Making gold

Foundations

A textbook that changed

If you learned any astronomy before about 2017, you were probably told that gold is made in supernovae. It is a good line, and until recently everyone said it, including textbooks and museums and science programmes. It now looks to be mostly wrong.

The correction is one of the cleanest examples in modern science of a question being settled by a single observation, and it is worth telling in full, because it shows what “we know this” actually costs.

The problem with heavy elements

Lesson 3.1 got the periodic table as far as iron by fusion, and about halfway up the rest by slow neutron capture in old giant stars. That leaves a problem. Slow capture cannot make gold, platinum, uranium or thorium in the amounts we see, because the path to them runs through nuclei so unstable that a slow process cannot cross the gaps: the nucleus decays before the next neutron arrives.

To get there you need the opposite: a flood of neutrons so intense that a nucleus swallows dozens of them in about a second, racing far out into the unstable region, and only then decaying back toward stability. This is the r-process, r for rapid, and B²FH identified it in 1957. What B²FH could not say was where in the universe it happens. Answering that took sixty years.

The obvious suspect

The obvious place was a core-collapse supernova. The collapsing core of a massive star is made almost entirely of neutrons, and the explosion should fling neutron-rich material outward. For decades this was the standard answer.

It never quite worked. Detailed simulations kept finding that the material blown out of a core-collapse supernova is not neutron-rich enough, because the flood of neutrinos from the newborn neutron star converts neutrons back into protons on the way out. The theorists kept adjusting; the answer kept refusing to come out right. Meanwhile a second suspect was proposed as early as the 1970s: two neutron stars, orbiting each other, spiralling closer over billions of years, and finally colliding. That would tear apart matter which is already almost pure neutron, at the density of an atomic nucleus, and throw it into space. But nobody had ever seen such a collision.

17 August 2017

At twelve minutes past two in the afternoon, Indian time, on 17 August 2017, the LIGO detectors in the United States and the Virgo detector in Italy registered a gravitational wave unlike the black-hole mergers they had seen before. This one chirped for a hundred seconds, rising in pitch: the signature of two objects far lighter than black holes, spiralling together. Two seconds later, a gamma-ray satellite recorded a short burst from the same part of the sky.

The alert went out. Within eleven hours a telescope in Chile found a new point of light in a galaxy a hundred and thirty million light-years away. Over the following weeks about seventy observatories on every continent and in orbit watched it fade, in radio, infrared, visible light, ultraviolet, X-rays and gamma rays. It was the first event in history observed in both gravitational waves and light.

And what they saw was a kilonova: a glow powered not by fusion but by the radioactive decay of freshly made heavy elements. Its colour changed from blue to red over days, exactly as predicted if a cloud of newly synthesised heavy nuclei was expanding and cooling. In 2019 a careful reanalysis of the spectra identified an individual element in it: strontium, made in that collision, in that second, a hundred and thirty million years ago.

The r-process had been caught in the act.

What it means for a wedding ring

The merger threw out something like a few hundredths of the Sun’s mass in heavy elements. Among it, by the proportions such events produce, was gold: plausibly several Earth masses of it, from one collision.

So the gold in a ring did not come from a supernova. It was made in the last second of a collision between two dead stars, somewhere in the Milky Way, before the Sun existed. It drifted in a cloud of gas for hundreds of millions of years, was swept into the disc that formed the solar system, and ended up in the Earth’s rocks. Nearly all of the gold that reached the Earth’s surface sank into the core during the planet’s molten youth; what miners find in the crust was delivered later, by asteroid impacts, which is a story for chapter four.

What supernovae do make

None of this makes supernovae unimportant. They are still the source of most of the oxygen you breathe, the magnesium and silicon in rock, and much of the iron in your blood, all made during the star’s life and in the explosion itself.

They also make the light show. The explosion is over in seconds, but a supernova shines for months, and the light comes from radioactivity. A core-collapse supernova makes about a tenth of a solar mass of nickel-56, which decays to cobalt-56 and then to iron-56 with a half-life of about seventy-seven days. The exploding white dwarfs of lesson 2.3 make six times more, which is why they are so bright and so uniform, and why they work as the standard candles that revealed dark energy. Between them, those two kinds of explosion made roughly half the iron in the universe each.

Honest uncertainty

One observation does not settle everything. The open question is whether neutron-star mergers alone can account for all the heavy elements, particularly in the very oldest stars, which formed early, when there had not been time for many neutron-star pairs to spiral together. Some of the oldest stars are surprisingly rich in r-process elements. That suggests at least one additional site, and the leading candidate is a rare kind of supernova from a rapidly spinning, strongly magnetised massive star. The accounting is not finished.

Deeper

The r-process path

The r-process requires neutron densities above roughly 10²⁰ to 10²⁴ per cubic centimetre, temperatures around a billion kelvin, and a duration of about one second. Under those conditions neutron capture is far faster than beta decay, so nuclei are driven to the neutron drip line, twenty or more neutrons beyond stability, where capture is halted by photodisintegration and an equilibrium is established. When the neutron flux ends, the highly unstable nuclei cascade back to stability through beta decays, and the abundance peaks at mass numbers 80, 130 and 195 mark the closed neutron shells at N = 50, 82 and 126 encountered along the path. Those peaks sit systematically below the corresponding s-process peaks, which is how the two processes are distinguished in observed abundance patterns.

Why supernova neutrinos spoil it

In the neutrino-driven wind above a proto-neutron star, the enormous flux of electron neutrinos and antineutrinos drives the ratio of neutrons to protons toward equality by converting neutrons to protons. Modern simulations find the outflow is at best mildly neutron-rich, giving a weak r-process that reaches strontium and yttrium but not the platinum peak or the actinides. This is the calculation that removed ordinary supernovae from the list of main sites.

The GW170817 event

The gravitational-wave signal implied component masses between about 1.2 and 1.6 solar masses, consistent with neutron stars, at a distance of about 40 megaparsecs. The gamma-ray burst arrived 1.7 seconds after the merger. The optical counterpart, in the galaxy NGC 4993, brightened within hours and reddened over days, consistent with two components: a fast, lighter, blue ejecta of lower opacity, and a slower, redder ejecta rich in lanthanides, whose complex atomic structure gives enormous opacity and shifts the emission to the infrared. Estimates of the total ejected mass cluster around 0.03 to 0.05 solar masses, with the split between the components model-dependent.

The strontium identification

Watson and colleagues modelled the kilonova spectra with the atomic data for candidate elements and found that the broad features around 350 and 810 nanometres in the first days were reproduced by singly ionised strontium at the observed velocities. Strontium is a light r-process element, made near the first abundance peak. Identifying individual heavy elements in kilonova spectra is difficult because the atomic data for lanthanides and actinides at these conditions are incomplete, which is itself an active area of laboratory work.

Nickel-56 and the light curve

Nickel-56 decays to cobalt-56 with a half-life of 6.1 days, and cobalt-56 to iron-56 with a half-life of 77.2 days. The gamma rays from these decays are absorbed by the expanding ejecta and re-emitted as visible light, producing the characteristic supernova light curve: a rise of about twenty days, then an exponential decline whose slope matches the cobalt-56 half-life. That match, seen in supernova after supernova, is the direct evidence that the iron of the universe is made this way.

The uranium clock

Uranium-238 and thorium-232, made only by the r-process, have half-lives of 4.5 and 14 billion years. Measuring their ratio to a stable r-process element such as europium in an extremely old star gives an estimate of how long ago that star’s r-process material was made. The technique gives ages of order 12 to 14 billion years for the oldest stars measured, consistent with, though less precise than, the ages from the main-sequence turn-off in lesson 2.7.

Frontier

How many sites

The abundance patterns of r-process elements in metal-poor stars are strikingly uniform above barium, suggesting a single dominant process, while the lighter r-process elements vary, suggesting more than one. Reconciling the rate and delay time of neutron-star mergers with the enrichment history of the Galaxy, particularly the presence of r-process elements in some very old stars and in some dwarf galaxies, is an open problem. Rare magnetorotational supernovae and collapsars are the leading additional candidates. More mergers, detected as gravitational-wave observatories improve, will constrain the rate directly.

Nuclear physics far from stability

The r-process path runs through thousands of nuclei that have never been made in a laboratory. Their masses, decay rates and neutron-capture cross-sections must be predicted by models, and different models give abundance predictions that differ by factors of several. Rare-isotope accelerator facilities are working outward toward the path, measuring the properties of nuclei that were previously only calculated.

Transcript and shot list

Target runtime about 6 minutes. Built as a correction with a date on it: the old answer, why it never worked, the single night that replaced it, and the part that is still not settled. Shot types: ANIM is rendered from code with Manim, AI is a generated shot from Higgsfield with the shared style reference, REAL is agency imagery with credit, CAM is the narrator.

[00:00] S01 · AI · A gold wedding ring on a dark surface, catching the light.

If you learned any astronomy before about 2017, you were probably told that the gold in a ring like this was made in a supernova. Everyone said it. Textbooks, museums, science programmes. It now looks to be mostly wrong. And the correction is one of the cleanest cases in modern science of a question being settled by a single night of observation.

[00:29] S02 · ANIM · Cosmic calendar; a marker before 1 September labelled “this gold already existed”.

Whatever made this gold, it did so before the Sun formed. Before the 1st of September on the cosmic calendar.

[00:38] S03 · ANIM · The periodic table with fusion filling up to iron, slow capture filling half of what’s above; gold, platinum and uranium stay grey.

Lesson 3.1 got us to iron by fusion, and about halfway up the rest by slow neutron capture in old giants. But slow capture can’t make gold, platinum or uranium in the amounts we see. The path to them runs through nuclei so unstable that a slow process can’t cross the gaps. The nucleus decays before the next neutron arrives.

[01:06] S04 · ANIM · A flood of neutrons; a nucleus swallows dozens in about a second, racing far off the stability line, then cascading back; label: the r-process.

To get there you need the opposite. A flood of neutrons so intense that a nucleus swallows dozens of them in about one second, racing far out into unstable territory, and only then decaying back. The rapid process. B²FH identified it in 1957. What they couldn’t say was where in the universe it happens. That took sixty years.

[01:32] S05 · ANIM · A core-collapse supernova; neutrinos stream through the outflow converting neutrons into protons; a “not neutron-rich enough” readout.

The obvious suspect was a supernova. The collapsing core is almost pure neutrons, and the blast should fling neutron-rich matter outward. For decades that was the standard answer, and it never quite worked. The simulations kept finding that the flood of neutrinos from the new neutron star converts neutrons back into protons on the way out. The material isn’t neutron-rich enough. The theorists adjusted. The answer kept refusing.

[02:04] S06 · ANIM · Two neutron stars orbiting, spiralling closer over a clock of billions of years.

There was a second suspect, proposed back in the 1970s. Two neutron stars, orbiting each other, spiralling closer for billions of years, and finally colliding. That would tear apart matter already at the density of an atomic nucleus and throw it into space. Perfect. But nobody had ever seen one.

[02:27] S07 · ANIM · 17 August 2017; a rising chirp waveform of 100 seconds; a gamma-ray burst 1.7 seconds later; a sky map narrowing; observatories light up around a globe.

Then, on the 17th of August 2017, at twelve minutes past two in the afternoon Indian time, the LIGO and Virgo detectors registered a gravitational wave unlike the black-hole mergers they’d seen. This one chirped for a hundred seconds, rising in pitch: two objects far lighter than black holes, spiralling together. Two seconds later a satellite caught a short gamma-ray burst from the same patch of sky. The alert went out. Within eleven hours a telescope in Chile found a new point of light in a galaxy a hundred and thirty million light-years away. Over the next weeks, about seventy observatories watched it fade. The first event in history seen in both gravitational waves and light.

[03:20] S08 · REAL · The optical counterpart of GW170817 in NGC 4993, before-and-after frames (observatory image; confirm credit line). S09 · ANIM · The kilonova’s colour shifting blue to red over days; a spectrum with the strontium feature marked.

And what they saw was a kilonova: a glow powered not by fusion but by the radioactive decay of freshly made heavy elements. Its colour shifted from blue to red over days, exactly as predicted if a cloud of newly made heavy nuclei was expanding and cooling. In 2019 a careful reanalysis found an individual element in the spectrum. Strontium. Made in that collision, in that second, a hundred and thirty million years ago. The rapid process, caught in the act.

[03:58] S10 · ANIM · The merger’s ejecta: a few hundredths of a solar mass; a scale showing several Earth masses of gold; the cloud drifts, joins a nebula, and becomes the solar system.

That merger threw out a few hundredths of the Sun’s mass in heavy elements. Among it, plausibly several Earth masses of gold. From one collision. So the gold in this ring wasn’t made in a supernova. It was made in the last second of a crash between two dead stars, somewhere in this galaxy, before the Sun existed. It drifted for hundreds of millions of years, was swept into the disc that made the solar system, and ended up in rock.

[04:35] S11 · ANIM · Nickel-56 decaying to cobalt-56 to iron-56; a supernova light curve rising and falling with the 77-day slope marked; two bars: core-collapse and type Ia, each making about half the iron.

None of which makes supernovae unimportant. They’re still the source of most of the oxygen you breathe, the magnesium and silicon in rock, and much of the iron in your blood. And they make the light show. The explosion is over in seconds, but a supernova shines for months, and that light is radioactivity: nickel-56 decaying to cobalt, then to iron, with a seventy-seven-day half-life you can read straight off the fading curve. Between exploding giants and exploding white dwarfs, they made about half the universe’s iron each.

[05:16] S12 · ANIM · Old metal-poor stars with unexpectedly high r-process content; a question mark; a rare magnetised supernova as a candidate.

One observation doesn’t settle everything. Here’s what’s still open. Some of the very oldest stars are surprisingly rich in these heavy elements, and they formed early, when there hadn’t been time for many neutron-star pairs to spiral together. That points to at least one more site. The leading candidate is a rare kind of supernova from a fast-spinning, strongly magnetised star. The accounting isn’t finished.

[05:46] S13 · ANIM · Labels: ESTABLISHED for the r-process and the merger detection; ACTIVE RESEARCH for how many sites and the nuclear physics far from stability.

The scorecard. That the heaviest elements need rapid neutron capture, and that neutron-star mergers do it: established, and seen directly in 2017. Whether mergers alone are enough, and the properties of thousands of nuclei on the path that have never been made in a laboratory: active research.

[06:07] S14 · AI into ANIM · The ring again; it dissolves into two neutron stars colliding; the calendar returns.

So next time someone tells you gold comes from supernovae, you can give them the better answer, and the date it changed. Every element now has an origin story, and they’re startlingly different from each other. Next lesson: the periodic table, rearranged by where each element was made.

Shot list

Two AI shots, eleven code-rendered shots, one observatory image. Estimated Higgsfield use: 3 to 6 generations including retries.