Interview with Bernard Schutz

0:00 So now I think we're actually on. And it's February the 12th at 2.30 and I'm talking with Bernard Schutz. So usually a good place to start is to ask how you've come to be working on gravitational waves and your history in the field. It goes back quite some time. As a graduate student, I got started on the theory of what one could call gravitational wave sources of a certain kind anyway, pulsating stars and working out stability and the mathematical theory of pulsation in the context of relativity. Then I got interested in a number of other things, binary systems. It seemed like anything that was interesting in general relativity to me was giving off gravitational waves almost sort of incidentally because I was studying more of the dynamics of the systems. And then in the mid-80s, the Glasgow group which had started out making bar detectors and then was just making the transition at that point to building an interferometer I think they were just building their first prototype 10 meter interferometer began to be asking for more money from the funding agency, it was then the Science Research Council I got consulted by that for my opinion as a sort of relativist, a theoretical relativist in Britain, as to whether this was a good idea. And the more I got to talk to them, the more I got enthusiastic about really trying to support their project. And they They had, as I said, I guess they had recently built this 10-meter interferometer, and they were moving toward making an application for money to build a large-scale detector.

2:30 The first application I made was for a kilometer-scale interferometer to be built in Scotland. And so I think over a period of about two years, I moved from being a kind of outside advisor or assessor to being not just a supporter but having my name on the proposal for the one kilometer as the theoretical house theoretician. And so I then felt that the only useful role I could play inside the project was to take responsibility for data analysis so that was the part of that proposal that was allocated to me plus then the writing of the scientific case for it and then doing the lobbying and the scientific politics that goes with trying to get a very big project approved thank you very much just prepared great, thanks So, we started out that way, and I just stayed, I've always had a very good working relationship with Jim Huff and with the people of Glasgow, and that's continued through the period in which they joined with the German groups to propose first the three-kilometer geodetector, and then when that didn't succeed in the end, the 600-meter geodetector, which is now under construction. So I've stayed in that alliance, that collaboration, since the mid-'80s. At the time when they were putting forward the proposal for their one-kilometre device within Britain, was Ron Rivers still in class at that time? I'd have to check dates, but my recollection now that the proposal for the one-kilometre is dated 86, and my recollection is that Ron already had a half-time position at Caltech then. and was making up his mind whether to go to stay in California permanently.

5:00 And I think round about that 1986-1987, he did decide he would go full-time to Caltech and nevertheless remained for a number of years after that, he remained nominally. I think he had a visiting professorship or something like that at Glasgow which meant that he was on proposals and he was a frequent visitor to the group and so on but it was just at the time he was making the transition I believe that the one kilometer proposal was the first proposal anywhere for scale detector. And it was the first time, therefore, the people had sat down to work out what the scaling was, really how things had to scale up from laboratory size to this kilometer size, and try to estimate costs and so on. And to show, part of the scaling was to show that the required precision of control could be achieved in positioning the mirrors and keeping them in one place and all that, that it wouldn't scale into such a difficult problem that you couldn't contemplate doing it. and that exercise proved very useful because I think then it was used by the Americans it had broken the ice, if you like, for the Americans and they then started up their much more ambitious LIGO proposal but it took the LIGO, it took a decade or so from the time it began but full funding, almost a decade from the time they began, the LIGO proposal. So, in that sense, you know, Robin sort of took the experience of the one-kilometer proposal to the United States and used that as the basis of the LIGO.

7:30 One of the interesting things about projects like certainly in the case of Geo and Virgo seems to be a fairly close association between the theorists and the experimentalists at least at some level and also the fact that so far at any rate the theorists seem to have been for the most part responsible for the data analysis effort. Kurt for instance was remarking the other day that it seemed unusual that it was theorists and not who are so interested in that side. Is there any kind of reason for that that you think of, or has it happened by chance? One, yeah, I think there's two reasons. One is that the experimentalists in this area have been so challenged by the technology and have been forced by those challenges and by the slowness with which they were able to get funding to take a very long view and to put in their minds that real data isn't going to come for another decade or something like that. They didn't really have to worry about data analysis. And so if you look at the experimental groups, they didn't, I think the geo-proposals are the only ones that actually built data analysis, built in a budget for data analysis. LIGO never had a budget that was approved by NSF in its funding for data analysis. Virgo doesn't. both of those projects expect outside people to do some of the work, I suppose. So one of the reasons was this long view that the experimental work was so hard and going to take such a long time that you could worry about data analysis later. Then the second thing was that the data analysis has to be sophisticated. it. And it goes beyond the kind of data analysis that every physicist, every experimental physicist learns about. You really had to be familiar with the mathematical theory of signal analysis

10:00 and the least maximum likelihood theory and filtering theory. And that's something that engineers, particularly electrical engineers, are fairly familiar with, but physicists don't get very much of. So the people who went into, the physicists who were designing these experiments, had no head start on the theorists, let's say. And since the physicists, the experimental physicists, were fully occupied with their lasers and their mirrors and so on, it left the whole subject open to the theorists. Why do you think the theorists were interested in moving into that? Well, from my point of view, it was the only useful thing I could do. I wanted to lend my name to the project, and I wanted to be able to help guide it. And there was nothing I could contribute on the hardware side, so it seemed a reasonable thing to do. So I particularly integrated myself into the project and took a responsibility and got research grants to support it and so on. In America, Kip worked very hard, but always from outside of LIGO. He would sit on an advisory board or a board of directors or something, but he would never actually work, associate himself as a member of the LIGO team. got students to work on things. He worked himself on a number of theoretical questions, including the baffles, the light scattering, right, which you know about. And that was not a question of data analysis, but it was a question of somebody with a good theory background had to sit down and think about this. And so Kip did that. So he's always worked on behalf of the project, but he's always held himself off from being a member of the LIGO. team itself so from your point of view was it a sense that you felt that the project was important enough for the

12:30 from the field's point of view that it was going to be a very significant thing for theorists if it was successful that was part of the motivation, I certainly felt it was but I don't remember in the mid-80s thinking about the world in terms of what's good for theory or what's good for experiment I thought of what's interesting for my subject and what really excited me was the fact that relativity had been such a theoretical subject here was the first time that you could get information conveyed to the Earth by means of purely relativistic phenomena in the gravitational waves, it turned relativity into an observational science. And I was very interested in astronomy. so that aspect of it really was the most exciting to me so I didn't think of it, I don't know if you meant to put it this way but I didn't think of it as what relativity ultimately will do what gravitational waves ultimately will do for relativity theory I just thought of it as how exciting to transform relativity into an observational field And then it was natural to move in that direction. But then I found, the more I studied questions of signal analysis and data analysis, that they were very interesting. This was a field that I'd never looked at. And in the beginning, I thought it was, you know, sort of stuff that's... Engineers, it was open and shut. You'd just write a textbook, and engineers had written the textbook, But I found that the theory was challenging, and inventing algorithms for things was challenging. So I enjoyed the subject just for its intrinsic interest. Also, and I should be quite frank about this, in the late 80s, well, in the mid to late 80s,

15:00 There were a number of things happening in Britain and in Cardiff that made it seem a wise, let's say, a wise career move to be associated with this project. Funding was getting tight. That is, I'd always had reasonable success in getting post-doctoral funding and so on. But it looked like that might get harder in the mid to late 80s. The spending cuts imposed by the Thatcher government were starting to bite. But specifically in Cardiff, the university was on the verge of going bankrupt. and there had been mistakes made by the university administration on spending too freely and then they found that the money they were getting from the government was suddenly being counted and they were being told they couldn't do what they were doing. They built up serious debts. Ultimately, what happened was there was a merger. Off we go again. Okay. Ultimately, there was a merger between two institutions, two universities in Cardiff, to form the present institution that's there. And they got their fiscal policy right, they got their accounting correct, and they were given loans and other concessions in order to get themselves back on the feet. but about a third of the staff at my old university lost their jobs in this process one way or another. Nobody was made redundant but people were encouraged to leave or encouraged to take early retirement and so on. And there were threats to close departments specifically my own because we had a small number of students and although our research was good we were not so well funded we weren't taking a lot of research grants and it seemed to me that it would be wise just from that point of view to propose rather than to stay as a theorist

17:30 to move into an area where I could get research funding from the research councils for people associated with an experimental project and get money that way, a large amount of money that way. And so that was in the back of my mind, that it was a way of securing my group and my research in Cardiff against the vicissitudes of what looked like a very, very uncertain situation at that time. Another side of your research, I know, in recent years has been a miracle of relativity work. Was that something that was also inspired by the idea of gravitational ray detection in some sense? Initially, no. Initially, it arose in a conversation, my involvement arose in a conversation that I had at Grecanog. I had been in the habit of organizing meetings roughly once every two years at a meeting house that the University of Wales owns in mid-Wales called Grecanog. And at one of these meetings, Chris Clark of Southampton and John Stewart of Cambridge and I were talking just about research in theoretical relativity. And we thought, well, numerical methods were going to become important for understanding theoretical relativity. they were going to supplement and maybe to some extent supplant the analytic methods that people had used up until fairly recently and computers, it was clear computers were getting better and better and bigger and bigger and that they were beginning to get to the point where it was reasonable to think about doing realistic

20:00 simulations in general activity. So the three of us put in a joint research grant SRC or SCRC, I can't remember what it was by that time. And we got money for some workstation equipment and for some postdocs different places at the three institutions as part of a collaboration to start working together on these things. And that was how my involvement started. Again, I wasn't motivated from the point of view of gravitational waves. I was motivated from the point of view of wanting to have a new tool for exploring solutions in general relativity, finding out what really happened in dynamical systems that were way beyond anything that could be done analytically. I didn't at the time think that black holes, and particularly binary black holes, would be one of the most interesting initial sources for gravitational waves. That wasn't part of my thinking at all. The binary black hole was simply the simplest system we could think of doing. It didn't have any fluids in it and so on. Initially, we thought in Cardiff that we wouldn't do binary black holes, that we would do maybe gravitational collapse and hydrodynamics, non-spherical gravitational collapse. But after studying the problem for about a year, I think we fixed on the direction that we took after that, which was to work hard on the binary black hole problem. So in some sense, the gravitational waves are as much a tool to study interesting dynamical systems from the relativistic astrophysic view point as the numerical things, and they both sort of feed in the same direction. But again, to be honest, in the sense of survival and keeping a research group going in this atmosphere of uncertainty that I was talking about when the university was just getting back on its feet,

22:30 And also, when there was continuing pressure on funding, and we had not got funding for the one-kilometer detector, and we were then putting together a proposal in 89 for the three-kilometer detector with the Germans. It seemed to me that by that time I had all my eggs in one basket, which was the gravitational wave detector. data analysis basket and I could see that this was a project that was going to have to struggle to get funding and might not so it seemed a useful diversification to put to get money for doing numerical relativity. Initially that money came from a totally different place in the funding setup, not the people who were the astronomers funding the gravitational wave detector, but it came from the mathematics Committee, which was a totally different section of the Funding Council. So it was a diversification, again, from the point of view that if you were getting money from two different places, you were less likely to be totally wiped out if things went against you. Fortunately, it was successful at both fronts. Fortunately, both things got funded and the research got renewed. And eventually we combined the two, and we transferred from mathematics into astronomy the work that we were funding on gravitational waves, on numerical relativity. So it actually was officially moved from the funding point of view from one research area to another? Yeah, gradually, that's right. So certainly also, I think, in America, where, for instance, the Binary Black Hole Grand Challenge Alliance, but I think it's justification to go as well. This is going to be very useful, hopefully, for gravitational wave astronomy. Where do you, in being interested as you are in both the data analysis and the numerical relativity side, where do you think the numerical relativity stands at the moment in terms of the usefulness that it can produce towards detection? in the long run it will be very

25:00 important in the short run there are some interesting in the short run meaning before they're able to solve and predict detailed waveforms of two merging black holes I believe that that's still some years away, partly because of computers, but partly because some of the algorithmic problems, boundary conditions and so on, have proved to be very difficult. But I believe that they should be able to produce intermediate results, and this is one thing I've been trying to focus people's attention on here. not doing the problem as well as you think you want to do it, but doing an approximation to the problem. Not working just on your toolkit in order to make sure that you can assemble the thing very well in the end, but putting what you have together now and trying to get some intermediate results. Because it does seem that even a limited amount of information could prove useful for enhancing the sensitivity of detectives. Whether the numerical relativity will do that or whether we'll do that just by guessing what waveforms we might get and just filtering for a few of the guesses, I don't know. But it's very hard to predict that because we also don't know how many of these black holes there are going to be in our initial data sets. So it could be very important early on, or it could be that there are not so many black holes. We have to have very detailed information from the marker relativity before we sort them out. If black holes are abundant, binary black holes are abundant, we'll probably detect them before we have really good simulations in the marker relativity. But even then, we need really good simulations in the marker relativity to infer what's really going on from the signal. So you might be able to say that looks like two black holes spiraling together. But to say that they really are black holes as described by general relativity, they really are the Kerr metric,

27:30 that's going to require a very good numerical simulation. that with the observations. So, if the, if the sort of signal to noise that you're liable to get based on, say, the end rates, as you say, is not very high, then even to identify the sources, you're probably, the situation would probably be that you're still likely to need. You'll need the numericality. But, on the other hand, if, as you say, you have good event rates and the signal to noise is much higher, then the relationship would be maybe much more like it was with the binary pulsar data in the early 80s, where now you sort of have quite a good experimental result, and the question is, can the theory show that this is a prediction of relativity? That's right, yeah. Talking to Alberto Vecchio early on today, he mentioned that one of the focuses for geo-death analysis would be pulsars, individual pulsars, if you're starting off and you're the only detector on the air then it's maybe not so good looking for coincidences with burst sources but to look for a periodic source. what are some of the particular challenges towards looking for that kind of source? What I'm sort of more familiar with is the burst sources. That's right. Well, the birth sources were, I think, what got these projects approved. It was really, in my view, it was Kip Thorne's realisation before any of the rest of us who were working closely in this field, anyway, that binary neutron stars, as the coalescing binaries, were not only detectable at interesting rates, but could convey interesting information. That understanding, I think, was the key step in getting these projects approved, because then we had something we could point to with a sensitivity limit that seemed achievable,

30:00 that would almost guarantee us regular deaths. And that broke the conservatism of the funding agencies. They didn't want to spend a lot of money on something that could turn up nothing. They wanted to be sure of getting at least something. And so I think that was the crucial thing. So the birth sources, and particularly the co-lossine binaries, guided the early development. And that's why the detectors were built. so we never really worried about the first phase I mean obviously when you build detectors there's always one detector that's ready before any others and what do you do with it but that's the situation that might face GEO and in fact it might face it for a year or some comparable period of time before LIGO starts taking data there will be a detector in Japan, TAMA idea how strongly they're going to try to push the sensitivity of TAMA. It'll be operating on a 300 meter baseline, but whether they will get it to be as sensitive as GEO or close to that, so that you could do useful coincidence searches for bursts, I don't know. So at the moment, we're thinking, well, GEO's function for the first year, let's say, might be to look for things that it can identify without running simultaneously with another detector. The key thing about continuous waves from neutron stars is that intrinsically the neutron star is giving off a very narrow signal, a very well-defined frequency. That frequency may drift with time, the star may spin down and so on. but it's a very sharply defined frequency. And the motion of the detector as the Earth moves around changes that, because the Doppler effect changes that. So you have a signature that you can look for, which is a particular pattern of changing frequencies. And it seems very unlikely to me that that will be mimicked by something in the experiment,

32:30 some terrestrial form of interference. So it seems to me that in looking for a signal like this, you're more likely to have confidence that you understand the detector noise. You're more likely to be fighting against truly random noise and not against systematic effects or interference. When you look for bursts, the reason for doing coincidence is that you can never be sure that a burst event in your detector didn't come from something you didn't understand that just occurs once in a while and you can't understand everything. So the way to be confident of that is that it occurs at the same time or at a correlated time somewhere else in another detector. But if you're looking for a continuous source, and you can only find it by filtering the data for a signal that is moving around in frequency at exactly the pattern you expect, because the Earth's very complicated motion, it, I think, gives you more confidence that when you see a strong signal, it's not interference. And in addition to that, if you're seeing a signal that lasts for a year, it'll last for another couple of years, too. So when LIGO starts operating or GRIGO starts operating, it should be there. So if you really want confirmation that you've seen it, you can get that later on. You don't need a coincidence experiment at the same time. Somebody else can replicate your result rather than... Yeah. So for that reason, we're putting a lot of effort into developing the data analysis methods for these sources. LIGO has a different perspective. They will start with two detectors operating at the same time, and they plan to do coincidence runs between those two detectors, regardless of whether the rest of us are on the air or not. They can schedule it to do that. So they're putting a lot of emphasis on the birth sources. So between the two of us and between the activities in Virgo as well, and hopefully we'll discover the right way to approach all of these things and won't miss anything. So while the motion of the Earth and so on that you have to deal with and the vagaries of the motion of our system and whatever the other system does

35:00 present a technical challenge from the point of view of the data analysis, is they provide you with a clear sign that the object you're looking at is extraterrestrial. Yes, that's right. Extra from outside the sun. That's right. What's more, the pattern depends exactly where the source is located on the sky. So something at one place and something at another place has a different pattern. So it gives the detector very good directional capability. So if you can find it, you know where it is. On the other hand, it means that you have to look for not just one pattern, but a huge number of different patterns, and that's why we have a big effort now to find efficient algorithms to do this. So are you likely to be limited by available computing power? Definitely, yeah. This is an aspect of gravitational wave detectors that was never contemplated in the beginning, but actually the sensitivity for this kind of thing, the sensitivity for, I should say for doing a search over large areas of the sky for systems that you don't have any information about that is neutron stars that are not pulsars not known as radio objects, not known as x-ray objects just old, quiet, whatever, I don't know that problem of searching and discovering it in gravitational waves is going to be limited by computer power So do you feel that you're likely to be able to come up with sufficiently efficient algorithms to permit all sky coverage, or that you'll maybe be forced to sort of pick your spot for one reason? I believe we'll come close. Clearly there are favorable spots, and you'll want to do as well as you can on favorable spots. And there are candidate locations, candidate objects, that you're going to look at even if you don't know what frequency, let's say, to look at or something, but they're probably good candidates for gravitational wave sources. But in terms of a search, I believe we'll get to the point where we can do a search, let's say, down to the theoretical achievable sensitivity

37:30 of the detector over a couple of months, anyway, provided we can get a big enough computer. But a computer that's achievable, not something on a wish list, a computer that in two or three years you will be able to get, the so-called teraflop computer. If we had a teraflop computer, and could use it for three months, I believe we could do a three-month search. If we had complete use of that computer for three months, we could reduce three months' worth of data and reach the theoretical sensitivity limits of a detector over three months. That means given the noise, the intrinsic noise of the detector, and allowing that the signal has to be quite strong anyway because you're doing effectively so many searches, you're looking at so many different locations in the sky, that there's a certain probability just on random noise that you're going to get a result. And therefore you have to set your threshold higher than that probability, high enough that the probability of getting it by pure noise is very small. but I believe we'll be so whatever that threshold is I believe we'll be able to reach that sensitivity over three months two or three months but it certainly is it's not clear where we're going to get a teraflop computer from yet there's also many other things There are many targeted sources, all the known pulsars, for instance, but also X-ray binaries, certain kinds of giant stars, that might be sources of gravitational waves. And we have to make sure that our search algorithms for those are good because it'll be easier to, let's say, achieve a one-year sensitivity on such things, and that will greatly improve our chances of detecting them. So, you know, we have, within the class of neutron star sources, we have a lot of possibilities.

40:00 What would be the, what do you think is likely to be the model for the use of the data that comes from detectors like GEO, will it at some point develop into a situation where the data will be available and different people will go and decide to look for different things with it, perhaps in the way that it is sometimes, well, somewhat on the model of, say, current astronomy, or will it be? This is much discussed within the detector groups. And in Europe and in America, models that are currently being operated are very different. In Europe, no one that I know of in the German or British astronomy communities, let's say, has said to us that they really want this data. We're not under pressure from the funding authorities to make data public. we will in fact I think we promised the British that after a suitable time lag there will be a publicly available data store so we have proprietary access for a certain amount of time to do our data analysis and after that it becomes available to astronomers but we're not under a lot of pressure to widen out this user community yet, I think in Europe that will change, I think get more prominent, the funding agencies will be asking for us to think about these issues. In America, they're already thinking about, although they haven't made real hard decisions, but they are already thinking about how they can make products, data products, available to a wider community. And they're signing up collaborators and getting MOUs and so on with different groups on a quid pro quo basis. You know, if you develop some data analysis theory or software for us, then you can participate in something. My own feeling is that it's actually not going to be in great demand. Let's say, particularly for the geo data,

42:30 from all we know about astronomy, gravitational waves are going to be rare and weak we'll do the best job we can searching through the geodata for everything we can think of when we supposing we find, let's say events, supposing, I mean if there's a Pulsar, obviously, we have to make data available to people to justify that we found it and so on. But if you're going to do a search for continuous waves from a particular source, it's much easier to propose to the geocollaboration that they should do an analysis for that particular source than to get the masses of data that would be required just to filter at once for a particular location in the sky. So I think we may run a data analysis mode in which we make our computers available in a responsive kind of way to users. So we don't just ship the data out to them, but we make computing cycles available to them. They can see what they want to look for. another thing is burst sources and there of course it involves all the different gravitational-like collaborations and how they want to make data public and share data with and they all have sort of different perspectives so it's quite a difficult problem but eventually I'm sure that data will be available let's say a few seconds either side of a coalescing binary signal so people can study it to what they want that's relatively small amounts of data that would be easy to put in an archive so people can get more problematic is is again if somebody says they have a new kind of burst source that they want to look for and they want to filter a year's worth of data we do that for them, take their filter and apply it, or we give them the data, that's not so easy to work out yet, how that would go. So, there's very large data sets, and

45:00 most of it is DOIs, and it's not like astronomy, where most of the data you get in big surveys is potentially of interest to people and you want to make it available to them. Most of the data we get is pure noise. And so should you expect to have to ship out large collections of tapes to people or should you just set up a system where people can ask you to do the searching? Certainly in situations where you, for instance, the teraflop computer, as it were, then the access to the computing would be as big an issue as the access to the data data. Yeah, that's right. The access to the computer, that's the sense of what I'm saying. The data handling is more difficult than the data analysis, so that it's probably easier to do the data analysis where the data is rather than shipping the data around to other people if you want a year's worth of data to work on. I forgot something that I was going to ask, but just as a matter of interest, while back to the subject of the birth, in the last year or two, the last couple of years, there's been a certain kind of debate over a numerical relativity result due to Wilson and Matthews, where they produced one of the first largely relativistic hydrodynamic neutron star binary merger codes and derived the rather unexpected result that the neutron stars appeared to be collapsing, tending to collapse to black holes before they actually went into merger. And of course, by and large, it would be safe to say the result is disbelief by most theories. But I was curious as to how robust you feel to any major theoretical surprises that arise in numerical relativity work? I mean, you know, supposing this or a somewhat unexpected result were to come out of future numerical work, is there any real way in which that could throw off the data analysis? Well, it's certainly, it's vulnerable, because to get the sensitivity, the interesting sensitivity,

47:30 is you really have to have a filter which matches the signal quite well. What's important is it has to match the oscillations in the signal, the phase of the signal. It doesn't really matter if you get the amplitude changes from one cycle to the next rolling bit, but you mustn't get out of phase with the signal. So you're certainly vulnerable to effects that would change orbits and change orbital frequencies. and things that you don't expect. It's hard to know, it's hard to know, it's hard to anticipate what those would be. But, you know, there's always unexpected things happening in physics. So, yes, we could lose a signal of noise if something really unexpected happens. The Wilson-Matthews thing wouldn't, because if a star collapses, What they found was the neutron stars collapsed to black holes. The orbit is essentially unchanged, and the masses are not changed very much, and so you don't, in fact, wouldn't lose signal to that particular effect. But it's possible for numerical relativity that would tell us that we've left something out of our calculations. I doubt very much if coalescing binaries would be much affected because the orbit, the sensitivity to coalescing binaries is dominated by the in-spiral phase of the orbit what happens when the stars get really close together Wilson and Matthews, whether they're right or wrong that was an effect that happened when they were nearly merging in the very strong field What happens at that point is not relevant to detecting the signal. So the radiation for detecting the signal comes from the orbit, which is a phase that is really well understood. I don't think there could be any reasonable chance of surprises because we would see them already in the binary pulsar system. If that was substantial, we're going to see any surprises there. So I think the detecting of coalescing binaries wouldn't be. But to the extent that detecting black holes depends upon the merger phase more than the inspiral phase, that could be an issue.

50:00 There could be other things to do with predictions about supernovae and so on that could come unstuck. I think we're very robust against changes or missing out things in theory for the cause of binary neutron stars. I guess the other question I was going to ask while we were discussing the question of availability of data and so on is at the present time, you have data available from prototype detectors such as ones at Glasgow and Caltech. And obviously it's of interest, I suppose, when one's doing data analysis work, theory of data analysis, to look at real data, how much has data been available from these projects? I suppose in your case, you're essentially associated with GEO that access to the data may not present the problem. Yeah, access to data is not a problem for us. One of the things that was very good that came out of our application for the three-kilometer detector with the Germans, the original GEO application, was that the funding agencies asked the two groups to do a 100-hour, a long coincidence run. So for 100 hours, they took data in Garching and in Glasgow at the same time. So we were able to get a lot of experience analyzing that data, and that shaped a lot of our plans now for data analysis for Geo. And we have since had smaller packets of data, smaller stretches of data from Glasgow. Getting LIGO data is... We haven't tried to get LIGO data through signing MOUs. We might get LIGO data through other routes. We have here data from the Roam detector. This is an aspect that, again, is part of the pulsar problem. If you're looking for neutron stars, the problems of looking for neutron stars with bar detectors are very similar to those with interferometers because the spreading out of the signal is still very even.

52:30 With all the frequency shifts, it's very narrow compared to the bandwidth of a bar detector. So bar detectors are broadband with respect to pulsar or neutron star signals. They just happen to have a bandwidth in the wrong place. It's still a very narrow part of the spectrum that they're sampling if you're looking for something that's unknown. And it's too high in frequency for what we think we understand about neutron stars. But nevertheless, groups, the Rome group and the Louisiana group, are very interested in doing studies. We have data from the Rome Group here, which we have been using for testing out ideas of how to analyze, how to do that. We have a very strong collaboration with the Rome Group. So in general, the experimentalists are fairly keen to promote the data analysis? Well, they're very keen to promote the data analysis, yeah. LIGO has a problem giving data away I think because of the NSF structure and because they're seen as a very big project there you really have to sign NMOU to get any prototype data from them if you're an ordinary user that's not because they want to discourage people it's just the way they're set up but generally the experimenters do want to see people playing with the data and getting used to the problems with it. Well, I would imagine so, at least in the sense that I noticed it. When I was at Caltech, you know, as a theorist, we always had a tendency to be talking about Gaussian noise. Yeah, that's right. And the experimenters would always be talking about non-Gaussian noise. That's right. And we learned from the 100-hour data that you do have to talk about non-Gaussian noise. The Gaussian noise model is a good first cut and a good way to help you get on with theoretical calculations and figure out what's the right way to move. But in the end, when you want to assess the achievable sensitivity and achievable signal noise, the non-Gaussian noise is very important.

55:00 And, you know, removing some of the work we've done here is directed at removing some of the sources of non-Gaussian noise so that you clean up the data and get more Gaussian. It's a very important problem. Have the experimentless formed much of an opinion at this stage about the theorist approach to the data analysis issues? Do they have any... large, my impression is that they're very positive. They see how difficult the problem is. The group, the project leaders are uniformly encouraging to people who want to study this problem and develop methods for data analysis. Like all experimenters, they believe that theoreticians sometimes get carried away too big an idealization and they often say that but nevertheless they feel they can see the difficulty of the problem they can see that it's important that theorists study it carefully there are some people who believe that it's being overanalyzed that we're in danger of overanalysis that we should just accept that if we can't see a gravitational wave sticking up like a sore thumb in the data stream that we haven't, you know, we shouldn't do very much more than that. But I think that's a minority of you. So some of the experimentalists would be inclined to feel that we're just going to look for a signal that's obviously... Really big, yeah, the first detector. I mean, fine, you filter, and that's the working way. But if you really want to announce the first discovery, I mean, this is a real problem for the field, is how to be confident the first time you've seen a gravitational wave. And it's a real question if you've had to use a very delicately constructed filter to find something that was only one-tenth of the noise amplitude in the unfiltered data stream. Do you believe it? Do you as a theorist think it's... How likely do you think as a theorist that there is going to be a spike sticking up out of the data? I don't think it's going to be that likely.

57:30 Unfortunately, I don't. I think we'll have to work pretty hard to get the first detections. Well, I was going to ask about the Institute here, because having been here for a few days, I'm impressed at the sheer size of the group here and the activity that's going on. I imagine that we can expect it to play an increasingly larger role in the sort of general activity community. I certainly hope so, yeah. It's sort of on a scale that I don't think has really been matched before. No, that's right. You have, I guess, the structure with three different sections, mathematical, physics, and quantum gravity being the two sections. are those sections I was hoping to talk to Jorgen Evans but I don't think he's here yeah that's right and Nikolai is also not here I was wondering if you have any sense in which the people involved in those sections see themselves as having a stake in the success of the gravitational wave experiments I think in the quantum gravity section they probably wouldn't think they have a big stake in it, but being fundamental physicists by nature, they're very interested in observing the graviton, if you like, in seeing this dynamical aspect. And there is always the chance that observations will turn up some information about particle physics, that we'll see extra gravitational fields or we'll see something that wasn't expected. So they follow it in a supportive way, but I don't think they have a stake in it, very much. The mathematical relevantists, on the other hand, have very closer interest and involvement. A number of them are actively working with Ed on the numerical relativity side, because they see that numerical relativity can help them solve problems that they've always been interested in, and the analytic techniques cannot solve. And they're also, I think, aware of the fact that if gravitational waves are detected this last

1:00:00 on not-quite-proven or something frontier of Einstein's predictions, that it has the potential for not only settling questions that are by and large settled now anyway, but the nature of gravitational radiation in relativity, but just giving relativity extra status, I think, in the world of physics. is something that they would benefit from. You were mentioning that the mathematical physics is interested in the numerical relativity side. When I spoke to Jeff Winokur on a couple of occasions, one thing he said, at the time I was interested in the quadruple formula controversy. And he was saying that well, maybe numerical relativity would introduce such a revolution in relativity as a whole that you wouldn't speak anymore in terms of constructs like the quadruple form that these would doubt, and when I was talking to Ed Seidel earlier in the week I think he, if I remember correctly, he quoted you as saying well maybe for instance Scry as another example of an existing construct in general relativity has no interest for numerical relativity so I was interested in asking if you have any sense of whether numerical relativity, just purely on the theory side, is likely to completely revolutionize the field in such a way and to eclipse older methods of doing that too? Yeah, I think it won't totally eclipse older methods, but I think it will definitely revolutionize things because it'll introduce a robustness into understanding relativity. It's in any case an approximation. But if you can solve the system one way, you can always tweak the parameters a bit in the beginning and see how it changes and make sure that the understanding you're getting out from looking at these things is robust. You're not being misled by coordinate effects or by other things that people are worried about in analytic methods.

1:02:30 is always that if you could solve a problem, often you could only solve it one way and you were always a little bit worried that gauge effects were misleading you or whatever. The scry is a very good example because it was, as a heuristic, as a way of thinking, it was extremely important for people that they could get away, really geometrical which didn't depend upon coordinate systems and described gravitational radiation and where it went it was important in laying to rest the question of is gravitational radiation real it was part of that whole progression in the 50s and 60s but as a calculation, as a tool for for making useful approximations to systems that give off radiation or that detect radiation for understanding the interaction, if you like, between a source and a detector. I don't believe that SCRI has a central role to play in that. The detector is very much closer to the source than scry is. Scry is infinitely far away. But even given that you'd like to make an approximation that it's not infinitely far away, scry is really, really far away from some of the effects that are delicate and are described there. So I don't think scry plays an important role in these real issues. And I think there has been too much of a demand by theorists, once they got comfortable with SCRI, that all theory of radiation and science should be described in terms of SCRI. I think that's not a helpful point, do you? Would I be right in thinking that your work in the early 80s, I guess, with futanity

1:05:00 I think that sort of more statistical approach to the radiation reaction problem was sort of one effort to get around it that's right, yeah it was an effort to introduce robustness for instance and to introduce a different way of getting an hour of time to get retarded radiation it was also enabled us to do something which was in some sense local, that is we didn't have to introduce a scribe, we didn't really have to far away from the system to say what was happening. So, yeah, that was very much part of that. So, even at that stage, was it partly inspired by an effort to answer, as you say, these old questions of, well, does gravitational radiation exist the way we think it does? Or was it? No, I regarded those as settled. I mean, like I said, scry was very important in settling those issues. But once you've agreed that, then you should look again at what are the appropriate methods for attacking a problem. I wasn't even particularly motivated by wanting to get rid of scry. What I was motivated by was wanting to get rid of retarded green functions. I wanted to have a way of looking at radiation reaction and the post-Newtonian limit and all of that, which didn't rely on this concept from linear field theory of a green function. Because in nonlinear relativity, the only way you can introduce that is by iterating and using it in different ways. Now, I mean, mathematically, you still, one way or another, you still wind up doing very similar operations because I was developing the post-Newtonian approximation. I certainly wasn't the first to do it, and I was going to get equations that were very similar to everybody else. But I didn't want to... I wanted to get away from a controversy which was current at the time of how do you define the green function that you're going to use? How do you define what you mean by retarded radiation or an isolated system? And there were several different prescriptions, over them, and I felt that they were arguing over things which had no physical importance, that is, the nature of radiation and scry, because I felt that you had sources, you had

1:07:30 detectors, and neither of them was that far away that these issues had to be important. I wasn't alone in that. Martin Walker, who worked in the Munich group in the 70s, early 80s did a calculation once in which he said if I look at the deviation between a flat space kind of light cone in some suitable coordinate system, an artificial light cone and the real light cone, they diverge by sort of logarithmic terms how far away do I have to get before that divergence is let's say one gravitational wavelength of a system waves, it turned out that it was immensely further away than the detector was. So you go, you know, the gravitational waves are going out, and they hit the detector from any reasonable source of detector separation. They hit the detector long before there's a worry, you have to worry about the difference between light comes. And yet, the dominant issue in the theory of the post-Lutonian structure of relativity at one point was, how do you icons? What's the suitable definition? So I wanted to get away from that. I wanted to introduce a completely different point of view. So I needed a different definition of retardation. I had to get causality from someplace else instead of saying I don't have incoming radiation or I have something somewhere. I realized that, and it was to me an experiment that was very satisfying, that I could get retardation from random initial conditions. That I could somehow make it seem more like the causality in the second law of thermodynamics, that it happens because things are so complex that it's bound to happen, rather than that it's built in by God who likes retarded green functions instead of advanced green functions. So I found that philosophically satisfying, but it was also just a totally different and more local way of introducing causality and not worrying about Scry. So really my dislike of Scry came from then

1:10:00 having to fight with people who said, well, you have to do it from Scry. This is the right way to do gravitational radiation. There's only one way to do it. You have to say what's happening at Scry. And I was amazed at that point of view, because it seemed to me that really you wanted to describe the interaction of a source and a detector, and scry was a mathematical construct that would or would not be useful in certain circumstances. But a lot of people were taking a very orthodox line that if you didn't say what was happening at scry, you were somehow confusing the issue. So my dislike of scry stemmed from that point. was one of the reasons for wanting to get away from the Green function that if you were interested in the issue of the hour of time, then with the Green function you were going to start off with, as you were saying, a linear superposition of an advanced and retired Green function, which would be out of place in a nominee of theory. Was that part of it? Yeah, this was the problem that in electrodynamics, which is a linear theory, at least it makes sense to say that it makes sense it's a philosophical issue really but it makes sense to say that this electron is governed purely by retarded fields because that doesn't depend upon what that other electron is doing you could say it for all electrons and then all the fields superpose into something that's fully retarded but in a non-linear system like general relativity this bit of matter here and that bit of matter over there, you don't know what fully retarded means or no incoming radiation means unless you know what they've done together and how the space-time was modified by other bits of radiation. You have to solve the whole thing at once. The definition isn't local at all. And given that we live in a cosmology and not in an isolated, Our stars are not in asymptotically flat space-time anyway. They are in a cosmology which doesn't have that kind of scry. In the past, it has a cosmological singularity, as far as we know. It seemed rather a strong idealization to have to make that you had to impose this analog of retarded green functions

1:12:30 to solve a problem when you knew that in any case in the real world it wasn't the right lecture So I presume that even though I think some of this work was going on while the quadrupleformer controversy was still in some degree of activity I presume that it wasn't so much your purpose wasn't so much to provide one more derivation of the quadrupleformer No, in fact we never pushed it far. We never pushed it nearly as far as D'Amour and Blanchet had done, although it would be interesting to me to do that if I had somebody who was willing to take up the immense amount of calculation that would be required. I would like to see it done. It was more an attempt to show that you could reformulate the problem to get away from some of these conceptual difficulties and expect to get essentially the same results as you'd get from these other approaches. It was an attempt to show the robustness of the post-Newtonian approximations. But I think it's very interesting. We haven't carried it out to high order and now Blanchet is getting results with tails of tails and things like that in the radiation, it would be really very interesting to see how robust those things are to a change in the paradigm, a change in the definition of what you mean by isolated and so on, because some of those things could change at a very high order. So I think it's an interesting question, but we don't, we're not pursuing it at the moment. Yeah, I should probably let you go. Yeah, I thought I would hear somebody knocking on my door. Let me just check and see if anybody yeah I think probably Here in a second, yeah. Coming, but I don't know why he hasn't turned up. I saw on your website that you started an electronic journal.

1:15:00 Yeah, yeah. That was interesting. That's a real passion, yeah, of mine. I was quite interested in it because I was... Here we go. Thank you.