A star is held up by its own weight

The usual account of a star has the causation the wrong way round. Nuclear reactions release energy, the energy heats the interior, as the furnace’s own arithmetic says, the heat provides pressure, the pressure holds the star up.

Run that argument and a star is unstable in an obvious way: any increase in the reaction rate adds pressure, which expands the star, which reduces the density, which changes the rate — and nothing in the story says which way that feedback goes or where it settles.

The correct account starts at the other end. A star is a self-gravitating mass of gas, and the requirement that it be neither collapsing nor exploding fixes the pressure at every depth. The pressure fixes the temperature, through the gas law. And the temperature then determines the reaction rate, which is a consequence rather than a cause.

The star does not burn at a rate and thereby acquire a temperature. It has a temperature, and burns at whatever rate that temperature dictates.

The balance, in one equation #

Take a thin shell at radius $r$, of thickness $dr$ and density $\rho$. Its weight, per unit area, is the mass inside pulling on it:

$$dW = \frac{GM(r)\rho}{r^2}\,dr,$$

where $M(r)$ is the mass interior to $r$ — and it is the mass interior because a spherical shell exerts no net force on anything inside it, which is what makes the problem one-dimensional at all.

For the shell not to accelerate, the pressure below must exceed the pressure above by exactly that amount:

$$\frac{dP}{dr} = -\frac{GM(r)\rho(r)}{r^2}.$$

That is hydrostatic equilibrium, and it is the single most important equation in stellar structure. Everything else in a star follows from it plus a statement about how pressure relates to density and temperature.

The equation is a statement about accelerations, and it is worth appreciating how good the approximation is. If the pressure support of the Sun were removed entirely, it would collapse on the free-fall timescale

$$t_{\text{ff}} \approx \sqrt{\frac{3\pi}{32G\bar{\rho}}} \approx 30\ \text{minutes}.$$

The Sun has been shining for 4.6 billion years, which is $8\times 10^{13}$ free-fall times. Any imbalance between pressure and gravity would have to be smaller than one part in $10^{13}$ to have gone unnoticed. Hydrostatic equilibrium is not an approximation in any practical sense; it is the most precisely satisfied statement in astrophysics.

An order-of-magnitude central pressure, from nothing #

The equation can be integrated crudely before it is integrated properly, and the crude version gets the answer nearly right.

Replace the derivative with a difference across the whole star: $dP/dr \approx -P_c/R$, take $M(r) \approx M$ and $\rho \approx \bar{\rho} = 3M/4\pi R^3$. Then

$$P_c \sim \frac{GM\bar{\rho}}{R} = \frac{3GM^2}{4\pi R^4}.$$

For the Sun that is $4.4\times 10^{14}$ Pa — four billion atmospheres. The correct answer from a solar model is $2.5\times 10^{16}$ Pa, so the estimate is low by a factor of 57, which for a calculation containing no physics beyond a dimensional argument is a reasonable outcome. (A rigorous version, the virial theorem, gives a strict lower bound of $GM^2/8\pi R^4$, and the shortfall is because the mass is far more concentrated than uniform.)

Now use the ideal gas law, $P = \rho k T/\mu m_H$, to convert that into a temperature:

$$T_c \sim \frac{GM\mu m_H}{kR}.$$

For the Sun, with $\mu = 0.6$ for ionised solar-composition gas, this gives about 12 million kelvin against the true 15.7. The relation $T_c \propto M/R$ is the useful part, and it contains no reference whatever to nuclear physics.

Why the fuel does not set the temperature #

That last point deserves its own statement, because it inverts the textbook order.

The central temperature is fixed by $M/R$. The main-sequence mass–radius relation is roughly $R \propto M^{0.8}$ over the range from a tenth of a solar mass to a few tens, so

$$T_c \propto M^{0.2}.$$

A star thirty times the mass of the Sun has a core barely twice as hot. Across the whole main sequence — a factor of a thousand in mass and ten million in luminosity — the central temperature varies by a factor of four.

Set against that, the reaction rate is a fierce function of temperature: the proton–proton chain goes as $T^4$ and the CNO cycle as $T^{17}$ or steeper. So a factor of four in temperature is a factor of $10^{10}$ in the CNO rate, which is exactly the enormous range of luminosities the main sequence spans.

The causal chain is therefore: mass fixes the balance, the balance fixes the temperature, the temperature fixes the burning rate, and the burning rate is the luminosity. The mass–luminosity relation is that chain, compressed into one power law.

The stability that makes a star a star #

The balance is not merely satisfied; it is stable, and the mechanism is worth stating because it is the reason stars are not bombs.

Suppose the core is momentarily too hot. The reaction rate rises steeply, energy output rises, the core expands. But an expanding core does work against gravity, and by the virial theorem a self-gravitating gas that expands cools. The temperature drops, the reaction rate falls back, and the perturbation is damped.

The essential ingredient is the negative heat capacity: adding energy to a self-gravitating gas makes it colder. That is what supplies the thermostat, and it exists only because the pressure comes from a gas whose temperature is free to respond.

Take that away and the thermostat fails. In a degenerate core the pressure no longer depends on temperature, so heating the core does not expand it, so the rate rises without limit. That is the helium flash in a low-mass red giant — a runaway that releases $10^{11}$ solar luminosities in seconds, all of it absorbed by lifting the degeneracy — and it is the same failure mode that detonates a Type Ia supernova, the candle the distance ladder leans on.

Solving it properly: the Lane–Emden equation #

To go beyond estimates, the balance needs a relation between pressure and density. Assume a polytrope, $P = K\rho^{1+1/n}$, and the two equations combine into one:

$$\frac{1}{\xi^2}\frac{d}{d\xi}\left(\xi^2\frac{d\theta}{d\xi}\right) = -\theta^n,$$

with $\theta$ a dimensionless temperature and $\xi$ a dimensionless radius. Integrating outward from $\theta(0)=1$, $\theta'(0)=0$ until $\theta$ reaches zero gives the surface, at $\xi_1$.

Three values of $n$ have closed forms — $n=0$, 1 and 5 — and everything else is numerical. The two useful ones are both numerical: $n = 3$ is the Eddington standard model, appropriate when radiation pressure and gas pressure keep a fixed ratio, and it is a fair likeness of a radiative star like the Sun; $n = 1.5$ describes a fully convective star, which is what a low-mass red dwarf and a red giant’s envelope are — and what decides how long either lasts.

The roots are worth quoting because they are checkable: $\xi_1 = 3.65375$ for $n=1.5$ and $6.89685$ for $n=3$. The second is the source of the density concentration in the hero figure: for $n=3$ the central density is 54.2 times the mean, which is why 90% of the mass sits inside half the radius and why “the interior of a star” means its innermost tenth.

The polytrope is not a solar model — it has no energy transport, no composition, no opacity — but it captures the one thing the balance alone determines, which is how mass is distributed when a gas of a given compressibility holds itself up.

The pressure that is not a gas pressure #

The balance says what the total pressure must be. It does not say what supplies it, and in the most massive stars a second contributor takes over.

Photons carry momentum, so a radiation field exerts a pressure

$$P_{\text{rad}} = \frac{1}{3}aT^4,$$

which rises as the fourth power of the temperature while gas pressure rises only as the first. In the Sun’s core radiation supplies 0.03% of the support. In a star of 10 solar masses it is about 10%; at 100 solar masses it is more than half.

Two consequences follow, and both are why the upper main sequence looks the way it does.

The first is the Eddington limit. If the outward radiative force on the gas exceeds gravity, the star cannot stay in balance and drives a wind:

$$L_{\text{Edd}} = \frac{4\pi G M m_p c}{\sigma_T} = 3.2\times 10^4 \left(\frac{M}{M_\odot}\right) L_\odot.$$

Since luminosity rises much faster than mass — roughly as $M^{3.5}$ — the two curves cross, and above about 150 solar masses a star exceeds its own limit. That is the observed upper end of the stellar mass function, and it is a hydrostatic statement rather than a statement about how stars form.

The second is that a star supported largely by radiation is close to being a polytrope of index $n = 3$ exactly, because that is what a fixed ratio of radiation to gas pressure produces. Eddington’s standard model is not an arbitrary choice; it is the structure a massive star is driven toward.

What was actually measured #

The interior of a star is not visible. Every statement above is an inference, and until the 1970s the only test available was that the models produced the right luminosity and radius at the surface — a two-number check on a theory with many parameters.

Helioseismology changed that. The Sun’s surface oscillates in millions of acoustic modes, standing sound waves trapped in cavities whose depth depends on the mode. Each mode’s frequency depends on the sound speed along its path, and the sound speed is $c_s^2 = \Gamma_1 P/\rho$ — which is exactly the ratio the balance fixes. Measuring the frequencies and inverting gives the sound speed as a function of depth.

The result is a direct measurement of the Sun’s internal structure, and it agrees with the standard solar model to better than 0.1% in sound speed over most of the interior. That is the check that hydrostatic equilibrium, together with the equation of state and the opacities, is right.

It also produced a genuine failure, which is more informative. The inversion fixes the depth of the convection zone at $0.713 \pm 0.001$ solar radii and the surface helium abundance at $0.2485 \pm 0.0035$. When the solar photospheric abundances were revised downward in 2005 — a spectroscopic result, from 3D model atmospheres — the resulting solar models disagreed with the helioseismic sound speed by up to 1.4%, ten times the observational error. That is the solar abundance problem, and it is still not resolved: either the spectroscopy is wrong, or the opacities are, or something is missing from the models. The point is that the interior is now measured well enough that a change to an input can be rejected.

The same technique now works on other stars. Asteroseismology with Kepler and TESS measures oscillation frequencies for hundreds of thousands of red giants, and two easily measured quantities — the large frequency separation $\Delta\nu \propto \sqrt{\bar{\rho}}$ and the frequency of maximum power $\nu_{\max} \propto g/\sqrt{T_{\text{eff}}}$ — give a mass and a radius for each. Both scaling relations come straight out of the balance.

The generalisation: any self-gravitating thing #

The equation is not about stars. It is about anything held up against its own gravity by a pressure gradient, and it applies unchanged to objects that have nothing else in common.

The Earth’s interior is in hydrostatic equilibrium to high accuracy, and its pressure profile — 364 GPa at the centre — comes from the same integration with a different equation of state.

A planetary atmosphere is the same equation with $M(r)$ constant, giving the exponential barometric law and the scale height $H = kT/\mu m_H g$.

A giant molecular cloud — the thing an initial mass function counts — is in the same balance until it is not, and the mass at which it fails is the Jeans mass, which is where star formation starts.

A galaxy cluster’s hot intracluster gas is in hydrostatic equilibrium in the cluster’s potential, and measuring its temperature and density profile in X-rays gives the total mass — which is how a great deal of the evidence for dark matter was obtained.

In every case the same structure holds: the balance fixes the pressure, an equation of state converts it to a temperature, and everything else follows. What changes between them is only the equation of state.

Three timescales, and why they are so far apart #

A star has three characteristic times, and the enormous ratios between them are what make the whole subject tractable.

The dynamical time is the free-fall time computed above: thirty minutes for the Sun. It is how long the star takes to respond mechanically to anything, and it is why hydrostatic equilibrium holds so precisely — any imbalance is corrected within half an hour.

The thermal time is how long the star could shine on its stored gravitational energy alone, which is the total gravitational binding energy divided by the luminosity. For the Sun that is about thirty million years. It is how long the structure takes to adjust to a change in the energy transport, and it is the timescale on which a star that stops burning would contract.

The nuclear time is how long the fuel lasts: ten billion years for the Sun. It is how long the star exists.

The ratios are $10^{-5}$ and $10^{-3}$, and each of the three is a different subject. Because the dynamical time is so short, the mechanical structure can be treated as instantaneous and the equations become a boundary-value problem rather than an initial-value one. Because the thermal time is short compared with the nuclear one, the thermal structure adjusts to the composition rather than the reverse, and a star’s evolution can be computed as a sequence of equilibrium models rather than as a dynamical simulation.

That separation is what makes stellar evolution a calculation anybody can perform. Remove it — as happens in the last hours of a massive star’s life, when the nuclear timescale falls below the thermal one and eventually below the dynamical one — and the star must be simulated rather than modelled, which is why the final stages of stellar evolution are the least settled part of the subject.

The thermal timescale also settled a nineteenth-century argument. Kelvin and Helmholtz computed it, found thirty million years, and concluded that the Sun could be no older — against a geological record that demanded far longer. The discrepancy stood until nuclear energy supplied a fourth reservoir, three hundred times deeper. The timescale was calculated correctly and the conclusion was wrong, because the energy source it assumed was not the one operating.

Where the model stops #

Spherical symmetry. A rotating or magnetised star is not spherical, and the equation becomes two-dimensional. For rapid rotators — Achernar is 56% wider at its equator than at its poles — this is not a small correction.

No energy transport. The balance says nothing about how energy gets out. Adding that requires the radiative transfer equation and the opacity, and it is where the real difficulty in stellar modelling lives.

A polytrope is not a star. The assumption $P = K\rho^{1+1/n}$ is a stand-in for the real relation between pressure, density and composition, and it is right only in special cases.

The figures show a static object, and a star is not one. The profiles here are a snapshot of one moment in a life of billions of years, during which composition changes and the structure changes with it. Nothing on these axes carries time, and a star’s whole story is what these curves do as it passes.

One more polytrope sits between the two the essay has drawn.

The ladder from here #

Later rungs on this anchor: the four equations of stellar structure, and the Vogt–Russell theorem. The virial theorem, and negative heat capacity in full. Radiative transport, opacity and the Kramers law. The Schwarzschild criterion, and where a star convects. The Eddington limit as a hydrostatic statement about radiation pressure. Helioseismology’s inversions, and the solar abundance problem. Asteroseismic scaling relations. Degenerate cores, and the thermostat’s failure.

Eddington put the argument in The Internal Constitution of the Stars in 1926, before anyone knew what powered a star at all. His central temperature was right; his energy source was wrong. That is the strength of the argument this essay is about: the balance fixes the interior whatever is happening in it, and the nuclear physics arrived a decade later to fill a slot whose dimensions had already been measured.