Neutron Star Mergers and Kilonovae: Inside GW170817

On 17 August 2017 a gravitational-wave chirp and a gamma-ray burst arrived 1.74 seconds apart. What two colliding neutron stars really told us.

Table of Contents

Eleven Hours in August

At 12:41:04 universal time on 17 August 2017 the two LIGO detectors recorded a rising wail that had been climbing through their sensitive band for about 100 seconds, starting at 24 Hz. Precisely 1.74 ± 0.05 seconds after it ended, the Gamma-ray Burst Monitor on NASA’s Fermi satellite triggered on a short, weak burst of gamma rays. Two completely different instruments, built on different continents by different communities to detect different things, had just reported the same event.

Then the real work started. Roughly 70 observatories across the world and in orbit were pointed at the search region. Eleven hours after the merger the one-metre Swope Telescope in Chile found a new point of light on the edge of the galaxy NGC 4993 that had not been there before. For the first time in the history of the subject, a gravitational wave source had a face.

What Virgo Barely Heard

The localisation is the part most summaries get wrong. Virgo did not miss GW170817 — it registered a weak signal, and that weakness is exactly what made the discovery possible. A two-detector network can only constrain a source to a long thin banana on the sky. The Virgo interferometer has blind spots where its antenna pattern nearly vanishes, so an almost-absent signal in Virgo, combined with a loud one in both LIGO instruments, said the source had to sit in one of those holes. That cut the search region to roughly 30 square degrees — a patch a few hundred times the area of the full Moon, which sounds enormous but is small enough for a survey telescope to tile in a night.

The gravitational waveform also gave a distance, because the amplitude of a chirp encodes how far away it was. With the position of NGC 4993 imposed as a prior, that distance is 43.8 Mpc. It is worth keeping separate from the independent electromagnetic distance to the same galaxy: surface-brightness fluctuations put NGC 4993 at 40.7 ± 1.4 ± 1.9 Mpc. The two agree, but they are different measurements of different things, and conflating them throws away the entire point of having both.

A Cloud of Fresh Nuclei, Glowing

What the telescopes were watching is called a kilonova, and the one from GW170817 got the designation AT2017gfo. Around 0.05 M☉ of debris — a mass comparable to all the planets in the solar system, several thousand times over — was flung off at 10 to 30 percent of the speed of light. Nothing in that cloud is burning. It glows because it is radioactive.

The merger squeezes matter so neutron-rich that nuclei capture free neutrons faster than they can beta-decay, racing up the chart of isotopes into territory that does not exist on Earth. Once the neutrons run out, those nuclei decay back towards stability, and their collective radioactive decay is the heat source. Because the cloud is also expanding at a fifth of the speed of light, its photosphere grows while cooling fast, and the colour tells you the temperature. AT2017gfo peaked near magnitude 17 about half a day in, with a photosphere above 10,000 K, and within five days had dropped below 2,500 K. The blue light collapsed by more than a magnitude a day. The infrared hung on for a week and then faded too.

GW170817 / GRB 170817A / AT2017gfo — one event, three messengers Logarithmic time axis, 72 pixels per decade on both sides of the merger. Light curves from an expanding-blackbody model at 40.7 Mpc. gravitational-wave frequency 2 kHz 200 Hz 20 Hz inspiral: 100 s of chirp from 24 Hz apparent AB magnitude of the kilonova 16 19 22 25 27 g (blue) i (red) K (infrared) Swope discovery, 10.9 h merger GRB 170817A, +1.74 s −100 s −10 s −1 s 1 s 10 s 100 s 17 min 2.8 h 1.2 d 12 d time from merger +1.74 s short gamma-ray burst LIGO and Virgo Fermi GBM optical, g band infrared, K band signal has passed below the band burst in progress no counterpart reported yet no counterpart reported yet
+1.74 s
Drag the scrubber from left to right and watch the order of events: the chirp climbs for 100 seconds, the gamma rays arrive 1.74 seconds after contact, and then the kilonova appears eleven hours later and reddens as its photosphere cools from above ten thousand kelvin to under two thousand.

The Element Question

This is where popular accounts of GW170817 go furthest off the rails. The headline claim — that neutron star mergers forge the heaviest elements — is sound. The claim that this event was watched making gold is not, because no gold line has ever been detected in AT2017gfo.

What was detected is strontium. In 2019 Darach Watson and colleagues identified a triplet of absorption features near one micron as singly ionised strontium, element 38, in the early spectra. That is the one secure identification of a specific heavy element in any kilonova to date, and it is enough to establish the rapid neutron capture process, because strontium at that abundance cannot be made any other way in that environment.

Gold and platinum were looked for deliberately. In 2021 James Gillanders and colleagues modelled their expected signatures and compared them with the observed spectra, and found nothing. Their paper reports upper limits of roughly 10⁻² M☉ of gold and a few times 10⁻³ M☉ of platinum. Upper limits are not measurements. Meanwhile the published and press estimates of how much gold the merger actually made span something like 1 to 200 Earth masses, depending on which ejecta mass you adopt, what you assume about the abundance pattern of the heaviest nuclei, and how generous the writer feels. A two-order-of-magnitude spread is not a number; a single figure quoted to three digits is a model, not an observation. The r-process in mergers is established nuclear physics; the gold bookkeeping is not.

Squeezing the Equation of State

The inspiral itself carried information that no telescope could have supplied. In the last orbits each star raises a tide on the other, and how much it deforms depends on how stiff its interior is. A squishier star bulges more, which drains orbital energy faster and makes the final cycles arrive slightly early. That effect is encoded in a quantity called the tidal deformability, and GW170817 put a firm upper limit on it.

Translated into geometry, the limit says a 1.4 M☉ neutron star cannot have a radius much above 13 km. The system had a total mass of 2.74 M☉ and a chirp mass — the particular combination the waveform is most sensitive to — of 1.188 M☉. Ruling out the stiffest equations of state in this way is the first direct laboratory constraint on how matter behaves at several times nuclear density, a regime that no accelerator can reach. Everything known about neutron star interiors before 2017 came from modelling their surfaces and their spin.

A Hubble Constant with No Ladder

Gravitational waves measure distance directly. The amplitude of a chirp depends on the luminosity distance and on the masses, and the frequency evolution pins the masses down, so the distance drops out of general relativity with no calibration step at all. Pair that with the redshift of NGC 4993 and you have a Hubble constant from a single event: 70 (+12/−8) km/s/Mpc.

The error bars are wide because one event is one event, and because orbital inclination is entangled with amplitude. But the method shares no systematics with the Cepheid ladder or the cosmic microwave background, which is exactly what makes it valuable to the argument over how fast the universe is expanding — the same virtue claimed by time delays in gravitational lensing. A few dozen sirens with counterparts would settle the question.

Still the Only One

Nine years on, GW170817 remains the only compact binary merger with a confirmed electromagnetic counterpart. Observing run O4 ran from May 2023 until 18 November 2025 and produced around 250 confirmed detections, almost all black hole pairs, and not one of them had a counterpart anybody could find. The next run, designated IR1, is scheduled to begin in November 2026.

The reasons are unglamorous. Neutron star mergers are intrinsically quieter than black hole mergers, so the volume they can be detected in is far smaller. Being detectable is not enough: the event must be close, well localised, above the horizon at night, and followed up within hours before the blue light is gone. GW170817 cleared all of those at once. That is luck, and the 1.74 second gap between its two signals also happens to be the tightest constraint we have on the speed of gravity against the speed of light, somewhere between −3 × 10⁻¹⁵ and +7 × 10⁻¹⁶ of a fractional difference.

The Value of Arriving Twice

Supernova 1987A hinted at what two channels can do: neutrinos arrived hours before the light, and the gap itself was the physics. GW170817 made that a method. The chirp gave masses, a distance and the stiffness of nuclear matter. The gamma rays timed gravity against light. The fading red glow showed the chart of the elements being extended in real time. None of those results could have come from any one of the three signals alone, which is the whole argument for building instruments that watch the same sky in different ways and then comparing notes within the hour.

Frequently Asked Questions

What happens when two neutron stars merge?

Two neutron stars in a tight binary lose orbital energy to gravitational waves and spiral inward, sweeping upward in frequency for the last minute or so before contact. GW170817 was tracked for about 100 seconds from 24 Hz up to several hundred hertz. At contact the two stars, each a little over 1.4 solar masses packed into roughly 12 kilometres, are torn apart by tidal forces and shock-heated. Around 0.05 solar masses of neutron-rich debris is flung out at between 10 and 30 percent of the speed of light, while the bulk collapses into a hypermassive remnant and almost certainly into a black hole within milliseconds. Jets launched from the remnant produce a short gamma-ray burst, and the expanding debris cloud, heated from within by the radioactive decay of freshly made heavy nuclei, glows for days to weeks as a kilonova. GW170817 produced all three signals and remains the only merger ever seen in both gravity and light.

What is a kilonova?

A kilonova is the optical and infrared glow of radioactive debris thrown out by a neutron star merger. The name comes from its brightness: around a thousand times brighter than a classical nova, but a hundred times fainter than a supernova. Its power source is unusual. Nothing is burning. The ejecta is packed with extremely neutron-rich nuclei built by rapid neutron capture during the merger, and those nuclei decay towards stability over hours and days, dumping heat into an expanding cloud. AT2017gfo, the kilonova of GW170817, peaked at about magnitude 17 roughly half a day after the merger and then reddened dramatically: the blue light faded by more than a magnitude a day while the near-infrared held up for a week, because the photosphere cooled from above 10,000 K to under 2,000 K as it expanded. That colour change from blue to deep red within days is the signature astronomers now look for.

Did GW170817 prove that neutron star mergers make gold?

It proved that neutron star mergers make heavy elements by rapid neutron capture, but it did not detect gold. The one secure identification in the spectra of AT2017gfo is strontium, reported by Watson and colleagues in 2019 from a triplet of absorption lines near one micron. Gold and platinum were searched for directly by Gillanders and colleagues in 2021 and were not found; that work set upper limits of roughly 10 to the minus two solar masses of gold and a few times 10 to the minus three of platinum, which are limits rather than measurements. Published estimates of the gold actually produced range from about 1 to about 200 Earth masses depending on the ejecta model, the mass fraction assumed for the heaviest nuclei, and which press release you read. The honest summary is that the r-process was confirmed, the specific gold yield is not yet pinned down to better than two orders of magnitude.

Why did the gamma-ray burst arrive 1.7 seconds after the gravitational waves?

The delay was 1.74 ± 0.05 seconds between the merger time inferred from GW170817 and the trigger of GRB 170817A in the Fermi Gamma-ray Burst Monitor. Almost all of that is astrophysics, not a difference in travel speed. The gravitational waves are emitted at the instant the stars touch, whereas the gamma rays need time for a jet to be launched, to push through the debris already ejected, and to accelerate to the point where it becomes transparent and radiates. Since both signals crossed about 40 megaparsecs, a 1.74 second offset over roughly 130 million years of flight constrains the fractional difference between the speed of gravity and the speed of light to lie between about −3 × 10⁻¹⁵ and +7 × 10⁻¹⁶. That single number eliminated whole families of modified gravity theories that had been proposed as alternatives to dark energy.

Read Next