then there would be no neutrals to absorb the Lyman limit photons that
are continuously streaming out of the star.

6
(a) Derive a symbolic expression for the radius of the Stromgren sphere using the above variables and
whatever variables were introduced in lecture.

Every Lyman limit photon is used to make a photo-ionization. The rate of photo-ionizations equals
the rate of radiative recombinations in steady state. Then

4
η = πRs3 np ne α (13)
3

where Rs is the radius of the Stromgren sphere and α is the recombination coefficient discussed in class.
Since the gas is nearly fully ionized, np = ne = n, and therefore the radius of the Stromgren sphere is

 1/3
3η
Rs = (14)
4πn2 α

Knowing that every Lyman limit photon is ultimately absorbed by the sphere, we can say that taking
α = αB (on-the-spot approximation) is a better choice for deciding the global size of the sphere.

(b) What is the timescale, trec , over which a free proton radiatively recombines in the sphere? That is,
how long would a free proton have to wait before undergoing a radiative recombination? Give both a
symbolic expression, and a numerical evaluation for n = 1 cm−3 .

At photo-ionization equilibrium, the rate η of Lyman limit photon emission equals the rate of radiative
recombinations in the sphere (of volume Vs ):

η ≃ αnp ne Vs ≃ αn2 Vs , (15)

where α is the net recombination coefficient for hydrogen (taking into account “on-the-spot” re-ionization
by photons from radiative recombinations into the ground level). At 10000 K, α = 2.59 × 10−13 cm3 / s
(Osterbrock).

A free proton would have to wait before undergoing a radiative recombination about

total # of protons in the sphere nVs 1

trec ∼ ∼ ∼ . (16)
rate of radiative recombinations in the sphere η αn

For n = 1/ cm3 and at 10000 K, trec ∼ 100,000 years.

(c) If the star were initially “off,” and the gas surrounding it initially neutral, what is the timescale for
the Stromgren sphere to develop after the star were turned “on”? That is, how long does the star take
to blow an ionized bubble? Think simply and to order-of-magnitude; you should get the same answer as
(a).

Initially surrounded by neutral gas, the star develops an ionized sphere around it over a timescale
roughly equaling (ignoring radiative recombination):
 # of H atoms to ionize to create the Stromgren sphere nVs 1
∼ ∼ . (17)
rate of Lyman limit photon emission η αn

Problem 4. Time to Relax in the Stromgren Sphere

It is often assumed that velocity distributions of particles are Maxwellian. The validity of this assumption
rests on the ability of particles to collide elastically with one another and share their kinetic energy. For a
Maxwellian to be appropriate, the timescale for a collision must be short compared to other timescales of
interest. This problem tests these assumptions for the case of the Stromgren sphere of nearly completely
ionized hydrogen of problem 3.

(a) Establishing the electron (kinetic) temperature: what is the timescale, te , for free electrons in the
Stromgren sphere to collide with one another? Consider collisions occurring at relative velocities typical
of those in an electron gas at temperature Te . Work only to order-of-magnitude and express your answer
in terms of n, Te , and other fundamental constants.

Establishing the electron (kinetic) temperature: The timescale te for free electrons to collide with one
another is: te ∼ 1/nσv, where σ is the collisional cross-section. For two charged particles to share their
kinetic energy, they must approach each other to within a separation r where their electric potential is
comparable to their kinetic energy: e2 /r ∼ mv 2 . So, σ ∼ r 2 ∼ e4 /(mv 2 )2 . With mv 2 ∼ kTe ,

√
me (kTe )3/2
te ∼ 4
∼ 10 days (for 104 K). (18)
ne

(b) Establishing the proton (kinetic) temperature: repeat (a), but for protons, and consider collisions at
relative velocities typical of those in a proton gas of temperature Tp . Call the proton relaxation time tp .

Establishing the proton (kinetic) temperature: Similarly the proton relaxation time is:

√
mp (kTp )3/2
r
mp
tp ∼ ∼ te ∼ 1 year (for 104 K). (19)
ne4 me

(c) Establishing a common (kinetic) temperature: suppose that initially, Te > Tp . What is the timescale
over which electrons and protons equilibrate to a common kinetic temperature? This is not merely
the timescale for a proton to collide with an electron. You must consider also the amount of energy
exchanged between an electron and proton during each encounter. Estimate, to order-of-magnitude, the
time it takes a cold proton to acquire the same kinetic energy as a hot electron. Call this time tep . Again,
express your answer symbolically.

Hint: you might find it helpful to switch the charge on the electron and consider head-on collisions
between the positive electron and positive proton.

Establishing a common (kinetic) temperature: Consider collisions in which the electron is significantly
deflected by the proton. These occur, to order-of-magnitude, over a timescale te . In such collisions, the
change in the electron’s velocity ∆ve is comparable to the original electron velocity ve . The proton’s
momentum changes by mp ∆vp ∼ me ∆ve by momentum conservation. Therefore ∆vp ∼ (me /mp )ve .
The amount of ENERGY gained by the proton is ∆Ep ∼ mp (∆vp )2 . Each collision imparts ∆Ep and
it takes N ∼ me ve2 /∆Ep number of collisions for the proton’s energy to rise up to the electron’s energy.
Putting it all together, N ∼ mp /me . Therefore tep ∼ Nte ∼ (mp /me )te . Numerically,

mp
tep ∼ te ∼ 60 years (for 104 K). (20)
me

What if instead of considering the energy impulse ∆Ep we considered

p the velocity impulse ∆vp ? Can we
reason that the final equilibrium velocity of
p the proton must be me /m p ve and therefore the number of
collisions required to reach equilibrium is me /mp ve /∆vp ? The answer is no; velocity has direction and
therefore undergoes a random walk. The root-mean-square of the velocity increases not as the number
of collisions but rather more slowly as the square root of the number of collisions. Since the velocity
random walks, the energy walks (regularly), a result that returns us to the derivation above.

(d) Numerically evaluate te /trec , tp /trec , and tep /trec , for Te ∼ Tp ∼ 104 K. Is assuming a Maxwellian
distribution of velocities at a common temperature for both electrons and ions a good approximation in
Stromgren spheres?

For Te ∼ Tp ∼ 104 K,

te 1 tp 1 tep 1
∼ 6.5 , ∼ 5, ∼ . (21)
trec 10 trec 10 trec 2000

Since the timescales for these collisions are shorter than that of radiative recombinations in Stromgren
spheres, the velocity distributions can be approximated Maxwellian at a common temperature for both
electrons and protons.

9
 Astro 201 – Radiative Processes – Solution Set 7

By Julia Kregenow and Eugene Chiang

Readings: Rybicki & Lightman Chapter 5

Problem 1. Seeing the Forest Through the Clouds

Optical spectra of high redshift quasars often exhibit a dense thicket of absorption lines
(see examples from Wolfe et al. 1993; ApJ, 404, 480). Many (but not all) of these
absorption features are thought to be from the same transition: Lyman α (n = 1 to
n = 2) in hydrogen. The absorption lines are located at different wavelengths because
they arise from different hydrogen clouds located at different redshifts between us and the
quasar (for the latter, read: convenient, bright, continuum light source). This so-called
“Lyman α forest” of absorption lines is diagnostic of the degree of clumpiness in gas at
high redshift; i.e., it is diagnostic of structure formation.

In this problem, we relate the observed width of each line to the column density of hy-
drogen in the corresponding “Lyman α cloud.” Most of this problem can be done in
ignorance of cosmology.

Take the Lyα line profile presented by a single cloud to be both Doppler broadened and
naturally broadened. As described by Rybicki & Lightman page 291, the convolved line
profile is described by the Voigt function.

(a) If the “a”-parameter in the Voigt function is much less than one, the Voigt function
can be described in two parts. Define u ≡ (ν − ν0 )/∆νD , where I am using the notation
of Rybicki & Lightman. If u ≪ 1 + a−2 , then

1 2
φ(u ≪ 1 + a−2 ) ≈ √ e−u , (1)
π∆νD

while if u ≫ 1 + a−2 , then

a
φ(u ≫ 1 + a−2 ) ≈ . (2)
π∆νD u2

In other words, the core of the line, near line center, is dominated by the Doppler Gaus-
sian profile, while the wings of the line, far from line center, are dominated by the
Lorentzian profile. You can verify that these expressions are nothing more than (10.68)
and (10.73) taken in the appropriate limits.

Calculate “a” for the Lyα line and verify that it is much less than one. Use a Doppler
width of ∆νD = (10 km/ s)ν0 /c.

1
For the Lyα line that is Doppler broadened [by width ∆νD = (10 km/ s)ν0 /c = 8×1010 Hz
for λ0 = 1216Å], and naturally broadened [Γ = A21 = 6 × 108 Hz],

Γ
a≡ ≈ 6 × 10−4 ≪ 1. (3)
4π∆νD

(b) The equivalent width of a spectral absorption line is, by definition,

∞ Fλ,0 − Fλ
Z
EW ≡ dλ (4)
0 Fλ,0

where Fλ is the actual flux density, and Fλ,0 is the flux density of the spectrum if the
absorption line were not present (i.e., the “continuum” flux density.)

Measurements of the equivalent widths of spectral lines can be used to infer column
densities of intervening material.

First suppose that the Lyα cloud is optically thin in the line, so that the line profile
looks Gaussian (we cannot probe the Lorentzian wings of the line). Derive a symbolic
expression between EW and the column density, NHI , of neutral hydrogen in the cloud,
using whatever fundamental constants you need.

Assuming that the hydrogen is purely absorbing, Fλ = Fλ,0 e−τ (λ) , where τ (λ) is the
optical depth to absorption. Then the equivalent width of a spectral absorption line is:

∞ Fλ,0 − Fλ ∞
Z Z
EW ≡ dλ = 1 − e−τ (λ) dλ. (5)
0 Fλ,0 0

The optical depth depends on wavelength through the absorption cross-section σ(λ):

A21 2
τ (λ) = NHI σ(λ) = NHI λ φ(ν). (6)
8π 0

If the Lyα cloud is optically thin in the line (τ ≪ 1),

∞ A21 λ20 ∞ A21 λ20 c ∞ φ(ν)

Z Z Z
EWthin ≃ τ (λ)dλ ≃ NHI φ(ν)dλ, ≃ NHI dν, (7)
0 8π 0 8π 0 ν2

1 2
and the line profile looks Gaussian: φ(u) ≈ √π∆ν e−u , where u ≡ (ν − ν0 )/∆νD . Now
D
φ(ν) is sufficiently narrow that the ν 2 -denominator in the above integral hardly changes,
so we can take the ν 2 outside as ν02 :

2
 A21 λ20 c ∞ A21 λ20 c A21 λ40
Z
EWthin ≃ NHI φ(ν)dν = NHI = N HI . (8)
8πν02 0 8πν02 8πc

This result is referred to as the “linear portion of the curve of growth.”

(c) Now suppose that the Lyα cloud is so optically thick in the line that we can make
out the naturally broadened Lorentzian wings of the line. That is, the core of the line
is completely black, but far from line center, the flux rises back up according to the
quasi-power-law Lorentzian. Derive a symbolic expression between EW and NHI in this
regime.

If the Lyα cloud is so optically thick in the line that the core is completely black, but
the flux rises back up in the naturally broadened wings of the line (φ(u) ≈ π∆νaD u2 ):

A21 2 aA21
τ = NHI λ0 φ(ν) ≈ NHI 2 λ2 (9)
8π 8π ∆νD u2 0

Notice we have pulled a trick here: we used only the asymptotic Lorentzian shape of
the line, even though we know the line shape in the core is Gaussian, not Lorentzian.
We can do this because the line core is pitch black, a feature that the Lorentzian shape
captures just fine—after all, using the Lorentzian at u = 0, we get infinite optical depth!

Letting α ≡ aA21 λ20 /(8π 2 ∆νD )

2
∞ 1 − e−αNHI /u
Z
EWthick ≃ c dν (10)
0 ν2

Again, the integral is dominated by small u, near line center. Provided the line width
in frequency is still smaller than the frequency (the line is not relativistically broad), we
can again take the ν 2 -denominator outside as ν02 :

c ∞
Z
2
EWthick ≃ 2 1 − e−αNHI /u dν (11)
ν0 0

Now it is a matter of non-dimensionalizing the integral. Note that

∆νD 10 km/ s
ν = ν0 + u∆νD = ν0 (1 + uβ), where β ≡ = , (12)
ν0 c

to write the above integral as

c ∞
Z
2
∆νD 1 − e−αNHI /u du (13)
ν02 −1/β

3
Make a change of variable to x such that x−2 = u−2 αNHI . Then the integral reads

cp ∞
Z
2
∆νD 2 αNHI √ 1 − e−1/x dx (14)
ν0 −1/(β αNHI )

The lower limit on the integral is ≪ −1, as can be verified for any particular problem
(we can verify it for the cases below, if we wish). Then the integral is insensitive to the
√
value of NHI and is approximately 2 π. We conclude that

A21 λ3
NHI √ 0 .
p
EWthick ≈ (15)
8πc

This is referred to as the “square root portion of the curve of growth.” We can derive
roughly the same result more simply by saying that the line is black all the way from line
center until the frequency where τ = 1 (say; you could choose another favorite number
> 1). Use the approximate Lorentzian form of the line profile function to decide at what
frequency τ = 1. Then take the equivalent width to be the area in the black square
well of the line, ignoring the wings of the line. This procedure will yield about the same
result as above.

(d) Clouds in regime (b) are referred to as members of the “Lyα forest.” Clouds in
regime (c) are referred to as “damped Lyα systems”—the “damped” refers to the fact
that we are sensitive to the “naturally damped” Lorentzian wings of the line.

In Wolfe et al.’s paper, we can distinguish Lyα forest clouds having observed EWobs ∼
1Å. Numerically estimate NHI for such clouds, assuming they are located at redshift
z = 3.

Remember that spectral line profiles get stretched with the expansion of the universe,
so EWobs = (1 + z)EWint , where EWint is the intrinsic equivalent width of the line
(the EW you would measure if you were positioned right behind the cloud). All of your
expressions for (a)–(c) pertain to EWint .

Assuming the Lyα forest clouds are at redshift z = 3, the intrinsic EWthin = EWobs /4 ∼
0.25Å.

8πc
NHIthin ≈ EWthin ∼ 1.4 × 1014 / cm2 . (16)
A21 λ40

We can check a posteriori whether our assumption that the cloud is optically thin is
valid or not. For the above value of N ≈ 1014 cm−2 , τ (ν0 ) ≈ 3, which means that our
assumption that this cloud is optically thin is marginally violated.

(e) In the same quasar spectrum, we can also distinguish damped Lyα systems having
EWobs ∼ 100Å. Numerically estimate NHI for such clouds, and compare your answer

4
to the column in our Galaxy, NHI,Gal ∼ 1021 cm−2 . The fact that they are comparable
argues that damped Lyα systems are full-fledged galaxies that happen to lie between us
and the quasar.

Assuming the damped Lyα systems are also at redshift z = 3, the intrinsic EWthick =
EWobs /4 ∼ 25Å.

√ !2
8πc
NHIthick ≈ EWthick ∼ 4 × 1021 / cm2 , (17)
A21 λ30

comparable to the column density in our Galaxy of ∼ 1021 / cm2 .

Problem 2. Protogalaxies

Why do galaxies have the sizes and masses that they do?

Here is a rule of thumb governing the sizes of objects that can collapse under the weight
of their own self-gravity: objects can only collapse if their cooling time is shorter than
their gravitational collapse time. The cooling time of gas is the time it takes to lose order
unity (say, 0.5) of its thermal energy. The collapse time is the time it takes to shrink its
radius by order unity. (This rule of thumb is subject to details regarding the exact cooling
mechanisms; i.e., the effective adiabatic index of collapsing material. Truth be told, the
adiabatic index doesn’t have to be exactly one (isothermal) for collapse to proceed.)

(a) Present an order-of-magnitude derivation, using only bremsstrahlung radiation to

cool off gas, of the maximum radius R of an object that can form out of a virialized
lump of hydrogen plasma. (Recall that if a gas is virialized, its kinetic temperature is
uniquely related to the depth of its gravitational potential well. Gas becomes virialized
as it collapses under its own self-gravity and shocks; the kinetic energy of collapse is
converted efficiently into heat in these shocks.)

Express in kpc.

By the virial theorem, 2 K.E. = -[Link] , so

3 GM 2 GM 2 GM mH
2( N kT ) = ⇒ T (t) = =
2 R 3N kR(t) 3kR(t)

Now investigating how long it takes the gas to cool:

amount of energy 1 E0 1 E0 1 3 N kT0

tcool ∼ ∼ ( ) = ( dE )≃ (2 )
rate of energy radiation 2 dE/dt 2 dt·dV ol · V ol. 2 j · V ol

5
Note this assumes that the cloud is optically thin. Further assuming that Bremstrahlung
is the only cooling mechanism, I use dW/(dV × dt), the frequency-INTEGRATED emis-
sion coefficient from Rybicki & Lightman (5.25) (or 5.15b would have done as well). I
M
assume g B ∼ 1, Z ≃ 1, and ne = ni = n, and substitute N/V ol = n = V ol·m H
.

 
3 s K1/2 )kT 1/2
.5 × 1027 ( erg s K1/2 )kT 1/2 R3 m
2 × 1027 ( erg
1 2 nkT0 cm3 cm3 H
tcool ∼  ≃ =

2 1.4 × 10 (
−27 erg cm 3
1/2
)T n 2

n M
s K1/2

Now investigating the collapse time using the very crude assumption that the gas is in
free-fall, so acceleration a = g. Taking advantage of the spirit of order of magnitude, I
will do dimensional analysis derivation:

d2 r GM R3 R3 1/2
a=g⇒ = ⇒ ∼ GM ⇒ t collapse ∼ ( )
dt2 r2 t2 GM

Now enforcing the condition that tcool < tcollapse ,

s K1/2 )kT 1/2 R3 m

2 × 1027 ( erg cm3 H R3 1/2
<( )
M GM

Squaring both sides and using the definition of T(t) from the virial theorem above,

K s2 )k2 R6 m2 10−54 ( erg cm

2 6
4 × 1054 ( erg6 erg2 H GM mH R3 2 )
( ) < ( ) ⇒ R 2
< Ks
M2 3kR GM m3H G2 k

Plugging in all the constants and converting to pc, we find that R < 6 × 1023 cm
= 200 kpc. What an eminently reasonable spacing between neighboring galaxies!

(b) For an object having such a maximum radius R, present an order-of-magnitude

derivation of its minimum mass M by considering how hot the gas must be to be a fully
ionized plasma.

Express in solar masses.

Getting the mass limit is a little trickier, because M cancels out of the tcool < tcollapse
condition above, so we cannot solve for it. (Indeed, we were only able to solve for R
in the first place because M cancelled out!) So, I will investigate the assumptions that
went into this calculation and see what falls out.

6
Assumption #1: Cloud is optically thin to Bremstrahlung: τ ≪ 1. With a little help
from R.&L. (5.20) for a frequency-averaged free-free absorption coefficient, we have

h i
τ = αfν f · R ≃ 1.7 × 10−25 K7/2 cm5 T −7/2 n2 · R ≪ 1
h i 3M
1.7 × 10−25 K7/2 cm5 T −7/2 ( )2 · R ≪ 1
4πmH R3

Substituting in T as a function of M and R from the virial theorem leaves an expression

with only M and R as variables.

h i 3kR 7/2 2
1.7 × 10−25 K7/2 cm5 ( ) M < 16m2H R5
GM mH

Solving for M, and then plugging in constants and the R we just derived gives

h i
(3k)7/3 (1.7 × 10−25 )2/3 K7/3 cm10/3 )
M> 11/3
≃ 9 × 1026 g ≃ 4 × 10−7 solar masses.
162/3 mH G7/3 R

Um, I’m gonna be bold and say this is not a terribly interesting lower mass limit.

Assumption #2: Cloud is all ionized Hydrogen. How hot does it need to be to be ionized?
Well, we know from Saha’s equation of ionization equilibrium that the average photon
energy should be larger than the binding energy of H’s electron, divided by a logarithmic
phase space factor (we looked at this in a previous problem set!)

hνpeak > 13.6 eV/χ

where χ ∼ 10 is the order of magnitude of all logarithms in the universe (this is known
as Fermi’s law of logarithms). And, using the fact that the Bremstrahlung spectrum
peaks at ν ∼ kT /h, this becomes a limit on Temperature:

kT 13.6 eV 13.6 eV 13.6 eV

h( )> ⇒T > ∼ ∼ 1 × 104 K
h χ χk 10 × 9 × 10−5 eV/K

Now, going back to the virial relation armed with R ∼ 200 kpc = 6 × 1023 cm and a
new lower bound on Temperature,

3kT R 3k(1 × 104 K)(6 × 1023 cm)

M= > ∼ 2 × 1043 g ∼ 1010 M⊙ (Solar Masses)
GmH GmH

7
Now this certainly seems like a more interesting lower limit on mass! It’s within striking
distance of the mass of a Galaxy!! Ah, bliss... This is why I can’t stop taking classes.

Problem 3. ESCAPE: A New Formalism by Sobolev

Imagine a homogeneous sphere of uniform density that emits and absorbs photons of a
certain frequency. Every volume element of the sphere produces a certain number of pho-
tons per second, and these photons are unleashed from each volume element isotropically.
The material is purely absorbing; neglect scattering.

The radial optical depth of the sphere to these photons is τ (integrated over the radius
of the sphere).

(a) What fraction of the photons that are unleashed per second (throughout the entire
sphere) actually escape the sphere per second (unabsorbed)? Express in terms of τ . You
have calculated what is called the “escape probability for a uniform absorbing sphere,”
useful for more technical analyses of radiative transfer.

Consider one small patch on the surface of the gas sphere. The area of the patch is dA.
Choose one arbitrary line of sight (L.O.S.) through the sphere starting at dA, with angle
θ to the equatorial plane (which contains dA). This L.O.S. passes through length L of
cloud material, where L is a function of θ. See figure below left.

L
R L

dA R
equatorial plane
R

The Law of Cosines gives the path length, L, in terms of the cloud radius (R) and θ.
See the figure above right.

R2 + L2 − 2RLcos(θ) = R2 ⇒ L(θ) = 2Rcos(θ)

Assuming a uniform density inside the sphere and vacuum outside, the source function
is constant and there is no intensity entering from outside. τ is defined as τ = αR, so

I(θ) = S(1 − e−τ (θ) ) = S(1 − e−αL(θ) ) = S(1 − e−2Rαcos(θ) ) = S(1 − e−2τ cos(θ) )

Now, to turn this into a FLUX through dA, consider what the patch dA “SEES” looking
into the sphere. dA sees her own celestial sphere of sorts, with herself at the center. Let’s
say she chooses one particular L.O.S. and its corresponding angle and path length, θ and

8
L(θ). We’ll call the emitting patch at the end of that L.O.S. “da”. See below left. To
turn the specific intensity of this emitting patch da into Flux, we simply multiply by the
solid angle subtended by the emitting patch at the point of interest: dΩ = cos(θ)da/L2 .

da

Ld
L L

L sin
dA

R
L

"celestial sphere" for dA anatomy of a ring

adding up Flux from rings

Now let us consider all such emitting patches ’da’ which lie at the same distance L from
dA. They lie in a continuous ring along the intersection of the 2 spheres (depicted above
middle), and the lines of sight to them describe a cone with opening angle 2θ and vertex
at dA. The solid angle subtended by this emitting ring (figure above right) is

daring 2π(Lsin(θ))Ldθ 2πL2 sin(θ)dθ

dΩring = cos(θ) = cos(θ) = cos(θ) = 2πsin(θ)cos(θ)dθ
L2 L2 L2

Now to get the TOTAL flux through dA from all rings, we integrate over θ.

Z π/2 Z π/2 Z π/2

−2τ cos(θ)
Fout = I(θ)dΩ = S(1−e )2πsin(θ)cos(θ)dθ = 2πS (1−e−2τ cos(θ) )sin(θ)cos(θ)dθ
0 0 0

j jR
Now substituting x = 2τ cos(θ) and writing the source function S = α = τ ,

jR
Z 0 x dx πjR
Z 2τ πjR
Z 2τ Z 2τ 
−x −x −x
Fout = 2π (1−e ) cos(θ)dθ = (1−e )xdx = xdx + −e xdx
τ 2τ 2τ −2τ 2τ 3 0 2τ 3 0 0

dE πjR h 2 −2τ −2τ

i
Fout = = 2τ + 2τ e − 1 + e (18)
dtdνdA 2τ 3

This is a flux: the rate that energy escapes per unit area per frequency.

To calculate at what rate energy is “unleashed” inside the sphere, I use the volume
emissivity, j, remembering that it has units of energy per time per volume per solid
angle per frequency: j = dtd(V dE
ol)dΩdν . To get the total energy created in the whole
sphere I multiply by its volume. To account for it radiating out isotropically in all
directions I multiply by 4π steradians. To turn this into a surface flux, I divide by the
surface area of the sphere.

9
 4
dE V ol × [Link] πR3 × 4π 4πjR
Ftot = =j× =j× 3 2
=
dtdνdA [Link] 4πR 3

This is the flux of photons that would pass the sphere’s surface if there were no absorption
or scattering. Now, taking the ratio of the two,

πjR
2τ 2 + 2τ e−2τ − 1 + e−2τ
 
Fout 2τ 3 3 h 2 −2τ −2τ
i
= 4πjR
= 2τ + 2τ e + e − 1 (19)
Ftot 3
8τ 3

(b) Explain why your answer in (a) makes sense in the limits that τ ≪ 1 and τ ≫ 1.

You may find it helpful to solve (a) to order-to-magnitude in these limits before trying
to attack (a) in full generality.

Now, evaluating this expression in its 2 extreme limits: very large and very small τ.

For τ ≫ 1, I expect that no photons will get out. They should all be trapped if the
sphere is infinitely thick, so the fraction should go to zero.

3 h 2 i 3 h 2 i 3 6 3
For τ → ∞, 3
2τ + 2τ e−2τ
+ e−2τ
− 1 ≃ 3
2τ − 1 ≃ 3 ·2τ 2 = ≃ → 0.
8τ 8τ 8τ 8τ 4τ

Indeed it goes to zero. But there is a bit more information to be interpreted. It goes
to zero roughly as 1/τ . This makes sense because in an optically thick sphere, the
photons that escape are those photons that are emitted within optical depth unity of
the surface. All the other photons are unleashed deep in the sphere, and there they are
releashed (buried). Therefore the fraction of photons that escape equals the volume of
a shell of optical depth unity, divided by the volume of the entire sphere. This equals
4πR2 ∆R/[(4π/3)R3 ] = 3∆R/R. But αR = τ and α∆R ∼ 1. Then we conclude, to
order-of-magnitude, that the escape probability equals 3/τ , which matches our exact
answer to order-of-magnitude. (With hindsight, we could choose α∆R = 1/4 to get the
answers to match exactly).

For τ ≪ 1, I expect that all of the photons will escape, so the fraction should go to 1.
Now both denominator and numerator go to zero in (19). So we use L’Hopital’s rule
and differentiate top and bottom, getting

10
 Fout 3 h i 3 h i
For τ → 0, → 2
4τ + 2e−2τ − 4τ e−2τ − 2e−2τ = 2
4τ − 4τ e−2τ
Ftot 24τ 24τ

which still runs into problems, so we apply the rule again and get

Fout 3 h i
→ 4 − 4e−2τ + 8τ e−2τ
Ftot 48τ

which still runs into problems, so we apply the rule again and get

Fout 3 h −2τ i
→ 8e + 8e−2τ − 16τ e−2τ = e−2τ − τ e−2τ → 1 − 3τ
Ftot 48

which most certainly goes to 1 as τ → 0 (and it is always less than one, as it had better
be!)

11
 Astro 201 – Radiative Processes – Solution Set 10

Eugene Chiang

NOTE FOR ALL PROBLEMS: In this class, we are not going to stress over the sin α
term in any expression involving synchrotron emission, nor are we going to worry about
other factors of order unity, like gamma (Γ) functions. If you see a sin α or gamma
function in your travels, just set it equal to 1.

Problem 1. Leaving on a Jet Frame1

Recall the images of relativistic jets shown in class, and how they appeared brighter on
one end than the other.

Consider a bipolar (two-sided), symmetric jet of material moving at BULK velocity v

and having BULK Lorentz factor γ. In the rest frame of the jet, the jet axis points at
an angle θ relative to our line-of-sight. One side of the jet is directed towards us, and
the other side is directed away.

The jet is composed of electrons (possibly also positrons; the situation seems unclear) that
radiate synchrotron emission. In the rest frame of the jet, the electrons are still gyrating
relativistically, and in that same rest frame, they emit a specific intensity Iν ∝ ν α between
frequencies ν1 and ν2 . For this problem, take α < 0.

(a) Sketch the specific intensity of the jet that points toward an observer on Earth, as
seen by that observer.

Overlay on this sketch the specific intensity of the counter-jet (again, as seen by the
Earth-bound observer).

Overlay also the specific intensity of either jet or counter-jet as measured in its rest
frame. Compare in particular the rest-frame jet spectrum with the counter-jet spectrum
observed at Earth.

As usual, annotate your sketch with scales and indices, expressed symbolically in terms
of the variables given above.

For the jet pointing towards us: the spectrum in the jet rest frame is a power law
from ν1 to ν2 . In the observer frame, every frequency gets Doppler shifted by D+ =
γ(1 + β cos θ) > 0, where β = v/c > 0 (RL equation 4.12b). The factor of 1 + β cos θ
is the usual Doppler shift (train-whistle effect), while the factor of γ arises from time
dilation. So the whole forward jet spectrum shifts to higher frequencies, between D+ ν1
and D+ ν2 .
1
I am indebted to Julia Kregenow for naming this problem.

1
 What about the height of the spectrum? Use the Lorentz invariant Iν /ν 3 . Denote
by I˜ν the specific intensity of the jet seen in the observer’s frame. Then by Lorentz
invariance:

I˜ν (D+ ν) Iν (ν)

3
= (1)
(D+ ν) ν3

implying I˜ν (D+ ν) = D+ 3 I (ν). So the forward jet spectrum gets “Doppler boosted” in
ν
intensity by a factor of D+3 (which can be  1). To sum up: the forward jet spectrum is a

power-law that extends from D+ ν1 to D+ ν2 , whose amplitude at D+ ν1 equals D+ 3 I (ν ),

ν 1
3
and whose amplitude at D+ ν2 equals D+ Iν (ν2 ). The spectral slope remains equal to α.

The answer for the counter-jet is the same, only now the Doppler factor equals
D− = γ(1 − β cos θ). Note that D− can still be  1, so that the emission from the
counter-jet still gets Doppler boosted to higher frequency and higher intensity (just not
quite as high as for the forward jet).

(b) By what (symbolic) factor is the forward jet brighter than the counter-jet in specific
intensity at a fixed frequency?

Both jet and counter-jet spectra are power laws with the same slope. Therefore we
can look at any single frequency and find the ratio of specific intensities. Let’s choose
ν = D+ ν1 . The forward jet’s specific intensity at ν = D+ ν1 equals D+ 3 I (ν ). The
ν 1
counter-jet’s specific intensity at ν = D+ ν1 equals D− Iν (ν1 ) × [(D+ ν1 )/(D− ν1 )]α , where
3

the correction factor arises because we have to scale from ν = D− ν1 to ν = D+ ν1 , and

we know that the counter-jet spectrum scales as ν α . Divide the forward jet specific
intensity by the counter-jet specific intensity to find the desired ratio:

3−α 3−α
D+ 1 + β cos θ
 
= (2)
D− 1 − β cos θ

which does not depend explicitly on γ.

(c) The dynamic contrast in optical surface brightness between jet and counter-jet in
M87 exceeds 450. If α = −1/2, what are the constraints on γ, v, and θ?

Use the answer in part (b). Define x = β cos θ and y = 450 to write

7/2
1+x

>y (3)
1−x

which can be solved for

2
 y 2/7 − 1
1 > x = β cos θ > = 0.7 (4)
y 2/7 + 1

The maximum of β is one, so cos θ > 0.7 (the jet can’t point too far away from the
head-on direction). For a head-on p
jet, cos θ = 1 and β > 0.7. So 1 > β > 0.7, with a
corresponding constraint on γ = 1/ 1 − β 2 > 1.4.

Problem 2. In Space, No One Can Hear You Scream

Take a look at Figure 3, Table 3, and Figure 4 of Carilli et al.’s paper on the radio galaxy
Cygnus A.

In Figures 4a and 4b, we can distinguish a break in the spectral index at about ν ≈ 104
MHz. The break is interpreted to represent the onset of synchrotron losses, as we studied
in a previous problem. The corresponding electron energy spectrum will also be broken.

In either regime, provided we don’t look at frequencies too low (lower than ν ≈ 102.75
MHz), the emission arises from optically thin material.

In Figure 3, one arcsecond equals one kpc.

Part (d) is independent of the other parts.

(a) Estimate the minimum magnetic field strength in “Hot Spot A” using Burbidge’s
minimum energy argument and using Figure 3, Table 3, and/or Figure 4, or any other
raw data from Carilli et al.’s paper. Express in Gauss.

You will have to decide whether to use the spectrum below the break, above the break, or
both. We want a minimum estimate of the magnetic field.

Interpret the unit of “(beam)−1 ” (per beam) to mean [π(2.25 arcseconds)2 ]−1 . This is the
angular area of the radio interferometer’s beam, or footprint on the sky. When Carilli
et al. cite “4.5 arcseconds resolution,” I interpret the 4.5 arcseconds to be equal to the
FWHM (full-width-half-maximum) of their beam.

Burbidge’s minimum energy argument relies on a measurement of jν , the volume

emissivity (erg/s/cm3 /Hz/sr). We can measured jν for optically thin media, and we can
certainly measure it given the data from Carilli et al. The first question is: at what
frequency should we measure jν ?

The frequency of choice is the frequency at which the electrons carrying the bulk of
the (kinetic) energy density are radiating. We stated in class that in many situations,
the bulk of the energy density is carried by the lowest γ electrons, because they are
overwhelmingly the most numerous. This is true whenever p (defined such that dN/dE ∝
E p ) is less than -2. What is p in the hot spot? Synchrotron theory relates p to the
spectral index, α (defined such that Fν ∝ ν α ), via

3
 1+p
α= . (5)
2

Now we can measure α for the hot spot. In Table 3, the spectral slope between 1350
MHz and 4995 MHz—the portion of the spectrum at frequencies below the break—is
α = −0.73, which implies p = −2.4 < −2. Therefore yes, the bulk of the energy density
(∝ (dN/dE) × dE × E ∝ E 2+p ) is carried by electrons having the lowest individual
energies E (or γ) that radiate at the lowest frequencies (γ 2 ωcyc ). From Figure 4c, the
lowest frequency for which the power-law extends is about ν1 = 102.8 MHz. That is the
frequency at which jν should be measured to obtain a lower bound on the total energy
density. (Note that we need not worry about the spectrum at frequencies higher than
the break because the spectrum there is still steeper than the spectrum at frequencies
lower than the break.)

Let’s go ahead and measure jν at ν1 . Now

Lν
jν ∼ (6)
V

where V ∼ (4/3)πr 3 is the volume of the hot spot. Looking at Figure 3, I’d say the hot
spot was a sphere about 5 arcseconds in radius, or r = 5 kpc. What is the luminosity
density, Lν , of this luminous sphere? (Note that density here refers to per Hertz, not
per volume!) Define d to be the distance to the source and Fν to be the flux density of
the source. Then

Lν ∼ 4πd2 Fν (7)

for an isotropic emitter. The flux density at ν1 is Fν ∼ 102 Jy/beam × ∆Ω, where
∆Ω is the solid angle subtended by the hot spot. Measured in beams, we estimate
∆Ω ∼ π(5 arcsec)2 /π(2.25 arcsec)2 ∼ 4 beams. So Fν ∼ 400 Jy. Since 1 arcsecond = 1
kpc, the distance to the source is d = 2 × 105 kpc. Then putting it all together,

jν ∼ 8 × 10−34 (cgs) (8)

This estimate is about 1 order of magnitude lower than curve 1 (labelled “Hot Spot A”)
of Figure 5a. I am not sure why the values are discrepant, but I suspect it has partly
to do with the definition of “beam” (which Carilli et al. never define) and partly due
to the modelled dimensions of the hot spot. It is possible that Carilli et al. are using a
smaller volume for the hot spot. In any case, logic is more important than numbers for
classwork, so we will proceed with our simpler and cleaner estimate.

Now we have to fill in the coefficients for Burbidge’s minimum energy arguments as

4
outlined in lecture. Start from first principles: the volume emissivity equals, to order-
of-magnitude,

1 1
jν ∼ ne × Psingle electron × × (9)
4π ν

where

4 B2
Psingle electron =σT cγ 2 (10)
3 8π
γ 2 ωcyc
ν= (= ν1 ) (11)
2π

and ne is the number density of electrons having γ. Burbidge’s argument is based on

minimization of total energy density

ptot = pe + pmag (12)

where the electron pressure pe and the magnetic pressure pmag are given by, respectively,

pe ∼ ne γmc2 (13)
B2
pmag = (14)
8π

Since Burbidge’s argument is based on minimization of total energy density given a

single measurement of jν , we must re-write jν in terms of pmag and pe . This was done
in lecture without the coefficients. All the coefficients are here now before us, as given
above. The same algebra that gave us the proportionality in lecture now gives us the
order-of-magnitude expression

jν ∼ pe p3/4
mag ν1 C (15)
−1/2

2σT
C= √ = 6 × 10−13 (cgs) (16)
3 2πemc(8π)1/4

We performed the minimization in lecture and found that if we want to minimize ptot
given jν , then pmag ∼ pe ∼ ptot /2, where

5
 1/2
!4/7
jν ν1
(min) ptot ∼2 (17)
C

Putting it all together, pmag ∼ 3.3 × 10−10 dyne/ cm2 , which means B ∼ 1 × 10−4 G, or
100 µG. This is a factor of 3 lower than Carilli et al.’s declaration (without derivation)
of 300 µG. The discrepancy can be traced mostly to the discrepancy in the estimate for
jν ; see above.

(b) Estimate the minimum total pressure—magnetic plus material (electron)—in hot spot
A. Express in cgs units. Compare to the pressure of air around you.

From part (a), ptot ∼ 7 × 10−10 dyne/ cm2 , compared to atmospheric pressure of
106 dyne/ cm2 . In space, even at the end of a relativistic jet, no one can hear you
scream.

(c) Estimate the total energy (magnetic plus material) in hot spot A in ergs.

The total energy is ∼ptot V ∼ 1 × 1058 erg. And that’s just for the hot spot, let alone
the rest of the jet. Accretion onto a supermassive black hole at the jet origin is thought
to give rise to such prodigious energies.

(d) Estimate an age of the hot spot given the observation of the spectral break. Compare
to the dynamical time of the jet (the time it took the jet to grow to its present size).

The break occurs at ν ≈ 104 MHz. The synchrotron lifetime of electrons is

γmc2
tlife ∼ (18)
(4/3)σT cγ 2 pmag

We use ν = γ 2 eB/2πmc ≈ 104 MHz and our estimate of B from part (a) to estimate
γ ∼ 6 × 103 . Then tlife ∼ 5 × 105 yr.

The dynamical time is tdyn ∼l/c, where l is the length of the jet and we assume that
it’s moving at very nearly the speed of light.2 Lengthwise, it’s about 6 seconds of time
in right ascension = 90 arcseconds of angle = 90 kpc = 3 × 105 light years. The true
length might be a factor of few longer than this, because we only see the projection of
the jet in the plane of the sky. In any case, a dynamical age of 3 × 105 yr is hearteningly
comparable to our estimate of the age of the source based on the synchrotron break,
tlif e .

Problem 3. Galactic Synchrotron Emission

2
Actually, as far as I can tell from the introduction of Carilli et al., we don’t actually know the
velocity of the jet. The velocity is INFERRED based on the synchrotron break, assuming equipartition
field, and the requirement that tlife < tdyn ! This is an example of extreme faith in models.

6
The brightness temperature of synchrotron emission in the Galactic plane is measured
to be

ν
 −2.8
Tb = 250 K, (19)
480MHz

valid for 8 GHz > ν > 480 MHz. This emission arises from cosmic ray electrons gyrating
in the Galactic magnetic field of strength B ∼ 3µG. Take the size of the emitting region
to be ∼10 kpc.

(Aside from being a nuisance foreground for CMB researchers, Galactic synchrotron
emission is thought to probe the supernova rate in the Galaxy, since cosmic ray electrons
and protons at the relevant energies are thought to be accelerated in supernova shock
waves. In addition, a famous “FIR-radio correlation” is observed to exist for normal
spiral galaxies; the radio traces electrons accelerated by supernovae, while the far-infrared
(FIR) emission traces warm dust grains. Where there are supernovae, there is active
star formation; where there is star formation, there are dust grains warmed by the ISRF.
Or so goes the traditional interpretation of the FIR-radio correlation.)

(a) Provide an approximate expression for the differential energy spectrum of cosmic ray
electrons. Express in units of particles m−2 sr−1 s−1 GeV−1 . Indicate the approximate
range of energies (Emin and Emax ) for which your expression is valid.

We solve this in three steps. The first step is estimate the number density of electrons,
ne (γmin ) responsible for emission at ν = νmin = 480 MHz. The second step is to calculate
the desired differential energy spectrum at electron energy Emin using ne (γmin ). The
third step is to calculate the slope of the differential energy spectrum.

First ne (γmin ). By definition of the brightness temperature, and for optically thin
media,

2kTb
∼ jν L (20)
λ2

where L ∼ 10 kpc and

( 34 γ 2 β 2 UB σT c) × ne (γ)
jν ∼ (21)
ν × 4π

The term in parentheses in the numerator is the total (bolometric) power radiated by
a single electron, ne (γ) is the number density of electrons having Lorentz factor γ (in a
logarithmically sized interval about γ), the ν in the denominator is thrown in because
we are interested in jν (per Hertz), and the 4π is thrown in because jν is per steradian
and we assuming isotropic emission from a tangled field.

7
 Now ν ∼ γ 2 νcyc ∼ γ 2 eB/(2πme c). Also, for synchtrotron radiation, β ≈ 1. We
evaluate (21) at ν = νmin = 480 MHz, which corresponds to γ = γmin and Tb = 250 K.
Solve for ne (γmin ):

2
2kTb νmin 4πνcyc
ne (γmin ) ∼ × 4 B2 (22)
c2 ( 3 8π σT c) × L

We set B = 3µG, as given. After the arithmetic clears, we find ne (γmin ) ∼ 7 ×

10−12 cm−3 . (This is awfully small density. Most electrons in the universe are not
of this high energy variety.)

Now the second step. Geometrical and dimensional considerations give

dNe ne (γmin ) × c
(Emin ) ∼ (23)
dE 4π × γmin me c2

The left-hand side is the desired differential energy spectrum for the electrons (subscript
“e”), evaluated at electron energy E = Emin , in units of electrons per m2 per s per sr
per GeV. The c in the right-hand side numerator is thrown in to get a flux, because a
number density times a speed equals a flux, and the speed of these synchrotron emitting
electrons is the speed of light. The 4π is thrown in because this electron flux is isotropic
(because we are assuming a tangled field, and an isotropic distribution of pitch angles),
and because the desired dNe /dE has units of per sr. The γmin me c2 is thrown in because
the desired dNe /dE is a differential spectrum, i.e., it measures number of electrons per
energy interval.

The only quantity we need to calculate in the above expression is γmin . That’s easy:
2 ν
recall that ν = νmin ∼ γmin 3
cyc . Solve for γmin ∼ 8 × 10 . Then (23) yields

dNe
(Emin ) ∼ 42 m−2 sr−1 GeV−1 s−1 (24)
dE

where Emin = γmin me c2 = 4 GeV.

Finally, the third step is to calculate the slope. The measured flux density Fnu =
2kTb /λ2 ∝ ν−2.8 ν 2 ∝ ν −0.8 . By definition of the spectral index α defined in class,
α = −0.8. Now we also saw in class (and in the reading) that α is related to the electron
energy spectral index by α = (1 + p)/2, where dNe /dE ∝ E p . So p = −2.6 and

dNe E
 −2.6
∼ 42 m−2 sr−1 GeV−1 s−1 (25)
dE 4 GeV

8
This answer is good between Emin = 4 GeV and Emax = 16 GeV (the latter is calculated
using νmax = 8 GHz.

(b) Compare your answer to the differential energy spectrum of cosmic ray protons, as
given in the reader on page 145. The units of the plot on page 145 were accidentally
chopped off; they are the same as those specified in part (a). By what factor do cosmic
ray protons outnumber cosmic ray electrons?

At E = 4 GeV, we compare our derived answer of 42 electrons (per everything) with

the plot from the reader which gives about 300 protons (per everything). At E = 16
GeV, we get 1 electron (per everything) vs. 20 protons (per everything). So it seems
the cosmic ray electrons are outnumbered by cosmic ray protons by a factor of 8–20, at
fixed energy.

Textbooks like Shu and on-line sources quote a factor of 100 instead. So it seems our
little calculation is off by a factor of 10. Perhaps the discrepancy can be partly resolved
by restoring the various factors of order unity that we have ruthlessly discarded, and
partly by setting the true size of the emitting region L to be larger than 10 kpc, say
closer to 30 kpc.

(c) Estimate the gyro-radii of cosmic ray electrons and of cosmic ray protons at Emin
and Emax . Compare to the size of the (baryonic disk of the) Galaxy, 10 kpc. From this
comparison you should be able to understand why cosmic ray astronomy is so difficult.

We know from class (and the reading) that the gyro-frequency ωB = ωcyc /γ. The
gyro-radius is just v/ωB = c/ωB = γmc2 /(eB). Notice this expression cares only about
the total energy of the charged particle, γmc2 , regardless of whether it is an electron or
a proton. So at Emin = γmin mc2 = 4 GeV, both the electron and proton gyro-radii are
4×1012 cm. At Emax = 16 GeV, both the electron and proton gyro-radii are 1.6×1013 cm
(about an AU). These gyro-radii are much smaller than object-to-object distances in the
Galaxy (measured in parsecs to kiloparsecs), rendering cosmic ray astronomy difficult
because the cosmic rays spewed out by some source (like a supernova or a magnetar)
don’t travel on straight lines to us (unlike photons).

Only when the gyro-radius exceeds the distance to the source is the curvature negli-
gible. This becomes possible at, say, E ∼ 109 GeV (near the “ankle” in the cosmic ray
proton spectrum shown in the reader), when the gyro-radius is ∼1 kpc. So to locate
astrophysical sources of cosmic rays, we must build detectors sensitive to those extreme
energies.

(d) Why do we ignore the contribution from cosmic ray protons to the Galactic syn-
chrotron emission at the above frequencies ν? Provide a quantitative answer.

Let’s say

9
 dNp dNe
=F (26)
dE E dE E

where F is the factor by which protons outnumber electrons at fixed energy E. We

derived F ∼ 10, but more careful treatments seem to give F ∼ 100, as noted in part
(a). Let’s say that F is a constant, and that both the protons and electrons obey
dN/dE ∝ E −2.7 (splitting the difference between the reader which gives p = −2.75 for
the protons and our answer in part (a) which gives p = −2.6 for the electrons). So
the shapes of both distributions are the same; it’s just the normalizations which are
different.

If F > 1, why do we neglect the contribution of protons to the observed emission?

Let’s compute the contribution to jν from electrons, and compare to protons. From
electrons,

ne (γe )σT,e cUB γe2

jν,e ∼ (27)
ν

So at fixed frequency ν,

jν,p np (γp )γp2 σT,p

= (28)
jν,e ne (γe )γe2 σT,e

Now σT ∝ m−2 . So

!2
jν,p np (γp )γp2 me
= (29)
jν,e ne (γe )γe2 mp

At fixed ν, we have ν = γp2 νcyc,p = γe2 νcyc,e , which implies (γp /γe )2 ∝ mp /me . So

!1
jν,p np (γp ) me
= (30)
jν,e ne (γe ) mp

Finally, np (γp ) ∝ (dNp /dE)|Ep Ep , so np (γp )/ne (γe ) = [(dNp /dE)|Ep /(dNe /dE)|Ee ] ×
(Ep /Ee ) = F (Ep /Ee )−2.7 × (Ep /Ee ) = F (Ep /Ee )−1.7 . Now Ep /Ee = (γp mp )/(γe me ) =
(mp /me )3/2 from the line immediately preceding (30). So np (γp )/ne (γe ) = F (me /mp )2.55 .
Therefore

!3.55
jν,p me
=F ∼ 2 × 10−10 (31)
jν,e mp

10
for F = 100. So the proton contribution at fixed ν is utterly negligible compared to the
electron contribution (a repeated theme throughout this class).

11
 Astro 201 – Radiative Processes – Solution Set 9

Joe Converse and Eugene Chiang

NOTE FOR ALL PROBLEMS: In this class, we are not going to stress over the sin α
term in any expression involving synchrotron emission, nor are we going to worry about
other factors of order unity, like gamma (Γ) functions. If you see a sin α or gamma
function in your travels, just set it equal to 1.

Problem 1. Synchrotron Losses

(a) Obtain an analytic expression for the energy of a single relativistic electron as a
function of time, E(t), taking into account its energy loss by synchrotron radiation. Your
expression should contain only the variables E(0) (the initial energy of the electron), B
(the magnetic field, here held fixed with time, following the rest of the world, though one
should worry in general about the field changing with time just as the electron energy
spectrum changes with time), time t, and fundamental constants. Assume sin α = 1 (the
electron pitch angle is 90 degrees) for simplicity.

For (synchrotron) problems of interest to us, the electron always remains relativistic. It
merely evolves from a large γ  1 to a smaller γ  1.

We know what the power emitted is:

dE 4
P = = − σT cγ 2 β 2 uB (1)
dt 3

where the minus is because the energy is being lost. Since γ  1 always, β ≈ 1 always.
So we have

dE 4
= − σT cγ 2 uB (2)
dt 3

E
where γ = me c2 . Solving this differential equation by separation of variables we get

Z E dE 4σT uB
Z t
2
=− dt (3)
E0 E 3m2e c3 0

where E0 is the initial energy. Integrating this and multiplying the result by E0 gives

E0 t
− 1 = , where (4)
E τ

1
 3m2e c3 γ 2 m2 c4 E02 E0
τ= = 4 0 2e = = (5)
4σT uB E0 3 σT cγ0 uB E0
P E
0 0 P0

which is the synchrotron lifetime (all subscript 0’s refer to the value at t = 0). Solving
for E we find

E0
E(t) = (6)
1 + τt

(b) How can you reconcile the loss of energy of the electron with the bald statement of
Rybicki & Lightman on page 168 that “γ is constant”?

Rybicki & Lightman say on page 167 that ∂t ∂ ~ = 0, and from there
(γme c2 ) = q~v · E
conclude that γ=constant because E ~ = 0. They are referring to the fact that the
external E-field is 0. The electron, however, also feels its own E-field, and so the total
~ 6= 0, and hence γ is not actually constant. It is, however, very nearly constant over 1
E
gyro-period (except near ultrastrong magnetic fields, ∼1015 G), and the dynamics over
1 gyro-period is all that concerns RL on page 167.

The electron feeling its own E-field is referred to as “radiation reaction.” All syn-
chrotron radiation is a consequence of radiation reaction.

(c) We have made arguments in class that power-law distributions of electrons in astro-
physical sources are maintained against synchrotron losses by continuous energization by
central engines (a.k.a. injection). The injection (input) spectrum of electrons is modified
by synchrotron losses to produce a steady-state (output) distribution.

Call η(E, t) = dN/dE the differential energy spectrum of electrons as discussed repeatedly
in class. Continuity of electrons in energy space reads

∂η(E, t) ∂
+ [Ėη(E, t)] = I (7)
∂t ∂E

where I is the rate of injection of electrons with some input distribution and Ė is the
rate of energy loss of a single electron by synchrotron radiation. This equation should
not mystify you; it merely describes how the number of electrons in a given energy bin
changes with time, taking into account a flux divergence (the second term on the left-
hand-side) and a source term (the right-hand-side).

We have assumed in class a steady-state distribution of electron energies for which η ∝

E p . Given p, how must I scale with E? Give only the scaling and forget about the
numerical coefficients.

2
As with most scaling problems, you don’t have to solve anything in detail. Ruthlessly
work to order of magnitude.

From the definition of I we have

∂η ∂ Ė ∂η
I ∝ Ė +η + Ė (8)
∂E ∂E ∂E

∂η
and we know η ∝ E p , so ∂E ∝ E p−1 . For Ė we have

E0
Ė = − (9)
τ (1 + τt )2

t E0
and from part (a) we know that 1 + τ = E . Then

2
E0 E E2

Ė = − =− ∝ E2 (10)
τ E0 E0 τ

∂ Ė
and ∂E ∝ E. Therefore

I ∝ E 2 E p−1 + E p E + E 2 E p−1 ∝ E p+1 (11)

(d) Electrons having a given energy must wait a characteristic time before synchrotron
losses become important. Before this time elapses for all such electrons, how does η scale
with E?

Before losses become important, t  τ , and so E ≈ E0 and Ė ≈ − Eτ0 ∝ E. Then we

have

∂η
I ∝ E p+1 ∝ E +η (12)
∂E

which gives that η ∝ E p+1 . In other words, it looks just like the injected spectrum.

An easier way to find the same result is to set the second term on the LHS of equation
(7) equal to zero, appropriate to the case of zero energy loss. Then η = It + η0 , and if
η0 = 0, then η ∝ I ∝ E p+1 . The spectrum looks just like the injected spectrum in slope,
but its amplitude rises linearly with time.

(e) The spectral index (α = d ln Fν /d ln ν) of radiation in a fixed frequency range from

a radio jet flattens with increasing distance from the central galaxy. That is, α = −0.5
at the remote edge of the jet (the “hot spot”), and α = −1 closer in. Given your

3
understanding in (c) and (d), where are the “freshest” electrons located, i.e., those newly
injected into the energy spectrum? Are they at the end of the jet, or are they closer in?
In other words, where is the principal site of particle acceleration?

Since α = 1+p2 , for the hot spot (α = −0.5), p = −2, and farther in on the jet
(α = −1), p = −3. Hence the hot-spot (where the spectrum is flatter) looks more like
the injected spectrum, and so it is where the ”young” elections are. It is when the jet is
stopped at the lobe that most of the acceleration occurs.

(f ) Sketch several profiles of η vs. E at various times, assuming I is constant in time.

Since the synchrotron lifetime goes as E0−1 , the higher-energy electrons lose their energy
first. Initially the spectrum is ∝ E p+1 (upper-left panel), but as time goes on, the high-
energy end steepens to be ∝ E p (upper-right panel). The point of steepening moves to
lower energies as time goes on (lower-left panel), until eventually the entire spectrum
reaches the steady state of being ∝ E p (lower-right panel).

4
log(η)
log(η)

log(E) log(E)

Problem 2. Great Balls of Relativistic Fire

Rybicki & Lightman Problem 4.1 (Don’t worry if you can’t reproduce the answer to within
a factor of 2; the answer is meant to be a rule of thumb.)

5
Since it is optically thick, we will only see light from the nearest side. Even then, because
of beaming, only those parts of the object within an angle of about γ1 will send a signal
to us. This gives the geometry shown in the figure below.

1
γ

R
R

y
1/γ
x

R
The observed time delay is c∆t > x. The geometry gives us y = R sin γ1 ≈ γ. Then we
see that x = y tan γ1 ≈ R 1 R
γ γ = γ 2 . So therefore

R < γ 2 c∆t (13)

Problem 3. Superluminal Motion, or Photons Chasing Photons

Rybicki & Lightman Problem 4.7

The geometry for this problem is very similar to that of problem 2. In this case, in a
time dt, the light coming straight down at us travels a distance of cdt, while the material
going along the jet at an angle θ (instead of γ1 above) travels vdt. The change in apparent
position on the sky is ∆x = vdt sin θ, and the distance it traveled parallel to our line of
sight is vdt cos θ. The apparent difference in distance between the original light pulse
and one emitted after dt is cdt − vdt cos θ = c∆t. Finally,

6
 ∆x vdt sin θ v sin θ
vapp = = = (14)
∆t vdt
dt − c cos θ 1 − vc cos θ

For v = 0.99c and θ = 0.1, vapp = 6.6c, and hence it can appear to move superluminally.
The maximum apparent speed occurs when

2
dvapp v cos θ − vc
0= = (15)
dθ (1 − vc cos θ)2

q
v2
which is true when cos θ = vc . At this angle, sin θ = 1− c2 = γ1 , and

v
γ vγ 2
vapp,max = vv = = γv (16)
1− cc γ

(c) Plot vapp /c vs. θ for γ = 102 . Does the viewing angle θ need to be especially small
for superluminal motion to be perceived?

For culture: see the beautiful illustration of superluminal motion in the optical M87 jet
by Biretta in the accompanying .jpg on the class website.

For γ = 102 , v = 0.99995c. As the plot below shows, almost all viewing angles below
θ = π2 exhibit superluminal motion if the jet is moving fast enough (dotted line is
vapp
c = 1).

7
 100

80

60
vapp/c

40

20

0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
θ

8
 Astro 201 – Radiative Processes – Solution Set 11

by Eugene Chiang

Readings: Article by Kellerman; Rybicki & Lightman 7.1, 7.2, 7.4, as much as you like
of 7.6, 7.7; however much of Blandford’s lecture notes but watch for errors.

Problem 1. Compton Catastrophe

We derived in class that for a synchrotron-self-Compton (SSC) emitting source that

is static in bulk, the ratio of the luminosity due to first-generation inverse Compton
scattering of synchrotron photons to the luminosity due to synchrotron processes is

LIC,1 5
f≡ = Cνm Tbm (1)
Lsync

where νm is the frequency at which the synchrotron spectrum peaks, and Tbm is the
brightness temperature of the synchrotron radiation at that frequency.

(a) Derive an estimate for the constant C in terms of fundamental constants. Assume
whatever geometry you like for the source. I found I had to use the (usual) constants e,
k, c, and me .

You might find the heuristic derivation of the optically thick portion of synchrotron
spectra helpful (very indirectly).

We derived in class that f = Urad,sync /UB , where Urad,sync is the energy density of
synchrotron radiation, and UB is the energy density of the background magnetic field.

Also as outlined in class, we obtain a minimum estimate of Urad,sync by taking the

energy density near the peak of the synchrotron spectrum (where optically thick tran-
sitions to optically thin). We imagine a sphere of synchrotron radiation of radius R,
energy density Urad,sync , and distance to the observer d. The luminosity of the source is

Lrad,sync ∼ Urad,sync c4πR2 (2)

since Urad,sync c is the internal flux of energy in the source (a flux is a density times a
speed; this internal flux is not to be confused with the flux of energy observed by the
distant observer), and 4πR2 is the area of the source.

Now the distant observer measures a flux

1
 Frad,sync = Lrad,sync /4πd2 (3)

so we combine (3) and (2) to write

R2
Frad,sync ∼ Urad,sync c (4)
d2

But we can also write

Frad,sync ∼ νFν (5)

πR2
∼ νIν (6)
d2
2kTbm πR2
∼ νm 2 (7)
λm d2

where Tbm is the maximum brightness temperature of the source measured at wavelength
λm . Combine (7) and (4) to write

2π 3
Urad,sync ∼ kTbm νm (8)
c3

Now we have to massage the denominator, UB , into terms of Tbm and νm . We know that

2 2 eB
νm ∼ γm νcyc ∼ γm (9)
2πme c

where γm is the Lorentz factor of the electrons that are most responsible for emission at
νm . We have argued in class that γm me c2 ∼ 3kTbm . Putting it all together,

UB = B 2 /8π (10)
πm2e c2 νm2
∼ (11)
2e2 γm 4

πm6e c10 νm 2
∼ 2 4 4 (12)
162e k Tbm

This finishes the massage. Our final answer is therefore

2
 324e2 k5 5
f= νm Tbm (13)
c13 m6e

(b) If νm ∼ 1 GHz (as it typically is for compact radio sources), what is the maximum
Tbm above which f > 1? Express in Kelvin.

We solve the last equation in part (a) for Tbm given f = 1 and νm = 109 Hz. We get
Tbm ∼ 1 × 1012 K, as advertised in class.

(c) Given f , what is LIC,2 /LIC,1 ? That is, what is the ratio of second-generation to
first-generation scattered power? Explain your answer.

Well

LIC,2 Urad,IC,1
= (14)
LIC,1 Urad,sync

where Urad,IC,1 is the energy density of radiation produced by the first-generation of

Compton scatterers. But the right-hand-side just equals f . Therefore the answer is f .

(d) Consider the situation after N scatterings, where N → ∞. What is the critical f
for which the total luminosity of the source becomes unbounded? Careful estimates will
be rewarded.

The total luminosity after N scatterings equals

X
N
Ltot = LIC,i (15)
i=0
XN
= f i Lsync (16)
i=0
X
N
= Lsync fi (17)
i=0

where LIC,0 ≡ Lsync . This sum diverges for f > fcrit = 1.

Problem 2. Compton Saturation

Rybicki & Lightman problem 7.1b (only b. We answered the other parts in lecture.)

See the solution in RL pages 351–352.

3
Problem 3. Typical vs. Average (or Lies, Damn Lies, and Statistics)

Consider a non-relativistic electron moving at speed v  c. A single photon of original

energy  Compton scatters off the electron and suffers an energy shift of ∆. Work in
the same Thomson limit that we have been working in throughout lecture.

(a) Estimate a typical value for the fractional change in photon energy, ∆/. Express
your answer symbolically. By “typical,” we mean any old photon that comes in from any
old direction.

From lecture, we wrote down the exact expression for the photon’s energy post-
scattering:

v
1 = 01 γ(1 + cos θ10 ) (18)
c

where the subscript 1 refers to any quantity post-scattering, and the superscript prime
refers to any quantity evaluated in the electron’s rest frame. The γ here refers to the
electron’s Lorentz factor (it is very nearly one since the electron is non-relativistic).

Now in the Thomson limit that we have been working in throughout lecture, 01  me c2 ,
which means that in the electron’s rest frame, the collision is elastic: 01 ≈ 0 . Then

v
1 = 0 γ(1 + cos θ10 ) (19)
c

But we know from the rules of Doppler shifting from the lab frame to the electron’s rest
frame that

v
0 = γ(1 − cos θ) (20)
c

Substitute this equation into the previous one to find

v v
1 = γ 2 (1 + cos θ10 )(1 − cos θ) (21)
c c

Now in a typical scattering, θ10 ∼ 1 and θ ∼ 1 and they are not likely to be the same.
Therefore, to order-of-magnitude, replacing γ 2 = 1, we have

v
1 ∼ (1 ± ) (22)
c

4
where the ± refers to the fact that either the cos θ10 term or the cos θ term might win in
any scattering; the photon might be upscattered OR downscattered, depending on the
geometry of a single collision! Then our final answer reads

∆ v
∼± (23)
 c

(b) Consider this same electron flying through a sea of photons all of energy . Many
photons get inverse Compton scattered to and fro.

Averaged over all the photons that got scattered, what is the mean fractional change in
photon energy, h∆/i? You may use our expression for the inverse Compton power
scattered by a single electron if you wish. Neglect the change in the electron’s velocity as
it is bombarded by photons.

This question was answered in class. The easiest way to derive this is to take the
net power radiated by a single electron (read: net power imparted to the photons) and
divide by the rate of collisions. This gives the mean energy shift for the photons:

Pcompton
h∆i = (24)
nphoton σT c
(4/3)Uphoton γ 2 β 2 σT c
= (25)
nphoton σT c
4 2 2
= γ β  (26)
3

where the energy density of the photons, Uphoton , divided by the number density of
photons, nphoton , equals the mean energy per photon, . Therefore to leading order in
v/c,

h∆i 4v 2
= 2 (27)
 3c

and it is always positive, since we said at the outset that the electron has more energy
than the photons at the beginning.

(c) Explain physically why parts (a) and (b) give different answers. (And then read
Blandford page 175, if you wish. It is better to think about this problem and achieve
your own understanding than to head straight for Blandford right away).

Answer (a) refers to the typical energy shift in a single scattering. A single scattering
can either upshift or downshift a photon’s energy. To order v/c, the probability of an

5
upshift equals the probability of a downshift. But to order (v/c)2 , the probability of an
upshift exceeds the probability of a downshift. That is why in part (b), the expectation
value of the energy shift is positive and of order (v/c)2 . Blandford calls upshifts and
downshifts “blueshifts” and “redshifts,” respectively.

6
 Astro 201 – Radiative Processes – Solution Set 12

by Eugene Chiang

Problem 1. Faraday Rotation

Consider the propagation of light through a magnetized plasma. The magnetic field is
uniform B~0 = B0 ẑ. Light travels parallel to ẑ.

An electron in the plasma feels a force from the electromagnetic wave, and a force from
the externally imposed B-field. Its equation of motion reads

~ − e ~v × B~0
m~v˙ = −eE (1)
c

where the electric field E~ can be decomposed into right-circularly-polarized (RCP) and
left-circularly-polarized (LCP) waves:

~ = E0 (x̂ ∓ iŷ)ei(k∓ z−ωt)

E (2)

where it is understood that the real part should be taken. The upper sign (-) corresponds
to RCP waves, while the lower sign (+) corresponds to LCP waves.

In the equation of motion above, we have neglected the Lorentz force from the wave’s
B-field, since it is small (by v/c) compared to the force from the wave’s E-field.

(a) Prove that the solution of the equation of motion reads

−ie ~
~v = E (3)
m(ω ± ωcyc )

where ωcyc = eB0 /mc. This is the hardest part of the derivation for the dispersion
relation of RCP and LCP waves. But it is fairly straightforward.

The x-component of the equation of motion (1) reads

eE0 i(k∓ z−ωt)

v̇x = − e − ωcyc vy (4)
m

where ωcyc = eB0 /mc, and the y-component reads

1
 ieE0 i(k∓ z−ωt)
v̇y = ± e + ωcyc vx (5)
m

These are coupled first-order ODEs. We take another time derivative of the x-equation,
and then plug in the y-equation. We get

ieE0 i(k∓ z−ωt) 2

v̈x = e (ω ∓ ωcyc ) − ωcyc vx (6)
m

Our ansatz (read: guess) for the solution reads

vx = Aei(k∓ z−ωt) (7)

which we insert into (6) to solve for A. We obtain

ieE0 (ω ∓ ωcyc )
A = (8)
m(ωcyc + ω) (ωcyc − ω)
−ieE0
= (9)
m(ω ± ωcyc )

Repeat the above analysis for vy . We get

vy = Bei(k∓ z−ωt) (10)

where

−eE0
B= (11)
m(ωcyc ± ω)

It is straightforward to finish the proof from here.

(b) Rybicki & Lightman Problem 8.3

See RL’s solution on page 357.

Problem 2. Combined Scattering and Absorption

Rybicki & Lightman Problem 1.10

2
 See solutions in RL.

Problem 3. Greenhouse Warming

Taken from Chamberlain & Hunten Problem 1.3

Assume that solar radiation is absorbed only at the Earth’s surface where the albedo is
40%. The re-radiated energy is absorbed mainly by water vapor, which we approximate as
a gray absorber with a density scale height of 2 km and total optical depth τ = 2. Plot the
temperature distribution with height for radiative equilibrium. What is the temperature
discontinuity at the ground? What is the gradient in the air temperature near the ground,
in K/km?

First calculate the effective temperature Teff , defined in terms of the absorbed flux:

L⊙ 2 4 2
(1 − A) × πR⊕ = σTeff × 4πR⊕ (12)
4πa2

where we have assumed that the atmosphere/earth is spherically symmetric. (You could,
alternatively, solve this problem just for a patch on the Earth at a certain latitude, longi-
tude, and time, in which case you would have to calculate the local angle of insolation.)
Keep in mind, the effective temperature is not an actual physical temperature; it is
merely shorthand for the radiative flux.

For a = 1 AU and A = 0.40, the above equation yields Teff = 247 K.

We know from lecture/reading that the temperature at infinity for an atmosphere in

radiative equilibrium is T0 = Teff /20.25 = 208 K.

The temperature profile according to the Eddington solution is

T = T0 (1 + 3τ /2)1/4 = 208(1 + 3τ /2)1/4 K (13)

So at the surface where τ = 2, the air temperature is T1 = 294 K, which sounds about
right (21 degrees Celsius).

Re: the gradient in air temperature:

dT dT dτ 3T dτ
= = (14)
dz dτ dz 8(1 + 3τ /2) dz

We are told that the density, and therefore the optical depth, falls with a scale height
of h = 2 km. So τ = τ1 exp(−z/h) where τ1 = 2, and so dτ /dz = −τ /h. Plugging in,

3
 dT 3T τ
=− (15)
dz 8(1 + 3τ /2) h

which at the surface gives dT /dz = −27.6 K/km.

Re: the temperature discontinuity with the ground. From lecture/reading, we

know that under conditions of M.R.E. (monochromatic radiative equilibrium) Bν (Tg ) =
Bν (T1 ) + Fν /2π, where Fν is the radiative flux (constant with height). We generalize
this statement for B.R.E. (bolometric radiative equilibrium), and write

σTg4 = σT14 + F/2 = σT14 + σTeff

4
/2 (16)

from which it follows that Tg = 311 K. So the temperature discontinuity is Tg − T1 = 17

K.