Interview with Curt Cutler

0:00 Okay, so the tape counter has actually started, and it seems to be recording. So, today is February the 11th, and it's quarter to ten, and I'm talking with Kurt Cook. So, let me see, Kurt, where did it begin? Well, you were telling me the other day a bit about work that you're starting on now, or engaged on now, with this GRASP package and designing data analysis methods for looking at not only known, but unknown others, so I suppose maybe we could certainly tell you more about that. Okay. Well, let's just give you a background. You've probably gotten this impression. So, like, more and more, I think, a lot of theorists are turning towards data analysis right now. Experimentally, I think it's unusual in science, actually. Usually, experimentalists also write their own data analysis packages and things. Yes. But I think they're understaffed and undermanned, And they've more or less said they're expecting the theoretical community to come up with algorithms and templates and other things to help them analyze data. So, in a sense, we've become adjuncts to the experimental effort. You can see that a lot at CalTAC, certainly, with Bruce Allen and everybody writing that grasp package and people pitching in. and so here at our institute we sort of have a connection to GEO the way Bruce does with LIGO and we're trying and Cardiff also has such a connection and we're trying to help them develop data analysis tools and a lot of the work has already been done by Bruce and others at Caltech but two big topics that I think have not been tackled yet are search for pulsars that's non-axiosymmetric neutron stars and the search for things that we haven't thought of yet. And we're just beginning, well, let me talk about pulsars first. So as you probably know, the problem with pulsars is we think the signals, if they exist at all, are very weak. So one needs very long integration times of order a year or months to have enough signal to noise. But in that time, the Earth has moved a lot, it's spun around a few hundred times, and the pulsar might have slowed down considerably in that period if it's a young pulsar.

2:30 If it's a 40-year-old pulsar, then in a year it probably slows down by 2%, 2-3%, we'd guess. So one has to fit out all those effects, which makes it a very big computational problem. So it's by far the hardest computational problem is the search for pulsars, much harder than all the others. Would you be focusing on young pulsars because there's a feeling that they might be more likely to be axisymmetric, or is it just to include all possible categories? Well, for one, one doesn't want to leave any good source untapped, so there's a sense that we might have to look for those. Certainly the oldest pulsars that are fast are not very non-axi-symmetric. 1957 plus 21, the very famous millisecond pulsar, for example, is clearly not slowing down very much, so it's not radiating much gravitational radiation. The idea is that a young pulsar would be both fast and maybe ratty, still have some seismic activity going on. So the hope would be that if I had to guess what's the most likely candidate for a pulsar that one could actually see with LIGO, I would say it'd be a younger brother of the crab. So the supernova rate in our galaxy is about one every 40 years, people think. So there should be about 25 pulsars out there that are younger than the crab. That is rapidly rotating neutron stars. Because we live in a dusty galaxy. And so all the supernova that people have seen with their eyes, even with telescopes, have actually been pretty close to us over in the galaxy. They're all in our little corner. So I've been talking to astronomers. I haven't had anyone tell me that it's impossible. That there's a supernova out there that's 40 years old, and we're at the center of the galaxy. So a supernova around us is somewhere there. We don't know where to look. Yeah. Would the amount, would the rate of slowing down of young pulsars be primarily due to admission of gravitation radiation or would it be... Well, nobody knows. Nobody knows. So you can't predict it too. That's right. So the strategies then are to try and take account of, um, of, try and be able to take

5:00 account of some change in the period without presuming what it's going to be. That's right. That's right. It's a variable to be searched over. But assuming it's fairly continuous and smooth, so that you could parameterize it just by a spin-down, a p-dot, a p-double-dot, maybe a p-triple-dot. I hope that that's good enough. You had a good acronym there for the unknown objects. Oh yeah, I'm trying, yeah, with my acronym, I'm trying to make it more popular. So UCOs, Unidentified Chirping Objects. So if you talk to experimentalists, then you ask them why they're building, spending twenty years building these things. They'll tell you, what Ray Weiss will tell you, for example, is that it's not to see pulsars, it's not to see binary coalescences, maybe it's to see the stochastic background, so they want to do something really exciting. They're not too excited by these pulsars, actually. I think they would be if they discovered one. But since they haven't discovered anything, they can set their sights really high and say they want to discover something that no one's thought of before. But then there's a... Okay, so let's assume their intuition is right, and there are things out there we haven't thought of. Well, it's possible that one will have to think of them to find them. That is to say, they'll be in this intermediate region where they're not sticking up out of the noise, but they're under the noise, but they have enough cycles that you can conceivably dig them out. So then, so one has this, then it's partly a statistics problem, which we're just starting to think about. What would you call a signal? And what techniques would you can use to dig it out if you don't know exactly what it's going to look like? Presumably. Presumably one has to imagine some family of things one's calling a signal. One would call a signal, you know, something that looks like it's sort of exponentially damped, but ringing for a while with some changing frequency. And one would, if someone just threw it on the board, you'd say, yeah, I think that's a signal. But there's some big parameter space of that. So does the kind of search that we're talking about entail coming up with a basic set of parameters, choosing basically the parameters that might define We're really just starting to do it so right now in fact we're starting off this gravitational wave search just as a literature search

7:30 so we're a bunch of general relativists but this just takes a question and so we're trying to read a bit of the literature so you can imagine other people who do this sort of people who do sonar and they want to hear funny chirps coming over their sonar, or, um, well, presumably, there are people, there are, we're not the only people in the world with this undefined problem of having lots of things that are out there that we're not sure of what they are, and having sit buried somewhere in noise, and people have developed algorithms for looking at these things. It's possible we're in a a fairly unique situation in that we're looking for such slow signal-to-noise events. That is to say, you know, so we really just don't know. So right now we're trying to talk to people and talk to statisticians and give us references and start reading. And it would be nice if what I would like to do, my vision of this, is rather than start out from scratch and reinvent what probably a lot of people have been working on for 20 or 30 years, is find four, five, or three likely candidates from the literature that people use for different sorts of problems, for telecommunications or for sonar or for something. Get them to ship us their code, tweak it, so that we can apply it to gravitational wave search, and then use GRASP or something that simulates some data. and just try it out. So take these packages, have some enemy, someone who doesn't like us too much, draw on his blackboard something that kind of looks like he would think of as a gravitational wave signal, have him hide it somewhere in our data, and then see which, if any, of these techniques is any good at picking out that sort of thing, and just try them up with things new on our own. Hopefully that strategy will work, but that's going to be our initial strategy. Just a matter of curiosity, is there any sense yet that some of the people, like, say, the military is interested in sonar and radar, that people have done this, but it's still classified and that kind of thing?

10:00 There might be. We actually, I mentioned sonar for a reason, we actually had, we met with the French sonar people in Paris a few months ago in December, they had heard about LIGO somehow, or a Virgo person, John Allen Mark, contacted them somehow and they arranged a day where they told us about their sorts of problems. Most of it didn't seem very helpful to us, though. They did emphasize a sort of techniques that are called time frequency techniques, a wavelet transforms that they said are useful to them. We've heard other people say that too, so we're looking at that. We're starting to look at that. But that's about the only thing I can say we got out of it. These people wouldn't know the classified techniques. Yeah, it's possible. I think that we couldn't have met, even though it wasn't very helpful, I think we couldn't have met with similar people in the U.S. because I more stringent rules about what's classified. But there's no evidence at the moment that somebody has a work on a technique that nobody knows about. Interesting. But the wavelets have been knocking around for a while. That's right. Yes, that's right. And they have been. No one's really looked at it very seriously. It's time. It's time. It's high time. Let me see. an interesting point that you made at the beginning about how the theorists are doing this work. You were saying that this is not something the theorists are doing. Is it something that you benightened yourself doing when you started off? Oh, not at all. Not at all. And I don't want to make it my full-time occupation. But it's important. And at least the problems I weren't thinking about right now, I'm also just actually just interested in. So it seems I don't mind learning these statistical techniques and Especially now that we've got this big group here and it's getting bigger. I can hopefully share the labor out. So it would be one thing for me to try to dig up five routines and try them all myself. It might take two years. My hope is that I can do one, and Alicia can do one, and Alberto can do one, and then we can compare results. Something like that. It becomes more, in my mind, it becomes more possible. We haven't progressed very far, but that's the idea.

12:30 So, how big is the group? Oh, we've got about five people in the group now, and we're going to hire a few more. So you actually hire in the Fuse at a time. Yeah, that's right, we've got... Because you just started. Yeah. So it'll be five plus a few more. And are the five all postdocs? Are you... I gather Ed Seidel's group has some graduate students for that. That's right. We could. But there's, as I said, we're not money limited, we're desk limited, so it's, I mean, the desks count differently, actually, because postdocs are sort of one to an office, graduate students are three or four to an office, and so it's a question of space and cramming, and what's the trade-off? I'm being selfish, I personally prefer to have postdocs than graduate students. students. It takes, they're already sort of trained. So, I would prefer mostly to use our spaces for post-docs, but we might take some graduate students. There's a sense in this institution here, the Albert Einstein Institute, that we owe some things to the government who gave us all this money in these buildings, and we should have some debt. We should somehow try to repay this debt. It's not completely... One way we can sort of repay it, the idea is, is to help train German graduate students. And I'm sympathetic to that. I mean, I feel we do, to some extent, have this debt that we should be repaying. And so that would be my primary reason, actually, for taking graduate students, is to help train some Germans. Sorry. Sure. Well, I was going to ask, so how active is Bernard Schutz with the group at the moment? Oh, Bernard hopefully will be more active in the future. He comes to our meetings sometimes. He's there hovering about listening to us and he's suggesting some ideas. He's not doing any real work himself, I would say, or very little. he's still commuting every week and one of the things he's very good at is keeping his fingers on lots of pies he likes that very much and he's good at it so he has connections with all the great different experimental efforts

15:00 and the nice thing about that for all of us is it keeps us informed we know all the latest rumors what's going on inside the closed door committees I'm not sure why that helps us at all But the downside is every week he seems to be jetting off to California, or the Netherlands, or Italy, or something like this for some directors meeting for some experimental project. And then he comes back and tells us the gossip. Is he your main connection with the G600 project? Yeah, absolutely. So, probably, since I just came here, I'm in the process of formalizing my own connections with these people, so I'll probably become a member of my own standing. Right. And he's been good at fostering that as well, he sort of, he sort of puts my name forward. You should really invite Kirk Petlaren to be a member of this team. And what are your connections with Ed Sidel's group then? Is there much interplay between the numerical values and the same? Let's see, there's been a little, but not really very much. to a large extent I think they can all they need from our half of the group is a little tiny bit of problem direction as to which problem we think is most important is it important what information do you need to know from say the two collision of 250 solar mass black holes or we'd like to find out and that you can describe that you know in a few minutes then they go away and spend 10 years trying to figure it out so it's a little bit and they're not at the point where they've can actually generate much information that's extremely helpful either as yet there hasn't been a great deal of mutual influence and it's hard to imagine can be. That's my feeling. So it's a little bit like the situation you've developed, you've guessed from reading, say, Hughes Flanagan, where you've sort of got a hole in the middle. That's exactly right. That's where we're going to hope that you're talking. That's right. That's right. That's all I have to know, more or less. But that's a useful paper because it points out. I was thinking of that in my answer because it says, well, if you can't give us the whole waveform, maybe you can tell us how long it lasts or what

17:30 frequency band it's in. So I guess an interesting question to ask is, in your opinion, how more difficult will it make the data analysis effort if, you know, the detectors go online for a while and there's not really still been any significant input from the or the relativity side. How much more difficult does that make just seeing anything or seeing things but getting a lot of useful information out of it? Well, one just doesn't know because one doesn't know exactly what's out there. It could be important. It could end up as if the strongest thing out there is the, say, collisions of 230 or 40 solar mass black holes. If that really is the most optimal, then that's very important information that they could provide. And if the detectors don't progress as fast as people want, then it's doubly important. So you can imagine scenarios where it's very important. But on the other hand, it could be that there are other sources out there that keep us busy and it's not so important. So basically, if the sort of source that they can tell you something about, such as what you described, are the most important, and if they're at a level where the signal can noise is relatively low, then obviously that would be a crunch situation, but other than that, you could get one. That's right. But there's a reasonable probability, I would say, you know, 20%, but that would be the situation. One of the questions that I mentioned yesterday, I was interested in asking, and it sort of reflected in what you're talking about, the sort of work that you're doing, and how it's a little bit unusual for what you're doing. when the last three minutes paper came out a few years ago now I guess the impression that I would have had was that it was trying to set an agenda or at least point out to people

20:00 in the field issues that were going to have to be addressed by theorists in order to make a success of projects like LIGO so I was interested in asking you as a major what sort of influence you thought it had, and how much the attitude it expressed has been borne by it. So that's a good experience. Well, yeah, fortunately I asked you that yesterday, so I had a little time to think about it. I'm not sure about that paper all by itself, but there was a whole bunch of papers that came out of Caltech exactly the same time, or roughly the same time, that was like perhaps maybe the first or one of the first. And I think together they did have an influence. So together they highlighted a bunch of problems that other people then did jump into. So around that time, Eric and I, mostly Eric, advanced the field, the idea of using a point particle around Schwarzschild, and that's what you worked on as well, to get high order post-Newtonian results in the limit of large mass ratio. sort of my idea, and then Eric really took the ball and ran with it. And then after he ran with it for a while, the Japanese group really ran with it even more. And so it's an idea that then left Caltech and was done most of the work, ended up being done in Japan. We pointed out, I think, the things we said in those papers were also lit a real fire They were underneath people who do post-Newtonian calculations, and I think there was a huge amount of progress after that in, let's say, mostly coming out of Dumb War and Cliff Wheels groups. And I think, I think knowing that these were important, they probably didn't quite understand before. They didn't know how important it was to get these high order corrections to the phase of the waveform. And I think it really, having somebody say, this is important, was a real spur to them. And I think that little area, that field, post-Newtonian calculations, advanced a lot farther a few years than it had done in many, many years before, just because it suddenly seemed important and that people were paying attention and saying, please tell me your results, as opposed to just publishing them and not having much reaction. Let's see, there's been a lot of work on coalescences of neutron stars, and I think we started

22:30 talking about that, and you do too as well, what sort of information you could get from the neutron, about neutron star radii, etc. And now there's been lots of people are simulating those things numerically. Various approximations. So I think a lot of work that was done at taken on and carried a lot further elsewhere, and I think it was those set of papers that came out around that time that was Spirita. What are the connections between the groups that are interested in that, and I was like the content group and here in Cardiff, and I noticed Kip was going to come here twice in the next quarter. Yeah. Um, we talk a bit on the phone, but that's about it. I mean, mostly, most of the work honestly has been done at Caltech and mostly by Bruce Allen lately. And some work has been done by Satsa Prakash in Cardiff, though Bruce has really done the most. And the rest of us are just, are jumping on somewhat late, I would say. Um, we've scoped out the problem I wrote a paper with Pat Brady and other people. So it's more or less, for us, it was just identifying things that still hadn't been done at all, but that were important, and were things we wanted to work on as well. So I don't think we'll need a very close connection with Caltech to work on the problems I told you about. They haven't been done. I don't think there's any plan to do them. they're very happy if we do them. It's just important that if they're working on them simultaneously they would probably want to call each other just to know. So the situation would seem to be quite different from the numerical people where there have been all these attempts to get grand alliances of different groups to work together with various reasons of success. That's right. People just work on different problems that are perceived to be part of the general. That's right. Well, there's enough problems to go around. and then presumably people talk to each other I mean for instance a lot of the bar groups have been also looking for pulsars so we have a postdoc here Maria Alessandra Papa who's sort of been doing some writing software for that so some of the ideas that are used there, they're different kind of instruments some of the ideas can maybe

25:00 there can be used by us There's some cross-fertilization in talking, but not too much direct collaboration. Yeah. Just talk about direct collaboration. The data analysis meetings and other connections that have been organized, are they useful for coordinating any of the effort? Or is it a case of just finding out who's doing what and so knowing what's left? I think, well, for me, they've mostly been good at providing inspiration or to remind myself that, yeah, someone should work on that. Or, yeah, I realized, boy, there's a big hole. Someone should fill that hole. So, for instance, Bruce, in the last gravity wave analysis workshop in Paris, had this large presentation where the whole Caltech group marched on and described their search for coalescing binaries. I thought, yeah, we could do that, too. There's no reason why a bunch of us couldn't collaborate together to search for pulsars. It's a harder problem, unfortunately. But we have a lot of people. If we all devoted a couple of months to it, maybe we could make a contribution, too. There's no reason why we shouldn't try. And then just as people talk about these things, one gets ideas that this is important, or that's important. Or, yes, this idea that not only is a search for UCOs important for finding astronomical things, but they can be useful, for instance, for finding funny chirps in the detector. So, presumably, the things you would find any such technique is going to find two sorts of things. Real astronomical things and funny things in the detector that go bing. And, uh, And on the one side, that's terrible. It's confusion. On the other hand, it's really helpful to the experimentalists. It helps them understand what's going on in the detector. Yeah, so the idea that, for instance, that was emphasized at this last conference, that any techniques used for searching for UCOs would also be good for looking for detector diagnostics, and maybe even better for looking at detector diagnostics. And that's not a bad thing. That's a good thing. part to it. So you'd be picking out... Funny ringing, that was due. Yeah, that's right.

27:30 Speaking of that kind of noise, one of the impressions I used to get at meetings was that, to a limited extent, experimentalists and theorists were talking about different kinds of noise, because theorists would be coming along and just talking about kind in the experiment, is that something that started changing a bit, now that you're looking for this kind of object? It's becoming more and more true that theorists are realizing that non-guessing is often the problem. It's one of the useful things about actually having real data to look at, is you see that it's actually a real mess, and that that's part of the problem. Yeah, so people's view of the data becomes more and more realistic if they have to actually sit down and write code, because it's very easy to talk about it just being this function that, you know, SOT with Gaussian noise properties, but then you get the real data that has gaps in it, and then the levels of noise change with time, and you have to somehow deal with it without having quite all the theoretical tools for doing that. Where do you get your real data from from Geo? Oh, so we haven't gotten any yet. Well, I know we do have some. We have some, sorry, I didn't say that. That's not true. We have real data from Geo. That's right. And we have some real data from the bar groups. And we're not sure yet. We've asked for real data from Caltech, just a minute of it, so we can start using the GRAS manual. That's how we haven't really started using grass very much yet. We're just starting to. But to even get a minute's worth, they want us to write a large MOU with lots of legal sounding things in it, all of which sound like they're completely to the advantage of LIGO and completely to the disadvantage of us. So we might just try to simulate our own data for a while until Geo can give us some in the right format. I'm not sure if we're going to end up playing with LIGO data for a while until there's some big understanding. The, uh, oh, do you have to go? I do, actually.