Matthew Young is a postgraduate student with the University of Western Australia and the Australia Telescope National Facility.  This is his very personal account of the circumstances that led to his discovery of a pulsar with a period so long that, according to accepted theories, it should not have been able to emit radio pulses - making it undetectable.  Matthew's story is an interesting counter to those who claim that scientists have a vested interest in their theories and unreasonably resist claims that fall 'outside the square'.  The excitement of the unexpected discovery, the careful thorough work which made it possible, and the shock of handling the resulting publicity (and Matthew's indifferent health) make this a fascinating account, summed up by Matthew's quote, attributed to Pascal, "In observational science, chance favours those with open eyes."

If you have any questions for Matthew, post them to the Astronomy News newsgroup, and I'll summarise them and forward them to Matthew for comments.
 

How do pulsars emit the radio waves that we observe?

This question has puzzled astrophysicists since the discovery of pulsars.  Just about all the physical conditions that go to make up pulsars are extreme.  As such they provide a wonderful opportunity for physicists to test their theories under conditions not currently achievable on Earth. Many theories of the radio emission mechanism have been proposed, but to date there is no consensus on which, if any, is correct.  Nevertheless, certain common elements have become popularly accepted, especially the idea that electron-positron pair-production is required for radio emission.  At least, this is how things stood at the beginning of October 1998.  Little did I realize then, that I was about to play a major part in challenging this long-held notion.

My name is Matthew Young and I am a PhD student in the Physics Department at the University of Western Australia where my supervisor is Dr Ron Burman.  I am also affiliated with the Australia Telescope National Facility (ATNF) where my supervisor is Dr Dick Manchester. Part of my research project has been the study of the variability of single pulses from pulsars.  In mid-1998 it had been agreed that I would re-examine data collected by Dr Simon Johnston as part of his study of pulsar scintillation, this time looking at single pulses from pulsars which had not previously been examined in that way.

In mid-October 1998 I arrived in Sydney and collected the data tapes from Simon.  They contain a vast amount of information from many pulsars.  Since interesting pulse-to-pulse behaviour is normally seen in the longer period pulsars, I went through the list of pulsars on the tapes and found the one with the longest period for which we had not undertaken a single pulse study. This was PSR J2144-3933 and the pulsar catalogue gave its period, P,  as 2.84 s.

This pulsar was discovered in the highly successful Parkes Southern Pulsar Survey (Manchester, Lyne et al.) earlier in the decade. The survey discovered 101 new pulsars and involved a huge amount of data processing--then equivalent to 50 years of workstation-computer CPU time.  Because of this huge computational task, many of the processes were automated, including the initial pulsar period determination. A pulsar's period, or rather its rotation frequency, was determined from the Fourier power spectral density (power spectrum). The survey data included a large component of low-frequency (long-period) noise. Because of this, frequencies lower than a cut-off, corresponding to a period greater than 4 or 5 seconds I think, were not searched for.  PSR J2144-3933 was discovered via its 3rd harmonic corresponding to a period of 2.84-s.  This wasn't realized at the time, perhaps in part because people weren't expecting to find a pulsar with an 8.5 second period (for reasons I will go into below).

At this stage, I did not realize that the catalogued value of a pulsar's period could be incorrect.  So I analyzed this pulsar as if it had the catalogued period of P = 8.51/3 = 2.84 seconds. This meant that only every third pulse I looked at was a real pulse, and the other two were just noise, but which I treated as pulses and which I will call, for now, pseudo-pulses.

In the analysis I did not look at every single pulse of which there were almost 1000.  Instead I looked at some of them, and then computed quantities for each pulse which would tell me about the others. The first thing that became apparent was that the pulse peak flux density was very variable and that most of the pulses had a peak radio flux density indistinguishable from noise.

This was not that surprising since the pulsar does not have a large average flux density.  In general individual pulses from a pulsar have variable peak flux density, in part due to scintillation because of the interstellar medium, and in part due to processes at the pulsar. If the signal to noise ratio in the data is not large, weaker pulses become indistinguishable from the noise and I took that to be the case in our data.

The next thing I noticed from the analysis was somewhat surprising: one strong pulse was often followed by another three pulses later. Now in retrospect this is obviously because the value for the rotation period that I was using was wrong by a factor of three: two out of three "pulses" were indeed only noise because of that.  However, at the time, I didn't know the period could be wrong, and this sort of modulation of pulse intensity is not entirely unknown - there are pulsars which exhibit such a modulation - though not in this extreme way.  It was also masked by the fact that many real pulses were indistinguishable from noise.  Furthermore, there was radio frequency interference in the data resulting in quasi-periodic variation of the noise.  Hence, some pseudo-pulses looked like real pulses.

Since I took the period of 2.84 s as an absolute given, I pursued the idea that there may be a pulse modulation going on, but I needed to rule out radio frequency interference as the actual source of this apparent modulation.  I also needed to consider the possibility that the pulsar may be nulling.  Nulling is the complete absence of detectable radio emission from a pulsar for a duration of one or more pulses due to a change in the state of the pulsar itself.  One simple explanation of nulling is that the pulsar temporarily ceases to emit radio waves.  If this pulsar nulled, I needed to be able to distinguish this, if possible, from the low-energy (pseudo-)pulses in the modulation. Nulling could also offer the possibility of introducing phase shifts into the periodic pulse modulation, thus helping to explain the assumed 2.84-s pulse period.

Over the next week I concentrated mainly on the broader question of nulling, investigating new schemes to distinguish nulls from pulses.  During this time I had gone to the Compact Array radio telescope to act as the on-call duty astronomer for a week.  This and the difficulties of doing some of the analysis at Narrabri where the computers are not set up for pulsar research of this sort, meant that progress was slowed during that time.

The morning after my arrival back in Sydney I arranged to have a meeting with Dick and Simon at 1:30 pm to discuss what I had found. During the morning I went back to looking at the pulse modulation and computed the time between all consecutive pulses stronger than a certain threshold.  I found that in nearly every case the interval was indeed a multiple of 3 P.  However, in a few cases there were pairs of such pulses with an interval of less than three pulses between them. I studied these and concluded that one of each pair was not clearly distinguishable from noise.  This opened up the surprising possibility that strong pulses only *ever* occurred a multiple of 3 P apart with no null-induced phase shifts in the modulation. So then it became clear that the period of the pulsar might indeed be 3 P = 8.51 s. However, I still wasn't certain.  Just because the weak pulses between the strong were not distinguishable from noise when viewed individually, when you add several hundred of them together, if they are there they should become evident.  By now it was lunch time, but I was becoming too excited and engrossed by what was unfolding before me to want lunch.

To test the idea of a 3 P period, I computed the average pulse profile using the 8.51-s period.  If the period were 2.84 s after all, this average pulse profile would show a triple pulse, each pulse separated by 2.84 s, and one of the pulses much stronger than the other two (to account for the 3 P modulation). The resulting plot showed a single narrow pulse, consistent with the new period. I started to become very excited.  I double checked all my reasoning and computations and then confirmed that the previous longest catalogued radio pulsar period was 5.09 s.  Then I was sure -- I had discovered by far the longest radio pulsar period.  I felt wonder and shock and joy at finding something so unexpected. I just had time to print out my results and then head down the corridor to my meeting with Dick and Simon.

I think it is fair to say they were both very surprised at my finding, but agreed that my conclusion was correct.  We then discussed the implications of this finding and some quick calculations showed that the pulsar was over the so-called death-line, making the finding even more significant and suggesting that a paper in Nature might result.

The death-line is a line on the plot of surface magnetic dipole field strength (B_s) vs period (P) which separates the parameter values for which electron-positron pair-production (and hence pulsar radio emission) is thought possible, from those for which it is thought impossible. Quoting from the resulting Nature paper (Young, Manchester & Johnston 1999, Nature, 400:848):

"Radio pulsars are rotating neutron stars which emit beams of radio waves from regions above their magnetic poles. As pulsars dissipate rotational energy, their rotational periods increase.   Popular theories of the emission mechanism require continuous electron-positron pair production above the magnetic poles. The accelerating potential responsible for this is inversely related to period.  Since pair-production stops when this potential drops below a threshold, the models predict that radio emission ceases when the period (P) exceeds a value dependent upon the surface magnetic field strength (B_s) and configuration. "

A death-line is the set of P-B_s values at which radio emission should cease for a given magnetic field configuration (i.e. curvature). The greater the magnetic field curvature, the longer the rotation period at which pair-production can still occur. B_s is estimated from P and dP/dt and depends on the assumed neutron star equation-of-state. Making the usual equation-of-state assumptions, and assuming the most curved magnetic field thought likely, we found that this slowly rotating pulsar should not be emitting a radio beam. Hence either the assumptions about the equation-of-state are incorrect, or current theories of radio emission must be revised. PSR J2144-3933 is a benchmark against which physicists and astrophysicists can test their theories.

Those couple of hours on 29 Oct 1998 when all this unfolded were very exciting ones in my life, though it was the excitement without the precursor of long anticipation.  In the following days I re-analyzed other data from this pulsar to get a more accurate measurement of its period and rate of change of period (dP/dt). Then in the following weeks I read around in the literature on related topics and did some other calculations to hopefully make sure there was nothing else I had over-looked. By the time I was starting to write the paper, I unfortunately had a relapse of old health problems, and it took a few months to get the paper finished and submitted to Nature.

The paper was published in Nature on Thursday 26 Aug 1999. Coincidently, I got married the previous Saturday and departed for a pulsar conference in Germany on Wed 25 Aug. The weeks and days leading up to my departure turned out to be quite frenetic and somewhat stressful at times.  Prior to the publication of the Nature paper, both Nature and the ATNF had made press-releases about it.  Press coverage was embargoed by Nature until the night of Wed 25 Aug. I was allowed to talk to the media, but I had to make it clear that the story could not be published before the end of the embargo.  Since I was leaving the country on Wednesday, most of the media contacted me on Tuesday, though I had been contacted by New Scientist and others at the end of the previous week.

Talking with the media was exciting.  I did not feel particularly nervous once things got under way.  It was rewarding to be able to share my excitement with others and to explain what we did and why we thought it was important.  My main concern was trying to fit in all the interviews before I left the country.  In fact, some of these carried on once I got to Singapore.

I arrived in Germany late Thursday night.  In the following days I started to receive email from people around Australia about the news stories they had heard.  Some were from long lost friends, others complete strangers. Several were asking for more information.  I tried to help out where I could, but email contact was becoming harder and harder as I travelled round. I presented a talk on the pulsar at the conference and this seemed well received. Afterwards a fellow astronomer told me that she had been studying this pulsar, but had not noticed the incorrect period.

Now I am back in Perth and settling back into a routine.  It has been a strange experience to do something which results in some international press coverage.  I have at times wondered about my place in it all. Recently the director of the ATNF came to Perth to give a public lecture.  As part of this talk he briefly mentioned our discovery of the longest pulsar period in the context of a theme he was developing based upon a quote of Pascal. The quote I think was: "In observational science, chance favours those with open eyes." Now I see more clearly.

Matthew Young

Back to the Guest Contributions Index