Interview with Doug Swesty (contd.)

0:00 Many, many, many years, and they're reluctant to abandon them. They're often forced to hobble along, constrained by the architecture of something that they've done in the past, and it creates problems for them. So I think that's an advantage. I think that's something we can do a lot easier than the Black Hole Grand Challenge can. There's a lot of software baggage, I think, that they carry along. There's a lot of fighting over. the other advantage I think we have over them is not an advantage because of experience again it goes back to a simpler problem I think we have a problem we have more physical intuition for I think we have some physical intuition for what should happen during a black hole or a neutron star merger as opposed to what should happen for a black hole merger I mean we can sit here and do Newtonian and post-Newtonian simulations and neutron star mergers no analog to that in the black hole case. Yeah. Right, so I can look, you know, we can pull up movies or plots or whatever and say, ooh, look, here's what we saw in the Newtonian case, here's what we saw in the post-Newtonian case, here's what we're going to see in the relativistic case, and we can look at the differences there. We can begin to understand what effect gives what difference. There's no analog to that in the black hole case. Right. That's right. And it's a completely, completely relativistic phenomenon only. And that, I think, is a big advantage. You know, I was doing... I was actually... I didn't even... I never thought of it this way. When I was doing supernova calculations, I was doing 1D spherically symmetric relativistic supernova calculations. I didn't realize I was doing what we're now called in numerical relativity, constrained evolutions. And I had developed all this intuition about what happens form right at neutron stars hydrodynamically and all these things. And I can rely on all that intuition now that I have that's not there in the black hole case. I personally think that's a big advantage. And I think that's where some of the disagreements with my colleagues come from about how easy things are going to be. You know, they proceed from that. In many senses, neutron stars are not terribly relativistic phenomena. An isolated neutron

2:30 You know, to a large extent, you know, it's a 10 or 20 percent level correctly described by Newtonian gravity. You know, if you want to go to a post-Newtonian theory, it's even better. And it's not to say that the merger itself isn't going to be well described at that level, but, you know, the space-time evolution from a single neutron star is, to me, much, much easier than the space-time evolution from a single black hole. have a big advantage. Also, you know, I I've done a lot of administration for this project it's much easier to do to handle a project that has fewer people. We have a smaller team than the Black Pole Grand Challenge does. everyone brings the appropriate area of expertise and because there's not such competition because we do have a smaller team um i think that it makes it much easier to manage for i i think this is kind of a stunning thing for theorists because we're not used to working this way and uh you know when you look at these bigger collaborations it starts to look more like particle physics experiment and um we just in some ways don't have the mentality to deal with And I have just developed a tremendously, you know, an absolutely newfound respect for experimentalists who work in these big collaborations because, you know, they manage them so well. I keep thinking, gee, if we had 50 theorists on this project, we'd have a war. Yet, you know, big experimental projects proceed along in this way and, you know, fairly good fashion. I mean, I don't know of a single large experiment because of politics, whereas I'm pretty sure if we tried the same thing in theory, we'd be in that situation. And, you know, I think when I actually look at the LIGO experiment itself, I mean, it's a marvel at how well this thing runs, at least from the outside, from me looking at it, thinking about, you know, what has to go into it and how much management there is and, you know, how much fighting there must be over directions to take on technical issues

5:00 And it's pretty amazing to me that these things work as well as you do. The thing I've learned from this is I think the next time I want to do a project like this on this kind of scale and go for an HPCC computing project, I probably want to deal with even fewer people. My objective is to spend the least amount of time I can doing administration and the least amount of time doing science. That has been a bit of a tough haul for us. Yeah, there's a fair bit of overhead. Well, it's also made it a bit difficult, too, the fact that this is the first time NASA has ever done anything like this with the idea of issuing a contract for work as opposed to issuing a grant, awarding a grant. And there were a lot of problems within the university here getting this accepted. The university didn't like the idea of us having a contract to deliver software where we wouldn't get paid if the software didn't work. That was a bit of a tough sell. There was a point about a year ago where I wasn't sure whether we were going to get necessary approvals. Even after NASA had said, yes, we're going to fund you, I wasn't sure that the university was going to sign off on the contract the way it was. The university was saying, no, we don't fund work in this fashion. It's not like delivering a car. It's to be built for you to your specifications. We can't fund science in this way. effort basis and if it doesn't work out well you've still got to be paid and NASA was saying that's unacceptable you know we won't pay if it doesn't work and that's a new paradigm and I think potentially a very dangerous one for funding science not getting paid if it doesn't work goes a long way to encouraging people to take shortcuts I think it's just a bad bad idea so that aspect of this has been very different the black hole grand challenge well yeah it does it also kind of holds our feet to the fire too we've had big yelling matches among collaborators over the phone basically saying look we have to do this we have to deliver this code by a certain date and it forces things to happen that may not happen if there weren't so much pressure I don't know there's some pluses and minuses to it you may find yourself

7:30 getting along better with collaborators when your feet are held to the fire over something than you would if you had the option of going your separate ways and doing what you want no that's a plus side to the whole thing interesting so I mean in that sense I suppose it's our the collaboration is organized quite differently from the black hole one It's really more coincidence that they both go under the name of the Grand Challenge device. The structure is very different. Well, no, they're both Grand Challenges. I mean, it was just that the black hole was funded by the NSF Grand Challenge Program under their HPCC effort. And we were funded by NASA's Grand Challenge Program under their HPCC effort. So NASA, characteristically, is different that way? Or this was unique for NASA, too? Well, NSF has had one round of Grand Challenges so far. of which, you know, there was the Cosmology Grand Challenge, there was a Radio Astronomy one, there was the Black Hole Grand Challenge, and some others that I'm probably not aware of. NASA, this is the second round of Grand Challenges we've had. The NSF Grand Challenges, I think, were funded for five years. Ours were funded for three. The first round of them was passed. And NASA's experience on the first round of Grand Challenges was that things didn't necessarily work out the way they had envisioned. A lot of people had submitted proposals not doing what they proposed to do. And so on the second round of grant challenges, NASA decided instead of awarding these things as grants, they were going to award them as contracts where they wouldn't pay if they didn't get what was proposed. So in that sense, it's a little bit different. The programs are run differently. They all fall under the federal HPCC program umbrella that they're administered by different funding agencies within the federal government. I mean, for example, DOE also has an HPCC program, and I think Department of Defense also does. And so there are many different, you know, Grand Challenge programs funded by different agencies. And so there are differences probably between each of them. Yeah, they're all slightly different. They have slightly different objectives. The idea, I mean, the idea behind all of them is that one is doing really computational science that push in engineering, that pushes the forefront, pushes the envelope on what's possible. And to be honest, I think we really are doing that.

10:00 I mean, five years ago, we would never have thought about trying to do a calculation like this. Computational, you know, technology wasn't there to do it. That's both algorithmically and hardware-wise. It just simply didn't exist. So, you know, we are doing the kind of science that you wouldn't have thought about doing a few years ago. Yeah, I mean, I would say it's a big step up from Newtonian hydrodynamics. Well, you know, the thing that you have to realize is even ten years ago, Newtonian hydrodynamics was not being done in three dimensions. Yeah. Everything's evolving here. I mean, all of astrophysics were doing the problems that 10 years ago wouldn't be doing. I mean, 10 years ago, we weren't doing 3D hydrodynamics. In 10 years, we will be doing 3D radiation hydrodynamics, which we're already starting to do, or we model radiation transport, which is really a much harder problem than doing hydrodynamics. And, you know, it's just enabling us to tackle the kinds of problems we couldn't tackle before, and I think it's really the future. Just the improvement in the computing technologies? And also, this is actually probably a lesser-known aspect, but I think it's in some sense even more important is the improvement in algorithmic technology to do this kind of thing. For example, in the supernova work I'm doing, radiation transport is very important. In multi-dimensions, radiation transport is a really hard problem. In theory, you should be solving a Bolton equation, but that's a six-dimensional equation. And even now, we just can't even begin to do that. So you make approximations to it. But even there, in three dimensions, before you couldn't do it, Because of advances in linear system-solving technology, you know, the cloud mathematicians are doing better and better jobs in finding ways for us to solve, you know, linear and non-linear systems of equations. We're doing things now in multi-dimensions that we couldn't have done even three or four years ago. And that, I really think, is a tremendous forefront to what's going on. I mean, it's right at the forefront, I'd have to say, of what's going on. It's a tremendous advance of our capabilities. And in some sense, that's made a lot of the things that we're trying possible. So I think, you know, it's a combination of a lot of things. In some sense, it's the right time for this to happen

12:30 because hardware advances alone wouldn't get us where we want them to go. It's hardware advances plus methodology advances plus, in some sense, a renaissance in the way that we think about a lot of these problems that's made possible. I mean, if you would have told somebody five years ago, I'm going to submit a grant asking for 60,000 hours of computer time at NSF Center, it'll have laughed at you. Now, I sit here and I talk to my grad students on my postdoc on the phone and say, okay, we estimate this run's going to take 5,000 hours. Okay, submit this and we'll see the result in a week. We would have never done that. We would have never done that five years ago. Not at all. And the idea of 5,000 hours for a single run of something would have been unheard of. That's been part of the problem catching on. People have just... You now have to think bigger. a lot of people aren't there yet. A lot of people haven't made that leap. I see it when I review proposals for computing time that people write these very timid proposals and they're very timidly ask for, like, 1,500 hours of time over the course of a year. And that's changing. You can see it. And, you know, we're beginning to do bigger problems because we now realize we can do bigger problems. And it's actually, that's lag behind the arrival of the machines that can do it. had the machines that can do that kind of thing for some time. People didn't think grand enough in what they could do. Also, I think that the field is maturing. The field of computational astrophysics and computational science in general is maturing in the level of what's done. It used to be very difficult in AppJ to get papers published that would reveal numerical methods. People thought it was kind of trite and irrelevant to the whole thing. Now we think nothing about, or people who are really concerned about this think nothing about, you know, writing a paper just on the numerical methods they're using alone, followed by a paper on results. And a numerical methods paper will show results and show, you know, what the quality of the algorithm is, how the algorithm performs on certain benchmark problems. And I think that's a sign of growing maturity in the field. Other fields in computational science and engineering had done that long ago.

15:00 and astrophysics and numerical relativity have been slow in following. Now it's starting to happen, and I think that's very important. In some sense, I was just going to say, and this is a point I often make, that if you went somewhere and you were an investment broker and you said, I want you to invest your money in this company, the person would demand to see the books on the company. They'd want to see what the account balances were, what the profits were, what the losses. but it's been very common for astrophysicists to run around and say I want you to invest your faith in my calculation yet they don't tell you anything about what's wrong with the calculation what its liabilities, what its assets are and I think that's got to change and it is changing to some extent until it changes to the point where everybody reveals their strengths and weaknesses the weaknesses mainly in their calculation we won't have a reliable paradigm for doing this kind of science and it's a struggle I mean you know everybody that's doing this sees papers published in AppJ or FizRev or whatever that show results and don't show the don't reveal nothing you always have to have your doubts about that well I thought to get back to where we kind of started off have projects like in some sense seem to demand a high level of precision from numerical work, which goes beyond what analytic, in many cases, what analytic work can provide. Has that helped in some way establish an audience for computational astrophysics work, or is Is it more just the technological advances that sort of brought the moment? I think both. I mean, we certainly wouldn't be doing computational astrophysics without advances in computing technology. I mean, we'd be doing 1D problems, maybe 2D. LIGO has upped the ante because it it represents a problem it represents

17:30 experimental knowledge or observational knowledge I think perhaps that would be a better way to say it given that there is the O at the end of log O the observational information that would constrain theoretical models of certain phenomena tell you where I see the two biggest advances in computational astrophysics, or the two biggest pushes, let's put it that way, going on computational astrophysics now. One is mergers, both black hole and, you know, I should say computational astrophysics and relativity. Black hole, black hole, and neutron star, neutron star mergers is obviously a big push. There are a lot of groups working on this. The other one is in numerical cosmology. And what has that been driven by? Well, it's been driven by COBE and the observations of the cosmic microwave background. The information we've gotten there have now provided things that constraints on cosmological models. There's been a tremendous push and acceleration of work in that area. I see the same sort of thing happening with LIGO, probably on a smaller scale. There are by no means as many numerical relativists as there are cosmologists, but these are the two biggest computational pushes I see in astrophysics right now and they're both driven it seems to me by a combination of advanced capabilities as far as the computation guide plus advances in observation that provide firmer constraints challenges for theorists I mean the computational technology provides the capability the observations provide the challenge And I think LIGO, in that extent, has, because if you stop and think about what LIGO could do, well, okay, first, you know, direct detection of gravitational waves. Okay, well, that's fundamental in and of itself, and if for no other reason, that would be a tremendous impetus to be LIGO by and of itself. The fact that, you know, we could actually watch the process of black hole formation occur, which is pretty marvelous when you think about it. I mean, everything we see in astronomy, we say, all right, we learn everything in astronomy, we look at whole populations of objects, and we see trends of those populations. infer evolutionary information about those objects from the trends in the data we see. Here's something we're going to sit and watch, and this is pretty cool. It could tell us something about the structure of neutron stars and the dense matter equation of state. It could tell us something about how nuclear synthesis occurs and how half the elements heavier than iron that we see in nature may occur.

20:00 It could tell us something about where the site of gamma ray bursts are, and that would be pretty nifty in and of itself. other thing is it's going to tell us something about stellar populations that we don't know we're going to be able to measure the rate of mergers and so you've got you know people running around there you know people doing estimates of the rate the observationalists like sterile finney all right and then there are people down the hall here like ego even who are doing theoretical population synthesis calculations of what those rates could be and they're very different they're different by orders of magnitude and either sterile's going to be wrong or eco's going to wrong, or probably likely both of them are going to be wrong, okay, and that's going to be pretty awesome in itself, it's going to be a direct measurement of how often these things occur, we'll be able to infer something about stellar population statistics on the basis of these, and that's pretty neat too, and that's not to speak of other phenomena which, you know, like supernovae or something like that, which LIGO will tell us something about, so there's a lot being driven by it, you know, the trouble is, I mean, you know, It can be, from my own experience in moving from different places to different places, it can be hard to see. I mean, at Caltech, I'd imagine, you know, everybody's very cognizant of LIGO. Here, it's less so. I mean, basically, the only people in this department that are probably very cognizant of LIGO are myself and, to a lesser extent, ECO. You go over to physics, and certainly Stu and Fred Lamb are, and other departments who are going to have people doing this kind of stuff may be more or less involved. But it has a tremendous influence on the whole field in my opinion. Yeah, that's interesting. Yeah, so I think that's, yeah. it's actually going to have an influence on the field in, you know, ways that I think people who are not directly involved with things that are related to life probably don't even appreciate at this point. Oh, yeah, and I failed to mention the potential for, you know, a possible measurement of a Hubble constant independent of the optical distance ladder or the electromagnetic distance ladder, I guess is the way I should describe it, which in itself would be really an awesome, awesome thing for cosmology and for everything an extra block in astrophysics yeah, sure there might be a cosmological

22:30 yeah, that's pretty neat so, well yeah, it's exciting and it's interesting to hear about your work on this this is an area that I've had interest Yeah, I guess, to me, maybe I've adopted it a little easier because I came out of a background, like I said, I came out of a nuclear theory group, and I've been interested in neutrino astrophysics for a long time, which is really a pretty nontraditional area of astrophysics. Nobody in this department does anything with this at all other than me. And so to start working with gravitational wave astronomy, well, okay, that wasn't such a big deal. I'm used to doing something different. But I think a lot of people who are traditional electromagnetic astronomers or astrophysicists this is a very different sort of thing I don't think everybody's fully appreciated how much influence the LIBE is going to have on the entire field It's pretty nifty Yeah, interesting Well, thank you very much Sure, no problem So, uh, this interview was recorded today between about a quarter past four and five thirty, the fourth of September, 1997, with Doug Swesty.