We saw it! A Diary of Discovery

The Swift UVOT team at MSSL find surprisingly bright UV emission from the first ever visible counterpart to a gravitational wave event. Here is their story… from Paul Kuin 


First Image: The Swift observations of the event started 15 hours after the event. The UV colours have been combined and changed to colours our eye sees better. The afterglow of the binary neutron star merger can be seen fading away next to its galaxy NGC 4993.  Credit: NASA/Swift

The Alert

On August 17th 2017, just after 1pm UK time, a brief message went out to the LIGO/Virgo follow-up team that a new Gravitational Wave event (GW) had just been detected. As part of that world-wide team the Mullard Space Science Laboratory sprang into action. Alice Breeveld, Sam Emery, and myself, Paul Kuin, received that message, and almost simultaneously another one from Fermi satellite that they had detected a short Gamma Ray Burst (GRB). This was the first time a GRB and a GW had been detected together. The GW signature looked like one from two neutron stars merging – with a false alarm probability of 1 in 10,000 years; this time it was real. The time delay between the Fermi and LIGO alerts was about 2 seconds, so there was a good possibility the LIGO and Fermi detections were connected and carefully planned follow-up observations were started by many telescopes around the world.

Initially, LIGO gave us a huge area on the sky to search for an event (LIGO itself being not very sensitive to the exact direction from which a gravitational wave comes from), covering nearly the whole sky. Although smaller, the Fermi sky search area was still fairly large. Only when the results from the VIRGO gravitational wave observatory in Italy were added the area to search was reduced to a 30.7 square degrees (for comparison: our Moon has a diameter of half a degree). From the GW, an estimate of the distance was also made of 40±8 Mpc (about 130 million light years).

So, we sprang into action. For the Swift satellite follow-up, Phil Evans from Leicester University and Jamie Kennea from Penn State University had prepared detailed plans over the last two years for following up on detections of gravitational waves. The initial search with Swift concentrated on the area on the sky identified by the GRB, first looking at the galaxies. Each time the search area changed the observing plan had to be re-generated by the operations team in Penn State, and uploaded to the satellite, so there is some delay there as well. Swifts’s strategy was to observe the area with the most massive galaxies first and to look for any new X-ray sources, because that is what we expected to see from a GW and GRB.


The first sign of the Aug. 17, 2017, neutron star merger was a brief burst of gamma-rays seen by NASA’s Fermi Gamma-ray Space Telescope (top). Shortly after, LIGO scientists reported detecting gravitational waves that arrived 1.7 seconds before the Fermi burst (middle). A short time later, scientists analyzing gamma-ray data from the European Space Agency’s INTEGRAL spacecraft also reported seeing the burst (bottom). Credit: NASA’s Goddard Space Flight Center, Caltech/MIT/LIGO Lab and ESA

The X-ray and UV data were processed automatically by the Swift processing centres at NASA-Goddard and in Leicester.  Just after midnight on August 18th, a new optical source was found by the Swope ground-based telescope at Las Campanas observatory in the search region close to a galaxy 40 Mpc away. The Swope source was quickly confirmed by other observatories.

At that time, Swift had started to observe the 30.7 sq. deg. area of high probability; this needed many individual observations, because each image covers only 10 percent of a square degree. We discussed the new Swope source in emails with the Swift follow-up team, and we decided  to observe the Swope source for 2000 seconds, prioritising this over about 14 other patched of sky that were planned be observed.  It took several hours before we got that data, and a few more before it was on the ground to process.  Around 15 hours after the initial discovery we saw that UVOT had captured bright UV and optical light, but that, if there were X-rays, they were too faint detect. We sent out a notice with preliminary UVOT magnitudes derived by Alice Breeveld. We were surprised because we had expected to see X-rays if the source was a short GRB.

Short GRBs always seem to have an X-ray afterglow, but we have rarely seen UV/optical emission. This was confusing. We wondered if this was really the optical counterpart of the short GRB. However, there was another anomaly: the gamma rays, if they were from the source located at 40 kpc (as indicated from the gravitational wave analysis) were much fainter than the short gamma-ray bursts observed before. So, there was some confusion at the time: was this really the optical counterpart of the gravitational wave event or just another transient that just happened coincidentally at the same time…

More data coming in: is this the one? or the two?

Throughout the morning of August 18th,  images were arriving from all the extra patches of sky that Swift observed in the target area, and Sam Emery diligently went through all of the most promising images, checking new source candidates reported by the NASA Goddard processing center. No good alternative was found.

Swift continued to re-observed the Swope candidate and, after some discussion about how to correct for the nearby galaxy, Sam Emery reported the UVOT magnitudes of the Swope source. Other people within the large electromagnetic follow-up collaboration reported magnitudes in the visible and infrared wavelength in LIGO/VIRGO communications.  Optical and IR spectra showed a featureless spectrum. Ryan Foley (UC Santa Cruz) used all the information so far to conclude that this was not a supernova, but a rare event.

Then, near midnight on August 18th, Daniele Malesani, Darach Watson, and Jens Hjorth (DARK/NBI) used the available reported magnitudes to derive a spectral energy distribution and fit that with a black body spectrum. Its brightness and the source distance could be used to infer the speed of the ejected matter, which was 0.2-0.3 times the speed of light (very fast!). This was consistent with the spectrum having no spectral lines – such high speeds would wash those out.

However a mystery remained. The ultraviolet data did not fit the black-body spectrum particularly well, and privately I suspected a combination of two blackbodies would be a much better fit, suggesting two components for the emission.

By the start of August 19th, when we were all getting excited about these discoveries, the results were still a bit strange. The Swope source had a very red spectrum, with no features consistent with the high speed, and looked different from any nova or supernova we’d ever seen, but the distances to the sources were a match. What we were observing was closer to what we would expect for the merger of two neutron stars: called a kilonova.

But still the data didn’t quite fit even the kilonova expectations. What differed from the expectation was the bright UV emission we had observed, 15 hours after the gravitational wave detection. In our subsequent observations the UV emission got fainter, while the red emission got stronger.

So by August 19th  we were increasingly convinced we had found the optical counter part to the gravitational wave event, that the short GRB somehow had to fit in data, but what we had observed must be a very strange event indeed…

Light in the darkness: understanding the event

During the following days Brad Cenko, the Swift Primary Investigator, who had been coordinating our efforts, got into touch with some of the people who had made the kilonova predictions with such amazingly fidelity for an event that had only ever been theorised.

Chris Frey and his group in Los Alamos quickly computed new models with different viewing angles. They explained that UV emission could not pass through matter saturated with newly created elements, Lanthanides and Actinides, heavier than Iron, but as long as the heaviest element produced by the neutron reactions was Iron, UV emission would be seen. This had actually been considered as a possibility by Brian Metzger at MIT a few years before, but had not been simulated in latest computer models. Because the atomic properties of the Lanthanides and Actinides are not well-known, it makes any modelling fairly complicated. On top of that, the models must also take into account nuclear reactions for building new elements out of neutrons, absorption of light (photons) by the newly created matter, dynamical effects due to tidal forces and heating, the gravitational waves themselves, and so on. However it looked like the event was a neutron-star-merger kilonova!

The smoking gun was that several weeks after the initial coalescence of neutron stars,  radio observations were predicted to be detected. It was just that no one knew for sure how long the wait was going to be.


Model SEDs for kilonovae compared to our data. The SEDs were created for four models with low neutron density, for different orientations and for two different epochs. The shaded regions display the range of model SEDs for different viewing angles. Thick solid lines show the spectrum for viewing angle θ ≈ 30◦, and thin lines represent gradually larger viewing angles going up to ≈ 60◦. Data points are shown as circles with error bars, or triangles for upper limits. The models span a range of wind masses: from a low wind mass of 0.005 M⊙ to 0.015 M⊙, 0.03 M⊙ and 0.1 M⊙. Clearly, to fit the observed data points, the wind mass needs to be at least 0.03 M⊙

Checking the results

For the measurements from the early UVOT data we had to make a correction for the extended emission from the nearby galaxy.  We kept collecting data after the (no strongly suspected) kilonova emission had disappeared and thus obtained a good measurement of the nearby galaxy. This allowed us to make a better correction to our observations of the kilonova, and we passed those to the people in Los Alamos to help in their modelling.

The new models from our Los Alamos collaborators were for a range of viewing angles and mass for the ejected matter, while keeping the electron content high, meaning a smaller amount of Lanthanides produced. The figure above shows the result.

Our companion paper which was led by Mansi Kasliwal focussed on the problem of the GRB jet and overall emission spectrum. Their models showed clearly that if the jet interacts with the ejected matter in a particular way, a cocoon is formed around the core of the jet. The result of the simulations is shown in the figure below.


Snapshots from a hydrodynamic simulation of a cocoon generated by a choked jet leading to  emission consistent with our observations. On the left the brightest parts of the image have the highest energy density where there is turbulence. On the right the brightness shows the places where most movement occurs. For the best fit we would be observing at an angle of 40 degrees, the ejecta mass would be 0.1 solar mass and the jet luminosity is 2.6 1051 erg/s. The jet would have been fully choked by 4sand the cocoon would have broken out around 10 s. 

So the result was confirmed, the first observed merging neutron star pair – resulting in a kilonova! Now we had to tell the world…

Collaborations and Top Secret Documents

Everyone involved in the electromagnetic (EM) follow-up of the LIGO/Virgo event had signed an agreement of confidentiality, so before we could use data from other groups in our analysis our leaders had to make collaborations. Brad Cenko set up a collaboration with the NUStar team and with Mansi Kasliwal at Caltech who was leading the GROWTH collaboration that included the Swope telescope which discovered the optical source and we shared our data for analysis. After about three weeks we had drafts of our publications ready and Science had agreed to publish – though of course our paper had to pass its standards and peer review.

It was difficult to keep all this secret, because as scientists we are not at all used to secrecy. We like to discuss ideas with friends and hear other opinions from colleagues. And, of course, some hints that there was a discovery leaked out. For example, some observatories listed the title of the observing programs, and even the target galaxy.  Mistakes were made putting information on the web, only to be hidden again – apparently without being discovered outside the EM collaboration. It was just a weird experience! Fortunately there was also an exciting Black Hole merger, and a Nobel Prize, which seemed to distract the news hunters for a while.

In Summary: Astronomy has changed

  • Before the discovery we knew of the existence double neutron stars, and also that some of these systems will lose enough orbital momentum through gravitational waves to eventually merge. Now, for the first time we had actually observed them merging and not just with photons, but with measurements of the distortion of space-time.
  • We also suspected that short GRBs might be caused by the merger of two neutron stars, and now we have a confirmation. This is an amazing triumph of theory predicting observation.
  • The observed faintness of the short GRB and the upper limit of X-ray emission point to the merger being seen off-axis i.e. the powerful jet was not pointed towards Earth.
  • The models predicting that ejected matter would create the heavy elements in the Lanthanides and Actinides, and this is consistent with the overall redness of the spectrum, especially at later times after the event, and so we now know that binary neutron star mergers are of one of the largest, or the largest, source of such elements (like gold).
  • And of course we have now seen gravitational waves from both Black Hole-Black Hole and neutron star-neutron star mergers.
  • The surprise was the bright UV emission. Maybe we were just lucky. Next time the viewing angle may be in the plane of the dense neutron ejecta with the Lanthanides blocking all the emission. It is likely that the emission at late time became dominated by the neutron-rich ejecta which due to the high opacity of the Lanthanides would only slowly release the generated heat.
  • The Fermi team reports in their analysis a weak, lower energy, second peak about a second after the first peak. This is intriguing and it will be interesting to see if that can be explained with the cocoon model.

I end with a link to a video from NASA. It is an animation of the binary neutron star merger, based on what we think probably happened 130 million years ago. Our models will continue to improve and confirmed by watching and studying more of these amazing events. But the first will always remain a special kilonova.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s