Paul Francis, Mt Stromlo Observatory, the Australian National University.

This is the story of pink black holes, and how they were discovered. As usual in science, it is a story of politics, confusion, errors and the occasional miraculous and wholly unexpected discovery.

The story begins several years before I was born. The first quasars were discovered early in the 1960s, using radio telescopes. Mysterious intensely powerful radio sources were found scattered all over the sky, and after years of effort, the radio astronomers were able to pin down their positions accurately enough for optical astronomers to take pictures of them and measure their spectra. As is now well known, these radio sources looked just like stars – hence the name "Quasi Stellar Radio Source" or quasar. On closer examination, however, they turned out to be monstrous black holes lurking in the middle of fantastically distant galaxies. In some bizarre way, these black holes were devouring the gas and stars around them, and spitting some of the debris out at nearly the speed of light in two powerful jets. The intense radio emission came from subatomic charged particles spiralling around magnetic field in these jets.

Measuring the optical position of quasars accurately enough was difficult, and soon the (in)famous American cosmologist Alan Sandage figured out an easier way to find quasars. He noticed that most quasars looked extremely blue when observed with an optical telescope. He then searched for other extremely blue dot-like things in wide-angle photographs of the sky. He would take an image through a UV filter, and through a blue filter. Objects which were abnormally bright in the UV compared to the blue image were quasar candidates. About half such objects turn out (when their spectrum is taken) to be quasars, while the rest are white dwarf stars. In one wide angle photographic plate you might find 300-400 quasars.

His discovery triggered a whole new industry: teams of astronomers combed the sky, looking for objects with these incredibly blue colours, and taking optical spectra of the ones they found to prove that they were quasars. Tens, then hundreds of quasars were found. As I write, over 20,000 quasars have been found, almost all of them by this technique. Literally hundreds of astronomers around the world work at finding and studying these quasars, and several thousand scientific papers have been published about them.

Why should black holes be incredibly blue? A lonely isolated black hole, all by itself in the depths of space would indeed look black – its intense gravity would swallow all light falling on it. We would never see a black hole like this: there could be any number of them out there and we’d never know. We can only see a black hole when it gets close to something, such as a passing star or gas cloud. Its intense gravity reaches out and snaffles the passing material, ripping it to shreds before stuffing it down the throat of the hole. Understandably, this ripping and gobbling process is extremely violent, and the material tends to get very hot in the process: typically 30,000 degrees or more. Anything at this temperature will emit intense blue-white light, just as you see from arc welding. That’s why quasars are thought to be blue: you are not seeing the hole itself, but the intense blue light from tortured matter about to fall in.

In a few unfashionable corners of the astronomical world, however, people persevered with the original way of finding quasars: the radio way. The survey that found the pink black holes started when I was two years old. It was carried out at a brand new telescope on the far side of the world to my London home: at Parkes, in the Western Plains of New South Wales, Australia.

The revolutionary Parkes radio telescope, designed in part by Barnes Wallace (of Dam Busters fame) had just been completed, and was commencing an incredible project: a complete two-year radio survey of the southern sky. It discovered thousands of previously unknown radio sources, hundreds of them with the particular characteristic radio spectrum of quasars (a so-called "flat" spectrum – one with lots of high frequency radio power and not so much at low frequencies).

The problem with using radio telescopes to find quasars isn’t with finding quasar candidates: even 30 year old surveys would find thousands of possible quasars. No – the problem is the optical follow up. Unless you can get an optical spectrum of each object, you cannot tell for sure that they are quasars, and you cannot measure their redshift (and hence the distance to them). Getting the spectrum of a quasar candidate is easy if it is optically bright: on a 2 metre telescope with modern CCD spectrographs you can get a good spectrum of a 16^th magnitude quasar in about ten minutes.

Unfortunately, most of the radio-selected quasar candidates were very faint. Some were fainter than 25^th magnitude. This means that you need vast amounts of observing time on really large telescopes to get their spectra.

About 10% of the probable quasars from the Parkes survey were relatively bright in the optical: between 12^th and 16^th magnitudes. These were rapidly observed, and found to be mostly quasars: normal blue quasars. But what of the other 90%? These were much fainter: from 16^th magnitude to fainter than 25^th magnitude. Getting spectra of these was very slow and painful work. In the mid 1970s however (while I was finishing junior school) the Anglo Australian 3.9m optical telescope (three hours drive north of Parkes) was completed, and for the first time Australian astronomers had a telescope capable of taking spectra of the brighter of these hundreds of radio sources.

A few brave astronomers (particularly Anne Savage) persevered with the identifications into the mid 1980s, but they only managed to identify about half the sources (those brighter than about 19^th magnitude). By 1985 (when I was starting my undergraduate degree in Physics at Cambridge), the project had ground to a halt. The committee that allocates observing time on the Anglo-Australian Telescope couldn’t see why anyone should bother with identifying radio quasars: surely Alan Sandage’s technique (looking for blue dots) was so much faster and easier.

The survey languished until the mid 1990’s, when Dr Rachel Webster, newly arrived at the University of Melbourne, had to dream up an idea for a grant application. In the commercial world of modern Australian universities, an academic’s career goes nowhere unless he or she can bring in lots of money. Rachel desperately tried to dream up a research project that she could apply for money to pursue. She hit upon the idea of reviving the long-dead Parkes quasar survey, and trying to finish it off. She wasn’t primarily interested in the quasars themselves: she was mostly interested in how galaxies in front of them would bend their light (gravitational lensing). But before she could look at this bending, she had to identify all the quasars and measure all their redshifts.

That’s where I came in. Rachel succeeded in getting money for this project, and used it to hire me. I moved from Arizona (where I was then working) to Melbourne, and started trying to figure out how to identify the faint Parkes quasars.

We were particularly worried about the faintest 10%. These radio sources were so faint at optical wavelengths that we couldn’t see them at all on photographic plates, so we didn’t even know where to point our spectrographs. And yet, these invisible sources were amongst the most powerful radio sources in the sky.

In desperation, we took pictures of the regions on the sky from which the radio waves were coming using infra-red cameras. We were dumbfounded when we realised that these optically invisible sources were almost always incredibly bright in the infra-red! One source, for example, was fainter than 23^rd magnitude in the optical, but a stonking 15^th magnitude in the near-IR! As we obtained more and more data, the pattern became clear: the difficult Parkes sources were only difficult in the optical: in the infra-red they were all quite bright.

We published this discovery, which was very controversial. After all, everybody knew that quasars were blue – ie. that their radiation was stronger in the blue and UV than it was in the red and IR. That was how people found quasars, after all. What we were saying was that many quasars are actually very red. These quasars would have been missed by all conventional surveys: nobody would have seen them before! 80% of all quasars in the universe could have been missed. In effect, we were telling the hundreds of people who’d been searching for quasars for decades that they’d been wasting their time: they’d missed most quasars.

We were bitterly attacked by some astronomers, and highly praised by others. Once, when I presented this result at a seminar in Cambridge, one eminent astronomer in the audience got up and started attacking this theory. A different astronomer then jumped up and started arguing with the first. Several more astronomers joined the fray, and before long the whole room was in anarchy, and everyone was ignoring me. The only person who remained quiet was the Astronomer Royal, Sir Martin Rees, who when we first told him about the red quasars just said "yes, I always knew that"!

As the data became clearer, it became more and more certain that there were vast numbers of quasars with red colours: just within the last few weeks, the newly launched Chandra X-ray satellite has been finding lots of them.

The question now is: why are they red? The blue light was thought to come from hot gas swirling down the throat. Hot gas never emits red light. Cold gas can, but cold gas is so faint that we’d never see it. So what are these things?

We knew that these quasars were very bright in the near IR (at 2.1m m wavelength) and very faint when seen through a blue optical filter (the human eye perceives as blue light of wavelength around 0.4m m). But what about in-between? How bright were they at wavelengths of (say) 0.7m m or 1.3m m? I hoped that this might give us a clue as to what these "red quasars’ were.

I decided to try and measure the brightness of a few of these quasars at every possible wavelength. The procedure was to take pictures of each quasar through seven different filters: a blue one (0.4m m, B-band), a green one (0.55m m V-band), a red one (0.7m m R-band), and four infra-red filters: 0.85m m (I-band), 1.3m m (J-band), 1.6m m (H-band) and 2.1 m m (K-band). The observations were carried out using two different telescopes at the Australian National University’s Siding Spring Observatory.

This type of observation is one of the hardest to make. You need to measure precisely how bright each quasar is as seen through each filter. To do this you have to allow for light losses in the atmosphere, in the telescope and in the detector. You do this by observing stars of known brightness frequently throughout the night, and measuring the ratio of the signals detected. This only works, however, if there is absolutely no cloud. Not even a wisp of cirrus – any cloud at all loses you some unknown fraction of the light, completely trashing your calibration. The trouble with this is that only 40% of nights are clear enough for this at Siding Spring Observatory.

To make matters worse, the optical and infra-red images had to be taken with different instruments, and changing instruments takes so long that it has to be done by day. So the BVRI images had to be taken on one night, while the JHK observations had to be taken a few nights later when the infra-red camera was installed. And both had to be taken during perfect conditions, or all the data is useless.

Worse still, these quasars change in brightness, on timescales of weeks. If you failed to get half the data on some quasar, you cannot just come back next month: the quasar will have changed. All the images on any given quasar have to be taken within a week at most, and preferably a couple of nights.

I thought that I had a snowball’s hope in hell of getting much useful data. I’ve had a great deal of experience with weather at Siding Spring (character building, is how one astronomer puts it). What, I thought are the odds of both the infra-red and the optical halves on an observing run being absolutely perfectly clear? Nil. So I asked for three times more nights than I needed, to make sure that there was a decent chance of getting some data.

I was awarded 26 nights of telescope time, spread over three different observing runs in 1997. 24 of those nights were perfectly clear –a staggering record. I took over 2000 images, getting complete data on 153 quasars. The nights were beautiful but I never got out of the control room to see – I was so busy sitting at the computer taking and analysing data, or running into the dome at ten minute intervals to change filters or to move to the next target. I was observing solo – a very lonely activity, so I turned the stereo in the control room up loud, alternating Beethoven, Bach, the Beetles and the Beach Boys. Two of the observing runs were gruelling winter runs, when you start work at 3pm (calibrating the telescope) and don’t get to bed until 7am.

The data took the best part of a year to analyse. The results were remarkable: if you plot the brightness of each quasar as a function of wavelength, you get a straight line. This result demolished all the rival theories. Our best guess at the moment is that the radiation is a curious mixture of light from the gas swirling down the throat, with light from the jet of particles that is emitting the radio waves. Perhaps the same spiralling particles that emit the radio waves have just enough energy to emit some red light as well, but not any blue light. We have no idea why such particles should exist. So this story ends badly: we’ve measured the colours of these weird quasars with exquisite precision, but we have no idea what cause these colours: all our models either fail to match the data or seem physically implausible.

The final surprise came when the director of Mt Stromlo Observatory nominated me for a science communication prize, and I had to present a talk on these quasars for the general public and the media. Until this point I had always called these quasars "red", because they have much more long-wavelength emission and much less short-wavelength emission than anything else in the night sky. But I’d never actually worked out exactly how they would look to the human eye. I decided to do so for my talk.

I had images taken through red, blue and green filters of all the funny quasars. The question was how to combine them such that the resultant image would match what the human eye would see, through a suitably powerful telescope. I talked to David Malin (the acknowledged expert) about this, and he told me that he uses Alpha Centauri as a standard. His reasoning is as follows:

When we look at a white wall in the middle of the day, it appears white. The light coming from this wall is reflected sunlight. The sun is a G-type star. So is Alpha Centauri. So, let’s take Alpha Centauri as our definition of white. Combine our images such that an object with the colours of Alpha Centauri will have an equal red, green and blue intensity – ie. will appear white.

When I did this to my images, I was astonished to find that the quasars actually looked a horrible shade of nauseous pinkish-purple. This is how they would appear to the human eye, if we were close enough.

So, many quasars are pink. We don’t know why. But we are still trying to find out!