lecture08
lecture08
Initial questions: What sets the temperature of the warm interstellar medium? What
determines the epoch from which we see the CMB?
We are now going to consider the statistics of equilibrium. Specifically, suppose we let
a system stand still for an extremely long time, so that all processes can come into balance.
We then measure the fraction of particles that are in each of several states (e.g., if we have
a pure hydrogen gas, we might want to know the ionization fraction). How can we compute
this fraction? We’ll take a look at this in several different ways. We will then step back
and ask ourselves when these equilibrium formulae do not apply. This is important! To
paraphrase Clint Eastwood, “An equation’s got to know its limitations”. In virtually all
astronomical problems we deal with approximations, so we always need a sense of when
those approximations fail. This serves a double purpose: it keeps us from making errors, but
it also keeps us from applying too complex an approach to a simple problem.
Let’s start out with a simple case. We consider a column of isothermal gas (temperature
T ) in a constant gravitational acceleration g. Each molecule has mass m. If we can treat
this like an ideal gas, Ask class: how do we figure out how the pressure varies with height
h? An ideal gas has pressure P = nkT , where n is the number density. In equilibrium the
pressure gradient balances the weight of the gas, so dP/dh = −ρg. Using ρ = nm = P m/kT
and the constancy of T , we have dP/dh = −(mg/kT )P , so P = P0 exp(−mgh/kT ), where
P0 is defined as the pressure at h = 0. This is the familiar Boltzmann formula. As usual, we
can check various limits: when h increases the pressure decreases (correct); when T decreases
the pressure decreases at a fixed h (correct); and so on. Note that the decrease is faster for
heavier molecules. This is one reason why at high altitudes we have an oxygen deficit: our
natural breathing rate is controlled by the level of CO2 in the bloodstream, and this decreases
rapidly at high altitude so we don’t breathe as often as we should until we acclimate.
This is one special case of a much more general formula of statistical distribution in
thermal equilibrium. I highly recommend you look at the Feynman Lectures on Physics if you
want physical insight on just about anything; Feynman had a remarkably original perspective
on physics, which is different enough that it often provides the key to understanding. In it,
he shows that in any potential, the distribution of states is proportional to exp(−∆E/kT ),
where ∆E is the difference in energy from some arbitrary level (it’s the ratio that matters,
so we don’t need to specify the reference level). There may be other factors (e.g., the
multiplicity of the state), but this basic one is always obtained in a classical equilibrium.
Another example is the Maxwell-Boltzmann distribution. At temperature T , the fraction in a
volume d3 p around momentum p is ∝ exp(−E/kT )d3 p, which (for nonrelativistic motion and
isotropic distribution in momentum space) is ∝ exp(mv 2 /2kT )p2 dp. Yet another example
is the population of states in an atom. Compared to some reference state (e.g., the ground
state), the fraction of atoms in a state of energy E is ∝ exp(−E/kT ).
Now let’s return to the subject of ionization. This is important because the ionization
fraction has consequences for the equation of state (pressure related to density and temper-
ature) and opacity (e.g., which sources matter most), among other things. Let us assume
that we have a cavity in which atoms and radiation are in thermal equilibrium with the same
temperature T . For simplicity, consider a hydrogen atom with a single energy level (ground
state), and consider the ionization reaction
H + + e− ←→ H 0 + χH , (1)
where χH =13.6 eV is the ground state binding energy. In equilibrium, what can be said
about the abundances? In this case, let the reference energy (E = 0) be for the free electron
and proton. Statistical weights: 2 for the hydrogen atom, two combined for the proton and
electron (spins in same direction, spins in opposite directions). From the previous lecture
notes, integration of the distribution function over the momenta yields the number densities
2[2πme kT ]3/2 µ− /kT
ne = e , (2)
h3
[2πmp kT ]3/2 µ+ /kT
n+ = e , (3)
h3
and
2[2π(mp + me )kT ]3/2 µ0 /kT χH /kT
n0 = e e . (4)
h3
Ask class: if we were being really accurate, would the actual mass in the third equation
be mp + me ? (No, because we’d have to subtract the binding energy. But this is only
13.6/511,000 of even the electron mass, which would be a 1/40,000 correction, and can be
dumped. In the same spirit, the electron mass is only a 1/1,800 correction and can be ignored
here.)
Forming the product n+ ne /n0 , we find
3/2
n+ ne
2πme kT +
e(µ +µ −µ0 )/kT e−χH /kT .
−
0
= 2
(5)
n h
Ask class: what is (µ− + µ+ − µ0 ) in equilibrium, which we’ve been assuming? It’s zero.
The numerical factor is about 2.4 × 1015 T 3/2 cm−3 . Ask class: is this equation sufficient to
determine the abundances of electrons, protons, and neutral hydrogen? No, because we have
only one equation for three unknowns. Ask class: what are reasonable auxilliary conditions?
Charge balance implies n+ = n− , and conservation of nucleons means that n0 + n+ = n, the
total number density of protons (which equals the total number density of electrons). With
these extra conditions, the degree of ionization y ≡ n+ /n = ne /n is, numerically,
y2 4 × 10−9 3/2 −1.6×105 /T
= T e , (6)
1−y ρ
where as usual quantities such as ρ and T are referenced to the appropriate cgs unit.
An important thing to remember about this equation and many like it is that the
exponential usually dominates. What this means specifically is that for “interesting” levels
of ionization (say, around 50%), T is so much lower than 1.6 × 105 K that a tiny change in
T makes a huge difference in the ionization fraction. Operationally, this means that you can
get a remarkably good first guess at the temperature for a specified value of y by ignoring
the T 3/2 factor, which doesn’t change as rapidly. If you like, you can then substitute your
value of T in the T 3/2 factor and solve again by iteration. This is an easy way to converge
rapidly to a solution without having to write a computer program.
In fact, let’s do a practice example. Suppose we take a number density of 1 cm−3 , which
is about average for our interstellar medium. Then ρ = 1.7 × 10−24 g cm−3 , so the Saha
equation becomes
y2 5
≈ 2.4 × 1015 T 3/2 e−1.6×10 /T . (7)
1−y
We would like to solve for the temperature where y = 0.5, which is about the halfway
ionization point. Let us also say that we only need the answer to within a factor of 3. Thus
we get
5
T 3/2 e−1.6×10 /T ≈ 2 × 10−16 , (8)
where T is in Kelvin. Getting an exact solution would require a computer, but the fact that
the right hand side is so small gives us an easy way to get an approximate answer. Let us
first ignore the T 3/2 factor. Then we can use the approximate equation
5 /T
e−1.6×10 ≈ 2 × 10−16 . (9)
Since e10 ≈ 2 × 104 , if the exponent is about -40 then we get something in the vicinity
of 10−16 . This would suggest T = 4000 K. Using that value, T 3/2 is about 43/2 (103 )3/2 ≈
5
8 × 3 × 104 ≈ 2 × 105 , so dividing through we would need e−1.6×10 /T ≈ 10−21 , or about -50
for our exponent. This would suggest T ≈ 3200 K, which is already within our factor of 3.
3300 K is an excellent approximation. Behold the power of exponentials! It should be said,
by the way, that in reality the warm neutral medium has a temperature of about 8000 K.
This is an indication that one or more of our assumptions is not correct. That assumption
is that the radiation field is in thermal equilibrium with the matter. In fact, the ionizing
radiation field is provided mainly by UV from hot stars, rather than a constant thermal
bath. Thus even though we did not get quite the right answer, (1) we did get within a factor
of a few, and (2) the fact that comparison with observations shows we’re not quite there tells
us that we need to know something extra.
Ask class: what does the Saha equation imply about the degree of ionization for a fixed
temperature but varying density? (degree of ionization goes up when density goes down).
Why is that? This is thermal ionization, so you can imagine some radiation field pervading
the region. The ionization rate per volume, therefore, just goes like the number density.
However, the recombination rate involves two particles, and hence must go as the product of
their densities. Therefore, the recombination rate increases more rapidly when the density
goes up, and the resulting ionization fraction goes down.
Let’s test how well this equation works in practice. Ask class to compute ionization
fraction in the center of the Sun. Suppose Ask class T = 1.5 × 107 K and ρ =
100 g cm−3 , for round numbers. Then y 2 /(1 − y) = 2.3. This has a solution of y = 0.75,
so the center of the Sun is 25% neutral (!). See if this poses a problem for anyone in
class!. This is nonsense, of course! The problem is that at such densities, the separation
between atoms is much less than the size of an individual free atom, so the electrons get
squeezed up to continuum states and pressure ionization dominates. In reality, the center
of the Sun has fully ionized hydrogen. As Shu points out, this also means that when you
sometimes hear people say that the full ionization is because of the high temperature, that’s
not completely accurate.
Okay, so let’s return to the Saha equation. We derived it by assuming that hydrogen
atoms are single-level systems (either ground state or ionized). ’Tain’t so in reality. There
is, in principle, an infinity of excited bound states. Each of these contributes a term that is
like the multiplicity of the state times exp(−∆E/kT ) times the original term, where ∆E is
the energy with respect to the energy of the ground state. But wait! This means we have an
infinite sum of finite terms. Ask class: how do we get around this? In reality, the size of
the orbital increases like n2 , so in most circumstances only the first term needs to be kept,
as we did.
The Saha equation is important enough that I’d like to come at it from another perspec-
tive, given in Shu (Chapter 7, pgs. 64-65). Let’s define the thermal de Broglie wavelength
of the electron
λT e ≡ h/(2πme kT )1/2 . (10)
Then the Saha equation between ionization states i and i + 1 can be written
ni+1 2Zi+1
= exp(−Ii /kT ) . (11)
ni (ne λ3T e )Zi
Here Ii is the ionization energy from state i to state i + 1 and Zi and Zi+1 are partition
functions (these are sums of the multiplicity of each atomic level times exp(−∆E/kT )). Even
better, we note that the chemical potential for nondegenerate free electrons is −µe = kT ξ,
where ξ ≡ ln(2/ne λ3T e ). Note that ne λ3T e is the number of electrons in a cubic thermal de
Broglie wavelength, which is really tiny if our assumption of nondegeneracy works. Typically
ξ ∼ 10 − 30 (remember, “all logarithms are 10”!). Then we rewrite the Saha equation as
ni+1 Zi+1 −(µe +Ii )/kT
= e . (12)
ni Zi
What does this tell us? The ratio of partition functions is usually of order unity, and as
before the exponential will dominate. This means that the two states are populated equally
roughly when −µe = Ii . This implies that the thermal energy at this transition temperature
satisfies kT ≈ Ii /ξ, which is a good order of magnitude less than one’s first guess kT ≈ Ii .
Why the difference? You can think of it like this: at a temperature kT = Ii /ξ there is a
substantial energy penalty to be paid by ionizing an electron, but the electron then has lots
more phase space to run around in. There are some additional consequences. First, the
transition between ionization states is usually pretty sharp, on the order of ∆T ≈ kT /ξ.
Therefore, if you fix the density and look at the state that is most populated, it changes
abruptly with the temperature (and there is usually one state that dominates). Second, this
means that it is in some sense “easier” to ionize an atom than to raise it to an excited bound
state (since that really does need kT ≈ Ii , if the excitation energy is roughly Ii as it is in
hydrogen).
The Saha equation is useful in lots of circumstances, but now we need to examine the
assumptions that go into it. Ask class: what are two circumstances under which the
Saha equation will give incorrect ionization fractions? One we just saw: when the density
is high enough to infringe on the atoms, thermal ionization isn’t as relevant as pressure
ionization. Ask class: what would this mean, physically, about the size of atoms and their
separations? Qualitatively, if the separation between atoms is comparable to the size of their
electron clouds, one would expect significant extra ionization. This is the same thing we used
to argue that we could cut off the partition function after a finite number of terms.
Ask class: what is another way in which the Saha equation could be invalid? If there is
any other source of ionization, it could modify the ionization fraction. Ask class: Examples?
Gas in a molecular cloud at 100 K that is being illuminated by UV from a hot star has a
much different ionization structure than if it were not illuminated. Cosmic rays help keep
protoplanetary disks far more ionized than they would be in thermal equilibrium at their
temperatures. In fact, this may be crucial to the birth of stars. There is strong evidence
that a nonnegligible level of ionization is crucial to allowing angular momentum transport
in protoplanetary disks, so without the influence of cosmic rays the matter would just pile
up until something catastrophic occurred.
Ask class: what is yet another way in which the ionization fraction could be different
from the Saha prediction? The ionization fraction could in principle be anything at all if
there hasn’t been enough time to establish equilibrium. Ask class: is this more likely in low-
density or high-density environments? Low, because interaction rates are less. An example
is the very early universe (prior to any star formation), where there is a residual ionization
fraction because the density drops due to universal expansion. Ask class: how, then, would
one compute the ionization fraction? Your only recourse is to have some initial conditions,
then make calculations based on the fundamental rate equations rather than being able to
assume equilibrium. As always, equilibrium is a great simplifier, but you have to be able to
justify that it is valid.