The explosion that never reaches the surface

A star is stable because it has a thermostat, and the thermostat is the ideal gas law. Burning too fast raises the temperature, the temperature raises the pressure, the pressure expands the layer, and the expansion cools it. Every nuclear burning stage in every star runs at exactly the rate that balance permits, and that is why the Sun has burned steadily for four and a half billion years rather than exploding in an afternoon.

There is one moment in a low-mass star’s life when the chain has a link missing.

The link that is missing #

Degeneracy pressure comes from the exclusion principle: electrons cannot share quantum states, so compressing a gas of them forces some into higher momentum states and the resulting pressure depends on the density alone. For a non-relativistic degenerate gas, $P = 1.0036\times10^{13}(\rho/\mu_e)^{5/3}$ dyn/cm², and the expression contains no temperature at all — this is the counting rule that holds a white dwarf up.

In a star that has finished hydrogen on the main sequence and is climbing the red giant branch, the helium core is exactly such a gas. It is inert, it is compressed by the weight of the envelope, and its pressure is degenerate. As the shell above it adds helium ash the core grows, and as it grows it contracts and heats — because a self-gravitating body has a negative heat capacity — until it reaches about $10^8$ K, which is where helium can burn — the same climb that makes the envelope swell as the core shrinks.

At that point the feedback loop is open. Heating raises the rate; the rate raises the temperature; the temperature does not raise the pressure; nothing expands; and the rate goes up again.

Why the rate is so steep #

The other half of the runaway is the temperature sensitivity, and helium burning is the most sensitive reaction in stellar physics.

Three helium nuclei must meet, which is a three-body reaction and would be hopelessly improbable if it happened all at once. It does not: two alphas make an unstable beryllium-8 nucleus that survives for $10^{-16}$ seconds, and a third arrives within that window and — because of a resonance whose existence was predicted from the fact that carbon exists — is captured.

The rate of a two-step reaction through a tunnelling barrier is dominated by the exponential of the Gamow factor, and for the triple-alpha the result is $\varepsilon \propto \rho^2 Y^3 T^{-3}e^{-4.4\times10^9/T}$. Its local logarithmic slope is $4.4\times10^9/T - 3$, which is 41 at $10^8$ K.

A one per cent rise in temperature is a fifty per cent rise in the burning rate. That number, meeting a pressure that does not respond, is the flash.

Where the energy goes #

The peak luminosity of a helium flash is around $10^{10}$ solar luminosities, briefly comparable to a small galaxy, and it happens inside a star that shows no change at all. That is the fact worth explaining.

The energy has three places it could go: out as radiation, into mechanical work, or into the thermal energy of the gas. The first is unavailable — the core is buried under half a solar mass of envelope whose radiative diffusion time is thousands of years, so on the flash’s timescale of seconds nothing escapes. The second is unavailable at first, because the core is degenerate and does not expand.

So it goes into heat, and heat is exactly what lifts the degeneracy. As the temperature climbs, the thermal energy per electron approaches and then exceeds the Fermi energy; the gas stops being degenerate; and at that moment the ideal gas law reconnects, the pressure responds, and the core expands and cools. The runaway is terminated by the thing it was creating.

The timescale of that termination is what makes the event survivable. A degenerate core can absorb an enormous amount of heat before its thermal energy becomes comparable to its Fermi energy — that is what “degenerate” means — so the temperature has a long way to climb before the pressure begins to notice, and the climb takes seconds rather than the milliseconds a genuinely unregulated exponential would take. By the time the gas has lifted, the temperature has reached about $3\times10^8$ K, the reaction’s sensitivity has fallen from 41 to 14, and the expansion that follows brings the core to a stable, non-degenerate, quietly burning state at roughly a tenth of the peak temperature. The total energy released is around $10^{41}$ J, which is enormous by the standards of a star’s ordinary output and is precisely what is needed to lift a $0.47\,M_\odot$ core out of degeneracy. Essentially none of it emerges.

What was actually measured #

The flash itself has never been observed and probably cannot be. What has been observed is its consequences, and they are sharp.

The tip of the red giant branch is at a fixed luminosity. Because the core mass at ignition is essentially the same for every low-mass star — about $0.47\,M_\odot$, set by the physics of degenerate helium rather than by the star’s total mass — the luminosity at which the flash occurs is nearly independent of mass and of composition. That makes the brightest red giant in an old population a standard candle, at $M_I \approx -4.05$, and it is now one of the primary rungs of the extragalactic distance ladder and one of the two independent routes to the Hubble constant.

The horizontal branch is horizontal. In a globular cluster’s colour–magnitude diagram, the post-flash stars lie along a nearly constant luminosity at $\sim 50\,L_\odot$, spread out in temperature. The constancy is the core mass being constant; the spread is how much envelope each star has left after its own mass loss on the giant branch.

The gap. Between the tip of the giant branch and the horizontal branch there are almost no stars in any cluster, because the transition takes about $10^5$ years against the $10^8$ years spent on either side. The emptiness of that region is a direct measurement of how fast the flash and its aftermath are.

The mass boundary #

Not every star flashes, and where the boundary lies is worth stating because it explains a discontinuity in the whole scheme of stellar evolution.

A star above about $2\,M_\odot$ reaches helium ignition temperature before its core becomes degenerate, because its core is hotter at every stage and its evolution is faster. It ignites helium quietly, in an ideal gas, with the thermostat intact — and it shows no flash, no drop in luminosity, and no horizontal branch. Instead it settles onto a red clump, and its subsequent evolution is the one that ends in a core-collapse rather than a white dwarf if it is heavy enough.

So the same element, ignited in the same reaction, produces a violent runaway in one star and a smooth transition in a slightly heavier one, and the difference is entirely a matter of whether the electrons got degenerate first.

The number that makes it a standard candle #

The tip-of-the-giant-branch distance method deserves its own paragraph, because it is the one place where this essay’s physics becomes a cosmological measurement.

The core mass at ignition is set by a competition that has nothing to do with the star as a whole. On one side, the degenerate core is heated by the shell burning above it and cooled by neutrino emission from its own centre; on the other, it is compressed as the shell adds ash. Both depend on the core’s own mass and on very little else, so the mass at which the temperature reaches $10^8$ K comes out at $0.47\,M_\odot$ for a star of $0.8\,M_\odot$ and $0.46\,M_\odot$ for one of $1.8\,M_\odot$ — a spread of two per cent across the whole range of stars old enough to be doing this.

A nearly fixed core mass gives a nearly fixed luminosity, because the shell’s output is a steep function of the core mass and of nothing else. In the I band, where the bolometric correction happens to run the other way and flattens the residual dependence on colour, the tip sits at $M_I = -4.05 \pm 0.05$. A five per cent distance from counting stars in a galaxy’s halo, requiring no variable stars, no light curves and no period measurement.

Where the model stops #

Three-dimensional simulations of the flash have changed the picture in ways one-dimensional models could not have shown.

The runaway does not begin at the exact centre. Neutrino losses are strongest where the density is highest, so the core is cooler at its centre than a little way out, and ignition occurs in a shell at about $0.2\,M_\odot$ from the middle. The burning front then has to propagate inward, which it does convectively, and the convection is violent enough to be genuinely three-dimensional.

Whether that convection reaches the hydrogen-rich envelope is an open question with consequences. If it does, hydrogen is dragged into helium-burning temperatures and a secondary flash follows, mixing carbon to the surface. Some models produce this and some do not, and the observational discriminant is a class of unusually carbon-rich metal-poor stars whose origin is still argued about. Three details of the picture are worth following, and each of them is a place where the physics comes from outside the star.

The neutrinos that decide where it starts #

The three-dimensional models put the ignition off-centre, and the reason is a process that removes energy from the core without any nuclear physics in it.

At the densities and temperatures of a degenerate helium core, energy is lost to neutrinos — produced not by fusion but by processes involving the electrons and the plasma itself, in which a photon-like excitation decays into a neutrino pair. Those neutrinos leave without interacting, so the process is a pure sink.

The loss rate depends steeply on density, so it is strongest at the centre. The heating from the shell above is strongest at the outside of the core. Between them they make the core’s temperature profile non-monotonic: it has a maximum somewhere partway out, and that is where the flash begins.

The same losses do something more consequential. They cool the core as it grows, so the core has to become more massive before it reaches ignition temperature than it would if it were merely being compressed — and the core mass at ignition is what fixes the luminosity of the tip of the giant branch.

That makes the standard candle described above a function of a neutrino emission rate. Change the rate and the core mass at ignition changes, the tip luminosity changes, and every distance measured with it moves.

The dependence has been used in the other direction. Because the tip luminosity is measured and the stellar physics is otherwise well constrained, comparing the two bounds any additional energy-loss channel — a neutrino magnetic moment, an axion coupling — that would cool the core faster than the standard processes do. Some of the tightest astrophysical limits on such couplings come from exactly this comparison.

A cosmological distance rung depends on a weak-interaction rate, and the same measurement constrains particle physics that no accelerator reaches.

None of the three is a detail of the modelling, and two of them are the largest uncertainties in the observational consequences.

The mass that is lost before it happens #

The horizontal branch is horizontal because the core mass is fixed, and it is spread out in temperature because the envelope mass is not. What sets the envelope mass is how much the star lost climbing the giant branch, and that is the least well determined quantity in the whole picture.

A red giant has a very extended, loosely bound envelope, and it loses mass through a slow wind driven by some combination of radiation pressure on dust and pulsation. The rate is parameterised by an empirical formula fitted to observed giants, with a coefficient that is adjusted to make cluster models match cluster observations.

The consequence is that the location of a star on the horizontal branch is determined by a fitted parameter rather than by a computed one. A star that lost more envelope arrives hotter and bluer; one that lost less arrives cooler and redder.

That would be a modest embarrassment if the spread were small. It is not: some clusters have horizontal branches concentrated on the red side, others on the blue, and some are bimodal with a gap in the middle. Metallicity explains part of the variation and demonstrably not all of it, and the remainder has been called the second-parameter problem since the 1960s.

Candidates for the second parameter include the cluster’s age, its helium abundance, the density of its core and whether the stars concerned are members of the multiple populations that globular clusters turn out to contain. Several probably operate together, and no combination has been shown to account for every cluster.

The event this essay is about is understood to a per cent and its most conspicuous observational consequence has an unexplained parameter in it, and the unexplained part is not the nuclear physics but the wind.

The carbon that stays put #

The flash makes carbon, and the star’s surface shows none of it. Where the carbon goes is worth following, because the answer is that it waits.

The convection during the flash is confined to the core: it extends outward from the ignition point to the core’s edge and stops there, at the composition discontinuity where the hydrogen-rich envelope begins. So the carbon produced is mixed through the helium core and nowhere else.

The star then spends its horizontal-branch lifetime burning helium quietly in that core, building more carbon and some oxygen, with the envelope entirely uninvolved.

What eventually brings the carbon out is a later and different mechanism. On the asymptotic giant branch, after core helium is exhausted, a helium-burning shell flashes repeatedly, and each flash drives a convective zone that reaches — in stars of the right mass — into the hydrogen-rich layers above. The envelope’s own convection then carries the enriched material to the surface.

That is dredge-up, and it is what turns a star into a carbon star: an object whose photosphere contains more carbon than oxygen, which changes its chemistry entirely, because the two lock up into carbon monoxide and whichever is in excess forms the molecules that dominate the spectrum.

So the carbon made in the flash reaches the surface tens of millions of years later, by a process the flash has nothing to do with, in a star that has by then left the horizontal branch. A nucleosynthetic event and its observational signature are separated by a whole evolutionary phase, which is one reason the connection took so long to establish.

A closing point about what the steepness in this essay is for. A rate that goes as the fortieth power of temperature is not merely sensitive; it is a thermostat with enormous gain. In a normal star, where pressure depends on temperature, a small overshoot expands the core and shuts the rate down, and the star sits at whatever temperature makes its luminosity match what the envelope can radiate. The exponent sets how tightly it is held there — and a star burning by the CNO cycle, at an exponent near eighteen, is held less tightly than one burning helium at forty and more tightly than one burning hydrogen by the proton–proton chain at four. The core temperature of a star is not a free parameter; it is the fixed point of a feedback whose gain is the exponent on this plot.

One more reading covers a denser core, where the degeneracy the flash depends on is stronger still.

Where this ladder goes next #

The rung below has the core shrinking and the envelope swelling as one event; this rung is what happens when the shrinking core stops being able to respond. The rungs above go two ways.

One follows the star: after the horizontal branch comes the asymptotic giant branch, where a shell of helium burns on top of a carbon core and does so unstably, in a series of thermal pulses — flashes again, but this time repeated every hundred thousand years for a million years, and this time with an observable consequence, since the pulses dredge carbon and s-process elements to the surface.

The other follows the mechanism. A thermal runaway in degenerate matter with no way to expand is also what a nova is, on a white dwarf’s accreted hydrogen layer, and what a Type Ia supernova is, on a whole white dwarf. The same disconnected thermostat produces an event nobody can see, an event visible across a galaxy, and an event visible across the universe, and the difference is only how much degenerate matter is available to run away at once.