Interview with Tom Prince

0:00 Okay, so now it's actually working okay. Well, I guess just to begin somewhere, because maybe you'd be able to give me some feel for this. I sometimes hear people say that it's unusual to have theorists, like maybe the people in Kip's group, working on signal analysis issues. I was wondering if that's your experience from other fields, if that's really true. Yeah, I would say that that is somewhat unusual, in that typically, well, one thing just generally is that the distinction, I think, between theory, it used to be that theory was a little bit more distinct as a discipline. I think that there's much more of a blurring these days between theory and even experimentation. You see that in just about all areas. I think it's been true in the past, but I think there's even more of a blending these days. And part of that, I think, is actually because of computation. in that computation is a newly developing tool. It's not the province of either experimenters or theorists exclusively. And so it's almost a middle ground. And I think data analysis is part of that computation thrust. And actually, one background for this is that I'm on the panel for theory, computation, and data exploration for the Storm and Astrophysic Survey Committee. And I'm the only non-theorist on that panel. But it is theory, computation, and data exploration. So you see there sort of the blending of all those various themes, and that indeed does stretch the entire realm. So theory is what we think it is. We all agree with theory. And then computation is traditional computational astrophysics, so using a computer or a CIST in order to do modeling and simulations. But then data exploration, then, is a different area entirely, which is much more like data analysis. I think it's interesting that this panel, which used to be a theory panel, now covers this range, and it's all in one panel.

2:30 And it's mostly theorists, except for me. But a lot of people, a lot of the theorists are doing a data exploration now. So it's not, it's a new phenomenon you think, feel, but not something that necessarily confined just to relativity? I think that's right. I think that in general in, well, I think that the depth to which theorists have gotten involved in the actual signal processing is somewhat deeper in gravitational wave astrophysics than in some of the other areas of astrophysics. And I can say by that I think that's true. But I think that this is only sort of, it's kind of out in the tail of the distribution of a wider trend, which is this fact that actually, as far as disciplines go, the distinctions between pure theory and analysis is starting to blur. And what would you see as the motivations from the theorist's point of view for them to get more involved? Is it just because they can now? I think they can, yeah. In other words, in order to, before, do an interesting measurement or an observation, you have to go out to the telescope, sit there at all hours in the night. You have to go build an instrument, solder things together, all things that are not necessarily very appealing to a theorist. But nowadays, for instance, if a theorist wants to do, is thinking about large-scale structure, they can go to some survey data and actually test out their theories right there at their desk. And so I think that's part of the trend that's blurring this distinction is that data and information are much more accessible now to everyone than they used to be. And correspondingly, I think that a lot more of the experimentalists now are doing, are also doing analysis and phenomenology, which is kind of going more towards the theoretical lines. Right. I think the middle is filling in. Right. They're moving in the other direction. Right. Indeed. Indeed. And I guess, well, that sort of maybe naturally leads into something I was curious about, which is if you have any sense of, well, for instance, the experimentalists feeling uneasy about the theorists moving in their direction.

5:00 If, for instance, they're worried that the theorists won't know enough about data or the instruments or whatever. Right. I think that there is that worry. But in some sense, I think it's all placed. Each, you know, sort of coming at it from the theory end, you know, misses some things in the data analysis. But also coming out from the data analysis end, or the experiment end, also misses some things. So I see it more as two parts of an integrated whole in which integration hasn't taken place yet. You see what I mean? Yeah. So let me give you an example. I think the standard thing, this is not what I'm saying, but that you hear from the experimentalists, oh, well, the theorists really haven't really worked with real data. And to some extent, that is true. And when you really do get into working with real data, things happen and you see things that you just didn't think about before. And oftentimes, 90% of your time is spent on issues which are not there in the first theoretical analysis. And so, but the theorists, if they continue to go to do data analysis, will learn those things as soon as there is relevant data there. So in that sense, the experimentalists are right. And also the experimentalists are saying, well, we have to get the instrument right before we do the data analysis. And indeed, that's absolutely right. Without the instrument pumping out data, you don't... So in some sense, the experimentalists then are working on the critical path. And they're saying, okay, we're doing the first thing that needs to get done in order to do the whole thing. And these theorists are out there. But then look at it from the theorist's point of view. I think that what everyone believes is that when the interferometers really start detecting sources, there will be a lot of new information and a lot of theory to be done on that. And so the theoretical community, then, in my opinion, is somewhat in limbo right now. Because they don't have the data that they need yet.

7:30 And they know it's coming, but it's always late. So what do they do in the meantime? time. Well, they can work on astrophysical sources or other aspects of relativity theory. In some sense, when you really get into that, you quickly realize that you'll learn so much more when the first observation comes about. For some of these items, that a lot of, I think, there are some are drawn to that sort of data analysis and trying to say, okay, ourselves and get mentally prepared to do the interesting data analysis and get that out as quickly as possible. Plus, you know, thinking about really what are the, can we do anything from our end to increase the sensitivity of the instrument by really looking deeply into the data analysis issues. So I think from the theory point of view then, what is the real value there is that one looks at the problem from a much more analytic point of view and tries to think about new techniques for doing data analysis that aren't necessarily the standard techniques, but may give you an enhanced sensitivity for the answer. And actually that work is very valuable, very beneficial. So for instance, the experimentalists will not confidently calculate the, you know, and generate the in-spiral templates that one needs to do with the analysis. So even though they are working indeed on the critical path, when they turn on the instrument, they won't have the in-spiral templates that they need, unless the theorists turn around them. And so then on the theoretical end, there's a lot of interest, as he talked generating those templates, how do you get them very accurately, getting them into regimes which we're not very confident about right now, namely neutron star black hole coalescences, black hole black hole coalescences, higher mass objects, spins and so forth, and how does the wind do all that properly? And so that's something that the retroimmunity can really bring to bear on the whole process. I'm curious, is the extent to which the experiment will depend on theoretical input. Do you think it's an aim unusual for LIGO, or is that fairly typical? I mean, for instance, needing theoretically derived

10:00 templates, for instance, for some of the processes? I think, actually, it is not totally unusual. Okay, so, for instance, you look at the cosmic microwave background, where the theoretical community in recent has developed the whole concept of the acoustic peaks or Doppler peaks in the high L moments of the CMB. And that's one place where definitely the theoretical guidance is very important in that, or the theoretical ideas about what one might see, or really defining the context in which the experimental results are being interpreted. Now, there it's a little bit different because it's more context rather than detailed templates. But let me just say that one of the things that is being proposed now by the panel on theory of computation day analysis is that with each mission, say space mission or major ground-based facility, that there be, right up front in the development of those projects or missions, a set of theoretical challenges defined for those missions, which then the theorists then concentrate on working on those theoretical challenges. Because there's a feeling now that perhaps theory hasn't been as active in the front end of missions and projects as they might have been. LIGO is not being reviewed in the Cato survey, but LISA is. And there are definitely, there's a whole set of theoretical challenges being developed for LISA. And in that sense, you know, the analogy to LIGO is almost one-to-one, right? There's templates, you know, what kinds of objects you see, what's the best way to get the signal noise, you know, all those kinds of things. And so that whole set of theoretical challenges. And for LIGO, one can also generate them also. So besides the templates, one certainly, if we ever see a black hole-black hole coalescence, I think it's well known that one doesn't know how to do the coalescence of two black holes. So that's a problem. So that's a big theoretical challenge that sits out there. So I would say that there's certainly a very strong component of that in LIGO,

12:30 but it's not necessarily unique. stronger than in some areas oftentimes one reason why it's strong I think is that it's new territory by and large where a lot of space missions or ground based observatories have had more of a theoretical foundation laid over many years more optical IR so you find a little bit less But, for instance, even like a NGST, Next Generation Space Telescope, the theoretical challenges there have to do with formation and evolution of galaxies. And so that's a big theoretical challenge, and there's a lot of theoretical input. It's not quite so detailed as templates, so it doesn't permeate quite so far down into the data analysis as for longer. Yeah, that's interesting. Interesting. So it's more of a general, there's a general movement towards trying to encourage theorists to work on issues that are... Well, this is theorists themselves talking. Yeah. Right, okay. So it's not in some sense encouraging, because this is a panel of theorists, and the only non-theorists in the panel, okay. So it's what they want to do. Yes, exactly. And part of it is financially motivated, is that by getting in at the beginning of missions, That's also where a lot of dollars are flowing around. But at the same time, there's also then just this feeling. I think the post-COBE map CMB was seen as a very interesting case for theory really defining some of the objectives in a major observational area. And the thought is that perhaps theory should be doing more of that. In other words, there's many, many issues in theory, but perhaps there should be an attempt to focus theory on topics in a way that matches a little bit more closely And benefits, then, major facilities are being thought about. Well, just briefly, you mentioned the new types or, you know, relatively novel types

15:00 of signal analysis techniques that might be developed for LIGO. What are the sort of things that are being thought of for LIGO that are new, relatively novel in your experience? Yeah, well, let's take one that is not necessarily a new and novel, but LIGO will push it further. And that's, for instance, right now, let me give you an example, in radio pulsar astronomy. So our group, when it started out doing radio pulsar astronomy, brought a significant component of high-performance computation to that. And we then introduced the first time that what are called acceleration searches were done in radio astronomy. So these are searches for objects that are accelerating, and so typically in binary orbits. And so doing that then, we found one of the three relativist neutron star-neutron star binaries, the one in the Kavya cluster. So, interesting enough, we worked on that, And those techniques are actually only now being applied widely in radio astronomy. So radio astronomy is a more traditional field. I'll see you later. Okay, yep, yep. And where is LIGO taking it? Well, LIGOs, Brady and Creighton, have now published a paper, which really, you know, they have the time and the energy, and perhaps even the lack of data, destruction of data, actually go at and really think about what that algorithm should really look like. People like Stuart and Anderson and myself certainly contributed to that process, but they really then formulated a hierarchical algorithm. How do you do acceleration searches or searches objects that don't have discontent velocity? How do you do those types of searches optimally? So there's a good example of something where, you know, I've been referring, you know, my colleagues in radio astronomy to their paper. And so there's a place where, you know, the theoretical community from LIGO is able to develop and enhance, you know, and actually codify and write down, you know, algorithms that are actually more advanced than are being used in a very traditional field.

17:30 namely radio pulsar astronomy and our views to that field. So that's one example. So the theorists in Simpsons catch on reasonably quickly. They're able to contribute something. I think that's definitely true. Secondly, I think that there is something to say that if you don't have the data there you can actually invest your time and analysis of the algorithmic content. Whereas, let me say, in a data-rich field like pulsar radio astronomy, more times the thrust is you develop the technique because you need it at the moment. So you're data-driven. You have the data that's there. You say, okay, I'm going to quickly go at the data and develop a technique that works in this case. But sometimes then you miss the more abstract analysis of the algorithm. And a lot of our processing algorithms, signal processing algorithms, came from, say, the EE community, which thought about it in a very theoretical and analytic way. But both are needed. So you take the Fourier transforming and windowing as an example. everyone who's ever worked with real data knows that there's a difference between those theoretical analysis of the algorithms and the writing them down as a theoretical construct in the equations that are there and actually applying them once you apply them to data they don't work so you have to do things because the data is, well it's got gaps in it well it has the frequency spectrum rolls off in the wrong way or the noise isn't really stationary surprise surprise All these things which then require, then, point solutions. But at the same time, a lot of times, the experimentalist or the social observer is worried about nothing but point solutions because they have to solve the problem now. It is today's problem. And they don't then take that broader sort of analytic view.

20:00 So I think, actually, the match of the two is a perfect situation. And it's unfortunate sometimes that the two, you know, there isn't as much respect perhaps between. Right. I guess that can happen, of course, when you have different fields mixing. Right, yeah. Well, you mentioned that the data is more widely available and so on. So I'm curious about that. It'll be interesting to see, I suppose, how that works out in the LIGO case. Well, in the LIGO case, actually, the data is not more widely available. So I was talking more about why that's a trend for astrophysics. I think LIGO is somewhat different. I think LIGO, it's, again, this suspension of the data. In other words, the data's out there, you know, sort of 2005 time frame, perhaps. We'll have data sooner on that, but I think most people are willing to admit to themselves that the likelihood of really getting astrophysic sources is out of the way. So I think for LIGO, it's more that there's a theoretical community that's very interested in this data to come. But in the meantime, what do they do? So in the field like radio pulsars, that's your field, do you take most of your own data, or do you also find yourself analyzing data that's more publicly available? There it's almost all our own data, but it's not exclusively that. Yeah, we go to the telescope, point at objects, look at the data. So this is sort of looking off into the future, but obviously in the case of LIGO, if theorists are going to be involved with real data, then they're not going to be taking it themselves in some sense because there's only a certain number of instruments. Right, and LIGO is unusual in that sense, and I think that's one area where... I think that actually perhaps people haven't quite understood do that, is that LIGO data really is different in many ways, in that it's really not much data when you come right down to it. I mean, it's huge in terms of how much the instrument is pumping out. But basically, the basic data stream is only an audio data stream. It's roughly, at most, a few kilohertz of the strain data. So eventually, once there's all kinds of data reduction, then you get to the strain data. So as far as the real science analysis,

22:30 that's audio data so it may be audio data for a year but that's not that much so that's kind of different than say some of the high energy physics experiments or data that telescopes put out which is two dimensional image data so it's really very different in that sense the fact that the theorists are not taking the data doesn't make much difference at all. Because one is the thing doesn't point. So if we go to Arecibo or wherever parks, we actually have to decide where to point the thing and which filter bank to use, what frequency do you observe at, all those other kinds of things. So that's part of the whole observing lore. It's not going to happen this time, right? It's basically the signal parameter that comes out as strain, you know, H versus T. And so because of that, it's a different type of data set. It's a data set that's actually somewhat perhaps similar in nature to radio astronomy data sets because in radio astronomy data sets, eventually one has pretty much a signal versus time. once you set up your basic parameters of your observation, what you're after. But I think one item is that perhaps hasn't been thought through, and this is something that the LSC is now starting to wrestle with, is when you have actually such a small data set, and eventually a very large number of people interested in analyzing that data set, what does that mean? And how does one actually do that and go about that? So with space experiments, you know, it's pretty easy. You chop up, you know, this person wants a point of telescope there, you're getting that data. This person wants a point of telescope there, you're getting that data. But here you only have one audio data stream. And then how do you then go about the analysis, efficient analysis of that data? As a team, do you do it as a team? Do you do it as a bunch of disconnected individuals that go off, do their own thing, and then come back and talk about it? How do you do that? So it's still something to be up in the air? Yeah, yeah, sure. Yeah, the legal science collaboration. It must eventually deal with that. Right. How does one do that?

25:00 Interesting to see. Well, do you see, you mentioned that your own perspective is a little bit between, say, the theorists and the experimentalists, maybe you're an observer, but with a good deal of... Yeah, so a lot of my work has been sitting in this intermediate position between, you know, sort of observer, I would start out as experimentalist, so experimental slash observational work, and then but then at the same time being astrophysically motivated, so that, you know, and signal processing motivated, you know, and so bringing sort of signal processing data analysis techniques sort of in the real domain. all the issues that have to involve with working on real data and instrumental signatures, and we've done a lot of that, but then also very much interested in the actual analytic techniques that have been developed by many, many different groups, especially out of, like I said, EE, geophysics, all those kinds of things. So, yeah, so that's the area I tend to play in right now. So it's that middle ground. And do you see people who are involved in gravitational wave experiments going in a similar direction up to now? The theorists have been more interested. Do you see the experiment? People that have actually come from the middle ground are few and far between. In fact, there aren't many. Stuart Anderson, who was a postdoc with me, is one such person. So there aren't actually a lot of people that have come into that middle area. And it is indeed. It's either theorists who have been relativity theorists, basically, or have worked in Kip's group who are sort of a relativity theorist, but have been perhaps trained in sort of this data analysis approach, which is a little bit different, but haven't done a lot of real data analysis. Or it's people that really are there and have done, you know, do the development of interferometers and control systems and all those types of things. And then there's also a few people that are coming in which is a high-energy physics perspective. So they have done data analysis on high-energy physics experiment, and they're more in the middle ground also. So you have people more on the experimental side

27:30 who have the right background from their previous experience? Well, again, I wouldn't say right background. I think that's kind of a value judgment. But I think traditional LIGO has been divided up into this theory coming into the middle ground and experiments sort of eventually moving up into the middle ground. I guess as far as this, You know, the traditional LIGO community has been sort of, you know, at the two ends, moving towards the center. But I think that, you know, there are certain, a very small number of people sort of, you know, from astrophysics, basically. People have worked in compact objects, radio pulsars, that sort of thing. Just from Caltech, you know, a few people now. And then the other people that are also part of that middle ground, but from a different perspective, are people that have done high energy physics experimentation. so they're experimenters but they have done data analysis so with real data well I have one other quick question before I let you go which is on a completely different topic about gamma ray bursts and that kind of issue one of the things I've been looking at recently was the results of Wilson and Matthews numerical simulations of neutron star binaries where they had this neutron star compression effect if you're familiar with this oh yes And I was curious because one of the things that they were saying is, well, you know, the relatives don't believe what we're saying, but what we want to try to do is develop a gamma ray burst model out of these results. And they say, well, the trouble is we kind of find it difficult to get an audience there because people say, well, the relatives have already rebutted your results and so on. So I'm sort of curious just to get somebody who has more of a perspective on whether, you know, people who are interested more in gamma-ray birth type of issues are aware of their model. Well, I'm certainly aware of their model, but more from, you know, hearing about it from the relativity side. And, you know, I can't judge this myself, but, you know, people I've talked to around here have tended to say that if I can't, you know, you can write down reasons why that doesn't work. As far as models go, I can't say whether or not, you know, the broader, say, Emory Burst community is aware of their models. Although, you know, they're certainly the only model of binary co-essences.

30:00 In fact, I think that someone like Stan Woosley is more widely known as a person who has a whole range of models of binary coalescence leading to gamma-ray bursts. So he's talked about black hole neutron star, neutron star, neutron star, and so forth, and formation, sort of tidally shredding the neutron star, creating then a short-lived and provides the energy for the gamma-ray burst. So probably the gamma-ray burst community is interested in possible models based on compact binaries, but there are already people who... Yeah, I would say that probably that particular model, Wilson Company, is one of a class, and I think that probably most of the gamma-ray burst community you know or a large amount of most of them would say okay yeah there's this class of models that the theorists are working on and and actually right now it's a most people are more interested in from an observational point of view of really pinning down the whole afterglow the scenario for gamma-reverse. And so, although people are very interested in the progenitors of gamma-reverse, so far, observationally, there haven't been a lot of handles on that. And yet, there's a lot of work to be done on the afterglow situation, just what kind of conditions and evolution is there after the initial energy release. So I think a lot of the people working in gamma-reverse right now. I think, okay, initial energy release, boom, and then what happens? You know, internal shocks, external shocks, you know, evolution of the radio, evolution of the optical, what's happening, reverse shocks, all those kinds of things, which are in some sense don't need to assume a lot about the progenitor. So I think that's where that field is. Although I think as time evolves, people become more and more interested in what the progenitors there. But at the same time, it'll be interesting to see how much information we get from the

32:30 observational community about that very early time. In fact, I think that's one place where LIGO can really play a role, is that if, and that's an if, if gamma ray bursts really are associated with gravitational wave events, then it may be that gravitational wave detectors will provide one of the few mechanisms for really, really pinning down what the progenitor of the system is. That'd be interesting. So that... Is that a possibility that the gamma-ray burst community is kind of interested in? I mean, the way it would bated breath as it were. Well, bated breath is probably not, because there's so much happening in there, you know. certainly proposals refer to LIGO as, okay, we're going to put this Gammary Burst mission and it will be operational at the same time LIGO is operational and looking for things. And that's kind of interesting. But there are yes about the whole, we don't know yet about all the all we can say is that there's a possibility that the same events which give Gammary Burst also produce gravitational waves. And then even beyond that, there's the idea, there's a question of whether or not, even if were true, whether or not you'd actually ever get a real coincidence between the two, because of distance. Right now, the pre-memory births that we know of are too distant for libraries. At least we think they are. Well, thank you very much. I hope that's helpful. It's very helpful, indeed. Thank you. I should actually say at the end, since I forgot to say at the beginning, that it's that we started at about 9 in the morning and I was talking with Professor Tom Prince. Okay. Just so we know. That interview was recorded on the 14th of July 1999 with Tom Prince.