ASTR 4017/8007: Diffuse Matter in the Universe
Class 17 Notes: H i: Thermodynamics and Phase Structure
We now turn to the next phase of the ISM: atomic hydrogen. This is the most abundant
phase of the ISM in the disc of the Milky Way, and in most other galaxies like it. Our goal
today will be to understand the heating and cooling processes in these gas that control its
thermodynamics, and to understand the reason we speak of H i as having distinct phases.
We will then examine the interfaces between those phases.
I. Heating processes
As with H ii, we will start by examining the heating processes that take place in this
gas. There are two dominant ones: cosmic rays and the grain photoelectric effect.
A. Cosmic rays
Galactic discs are filled with energetic cosmic rays that have long mean free paths.
When these strike atoms, they cause ionisations, a process we studied several
weeks back in our discussion of ionisation processes. The free electrons created by
this process typically have energies around 35 eV. Some of this energy will go into
secondary ionisations, or into collisional excitation of H, H2 , or He followed by
radiative de-excitation, but some of it will go into thermal energy as the electron
scatters off other particles.
The amount that goes into heat depends on the chemical state of the gas with
which the electron is interacting. If the gas is wholly or partially ionised, the
cosmic ray-produced electron will Coulomb scatter off the large number of free
electrons, rapidly thermalising. As a result, almost all of its energy will go into
heat. In predominantly neutral gas, the interactions will be charged-neutral in-
stead. If the scatterings do not excite the target atom or molecule, then there
will be negligible energy transfer. The electron loses energy mainly by causing
excitations in the atoms or molecules it hits.
In atomic gas, the primary collision parters for the electron will be H atoms.
When such an atom is excited, it is far more likely to radiatively de-excite than
to collisionally de-excite, and if this happens the photon will usually escape and
thus no heat will be added to the gas. Heat is added only on the rare occasions
when the H atom collisionally de-excites. Calculations of the many relevant cross
sections and branching ratios by Dalgarno & McCray suggest that the energy
yield per cosmic ray-produced electron may reasonably be approximated by
1/2
xe
Eh = 6.5 eV + 26.4 eV , (1)
xe + 0.07
where xe = ne /nH . In the atomic ISM xe is generally quite small, so the yield is
close to 6.5 eV per CR ionisation.
1
Glassgold et al. (2012) find that the yield is a factor of ∼ 2 larger in regions with
significant molecular content, because the many rotational and vibrational levels
of H2 provide additional conduits into which electrons can dump their energy, and
because the lower Einstein A’s of these transitions make it much more likely that
molecules will collisionally rather than radiatively de-excite. Chemical reactions
facilitated by the presence of free electrons and H+
2 ions (created when the primary
electron is knocked free) provide an additional heating source.
In the atomic gas, the cosmic ray heating rate per unit volume due to interactions
with neutral atoms is
ΓCR,n = (nH0 + nHe ) ζCR Eh , (2)
where ζCR is the primary cosmic ray ionisation rate (i.e., not including secondary
ionisations). For the observed galactic cosmic ray population, this is about 10−16
s−1 . Plugging this in, we obtain
" 1/2 #
xe
−27 −1
ΓCR,n ≈ 1.0 × 10 nH erg s ζCR,−16 1 + 4.06 , (3)
xe + 0.07
where ζCR,−16 is the cosmic ray ionisation rate in units of 10−16 s−1 .
Cosmic rays can also interact with free electrons and heat them, but this effect is
small compared to heating due to interactions with H and He atoms unless the
ionisation fraction is high.
B. Dust photoelectric heating
Another source of free electrons that can heat gas is dust grains. These have the
great advantage over atoms and molecules that they are continuum absorbers,
and the great advantage over cosmic rays that photons are much more abundant
than cosmic rays in the ISM. The work function (the minimum energy required to
eject an electron) for graphite is about 4.5 eV, so all photons in this energy range
can produce photoelectrons, although the yield is very low until photon energies
reach about 8 eV.
As a rough estimate, let n(8 − 13.6 eV) be the number density of photons from
8 eV, where the photoelectric yield becomes significant, up to 13.6 eV, where the
spectrum cuts off in the neutral ISM due to absorption by neutral hydrogen. For
the observed interstellar radiation field of the Milky Way, this is about 3 × 10−3
cm−3 . The grain cross section to photons in this energy range is about hσabs i ∼
10−21 cm2 per H atom at Milky Way dust abundances, and the photoelectric yield
per absorption is typically hY i ∼ 0.1. Note that it is this small largely because
the grains are on average quite negatively charged; for neutral grains it would be
closer to unity In this case the rate of dust photoelectric heating per unit volume
is
! !
−26 −1 n(8 − 13.6 eV) hσabs i hY i hEpe i − hEc i
Γpe = 1.4 × 10 nH erg s ,
3 × 10−3 cm−3 10−21 cm2 0.1 1 eV
(4)
2
where hEpe i and hEc i are the mean kinetic energy per photoelectron and the mean
kinetic energy of electrons captured from the gas onto the grain, respectively.
These depend on somewhat uncertain energy-dependent electron capture rates
and photoelectric yields, and the estimate of a 1 eV difference is based on a
combination of laboratory experiments and simple models. More sophisticated
approaches first model the charge distribution of the grains, then use the charge-
dependent photoelectric yields to evaluate this function, but our simple estimate is
accurate to the order of magnitude level. The heating rate tends to be dominated
by small grains, both because these provide most of the UV absorption, and
because the large charges on larger grains tend to reduce their photoelectric yield
and photoelectron energies.
The interesting thing to notice here is that this rate is an order of magnitude
larger than the cosmic ray heating rate. Thus the grain photoelectric effect is the
dominant heat source in the atomic ISM, at least under Milky-Way like conditions.
Because the grain photoelectric heating rate scales linearly with both the number
density of hydrogen atoms (assuming a fixed dust to gas ratio) and the number
density of UV photons, it is common to write it as
Γpe = nH G0 g(G0 /ne , Te ), (5)
where G0 is the normalised UV radiation field strength discussed earlier in the
course. The function g incorporates all the messy grain behaviour. It depends
on G0 /ne , where ne is the free electron number density, and on the electron
temperature Te , because these two quantities determine the charge equilibrium of
the grains. The function g also depends on the grain size distribution.
II. Cooling processes: atomic lines
A. Dominant cooling lines
The main cooling process in the atomic ISM is, as in ionised gas, collisionally
excited lines. Which lines dominate is determined by a number of factors
• More abundant species are obviously more important than less abundant
ones.
• The ambient starlight spectrum contains many photons below 13.6 eV, and
almost none above it. Thus the relevant ionisation state is neutral for elements
with first ionisation potential above 13.6 eV (e.g., O) or once ionised for
elements with first ionisation potential below 13.6 eV (e.g., C and Si).
• In order to cool effectively, the energy of the upper state of a particular line
must be well matched to kT . Higher energy levels are not collisionally excited
often enough to contribute significantly, and lower energy levels are excited
often but do not remove much energy per collision. As a result, the dominant
lines tend to be forbidden infrared fine structure lines, which generally have
upper state energy levels at ∼ 100 − 1000 K.
3
The most important coolant at low temperatures ( 104 K) is generally C ii.
Carbon is the most abundant element after H and He, and its first ionisation
potential is 11.26 eV, so in the atomic ISM most of the carbon is C ii. This
species has five electrons, so its ground electronic state is 1s2 2s2 2p1 . Since there
is only one electron in the outer shell, the L−S state corresponding to this ground
electronic state is trivial. The one electron has orbital angular momentum ` = 1
and spin angular momentum s = 1/2, so the total orbital angular momentum
is L = 1 and the total spin angular momentum is S = 1/2. The total angular
momentum is either 1/2 or 3/2, depending on whether the spin and orbit are
aligned or anti-aligned, so the L − S state corresponding to the ground electronic
state consists of a doublet 2 Po1/2 and 2 Po3/2 . As one expects for a fine structure
splitting, the energy difference is small: the J = 3/2 level lies 92 K above the
J = 1/2 state. This energy difference is well matched to temperatures ∼ 100 K.
The second most important low-temperature coolant is O i. Like C it is very
abundant, and its first ionisation potential is 13.618 eV, as opposed to 13.598
eV for H, so there are few photons capable of ionising oxygen. Moreover, the
nearly identical ionisation potentials means that O and H can charge exchange
easily, and this generally forces the oxygen ionisation fraction to be closely tied
to the H ionisation fraction. O i has 8 electrons, so the ground electronic state
is 1s2 2s2 2p4 . The ground electronic state in this cases consists of three terms:
3
P, 1 D, and 1 S. The 1 S and 1 D terms are singlets, and both lie ∼ > 104 K above
ground, so they are not important at neutral gas temperatures. The 3 P state
is a triplet that breaks into 3 P0 , 3 P1 , and 3 P2 , with the latter being the lowest
energy state. The 3 P1 is 228 K above ground, and the 3 P0 is 326 K above ground.
Since the direct 3 P0 → 3 P2 transition has ∆J = 2 it is strongly suppressed, and
collisional excitations to 3 P0 almost always decay by going into 3 P1 and then 3 P2
The first step produces a 14 µm photon, but most of the power comes out in
the 3 P1 → 3 P2 decay, which produces a photon at 63 µm. This process starts
to contribute significantly compared to carbon once the temperature is above a
several hundred K.
At temperatures that approach 104 K the cooling rate shoots way up because non-
fine structure transitions and electronic transitions become possible. In particular,
the Lyα line of hydrogen is 1.2 × 105 K above ground, and can provide a huge
cooling effect for gas that reaches temperatures approach 104 K. This is generally
the dominant cooling line for the warmest phases of H i.
In addition to these major lines, there are a host of minor lines that contribute
at the ∼ 10% level. Numerical calculations can take them into account.
B. Critical densities
The two dominant low-temperature cooling lines we have discussed have fairly
high critical densities. For C ii, the critical density for collisions with neutral H
atoms is 3 × 103 cm−3 , and for collisions with free electrons it is 10 cm−3 . For
O i, the critical densities are 2 × 105 cm−3 , 1 × 105 cm−3 , and 1 × 105 cm−3
4
for neutral hydrogen atoms, free electrons, and free protons, respectively. All of
these densities are significantly higher than typical densities in atomic regions, so
to good approximation we can simply assume that every collisional excitation is
followed by an immediate radiative de-excitation that removes energy.
Since the collision rate varies with density as n2 , it is common to write the cooling
rate as
Λ = n2H λ(T ), (6)
where λ(T ) is a function that depends on temperature only, and incorporates
the temperature dependence of the collisional excitation rate coefficient for a
particular line. Note that, in some sources, one instead sees the cooling rate
written Λn2 , with Λ being the quantity that we have written λ(T ). Unfortunately
there is no uniformity on this in the literature, so check units carefully to make
sure you know which cooling function you’re dealing with!
The temperature dependence λ(T ) behaves in the manner one would expect. For
the neutral-neutral collisions λ(T ) ∝ T 1/2 e−Elev /kT from the usual integration of
the collision rate coefficient over the Boltzmann distribution. For ion-neutral
collisions the variation is λ(T ) ∝ e−Elev /kT , while for ion-electron collisions it is
λ(T ) ∝ T −1/2 e−Elev /kT .
III. Thermal equilibrium and the two phase model
A. The three phases
Given these heating and cooling processes, we can solve for the equilibrium tem-
perature for a given density. In the process we must solve for the equilibrium
ionisation state too, since excitations due to collisions with free electrons can be
significant even at low electron fractions, due to the higher thermal velocities and
collision strengths of the electrons compared to the H atoms. Thus we simulta-
neously balance the heating rate against the cooling rate, and the ionisation rate
against the recombination rate, for a given density and abundance of elements.
Such calculations have been carried out by a number of authors, most recently by
Mark Wolfire and collaborators.
[Slide 1 – temperature vs. density from Wolfire+ 1995]
We can understand the general trend here by recalling that the main heating
source, cosmic rays, provides an energy input per unit volume that varies as nH ,
whereas the main cooling effects from collisionally-exicted lines remove energy at
a rate that varies as n2H . We roughly expect
G0 g(G0 /ne , Te )
nH G0 g(G0 /ne , Te ) ≈ n2H λ(T ) =⇒ λ(T ) ≈ . (7)
nH
The UV field G0 and the grain properties represented by the function g do not
vary much with density or temperature, so the temperature-density relation is
essentially dictated by the shape of λ(T ).
5
In the regime from ∼ 102 − few × 103 K where C ii and O i cooling dominate,
λ(T ) is a slowly increasing function of T . This means that the temperature must
drop as the density increases, and that it must do so faster than 1/nH .
This trend breaks at both low and high T . On the high T side, the equilibrium
temperature cannot get much higher than ∼ 104 K, because at that point the
immense effects of Lyα cooling kick in and prevent the gas from getting any
warmer as long as neutral H is present. Similarly, the equilibrium temperature
in atomic gas stops declining with density once the temperature is ∼ 30 − 50 K
because the C ii cooling rate has a temperature dependence e−92 K/T . Once T
drops significantly below 92 K, the line ceases to be able to cool effectively, and
λ(T ) becomes an extremely steep function of T . As a result, only a small change
in temperature is required to offset a large change in density.
It is instructive to plot this in a somewhat different way: pressure (∝ nT ) versus
density rather than temperature versus density. The motivation for this is that
gas at a given position in a galaxy generally has a pressure that is fixed by
its environment, e.g. the weight of the gas on top of it. Thus the density and
temperature of the gas at that point have to adjust to produce the required
pressure. If they cannot, the gas density will be changed by hydrodynamic motions
until an equilibrium is found. The resulting plot of pressure versus density is a
characteristic S-shaped curve, the general form of which was first obtained by
Field, Goldsmith, and Habing in 1969. As a result, this is sometimes referred to
as an FGH curve.
[Slide 2 – FGH curve from Wolfire+ 1995]
< − 0.3, the temperature is not far from its ∼ 104 K ceiling,
At low density, log n ∼
and the pressure increases close to linearly with density, since P ∝ nT . As the
density reaches log n ∼ −0.5, the temperature drops below 104 K, mostly due to
the [O i] 63 µm line. Thus the pressure stops rising linearly with temperature. As
the density continues to rise the cooling gets stronger and stronger, and eventually
the equilibrium temperature declines faster than 1/n. Thus the pressure decreases
as the density increases. This continues until the temperature gets to ∼ 100 K
at a density log n ∼ 0.5. At this point the exponential dependence of the C ii
cooling curve starts to play a role, and λ(T ) varies sharply with T . Thereafter the
temperature changes only fairly slowly, so the pressure-density relation becomes
close to P ∝ n again.
If we move vertically above the line then T increases at fixed n. This does not
change the heating rate, but it does raise the cooling rate, so in the region above
the line cooling occurs faster than heating and the gas tends to cool. Conversely,
in the region below the line, the reverse holds, and the gas tends to heat faster
than it cools.
We can identify three phases based on the maxima and minima in this curve. The
low density, higher temperature phase is known as the warm neutral medium, or
WNM. The high density, lower temperature phase is the cold neutral medium,
6
CNM. The intermediate density and temperature phase, we will see in a moment,
is unstable.
Which phases are actually present depends on the pressure in the ISM, which is a
function mostly of the large scale properties of the galaxy, i.e. how much pressure
is required to maintain hydrostatic equilibrium for a given galactic potential and
total amount of gas. Of course there are significant local variations around this
number. For the Milky Way near the Solar circle the estimated mean ISM mid-
plane pressure is P/k ∼ 3000 − 4000 K cm−3 , and in this regime all three phases
can exist. That appears to be the case for most spiral galaxies today.
B. Stability considerations
For a given pressure we have seen that there are three possible densities that can
be in thermal equilibrium. However, only two of those equilibria are stable. To
see this, it is helpful to write down the first law of thermodynamics:
de = T ds − P d(ρ−1 ), (8)
where e is the specific internal energy of a fluid, T is the temperature, s is the
specific entropy, P is the pressure, and ρ is the density (so that 1/ρ the volume
per unit mass).
The P d(ρ−1 ) term represents work done by or on the gas, while the T ds term
represents heating or cooling processes, such as radiation. Thus for our H i that
is subject to radiative heating and cooling, in a time dt during which radiation
adds or removes an amount of heat dq, we can write
dq L
ds = = − dt, (9)
T T
where L = Λ − Γ is called the loss function. It is the net rate of radiative energy
loss, with units of erg s−1 g−1 in cgs.
For a parcel of gas in thermal equilibrium, L = 0, since heating and cooling
balance, so the change in specific entropy ds is zero. Now consider perturbing a
parcel of gas away from equilibrium by changing its entropy by an amount δs at
time t = 0, in such a way as to leave some other thermodynamic variable A (for
example pressure or density) fixed. How will the gas respond? Based on what
we have just written down, we can write an evolution equation for the specific
entropy perturbation:
!
d ds L
(δs) = δ = −δ . (10)
dt dt T
In words, the rate at which the entropy perturbation grows or shrinks is simply
minus the rate at which the perturbation induces the gas to lose energy, divided
by the gas temperature.
Suppose that δs and δ(L/T ) have the same sign, meaning that when we increase
the specific entropy of the gas, the loss rate increases as well. In this case the
7
evolution equation tells us that δs goes down. Thus radiation acts like a restoring
force, and the entropy oscillates about its equilibrium value. On the other hand
suppose that δs and δ(L/T ) have opposite signs. In this case an increase in
entropy causes a drop in the loss rate, meaning that the gas heats up. In this case
the gas is pushed further from equilibrium, and is unstable.
Thus the condition for instability is simply that
" #
d L
< 0, (11)
ds T A
where the subscript A indicates that the derivative is to be taken while holding
the quantity A constant. This just says that the gas is unstable if L/T decreases
with increasing specific entropy. Since T is always positive, we can equivalently
write the stability condition as
!
dL
< 0. (12)
ds A
To apply this to the H i in the ISM, let us consider perturbing the entropy of the
medium at constant pressure, since the constancy of the pressure, at least over
long time scales and on average, will be enforced by hydrostatic balance in the
galactic disc. In our FGH diagram, this corresponds to starting from a point on
the equilibrium curve at a given pressure, then sliding left or right off the curve.
At constant pressure, a change in entropy ds is related to a change in temperature
by
T ds = CP dT, (13)
where CP is the specific heat at constant pressure. However, in order to keep
the pressure constant, a change in temperature dT must also involve a change in
density. Since P = ρkT /µ, where µ is the mean particle mass, we must have
ρ
T dρ = −ρ dT =⇒ dρ = − dT. (14)
T
Thus a change perturbation to the entropy ds produced a corresponding pertur-
bation to the temperature dT = (T /CP ) ds, and a perturbation to the density
dρ = −(ρ/T ) dT = −(ρ/CP ) ds. Plugging this in, we have
! ! !
dL T ∂L ρ ∂L
= − , (15)
ds P
CP ∂T ρ
CP ∂ρ T
and the condition for instability becomes
! !
∂L ρ ∂L
− < 0. (16)
∂T ρ
T ∂ρ T
8
We can write this in an even more compact form if we recall that, since P =
(k/µ)ρT , so
dP dρ dT
= + . (17)
P ρ T
If we make this substitution, then it is clear the instability condition is simply
!
∂L
< 0. (18)
∂T P
Now let us apply this to the FGH diagram. Suppose we start at the leftmost
equilibrium point, labelled F. We slide to the right at fixed density, so n increases.
Since the pressure is fixed, this means T decreases. We move into a region where
heating is stronger than cooling, so L decreases as well, and we conclude that, at
this point (∂L/∂T )P > 0. This point is stable. If we slide left, the conclusion is
the same: density decreases, temperature increases, and the loss function increases
as well.
Clearly point H behaves exactly the same as point F. Point G is a different story.
From there if we slide right n increases, T decreases, and L increases, because
we move into the region where cooling is stronger than heating. Thus this point
is unstable. Moreover, this analysis immediately provides us with the general
principle: anywhere the curve of equilibrium pressure versus density has a negative
slope is unstable. Thus we can identify that the central region of the curve is
unstable.
C. Structure of shocks
A final topic for today is the structure of the shocks produced by thermal in-
stability and other drivers of supersonic motion in the neutral ISM. These turn
out to be somewhat complicated by the presence of ions like C ii and their as-
sociated electrons. The significance of the free electrons and ions is that they
respond to the magnetic field that is present in the gas. We may think of the
gas as consisting of two separate fluids: a magnetically-inert one consisting of
neutral atoms and molecules, and a magnetized plasma consisting of the ions,
electrons, and magnetic field. When the ionisation fraction is low, the neutrals
will encounter one another much more often than they encounter ions, and as a
result they develop a Maxwellian velocity distribution. The ions interact with one
another via long-range Coulomb forces, so they also adopt their own Maxwellian
velocity distribution. The neutral and ions are only coupled weakly, via the rare
ion-neutral scattering events.
Now consider the implications of the presence of this magnetized fluid. Since
the ionisation fraction is low, its density ρi is much lower than the density of the
neutral fluid, ρn , and this gives it a very large signal speed in the ions. The Alfvén
speed is
!1/2
B −1 B 100 cm−3 10−4
vA = √ = 112 km s , (19)
4πρi 5 µG nH x
9
where x is the ion fraction and we have adopted an ion mass of mi = 12mH ,
appropriate if most of the ion mass comes from carbon atoms. In contrast, the
neutral gas has a signal speed equal to the sound speed,
s
kT 1/2
cn = = 0.8 km s−1 T2 , (20)
m
where we have used a mean particle mass m = 1.4mH , appropriate for H and He
in the standard cosmic abundance.
Thus we see that motions at a few km s−1 , the typical warm gas sound speed, are
supersonic with respect to cold neutrals, but are highly sub-Alfvenic with respect
to the ions. Motions at a few km s−1 in the ions are a factor of ∼ 100 slower than
the signal speed, comparable to speeds of ∼ 10 m s−1 in air – comparable to the
speed of a bicyclist. There are certainly no shocks associated with such motions.
The question then arises: what happens in this ion-neutral fluid when gas parcels
collide at speeds that are supersonic with respect to the neutrals but sub-Alfvenic
with respect to the ions? Consider an idealized problem in which we sit at a
stationary interface between an upstream region and a downstream region. Up-
stream the ions and neutrals move together at speed vs into the interface, at which
they decelerate. The magnetic field strength is B0 . Far downstream of they move
away from the interface at some smaller speed vp < vs , and the magnetic field
strength is Bp > B0 . The difference in speeds vs −vp is greater than cn but smaller
than vA .
If the neutral and ion gasses were uncoupled, what happens at the interface would
be clear. The ions would smoothly decelerate across the interface, since the mo-
tions are below their signal speed. The neutrals, since they would not find out
about the change in velocity until they hit the interface, would experience a shock,
producing a sharp jump in velocity and density at the interface. In reality, how-
ever, the ions and neutrals do collide. As a result, the neutrals upstream get
an advanced warning about the upcoming interface via collisions with the ions,
which find out via magnetically-mediated waves.
The results in this case can be calculated rigorously, but we will content ourselves
with an order-of-magnitude calculation that exposes the basic physics. First, we
can estimate the distance in front of the shock over which the ions and neutrals
communicate by writing down the equation of momentum conservation for the
ions. We do so treating them as a separate fluid, with the ion-neutral collisional
coupling handled as a source term. Thus the equation is
B 2 (B · ∇)B
" #
∂vi
ρi + (vi · ∇)vi = −∇pi − ∇ + + nn ni hσviin µ (vn − vi ) . (21)
∂t 8π 4π
Here pi is the ion pressure, B is the magnetic field, v is the velocity for ions or
neutrals as indicated, and the final term represents the rate at which collisions
between ions and neutrals exchange momentum between them. The term in
10
angle brackets is the ion-neutral collision rate coefficient, and µ is the ion-neutral
reduced mass. Taking the simplest case where the velocities are solely in the x
direction, and magnetic field is parallel to the interface (so it is in the yz plane),
this reduces to
dvi dpi 1 d 2
ρi vi =− − B + nn ni hσviin µ(vn − vi ), (22)
dx dx 8π dx
where the v’s now refer to the x components of the velocity.
Since the ions have very little inertia or pressure, the two dominant terms are
the ion-neutral collision term and the magnetic pressure term, and they must
approximately balance:
1 d 2
B ≈ nn ni hσviin µ(vn − vi ). (23)
8π dx
If the region over which the ions and neutrals interact has characteristic size L, the
characteristic magnetic field strength in this region is ∼ B0 , and the characteristic
difference in speeds between ions and neutrals in this region is ∼ vs , then at the
order of magnitude level we have
B02 mn mi
∼ nn ni hσviin vs . (24)
8πL mn + mi
B02
!
1 mn + mi 1
L ∼ (25)
2 4πnn mn mi ni hσviin vs
2
v mn + mi 1
= A,n (26)
2vs mi ni hσviin
vA,n 2 10 km s−1 0.01 cm−3
! !
15
≈ 10 cm , (27)
km s−1 vs ni
where vA,n is the Alfvén speed computed with respect to the neutral gas density
(rather than the ion density), and in the numerical evaluation we have taken
mn = 1.4mH , mi = 12mH , and hσvi ≈ 10−9 cm3 s−1 . Thus we see that the
interaction region gets larger as the magnetic field strength increases (larger vA,n ),
the shock velocity decreases, or the ion density decreases.
The total momentum transmitted by the ions to the neutrals in this interaction
region is, to order of magnitude, just the ion-neutral momentum exchange rate
multiplied by the time it takes the neutrals to traverse the interaction region.
Thus the change in neutral velocity in this region is roughly
µ L vA,n
∆vn ∼ (ni hσviin ) vs ∼ vA,n . (28)
mn vs vs
We may think of the first term as representing the ion-neutral collision rate,
the second as the mean velocity change per collision (where we assume that the
relative ion-neutral speed is of order vs ), and the third as the amount of time for
which a neutral is in the interaction region.
11
Thus we see that if vA,n ∼ > vs , the change in velocity in the interaction region is
comparable to the total velocity change vs required to traverse the interface. In
this case all the momentum transfer required to decelerate the neutrals occurs in
the interaction region, and there is no shock at all. Instead, the momentum is
transferred by ion-neutral collisions. We refer to this case as a C-shock, for con-
< vs ,
tinuous, since in this case all the fluid variables change continuously. If vA,n ∼
the momentum transfer in the pre-shock region is insufficient to decelerate the
neutrals fully, and there is still a shock in the neutrals. We refer to this as a
J-shock, for jump, since the neutral velocity still jumps discontinuously.
[Slide 3 – C and J shocks from Draine & McKee (1993)]
Radiative cooling can further complicate this picture, allowing a third type of
shock, called C*. We will not discuss shocks of this type here.
12