The Polytropic Equation of State of Interstellar Gas Clouds

Marco Spaans1

Harvard-Smithsonian Center for Astrophysics, 60 Garden Street MS 51, Cambridge, MA

02138
arXiv:astro-ph/0002483v1 25 Feb 2000

Joseph Silk

Astrophysics, Department of Physics, University of Oxford, Keble Road, Oxford OX1 3RH,
U.K.2 and Departments of Astronomy and Physics, University of California, 601 Campbell
Hall, Berkeley, CA 94720

Received ; accepted

1
Hubble Fellow
2
current address
 –2–

ABSTRACT

Models are presented for the polytropic equation of state of self-gravitating,

quiescent interstellar gas clouds. A detailed analysis, including chemistry,
thermal balance, and radiative transfer, is performed for the physical state
of the gas as a function of density, metallicity, velocity field, and background
radiation field. It is found that the stiffness of the equation of state strongly
depends on all these physical parameters, and the adiabatic index varies between
∼ 0.2 − 1.4. The implications for star formation, in particular at high redshift
and in starburst galaxies, and the initial stellar mass function are discussed.

subject headings: ISM: clouds - ISM: molecular process - radiative transfer - cosmology:
theory - galaxies: starburst
 –3–

1. Introduction

A fundamental problem in the theory of star formation is the physical structure of

molecular gas clouds, such as dense cores and Bok globules, from which (low-mass) stars
are formed. Early work on the thermal and chemical balance of interstellar clouds has
been performed by de Jong, Dalgarno & Boland (1980) and Falgarone & Puget (1985). A
framework for active regions such as Orion has been developed by Tielens & Hollenbach
(1985). The above work has been extended and refined by many authors (see Hollenbach
& Tielens 1999 for a review), driven to a large extent by recent advances in observational
techniques.

Examples of such observational progress include the catalog of 248 optically selected
Bok globules of Clemens & Barvainis (1988), the NH3 survey of isolated cores in Taurus
by Benson & Myers (1989, and references therein), the measurement of non-thermal line
widths in regions surrounding ammonia cores (Fuller & Myers 1992, Goodman et al. 1998),
and the use of spectroscopy of various atoms and molecules as temperature and density
diagnostics of different types of interstellar clouds as summarized in van Dishoeck (1997).

Despite this wealth of observational data, the nature of their equation of state (EOS)
remains a major theoretical problem that concerns the stability and collapse of these
molecular clouds The stiffness of the EOS can be largely responsible for the resulting density
probability function of interstellar gas in the turbulent ISM (Scalo et al. (1998) and Passot
& Vazquez-Semadeni 1998). In particular, the value of gamma affects the densities that
can be attained behind shocks (Vazquez-Semadeni, Passot & Pouquet 1996), the fraction
of gas driven to high densities by turbulent interactions (c.f. Scalo et al. 1998), and the
stability of clouds being overrun by shocks (e.g. Tohline, Bodenheimer & Christodoulou
1987; Foster & Boss 1996). In the present paper we concentrate on the effects of the EOS
for the fragmentation of individual collapsing clouds. Before embarking on an analysis of
 –4–

the EOS of interstellar gas clouds, say of the form P = P (ρ, Tg , Td , V, B, I, C) for pressure
P , density ρ, gas temperatures Tg , dust temperature Td , velocity V , magnetic field B,
radiation intensity I, and chemical composition C, one should assess which of the possible
physical quantities entering the EOS dominates in the parameter ranges under study.

The importance of ρ and Tg is self-explanatory. The presence of warm dust can play a
role since it can heat the gas through gas-grain coupling and radiatively through optically
thick molecular lines (Takahashi, Hollenbach & Silk 1983). The velocity field, even when
it does not contribute to the turbulent pressure, strongly influences the optical depth of
atomic and molecular cooling lines, and hence the thermal balance of the medium. The
magnetic field plays an important role in the support of interstellar gas (McKee 1999). The
presence of ultraviolet or other hard radiation sources strongly influences the chemical,
thermal and ionization balance of the ambient medium, and hence the overall EOS. Finally,
the chemical composition of the gas also depends on the metallicity since it is the abundance
of atoms and molecules which influences the local cooling rate.

In the present work, a polytropic equation of state, P = Kργ , is assumed to describe

the physical state of the gas. The exponent γ is considered to be a function γ(Tg , Td , V, I, C).
That is, the magnetic pressure is assumed to be not dominant in the support of the model
cloud. The radiation field I, in addition to the radiation emitted by the dust grains, is
assumed to be given by the cosmic ray ionization rate. In other words, the visual extinction
of the model cloud is assumed to be sufficient to shield the gas from any ultraviolet (stellar)
sources.

We note that γ is the logarithmic derivative of P with respect to ρ. That is, since
dlogT
P ∼ ρT it follows that γ = 1 + dlogρ
. Through implicit differentiation the above derivative
is related to the logarithmic derivatives of the heating and cooling functions, which can be
quite complex. It is this dependence of γ on the details of heating and cooling that is the
 –5–

main topic of the present paper.

In the remainder of the paper, a useful guide is the density dependence of the heating
and cooling rate. If the density dependence of the heating rate is steeper (shallower) than
that of the cooling rate, then γ is larger (smaller) than unity. A similar result holds for
the kinetic temperature. Here, the cosmic ray ionization rate is always linear in the density
and independent of temperature, whereas gas-grain heating and cooling is quadratic in the
density and proportional to Tg1/2 (Tg − Td ). Optically thin, subthermally excited cooling
lines have a quadratic density dependence which becomes linear in the thermalized (high
density) limit and even effectively sub-linear when radiative trapping (optical depth effects)
becomes important. Finally, collisional de-excitation rates are characterized by powers of
the temperature on the order of ∼ 0.5 − 1.5, whereas excitation rates are modified by an
exponential factor which is of the order of unity for kinetic temperatures larger than the
pertaining excitation temperatures.

In general, our results are applicable mostly to dense and/or well shielded molecular
clouds/cores where the pressure appears to be close to thermal, with only a small additional
contribution to the line-widths from turbulent motions, and the ionization fraction is small
(Goodman et al. 1998, and references therein). That is, we assume that the clouds are
outside the regime where the so-called Larson (1981) laws apply, which indicate an increase
in the non-thermal line width with length scale.

One should note that polytropic models can also be extended into the regime where
there is a significant contribution to the pressure from turbulence and magnetic fields. The
polytropic temperature T should then be interpreted as representing the total velocity
dispersion σ, i.e., T ∝ σ 2 (Maloney 1988). A complete description of polytropic models,
including composite ones for non-thermal pressure effects and bulk properties like mass,
radius and density contrast, is presented in Curry & McKee (1999). The aim here is to
 –6–

study the thermal regime and to investigate the specific effects of metallicity, radiative
transfer and internal sources on the stiffness of the resulting local EOS.

2. Model Description

The results presented here were obtained by application of the numerical code of
Spaans (1996), described in detail in Spaans & Norman (1997); Spaans & van Dishoeck
(1997); and Spaans & Carollo (1998). The interested reader is referred to these papers
for a description of the underlying algorithm. The code has been specifically designed to
solve large chemical networks with a self-consistent treatment of the thermal balance for
all heating and cooling processes known to be of importance, and to be computable with
some sense of rigor, in the interstellar medium (Spaans & Ehrenfreund 1999). Care has to
be taken in the treatment of line trapping for the atomic and molecular cooling lines which
dominate the thermal balance. Collisional de-excitation and radiative trapping is most
prominent in the density range ∼ 103 − 104 cm−3 (Scalo et al. 1998). Our result were found
to be in good agreement with those of Neufeld et al. (1995) for their, and our, adopted
linear velocity gradient model (see further details below).

The adopted chemical network is based on the UMIST compillation (Millar, Farquhar
& Willacy 1997) and is well suited for low temperature dense molecular clouds, with the
latest rates for the important neutral-neutral reactions. The radiative transfer is solved by
means of a Monte Carlo approach (Spaans & van Langevelde 1992) with checks provided
by an escape probability method for large line optical depths. The thermal balance
includes heating by cosmic rays, absorption of infrared photons by H2 O molecules and
their subsequent collisional de-excitation, and gas-grain heating. The cooling includes the
coupling between gas and dust grains (Hollenbach & McKee 1989), atomic lines of all
metals, molecular lines, and all major isotopes, as described in Spaans & Norman (1997)
 –7–

and Spaans et al. (1994), and also in Neufeld, Lepp & Melnick (1995) for the regime of very
large line optical depth. All level populations are computed in statistical equilibrium at a
relative accuracy, also imposed on the ambient line radiation field, of no less than 10−3 .
The combined chemical and thermal balance is required to obey a convergence criterion
between successive iterations of 0.5%.

For the low metallicity computations, the relative contributions to the total cooling
by H2 and HD can become significant. Since the value of γ depends on the logarithmic
derivative of the cooling rate with respect to temperature and density, the end result could
become quite sensitive to the (quantum mechanical) treatment of H-H2 , H2 -H2 and He-H2
collional processes.

The main differences between calculations of H-H2 and H2-H2 collisions are due to
different potentials (surfaces) and whether reactive collisions, i.e., the exchange of an H
atom, are included (important above 1000 K). The shape of the potential surface is quite
important for the lower temperatures, where rotational quantum effects dominate. This
situation has improved greatly over recent years (c.f. Forrey et al. 1997). Nevertheless, we
have restricted our calculations to Z ≥ 0.01 Z⊙ . In this regime, there are other atomic
and molecular coolants which still contribute, at the 30% level, diminishing the relative
importance of the aforementioned sensitivities.

As an additional check, we have performed runs with the recent le Bourlot et al. (1999)
H2 rates. We have found no significant changes in γ for our adopted metallicities. Finally,
for zero metallicity gas, and in the absence of HD cooling, we found (albeit tentatively
given the above caveats) that γ is always less than unity. The inclusion of HD restricts this
result to densities less than ∼ 105 cm−3 .
 –8–

3. The Initial Mass Function

Before discussing the results of our computations, we want to outline how the value of
γ could related to the process of star formation, i.e., how a large value of γ might suppress
the formation of stars.

The polytropic equation of state of interstellar gas clouds may be relevant to

determinations of stellar masses. Our starting point is to pose the question: is there a
characteristic stellar mass? Observationally, the answer appears to be positive. The IMF,
dN/dM ∝ m−1−x , for the number N of stars per unit of mass M, has x > 1 above ∼ 1M⊙ .
< x < 0. The critical value
Below ∼ 0.3M⊙ , virtually all data suggests a flattening to −1 ∼ ∼
that defines a characteristic mass is x ≈ 1 for a logarithmic divergence in total mass toward
decreasing mass, and there is no doubt that there is a transition from x > 1 to x < 1 in the
range 0.2 − 1 M⊙ . Hence there is a characteristic stellar mass. Whether this varies with
environment is uncertain, but mchar ≈ 0.3 M⊙ fits most data (cf. Scalo 1998).

The status of theoretical discussions of a critical mass often centers on the thermal
Jeans mass, mJ ∝ T 3/2 ρ−1/2 , generalized to include external pressure, mJ,P ∝ T 2 P −1/2 ,
turbulence mJ,t ∝ ∆V 4 P −1/2 , or magnetic flux support mJ,B ∝ B 3 ρ−1 (c.f. Elmegreen 1999,
Larson 1998). The generalized Jeans scale most likely applies to the masses of atomic and
molecular clouds or to self-gravitating clumps. However, the fact that in the interstellar
medium, the Jeans mass in diffuse atomic or molecular clouds generally exceeds stellar
masses by a large factor, apart from the very coldest clumps, argues that the physics of the
IMF is more complicated than Jeans instability. The Jeans mass is a fragmentation scale
according to linear theory, and three-dimensional simulations indeed find that clump masses
are described by the Jeans scale. Nevertheless, simulations cannot currently resolve the
scales of individual protostars on solar mass scales. Analytic arguments, on the contrary,
of opacity-limited fragmentation find scales of ∼ 0.01 M⊙ , or far less than the characteristic
 –9–

stellar mass scale.

All of this seems to suggest that the Jeans mass is not very relevant to the IMF,
not withstanding the Elmegreen (1999) and Larson (1998) results. If it were, of course,
the critical polytropic index for fragmentation to occur is 4/3. This has the following
significance: one can write mJ ∝ ρ(3/2)(γ−4/3) . As collapse occurs, smaller and smaller
masses fragment, and one plausibly ends up with a powerlaw IMF only if γ < 4/3. This
argument could apply to the clump mass function within molecular clouds. The physics
of nonlinear fragmentation is complicated, and this is at least a necessary condition for a
power law IMF to continue to very small mass scales (compared to the initial Jeans mass).

But what really determines stellar masses? Physical processes that must play a role
in determining the characteristic stellar mass scale include clump coagulation (Allen &
Bastien 1995), clump interactions (Klessen, Burkert & Bate 1998) and protostellar outflows
(Nakano, Hasegawa & Norman 1995; Silk 1995) that halt accretion. One conjecture is that
feedback from protostars plays a crucial role in limiting accretion. The latter depends
crucially on the accretion rate to protostellar cores (various authors relate inflow to outflow
rates on energetic grounds), which in turn depends sensitively on the turbulent velocity.
The maximum embedded young stellar object luminosity correlates with NH3 core line
width ∆V (Myers & Fuller 1993) and clump mass (Kawamura et al. 1998) over a wide
range. Myers & Fuller (1993) concluded that variations in the initial velocity dispersion
from core to core is a key determinant of the stellar mass formed from such a core. Also,
Fuller & Myers (1992) concluded that the physical basis of line width-size relations (the
so-called Larson laws) is part of the initial conditions for star formation. The theoretical
justification incorporates disk physics and, at least conceptually, magnetic fields (Adams
& Fatuzzo 1996; Silk 1995). So if the above holds (theoretically and observationally, both
aspects are relevant), then one might argue that the characteristic mass of a young stellar
 – 10 –

object depends primarily on ∆V . The latter is determined by the kinetic temperature if

the thermal line width dominates in a prestellar core. This situation holds for low mass
star formation but less so for the high mass cores, where the nonthermal contribution to
the line width (i.e. turbulent or magnetic) can be as large as 50-80% (c.f. Myers & Fuller
1993). In any case, γ = 1 appears the most natural critical value of interest.

If γ = 1 is indeed a critical value in this sense, there is the possibility that the
primordial EOS with a stiffer polytropic index, γ > 1, leads to a peaked IMF, i.e. one
biased toward massive stars, whereas current star formation, where the molecular gas is
characterised by an EOS with γ < 1, results in runaway fragmentation and hence a power
law IMF.

4. Results

The model results3 are obtained for self-gravitating, quiescent spherical clouds with
total column densities per unit of velocity given by

N SIS (H2 ) = 5.1 × 1019 n(H2 )1/2 cm−2 per km s−1 , (1)

(the singular isothermal sphere value; Neufeld et al. 1995) to set the optical depth parameter
for any species, total hydrogen densities nH = 102 − 106 cm−3 , metallicities Z = 1 − 0.01Z⊙
in solar units, and a possible homogeneous background infrared radiation (IR) field given
by 100 K dust grains. The total visual extinction through the region containing the model
cloud is fixed at AV = 100 mag, although this is only important for determining the
dust emission optical depth in the IR background case. Models are also constructed for a
constant velocity gradient field of 3 km s−1 pc−1 , which is a factor of three larger than the

3
Extensive tables can be obtained from MS.
 – 11 –

value adopted in Goldsmith & Langer (1978), corresponding to an optical depth parameter
of
N VG (H2 ) = 1.0 × 1018 n(H2 ) cm−2 per km s−1 . (2)

It has been shown by Neufeld & Kaufman (1993) that this choice violates the virial theorem
in the limit of high density. Therefore, (1) is interpreted to apply to a (statistically) static
cloud supported by thermal pressure, whereas (2) should be seen as representative of a
dynamic medium in a state of infall or outflow. The cosmic ray ionization rate is taken
equal to ζCR = 3 × 10−17 s−1 .

4.1. Variations in γ: Competition between Heating and Cooling Processes

Figures 1 and 2 present results which do not include an IR background radiation field,
e.g. little or no ongoing star formation, or a warm CMB. Figure 1 displays the results for a
quiescent velocity field with a velocity dispersion of ∆V = 0.3 km s−1 for H2 . Figure 2 is
for the corresponding case of a velocity gradient of 3 km s−1 pc−1 . The value of γ becomes
of the order of unity for low metallicities, Z < 0.03Z⊙ , and densities nH < 105 cm−3 .
The reason is that the cooling of the ambient medium becomes dominated by H2 and HD
(Norman & Spaans 1997), which emit optically thin line radiation. At low densities, the
level populations are not thermalized and the cooling rate has a density dependence which
is steeper than linear and typically quadratic, whereas the cosmic ray heating rate is linear
in the density.

At higher densities gas-grain cooling becomes important. This has a quadratic density
dependence, and the equation of state softens in the absence of any embedded sources
(Scalo et al. 1998), but to a lesser extent due to the lower dust abundance, for small
metallicities. Even though no computations were performed beyond densities of ∼ 3 × 106
cm−3 , the gas-grain coupling at very large densities is so strong that the gas temperature
 – 12 –

follows the grain temperature and the value of γ will increase again and attain a value of
unity if embedded protostars are present and the ambient radiation field is not coupled to
the local gas density (see below).

The high γ region between ∼ 103 and ∼ 104 cm−3 is the result of line trapping and
collisional de-excitation. Line trapping causes the cooling rate to depend on the density with
an effective power less than unity due to absorption and subsequent collisional de-excitation,
while the cosmic ray heating rate is always linear in the density, independent of the kinetic
temperature. It should be emphasized that the effect of radiative trapping is stronger for
the quiescent velocity field, but also depends crucially on the specific implementation that
we have chosen. A truly turbulent velocity field (Kegel, Piehler & Albrecht 1993), i.e. one
with a finite correlation length, may lead to a quite different value of γ in the density range
∼ 103 − 104 cm−3 .

4.2. High Redshift Star formation

As can be seen in Figure 3, the inclusion of an IR background leads to a much

stiffer polytropic equation of state for high, Z > 0.1Z⊙ , metallicities. The reason is that
the abundant presence of water leads to heating through absorption of IR radiation and
subsequent collisional de-excitation (Takahashi et al. 1983) for densities between ∼ 3 × 103
and ∼ 3 × 104 cm−3 . Temperature effects play a role here in the determination of the line
opacities, but this is essentially a heating term that is between linear and quadratic in the
ambient density. Since this heating rate is roughly proportional to the abundance of H2 O,
it is negligible for metallicities Z < 0.1Z⊙ . At very high densities, nH > 3 × 106 cm−3 , the
gas and dust become completely coupled, and γ approaches unity since the gas temperature
will simply follow the dust temperature. This only follows provided that the strength of the
IR radiation field is independent of the ambient density. This point will re-addressed in the
 – 13 –

next subsection on starburst galaxies.

Note also that γ decreases to unity from above for densities larger than ∼ 105 cm−3
because of the high temperature of the FIR background. This causes the dust to heat
the gas and results in a relatively stiff EOS, even for low metallicities where H2 and
HD dominate. This is because at these high densities the overall cooling rates are at
their thermodynamic limit, roughly linear in the density, while the dust-gas heating rate
goes like the square of the density. Note that for T < 200 K, HD quickly becomes the
dominant coolant of pristine gas. The lower equilibrium temperature of the gas facilitates
the thermalization of the level populations. Finally, for densities larger than 3 × 105 cm−3
radiative trapping becomes important for HD, leading to a density dependence of its cooling
rate of less than unity.

Even though no results are shown for intermediate (30-60 K) temperature dust
backgrounds, γ first decreases to somewhat below unity around 105 cm−3 in these cases,
if the metallicity is less than 0.1 Z⊙ . It then approaches unity from below for the reasons
given above. Finally, in the limit of a large velocity gradient, H2 O plays a minor role and its
heating (or cooling) does not strongly influence the thermal balance of the gas (Takahashi
et al. 1983).

Since an IR background is naturally present in the form of the CMB for redshifts larger
than z ∼ 10 − 30, one can speculate on the nature of very early star formation. That is,
any dense shielded region which has been enriched by metals through the first supernova
explosions, viewed here as the product of the very first population III stars, will become
more stable in the presence of an intense CMB background. Conversely, metal-poor regions
will retain a polytropic exponent of ≈ 1. Therefore, high redshift star formation may
exhibit the counter-intuitive property that metal enrichment and a warm CMB together
“halt” the process of star formation for a metallicity Zc > 0.1Z⊙ , even though the gas is
 – 14 –

intrinsically capable of cooling efficiently (Spaans & Norman 1997).

Once the CMB temperature drops below ∼ 40 K, at z ∼ 15, any dense, nH > 104
cm−3 , cloud of high or even modest metallicity rapidly develops a γ-value smaller than
unity and is prone to collapse. In effect, although more detailed calculations are required
to substantiate this, the CMB may play an integral part in determing the epoch of efficient
early star formation. One also sees that the primordial IMF can change strongly as one
goes from a γ > 1 to a γ < 1 EOS. In the former regime one expects only large density
excursions to collapse, leading to an IMF biased toward massive stars.

4.3. Starburst Galaxies

A similar phenomenon can occur in luminous (IR) starburst galaxies in that warm,
∼ 50 − 200 K, opaque dust hardens the polytropic equation of state of a nuclear starburst
region. The system Arp 220 is a possible example, being optically thick at 100 µm. From
the results presented here, one concludes that such IR luminous opaque systems have
an IMF which is quite shallow, favoring high mass star formation. This may in fact be
desirable to maintain the required power of such galaxies through supernova explosions (c.f.
Doane & Mathews 1993).

Note though that in these extreme systems there is a strong coupling between the
ambient radiation field and the star formation rate R, and hence with the gas density.
Typical Schmidt star formation laws adopt an observationally motivated value of 1 − 2 for
the dependence of R on gas density (Kennicutt 1998) in star-forming spiral galaxies, with a
best fit value of r = 1.4. Since the resulting dust temperature depends on the 1/5 power of
the ultraviolet energy density (Tielens & Hollenbach 1985), the effective γ for r = 2 (r = 1)
is 1.4 (1.2), consistent with the above results.
 – 15 –

5. Discussion

The polytropic equation of state of interstellar gas clouds has been computed for a
range of physical parameters, and its relation to the IMF, high redshift star formation,
and starburst galaxies has been investigated. A much needed extension of the results
presented here lies in MHD, and particularly turbulence. The magnetic and velocity fields
are important for pressure support away from the most centrally condensed part of a
molecular cloud core, and radiative transfer effects in a turbulent medium are non-trivial.
More to the point, the polytropic EOS indices computed here, with their relevance to
the IMF as discussed in Section 4, are to be applied to a medium with a given cloud
density probability distribution, as investigated in Scalo et al. (1998). Such a cloud density
probability distribution can only be derived from detailed hydrodynamic simulations which
include magnetic pressure and Coriolis forces. Furthermore, γ can also affect, in a way
still to be investigated, star formation and the underlying IMF in other scenarios like
turbulence-induced condensation or shocked-cloud induced star formation. We would like
to mention as well the work by Ogino, Tomisaka & Nakamura (1999) that shows that the
mass accretion rate in their polytropic simulations of collapsing clouds depends a lot on
whether gamma is taken as 0.8 or 1.2.

Extensions in terms of multiple polytropic models for the core and envelope of a
molecular cloud, such as discussed in Curry & McKee (1999), are another avenue to explore
for the bulk properties of interstellar clouds. Caveats in the approach presented here are
grain mantle evaporation which can influence the abundance of H2 O and hence the heating
rate in the presence of an IR background, uncertainties in the chemical reaction rates of
neutral-neutral reactions at low temperatures, and details of the freeze-out onto grains
of molecular species. We hope that through a synthesis of the approach adopted here
and those of others mentioned above, one can address the question of cloud stability and
 – 16 –

collapse in a more general (cosmological) framework.

We are grateful to the referee, John Scalo, for his constructive comments, which
improved the presentation of this paper, and his emphasis on the computation of the zero
metallicity case. Discussions with Ewine van Dishoeck and John Black on the collisional
cross sections of H2 are also gratefully acknowledged. MS is supported by NASA through
grant HF-01101.01-97A, awarded by the Space Telescope Institute, which is operated by the
Association of Universities for Research in Astronomy, Inc., for NASA under contract NAS
5-26555. J.S. has been supported in part by grants from NSF and NASA at UC Berkeley.
 – 17 –

REFERENCES

Adams, F.C. & Fatuzzo, M. 1996, ApJ, 464, 256

Allen, E.J. & Bastien, P. 1995, ApJ, 452, 652

Benson, P.J. & Myers, P.C. 1989, ApJS, 71, 89

Clemens, D.P. & Barvainis, R. 1988, ApJS, 68, 257

Curry, C.L. & McKee, C.F. 1999, astro-ph/9908071

de Jong, T., Dalgarno, A. & Boland W. 1980, A&A, 91, 68

Doane, J. & Mathews, W.G. 1993, ApJ, 419, 573

Elmegreen, B.G. 1999, ApJ, 515, 323

Falgarone, E. & Puget, J.-L. 1985, A&A, 142, 157

Forrey, R.C., Balakrishnan, N., Dalgarno, A. & Lepp, S. 1997, ApJ, 489, 1000

Foster, P.N. & Boss, A.P. 1996, ApJ, 468, 784

Fuller, G.A. & Myers, P.C. 1992, ApJ, 384, 523

Goldsmith, P.F. & Langer, W.D. 1978, ApJ, 222, 881

Goodman, A.A., Barranco, J.A., Wilner, D.J. & Heyer, M.H. 1998, ApJ, 504, 223

Hollenbach, D.A. & McKee, C.F. 1989, ApJ, 342, 306

Hollenbach, D.A. & Tielens, A.G.G.M. 1999, Rev. Modern Phys., 71, 173

Kawamura, A., Onishi, T., Yonekura, Y., Dobashi, K., Mizuno, A., Agawa, H. & Fukui, Y.
1998, ApJS, 117, 387
 – 18 –

Kegel, W.H., Piehler, G. & Albrecht, M.A. 1993, A&A, 270, 407

Kennicutt, R.J., 1998, ApJ, 498, 541

Klessen, R.S., Burkert, A. & Bate, M.R. 1998, ApJ, 501, L205

Larson, R.B. 1998, MNRAS, 301, 569

Larson, R.B. 1981, MNRAS, 194, 809

le Bourlot, J., Pineau des Forêts, G. & Flower, D.R., 1999, MNRAS, 305, 802

Maloney, P. 1988, ApJ, 334, 761

McKee, C.F., 1999, astro-ph/9901370

Millar, T.J., Farquhar, P.R.A. & Willacy, K., 1997, A&AS, 121, 139

Myers, P.C. & Fuller G.A. 1993, ApJ, 402, 635

Nakano, T., Hasegawa, T. & Norman, C.A. 1999, ApJ, 450, 183

Neufeld, D.A. & Kaufman, M.J. 1993, ApJ, 418, 263

Neufeld, D.A., Lepp, S. & Melnick, G.J. 1995, ApJS, 100, 132

Norman, C.A. & Spaans, M. 1997, ApJ, 480, 145

Ogino, S., Tomisaka, K. & Nakamura, F. 1999, PASJ, 51, 637

Passot, T.& Vazquez-Semadeni, E. 1999, Phys. Rev. E, 58, 4501

Scalo, J. 1998, in The Stellar Initial Mass Function (38th Herstmonceux Conference) ASP
Conference Series, Vol. 142, 1998, ed. G. Gilmore and D. Howell.

Scalo, J., Vázquez-Semadeni, E., Chappell, D. & Passot, T. 1998, ApJ, 504, 835
 – 19 –

Silk, J. 1995, ApJ, 438, L41

Spaans, M. 1996, A&A, 307, 271

Spaans, M. & Ehrenfreund, P. 1999, in “Laboratory Physics and Space Research”,

ASSL Vol. 236, eds. P. Ehrenfreund, K. Krafft, H. Kochan, and V. Pironello
(Kluwer:Dordrecht), p. 1

Spaans, M. & Carollo, C.M. 1998, ApJ, 502, 640

Spaans, M. & van Dishoeck, E.F. 1997, A&A, 323, 953

Spaans, M. & Norman 1997, ApJ, 488, 27

Spaans, M., Tielens, A.G.G.M., van Dishoeck, E.F. & Bakes, E.L.O., 1994, ApJ, 437, 270

Spaans, M. & van Langevelde, H.J. 1992, MNRAS, 258, 159

Takahashi, T., Hollenbach, D.J. & Silk, J. 1983, ApJ, 275, 145

Tielens, A.G.G.M. & Hollenbach, D.A., 1985, ApJ, 291, 722

Tohline, J.E., Bodenheimer, P.H. & Christodoulou D.M. 1987, ApJ, 322,787

van Dishoeck, E.F., ed. “Molecules in Astrophysics: Probes and Processes”, 1997, IAU
Symp. 178, Kluwer

Vazquez-Semadeni, E., Passot, T.& Pouquet, A. 1996, ApJ, 473, 881

This manuscript was prepared with the AAS LATEX macros v4.0.
 – 20 –

Fig. 1.— Density and metallicity dependence of the polytropic index for model I, discussed
in the text, for no IR background and a quiescent velocity field with ∆V = 0.3 km s−1 .

Fig. 2.— Density and metallicity dependence of the polytropic index for model II, discussed
in the text, for no IR background and a spherically symmetric constant velocity gradient
field of 3 km s−1 pc−1 .

Fig. 3.— Density and metallicity dependence of the polytropic index for model I, discussed
in the text, for a Td = 100 K IR background from dust, AV = 100 mag, and a quiescent
velocity field.
– 21 –
– 22 –
– 23 –