The popular version of where gold comes from has become almost a set piece. The metal in a wedding ring was forged in the collision of two neutron stars, drifted through space for billions of years, and ended up in the Earth’s crust to be mined. It is a good story, and parts of it are well supported. But it states as settled fact several things that are still being argued over, and the careful version is the more interesting one.

Two separate questions are worth keeping apart. Where is gold made in the universe, and how did the particular gold now in the Earth’s crust get to where a miner could reach it. Neither has the clean single answer the set piece implies.

What the 2017 neutron star merger actually showed

Gold is one of a group of heavy elements built by the rapid neutron capture process, or r-process, in which atomic nuclei are flooded with neutrons faster than they can decay. This requires an environment dense with free neutrons, which rules out the ordinary fusion that powers stars. For decades the site of the r-process was an open question.

In August 2017, the gravitational-wave observatories LIGO and Virgo detected the merger of two neutron stars, an event catalogued as GW170817, and telescopes across the world recorded its afterglow, a kilonova. Analysis of that glow, including modelling reported by Daniel Kasen and colleagues in a 2017 paper in Nature, showed the merger had produced heavy r-process elements, the family that includes gold and platinum. This was the first direct confirmation that neutron star mergers make these elements at all.

That is a real result, and it is worth stating precisely. The 2017 event confirmed that mergers are an r-process site. It did not establish that they are the only source, and the size of their contribution is still being tested.

Why mergers cannot be the whole story

Several problems sit between that single confirmed event and the claim that your gold came from a merger.

Neutron star mergers are rare, and depending on how quickly binary systems merge after forming, they can struggle to explain how r-process elements appeared so early in some ancient stars and young galaxies. In GW170817 itself, some analyses found the heaviest elements appeared underproduced relative to the pattern seen in ancient stars in our own galaxy’s halo. The arithmetic does not obviously close.

This has pushed attention toward other candidate sites. One is a class of rare, energetic supernovae from rapidly rotating, highly magnetised collapsing stars. Another emerged in 2025, when Anirudh Patel, Brian Metzger and colleagues argued that giant flares from magnetars, neutron stars with extreme magnetic fields, can also drive the r-process. They tied the idea to a bright flare from the magnetar SGR 1806-20 recorded in 2004, and estimated that such flares might account for somewhere between one and ten per cent of the galaxy’s r-process material. The work was reported through the Flatiron Institute’s Center for Computational Astrophysics and published in the Astrophysical Journal Letters.

This is recent work resting on a single historical flare and a model, not a settled share of the cosmic budget. The honest position is that gold is made by the r-process in more than one kind of violent, neutron-rich environment, and the relative contribution of each is genuinely unresolved.

You cannot trace one ring to one event

There is a further limit that holds even if the budget is one day worked out. The gold in any particular object is a mixture, drawn from material that was stirred together in the gas of the galaxy over billions of years before the Sun formed. A single wedding ring’s atoms did not arrive in a parcel from one collision. They are an average of many events of possibly several different kinds. The claim that a given ring’s gold came from a neutron star merger is not something that could be checked for that ring even in principle.

How gold reached the part of Earth we can mine

The second half of the story has its own complication, and it is less widely known.

When the Earth formed and its iron core separated out, gold should have gone with it. Gold is siderophile, meaning it binds readily with iron, so the great majority of the planet’s gold is thought to have sunk into the core during formation, out of any conceivable reach. By that logic the crust and mantle should hold far less gold than they do.

The leading explanation for the excess is a late veneer: a flux of meteorites that struck the Earth after the core had already formed, depositing fresh metal into the outer layers where it stayed. Matthias Willbold and Tim Elliott of the University of Bristol set out evidence for this in a 2011 paper in Nature, comparing tungsten isotope ratios in nearly four-billion-year-old rocks from Greenland against modern rock. The difference they measured fits the addition of meteoritic material after core formation.

On that reading, the accessible gold a miner pulls from the crust was delivered by asteroid bombardment, long after the Earth itself had taken shape. The late veneer is a well-supported hypothesis rather than a closed case, and the details, including how much arrived and exactly when, are still discussed. But it complicates the tidy image of gold drifting in untouched from the pre-solar cloud and simply settling into the ground.

What survives the scrutiny

Strip out the overstatement and a sound version remains. The gold in a ring is older than the Sun. It was built by the r-process in one or more violent, neutron-rich settings, of which neutron star mergers are a confirmed example and magnetar flares and rare supernovae are live candidates. The portion now within reach in the Earth’s crust was probably delivered to the surface by meteorite bombardment after the planet’s core formed.

That is less neat than a single collision before the Sun, and it leaves the central question, which sources made most of the gold, still open. What will narrow it is more events of the kind seen in 2017 and 2004, observed in finer detail. A future nearby magnetar flare, caught quickly enough across gamma-ray, ultraviolet and optical wavelengths, would give astronomers a much cleaner test of the idea than the distant 2017 merger allowed.