Where did gold come from?

Fusion inside a star releases energy up to iron and costs energy beyond it. That is why a massive star’s core collapses once it has built an iron core: there is no fuel left that pays.

So nearly everything heavier than iron is assembled by neutron capture rather than fusion. Ordinary stars do some of it slowly, over hundreds to thousands of years, and make about half of those elements that way. The other half, gold among them, needs a flood of free neutrons captured within seconds, in something more violent than an ordinary star. The candidates are merging neutron stars and rare kinds of collapsing star.

LIGO and Virgo recorded a gravitational-wave signal from two neutron stars spiralling together. The discovery paper gave a combined signal-to-noise of 32.4 and a false-alarm rate below one per eighty thousand years; in the collaboration’s 2019 reanalysis, one of its searches puts the rate below one per ten million years.

Less than two seconds later, a gamma-ray burst. Within hours, telescopes across the world found the source in a galaxy about 130 million light years away, and watched it for weeks.

The glow that followed is called a kilonova, and its colour changed in the way models of freshly made heavy elements predict: blue at first, reddening over days, which the models read as the heaviest and most opaque material taking over.

One well-observed merger does not fix a cosmic budget. Estimating the total contribution means knowing how often neutron stars merge, how much each event throws out, and how that rate has changed over the history of the galaxy.

Those numbers are improving and are not settled. Some models need an additional source, such as a rare kind of collapsing rapidly rotating star, to explain the heavy elements found in the oldest, most metal-poor stars, where mergers have struggled to account for them.

Neutron star mergers are not prompt: two massive stars must live, die, and then spiral together, which can take billions of years. Yet very old, metal-poor stars already contain heavy elements.

This is a genuine tension rather than a settled story with a footnote. A 2019 study could not reproduce how europium, a heavy element made the same way, built up in the galaxy’s disk with mergers alone and the spread of delays short gamma-ray bursts imply. It proposed an extra source, active early and fading later. A 2020 model of the elements from carbon to uranium reached a similar verdict: mergers make the heaviest elements, but too slowly to explain the most metal-poor stars. Another team, reading the delays off the neutron star pairs in our own galaxy, finds that at least 40 per cent of all such pairs merge within a billion years, which it argues is enough to match how heavy elements built up in the Milky Way.

Nobody saw the remnant of the 2017 merger. What it was gets argued backwards, from what the light that followed rules out. In Margalit and Metzger’s version, it cannot have collapsed straight to a black hole, because that leaves too little ejecta to make the kilonova that was seen. It cannot have been a long-lived neutron star either, because the rotational energy such a star must shed on its way down would have gone into the jet or the ejecta, and the observations do not have room for it. What is left in the middle is a massive neutron star that lasted a short time and then collapsed. That middle is not unanimous: a 2024 model hides a long-lived star’s lost spin in a narrow ring of outflow that Earth would not see until late, and finds it fits the observations so far.

That reasoning cuts both ways: run in the other direction it is how Margalit and Metzger put a ceiling on how massive a neutron star can be. The ceiling depends on how matter behaves at the densities inside a neutron star, which is not known, and that is why an object nobody observed can constrain nuclear physics.

Before this, no single source had been caught in both gravitational waves and light. This event was observed in gravitational waves, gamma rays, visible light, infrared, ultraviolet, X-rays and radio, and each channel constrained a different part of the story.

The result is a template. Ask what a violent event should look like in every channel at once, then wait, and a single occurrence can settle a question that decades of indirect argument could not.