II Neutral Atomic Hydrogen (HI) Regions

This chapter discusses the physics of regions dominated by neutral (or only weakly ionized) atomic
species. Since neutral atomic hydrogen is the dominant species, we generically refer to such gas as
“Neutral Hydrogen” or “HI” Regions, but bear in mind that the gas also contains metals in neutral and
weakly ionized forms that play an important role.

II-1 Interstellar UV & Visible Absorption Lines

The first observational evidence for an all-pervasive ISM came from observations of visible-
wavelength absorption lines. These lines were the principal objects of general ISM studies before
radio and space-borne observations became possible during the 1950s and later.
The strongest visible-wavelength absorption lines are:
NaI 3 2S3 2Po1/2,3/2 5890, 5896Å “D” lines of neutral sodium
CaII 4 2S4 2Po1/2,3/2 3933, 3968Å “H & K” lines of singly-ionized calcium
Both of these are resonance lines arising from the ground state in these ions. Other, weaker, visible
lines of importance (all discovered in the 1930s and 40s) include TiII, CaI, KI, LiI, CH, NH, CN, CH+,
and C2. Notice that there were about as many interstellar diatomic molecules known to early visible-
wavelength studies as atomic species.
The first UV satellites (e.g., Copernicus) and later IUE and HST have observed strong UV absorption
lines from the ISM. Because the typical excitation energies of ground-state resonance transitions are a
few eV, most atomic species have resonance absorption lines (electric dipole transitions out of the
ground state with S=0, L=±1) in the near UV (longward of 1000Å). These include:
MgII 2800Å HI Lyman series
(analog of CaII & NaI lines) (primarily Ly)
OI, OII, OIII, OIV, OVI, OVII H2 Lyman & Werner bands
CI, CII, CIII, CIV
In addition, rare elements like Kr, Ga, Ge, As, Se, Sn, Te, Tl, Pb, Cu, Co, Mn, Zn, and Al are all seen
in (weak) absorption lines. In general, the UV is the best place to study the gas-phase contents of the
general ISM.
The Diffuse Interstellar Bands (DIBs) are the final and most mysterious of the UV/Visible
absorption components of the ISM. Since their discovery by Merrill in 1938, about 200 DIBs have
been identified in stellar spectra, with the strongest appearing at 4430Å. They have not been
identified conclusively with any atomic or molecular species (neutral or ionized). They are
characterized by being extremely broad (by the standards of interstellar absorption lines). Some ideas
are exotic molecular bands, transition from stuff on dust grain surfaces, exotica like ionized Fullerenes
(3-D aromatic C molecules shaped like geodesic spheres), but none have produced consistent
predictions of wavelengths.

Observations of Interstellar Absorption Lines

At high spectral resolutions (R=/104) interstellar absorption lines resolve into narrow absorption
lines that are Doppler shifted relative to each other.

II-1
Neutral Atomic Hydrogen (HI) Regions

For example, towards the star  Orionis, the NaI D lines resolve into 5 distinct radial velocity
components with velocities of [+3, +11.3, +17.6, +24.7, +27.7] km s1 relative to the Local Standard
of Rest (LSR). This observation by Adams in the 1940s was the basis of the “cloud picture” of the
ISM. Some features of the clouds of absorption lines are that the strongest lines are associated with
Galactic rotation, and that there is no dependence of line strength on Galactic longitude.
The distribution of cloud velocities with respect to the LSR seems to be empirically well-described by
a simple exponential velocity distribution:
 (v)  e  v 2 / 
where (v) is the number of clouds seen in the velocity range (v,v+dv), and  is the rms dispersion
among cloud velocities
  v 2  1 / 2
The observed dispersion is ~8 km s1, which is small compared to most populations of stars except for
O and B stars.

Interstellar absorption lines towards the halo star HD93521 observed by Spitzer & Fitzpatrick [1993, ApJ,
409, 299] with HST and the Goddard High-Resolution Spectrograph (GHRS). These spectra show
multiple velocity components and effects of line saturation in different species.

II-2
Neutral Atomic Hydrogen (HI) Regions

Radiative Transfer in Lines

The transfer equation is
dI 
   I   j
ds
The right-hand side consists of 2 parts: the absorption term I and the emission term j.
We can rewrite this in a useful form by defining the Optical Depth
d     ds
and the Source Function
S  j /  
which give us:
dI 
  I   S
d 
Integrating this equation with respect to optical depth gives the formal solution:

I  (  )  I  (0)e    S (  )e  (  ) d 
0
Depending on the circumstances, this is the form we will see used throughout this course.
In this section we are primarily interested in absorption and emission by atoms and molecules in the
neutral phases of the ISM, so we need to introduce two useful quantities: the line absorption
coefficient, lu, and the line emission coefficient, jul.

Line Absorption Coefficient (lu)

The line absorption coefficient describes absorption of a photon by radiatively exciting an electron in
an atom or molecule in a lower energy level (l) into an upper energy level (u):
 lu    d
line
 n  s d  n s
l l lu
line

Here we have introduced the atomic absorption cross-section s=/nl. The line absorption
coefficient has two components:
h ul
 lu   nl Blu  nu Bul 
c
The first term on the right-hand side is the rate of absorption and the second term is the rate of
stimulated emission. Since stimulated emission adds a photon back into the system, it enters as a
“negative” absorption term – absorption because it involves interaction of the atom with the radiation
field. This is to be distinguished from spontaneous emission, which is the electron spontaneously de-
exciting independent of the incident radiation field.
The B’s are the Einstein Coefficients for absorption and stimulated emission, respectively. They are
related via the Einstein Relation in thermal equilibrium
gu Bul  gl Blu

II-3
Neutral Atomic Hydrogen (HI) Regions

The Einstein B coefficients can in turn be written in terms of the Einstein A Coefficient for
spontaneous emission (rate of radiative de-excitation) by
c3
Bul = Aul
8phn ul3
where
8p 2 e 2n 2
Aul = f ul
me c 3
here ful is the emission oscillator strength of the transition which is related to the absorption oscillator
strength via the statistical weights of the levels:
gu f ul = gl flu
By convention there is no Alu term.
This gives us an equation for the atomic absorption cross-section:
kn hn æ ö
d n = ul çç Blu - u Bul ÷÷÷
n
slu = ò sn d n = ò
n c çè nl ÷ø
line line l

hn ul æ n g ö
= Blu ççç1- u l ÷÷÷
c è nl gu ÷ø
Recall from Chapter I that the departure coefficients relate the true level populations (n) to the LTE
level populations (n*) via
nl  bl nl*
and the LTE level populations are related via the Boltzmann Equation:
nu* gu Eul / kT
 e
nl* gl
Putting all the pieces together gives:
æ b ö
slu = sabs çç1- u e-hnul / kT ÷÷÷
çè bl ÷ø

Where we have defined the integrated atomic absorption cross-section, sabs, to be:
hn ul pe 2
sabs º Blu = flu
c me c
Written this way, slu is just the integrated atomic absorption cross-section modified by a stimulated
emission correction expressed in terms of the departure coefficients and an exponential in hv/kT,
where T is the kinetic temperature of the system. The limiting behaviors are instructive:
Case 1: hn ul  kT
In this limit the stimulated emission term vanishes and the line formation is dominated by pure
absorption. Since in the ISM we expect most species to be in the ground state, very few species will be

II-4
Neutral Atomic Hydrogen (HI) Regions

in the excited state, contributing nothing to stimulated emission, so slusabs. This is the situation for
most ISM absorption lines at UV through near-IR wavelengths.
Case 2: hn ul  kT
In this limit stimulated emission becomes very important. If we expand the exponential in the
stimulated emission term to lowest non-trivial order in the exponent we get
 b  h ul 
slu  sabs 1  u 1  
 bl  kT 
this can be re-arranged to give:
æ hn ö é bu kT æç bu ö÷ùú
slu = sabs çç ul ÷ê
÷ - ç -1÷÷ú
çè kT ÷ê
ø ëê bl hn ul çè bl ÷øú
û
This hardly seems much of an improvement until you consider two extreme limiting cases:
In Local Thermal Equilibrium (LTE), bl=bu=1 (by definition), and we find
æ hn ö
slu = sabs çç ul ÷÷÷
çè kT ø

i.e., the effective pure absorption cross section is reduced by a factor of h/kT by stimulated emission.
For example, for HI 21cm absorption at T=100K, h/kT6×10−4.
In Extreme non-LTE:, 1- bl / bu > hn ul / kT , so that

 bu kT  bu 
    1   0
 bl h ul  bl 
Now the “absorption” term has become emissive! This condition occurs in a maser, when the level
populations are driven so far out of TE that nu>>nl.

Line Emission Coefficient (jul)

The line emission coefficient describes radiative transitions from an upper excited level into lower
levels:
jul  
line
j d

This is usually expressed in terms of the Line Emissivity:

4 jul  nu h ul Aul
This has units of erg s1 cm3. The factor of 4 is the number of steradians in a sphere, removing the
usual per-steradian unit from jul. The Einstein Coefficient Aul is the radiative de-excitation rate for
transitions from the upper to lower levels in units of s1.
This latter is the most common form in which we will express the line emissivity in a variety of
situations in the ISM.

II-5
Neutral Atomic Hydrogen (HI) Regions

UV/Visible Absorption Line Formation

In Visible and UV interstellar absorption lines stimulated emission is unimportant because hn ul  kT
for typical interstellar kinetic temperatures. These lines are therefore formed in the “pure absorption”
limit, and the equation of radiative transfer has the simple solution:
I  I ,0 e 

Alternatively, we can express this in wavelength units; since UV and visible-light spectra are usually
plotted in wavelengths (we’ll see frequency again in the radio regime):
I   I  ,0 e  

Ideally, observation of an absorption-line profile can be turned into a measurement of the optical
depth, , for the line species. In practice, however, effects of finite instrumental resolution compared
to the actual line width, limits on signal-to-noise, and so forth are such that we need to express the line
strength in terms of an integrated observable, the Equivalent Width, W, which is independent of
spectral resolution:
I  ,0  I 
W   d
I  ,0
Equivalent widths have units of Å or mÅ in the UV and visible bands. In words:
The Equivalent Width of an absorption line is the width that a line would have if it had a
rectangular profile with zero intensity at the line center.

1
I/I,0

W
0

The “area” of the line here is defined as the integrated area of the absorption profile measured from
the local continuum level, I,0. Note the operative term here is “local”: equivalent widths are defined in
terms of the unabsorbed continuum located immediately surrounding the absorption line of interest. In
practice one does not normally measure the true “global” continuum shape of a spectrum, but instead
estimates the local shape around the spectral features of interest. Note also that all information about
the detailed line profile shape is lost in measuring an equivalent width (e.g., the line in the figure
above is not symmetric).
In effect, this definition of the equivalent width “divides out” the spectrum of the background source.
In practice, equivalent widths are measured by integrating the spectral line numerically after fitting a

II-6
Neutral Atomic Hydrogen (HI) Regions

local “pseudo continuum” using adjacent unabsorbed regions of the spectrum. If the spectrum is very
complex (may stellar and/or interstellar absorption lines close to one another), defining this local
continuum can be problematical. In general, uncertainty in how to measure the local continuum is the
principal source of systematic error in measuring equivalent widths.
The equivalent width is an extremely useful quantity because what matters is the fraction of the light
absorbed, not the total count of the photons absorbed, by the intervening material. By dividing out the
spectrum of the background source, we have normalized the absorption-line profile. The equivalent
width, then, measures the effective area of this normalized absorption-line, integrating over the
detailed line profile shape. As a result, two stars with very different apparent brightnesses and intrinsic
spectra, viewed along the same line-of-sight and path length through and the same interstellar cloud,
will yield the same equivalent widths. In many cases, we will find that relative quantities are more
useful to us than absolute measurements.

The Equivalent Width Curve of Growth

A traditional method of analyzing absorption line data is via the Curve of Growth.
Consider an atom at rest. The absorption coefficient, , for a lower-to-upper transition will be:
o4 gu  u 
  Aul  2 2 
8 c gl
2
  u  (  0 ) 
The term in []’s is the Lorentzian or “Damping Profile” that characterizes the natural broadening of
the line due to quantum mechanical uncertainty. The damping width is characterized by the radiation
damping width, u, for the upper state is defined as the sum of all allowed downward radiative
transitions out of that state:
02
u   Aui
4 c i u
The units of u are typically given in Angstroms or microns as appropriate to the transition being
considered. Each upper level has a different damping width.
An interstellar cloud is an ensemble of atoms all moving about with some combination of thermal and
non-thermal motions (e.g., turbulence, bulk flows, etc.). The natural absorption-line profile is therefore
Doppler broadened by the combination of all of these motions along the line of sight to the observer.
The distribution of line-of-sight velocities is (y), such that

 ( y)dy  1
Here y is the dimensionless velocity parameter, y=v/b, the ratio of the line-of-sight velocity to the
internal velocity dispersion, b. For a Maxwellian distribution of velocities in 3-space, the velocities
project onto the line of sight with a Gaussian distribution:
1 2
 ( y)  ey

For purely thermal motions with kinetic temperature T the Doppler Velocity Dispersion is:
1/ 2
 2kT 
b 
 m 

II-7
Neutral Atomic Hydrogen (HI) Regions

Adding random turbulent velocities with a characteristic velocity dispersion of 2turb gives
1/ 2
 2kT 2 
b   turb 
 m 
[NOTE: be careful not to confuse the dimensionless “bj” departure coefficient with “b” the Doppler
velocity dispersion. Sadly there are only 26 letters in the alphabet and so there has been considerable
re-use – the key to survival in reading the ISM literature is to be careful of carrying forward
definitions from one ISM phase to another and keeping aware of the context. The aggregation of a
century of nomenclature doesn’t make it easy.]
The effect of the non-zero line-of-sight velocity of an individual atom is to Doppler shift the natural
profile’s line center from 0 to (1+v/c)0. The resulting optical depth, , for the ensemble of atoms is
the average of the individual ’s over (y), hence:

   N l    ( y ) ( y )dy


Here Nl is the column density of atoms with electrons populating the lower state out of which we
observe an absorption line:
L
N l   nl ( s )ds
o

This integral is taken along the line of sight (s) between the observer and the background source (e.g.,
a star) at distance L. Writing this out in full detail:
+¥
l04 gu gu
tl = N l ò Aul 2 y ( y )dy
-¥
8p 2
c g l g u + [l - l0 (1 + v / c )]2

This is rather complex, but we can simplify it by introducing four parameters:

y v/b
0b
b 
c
( -0 )
u
b
a   u / b
These are the dimensionless velocity parameter, y, defined as before; the velocity dispersion, b,
rewritten in units of wavelength, b; a dimensionless Doppler parameter, u, and the ratio of the natural
(damping) width to the Doppler width, a. Substituting these definitions and assuming a Gaussian line-
of-sight velocity distribution gives:
04 gu Aul   a  
2
 e y
    N l 3/2    2 dy 
 8 c gl b    a  (u  y ) 2

The first term in []’s above is the optical depth at line center, 0:
04 gu Aul
 0  N l 3/ 2
8 c gl b
The second term in []s is the Hjerting Function, the convolution of the Lorentzian damping profile
convolved with the Gaussian line-of-sight velocity distribution:
II-8
Neutral Atomic Hydrogen (HI) Regions
2
ey



a
H ( a, u )  dy
   a 2  (u  y ) 2
The Hjerting function has no analytic solutions, and it is usually evaluated by numerical integration.
We can, however, gain some useful insights by making a power-series expansion of H(a,u) in a:
H (a, u )  H 0 (u )  aH 1 (u )    a n H n (u )  
The first two terms are:
2
H 0 (u )  e u
1
H 1 (u ) 
u2
H0(u) is a Gaussian profile describing the line core. H1(u) is the first “damping term” that describes
the growth of the line wings (aka the “damping wings”) as the optical depth increases.
Recall that the equivalent width, W, of the line is defined as:
I  ,0  I 
W   d
I  ,0

For pure absorption I   I  ,0 e   this becomes


W  1  e   d

Substituting in the optical depth in terms of the Hjerting Function, we get the useful general form:
Wl = bl ò 1 - e-t0 H ( a ,u )du

where this integral is evaluated over the line-of-sight velocities rather than over wavelength.
What we measure is W, but what we want to derive is the optical depth with wavelength, , which in
turn measures how much of the given species is producing the absorption we see along the line of
sight. This conversion is described by the Equivalent Width Curve of Growth.
There are two limiting cases of interest that describe the properties of the Curve of Growth:
Case 1: The natural width u is much smaller than the Doppler width, b (a = u/b  103).
If the optical depth at line center, 0, is relatively small (<103), then only the first term in the
Hjerting function is important:
2
H ( a, u )  e  u
In this case the line profile is the Doppler (Gaussian) core with no significant contribution from
the damping wings. There are two regimes of interest:
a) Optically Thin ( t 0  1 ):

In this case, we can expand e   to lowest non-trivial order in :


W  1  e   d    d



u 2
 b  0e du


II-9
Neutral Atomic Hydrogen (HI) Regions

Evaluating the integral, and dividing by the Doppler width, b, leads to:
W
0 
b
In this case the equivalent width (in dimensionless units of the Doppler width) grows linearly with
optical depth 0. We call this the Linear Part of the Curve of Growth.
b) Optically Thick, but t 0 < 103 :
The lines are optically thick, but not so optically thick (in the limit a<103) that the damping wings
become important. In this case, we have:



W u 2
 1  e  0 e du
b 

This equation is not analytically integrable, but the limiting behavior is

u 2 1 for u small (line core " saturates" )
1  e  0e 
0 for u large
The turnover point occurs when
2
 0 e  u  1, or equivalently :
0

ln  0  u 02  0
 u 0  ln  0
Hence
W
 2u 0    2 ln  0
b
In this regime the equivalent width of the line grows very slowly even with large changes in the
optical depth. We call this the Flat Part of the Curve of Growth.
Case 2: Large optical depth 0 and large column density Nl
In this case, we can no longer ignore contributions from the damping terms in H(a,u). For
example, HI Lyman series lines, like Ly, have a103 but it is very abundant so that the column
density is very high. In this case
 2 a 
    0 H ( a , u )   0 e  u   
 u 2

The damping wings become important when
a 2
 e u
u 2

For Ly, damping becomes important for u29.8 by this criterion.

Because the line core has become “saturated”, all subsequent “growth” of the line equivalent
width will be due to the increasing contributions from the unsaturated line wings far from line
center, hence:

II-10
Neutral Atomic Hydrogen (HI) Regions

a
 0
u2
The equivalent width is then:
 

 
W u2
 (1  e   ( u ) )du  (1  e  0 a / )du
b  

Rewriting this by setting x 2   0 a /  u 2 gives:

1/ 2
W   0 a  

 (1  e  x )d  1x 
2
 
b    

This is analytically integrable, hence

1/ 2
W  a 
 2   0 
b   
Here the equivalent width of the line grows like the square root of the optical depth and we say we
are on the Square-Root part of the Curve of Growth.
These three regimes describe the behavior of the Equivalent Width Curve of Growth. At the transition
regions, we of course need to evaluate the integrals numerically. A Curve of Growth computed
numerically for interstellar HI Ly is shown in the figure below, illustrating the three regions (linear,
flat, and square root) derived above. In addition, we show spectra of the interstellar NaI D lines that
exhibit the range of behaviors.
The most useful measurements occur at the extremes of the curve in the linear and square root parts of
the curve. In the flat part, between the onset of saturation of the line core and the onset of significant
growth of the damping wings, large changes in optical depth lead to very small changes in the
measured equivalent width. Thus nearly saturated but undamped lines provide little useful data on the
line-of-sight column densities.

Equivalent Width Curve of Growth for HI Ly, showing the main regions derived in the text

II-11
Neutral Atomic Hydrogen (HI) Regions

Interstellar NaI-D absorption lines from Welty et al. [1994, ApJ, 436, 152]. These profiles show a
mix of linear ( Cyg), flat ( Per &  Ori), and square-root ( Oph) absorption lines.
Practical Considerations:
In practice, curve of growth methods are quite powerful, as they can relate the product Nf to a direct
observable, W, that is relatively insensitive to the choice of spectral resolution or fine details of the
spectroscopic experiment. In principle, two different spectrometers working at very different
resolutions and on different telescopes with different detectors should be able to measure the same
equivalent widths to within the irreducible measurement uncertainties. However, because the
equivalent width integrates over the detailed line profile shape, we do lose some information that
might be useful.
Real interstellar absorption lines are often highly structured with a mixture of both saturated and
unsaturated components because the line of sight to a particular star will often intersect a number of
interstellar clouds with a wide range of column densities. While technically the equivalent width is
insensitive to the spectral resolution (modulo effects of signal to noise which affects mainly the
contrast of the line against the continuum), at lower spectral resolutions the saturated and unsaturated
components will be blended, making interpretation of the composite line’s equivalent width in terms
of a single column density problematic at best.

II-12
Neutral Atomic Hydrogen (HI) Regions

In the case where lines are heavily saturated or show measurable damping wings (e.g., damped HI
Lyman absorption systems), the equivalent width curve of growth method is unreliable. One cannot
tell where the continuum should be placed, which leads to large systematic errors in measuring W. In
this case, various alternative methods have been used, for example the “Continuum Reconstruction
Method” described by Bohlin et al. (1975, ApJ, 200, 402).
At very high spectral resolutions (R>104), an alternative is to use the observed line profile and fit
models to account for instrumental effects and the contributions from multiple absorption components
with different column densities. The problem here is a lack of a complete database of UV spectra with
sufficient resolution to employ these methods. One usually has to fall back on making assumptions
about the intrinsic line profile shape.
The intermediate case occurs when the lines are fully resolved (i.e., when the line width is larger than
23 instrumental widths), but not necessarily resolved into fine velocity structure, a particularly
powerful class of techniques has emerged to deal with these data. These techniques use direct
integration of the observed optical depth profiles. Such methods make no a priori assumptions about
either the detailed line shapes or the velocity distribution of the gas (unlike the case with the curve of
growth, continuum reconstruction, and line profile modeling approaches).
A particularly good example of this type of analysis is the “Apparent Optical Depth Method”
described by Savage & Sembach (1991 ApJ, 379, 245). This method does an excellent job of allowing
discrimination of saturated structures in velocity space. When many different species are present in a
spectrum, it provides a complete N(v) profile by letting the different unsaturated parts fill in the gaps
left by saturated species. This method provides a modern alternative to the traditional Curve of
Growth analysis, and has been used in a number of recent absorption-line studies of the ISM.
Despite the practical caveats, the curve of growth still permits us to address a number of problems
quantitatively in a way that illuminates what can be learned from IS absorption lines. The new
methods lend us a greater degree of measurement precision, but no additional basic physical insight.

Applications of Interstellar Absorption Lines

1) Interstellar Ly
The HI Lyman series lines are UV resonance lines that arise in radiative excitations out of the n=1
ground state of HI. First seen in the ISM in the early 1970s with the launch of the first UV satellites,
they are seen along every line of sight, from early-type stars to late-type stars with chromospheric
activity (you need some UV flux to observe them).
The optical depth of the HI Lyman series is very high, and on the damping (square-root) part of the
Curve of Growth for most interstellar clouds, hence:
1/ 2
W   0 a 
  2 
b   
The optical depth at line center, 0, is given by
04 gu Aul
 0  Nl
8 3/ 2 c gl b
From this result and the definition a=u/b, the Doppler width, b, cancels out, and we can write the
equivalent width of the line as a function of the column density, Nl alone:

II-13
Neutral Atomic Hydrogen (HI) Regions
1/2
æ l4 g ö
Wl = ççç N l 0 u Aul gu ÷÷÷
è 2pc gl ÷ø

The u’s are computed by summing over all the A’s for the downward transitions out of the upper
excited state of the particular absorption line of interest. For Ly, which is the 2p1s line, there is
only 1 term in the computation of 2p (only one place for the electron to go, back to 1s). For Ly
(3p1s) there are two terms (3p1s and 3p2s are possible radiative decay channels), and so forth.
Putting in numbers for Ly, we can solve for the column density along the line of sight as a function
of the equivalent width:
N Ly  1.867  1018 W2 (cm -2 )

for W in units of Angstroms. Observations of stars at d100pc give W(Ly)10Å, which implies a
column density of Ly absorbers of NLy1.91020 cm2. A line of sight of 100pc31020 cm implies
a mean hydrogen column density of nH0.6 cm3. For nearby stars, however, nH0.1 cm3 or less is
typical, indicating that we reside in a “local bubble” characterized by a lower average density.
We can also estimate the mean abundance of Deuterium in the ISM by measuring the Lyman-series
absorptions from DI. The DI Lyman lines are shifted blueward of their respective HI counterparts by a
small isotopic shift due to the neutron in the nucleus (the Rydberg constant is proportional to the
reduced mass of the nucleus). The table below gives the isotopic shifts for the first 3 Lyman series
lines of HI and DI:
HI DI 
Ly 1215.67Å 1215.34Å 0.33Å
Ly 1025.72Å 1025.44Å 0.28Å
Ly 972.54Å 972.27Å 0.27Å
Since HI Ly is so strongly saturated in the ISM, with very strong damping wings (W10Å), the DI
line is lost in the saturated HI line core. You therefore need to measure the DI lines associated with
higher-order HI Lyman lines, like Ly, Ly, etc., that have less strongly damped lines. For example,
measurements of HI and DI absorption along the line of sight to  Centauri (where there is no detected
H2 absorption), one measures D/H(1.40.2)105, smaller than 2104 measured on Earth from the

Interstellar DI and HI Ly absorption lines towards the white dwarf star WD0621-
376 observed with FUSE. From Lehner et al. (2002, ApJS, 140, 81).

II-14
Neutral Atomic Hydrogen (HI) Regions

solar wind or ocean water. You can also measure HD molecular absorption bands (the analog of the
H2 Lyman bands), but it is hard to convert from HD/H2 to D/H.
The Copernicus satellite (mid 1970s) was the first UV satellite that was sensitive to the higher-order
Lyman series lines in the Far-UV (most UV satellites, except EUVE, were not sensitive below about
1100Å). Subsequent Far-UV studies have relied on either sounding rockets or short-duration space
missions (e.g., ORFEUS-SPAS I and II, or IMAPS on the Shuttle). Lyman/FUSE (launched June
1999) is the first long-duration satellite mission since the Copernicus satellite 1970s to work in this
Far-UV region (and is ~105 times more sensitive). So far, it has provided a number of good
measurements of D/H in the local ISM from HI and DI Lyman series lines out to distances of 100pc
from the Sun (and a few sight lines out to ~1kpc with IMAPS). The data show a pretty constant
D/H1.21.7×105, with the dispersion in values increasing with distance from the Sun.
2) ISM Gas-Phase Abundances
So many different species produce UV absorption lines in the ISM, from Hydrogen to rare metals, that
we can get a pretty clear picture of the relative gas-phase abundances of the various elements from UV
absorption-line studies. Among the scientific questions these permit us to address are chemical
evolution of the ISM (in particular the mix of elements from different nuclear processes like -
process, neutron capture and proton capture) and the depletion of refractory elements onto dust grains.
The current status of UV absorption-line abundances, especially as learned from the Hubble Space
Telescope GHRS, has been reviewed by Savage & Sembach (1996, ARAA, 34, 279). A table of
measured sight lines and the elements seen along them is reproduced on the following pages. The
lines of sight to 2 stars,  Ophiuchus and  Persei, are particularly rich in interstellar absorption lines.
These sight lines cross through regions of high column density, and many rare metal species (Cu, Zn,
etc.) are very well studied there. Examples of spectra along the  Oph sight line are reproduced on the
next page. Other sight lines have at least 5 or more different atomic species available for study.
A particular achievement of GHRS, which flew from 19901996, was the detection of unsaturated CII
absorption (e.g., CII]2325Å absorption). In previous studies, only one line of sight, towards 
Scorpii, had unsaturated CII lines, the rest were all on the flat part of the curve of growth and produced
no useful information (limits only). As a result, GHRS has permitted the first definitive interstellar
gas-phase carbon abundances.
A particularly important result of the HST/GHRS data has been to greatly refine measurements of the
depletion of gas-phase elements onto dust grains for sight lines that pass through both warm and cold
phases of the ISM in both the disk and the halo of the galaxy. We will treat this subject in more detail
later in the course when we discuss the properties of dust grains.
3) Thermal Balance
The CII 1334.5,1335.7Å doublet is produced by absorptions out of the C+ ground state, which is
split into 2 fine structure levels, 2 P3/2
o
and 2 P1/2
o
, into the same upper level. The strengths of these

2 3/2
D 5/2

1334.5Å 1335.7Å

2 o 3/2
P 1/2
158m

II-15
Neutral Atomic Hydrogen (HI) Regions

absorption lines give the relative populations in each of these levels. This is of particular interest as
transitions within this fine structure ground state are responsible for producing the [CII]158m
emission line. Morton’s 1975 observations of the CII 1334.5Å and 1335.7Å ratio showed a
significant population of atoms in the 2 P3/2
o
state (the upper most of the fine structure level). This
suggested that [CII]158m line could be a strong cooling line in HI regions. It was not until the 1980s
that the Kuiper Airborne Observatory detected Far-IR [CII] emission from the ISM, as predicted by
Morton.

II-16
Neutral Atomic Hydrogen (HI) Regions

The HI Hyperfine Structure Line

Most of our knowledge of the distribution of neutral atomic hydrogen in the ISM of the Milky Way
and other galaxies come from observations of the strong 21-cm line. This line arises from transitions
between the hyperfine structure levels in the ground state of Hydrogen, and is seen in both emission
and absorption. This section reviews the physics of the HI hyperfine transition and of HI 21-cm line
formation in both absorption and emission.

Hyperfine Splitting of HI: A little light quantum mechanics.

The 1s2S1/2 ground state of HI is split by the interaction between the magnetic moments of the proton
and electron spins.
The electron magnetic moment is:
 
e  ge B J
where:
g e  electron Lande factor   2.00232
e
B  Bohr Magneton 
 2me c  
J  electron quantum number  L  S
The proton magnetic moment is:
 
 p  gN N I
where:
g N  proton Lande factor  5.58 (experimentally)
e
 N  Nuclear Magneton 
2m p c
 
I  proton quantum number, I  1/ 2

A way to think of the Landé factors are that they express the difference between the classical and
quantum descriptions of the magnetic moment of a sphere of uniform mass and charge density. The
different Landé factors reflect the fact that the electron is a single particle with charge 1, but the
proton is a combination of 3 quarks with fractional charges that add up to +1. The quantization of the
coupling of the ’s results in the HI ground state being split into two states corresponding to parallel
and anti-parallel spins of the electron and proton. The difference in energy, W, between the ground
state and the hyperfine levels is given by:
 
W  e   p

Following Condon & Shortly, for the 2S1/2 terms, this is exactly
8   I  1 
W  e  p 1, 
3   I 
For HI, with I=1/2, this becomes
8
W   e  p 1,3
3

II-17
Neutral Atomic Hydrogen (HI) Regions

The result is that the hyperfine levels are split into two levels above and below the 1s ground level, as
shown schematically below.


+1
2
1s S1/2

 3

Schematic of the HI ground state hyperfine splitting, showing the electron and proton spins.

The hyperfine states are assigned a hyperfine quantum number, F:

  
F I J
by analogy with L-S coupling for fine structure states. For the 1s 2S1/2 state of HI is F=(1/21/2)=(1,0)
for the upper and lower states, respectively. The hyperfine degeneracy, gF=(2F+1) is gF=(3,1) for
upper and lower hyperfine states, respectively. The energy difference, E, for a hyperfine state with
quantum numbers (n,J,L,I,F) is given for any hydrogenic atom with nuclear charge Z and atomic
number A by:
h 0 Z 3  F ( F  1)  I ( I  1)  J ( J  1) 
E 
n3  J ( J  1)(2 L  1) 

m 
 0  g N 2cRA  e 
 mp 
2 2 e4 me  mA 
RA   
ch3  mA  me 
These properties are illustrated schematically below:
F=1
Eu gF=3
1s 2S1/2

El

F=0
gF=1

Arrangement of the hyperfine energy levels in the HI ground state.

For the F=10 transition in HI (Z=1, A=1), the total energy splitting (Eu+El) in frequency units has
been measured in the lab using Hydrogen masers to be:
  1,420,405,751.786  0.010 Hz

II-18
Neutral Atomic Hydrogen (HI) Regions

This is one of the few times you can quote a number in astrophysics to that kind of precision! In more
convenient frequency and wavelength units, however:
  1420.4 MHz
  21.106 cm
The F=10 transition is a magnetic dipole transition, with transition probability
4 2 e 2 h  1  1  2
A10   3  0 M 1
3m e c    2 F1  1
2 2

 2.85 10 15 s -1

This relatively low transition probability gives a mean lifetime of the state of 1.1107 years. The
natural width of this transition is
A10
 10   4.5 10 16 Hz
2
Compared to the thermal Doppler width expected from an ensemble of HI atoms with kinetic
temperature of 100K:
3kT
 th   0  4700 Hz
mc 2
For most astrophysical conditions, we expect the width of the HI 21cm line (either in emission or
absorption) to be dominated by the Doppler width due to thermal and/or turbulent motions.

Level Populations
We expect the level populations in the HI hyperfine levels to be far from LTE in most cases. As a
starting point, the LTE populations are given by:
n1* g 1  h / kT
*
 e
n 0 g0
where “1” and “0” refer to the hyperfine quantum number, F, of the states, T is the kinetic temperature
of the HI cloud, and  is the frequency of the hyperfine transition. The non-LTE level populations are
expressed in terms of an excitation temperature, TS, which is traditionally called the “Spin
Temperature” of the system:
n1 g 1  h / kTS
 e
n0 g 0
Putting in the values of the g’s, and using hn / k = 68 mK , the transition energy in units of Kelvin:
n1
= 3e-0.068 / TS
n0
For most astrophysically interesting conditions, TS68mK, and so we have to rely on subtle
absorption effects in the line formation physics which, as we shall see, are very sensitive to small
deviations in n1/n0.

II-19
Neutral Atomic Hydrogen (HI) Regions

Optical Depth
In general, as we saw previously, the optical depth in a transition is given by
 40 g 1  10
   N0
  ( y) 2
8 c g 0
A10 2
 10  (   0 ) 2
dy

where
 0   0 (1  v / c)
Integrating over wavelength gives the integrated line optical depth commonly used in radio astronomy

04 g1
  d   N0
0
8 c g 0
A10

This equation relates total HI absorption to the column density of HI along the line of sight, N0.
Just to be sure of confusing everything, the traditional way of writing this in radio astronomy is to
recast the distribution of velocities, (y), in frequency units:
 ( y )dy  f ( )d
and so we can write the optical depth at a particular wavelength, , as
 40 g 1 d
   N0 A10 f ( )
8c g 0 d
since d/d=c/2 (eliminate the  sign by integrating backwards) we can write:
3 20
   N0 A10 f ( )
8
or, finally purging the mixed wavelength/frequency units in favor of frequencies:
3 c2
  N 0 A10 f ( )
8  2
where in all of the equations above, N0 is the column density of HI in the F=0 (unexcited) hyperfine
state. The gF degeneracy values (3 and 1, respective for F=1,0) tell us that the mix of states in thermal
equilibrium should be (modulo a TS correction <<1):
1
N0  N HI
4
3
N 1  N HI
4
So that:
3 c2
  N HI A10 f ( )
32  2
This is the form of the HI 21cm optical depth most often used by radio astronomers.

II-20
Neutral Atomic Hydrogen (HI) Regions

Stimulated Emission
The calculation of the optical depth in the previous section was assuming the pure absorption case.
However, while stimulated emission is insignificant at UV, optical, and near-IR wavelengths, it is
very important at radio, mm, and (sometimes) far-IR wavelengths.
The effect of stimulated emission is for the radiation field to induce downward transitions from the
upper excited states at a rate proportional to the local density of photons. This adds photons in the
direction of the radiation field, making these photons coherent.

F=1

F=0

Schematic of Stimulated Emission

The intensity of the stimulated emission component, ISE, is proportional to the intensity of the local
radiation field:
I SE  nu Bul I   n1 B10 I 
The absorption component is also proportional to the local radiation field:
   nl Blu I  n0 B01 I
The B’s are related via the statistical weights:
gu Bul  gl Blu
So that, the net upward transitions (pure absorption out of F=0 less stimulated emission out of F=1 to
F=0) is thus:
 g 
    n 0 B01  n1 0 B01  I 
 g1 
 n g 
 n 0 B01 I  1  1 0 
 n0 g1 
Recall that the relative non-LTE level populations, n1/n0, were written in terms of the spin
temperature, TS, as a Boltzmann-like equation
n1 g 1  h / kTS
 e
n0 g 0
hence
     1  e  h / kT S

Recall now that h/k=68 mK << TS, so expanding the exponential to lowest non-trivial order gives:
 h 
      
 kTS 

II-21
Neutral Atomic Hydrogen (HI) Regions

To account for stimulated emission, we must multiply the optical depth derived in the pure-absorption
case (above) by a correction factor h/kTS, hence:
3 hc 2 N HI
  A10 f ( )
32  0 kTS
n HI f ( )
 7.95  10 3

los
TS
ds

Here I’ve substituted the definition of the column density of HI, NHI, in terms of the integral of the HI
density, nHI, along the line of sight

N HI 
n
los
HI ds

and put in the values of the physical constants, nHI is in units of cm3, TS is in K, and ds is in parsecs.
Note that in general the HI density, spin temperature, and distribution of cloud velocities f() are all
functions of position, s, along the line of sight.
21-cm Line Formation
HI 21-cm Emission Lines
The equation of radiative transfer is
dI n j
= -I n + n ;
d tn kn
d tn = kn ds
which has the solution:

I   j e  ( s ) ds;
0
s
 ( s)    ( s)ds
0

In radio astronomy, it is conventional to express the observed intensities in units of a Brightness

Temperature, TB (n ) , because at =21 cm, we are in the Rayleigh-Jeans limit:

2k n 2
In = TB (n )
c2
The radiation is not really blackbody, but the R-J limit allows us to express any I in terms of an
equivalent temperature for a blackbody that would give the same intensity. Brightness temperatures
are actually what radio astronomers measure in their receivers at radio wavelengths, unlike the case in
UV, visible, and IR wavelengths where (ideally) we count incident photons.
The emission coefficient expressed in temperature units (K/cm), J, is given by
c2
J  j
2k 2

II-22
Neutral Atomic Hydrogen (HI) Regions

Kirchoff’s Laws relate  and J through the spin temperature, TS by:

J     TS
Multiplying the transfer equation by a factor of c2/2k2, and re-arranging gives:
dT B
 TS  T B ( )
d 
for the Rayleigh-Jeans limit. This equation has the solution:

TB ( )   TS  e  ( s ) ds
0

  TS e  d

where
s
 ( s )    ( s)ds
0

Thus  is the optical depth to infinity along the line of sight, while  is the optical depth to distance s
along the line of sight.
Assuming that TS is constant along the line of sight (i.e., that we have an isolated isothermal cloud),
the solution of the radiative transfer equation reduced to

TB ( )  TS 1  e  
In general, however, both TS and  will be expected to vary with position, and solutions of the full
transfer equation are required (they are not pretty).
Recall from our earlier discussion of UV and visible lines that in the limit h<<kT, we found
 h 
  nl slu  nl sabs  
 kTexc 
where sabs is the integrated atomic absorption cross-section for the transition. In this limit, the
absorption is nearly completely compensated for by stimulated emission out of the excited state, so
that the effective absorption coefficient depends on small variations of n1/n0 with spin (excitation)
temperature, TS. Numerically (from Kerr 1968 in Ch 10 of Middlehurst & Aller):
TS (K) n1/n0
10 2.9806
100 2.9981
1000 2.9998
These numbers show that as the spin temperature rises, the ratio of the hyperfine level populations
approaches the ratio of the statistical weights (3.00). For the purposes of deriving column densities
from HI line measurements, this means that assuming n1/n0=3 is a reasonable approximation.

II-23
Neutral Atomic Hydrogen (HI) Regions

Writing  in terms of the transition probability A10:

c2 g1  h 
   n0 A10   f ( )
8 2 g0  kTS 
Since nHI=4n0, the optical depth is:
N HI c 2 g1  h 
s
    ds  A   f ( )
4 8 2
10
0
g 0  kTS 
NHI is the HI column density. Putting in the numbers for the HI 21-cm transition:
N HI
   2.59  10 15 f ( )
TS
Solving for NHI:

N HI  3.88 10  T  g ( )d cm
14 -2
S


(the function g() is the 1/f(), but both are arbitrary placeholders for “line profile” in different units,
so the notation can be similarly arbitrary). Since emission lines are often observed in frequency bins or
“channels”, radio astronomers will often derive the column density in a particular frequency bin:
N HI ( )  3.88  1017 TS    KHz cm -2
Alternatively, one often sees the column density measured in a particular radial velocity bin:
N HI (v)  1.82  1018 TS  (v)v kms cm -2
Here (v) is the optical depth in the radial velocity interval [v,v+dv].
To measure the column density of HI along the line of sight, we need to measure TS and , but the
only observable is TB (n ) . To see how optical depths, and hence column densities, are measured in
practice, it is useful to examine two limiting cases:
Case 1: Optically Thin Limit, (v)<<1
In this case,
TB (v)  TS  v
so that
N HI (v)  1.82  1018 TB (v) in 1 km/sec bins
or
N HI (tot )  1.82  1018 TB (v)dv
line

where the integral is often written as the “line intensity” expressed in units of “K km s1” (the
somewhat odd unit you will find in many HI radio observation papers, especially older papers).
Most extragalactic observations of the total HI content of galaxies determined from integrated 21-
cm line measurements assume the optically thin case, and this is justified except for edge-on
galaxies viewed in their mid-plane or large gas-rich merger remnants where lines of sight through
the thickest parts of the galaxy are optically thick. It is also true of many, but not all, sightlines

II-24
Neutral Atomic Hydrogen (HI) Regions

through the disk of our own Galaxy – the exception being sight lines within a few degrees of the
Galactic Center which are all optically thick.
Case 2: Optically Thick Limit, (v)>>1
In this case
TB  TS
This happens because 21-cm photons emitted inside the cloud are absorbed within the cloud, and
only those photons emitted from within 1 optical depth of the cloud surface escape. In this case the
observed TB is independent of column density, and depends on TS. The column density must
therefore becomes:

N HI  1.82 1018 TS   (v)dv

where now we must integrate the optical depth over velocity instead of integrating over the
observed emission-line profile in brightness temperature.
Line Broadening
Because the natural width, 10 is so small, Doppler broadening will dominate the line profile. This
broadening takes two forms:
Thermal Broadening within a single cloud:
 kT 
 D2  2 
 H 
m
in which case the optical depth profile will be:
2
 ( )   0 e  (  0 /  D )
Bulk Motions of individual clouds:

 ( )  i
0,i e
 (  0 /  D , i ) 2

Here the D,i are the thermal Doppler widths of the individual clouds.
In practice, we often add an arbitrary turbulent velocity term in quadrature with the thermal term to
specify the line width, much like what is done at UV and visible wavelengths with interstellar
absorption lines.
The narrowest HI 21cm absorption lines seen in the Galaxy are roughly Gaussian in shape, but most
emission lines are decidedly non-Gaussian, appearing as a superposition of many blended Gaussian
components.
HI 21-cm Absorption Lines
Consider a background radio continuum source (e.g., a quasar or radio galaxy) with a brightness
temperature Tbg. The solution of the equation of radiative transfer in the R-J limit is thus:

TB ( )  Tbg e 
  TS e  d
0

For a single, isothermal cloud, this reduces to


T B ( )  Tbg e   TS 1  e  
 (absorption)  (emission)

II-25
Neutral Atomic Hydrogen (HI) Regions

We have 2 unknowns to measure from these data, TS and . Three observational methods are used:
Method 1: Beam Switching
TB() is measured at two positions: on and off the source. The background source is also
measured at adjacent line-free frequencies and interpolated to get Tbg at the line frequencies
(“measuring the baseline”).
In the on-source beam, we measure
 
TB ( ) ON  Tbg e   TS 1  e   Tbg 
 TB  Tbg

In the off-source beam, we measure


TB ( ) OFF  TS 1  e  
We can then solve these two equations for TS and :
 T B ( ) OFF  T B ( ) ON 
    ln 1  
 Tbg 
 T B ( ) OFF 
TS  Tbg  
 T B ( ) OFF  T B ( ) ON 
The success of this beam-switching technique depends on the absorption region varying smoothly
over very small scales, so that TS and  are not substantially different on the on- and off-source
beams.
Method 2: Observe Variable Sources
Observe a background source that is time variable, for example, a pulsar that turns off, a variable
radio quasar or BL Lac object, etc. Pulsars tend to be very weak at 1420MHz, so this has only
been done for ~40 of the 1000-odd known radio pulsars.
This provides you with a measurement of the TS and  at the same physical location, obviating
the assumption that the absorbing cloud varies smoothly between beam positions as in the beam-
switching technique.
Method 3: Interferometry
Interferometric techniques (e.g., with the VLA) average out the diffuse emission, giving a direct
measurement of Tbg e  . Another way of stating this is that interferometry acts as a spatial filter
that resolves out most of the structure in the diffuse emission from the absorption-line cloud.
Interferometers are only sensitive to structures on scales of B/ (the spatial frequency), where B is
the baseline of the interferometer. The “imaging” capabilities of the antenna array are not used
(e.g., one often observes by “phasing” the VLA antennas, reducing the array to a 3-element
interferometer to boost sensitivity). Most Galactic HI clouds do not show significant structure on
<1’ scales, whereas most [useful] background radio sources have typical angular scale sizes of an
arcsecond (or less). Interferometry is a very powerful technique for probing HI absorption along
lines of sight to radio quasars.

II-26
Neutral Atomic Hydrogen (HI) Regions

HI 21cm emission and absorption profiles towards 8 extragalactic radio sources from Radhakrishnan
et al. (1972, ApJS, 24, 15). The dashed lines are fits to an optically thin emission-line component
without a corresponding absorption component. The vertical lines delineate the velocity limits of the
optically-thick absorption components. These are beam switching observations with the Parkes
Telescope, so the vertical axis is TB(v). The narrow absorption lines correspond to discrete CNM
clouds, and the broad optically-thin emission is diffuse WNM (intercloud) gas.

Excitation of 21-cm Emission

How do you excite HI into the excited hyperfine structure level?
The competing processes for determining the population of the upper excited hyperfine structure level
are Collisional Excitation and Collisional De-excitation.
X  H ( F  0)  X  H ( F  1)
Because the lifetime of the excited state is so long (t01.1107 years), radiative de-excitations are
relatively rare. Collisions with other HI atoms and electrons dominate.
The cross-sections for collisional excitation/de-excitation are:
Q01 (v0 ) : excitation, with threshold hn / k = 0.068 K
Q10 (v1 ) : de-excitation (no threshold)
Energy conservation demands
1 1
mv 02  mv12  h ,
2 2
2 h
with threshold : v0 
m
Note also that electrons have a much larger speed than H atoms by a factor of (mp/me)1/2  45.

II-27
Neutral Atomic Hydrogen (HI) Regions

The Principle of Detailed Balance

In LTE, any microscopic process is precisely balanced by its inverse, thus collisional excitation with
rate
n0*nx f (v0 ) v0Q01 (v0 )dv0
is precisely balanced by collisional de-excitations with rate
n1*nx f (v1 ) v1Q10 (v1 )dv1
Energy conservation demands that
v 0 dv 0  v1 dv1
so that the collisional equilibrium condition (no radiative processes) becomes:
n 0* f (v 0 )Q01 (v 0 )  n1* f (v1 )Q10 (v1 )
For a Maxwellian velocity distribution
2
f (v)  v 2 e  mv / 2 kT

which means that the equilibrium condition becomes

2 2
n 0*v 02 e  mv0 / 2 kT Q01 (v 0 )  n1*v12 e  mv1 / 2 kT Q10 (v1 )
In LTE, the Boltzmann equation tells us
n1* g 1  h / kT
*
 e
n 0 g0
but now
1 1
h  mv 02  mv12
2 2
2 2
 e  h / kT  e  mv0 / 2 kT  e  mv1 / 2 kT
Thus the exponentials cancel, leaving us with the Milne Relation:
g 0 v 02 Q01 (v 0 )  g 1v12 Q10 (v1 )
Since h/k=68mK is small for typical kinetic temperatures T, v0v1, and we have:
1
Q10 (v)  Q01 (v)
3
Electron Collisions
Electron collisions dominate the excitation balance in HI if the electron density, ne>0.03nH.
H ( F  0)  e   H ( F  1)  e 
There are two modes:
1. Spin-Flip Collisions due to the magnetic field of the passing electron flipping the spin of the
bound electron. This has a very small cross-section and is thus rare.

II-28
Neutral Atomic Hydrogen (HI) Regions

2. Electron-Exchange Collisions in which the incoming electron is captured and the bound electron
is kicked out. If the spins are opposite, the process looks schematically like this:
p e e p e e
      
The cross-section for this process is Q10=4.71016 cm2 at T=100 K.
H-Atom Collisions
If the electron density is very small (ne<<nH), Hydrogen-Hydrogen spin-change reactions will
dominate the collisional excitation/de-excitation of the hyperfine states:
p e p e p e p e
        
or
H ( F  0)  H ( F  0)  H ( F  1)  H ( F  1)

Mixed collisions result in no net change in excitation. The process was first described by Purcell &
Field (1956, ApJ, 124, 542), with subsequent re-calculation by Allison & Dalgarno (1969, ApJ, 158,
423). The cross-section for this process is Q1061016 cm2 at 100K, and is a complicated function
of the kinetic temperature (e.g., Allison & Dalgarno 1969, figure 2).
Collisionally-Dominated Excitation
The collisional de-excitation rate is
n1 n X v1Q10 (v1 ) Maxwellian
 n1 n X q10 (T )

Here we have introduced a new quantity, q10(T), the Collision De-excitation Coefficient, which has
units of cm3 s1. The collisional excitation rate can be written similarly in terms of Collisional
Excitation Coefficient q01(T):
n 0 n X q 01 (T )
The q’s are related by the integral of the Milne Relation:
g1 -hn / kT
q01 (T ) = q10 (T )´ e
g0
= 3e-0.068/ T q10 (T )
where T is the kinetic temperature of the gas.
In the collisional limit (radiative processes are unimportant) detailed balance only includes the
collisional rates:
n0 n X q 01 (T )  n1 n X q10 (T )
The Milne Relation above gives the ratio of the q’s as a function of the Kinetic Temperature, and The
ratio of the level populations is given in non-LTE by
n1 g 1  h / kTS
 e
n0 g 0

II-29
Neutral Atomic Hydrogen (HI) Regions

where TS is the Spin Temperature1. Substituting both into the equation of detailed balance, we obtain
the result
TS  Tkin
When collisions dominate the excitation, the Spin Temperature is driven to the Kinetic Temperature
of the gas. In such circumstances we say that the level populations have “thermalized”.
Radiatively-Dominated Excitation
The opposite limit occurs at low densities when collisions are unimportant relative to radiative
processes in the excitation equilibrium. In this case, the equation of detailed balance becomes
n1 A10  4J  n1 B10  4J  n 0 B01
where 4J is the radiation field density. With the HI 21-cm line we are in the Rayleigh-Jeans limit for
most radiation fields, and so we can express the radiation-field density in terms of TR, the Color
Temperature of the radiation field:
2 h
J  kTR
c3
From the Einstein Relations, the stimulated emission coefficient, B10 can be expressed in terms of the
radiative transition probability, A10:
 kT  T 
4J  B10  A10  R   A10  R 
 h   T* 
where
T*  h / k  68 mK
Using the Einstein relation g0B01=g1B10 the equation of detailed balance in the radiation-dominated
case becomes:
g1
n1 A10  n1 A10 T R / T*   n 0 A10 T R / T* 
g0
This can be solved for the ratio of the level populations:
n1 TR / T* 
3
n0 1  TR / T* 
Recalling that the non-LTE (“true”) level populations are given by:
n1
 3e 0.068/ TS  3e T* / TS
n0
and given that we are in the limit that TR  T* the relative populations reduce to

n1  T 
 31  * 
n0  TR 
1
In keeping with the convention we established in Chapter I this should be called the "Excitation Temperature", but it is
traditional to refer to it as "Spin Temperature" in radio astronomy. This is a recurrent theme we will encounter in ISM studies:
each specialty has its own traditional nomenclature and notation that makes crossing between disciplines a challenge.

II-30
Neutral Atomic Hydrogen (HI) Regions

However, TS  T* , so we also have from the Boltzmann equation

n1  T 
 3e T* / TS  31  *  
n0  TS 
Thus TS  TR , and in the radiation-dominated limit the spin temperature is driven to the color
temperature of the ambient radiation field by the effect of stimulated emission. Under typical
astrophysical conditions, TR is the temperature of the cosmic microwave background (~2.725 K).
General Excitation
Reality lies between the two limits just discussed, and the excitation and de-excitation of the HI
hyperfine levels are a mixture of collisional and radiative processes. If we assume for simplicity that
only atomic collisions are important (ne0.03nH), the condition of detailed balance becomes:
n1 (n H q10  A10  4J  B10 )  n 0 (n H q 01  4J  B01 )
Following a procedure similar to the one used in the limiting cases discussed above, this equation can
be solved for TS in the general case:
 1 A T  1 
  10  R  
1  Tkin nH q10  T*  TR 

TS  A10  TR  
 1   
 nH q10  T*  
This form is somewhat more complicated than is usually found in most textbooks, but it has the virtue
of emphasizing that the spin temperature is in effect a harmonic mean of the kinetic and radiation
temperatures, with the radiation temperature weighted by a factor that includes the influence of
stimulated emission and the relative importance of collisional vs. radiative de-excitation.
The relative importance of collisional and radiative de-excitation can be parameterized in terms of a
Critical Density, ncrit, at which the rates are roughly the same
A
ncrit  10 .
q10 (T )
When nH>ncrit, collisional de-excitation dominates and TSTkin, whereas in the low-density limit
radiative processes dominate and TSTR as before. Because of the temperature dependence on q10,
the critical density depends on temperature and does not have a simple analytic or even empirical
power-law form (despite what you may find in some references). See for example the tabulation of
Allison & Dalgarno. In general, the critical density is of order 10(45) cm3 for typical ISM conditions
(T= few 100K).
The figure below shows a plot of TS/Tkin for Tkin=10 and 100K for the collisional rates tabulated by
Allison & Dalgarno (1969). Note that TS does not exceed the background radiation temperature until
well above the critical density. For most ISM conditions, this won’t be an issue as nH is nearly always
much larger than ncrit, but this is not true at the very low densities of the intergalactic medium.

II-31
Neutral Atomic Hydrogen (HI) Regions

Plot of TS/Tkin for Tkin=10 and 100K.

Lyman- Pumping
In HI, absorption of Ly photons is almost immediately followed by re-emission (resonant scattering).
This process, however, does not necessarily return the electron to the same hyperfine structure level as
before absorption of the Ly photon.
Because most HI clouds are very optically thick to Ly photons, a single Ly photon is resonantly
scattered many times before exiting the cloud. Field (1959, ApJ, 129, 551) showed that if the tiny
energy difference of the hyperfine levels (6106eV compared to 10eV for Ly transition) is
accounted for, a very slight slope in the center of the Ly line profile develops that will drive
TSTkin.
Watson & Deguchi (1985, ApJL, 281, L5) have shown that in low-density extragalactic HI clouds this
process can occur, and can lead to small changes in the observed TS that increases it compared to the
expectations for the low-density limit.
In general, in the low-density limit (nHncrit) if no other radiation sources are present:
TS  TR  TCMB  2.725 K
As other sources of TR come into play, nHncrit, both collisional processes and Ly pumping will
start to drive TSTkin. In the Galaxy,
ne » 0.03 cm-3
nH » 0.1- 1 cm-3
Tkin » 100 K
Thus TSTkin is essentially always the case in HI regions in our Galaxy and other galaxies.

II-32
Neutral Atomic Hydrogen (HI) Regions

The CNM and WNM

HI 21cm studies of the Galaxy reveal that 21cm emission is virtually ubiquitous along every line of
sight, but 21cm absorption is not. The typical N(HI)>51019 cm2, and the general run of emission
lines are broader than the absorption lines (see the spectra from Radhakrishnan et al. above). This is
taken as evidence for two physically distinct neutral thermal phases coexisting in the ISM:
Cold Neutral Medium (CNM), composed of cold (T<100K), dense (n=20–60cm3), clouds and
filaments with a small filling factor (~1–4%).
Warm Neutral Medium (WNM), composed of diffuse, relatively low-density (~0.3cm3) gas
with temperatures of ~5000K with a much larger filling factor (~30%).
Observations suggest that the CNM has a typical column density of 51019 cm2, with a median
temperature of TS=58K, much colder than the canonical 100K. A column-density weighted average is
~70K, but some regions can be as cold as 15K and as warm as 250K. Detailed line profiles show that
the CNM is highly turbulent and supersonic, with a typical Mach numbers of ~3 with considerable
variation, and its morphology may thus be more sheet-like. The exponential scale height of the CNM
is about 100pc. While the CNM occupies only a few percent of the volume of the ISM, it probably
contains ~30% of the total mass of HI in the Galaxy.
The WNM fills about 30% of the ISM, with a higher typical column density of ~1020 cm2. Along
sight lines with no detectable absorption, TS=5000K, but using local UV absorption lines to
distinguish between thermal and turbulent line broadening suggests temperatures could be as high as
7000K. Two vertical scale-height components are seen, one is roughly Gaussian with a scale height of
~250pc, and the other is exponential with a scale height of ~500pc. A thermal stability analysis
analogous to the FGH criteria discussed in Chapter I suggests that about 50% of the WMN is
thermally unstable (T=500-5000K). Overall, the WMN contains about 40% of the total mass of HI in
the Galaxy.

II-33