Enigmatic bursts of γ-rays from cosmic sources are detected about once a day from every direction on the sky. During their brief duration, typically between 10-2 and 103 seconds1, these bursts often have the brightest γ-ray intensities ever seen. Now, after almost three decades of observation and speculation, the distance scale to γ-ray burst (GRB) sources seems at last to have been settled. Evidence has been growing at a remarkable rate in the past few months — the most striking pieces appearing on pages and of this issue2,3 — to support the proposal that most, and perhaps all, of the bursts come from explosions involving neutron stars or black holes near the edge of our observable Universe. If this is indeed the case, GRB sources are among the most powerful events in the Universe. This was the view of most of the participants in a GRB workshop last month.

In 1991, the Burst and Transient Source Experiment (BATSE), on board the Compton Gamma-Ray Observatory, began to accumulate what has now grown to almost 2,000 burst observations. These show us to be at the centre of a nearly homogeneous distribution of bursting sources, and the paucity of faint bursts implies that we are seeing almost to the edge of the source population4. These two features are natural consequences of any uniform source distribution in the simplest cosmological models for our expanding Universe5. Other suggestions, for example that GRBs are from relatively small, recurrent explosions on the surface of neutron stars now in the halo of our own Galaxy6, needed some ad hoc tuning of parameters such as delays in the onset of explosions. This made such models less appealing, even though they more easily explained certain controversial features of some bursts (such as cyclotron resonance spectral features, periodicities of several seconds, and possible burst source repetitions).

A ‘smoking gun’ was needed, to identify some GRB source with a familiar type of object which could determine the distance-scale. The problem was that γ-ray telescopes such as BATSE have an angular resolution of several degrees, and so cannot pin down a burst location finely enough for other kinds of telescopes to identify a unique counterpart at other wavelengths. The Interplanetary Network of satellites, started in the 1970s, could triangulate the angular position of a burst from arrival-time differences at different satellites, but gave only a handful of arc-minute-sized error boxes, generally months after a burst — far too long an interval to allow the identification of other transient emissions associated with it.

Since its launch in 1996, the Italian-Dutch satellite BeppoSAX has improved on both position and time constraints enormously: it can locate X-ray emission from a GRB to a few arc minutes, and then tell optical and radio telescopes where to look within a few hours. So far, it has identified three fading X-ray sources as probable GRB afterglows. The 28 February 1997 burst7 was followed by a fading optical afterglow from a source very near the centre of an apparently distant galaxy8, a result so exciting that soon afterwards the workshop was organized by F. Pacini (Arcetri Astrophys. Obs., Florence) and colleagues to review the new multi-wavelength evidence and its implications for GRB models. And, as if to reward the participants, nature provided the most exciting burst so far on 8 May, barely two weeks before the workshop, but time enough to identify an afterglow in X-ray, optical and, for the first time, radio bands.

The most important new evidence presented at this meeting was of redshifted absorption-line features in the optical spectrum3, at wavelengths corresponding to absorption between the burst source and us by iron and magnesium, at a distance of around five billion parsecs. Then this burst came from an explosive emission of about 1051 erg of γ-rays (approximately the total radiation from our Sun during the 1010-year age of the Universe) near the edge of the Universe. There is still a small possibility that the fading optical source is not associated with the γ-ray burst, but for most of us it has finally ended the distance-scale debate of the past three decades.

The huge GRB-source luminosities imply a very rapid conversion of energy, equivalent to 0.1 per cent of the total rest-mass energy of a solar-mass star, into γ-rays and afterglow. Observed time structures of γ-ray intensities indicate that, initially, the sources are typically not more than 100 km across (ten times the radius of a neutron star). The merger of two collapsed stars in a binary (for example, two neutron stars or a neutron star and a black hole) is the most popular proposed model for a cosmological burst source.

The enormous explosive local energy release would then initiate a relativistically expanding fireball whose properties and evolution are not sensitive to details of the explosion that caused it. Almost all of the initial fireball energy is eventually converted into a blast whose energy is carried mainly by its relativistically expanding nucleon flow. It is the later interaction of the relativistic blast wave with the ambient medium around the exploded source that is thought to be the origin of the observed X-ray, optical and radio afterglows. Progressively lower-energy spectra are emitted as the blast is slowed by this interaction. There was a consensus among modellers at the workshop that, to account for what is observed in the optical and X-ray afterglows, the Lorentz factor of the blast before such slowdown must be between 100 and 1,000 (that is, the blast's energy was 100 to 1,000 times its nucleon rest-mass energy). To achieve blast-wave Lorentz factors exceeding 100, no more than 10-5 of a solar mass of nucleons could have been included in an initial 1051-erg fireball. The observed afterglows would be expected when the blasts expand to a radius exceeding 1016 cm, the precise value depending on the assumed ambient density.

Radio emission (with a peak flux at 8.46 GHz), first detected five days after the 8 May burst9 and flaring to twice its initially detected magnitude in one day (S. Kulkarni, Caltech), gives a special clue for just such a blast wave: a strong, incoherent radio flux implies an emitting region at least 1016 cm across to avoid the intensity limitation from re-absorption of the radio waves by the electrons that emitted them.

Some were still unpersuaded about the cosmological distances. Particularly controversial10 was a report that the optical afterglow associated with the 28 February GRB moved by about 1.5 milliseconds of arc per day11 — very hard for cosmological models to explain (T. Bulik, Copernicus Center). Unfortunately, this source is now behind the Sun and we must wait until August to resolve this question. (No proper motion was observed for the very precise radio-location of the 8 May GRB radio afterglow.) And there are claims that the nearby nebulosity — the supposed galaxy host to the burst — may have changed in either size (D. Lamb, Univ. Chicago) or brightness.

If the relativistic-fireball model is established, GRB afterglows will become a tool to explore properties of the ambient gas in the early Universe. Perhaps, when enough becomes known about optical-afterglow absorption line redshifts and GRB intensities, these events may also yield the necessary relationship between redshift and distance to establish crucial cosmological parameters such as the density of the Universe, or even the cosmological constant (T. Piran, The Hebrew Univ.).

More BeppoSAX designated positions will come in roughly once a month, and we expect to see afterglows from many of them. Afterglows may also be identified in some cases where the GRB is not, perhaps because the γ-rays are mainly beamed away from us. It is hard to point to any other such important astrophysical problem that should be so quickly solved as the origin of γ-ray bursts. But that will probably only be the first extraordinary chapter of a book that promises to become even more exciting.

Workshop on the Latest Developments in Gamma-Ray Bursts,Elba, Italy, 26-27 May 1997.