Interview with Bangalore Sathyaprakash

0:00 So it's the 25th of January, 2000. It's a quarter past noon, and I'm speaking with Professor Satya Prakash. So, Satya, I suppose a good place to start is how you first got interested in the problem of gravitational waves, and what your background was to begin with. Right. My own PhD was actually in generativity and cosmology and sort of symmetry breaking in the early universe of scalar fields coupled non-linear to gravity. So I basically did a PhD on Singularity-Free Cosmology. So I graduated Hattini Institute of Science, and after that I went to Ayuka, and I was hoping to work on black holes because that was the area of interest to me. But it turned out that where I did my PhD, people were not interested in working in this area. In fact, it was even difficult to find a supervisor who would supervise me in the area of general nativity. Some condensed matter physicists agreed to be the supervisor, but it really was not of any help. so I just read articles and really grouping in the dark and did what I thought was the best at that time so when I went to Ayuka for my post-doc that was the time I had some seniors who had much more experience in the field so I started working on black hole physics and then at that time, even though theoretically I had read about gravitational waves earlier, I came to know only then that there were efforts being made to build these big detectors, that was in 1989 build these interferometry detectors, prototypes existed already and there were efforts in America and in Europe to build these big detectors and thought that one way of understanding black holes is perhaps observationally and then had just written from a post-doc with Bernard Schultz, not a post-doc sorry, I think it was a visiting

2:30 scientist PPARC, I think it used to be called EPSC, no what is it, SERC, it used to be called SERC at that time, so he had visited Bernard Schultz for about a year he had himself gotten into gravitational wave data analysis so just joined forces with him and started working on gravitational waves since about 89 or so So, mine is really an entry into gravitational waves from the data analysis side. Even though I was motivated to work on black hole physics, black hole perturbations, it did not happen so much. It's only now that I'm beginning to take active interest and trying to work on these issues now. So for you, being interested in general activity was very much a personal motivation. There actually wasn't much activity in that area when you were a graduate student, where you were a graduate student. Why was it so attractive to you personally? To general activity. To general activity. Yeah. That's a wider question. Just as a curious… It was really one single book written by Landa, which really attracted me to generativity, I think even when I was doing high school, I think it was in 11th or 12th standard. At that time we had an essay competition or something, and I don't know if you have seen this book, there is a little blue book with stripes on the front. it used to be it was one of these MIR publications, have you heard of MIR publishers? They are from Russia they used to publish books at a very very low cost and distributed in friendly states like in India India was always a friend of Soviet Union so they used to sell very very cheaply so I went and bought one of these books it was written by Landa It was on gravity, and special theory of relativity, and general relativity. It was on that topic. But the things that I learned there were really astounding. They were so fascinating that I immediately started reading more and more about it. and in addition to that I think in the same year when I was in Sturl Standard you know Vishu

5:00 Vishwishvada he came and gave a popular lecture on black books so I think those two things must have been sort of instrumental in changing the direction or setting the direction not changing so much I don't think there was any direction at that time in my view they were all, you know, in India, either you went and studied for medicine or you went and studied for engineering, you know, if you are coming from a middle-class background, that's what the parents, they hoped that you would do. Right. So I had already applied for engineering, got a seat, but then rejected it and went into pure science, mainly because of, I think, these two things happening. So, well, that just makes you personally curious again. Yeah. experience there was that that was the big push was for people to do medicine and engineering I think even now it is it's mainly because you know families normally don't have much of a base, financial base many people live in rented homes with very uncertain futures and what not and the parents really aspire, the middle class people aspire that at least their children would do better So the educated class now are having smaller families with a vision to have a good future for their offsprings, I would say. So that's the reason why I think there is a big push for children to get into either medicine or engineering. It's because those two people somehow thought they were guaranteed to give you a job. whereas if you did something like just a BSc an undergraduate degree you had to try several different avenues before you got a job there was no set job to do so it was really very hard to get into anything at all but I think things are changing now a little bit there are more focused courses job oriented courses but still there is a very big push I was curious because I think in Ireland there's a similar experience in that way, so I was just interested to see if there was a parallel there.

7:30 Yeah, you mean Paula described a similar? Well, yeah, I was just wondering if it's connected with the colonial experience. it could be somehow for the middle class education has been always very important not just way back perhaps during the colonial time getting a PA degree was the most important thing for middle class people but I don't know why it happened maybe due to huge migrations I think my own ancestors when they were wherever they were they had their own ancestral homes but when there was this migration from villages to urban areas rural to urban areas people did not have any good financial base and have suffered a lot and they thought that education is the best way to then start earning money and establish a firm base that may have been the reason an interest in GR and in a way when Durandar was working in data analysis because it was his time spent in Cardiff that sort of provided an avenue for working in this area. In fact, what happened soon after I started working with Durandar Bernard Schutz invited me to Cardiff for about three months. I had no background Bernard Schutz was trying to establish a contact with Ayuka and spread this area and Durandar suggested that I could come and visit Cardiff for about three months and Bernard Schutz made the offer. So I also came as a CRC visiting fellow for about three and a half months and And in fact, it was here in Cardiff that I learned computer programming. I had no background in computers until then. Not much anyway. I used to use it as a text processor, but nothing else. I mean, I did not know any Fortran or C or anything like that. So it was that visit in 1989, 10 years ago, that was really, really useful in starting off.

10:00 And then I think the problems in data analysis, at least at that time when we started off, were really nothing compared to the challenging problems that everyone does in general activity. So it was pretty straightforward to start off. The advantage we had was that I think not many people were working in that field. So we could, we had no competition, so we could leisurely, you know, develop our ideas and write papers. It may not be true anymore, but certainly at that time, it was quite straightforward. It was a new field, at least. It was a new field, yeah, that's right. Yeah. so I'm curious to know if between that period and now in addition to the fact that there are a lot more people interested in this subject whether there's also been a change in the way I guess in the way the problems are directed in particular if there is more emphasis on the particular needs of specific detentors I mean 1989 when things were a little more clear as to what detectives there would be, whether it was more general. That's right. Yeah, you're right in saying that when this business was started off in CarTech, Cardiff and Ayuga, certainly it was not very focused. Things were completely general. People did not address issues about what would happen if noise is not Gaussian, if noise what sorts of so certainly those questions were not addressed but I think what is currently being done more and more and I think also there is greater effort needed really in this direction namely we got to address the question how are we going to deal with non-stationality and non-gaussianity in detectors so in that sense there is a shift in focus and the problems that we are facing now are nothing to what they were 10 years ago that's the reason why I said in the beginning it was all very neat, you could work with fairly general assumptions about detectors and what not, but real detectors are much more dirtier than what you can simple things that you

12:30 assume to begin with and we need to tackle those issues now what I feel the way this will develop is I mean about 30 or 40 years ago when astronomers started collecting images photometric images they were dealing with raw data but as time went on they learned methods of cleaning the data and making images sharper or more well defined by just software by using cleaning techniques within codes So I think that's the direction in which we will be going most probably for the next four or five years. Eventually to dig out the useful signals, you've got to clean the data, you've got to reject things that you are confident is not due to any real gravitational waves. And that comes out of experience only. Unfortunately, we've got to wait until real detectors come up. limited. If you go and ask an experiment, should I really look at this and try to make some improvements on the algorithms that I am developing, some of them say the data from Geo 600 is not going to be anything like this. It will be of much better quality and perhaps much better stationarity and whatnot. So it is, the field is in a state of flux where different people are doing different things and how much of that will be useful eventually I don't know but I think the direction is right the direction that many groups have taken addressing data analysis problems which will clean the detector data and look for gravitational waves I think is the right way to go about my question the first group to actually look at the outputs of a real proton detector. Do you, in spite of the experiment you're saying that maybe there's not such great value in that or maybe there isn't, do you think it was very useful to do that kind of exercise for...? Yes. I think there were, it was not just of scientific value, there's something more

15:00 that people have learned from that experience you know Cardiff together with Glasgow and Munich was involved in this analysis of the 100 hour run coincidence run between Glasgow and Munich it has taught us how to organize and I think it was one big collaboration which taught us about kind of communication you need need for common data formats, the need for agreements on how you are running your detectors in coincidence, the need for storing the data in a proper manner, the need to store environmental monitors. Several things came out of that, which has become sort of social knowledge. I mean, it's not something that's in published literature, but it is something that people have experienced, it is not written anywhere, but nevertheless it has taught us many lessons which otherwise we would not have had. So I would not call it a waste because of, simply because the final data that we are going to have is going to be different. And for the same reason I think the groups which are engaged in analysing some of this prototype data are also learning very useful lessons. Another example is this 40 meter data analysis that was carried out. You know, it started initiating so many different things within the LSE like setting up committees which will go and review a piece of proposal, the way of accepting a certain piece of work or how to address the question of authorship in papers which These things all came up within the LSE, whereas in the GEO 600 case authorship and all those things were not a problem at all. It was simply an experiment to see how difficult it is to manage a data analysis of that kind, of that proportion. In fact, it has not been repeated even now, that kind of data analysis. It was 100 hours of run. The CARTEK 40 meter data was 45 hours, but one single detector. but this one was 100 hours from two different detectors so it was a horrendous thing it took about four four and a half years to analyze the data the chaotic 40 meter one took only about

17:30 about six months to analyze but it initiated so many other things it took another one year to finally you know put the paper on the web so it took only one and a half years but today if we were to do a similar thing perhaps it takes only about three or four months and finally we want to be in a stage where data analysis goes online. So it was useful. So I think in short, the answer to that question is that, yeah, I think it was very useful. Is one of the ways in which it helps, for instance, does it, I guess I imagine that for instance it can help in actually forming closer ties, say for instance between the theorists and the experimenters, was that an aspect of the thing for instance in the Yeah, that particular analysis, which I was not very closely involved, I was a distant observer, I was still at Ayuka, but what I have heard from other people, including Bernard Schiff's, is that it helped them establish relationships with experimentalists, which is very crucial. You see, normally in astronomy or even in physics, people who take data are the ones who analyze it. But we are in a sort of a strange position here where people who analyze the data are almost never near the instrument. There are exceptions to this. Definitely there are people who are getting themselves familiarized with the experiment on its day-to-day operation. But the majority of the people who are doing gravitational wave data analysis, even now, are not experimentalists. But in that, I think there is a danger. And people have to go and visit sites. They have to know how the experimentalists set up their system, well do they know about that I mean you know you don't have to know it in complete great detail but you have to have a feel for it how much can I rely on my data and it's like saying that you measure the redshift your quasar and how certain are you that kind of a feel also one has to have and I think it did help

20:00 in that help in some way but it has to happen more that's my feeling more and theorists will have to get conversant with what goes on in a laboratory. So that still needs to happen more. What kind of level of contact do you think is necessary for theorists to get a reasonably working feel for this? I mean, meeting the experimenters, seeing the instrument, being at the instrument for a while, or what kind of... I think it would be useful if theorists can go and actually work at a site once in a while, maybe about one week in six months. Go and spend your time with the experiments. Talk to them, you know, touch the detector and get a feel for it. And maybe you'll help move things and connect things and whatnot. This is really needed. You know, how is data recorded? How is data calibrated? You've got to get a feel for this on a day-to-day basis, how they do these things, I think most importantly, you have to be taking part in all the meetings that they have. Once in a while, they have these big meetings where they discuss about their problems. You've got to go and listen to them and be a part and involved in their discussion. So at GEO, normally we have these internal GEO meetings. It doesn't read normally. It's not even advertised. For instance, I was there in a meeting at Hanover this Friday, Saturday, Sunday, three days. So it is a geo-internal meeting, and all the experimenters come. Most of the theorists also go to that. I would say all the theorists also go to that. And then it's a very rigorous, very technical discussion for about two or three days. Theorists who are worried about, you know, initially, oh, all this technical jargon, And initially, I mean, initially it will be, there will be some barrier, but you've got to talk to them, mix with them, find out, you know, get over the jargon, and finally you will get a perception of what goes on behind the scenes. So I think once in about six months, if not any more frequent, is certainly essential. My impression is that Cardiff is more closely tied in with the GEO experiment than say theory groups in the US are with LIGO.

22:30 There are a few exceptions there again. I think there are one or two groups which are really taking special interest in going visiting the sites, that's helping with the experiments and whatnot. So there are some exceptions, like Sam Finn does that, Bruce Allen does that. But there are others who are sort of very remotely doing the analysis. I wouldn't encourage that. In Virgo, there are no theorists who are doing any serious data analysis. And the reason for that may be that the actual data analysis is anyway about two or three years from now and somehow in work of there has not been a serious effort in algorithm development even though data analysis there are they have been putting some efforts they are developing software they're developing you know common data formats with LIGO so there are efforts at different stages simply because they are not doing any algorithm development it doesn't mean they are not doing any data analysis at all they are doing some sort of data analysis but certainly one needs to have the theorists taking part in this because of the special circumstances in which we are in we cannot see all gravitational wave signals very clearly in data so theorist input is needed So that has not happened very much in Worker. But in LIGO, I think there is certainly some effort, but that has to perhaps enhance a little bit more. I think it's fair to say that. So in your mind, is that the main reason why theorists play this, perhaps this unusual role in the case of gravitational waves where they're involved so heavily in the data analysis, that the reason being that because nobody's actually seen only image of what they look like is the theoretical one. Is that probably the primary reason? I think the primary reason is even at the level at which we are constructing, the level of sensitivity at which we are constructing our detectors,

25:00 gravitational waves from astronomical sources are not going to be overwhelmingly strong. are going to detect them is by pulling them out of noise as you would even know. There is no other way at least in a theorist mind you can do it with the present day detectors. To contrast that I think something like LISA whenever there is a supermassive black hole coalescence or even galactic binary, it is not even inspiring, the signals will be so much stronger that you would not need much of theorist's help. You can see it in your data whenever such an event takes place. And in fact at that level, the theorist can learn from the data. But the situation we have for the present set of detectors, for the initial at least for the next about five to six years, is such that theorists ought to give their input. What is the best understanding that they have about gravitational waves from known sources? And how can that be used in enhancing the signal-to-noise ratio? Once a detection has been made, people will be more adventurous. I mean, you can explore other avenues. The reason why this input from what we know has to go into data analysis is because of the history of gravitational waves itself, which you know very well that it has not earned a very good name. once we have had some reliable detections once we have established that our instruments are capable of seeing these waves the rest of the community will accept, the rest of the academic community will accept us and then we can be more adventurous and then make and then theorist input is not so much needed I suppose I mean it will be needed continuously but not for other things. I mean, you could look for these unknown sources, etc. Right. But the way I see it is that general relativity has remained as a theoretical science for such a long time, and theorists want to get inputs

27:30 into their modeling from gravitational wave observations in several different areas, like black hole physics, binary evolution and neutron star physics, etc. theorists will be taking part in data analysis all the time, but I think this one-sided effort may disappear. Once experimentalists have built their detectors, once their responsibility of having constructed these things are over, they will also take part, I think, because eventually they are also interested in science. I think at the moment there is so much pressure on them, first they want to build an instrument, and then I think they will also start taking part in data analysis. Is there any difference there between say the Geo 600 and the LIGO experimenters, or are they both about equally taken up with the task of actually getting the detector? I think even in LIGO, there is quite a lot of effort needed because it's a huge project, you know, two and a half interferometers being built, and there's a tremendous amount of public pressure on them, political pressure on them. So I think the experimentalists, even in the U.S., have a lot of responsibility to finish the detector first. And once that is over, I'm sure they will start concentrating on doing science with the data. So since the group here is interested, is working closely, with the G600 project, of course, and also with LIGO through the LIGO scientific collaboration. Is there a difference in focus between data analysis for the two detectors? Are they sufficiently different that there are different issues that arise or did the same issues come up? Yeah, okay. That's a very interesting question and we have discussed within the collaboration about what should be the priority for us. As you know, GEO has very little hope of seeing binary coalesces, whereas for LIGO, binary coalesces are the most important events that they are going to see. So, as a result, what we have done is that if there is a neutron star,

30:00 or there are neutron stars in our galaxy, with an eccentricity of something like 10 to the minus 5, 10 to the minus 6 we are bound to see it using GEO and with a good signal to noise ratio using signal recycling that is narrow banding so we will definitely operate GEO whenever we can in a narrow band mode of operation so let me spell out the strategy in some detail Initially, when Geo comes online, it will be on a par with LIGO. It might be worse by a factor of 2, a factor of 2 to 3, depending on which part of the frequency spectrum we are talking about. So when that is the case, then even though Geo all by itself will not be able to see any binary in spirals and coalescences, coincidence observations with LIGO. So our own feeling at the moment is that we should have coincidence analysis with LIGO for binary in spirals when LIGO is running in coincidence with GEO. So we will operate the detector in broadband. But when LIGO becomes much more powerful or when LIGO goes down, let's say for instrumental upgrades and things like that, DOCAN cannot have the hope of detecting binary in spirals all by its own, then we will operate the detector in narrow band mode of operation to make targeted searches. Targeted searches meaning we have identified X-ray binaries, low mass X-ray binaries, super and we will go and specifically tune the detector to the frequencies at which we expect to see gravitational waves from these sources and try to make a long period of observations, like three to four months of observation. So that's the strategy really. So whenever we can, we help in coincidence observation. We would like to take part in coincidence observation and whenever there is no hope of, you know, single-handedly detecting, we go and operate

32:30 the detector in Narabai. The, so, well, obviously, I guess the analysis issues are very different than those two cases. So you, is the, I was trying to think of which is a more difficult problem. I guess in the case of targeted search for pulsars, as I understand it, you have a big problem because of having to account for all sorts of issues like the orbit of the Earth and so on as you're moving around. That's right. But targeted search is not such a problem. Because you know the direction and perhaps you know something about the source. Incidentally, even when we are operating the detector in a broadband mode, we will be carrying out... Uma, can you come after about 5 minutes? Or 10 minutes? Even when we are operating the detector in a broadband mode of operation, we will be doing all sky searches. When you are doing a narrow band mode search, then the in other frequencies will be so bad, we are not going to do all frequency search, but we might still do all sky searches. Right? So it's going to be horrendous even at that stage. If you are looking for a particular source, then it's not computationally demanding. Yeah. So all of the computers will be running all the time, whatever we do in any case. I think that's what's going to happen. Yeah. Let me just go back a little bit to this question of the experience built up of cooperating with, collaborating with experimenters and getting experience of noise and in particular to something that you touched on earlier, that is when things began and theorists were looking in general terms at data analysis issues, they tended to talk about Gaussian noise, and now the focus is much more on non-Gaussian noise, and so I was struck at the time of meeting by the fact that people who, theorists who were working on data analysis issues,

35:00 would keep looking, you know, look very distressed and say that all the noise is so non-Gaussian and so non-stationary. Was there a point for you at which you were struck by how different the real noise from the texture was different from the type of noise that theorists hadn't accustomed to speak about? Right, I think it was not such a big surprise mainly because we had input from the 100-hour experiment. I had looked at the data and we had characterized it. I visited, we had looked at statistics and what not, and we had known about this non-gaussianity and non-stationarity. What we have been told by experimenters is that finally the data, when it comes off from these big detectors, they are hoping that it is not going to be as non-stationary and as non-gaussian. Now you mentioned about TAMA, which is a finished detector, which still has all this non-gaussianity and non-stationarity. I think they have still not reached their targets sensitivity right you know that at high frequency they have already reached their target sensitivity i think the interesting part is the low frequency i think if we wait for another about six months which is i think their schedule they may actually improve their noise in the low frequency much better at which time i still don't expect it to be gaussian or anything but it may be manageable but we still don't have algorithms to deal with all sorts of non-gaussianity. One of the things that has happened in analyzing this 40 meter data is that we have learned how to handle some of the non-gaussianity issues or non-stationality issues but a lot more needs to be done. We have learned about techniques to remove lines, we have learned about techniques to remove transients by using Qt algorithms we know about veto techniques now which can remove these non-stationary or non-Gaussian signals transients but a lot more will be needed so when the detectors come online let's say even 5 or 6 years from now when we have a very good hope the first detections are going to take quite a long analysis it's not going to be an easy task anyway

37:30 so the answer is yes I mean there is some amount of worry we have not reached the final stages detector construction to be completely disappointed it is worry but it is not a complete disappointment as in for instance on February 17th Jio is taking its first ever data from the are looking forward to analyzing that and we want to see how good the behavior is whether there are long chains and longer chains of non-stationarity and we want to characterize the noise first so that that will give us some understanding how these detectors are going to behave because experimenters tell us that you know motley is locked it can lock for a very long time and whatnot. So we want to take about at least two to three days' worth of data and see how the instrument is working. Is there an issue of who is best able to deal with non-Gaussianity in the noise, say the experimenters or the theorists? I mean, you mentioned cleaning the data as it were who would be best I think it's a combination the way we are hoping to deal with this that since at the moment experimentalists are not doing the data analysis directly we will help them to identify instrumental transients environmental transients etc and then they will go back and try to rectify To give you one example of that, we have been seeing these mains and harmonics and how it is shifting and what not. You may have heard of that, you know, the mains frequency gets into the detector and then there are harmonics and these harmonics keep on shifting. The reason why these harmonics appear, first of all, is because of rectifiers and other complicated circuitry inside the instrument, in the electronics of the instrument. And now they have found that there is something called hum in the detector. And this can be removed if you make your circuit connections in an intelligent way. It's not really intelligent. It's a little bit of care.

40:00 don't create any loops make proper soldering etc so there was a whole session just on Saturday at the Geo 600 meeting where we discussed the experimenters warned the whole collaboration one set of people who have experience in this told why this could happen and asked them to exercise caution and we have now set inspection dates, hum inspection dates three months or four months when people with experience will go and check all the electronics so it's I think a lesson that we will we will learn from data analysis, we will give our feedback into the experimenters and they will rectify, it's an ongoing learning process and that's the reason why all the instruments will have the so called engineering run I think LIGO calls it engineering run And we call it, I think, test run or something like that, for about a period of one and a half to two years. That's what we will be doing mostly. And you think it will work somewhat like that, that in addition to the experimenters, I suppose, you know, just working with the instrument directly and looking at their own, you know, working in their own sphere, that also they'll be sending data to the theorists who are doing analysis and spotting things such as transients. the more cleaner data it's not going to have any gravitational wave data but nevertheless we are interested in looking at it and seeing the quality of the instrument and we will give the feedback and even they are doing the detector characterization all the time it's not that they are not looking at any data they are in fact they are not archiving the data that's the only difference the reason for that is you know when you talk to an experimenter what they say at the developmental stage instrument is evolving so fast it's meaningless to archive any data you know the data that's one week old doesn't represent the instrument at all it's completely different so the data that they are looking is on time scales of hours or days nothing more than that so that's the reason why they don't take part in this data that's taken off from the detector that then we will give the feedback it's also because you know the time time constraints that I mentioned

42:30 really involved in instrument construction and they don't want to do long stretches of data analysis. Yeah. But I think they are quite good in what they do. There's no doubt about it. Sure. Yeah. Having seen these guys in action. Sure. It's a phenomenon. Yeah. Yeah. To build such a sense of instrument. Yeah. Yeah. I should let you go at this point, I guess, since we're up to one o'clock and maybe we can continue. No, it was a pleasure. Yeah. Yeah.