Astronomy & Celestial Mechanics
the sky, computed
Stars

The resonance that had to exist

Three alpha particles cannot meet at once, and the two-body intermediate falls apart in 8×10⁻¹⁷ seconds. That is a hundred thousand times longer than a crossing time, which is just enough — provided a nuclear level sits at 7.65 MeV. Hoyle argued from the existence of carbon to the existence of the level, and it was found where he said.

Assumes Fusion and Degeneracy.

16 min read 8 figures Scales that defeat intuitionLight is the only messenger

Hydrogen burning is straightforward in outline: protons collide, and there is a stable nucleus at every mass number along the way from 1 to 4. Helium burning is not, because there is nothing at 5 and nothing at 8.

Add a proton to helium-4 and the result, helium-5 or lithium-5, is unbound: it comes apart in 10−2110^{-21} seconds, which is about the time it takes light to cross a nucleus. Add another alpha instead and beryllium-8 is also unbound, though not by nearly as much. So the ladder from helium upward has its first two rungs missing, and a star that has exhausted its hydrogen has a core of the most tightly bound light nucleus there is — the ash of the reaction that ran for billions of years — with no accessible way forward.

Ninety-three thousand electron-volts

The margin by which beryllium-8 is unstable is the smallest important number in stellar physics.

Two helium-4 nuclei have a combined binding energy of 8×7.074=56.598 \times 7.074 = 56.59 MeV. Beryllium-8 has 8×7.0624=56.508 \times 7.0624 = 56.50. The difference is 93 keV in favour of staying apart — which on the scale of nuclear energies is nothing at all, one part in six hundred.

But it is positive, so ⁸Be is unbound and decays. And it is small, so it decays slowly by nuclear standards: 8.2×10−178.2\times10^{-17} seconds, against the 10−2110^{-21} or so that an unbound nucleus with megaelectron-volts to spare would take. Beryllium-8 lives about a hundred thousand times longer than it has any right to.

That factor of 10510^5 is the entire mechanism. In a helium core at 10810^8 K, alpha–alpha collisions are constant, and each one produces a ⁸Be that survives for 8×10−178\times10^{-17} s before falling apart. At equilibrium there is a standing population — about one ⁸Be for every 10910^9 alphas — and a third alpha arriving during that window can be captured.

Nothing in this is a three-body collision. A genuine simultaneous encounter of three nuclei is vanishingly rare. What happens is two sequential two-body reactions with a very short-lived intermediate, which is a different thing and is many orders of magnitude more probable.

An exponent of 41 at 10⁸ K — and 1 at 10⁹ K. The triple-alpha rate against temperature, normalised to 1 at 10⁸ K, on a logarithmic scale — with the proton–proton chain and the CNO cycle drawn on the same axes for scale. The exponent is not a constant: the local logarithmic slope of ε ∝ T⁻³ exp(−4.4027×10⁹/T) is 4.4027×10⁹/T − 3, which runs from 108 at 4·10⁷ K down to 1 at 10⁹, passing 41.0 where helium actually burns. It is differentiated here rather than quoted, and checked against a finite difference on the drawn curve. A slope of 41 means a 1.7% rise in temperature doubles the energy output, and that is the whole of why the helium flash is a flash: in a degenerate core the temperature rise does not push the gas apart, so nothing turns the rate back down, and the run-away stops only when degeneracy is lifted. On the main sequence, where the gas is ideal, the same steepness is what keeps a star's centre at almost exactly one temperature.
Fig. 1 The rate the mechanism produces, against the two hydrogen-burning chains for scale. The exponent is not a constant: the local logarithmic slope of ε∝T−3e−4.4027×109/T\varepsilon \propto T^{-3}e^{-4.4027\times10^9/T} is 4.4027×109/T−34.4027\times10^9/T - 3, running from 108 at 4×1074\times10^7 K down to 1 at 10910^9, and passing 41 where helium actually burns. It is differentiated here rather than quoted, and checked against a finite difference on the drawn curve. A slope of 41 means a 1.7 per cent rise in temperature doubles the output.

The level that had to be there

Even with the equilibrium population of ⁸Be, the arithmetic did not work. Salpeter’s calculation in 1952 gave a triple-alpha rate far too small to produce the observed abundance of carbon — short by orders of magnitude.

Fred Hoyle’s argument, in 1953, ran in the direction that makes it famous. Carbon exists; it is 0.5 per cent of the mass of the universe and it is the fourth most abundant element. There is no other route to it. Therefore the reaction must be far faster than Salpeter’s estimate. The only thing that can make a nuclear reaction orders of magnitude faster is a resonance — a state of the compound nucleus at very nearly the energy the incoming particles bring. Therefore carbon-12 has an excited state at about 7.7 MeV, with the right spin and parity, and nobody has found it.

He took the argument to Willy Fowler’s group at Caltech, who were reluctant and then did the experiment. The state is at 7.654 MeV, spin-parity 0+0^+, and it is 287 keV above the ⁸Be + α threshold.

This is one of a very small number of successful predictions made from the existence of an observer’s raw material. It is often described as an anthropic argument, and it is worth being precise about what kind: the premise is not “life exists” but “carbon is abundant”, which is an ordinary astronomical observation. What makes it unusual is the direction — from an abundance measured in stars to a property of a nucleus, across a gap of thirty orders of magnitude in scale, with a falsifiable number at the end.

Energy generation against core temperature. The proton–proton chain and the CNO cycle, in solar units, against core temperature on logarithmic axes. The CNO curve is far steeper, so the two cross at 18.8 million kelvin — above that temperature a star runs mostly on CNO, and below it mostly on pp.
Fig. 2 The two chains helium burning has to follow, and why the transition is abrupt. The furnace that runs cooler than a compost heap does so because the proton–proton chain goes as T4T^4 and can therefore run gently at 15 million K. The CNO cycle goes as T18T^{18} and takes over above about 17 million. Triple-alpha goes as T41T^{41} and does not run at all until 10810^8 K, at which point it runs violently. Each successive fuel needs a higher temperature and has a steeper rate, so each stage of a star’s life is shorter and more sudden than the one before.

Why the flash is a flash

An exponent of 41 is not merely large; it is large enough to change what kind of process the burning is.

In an ordinary gas, a rise in temperature raises the pressure, the gas expands, and the expansion cools it. That is a thermostat, and it is why hydrogen burning on the main sequence is stable: the core cannot run away, because running away expands it and the expansion turns the rate back down. A T4T^4 rate under that thermostat is exceptionally placid.

In a degenerate gas the thermostat is disconnected. The pressure of a degenerate electron gas comes from a counting rule and not from the temperature, so raising the temperature raises the pressure hardly at all, the core does not expand, and nothing turns the rate back down.

Now put a T41T^{41} reaction into that. A one per cent rise in temperature raises the energy output by half again; the extra energy raises the temperature further; the rate rises again. The core of a low-mass star igniting helium under degeneracy reaches a luminosity of about 101110^{11} solar luminosities within seconds — comparable to a whole galaxy — for about a hundred seconds, and none of it escapes: it all goes into lifting the degeneracy. Once the electrons are non-degenerate the thermostat reconnects, the core expands, and helium burning settles down to something orderly.

A star whose core is degenerate at helium ignition therefore has a completely different history from one whose core is not, and the dividing line is about 2 solar masses, one more place where mass decides everything by deciding which equation of state applies. Below it, the helium flash; above it, a quiet ignition. That is one of the sharpest branch points in stellar evolution and it comes out of the interaction between a counting rule and an exponent.

A furnace with the thermostat taken out. Left, the two pressures in a red giant's helium core at 10⁶ g/cm³. The ideal-gas pressure rises with temperature, as the whole regulation of a star depends on it doing; the degenerate electron pressure is a horizontal line, because the exclusion principle does not know the temperature. At the ignition point near 10⁸ K the degenerate term is 5.1 times the gas term, so a temperature rise there raises the total pressure by almost nothing and the core does not expand. Right, what that costs. Both curves start at the same temperature with the same triple-alpha rate, whose logarithmic slope is 41 — the famous T⁴⁰, differentiated here rather than quoted. The regulated core settles; the degenerate one, with nowhere to put the extra energy, goes vertical. The flash reaches 10¹⁰ solar luminosities for a few seconds and not one photon of it is seen: every erg goes into lifting the degeneracy, and the star's visible response is to become fainter and settle on the horizontal branch.
Fig. 3 What the resonance does once the temperature reaches it. In a degenerate core the pressure does not depend on the temperature, so the energy the reaction releases raises the temperature without expanding the gas — which raises the rate, which raises the temperature again. The loop closes only when degeneracy lifts, some hundred million kelvin later. The triple-alpha reaction’s enormous temperature exponent is what makes it close so violently, and that exponent is a direct consequence of the resonance Hoyle predicted.

What was actually measured

Three separate measurements underwrite everything above, and none of them is astronomical.

The mass of beryllium-8. Measured by nuclear mass spectrometry and by reaction Q-values, and the 93 keV instability is the difference of two numbers each known to a fraction of a keV. Its width, and therefore its lifetime, comes from the energy spread of the ⁸Be resonance observed in alpha–alpha scattering: a resonance 5.65.6 eV wide corresponds by the uncertainty relation to 8.2×10−178.2\times10^{-17} s.

The Hoyle state. Found in 1953 by Ward Whaling’s group at Caltech, using the ¹⁴N(d,α)¹²C reaction and measuring the energies of the emitted alphas. The state’s energy, 7.654 MeV, is now known to about 0.5 keV; its spin-parity was established later, and 0+0^+ is what the reaction requires because two spin-zero particles at low relative energy bring in zero angular momentum.

The radiative width. The rate depends not just on the state existing but on how often it decays to the ground state of ¹²C by emitting a gamma ray rather than falling back apart into three alphas. That branching ratio is about 4×10−44\times10^{-4} — the state overwhelmingly falls apart — and measuring it is what actually pins the reaction rate. It is still the dominant uncertainty in the triple-alpha rate today, at the few-per-cent level, which propagates into every calculated carbon-to-oxygen ratio.

The astronomical observation is the abundance itself, and it is read from what is missing in a spectrum like every other abundance. Carbon is measured in stellar photospheres by the strengths of the CH and C₂ molecular bands and of the neutral carbon lines, in the interstellar medium by absorption against background stars, and in meteorites by direct chemistry. All three agree that carbon is about 3×10−33\times10^{-3} of the mass — and that number, which is what Hoyle argued from, is a spectroscopic measurement subject to all the usual difficulties of turning a line strength into an abundance. The exponent that makes the flash possible is worth reading over a wider range of temperature than the argument strictly needs.

An exponent of 41 at 10⁸ K — and -1 at 2.5·10⁹ K. The triple-alpha rate against temperature, normalised to 1 at 10⁸ K, on a logarithmic scale — with the proton–proton chain and the CNO cycle drawn on the same axes for scale. The exponent is not a constant: the local logarithmic slope of ε ∝ T⁻³ exp(−4.4027×10⁹/T) is 4.4027×10⁹/T − 3, which runs from 41 at 10⁸ K down to -1 at 2.5·10⁹, passing 41.0 where helium actually burns. It is differentiated here rather than quoted, and checked against a finite difference on the drawn curve. A slope of 41 means a 1.7% rise in temperature doubles the energy output, and that is the whole of why the helium flash is a flash: in a degenerate core the temperature rise does not push the gas apart, so nothing turns the rate back down, and the run-away stops only when degeneracy is lifted. On the main sequence, where the gas is ideal, the same steepness is what keeps a star's centre at almost exactly one temperature.
Fig. 4 The triple-alpha rate taken up to two and a half billion kelvin. The exponent of 41 that dominates at ignition has fallen away entirely by the top of the range, where the rate is nearly flat and then falls: all the helium in the way has already been consumed. The runaway is a property of the ignition temperature and not of the reaction.
A furnace with the thermostat taken out. Left, the two pressures in a red giant's helium core at 10⁶ g/cm³. The ideal-gas pressure rises with temperature, as the whole regulation of a star depends on it doing; the degenerate electron pressure is a horizontal line, because the exclusion principle does not know the temperature. At the ignition point near 10⁸ K the degenerate term is 5.1 times the gas term, so a temperature rise there raises the total pressure by almost nothing and the core does not expand. Right, what that costs. Both curves start at the same temperature with the same triple-alpha rate, whose logarithmic slope is 41 — the famous T⁴⁰, differentiated here rather than quoted. The regulated core settles; the degenerate one, with nowhere to put the extra energy, goes vertical. The flash reaches 10¹⁰ solar luminosities for a few seconds and not one photon of it is seen: every erg goes into lifting the degeneracy, and the star's visible response is to become fainter and settle on the horizontal branch.
Fig. 5 The two pressures at solar metallicity rather than at a globular cluster’s. The crossing where degeneracy takes over the pressure moves, and with it the mass at which a star’s helium core is degenerate when it ignites — which is why the flash happens in stars below about two solar masses and not above.

The flash happens at a fixed luminosity, and that is a distance

The helium flash has an observational consequence that has nothing to do with nuclear physics and a great deal to do with cosmology, and it follows from one property of degenerate matter.

A degenerate core’s temperature depends on its mass rather than on the mass of the star around it, because the pressure holding it up is set by density alone and the density is set by weight. So every low-mass star ignites helium when its inert helium core reaches essentially the same mass — about 0.47 solar masses — regardless of whether the envelope around it is half a solar mass or two.

The star’s luminosity while ascending the red-giant branch is set by the shell burning just outside that core, and it too depends almost entirely on the core mass. So the luminosity at the instant of ignition is nearly the same for every star that gets there: the ascent stops at a fixed brightness.

That is the tip of the red giant branch, and on a colour–magnitude diagram of any old stellar population it appears as a sharp edge — stars all the way up to a certain absolute magnitude and nothing above it, because everything that reached that point left within a few million years for the horizontal branch.

An edge at a known absolute magnitude is a standard candle. Measuring the apparent magnitude of the tip in a galaxy’s halo gives its distance, with an absolute calibration around −4.0-4.0 in the infrared and a scatter of a few hundredths of a magnitude. The method works in any galaxy with an old population resolved into stars, which is most of them within about 20 megaparsecs, and it needs no period, no light curve and no repeat observation.

It is now one of the two principal routes to calibrating type Ia supernovae, and the two routes — this and Cepheids — give expansion rates that differ by more than their stated errors. The tip’s sharpness is a direct consequence of the degeneracy that makes the flash a flash, and a disagreement in cosmology is currently resting on it.

Energy generation against core temperature. The proton–proton chain and the CNO cycle, in solar units, against core temperature on logarithmic axes. The CNO curve is far steeper, so the two cross at 36.3 million kelvin — above that temperature a star runs mostly on CNO, and below it mostly on pp.
Fig. 6 And the hydrogen chains in a star made before any carbon existed. The CNO cycle is catalytic — it needs the carbon the triple-alpha reaction makes — so the first generation of stars had no access to it and burned hydrogen by the slower proton–proton chain alone. That is the ordering the resonance imposes on cosmic history: carbon before the cycle that uses carbon, and the resonance before either.

What Hoyle actually argued

The prediction is told so often as an anthropic argument that it is worth separating what was claimed from what was inferred.

Hoyle’s premise was an abundance: carbon exists, at about three parts in a thousand by mass, and it is made in stars. His chain was that the only route to carbon runs through the triple-alpha reaction; that the reaction’s rate without a resonance is far too slow to produce the observed amount in the available time; and that a resonance therefore exists, at an energy his arithmetic placed near 7.7 MeV, with the spin and parity the reaction requires.

Every step of that is a deduction from a measured quantity through a computable rate. The conclusion is a statement about the energy levels of the carbon nucleus, and it was checked at Caltech within months by an experiment designed to look exactly there — and Fowler, who ran the group, is said to have thought the idea absurd before agreeing to test it.

What the story is usually reshaped into is different: that the resonance is finely tuned, that a small change in it would produce a universe without carbon and therefore without observers, and that its existence is evidence about the universe rather than about the nucleus. That is a separate claim, made later and by others, and the nuclear calculations mentioned below do not support it strongly.

The distinction matters because the two arguments have opposite structures. Hoyle’s runs from an observation to a prediction that could have been refuted by an experiment. The anthropic version runs from an existence to an explanation and predicts nothing. The first is one of the cleanest pieces of scientific reasoning in this collection; the second is what it turned into on the way to being famous.

The two are not even in competition, since a prediction that survives its test needs no further defence. What the retelling costs is the example: a piece of reasoning that shows exactly how an abundance measured in a spectrum can constrain a nuclear energy level is worth more as a method than as an anecdote about luck.

Where the model stops

The rate is not one number. At temperatures below about 10810^8 K the equilibrium assumption for ⁸Be breaks down — the population has not had time to establish — and the reaction has to be treated as a genuine three-body process, which is much slower and much harder to compute. That regime matters for accreting white dwarfs and for the coolest helium burning, and the standard tabulated rate is not valid there.

Resonances above the Hoyle state. At higher temperatures further ¹²C levels contribute, and the simple single-resonance formula is a low-temperature approximation. The tabulated rates carry additional terms for exactly this.

Carbon does not stay carbon. The reaction ¹²C(α,γ)¹⁶O competes for the same alphas, and its rate is the single largest uncertainty in all of stellar nucleosynthesis — known to perhaps 20 per cent, because the relevant energy is far below anything a laboratory can reach directly and the cross-section has to be extrapolated across three orders of magnitude. The ratio of carbon to oxygen that a star produces depends on the ratio of two rates, and one of them is not well known. Everything downstream — the composition of a white dwarf, the yield of a supernova, the amount of carbon available to make anything — inherits that uncertainty.

And the resonance is not fine-tuned in the way the story suggests. Modern calculations that vary the fundamental constants and re-derive the ¹²C spectrum find the Hoyle state’s position is more robust than the anecdote implies: substantial changes in the strength of the nuclear force move it by less than the width of the window in which carbon production works. The prediction was correct and the inference from it to a finely balanced universe is a separate claim that the nuclear physics does not obviously support.

What the picture cannot show

The intermediate. Nothing on this page draws the ⁸Be. It exists for 8×10−178\times10^{-17} seconds, at a concentration of one part in 10910^9, and it is the object the entire mechanism turns on. A figure of nuclear masses shows a gap where it would be, which is the closest a drawing gets.

The competition. The triple-alpha curve is drawn alone, and in a real helium core it is racing ¹²C(α,γ)¹⁶O for the same fuel. The outcome — how much carbon survives — is a ratio of two rates, and a plot of one of them says nothing about it.

The flash. A hundred seconds at 101110^{11} solar luminosities, entirely absorbed by the core it happens in, visible from outside as nothing at all. There is no observable to plot. What is seen is the star’s position on the horizontal branch afterwards, which is a consequence several thousand years downstream. The same machinery drawn on the two chains that come before helium closes the argument.

Energy generation against core temperature. The proton–proton chain and the CNO cycle, in solar units, against core temperature on logarithmic axes. The CNO curve is far steeper, so the two cross at 18.8 million kelvin — above that temperature a star runs mostly on CNO, and below it mostly on pp.
Fig. 7 The proton–proton chain and the CNO cycle over the range in which massive stars burn. Both are steep and CNO is far steeper, so above the crossing at 18.8 million kelvin a star’s luminosity is set almost entirely by a cycle that needs carbon to exist — which is the reaction the resonance in this essay is what made possible.
An exponent of 41 at 10⁸ K — and 1 at 10⁹ K. The triple-alpha rate against temperature, normalised to 1 at 10⁸ K, on a logarithmic scale — with the proton–proton chain and the CNO cycle drawn on the same axes for scale. The exponent is not a constant: the local logarithmic slope of ε ∝ T⁻³ exp(−4.4027×10⁹/T) is 4.4027×10⁹/T − 3, which runs from 108 at 4·10⁷ K down to 1 at 10⁹, passing 41.0 where helium actually burns. It is differentiated here rather than quoted, and checked against a finite difference on the drawn curve. A slope of 41 means a 1.7% rise in temperature doubles the energy output, and that is the whole of why the helium flash is a flash: in a degenerate core the temperature rise does not push the gas apart, so nothing turns the rate back down, and the run-away stops only when degeneracy is lifted. On the main sequence, where the gas is ideal, the same steepness is what keeps a star's centre at almost exactly one temperature.
Fig. 8 The same three rates at two per cent of solar metallicity. The CNO curve drops in proportion to the catalyst available and the proton–proton chain does not, so the crossing moves to a higher temperature and the first generation of stars had no CNO cycle at all. The triple-alpha rate is unmoved, because it needs no catalyst — only helium, and three of it.

Where the ladder goes next

Later rungs on this anchor: the Gamow peak, and why a nuclear reaction rate in a star is set by a small overlap between a falling Maxwell tail and a rising tunnelling probability. The ¹²C(α,γ)¹⁶O rate as a measurement problem, and the underground accelerators built to attack it. The alpha ladder beyond oxygen — neon, magnesium, silicon — and where it stops. Photodisintegration, and why silicon burning is an equilibrium rather than a chain. The s-process and r-process, which build everything past iron by neutron capture rather than by fusion at all. And the helium flash’s fate in a star of two solar masses, where the core is not degenerate and nothing dramatic happens.

Hoyle’s prediction is sometimes told as a lucky guess and it was nothing of the kind. It was a deduction from an abundance, through a rate, to a nuclear level, in a chain each link of which was quantitative — and it was published as a prediction, tested by an experiment designed to falsify it, and confirmed within months. The observation it started from was the amount of carbon in the universe, which is not a subtle thing to notice and had been sitting there the whole time.

Where this goes next

Energy generation against core temperature. The proton–proton chain and the CNO cycle, in solar units, against core temperature on logarithmic axes. The CNO curve is far steeper, so the two cross at 18.8 million kelvin — above that temperature a star runs mostly on CNO, and below it mostly on pp. The furnace that runs cooler than a compost heap The Sun's core produces 276 watts per cubic metre. A human body produces more. The Sun is bright because it is enormous, and its fusion is astonishingly slow — which is exactly why it lasts. The mirror: an envelope 62 times larger around a core 1.48 times smaller. Envelope radius and core radius against helium-core mass for a 1 solar-mass star, both in solar radii on one logarithmic axis. Two curves with opposite slopes at every point: the envelope grows from 2.0 to 122 solar radii — 0.57 astronomical units — while the degenerate core contracts from 13,848 to 9,368 kilometres, and the two are separated by a factor of 364 in the middle of the range. The figure is drawn from two stated relations rather than from a stellar model: a core-mass–luminosity law of the sixth power, calibrated to 2,500 solar luminosities at 0.48 solar masses of core, and a Hayashi temperature falling from 5000 to 3700 kelvin across the range. Everything else follows exactly — the radius by the Stefan–Boltzmann law, the core radius by the cold degenerate relation R ∝ M⁻¹ᐟ³. Note what the tip radius is and is not: it is the red-giant branch, and the later asymptotic-giant tip is a different and larger number. The star that swells because its centre shrank A helium core contracts, and the envelope around it expands by a factor of sixty. The central density rises at the same time as the radius, which is the opposite of what a self-gravitating body is expected to do, and a burning shell between the two is the whole reason it happens. Radius against mass for a degenerate star. The mass–radius relation for electron-degenerate matter. More mass gives a smaller star, and the radius reaches zero at 1.46 solar masses — the Chandrasekhar limit, solved from the same expression that draws the curve rather than quoted alongside it. A star held up by a rule about counting A white dwarf makes no energy and does not collapse. What holds it up is not heat or pressure in any ordinary sense — it is a quantum rule forbidding two electrons from occupying the same state.

About the same objects

Not linked from either essay — found by the objects both name.

What links here

Essays that link to this one from their own argument.

The objects this essay names

Each one links to every other essay that touches it.

Binding energyDegeneracyFusionHelium flashHoyle stateNuclear resonanceNucleosynthesisStellar evolutionTemperature sensitivityTriple alpha