Guest post: finding the most distant quasar

A couple of weeks ago, a few of my astrophysics colleagues here at Imperial found the most distant quasar yet discovered, the innocuous red spot in the centre of this image:
One of them, Daniel Mortlock, has offered to explain a bit more:

Surely there’s just no way that something which happened 13 billion years ago — and tens of billions of light years away — could ever be reported as “news”? And yet that’s just what happened last week when world-renowned media outlets like the BBC, Time and, er, Irish Weather Online reported the discovery of the highly luminous quasar ULAS J1120+0641 in the early Universe. (Here is a longer list of links to discussions of the quasar in the media: although at least this discovery was generally included under the science heading — the Hawai’i Herald Tribune some how reported it as “local news”, which shows the sort of broad outlook not normally associated with the most insular of the United States.) The incongruity of the timescales involved became particular clear to me when I, as one of the team of astronomers who made this discovery, fielded phonecalls from journalists who, on the one hand, seemed quite at home with the notion of the light we’ve seen from this quasar having made its leisurely way to us for most of the history of the Universe, and then on the other hand were quite relaxed about a 6pm deadline to file a story on something they hadn’t even heard of a few hours earlier. The idea that this story might go from nothing to being in print in less than a day also made a striking contrast with the rather protracted process by which we made this discovery.

The story of the discovery of ULAS J1120+0641 starts with the United Kingdom InfraRed Telescope (UKIRT), and a meeting of British astronomers a decade ago to decide how best to use it. The consensus was to perform the UKIRT Infrared Deep Sky Survey (UKIDSS), the largest ever survey of the night sky at infrared wavelengths (i.e., between 1 micron and 3 microns), in part to provide a companion to the highly successful Sloan Digital Sky Survey (SDSS) that had recently been made at the optical wavelengths visible to the human eye. Of particular interest was the fact that the SDSS had discovered quasars — the bright cores of galaxies in which gas falling onto a super-massive black hole heats up so much it outshines all the stars in the host galaxy — so distant that they are seen as they were when the Universe was just a billion years old. Even though quasars are much rarer than ordinary galaxies, they are so much brighter that detailed measurements can be made of them, and so they are very effective probes of the early Universe. However there was a limit to how far back SDSS could search as no light emitted earlier than 900 million years after the Big Bang reaches us at optical wavelengths due to a combination of absorption by hydrogen atoms present at those early times and the expansion of the Universe stretching the unabsorbed light to infrared wavelengths. This is where UKIRT comes in — whereas distant sources like ULAS J1120+0641 are invisible to optical surveys, they can be detected using infrared surveys like UKIDSS. So, starting in 2005, UKIDSS got underway, with the eventual aim of looking at about 10% of the sky that had already been mapped at shorter wavelengths by SDSS. Given the number of slightly less distant quasars SDSS had found, we expected UKIDSS to include two or three record-breaking quasars; however it would also catalogue tens of millions of other astronomical objects (stars in our Galaxy, along with other galaxies), so actually finding the target quasars was not going to be easy.

Our basic search methodology was to identify any source that was clearly detected by UKIDSS but completely absent in the SDSS catalogues. In an ideal world this would have immediately given us our record-breaking quasars, but instead we still had a list of ten thousand candidates, all of which had the desired properties. Sadly it wasn’t a bumper crop of quasars — rather it was a result of observational noise, and most of these objects were cool stars which are faint enough at optical wavelengths that, in some cases, the imperfect measurement process meant they weren’t detected by SDSS, and hence entered our candidate list. (A comparable number of such stars would also be measured as brighter in SDSS than they actually are; however it is only the objects which are scattered faintward that caused trouble for us.) A second observation on an optical telescope would suffice to reject any of these candidates, but taking ten thousand such measurements is completely impractical. Instead, we used Bayesian statistics to extract as much information from the original SDSS and UKDISS measurements as possible. By making models of the cool star and quasar populations, and knowing the precision of the SDSS and UKIDSS observations we could work out the probability that any candidate was in fact a target quasar. Taking this approach turned out to be far more effective than we’d hoped — almost all the apparently quasar-like candidates had probabilities of less than 0.01 (i.e., they were only a 1% chance to be a quasar) and so could be discarded from consideration without going near a telescope.

For the remaining 200-odd candidates we did make follow-up observations, on UKIRT, the Liverpool Telescope (LT) or the New Technology Telescope (NTT) and in fewer than ten cases were the initial SDSS and UKIDSS measurements verified. By this stage we were almost certain that we had a distant quasar, although in most cases it was sufficiently bright at optical wavelengths that we knew it wasn’t a record-breaker. However ULAS J1120+0641, identified in late 2010, remained defiantly black when looked at by the LT, and so for the first time in five years we really thought we might have struck gold. To be completely sure we used the Gemini North telescope to obtain a spectrum — essentially splitting up the light into different wavelengths, just as happens to sunlight when it passes through water droplets to form a rainbow. The observation was made on Saturday, November 2010 and we got the spectrum e-mailed to us the next day and confirmed that we’d finally got what we were looking for: the most distant quasar ever found.

We obtained more precise spectra covering a wider range of wavelengths using Gemini (again) and the Very Large Telescope, the results of which are shown here:
The black curve shows the spectrum of ULAS J1120+0641; the green curve shows the average spectrum of a number of more nearby quasars (but redshifted to longer wavelengths to account for the cosmological expansion). The two are almost identical, with the obvious exception of the cut-off at 1 micron of ULAS J1120+0641 which comes about due to the absorption by hydrogen in front of the quasar. We had all the data needed to make this plot by the end of January, but it still took another five months for the results to be published in the June 30 edition of Nature — rather longer than the 24-hour turn-around of the journalists who eventually reported on this work. But if we’d given up on the search after four years — or if the Science and Technology Funding Council had withdrawn funding for UKIRT, as seemed likely at one point — then we never would have made this discovery. It was a long time coming but for me — and hopefully for astronomy — it was worth the wait.