28.07.01 · astronomy / ism-star-formation

The interstellar medium and star formation

shipped3 tiersLean: none

Anchor (Master): Stahler and Pala, The Formation of Stars (2004); Draine (2011); McKee and Ostriker (2007), ARAA 45, 565

Intuition Beginner

The space between the stars is not empty. It is filled with a thin, dusty gas called the interstellar medium, or ISM. This gas is the raw material from which every new star is built. By mass it is about three quarters hydrogen, one quarter helium, and a sprinkling of heavier elements forged in earlier generations of stars 28.02.01. Cold and clumped in some places, hot and spread out in others, the ISM is anything but uniform.

A star is born when a region of this gas becomes dense enough that its own gravity overwhelms its internal pressure. Picture a cloud as a tug of war. Gravity pulls the gas inward, trying to collapse it; the gas pressure (from heat, turbulence, and magnetic fields) pushes outward, trying to disperse it. When gravity wins, the cloud fragments and contracts, and at the centre of each fragment a protostar gathers mass, heats up, and eventually ignites nuclear fusion. The whole journey, from diffuse cloud to shining star, takes anywhere from a hundred thousand to tens of millions of years.

Not every patch of gas can make a star. There is a threshold: a cloud collapses only if it is cold enough and massive enough that pressure cannot hold it up. This balance is captured by a single number called the Jeans mass. Below it the cloud sits quietly; above it the cloud falls in on itself. This threshold is why star formation happens in cold molecular clouds and not in the hot, diffuse gas that fills most of the galaxy.

Visual Beginner

The interstellar medium is divided into phases by temperature and density. The table below summarises the four phases that account for most of the gas.

Phase	Temperature (K)	Number density (cm $^{- 3}$ )	State	Filler of
Cold molecular	10 - 20	$1 0^{2}$ - $1 0^{6}$	H $_{2}$ molecules, dust	dense clouds, star-forming cores
Cold neutral atomic (CNM)	50 - 100	$1 0^{1}$ - $1 0^{2}$	neutral H atoms	filaments, cloud envelopes
Warm neutral/ionised (WNM/WIM)	$1 0^{3}$ - $1 0^{4}$	$1 0^{- 1}$ - $1 0^{0}$	diffuse H, some ionised	most of the disk volume
Hot ionised (HIM)	$1 0^{6}$	$1 0^{- 3}$	fully ionised plasma	supernova-carved bubbles

Only the cold molecular phase is cold and dense enough to collapse into stars. The warm and hot phases are pressurised, buoyant, and fill most of the volume between the cold clouds.

Worked example Beginner

Take a typical molecular cloud with number density $n = 100$ particles per cubic centimetre and temperature $T = 10$ K. We will compute the minimum mass this gas needs in order to collapse. This threshold is the Jeans mass, and it comes from comparing the sound-crossing time (how fast pressure can respond) with the free-fall time (how fast gravity pulls the gas together).

First, the sound speed in the gas is $c_{s} = k T / (μ m_{H})$ , where $k$ is Boltzmann's constant, $μ = 2.3$ is the mean molecular weight for molecular hydrogen, and $m_{H}$ is the mass of a hydrogen atom. Plugging in $T = 10$ K gives $c_{s} \approx 0.19$ km/s, about the speed of a slow bicycle.

Next, the Jeans length is $λ_{J} = c_{s} π / (G ρ)$ , where $G$ is Newton's gravitational constant and $ρ = n μ m_{H}$ is the mass density. With $n = 100$ cm $^{- 3}$ , the density is $ρ \approx 3.8 \times 1 0^{- 22}$ g/cm $^{3}$ , which gives $λ_{J} \approx 2.1$ parsecs (about 7 light-years).

The Jeans mass is the gas enclosed in a sphere of that radius, $M_{J} = (π /6) ρ λ_{J}^{3}$ . Substituting the numbers above gives $M_{J} \approx 30$ solar masses. So a 100 cm $^{- 3}$ , 10 K cloud must gather at least about 30 solar masses of gas inside a 2-parsec region before it can collapse.

The lesson is that collapse favours the cold and the dense. Heat the cloud to $T = 100$ K and the Jeans mass rises by a factor of $10 0^{1.5} = 1000$ . Squeeze the density up by a factor of 100 and the Jeans mass falls by a factor of 10. This is why stars form in cold molecular cores and not in the warm gas that fills most of the galaxy.

Check your understanding Beginner

Formal definition Intermediate+

Let $ρ$ , $P$ , $T$ , and $v$ denote the mass density, pressure, temperature, and velocity field of the gas. The interstellar medium is a multi-phase fluid governed by the equations of ideal (magneto)hydrodynamics coupled to gravity, radiative cooling, and chemistry. Following McKee and Ostriker ^{[McKeeOstriker1977]}, the gas occupies several phases in approximate pressure balance.

A phase is a thermodynamic state of the gas, specified by $(T, n, x_{e})$ where $n$ is the total particle number density and $x_{e}$ is the electron fraction. The cold neutral medium (CNM) and the molecular medium occupy $T ≲ 100$ K with $n ≳ 10$ cm $^{- 3}$ ; the warm neutral medium (WNM) and warm ionised medium (WIM) occupy $T \sim 1 0^{3}$ to $1 0^{4}$ K with $n \sim 0.1$ to $1$ cm $^{- 3}$ ; the hot ionised medium (HIM), carved out by supernova remnants, sits at $T \sim 1 0^{6}$ K with $n \sim 1 0^{- 3}$ cm $^{- 3}$ . The CNM and WNM are thermally stable branches of a single cooling curve; the warm and cold phases can coexist in pressure equilibrium because the cooling function produces a thermally bistable regime.

A molecular cloud is a region in which hydrogen is predominantly in molecular form (H $_{2}$ ), traced observationally by CO emission, dust extinction, and far-infrared continuum. Typical giant molecular clouds (GMCs) have masses $1 0^{4}$ to $1 0^{6} M_{⊙}$ , sizes 10 to 100 pc, temperatures 10 to 20 K, and mean densities $n \sim 50$ to 500 cm $^{- 3}$ . They are gravitationally bound, turbulent, and magnetised; only a small fraction of their mass ( $\sim$ a few percent per free-fall time) is converted into stars.

The isothermal sound speed of a gas of mean molecular weight $μ$ and temperature $T$ is $c_{s} = k_{B} T / (μ m_{H})$ . For molecular gas at $T = 10$ K with $μ = 2.3$ , this gives $c_{s} \approx 0.19$ km/s. In magnetised gas the relevant signal speed is the magnetosonic speed, and in turbulent clouds the effective pressure is dominated by velocity dispersion rather than thermal pressure.

The Jeans length and Jeans mass of a uniform isothermal sphere of density $ρ$ are

λ_{J} = c_{s} \frac{π}{Gρ}, M_{J} = \frac{π ^{5/2}}{6} \frac{c _{s}^{3}}{G ^{3/2} ρ ^{1/2}} .

A perturbation with wavelength longer than $λ_{J}$ and mass larger than $M_{J}$ is unstable to gravitational collapse. Equivalently, $M_{J}$ sets the minimum mass that can collapse at fixed $(T, ρ)$ .

The initial mass function (IMF) $ξ (M)$ is the number of stars formed per unit mass interval, normalised per unit star-forming volume. Salpeter ^{[Salpeter1955]} showed that over $0.4 ≲ M / M_{⊙} ≲ 10$ the IMF is well described by a power law

ξ (M) d M = A M^{- α} d M, α = 2.35.

Modern determinations (Kroupa, Chabrier) flatten $ξ$ below $0.5 M_{⊙}$ and steepen the high-mass tail, but the Salpeter slope $α = 2.35$ remains the canonical high-mass index.

The Kennicutt-Schmidt relation connects the disk-averaged star-formation-rate surface density $Σ_{SFR}$ to the total (atomic plus molecular) gas surface density $Σ_{gas}$ . Following Kennicutt ^{[Kennicutt1998]},

Σ_{SFR} = A Σ_{gas}^{N}, N \approx 1.4 \pm 0.15.

When only the molecular gas is considered, the relation is closer to linear, $N \approx 1$ , suggesting that the formation of H $_{2}$ is the rate-limiting step. The depletion time $τ_{dep} = Σ_{gas} / Σ_{SFR}$ is of order 1 to 2 Gyr in normal spirals and an order of magnitude shorter in starbursts.

Key derivation: the Jeans criterion and gravitational collapse Intermediate+

Consider a uniform, stationary, self-gravitating ideal gas with density $ρ_{0}$ , pressure $P_{0} = c_{s}^{2} ρ_{0}$ , and gravitational potential $Φ_{0}$ satisfying $\nabla^{2} Φ_{0} = 4 π G ρ_{0}$ . Perturb each quantity about the background: $ρ = ρ_{0} + ρ_{1}$ , $P = P_{0} + P_{1}$ , $v = v_{1}$ , $Φ = Φ_{0} + Φ_{1}$ , and linearise the continuity, Euler, and Poisson equations:

\frac{\partial ρ _{1}}{\partial t} + ρ_{0} \nabla \cdot v_{1} = 0, ρ_{0} \frac{\partial v _{1}}{\partial t} = - \nabla P_{1} - ρ_{0} \nabla Φ_{1}, \nabla^{2} Φ_{1} = 4 π G ρ_{1} .

For an isothermal equation of state $P_{1} = c_{s}^{2} ρ_{1}$ . Seeking plane-wave solutions $ρ_{1} \propto exp [i (k \cdot x - ω t)]$ and eliminating $v_{1}$ and $Φ_{1}$ gives the Jeans dispersion relation,

ω^{2} = c_{s}^{2} k^{2} - 4 π G ρ_{0} .

For wavenumbers $k > k_{J} = 4 π G ρ_{0} / c_{s}$ , $ω^{2} > 0$ and the gas supports propagating sound waves. For $k < k_{J}$ , $ω^{2} < 0$ , $ω$ is imaginary, and the perturbation grows exponentially with growth rate $γ = 4 π G ρ_{0} - c_{s}^{2} k^{2}$ . The longest wavelengths grow fastest, with the free-fall rate $γ \to 4 π G ρ_{0}$ . The Jeans length $λ_{J} = 2 π / k_{J} = c_{s} π / (G ρ_{0})$ is the threshold: perturbations larger than $λ_{J}$ collapse, smaller ones oscillate. The Jeans mass $M_{J} = (π /6) ρ_{0} λ_{J}^{3} = (π^{5/2} /6) c_{s}^{3} / (G^{3/2} ρ_{0}^{1/2})$ collects the gas inside one Jeans length and is the minimum mass able to collapse at fixed $(T, ρ)$ .

The Jeans criterion was first written down by Jeans ^[Jeans1902] and remains the conceptual backbone of star formation theory, though the assumption of an infinite uniform medium is unphysical: a real self-gravitating cloud cannot sit in static equilibrium (the so-called Jeans swindle). The criterion is nevertheless recovered rigorously in expanding backgrounds and in pressure-confined clouds, where it generalises to the Bonnor-Ebert mass,

M_{BE} = \frac{1.18 c _{s}^{4}}{( G ^{3} P _{ext} ) ^{1/2}},

the maximum mass of an isothermal sphere in hydrostatic equilibrium confined by external pressure $P_{ext}$ . For $T = 10$ K and $P_{ext} / k \sim 3000$ cm $^{- 3}$ K (typical of the CNM), $M_{BE} \sim 1$ to $3 M_{⊙}$ , which matches the masses of dense prestellar cores and aligns with the peak of the IMF.

Once a region exceeds $M_{J}$ or $M_{BE}$ it detaches from pressure support and falls inward. Collapse is approximately isothermal at first because dust radiates away the compressional heat efficiently; the density rises as $ρ \propto r^{- 2}$ , and because $M_{J} \propto ρ^{- 1/2}$ falls as the density rises, the collapsing region itself becomes Jeans-unstable on ever smaller scales. This fragmentation cascade continues until the central regions become optically thick, trap their heat, and the gas heats up. At that point $M_{J}$ stops falling and a hydrostatic protostellar core forms at a mass comparable to the opacity limit, of order $0.01 M_{⊙}$ . The protostar then grows by accretion through a disk, driving bipolar outflows that clear away the envelope.

Bridge. The Jeans criterion is the foundational reason individual star-formation thresholds exist at all: it builds toward the Kennicutt-Schmidt law of galaxy-scale star formation in 28.03.01 and appears again in the growth of primordial density fluctuations in 28.04.01. This is exactly the same pressure-versus-gravity competition that sets the Chandrasekhar and Bonnor-Ebert scales in stellar evolution 28.02.01; it generalises to magnetised, turbulent, and cosmological media, and putting these together fixes the characteristic stellar mass from first principles. The bridge is that collapse onset, threshold density, and the peak of the IMF all reduce to one comparison between the sound-crossing time and the free-fall time.

Exercises Intermediate+

Exercise 3 (medium, numeric).

A Salpeter IMF, $ξ (M) = M^{- 2.35}$ (in solar masses), runs from $M_{lo} = 0.1 M_{⊙}$ to $M_{hi} = 100 M_{⊙}$ . Compute, to two significant figures, the fraction of the total stellar mass that is locked up in stars more massive than $1 M_{⊙}$ . (Take the IMF valid across the full range for this estimate.)

Hint

The mass in stars between $M_{a}$ and $M_{b}$ is $\int_{M_{a}}^{M_{b}} M ξ (M) d M = \int_{M_{a}}^{M_{b}} M^{- 1.35} d M$ . Evaluate it from $1$ to $100$ and from $0.1$ to $100$ , then take the ratio.

Answer

0.39 (about 39%). The mass above $1 M_{⊙}$ is $\int_{1}^{100} M M^{- 2.35} d M = \int_{1}^{100} M^{- 1.35} d M = \frac{1 - 10 0 ^{- 0.35}}{0.35} = \frac{1 - 0.200}{0.35} = 2.29$ . The total mass is $\int_{0.1}^{100} M^{- 1.35} d M = \frac{0. 1 ^{- 0.35} - 10 0 ^{- 0.35}}{0.35} = \frac{2.24 - 0.200}{0.35} = 5.83$ . The fraction is $2.29/5.83 = 0.39$ . So under a pure Salpeter law from $0.1$ to $100 M_{⊙}$ , stars above $1 M_{⊙}$ hold about two fifths of the mass but contain only $\int_{1}^{100} M^{- 2.35} d M / \int_{0.1}^{100} M^{- 2.35} d M \approx 0.024$ , about 2.4%, of the number. (A real Kroupa or Chabrier IMF turns over below $0.5 M_{⊙}$ , shifting more mass to the low end and lowering this fraction further.)

Exercise 4 (medium, numeric).

A dense prestellar core has $n = 1 0^{5}$ cm $^{- 3}$ and $T = 10$ K. Using $μ = 2.3$ , estimate the Jeans length in parsecs and the Jeans mass in solar masses. Take $m_{H} = 1.67 \times 1 0^{- 24}$ g, $G = 6.67 \times 1 0^{- 8}$ cgs, $k_{B} = 1.38 \times 1 0^{- 16}$ cgs, and $1 pc = 3.086 \times 1 0^{18}$ cm.

Hint

Compute $ρ = n μ m_{H}$ , then $c_{s} = k_{B} T / (μ m_{H})$ , then $λ_{J} = c_{s} π / (Gρ)$ , then $M_{J} = (π /6) ρ λ_{J}^{3}$ , converting to parsecs and solar masses at the end.

Answer

$λ_{J} \approx 0.068$ pc; $M_{J} \approx 0.94 M_{⊙}$ . The sound speed is unchanged at $c_{s} \approx 0.19$ km/s. With $n = 1 0^{5}$ cm $^{- 3}$ , $ρ = 1 0^{5} \times 2.3 \times 1.67 \times 1 0^{- 24} = 3.84 \times 1 0^{- 19}$ g/cm $^{3}$ . Then $λ_{J} = c_{s} π / (Gρ) = 1.9 \times 1 0^{4} \times π / (6.67 \times 1 0^{- 8} \times 3.84 \times 1 0^{- 19}) \approx 2.1 \times 1 0^{17}$ cm $\approx 0.068$ pc, and $M_{J} = (π /6) ρ λ_{J}^{3} \approx 1.9 \times 1 0^{33}$ g $\approx 0.94 M_{⊙}$ . Dense cold cores collapse on solar-mass scales, which is why the IMF peaks near a solar mass.

Exercise 5 (medium, short answer).

The Kennicutt-Schmidt relation $Σ_{SFR} \propto Σ_{gas}^{1.4}$ is steeper than linear. Give a dynamical argument for why an exponent near $1.4$ , rather than $1.0$ , is expected if star formation tracks the free-fall rate of the gas.

Hint

The disk-averaged free-fall time scales as $t_{ff} \propto ρ^{- 1/2}$ , and $ρ \propto Σ_{gas}$ when the scale height is fixed by hydrostatic balance with the same midplane pressure.

Answer

If the star-formation rate per unit volume tracks the gas mass divided by the local free-fall time, $\overset{ρ}{˙}_{⋆} \approx ϵ_{ff} ρ / t_{ff}$ with constant efficiency $ϵ_{ff}$ , then since $t_{ff} \propto ρ^{- 1/2}$ we get $\overset{ρ}{˙}_{⋆} \propto ρ^{3/2}$ . Integrating through a disk of roughly fixed scale height gives $Σ_{SFR} \propto Σ_{gas}^{3/2}$ , i.e. $N = 1.5$ , close to the observed $1.4$ . The mild offset is attributed to the gas scale height itself depending weakly on surface density and to the molecular fraction varying across the sample. The dynamical content is that the superlinear slope is a free-fall signature, not a coincidence.

Exercise 6 (medium, short answer).

Why does the fragmentation cascade in a collapsing isothermal cloud halt at roughly the opacity limit rather than continuing to ever smaller masses?

Hint

Track what happens to the temperature, and therefore to $M_{J} \propto T^{3/2} ρ^{- 1/2}$ , once the gas becomes optically thick.

Answer

While the cloud stays optically thin, compressional heat is radiated away and the collapse is isothermal, so $M_{J} \propto ρ^{- 1/2}$ falls as the density rises and subregions fragment. Once the column becomes high enough that dust traps the infrared photons, the gas heats up adiabatically. The temperature then rises faster than $ρ^{1/3}$ , so $M_{J} \propto T^{3/2} ρ^{- 1/2}$ starts to increase rather than decrease, and further fragmentation stops. The minimum fragment mass set by this transition, the opacity limit, is of order $0.01 M_{⊙}$ , near the bottom of the observed IMF and the mass of the smallest brown dwarfs.

Exercise 7 (hard, symbolic).

For the Jeans dispersion relation $ω^{2} = c_{s}^{2} k^{2} - 4 π G ρ_{0}$ , show that the maximum growth rate of an unstable mode is $γ_{m a x} = 4 π G ρ_{0}$ , and state the wavenumber at which it is achieved. Then explain why this maximum is unphysical in a finite, pressure-confined cloud.

Hint

An unstable mode has $ω = iγ$ with $γ = 4 π G ρ_{0} - c_{s}^{2} k^{2}$ ; maximise over $k$ in the unstable range $0 \leq k \leq k_{J}$ . Think about the size of the cloud relative to the longest wavelength.

Answer

In the unstable range $0 \leq k < k_{J}$ , the growth rate $γ (k) = 4 π G ρ_{0} - c_{s}^{2} k^{2}$ is a decreasing function of $k^{2}$ , so the maximum is at $k = 0$ , giving $γ_{m a x} = 4 π G ρ_{0} = 2 / t_{ff}$ where $t_{ff} = 3 π / (32 G ρ_{0})$ is the free-fall time. The maximiser is a perturbation of infinite wavelength, $λ \to \infty$ , which cannot fit inside any finite cloud. In a pressure-confined cloud of radius $R$ the longest allowed mode has $λ \sim 2 R$ , and the correct growth rate is finite; the $k = 0$ divergence is a symptom of the Jeans swindle (an infinite uniform medium cannot be in true static equilibrium). The physical statement, recovered in expanding or confined geometries, is that the collapse time is of order the free-fall time, not less.

Exercise 8 (hard, short answer).

A real molecular cloud has a one-dimensional velocity dispersion $σ_{v} \approx 1$ km/s, far above the thermal sound speed $c_{s} \approx 0.2$ km/s. Explain how replacing $c_{s}$ by $σ_{v}$ in the Jeans analysis (the turbulent Jeans mass) both raises the mass threshold for collapse and, for supersonic turbulence, produces a spectrum of dense shocks whose masses are lognormally distributed.

Hint

Consider separately (i) the support role of turbulent pressure on large scales and (ii) the compressive role of supersonic flows on small scales. The mass variance in isothermal turbulence grows as $σ_{s}^{2} = ln (1 + b^{2} M^{2} β / (β + 1))$ .

Answer

On large scales the turbulent velocity dispersion acts as an effective pressure, $P_{turb} \sim ρ σ_{v}^{2}$ , raising the signal speed in the Jeans mass from $c_{s}$ to $σ_{v}$ and increasing $M_{J}$ by the factor $(σ_{v} / c_{s})^{3} \sim 1 0^{2}$ , which is why GMCs of $1 0^{5} M_{⊙}$ are marginally bound rather than fragmenting thermally into $100 M_{⊙}$ pieces. On small scales, supersonic turbulence (Mach number $M = σ_{v} / c_{s} \sim 5$ ) produces a field of shocks that compress gas into dense sheets and filaments; the density field of isothermal turbulence is approximately lognormal, with variance $σ_{s}^{2} = ln (1 + b^{2} M^{2} β / (β + 1))$ in natural log of density, where $b \sim 0.3$ to $0.5$ depends on the mixture of compressive and solenoidal forcing and $β$ is the plasma beta. Gravitational collapse then selects the high-density tail of this lognormal, which gravitationally steepens into a power law. This turbulent Jeans framework is the basis of modern analytic IMF theories (Padoan and Nordlund; Hennebelle and Chabrier), which reproduce the Salpeter slope from the statistics of the turbulent density field.

Lean formalization Intermediate+

This unit has lean_status: none. Mathlib does not formalise the continuum mechanics, radiative transfer, or empirical scaling laws that appear here, and there is no current Lean path through the Jeans dispersion relation or the Salpeter and Kennicutt-Schmidt fits. A formalisation would first require axiomatising a self-gravitating barotropic fluid on a manifold with the Euler, continuity, and Poisson equations, then proving existence and the linearised spectrum; the IMF and Kennicutt-Schmidt relations, being observational fits, would enter as uninterpreted power laws rather than theorems. The content below is therefore verified by the human reviewer documented in the unit metadata and by the primary literature cited in the bibliography, not by a mechanised proof.

Advanced results Master

The free-fall time and the rate of star formation

The characteristic timescale on which an unopposed pressure-free cloud collapses is the free-fall time,

t_{ff} = \frac{3 π}{32 G ρ _{0}} .

Its derivation, given in the full proof set below, integrates the pressure-free collapse of a uniform sphere and matches the gravitational acceleration at the edge to the equation of motion. The free-fall time depends only on the density, not on the temperature, because in the absence of pressure every shell falls ballistically. Empirically, the star-formation-rate per free-fall time, $ϵ_{ff} = \dot{M}_{⋆} t_{ff} / M_{gas}$ , is small and remarkably constant across environments, $ϵ_{ff} \sim 0.01$ averaged over whole GMCs and $\sim 0.3$ to $1$ in individual dense cores. The constancy of $ϵ_{ff}$ across five orders of magnitude in galactic gas density is one of the deepest empirical regularities in star formation theory and is the principal observational constraint that any complete theory must reproduce.

The origin of the IMF

The near-universality of the IMF across the Milky Way and nearby galaxies (with the possible exception of the most metal-poor or most extreme starburst environments) is striking given the enormous range of cloud masses and metallicities involved. Three families of theory compete to explain it. Turbulent-fragmentation theories (Padoan and Nordlund 2002; Hennebelle and Chabrier 2008) derive the IMF from the lognormal density distribution of supersonic turbulence, supplemented by a magnetised shock-collapse criterion; the high-mass tail steepens into a Salpeter slope naturally. Competitive-accretion theories (Bonnell and collaborators) argue that the mass spectrum arises from accretion in a cluster potential well in which protostars near the centre accrete more. The opacity-limit theory (Low and Lynden-Bell 1976; Rees 1976) pins a characteristic mass to the thermal physics of the dust-gas transition and has been extended by Bate and others using radiation-hydrodynamics simulations. The modern consensus is that all three mechanisms operate, with turbulence setting the shape of the high-mass slope and thermodynamics (the opacity limit and the dust-gas coupling temperature) fixing the characteristic mass near the peak. The universality then follows from the universality of supersonic isothermal turbulence and of dust-gas thermodynamics across most galactic environments.

Feedback and self-regulation

Once the first massive stars form, they reshape their surroundings. Photoionisation, radiation pressure, stellar winds, and supernovae inject energy and momentum that heat and disperse the remaining gas, raise the local Jeans mass, and quench further star formation. On galactic scales this feedback sets the depletion time and the low value of $ϵ_{ff}$ : were it not for feedback, a molecular cloud would turn its gas into stars in a single free-fall time, $ϵ_{ff} \approx 1$ , whereas observations give $ϵ_{ff} \approx 0.01$ . Momentum-driven feedback from supernova explosions is thought to be the dominant regulator in Milky-Way-like disks, while radiation pressure and photoionisation dominate in the densest starbursts. The self-regulating balance between turbulent driving (by feedback and by galactic shear) and turbulent dissipation produces a quasi-equilibrium in which the star-formation rate adjusts until energy injection balances dissipation. This picture is the conceptual link between the microphysics of individual clouds and the macroscopic Kennicutt-Schmidt law that governs entire galaxies 28.03.01.

The Kennicutt-Schmidt law from cloud physics

The superlinear Kennicutt-Schmidt slope $N \approx 1.4$ emerges from combining the molecular fraction, the free-fall rate, and the scale-height structure of galactic disks. In the inner, molecule-dominated regions of disks the relation is approximately linear, $Σ_{SFR} \propto Σ_{H_{2}}$ , with a constant molecular depletion time of about 1 to 2 Gyr. The superlinearity on global scales reflects the dependence of the molecular fraction $f_{H_{2}}$ on total gas surface density and metallicity: at low $Σ_{gas}$ the gas is mostly atomic and forms stars slowly, while at high $Σ_{gas}$ it becomes molecular and forms stars at the dynamical rate. This decomposition explains why starburst galaxies, which are molecule-dominated, sit on the steep extension of the relation while quiescent disks span the shallow, molecule-regulated part. The dynamical content is that the cosmic star-formation history is set by the rate at which galaxies convert atomic gas into molecular gas and then into stars on the local free-fall time.

The role of magnetic fields

The ISM is magnetised, with field strengths of a few $μ$ G in the diffuse medium and up to milligauss in dense cores. Magnetic fields modify the collapse in two ways. On large scales they provide anisotropic support, allowing gas to flow along field lines but resisting compression across them; this slows star formation and lowers $ϵ_{ff}$ . On small scales the question of whether magnetic fields can hold up a collapsing core altogether is captured by the mass-to-flux ratio, $M /Φ$ , compared to the critical value $(M /Φ)_{crit} = 1/ (2 π G)$ . Cores with $M /Φ > (M /Φ)_{crit}$ are supercritical and collapse; subcritical cores cannot, unless ion-neutral drift (ambipolar diffusion) slowly lets neutrals slip past the field. Observations of dust polarization in dense cores suggest most are marginally supercritical, so magnetic fields delay but do not prevent star formation. The magnetic-field-mediated picture, the standard theory of the 1980s and 1990s (Mouschovias, Shu), has been partially superseded by the turbulent picture as observations revealed supersonic turbulence to be the dominant supporter of GMCs.

Synthesis. The star-formation problem is the foundational reason galaxies regulate their gas consumption across cosmic time: putting these together, the Jeans criterion, the IMF, and the Kennicutt-Schmidt relation form a single self-consistent pipeline from molecular-cloud microphysics to galactic luminosity. The central insight is that the SFR per free-fall time is nearly invariant across environments; this is exactly the mean-field limit of the turbulent Jeans analysis developed above, and it generalises to high-redshift starbursts, clump-cluster systems, and the first metal-free Population III stars 28.04.01. The bridge is that local physics, not global dynamics, sets the normalisation of the cosmic star-formation history that appears again in the chemical enrichment and reionisation of the universe.

Full proof set Master

Proposition (free-fall collapse time)

Let a uniform sphere of initial density $ρ_{0}$ and zero pressure begin at rest and collapse under its own gravity. The time for a shell initially at radius $r_{0}$ to fall to the centre is

t_{ff} = \frac{3 π}{32 G ρ _{0}},

independent of $r_{0}$ .

Proof. By spherical symmetry the mass interior to a shell is constant during collapse, $M (r) = (4 π /3) ρ_{0} r_{0}^{3}$ , so the shell obeys the radial equation of motion

\frac{d ^{2} r}{d t ^{2}} = - \frac{G M ( r )}{r ^{2}} = - \frac{4 π G ρ _{0} r _{0}^{3}}{3 r ^{2}} .

Multiply both sides by $d r / d t$ and integrate once, using the initial condition $\overset{r}{˙} (t = 0) = 0$ and $r (0) = r_{0}$ :

\frac{1}{2} \overset{r}{˙}^{2} = \frac{4 π G ρ _{0} r _{0}^{3}}{3} (\frac{1}{r} - \frac{1}{r _{0}}) .

Taking the negative root (collapse) and substituting $r = r_{0} cos^{2} θ$ with $θ$ running from $0$ at $t = 0$ to $π /2$ at $r = 0$ , we have $d r = - 2 r_{0} cos θ sin θ d θ$ and

\overset{r}{˙} = - \frac{8 π G ρ _{0} r _{0}^{3}}{3} (\frac{1}{r _{0} cos ^{2} θ} - \frac{1}{r _{0}}) = - \frac{8 π G ρ _{0} r _{0}^{2}}{3} tan θ .

Then $d t = d r / \overset{r}{˙} = 3/ (2 π G ρ_{0}) cos^{2} θ d θ$ , and integrating from $0$ to $π /2$ ,

t_{ff} = \frac{3}{2 π G ρ _{0}} \int_{0}^{π /2} cos^{2} θ d θ = \frac{3}{2 π G ρ _{0}} \cdot \frac{π}{4} = \frac{3 π}{32 G ρ _{0}} . □

Note that $r_{0}$ cancels: every shell reaches the centre simultaneously, which is the signature of pressure-free self-similar collapse and the reason the free-fall time depends only on the density.

Proposition (Bonnor-Ebert critical mass)

An isothermal sphere of sound speed $c_{s}$ confined by external pressure $P_{ext}$ has a maximum mass in hydrostatic equilibrium given by

M_{BE} \approx \frac{1.18 c _{s}^{4}}{( G ^{3} P _{ext} ) ^{1/2}} .

Sketch of proof. The isothermal Lane-Emden equation, obtained by combining hydrostatic equilibrium $d P / d r = - GM (r) ρ / r^{2}$ with $P = c_{s}^{2} ρ$ and mass continuity $d M / d r = 4 π r^{2} ρ$ , admits a one-parameter family of solutions parametrised by the central density $ρ_{c}$ . Multiplying the Lane-Emden equation by $ξ^{2} d ψ / d ξ$ (with $ρ = ρ_{c} e^{- ψ}$ , $ξ = r / r_{0}$ , $r_{0}^{2} = c_{s}^{2} / (4 π G ρ_{c})$ ) and integrating gives the first integral. Analysing the external pressure $P_{ext} = c_{s}^{2} ρ_{c} e^{- ψ (ξ_{max})}$ and the enclosed mass as functions of $ξ_{max}$ shows that $M$ at fixed $P_{ext}$ reaches a maximum at $ξ_{max} = 6.5$ , where the dimensionless combination $M^{2} G^{3} P_{ext} / c_{s}^{8}$ equals $1.40$ , i.e. $M_{BE} = 1.18 c_{s}^{4} / (G^{3} P_{ext})^{1/2}$ . Beyond this maximum the configuration is dynamically unstable to collapse, since $d M / d ρ_{c} < 0$ : lowering the central density requires raising the mass, so the equilibrium has no stable continuation. This critical mass is the rigorous analogue of the Jeans mass for a pressure-confined cloud and matches the observed masses of dense prestellar cores. $□$

Connections Master

To stellar evolution 28.02.01. The Jeans mass sets the boundary condition that stellar evolution inherits: the distribution of initial stellar masses is forged in the fragmentation cascade, and the protostars that emerge enter the main sequence governed by the structure equations of stellar evolution. The endpoint of every star, white dwarf, neutron star, or black hole, then returns processed gas to the ISM, seeding the next generation in a closed ecosystem. The same pressure-versus-gravity competition that decides whether a cloud collapses also decides whether a stellar core is supported, scaled to degeneracy pressure rather than thermal pressure.
To galaxies and galaxy formation 28.03.01. Star formation integrated across a galaxy produces the Kennicutt-Schmidt relation, which connects molecular gas reservoirs to luminosity and fixes the depletion time of galactic disks. Active galactic nuclei 28.03.03 pending and galaxy mergers can trigger or quench star formation by compressing or expelling molecular gas, so the small-scale physics of clouds modulates the large-scale morphology and colour of galaxies. The cosmic star-formation history, peaking near redshift 2, is essentially the integral of the Kennicutt-Schmidt law over the gas accretion history of the universe.
To cosmology and the first stars 28.04.01. The Jeans criterion appears again in the growth of primordial density perturbations in the expanding universe, where it sets the baryonic acoustic horizon and the minimum mass of the first dark-matter haloes able to cool and collapse. Population III stars, the first generation, formed in metal-free gas with much higher temperatures (set by molecular hydrogen cooling near 200 K rather than dust near 10 K), giving Jeans masses of order $100 M_{⊙}$ and producing the top-heavy IMFs inferred from early-universe simulations. The same physics that governs a local GMC therefore also governs reionisation and the chemical enrichment of the intergalactic medium.
To exoplanets and planet formation. The protoplanetary disks that surround young stars are the byproduct of the angular momentum retained during cloud collapse: a core contracts along the axis of rotation but cannot collapse perpendicular to it, leaving a disk of $0.01$ to $0.1 M_{⊙}$ from which planets assemble. The conditions in the disk (temperature, density, dust-to-gas ratio) are inherited from the molecular cloud, so the chemistry of the solar nebula, and hence the composition of Earth, traces back to the cold molecular phase of the ISM.
To chemistry and the origin of complex molecules. Cold molecular clouds are the most efficient chemical factories in the universe. On the surfaces of dust grains, atoms combine into H $_{2}$ , water, ammonia, methanol, and increasingly complex organic molecules; some 300 interstellar molecules have been identified by radio and infrared spectroscopy. The same radiative and chemical networks that set the cooling rate (and hence the temperature, and hence the Jeans mass) also produce the prebiotic inventory delivered to forming planetary systems, connecting interstellar chemistry to the origin of life.

Historical and philosophical context Master

The first quantitative theory of gravitational collapse is due to Jeans ^[Jeans1902], who in 1902 linearised the equations of a self-gravitating gas and obtained the instability that bears his name. Jeans's analysis contained what later authors called the "Jeans swindle," the assumption that a uniform infinite medium could be in static equilibrium; the criterion nevertheless survived because it is recovered rigorously in expanding cosmological backgrounds and in pressure-confined finite clouds. The modern, physically clean derivation of the Bonnor-Ebert critical mass for a confined isothermal sphere was given independently by Ebert (1955) and Bonnor (1956), placing the Jeans criterion on firm footing.

The empirical foundation of star-formation studies was laid by Schmidt ^{[Schmidt1959]}, who in 1959 proposed that the star-formation rate per unit volume scales as a power of the gas density. Schmidt's relation sat at the level of a plausible hypothesis until Kennicutt ^{[Kennicutt1998]}, in a 1998 paper combining star-formation tracers and molecular and atomic gas measurements for 97 galaxies, established the global surface-density relation with slope $1.4 \pm 0.15$ . The Kennicutt-Schmidt law transformed the subject by tying star formation to observable gas reservoirs on galactic scales and underpins modern models of galaxy formation.

The shape of the stellar mass distribution at birth was first measured by Salpeter ^{[Salpeter1955]} in 1955 from main-sequence star counts in the solar neighbourhood. Salpeter found a single power law of slope $- 2.35$ in $lo g N$ versus $lo g M$ . Later, Miller and Scalo (1979), Scalo (1986), Kroupa (2001), and Chabrier (2003) refined the low-mass end, showing that the IMF flattens below about $0.5 M_{⊙}$ and turns over near $0.3 M_{⊙}$ , but the high-mass Salpeter slope has survived essentially unchanged for seventy years. The apparent universality of the IMF across vastly different environments remains one of the strongest constraints on any theory of star formation.

The three-phase picture of the ISM was crystallised by McKee and Ostriker ^{[McKeeOstriker1977]} in a 1977 paper that modelled the ISM as cold clouds embedded in a warm medium and permeated by a hot, supernova-carved component. Their model explained the coexistence of widely separated temperature phases as a consequence of the cooling curve's bistability and of dynamical equilibrium between supernova heating and radiative cooling. The recognition that molecular clouds are supersonically turbulent, rather than quiescent and magnetically supported, came later, driven by observations of linewidth-size relations (Larson 1981) and by simulations of driven turbulence (Stone, Ostriker, and Gammie; Mac Low and Klessen). This turbulent paradigm, combined with the thermal physics of dust-gas coupling at the opacity limit, is the basis of the modern theory that ties the IMF to the statistics of the turbulent density field.

Philosophically, the study of star formation marks the point where astronomy crosses from describing objects to explaining their origin. The Jeans criterion is one of the rare places where a deep astronomical question, why stars have the masses they do, reduces almost completely to a comparison of two timescales. It is also where the universe becomes historical: every star more massive than the Sun was born after the Sun, and the cold molecular clouds we observe today are the assembly lines that will build the stars of the remote future.

Bibliography Master

@article{jeans1902,
  author  = {Jeans, J. H.},
  title   = {The Stability of a Spherical Nebula},
  journal = {Philosophical Transactions of the Royal Society of London Series A},
  volume  = {199},
  pages   = {1--53},
  year    = {1902}
}

@article{salpeter1955,
  author  = {Salpeter, Edwin E.},
  title   = {The Luminosity Function and Stellar Evolution},
  journal = {Astrophysical Journal},
  volume  = {121},
  pages   = {161},
  year    = {1955}
}

@article{schmidt1959,
  author  = {Schmidt, Maarten},
  title   = {The Rate of Star Formation},
  journal = {Astrophysical Journal},
  volume  = {129},
  pages   = {243},
  year    = {1959}
}

@article{kennicutt1998,
  author  = {Kennicutt, Robert C.},
  title   = {The Global {Schmidt} Law in Star-Forming Galaxies},
  journal = {Astrophysical Journal},
  volume  = {498},
  pages   = {541--552},
  year    = {1998}
}

@article{mckeeOstriker1977,
  author  = {McKee, Christopher F. and Ostriker, Jeremiah P.},
  title   = {A Theory of the Interstellar Medium --- Three Components Regulated by Supernova Remnant Velocities},
  journal = {Astrophysical Journal},
  volume  = {218},
  pages   = {148--169},
  year    = {1977}
}

@book{draine2011,
  author    = {Draine, Bruce T.},
  title     = {Physics of the Interstellar and Intergalactic Medium},
  publisher = {Princeton University Press},
  year      = {2011}
}

@book{stahlerPalla2004,
  author    = {Stahler, Steven W. and Palla, Francesco},
  title     = {The Formation of Stars},
  publisher = {Wiley-VCH},
  year      = {2004}
}

@article{mckeeOstriker2007,
  author  = {McKee, Christopher F. and Ostriker, Eve C.},
  title   = {Theory of Star Formation},
  journal = {Annual Review of Astronomy and Astrophysics},
  volume  = {45},
  pages   = {565--687},
  year    = {2007}
}

@article{kennicuttEvans2002,
  author  = {Kennicutt, Robert C. and Evans, Neal J.},
  title   = {Star Formation in the {Milky Way} and Nearby Galaxies},
  journal = {Annual Review of Astronomy and Astrophysics},
  volume  = {50},
  pages   = {531--608},
  year    = {2012}
}

@article{larson1981,
  author  = {Larson, Richard B.},
  title   = {Turbulence and star formation in molecular clouds},
  journal = {Monthly Notices of the Royal Astronomical Society},
  volume  = {194},
  pages   = {809--826},
  year    = {1981}
}

Prerequisites

28.02.01

Tier anchors

beginner: Bennett et al., The Cosmic Perspective (8e), Ch. 7 and 16 (the interstellar medium and star birth in plain English)
intermediate: Carroll and Ostlie, An Introduction to Modern Astrophysics (2e), Ch. 12; Draine, Physics of the Interstellar and Intergalactic Medium (2011), Ch. 1-2, 32-41
master: Stahler and Pala, The Formation of Stars (2004); Draine (2011); McKee and Ostriker (2007), ARAA 45, 565

References

Jeans 1902 — The Stability of a Spherical Nebula · Phil. Trans. R. Soc. Lond. A 199, 1-53 (1902). Linearises the continuity, Euler, and Poisson equations about a uniform self-gravitating gas and obtains the critical length scale $λ_{J} = c_{s} π / (Gρ)$ below which pressure dominates and above which gravity runs away.
Salpeter 1955 — The Luminosity Function and Stellar Evolution · ApJ 121, 161 (1955). From star counts in the solar neighbourhood the initial mass function follows $ξ (M) d M \propto M^{- 2.35} d M$ over $0.4 ≲ M / M_{⊙} ≲ 10$ .
Schmidt 1959 — The Rate of Star Formation · ApJ 129, 243 (1959). Proposes that the star-formation rate per unit volume scales as a power of the gas density, $\overset{ρ}{˙}_{⋆} \propto ρ_{g}^{n}$ , the relation later refined into the Kennicutt-Schmidt law.
Kennicutt 1998 — The Global Schmidt Law in Star-Forming Galaxies · ApJ 498, 541 (1998). Fits disk-averaged star-formation-rate and gas-surface-density measurements for 61 normal spiral and 36 starburst galaxies to $Σ_{SFR} \propto Σ_{gas}^{1.4 \pm 0.15}$ .
Draine 2011 — Physics of the Interstellar and Intergalactic Medium · Princeton University Press (2011). Chapters 1-2 survey the hot, warm, and cold phases; chapters 32-41 develop molecular clouds, the Jeans criterion, protostellar collapse, and accretion.
McKee and Ostriker 1977 — A Theory of the Interstellar Medium · ApJ 218, 148 (1977). The three-phase ISM model: a neutral warm medium, a hot ionised medium carved by supernovae, and cold clouds confined by pressure balance.

Estimated time

beginner: 25m
intermediate: 55m
master: 85m