Interview with Ray D'Inverno

0:00 It's the 8th of May, 2011, o'clock in the morning, and I'm speaking with Professor Ray D'Inverno. Well, why not start at the beginning, since I mentioned that I'm interested in numerical relativity approaches to particularly things like binary black hole problems and gravitational waves. and one of the approaches to that problem that's been worked on a lot in recent years, especially at St. Hanton, is coach a characteristic matching to deal with the radiation far from the source and that kind of issue. And I was curious as to the history of this approach. Well, if you're curious in the history, should I just give you a little bit of background before I head into that? Because it's a bit specific. Right, well, my original interest was in exact solutions. Exact solutions has been, for many years, the sort of heart of general relativity. which is solving Einstein's equations. You probably know that Einstein originally thought they'd never be able to be solved because they're so complicated. And if you write them out explicitly, say, taking a metric as a function of... But if I say anything that isn't clear, if you take a metric as a function of four arbitrary variables and you write out the full field equations, you get a piece of paper which is about 10 foot high. They're just horrendously complicated because we write, you know, g mu nu equals 0 or whatever. It cloaks this complexity. sceptical about there ever being an exact solution so it was a bit of a shock that within a year Shrashro came up with the famous solution which was the precursor to black holes then there was a sort of pause in the sense that there weren't many exact solutions that were suffered in Syrian decades and then in the kind of 60s there was an explosion of new solutions quote new solutions as people used different techniques, invariant techniques, especially things like Petrov classification optical scalars and in particular my area in those days which was computer algebra because i was a pioneer of bringing computers in to do these vast calculations to solve einstein's equations so the good news was that there were literally hundreds thousands of new solutions which were discovered you have to be careful about what you mean by solutions because the schwarzschild solution has one parameter in the m whereas a lot of these other solutions are defined modulo ordinary differential equations or even partial differential equations so they're sort of like family's solutions so the good news is there's a lot of solutions around the bad news is that most of them seem to be pathological they're solutions which are just thrown up by

2:30 the fact that you're dealing with partial differential equations you know that if you take just the wave equation in one time dimension one space equation you get what's called down better solution which consists of two arbitrary functions and so it means that there is this tremendous wealth of solutions that's available but it's only when you start putting in realistic things like initial conditions and boundary conditions and so on that you then tie it down to sorts of solutions which have some relevance to the world we live in yes it's much the same sort of thing with relativity that you get if you don't put any boundary conditions or initial conditions then you get this host of solutions most of which are pathological they've got weird which was weird asymptotic structure, and so on. So I was interested in trying to, first of all, make some geometrical statements about the solutions which were invariant, because one of the problems about solutions is you exhibit them in a particular coordinate system, and that is both a kind of strength and a weakness. It often cloaks what it is you want to understand, so you need to get invariant characteristics. And the other thing is, because they're in particular coordinate systems, you don't know whether if you have two solutions whether in fact they're the same solution modulo a local coordinate transformation that you can map one into the other you probably know that historically the Schwarzschild solution has been apparently discovered on at least 20 different occasions subsequent to Schwarzschild just because these things were found in different coordinate systems. In fact when I first got my system working the person who taught me relativity at Oxford I won't reveal who it was, but they submitted a paper for publication with a new solution, and I used my package to check it. It turned out it was flat space. It was Minkowski space in a very esoteric corner. So computer algebra certainly has a role in exact solutions in dealing with these complex computer algebra systems. But the key question for me are, what do they represent? Are they distinct? and do we have solutions for scenarios that we're interested in? Well, what do they represent? As I say, it turns out to be a lot of bad news and as far as we are able to interpret them, there's still a lot of problems with that area, but the last majority of them, the exception of things like the black hole solutions, plain gravitational waves perhaps,

5:00 and cosmological solutions, so it's a bit disappointing. Are they distinct? Well, there was a breakthrough in the 70s with a Swedish mathematician called Carl Hader who, building on previous ideas of Brandts, was able to show that you could make a decision algorithm or part algorithm that enabled you to compare two solutions and to say, well, what you do is you look at an invariant classification of the two of them. If they're different, then the solutions are different. for having transformations which could map one into the other but then it becomes a different problem and it's a question of solving four algebraic equations and four unknowns which is not algorithmic but at least it makes the problem tractable. And this Carl Hader classification tells you that you have to go as far as the seventh covariant derivative will, even with computer algebra systems, the complexity goes sort of polynomial-wise, goes almost exponential. So the more derivatives you have to calculate, the huger the expressions become, if you excuse my English. So seventh is probably prohibitive. There are some conjectures around that maybe you probably only need the first covariant derivative. And James Vickers and I looked at this, and we did some theoretical work that tightened up these bounds that lowered them. And there's sort of good news and bad news. There's now a counterexample which I think needs the fourth covariant derivative, so you know you have to go that far, which is a bit unfortunate. it's probably only for a class of solutions of measure zero, and in the generic case, one or two derivatives is likely to be enough. So there's certainly progress on this business of understanding solutions, characterising them, and deciding whether or not they are new, or different from existing solutions. But the problem is that we don't have any solutions that model the things that we're really interested in. So, although we've got Schwarzschild, which is the one-body solution, we don't have the two-body solution, or the n-body solution, or a radius solution, you know, with a sort of interior that's doing something, an exterior that's mapping at that. We have some asymptotic solutions, I mean, in a sense, black holes are asymptotic solutions, in that they're the solutions which physical configurations will evolve to asymptotically in time. and we've got gravitational wave solutions which are sort of idealised

7:30 in a sense that they're what waves would look like a long way away from sources but I mean since that's how we see ourselves probably you know cosmologically sitting on the earth and we're going to get signals from a long way away they are useful so it became clear that if one was interested in wanting to really model astronomically significant configurations exact solutions was unlikely to be the way forward the chief example of the limitations of exact solutions has been the Kerr solution where for years people have tried to fill in to put in an interior solution to Kerr as a black hole solution just like Schwarzschild you've got twist and you've got rotation as well and as far as I understand it although different people might argue about this they fail to do it, we don't have something which has physically realistic matter and a Kerr solution outside, which is quite a disappointment after, you know, decades of effort. So it became clear to me that the way to go, having had a sort of background in computers, but using computers as doing computer algebra, was to use computers to solve Einstein's equations numerically. Because if you could do it, in principle, it was almost like experimental relativity, because the way that you proceed is you start off with some initial data some initial configuration and then you let that evolve in time you let the system rip so in principle you could take you know all sorts of weird and wonderful configurations of matter and then just see what happens to them so it seemed like that was a the right scenario to go and especially as bit by bit it looked like we might be on the edge of the new area of gravitational waves and the need to match signals if we were able to detect them to possible sources. So this is the direction that we went and when you start to look into it, of course the story gets much more complicated, but in essence The vast majority of workers in this field, and there's not that many of them, I suppose, although it's a growing field, have worked with finite difference methods on a finite grid. They have a finite Cauchy grid, so what they do is they take a time slice T equals constant, and they work on a grid that has finite dimensions. So you have a centre and you go out to some radius, say, for example.

10:00 And when you put a finite grid on that, you put matter, whatever the configurations you're interested on, and you try and evolve it in time to see what happens to it. So, for example, you might try and put two black holes in there and see how they interact. Now, there's a lot of problems with that. One of the key problems in this business, because it's finite, you have to know what to do at the edge, and you have to know what at the centre, and you have to know at the edge. So there's these boundary problems, and particularly the outer boundary problem, because if you take the outer boundary outside the matter configuration, then you assume they're going to interact in some ways, and that's going to give rise to gravitational waves. These gravitational waves are going to flow out of the system, and you want them to pass out this outer boundary out indefinitely in an idealised configuration where you just have an isolated system sitting in an otherwise empty universe sort of thing. And the problem is that there isn't a way of characterising gravitational waves locally. There are all sorts of problems about characterising energy locally in general relativity. You can't do it exactly. You can linearise a theory, You can say that if the field is weak enough, then you forget the non-linear interaction. You just look at the linearized theory, and then it becomes a bit like sort of Maxwell's theory or something. And then you have ways of characterizing gravitational waves, and you've probably seen the standard thing about the two polarization states and the cross mode and the plus mode and so on. That's a linearized picture, which probably only applies asymptotically a long way away from the source. grids of finite dimension the waves would come out they'd try and impose an outgoing wave condition a summer-filled radiation condition a condition that is taken from electromagnetic theory, which is a linearized condition. It's not an exact condition. It doesn't apply and what happens is you get these spurious reflections so the waves come out, hit this inappropriate boundary condition, being the best you can do. The waves reflect back as they come back, they hit the source to radiate and so on. So, you know, you're getting noise into the system that you can't really control. So there's a number of problems there. One, you've only got a finite, I can't

12:30 think of what the word is, a finite representation of the configuration. Secondly, you can't deal with gravitational waves, you can't detect them properly because you're using linearised approximations at points where it isn't appropriate to use it. And moreover, you've got this spurious reflections that are going on that are spoiling your solution. So, other people, particularly Geoffrey Winokur in the States at Pittsburgh, had been working with characteristic codes where instead of putting codes on time-like slices essentially you put them on null slices in fact in his case he was very specific he put them on null cones so he envisaged a situation where you have a central world line, time-like world line you had null cones emanating from it and they went all the way up to what's called future null infinity is that meaningful to you? I don't know what your background is ok, future null infinity which for thought might be a class of interesting space-times which Penrose calls simple so taking Schwarzschild in the sense is a paradigm for what a lot of solutions will look like a long way away and you assume it has this structure and it's called future non-infinity and there you can characterize gravitational waves exactly if you like that's the limit where linearized theory holds or anyway it can be done in the full theory asymptotically so you know what gravitational waves are this is after Bondi's pioneering work with what he called news functions which tell you what the sources are doing so he had experimented with these characteristic approaches based on null cones emanating from a central null, central time light world light because the problem with this is that matter in. This matter causes distortion of space-time, causes curvature and the curvature in its time will screw up the null cones so they will start to have caustics. They're ruled by null geodesic photons which are going out to the speed to null infinity and what will happen is that because of the matter it will cause these caustics to refocus, they cross, and then you have all sorts of problems numerically

15:00 because it's very, very difficult when, you know, you have one point that corresponds to sort of like two values to track this in a numerical regime. So he just assumed that he had such a sort of weak source of gravitational radiation that this wouldn't happen. now the other approach which I and others looked at was to go the other way and start out from null future null infinity and set up null cones or null hypersurfaces there where you know that in the limit they're not going to have and follow them in track them back in and then when you just look at the region assuming it's big enough which it should be where the null hypersurfaces behave, they're caustic free. And so the idea then is to match an interior Cauchy code on a finite grid with an exterior characteristic code that goes all the way out to future 9 infinity, 5 plus as it's called in the trade. And in some sense get the advantage of both approaches, because there's a lot of work that's been done on finite grids, the time like slices, there's a lot of insight and understanding, but you've got the big boundary problem at the edges. And then this characteristic stuff, it seems much more appropriate for discussing gravitational radiation, because after all, the characteristics of the theory are the characteristics of the null geodesics that rule these null hyperseptics. I mean, the gravitational information travels out on these null hyperseptics, so it seems the appropriate way to track gravitational radiation. And moreover, mathematically, it's much nicer because essentially you end up solving ordinary differential equations rather than partial differential equations. When you're inside, you have a very complicated setup having to solve what's called the constraint equations, and these usually give rise to elliptic equations or parabolic or mixed parabolic elliptic equations, and they're computationally very expensive to solve, you've got to do a lot of computation before you can evolve, whereas when you go into the characteristic regime it's very straightforward very simple, the equations decoupled in a nice way, there's a hierarchy of ordinary differential equations and a dozen ordinary differential equations which you solve which are much easier to solve so the good news is that you've got

17:30 the wealth of experience and expertise with interior Cauchy problems, you've got this simpler precise understanding of how to take gravitational radiation exterior. The problem is, you've got to match these two, and the way we do it here is we match them at a hard interface. So essentially we just have a time-like tube, shall we say defined by R equals constant, where R is some sort of radial parameter. And on this tube, we match these two solutions. So we have an interior solution, which is time-like, an exterior solution, which is characteristic. The interior time-like bit is called Cauchy. it's called characteristic this interface is meant to reside in the vacuum surrounding some sources and far enough out so that the characteristics don't focus so you sort of adjust this distance because constant, the constant is going to be something that depends on the physical problem that you're looking at so the question is how do you match these solutions at this interface so you can have a flow of information in both ways that you can match. Essentially, you take the information that comes from the coaching problem's initial values to go out to the characteristic problem, and you do the same thing in the other direction. So going out is called extraction, and going in is called injection. So you've got to have extraction and injection in such a way, really, that what's going on is you're treating the interface as an outer boundary, but you're making it transparent. So what you want to do, if it works, is to see the waves go through the interface and make it all the way out to feature null infinity. words, you have three advantages. This way, you have a global solution. Secondly, you overcome the outer boundary condition, in that you don't get these spurious reflections. And thirdly, by going all the way out to the future null infinity, you are able to characterize gravitational radiation in an invariant manner, so you can make invariant statements about say, where a lot of people plot what's happened to metric functions or something and, you know, what's that got to do with what's happening to gravitation and radiation. I mean, there's obviously a relationship, but because it's coordinate-dependent information, you know, it's hard to know what's physics and what's coordinate junk, so to speak. Right, so we started on this program quite a few years ago and it's proved to be a bit of a monster, really.

20:00 my idea was and looking back on it it probably was an incorrect idea in that it's caused us to go at snail's pace in comparison to our competitors was that we'd start as simple and get complicated so in numerical relativity you usually talk about dimensions in terms of spatial dimensions so a full four dimensional problem is called three dimensional because it's got three spatial dimensions and one time-like dimension so the time-like dimension is taken for granted you're interested in time evolutions for systems. So my idea was to go one-dimensional, two-dimensional, three-dimensional, hoping that the mathematics would be much simpler, we'd understand it, and we'd be able to move up. So first of all, we did cylindrical symmetry. We've done a lot of work on cylindrical symmetry. We've got a lot of results. We've shown the method works. We've shown its efficacy. One of the major problems with numerical relativity is, of course, we don't have any exact solutions. from the exact solutions that you know you can properly test out the codes by saying we've got this configuration here it does this it does that I mean in the end you have flat space when there's nothing around you've got Schwarzschild asymptotically if you have a single black hole stratically symmetric thing you know there's not that much to match against them even when you do have exact solutions they have problems associated with them so we really what you need these solutions to give you confidence in the code to calibrate the code to make you feel that you know i mean the thing about a numerical evolution is that you'll always get results but whether the results have got anything to do with physics or whether they're just you know artifacts and spurious things it's difficult so you need to calibrate these codes to have confidence in them so now we have had three families of solutions in cylindrical symmetry which we've been able to match the codes against. The first one is called Weber-Weir, which represents a sort of pulse of gravitational radiation which comes in from past I-infinity, sort of gets to its maximum on the symmetry axis and expands out again. And if you compare that to sort of doing a finite Cauchy grid and you look at the standard Somerfield conditions, you You can see how as you move it up, I mean, it gets better because, you know, the linearized approximation comes better, but even so, you still get these reflections, whereas ours goes straight through,

22:30 and, you know, there's certainly less than 0.1% of error from the exact solution. If you run this for large amounts of time, large amounts of time being defined by what's called the crossing time, So the time it takes for a light signal to cross the zone of interest. So there's Weber Wheeler, then there's the family of solutions due to Safia, Stark and Paran. They've got rotation in, so they're much more physically interesting, but unfortunately they're not asymptotically flat, not asymptotically simple. This rotation blows up if you go in asymptotic directions. Nonetheless, they are exact solutions. and we've been able to modify our code to cope with this sort of infinity. It's a kind of weak infinity, an infinity that you can control, factor out of the problem. So we've been able to show that that works, and that works for long times, and I think, you know, like hundreds of crossing times and so on. And then there's another solution, which we're not as sure what it probably represents. Some sort of gravitational waves doing some weird things due to Xanthopolis. and more recently some of my colleagues at the other end of this corridor have been using this code to look at cosmic strings and they've come up with some really nice physics showing what happens when you take a cosmic string and hit it with a gravitational wave and what happens to the waves and how the string is excited and it rings on so cylindrical symmetry we've done, we've milked we've shown it works so the next thing was to move to actual symmetry that cylindrical symmetry would give us a lot of insight into axial symmetry. It turns out it doesn't. It turns out cylindrical symmetry is pathological. It's sort of strange because it's it doesn't lead naturally to axial symmetry. It's a different story. What we believe, however, is that axial symmetry is close to full generality, full three-dimensional stuff. So that if we can get that working, it should in principle be a small step for us to go to three-dimensional. So all the effort's gone on to that. It's taken us a long just about have a master code which has got a bit in it for doing Cauchy interior, a bit in it for doing characteristics interior, a bit, a module for doing injection, a module for doing extraction and you've got to screw it all together and my most recent research student called Dennis Polny

25:00 who's now at AEI Albert Einstein Institute in Potsdam in Germany as a postdoc before he finished sort of got this master code working. I think it's still probably got some bugs in it, but it can evolve systems for a few hundred time steps and then things start to go wrong. So we're on the verge of getting that working. And then, but we have the same problems that everybody else does with axial codes, two-dimensional codes. There's real problems at the interior. it's not problems at the exterior, it's problems at the interior how you deal with the origin how you deal with the problems with the polar coordinates having singularities so you get these what they call artefacts in the train they've kind of haunted previous generations of codes because there's been two dimensional codes around since the middle of the 80s I suppose so we've got the same problems but that's not what's important to us what's important to us is to show that we can Cauchy characteristic matching at the interface there was a big effort in doing that it turns out the geometry of the interior and the geometry of the exterior are completely unrelated so there's a big price you have to pay a lot of work you've got to do at the interface some people would say it isn't merited I don't know about that in the end you'll have to see how well these codes do but again it seems to work and the idea is then that we should make the transition to the three-dimensional case. Now, we've just been looking at vacuum, because already that's hard enough, and we don't want to do too much on non-vacuum on the interior, because that's a big problem. It's only right, you know, specialist stuff. You have to get into hydrodynamics, and hydrodynamics in the absence of relativity is hard going, let alone when you've got non-linear fields as well. So the idea was to show the efficacy of it. interested in the characteristic bit and the matching bit and not so much interested in the middle and assume that other people have got that and will give it to us. And that's where this EU network grant comes in because that was the hope that in this 10 institution collaboration our role would be to do precisely what I've just said, which is take codes that already exist for interior finite Cauchy

27:30 codes and be able to glue onto it an interface in a characteristic mode that enables you to take the solution out so you have these things of global solution no reflections and invariant characterization of gravitational radiation you know about cactus do you yes right okay so the idea was that this this should probably since cactus is the kind of workhorse of the network it's the central code everyone is supposed to be contributing to then our part in this was to try and you know cactus a sort of central flesh and it's got these things called thorns on the idea is that these thorns are supposed to work as independent modules so you can just screw in different thorns depending on your problem so the idea was to write a ccm thorn a three-dimensional ccm thorn which you could screw on and then people could use that as a way of if you like coping with the outer boundary problem right uh so that was the But it turns out that when we advertised the post, so there were seven of us that wanted to post-docs and the others I think wanted what they call pre-docs in the application, graduate students. We had a very thin field of applicants in terms of people with expertise in numerical relativity. I mean, they had expertise in other areas, but they didn't have the background. It turns out you need a lot of background to do this, because you need quite a bit of mathematics to understand the mathematical side of it, and quite a degree of computing expertise to handle all the various bits of computing. And there was one person who had that expertise, but not surprisingly, in the end, he decided to go to AEI, Potsdam, which is now the World Centre for General Relativity. and we had two major pieces of work that we wanted to contribute to the network project one was the CCM Thorne and the other was to do with neutron stars and neutron star modelling using linearised approaches to perturbational approaches which is headed up by my colleague in the neighbouring office Nils Anderson and in the end, given the field of applicants we had someone who was very strong in that area and we made an offer to him and he's accepted it called Reinhard Pricks

30:00 who's actually coming here tomorrow to meet the group and is starting, we think, in September if there's any money for it so that means that I'm not really sure how things are going to develop on the CCM front as far as Southampton is concerned I do have a very able student who's in his third year who I groomed to do this, and he was the one that did a lot of work on this cosmic string, so he used the cylindrical code as a way of kind of tauling up, and he's done some good work. and the idea was to move on to the Axel Cove but in the meantime he's sort of been persuaded it's such a big task that he's not sure he wants to do it and so he's gone more in the sort of astrophysical model interaction and he's involved in a joint project with myself, Nielsen, a guy at Portsmouth University called Philippus Papadopoulos who funnily enough comes out of the CCM background, I mean it's a very small field there's only a handful of people that are doing it but he's one of Jeff Winneker's research His PhD was on that, so he actually knows quite a bit about it. So I don't know where we're going to go. The idea is for me to go probably in July and spend a couple of weeks with Dennis Polney in Potsdam. I'll probably have to take over the code because it's looking like he's going to be so busy with his new projects that he's not going to be able to really take it a lot further. I don't know. He's still interested in doing it. how much spare time he has he already has a sort of another unofficial project because he's a co-author of what's called gr tensor gr tensor is a computer algebra system that sits on top of maple um which is one of the reasons he came here because of the computer algebra thing and he does quite a lot of development work on that and well he would say i suppose it's more maintenance work keeping the system going so that's a private project but she keeps going But if he does CCM, it's not clear how that's going to interact with his major projects at Potsdam. So I may have to take it over, I may have to do it myself. And if I get a talented research student in the meantime, I might move them on. I don't know. So it's taken a lot longer for us to get here than I ever imagined.

32:30 It's turned out to be much more complicated than I ever thought. The code is a monster. I still believe the transition from axial to fully three dimensional is not big, in fact all you do is you just say all the functions are functions of UR theta and phi, whereas at the moment they're all functions of UR theta and phi is invariantly defined it should be straightforward, you just have more information essentially to process since we haven't really got axial working, fully enough to persuade people that it is working there's still quite a bit of work to be done there so that's about the length and breadth of it interesting so it's actually a tricky a difficult problem finding the people with the right background yeah it is the mathematics is a bit heavy turns out because it's to do with the geometry of the two regimes is just completely unrelated so you have some very very tricky i mean in the end we essentially have to solve a mixed parabolic hyperbolic parabolic elliptic equation to get our transfer of information basically the news but out of infinity it sort of shoots right back and pushes itself into the interior and solving that equation in a way where we can handle these problems at the origin it's again it's not the it's not the outer bit it's the because there's problems dealing with the interior boundary problems that we have to have this equation of mixed height that changes its character as you go in it's a bit dirty really i'm disappointed it's it's not ccm that makes it dirty it's axial symmetry that makes it dirty so you know you've got to show that it works in that environment well this was the theory before you move on to have said why you cut your losses and just go straight to three dimensions maybe that's what the EU network might project us in that direction but you haven't put quite a lot of effort to get the actual code to the point it is I feel since in the end we did milk cylindrical code we got a lot of results out of it we got quite a few publications if that's your criteria you know it'd be nice to get someone active symmetry after all this time but in the meantime you know the pressures on to try and get some results in three dimensions, because if what we claim, we can do these three things, and not only that, but we can do it with great accuracy.

35:00 If you want more and more accuracy, basically, you've got to go the characteristic route. People want exact templates, that's what the game's about, if you're going to try and find them in the early signals. You mentioned, or I got the impression that with the Axial code, you have your own internal Cauchy code, and then the idea with the current project is that you'll be using another code related to the Cactus code. Is that likely to be an advantage to have somebody else doing the interior work, or a disadvantage because you have to sort of adapt yourself to whatever way they've done Well, the trouble is I don't know very much about Cactus. I mean, Dennis Polly and my third-year student both went on a Cactus course, the first course in the States, and they have a bit more expertise about it. And it's the same problem about the interface problem within Cactus, about, you know, what the information is that is going to be needed in order to interface CCM to existing Kochi codes. So I guess it's, I mean, in principle I think it's straightforward, but my guess is that when you try to do it in detail, going to be a bit horrendous because of the way in which information has to be passed. It will have to be in the format that's needed for each of the two problems. And there's difficulties about the fact that most people are working on Cartesian grids now for doing interior codes, and ours is very much kind of like topologically spherical polar coordinate type coordinates with a characteristic regime so there might be problems about sort of having to use interpolation to get information at the grid points that you need for the various codes so I'm expecting there to be some sort of dirty detail but in principle the idea is that I know the information that you need to pass backwards and forwards with, not surprisingly, the metric and first derivatives of the metric, appropriately defined. Was it a big part of the labor with the actual code developing your own Cauchy part, or were you just able to draw on the experience? It was, actually. No, we tried to mimic. The best work that's been done to date has been done by Stark and Peran.

37:30 Peran has moved on to other things, but Richard Stark is still up to quite recently interested in the actual code that was originally developed. That's where most of the results have come from. And he was developing a three-dimensional version of it in much the same sort of setting. So the idea was that we were going to mimic Stark and Paran for the Cauchy code. One, because they had some tests that we could calibrate our code against. and two because Stark was still active and was working on three-dimensional stuff and in an earlier version of the EU network we were going to screw our code onto Richard Stark's three-dimensional code when it came to it there's some sort of rather nasty detailed things that they do to make the code work and when we asked them for guidance about how to cope with it they didn't give it to us I mean for whatever reasons they said they'd forgotten or they lost the stuff so we had to do it ourselves longer uh i think they weren't totally secure about some of the things they had done to get the codes to work you know kind of uh there's a certain amount of cookery in america relativity and they weren't able to pass it on to us so that that took quite a time to reproduce it and in the meantime i'm still unsure of what has happened to richard stark up until a year ago he was still at trieste where he's been for some years working on this code though i think he'd I think he was actually married to an Italian, had children there, was living off unemployment benefit from the state. And then a few months ago, he got a posting in somewhere in South America, I can't remember where, Argentina perhaps. And apparently he has this code and it's working, but he hasn't given us access to it or passed it on to us. We're a little bit nervous about it. In the meantime, the mainstream of the community has gone Cartesian and, you know, centred at Potsdam, and we are now involved with them. So the likelihood is that we will ditch that aspect and try and do something from the point of view of the Cactus forming the framework of the main code development workhorse

40:00 project so in some sense the way you've been working previously was a better fit from the technical detail point of view with this code of Stark and Paran but that well perhaps the main reason for going with Cactus at this point is that it's the main game in town yeah yeah I imagine if there's going to be a template matching at some stage, the templates are likely to emerge from parts done. And there's two major groups, there's them and there's Penn State, sort of all the people out, Pablo Laguna, or Jorge Pullen, there's a bunch of them there. so that's that's that the two main areas I think my codes are likely to produce templates you've mentioned the the sort of cookery that's required in American activity is that and the fact that therefore can be difficult to for one group does it were communicate the real details of what they did to another group, is that something that you think might continue to be a problem in a big collaboration like this with ten groups? I do, I do. The devil's in the detail is what I've found. I mean, cookery is a bit unfair. These are well-recognized techniques for showing them with problems. I mean, you You know, you get problems with shock waves in certain scenarios where if you don't have shock capturing techniques and you need to sort of dissipate them, so you put in this thing called artificial viscosity, which is kind of a bit of damping to stop the shock ways for me and it's an accepted technique and you know it doesn't massively alter the results you get but it does mean that the codes run longer you don't have these again these infinities cropping up that make them crash the big problem with codes is stability having them run long enough i mean the problem is that you want really these codes to run almost indefinitely

42:30 you know the bribery black hole thing you've got two black holes that fall between each other and they eventually merge you're talking about hundreds of revolutions to try and capture the interaction which is a long time so you've got to have long term stability which is in fact what we've shown we've got with the cylindrical set up so that's the good news and only that the longest running code ever Moniker's characteristic code that went for 1,000 M. It might have even been more. M is a sort of mass parameter that characterises the solution. So I think the characteristic approach has proved itself as long-term stable. How it interfaces with the Cauchy code. In Cauchy code, there's not any hard theorems around that tell you that these things are long-term stable. be very much suck it and see techniques you know whereas the characteristic there are some theorems around in fact there is another approach which um a guy called uh frown dina was pioneering although i'm not sure if he's still continuing that work because the last i heard he didn't have a post and was beginning to get a bit long in the tooth and was talking about leaving science which is a shame because his stuff is very promising that's one of the problems with numerical relativity up until recent years is a very small field there's only a handful of people doing it so you know the idea of people checking each other's codes there just weren't enough people around to do it so freundin is using theoretical he's implementing a theoretical approach of Friedrichs, Helmut Friedrichs, which uses hyperboloidal decomposition. I mean, we've got the sort of picture that we have is these Cauchy regions in the middle And then these characteristic regions, these neural regions, that go out asymptotically. So this is type-like and this is null. And I don't know how to draw it, but Friedrichs is working on things called hyperboloidal surfaces, which asymptotically become null, so on the limit, and you get it out far out of that null.

45:00 But they're just one surface, and he's able to formulate GR, you know, as one problem. have the advantages geometrically, you're just, you know, involved in one set of data for us, probably there's no interface. And you get out the right things asymptotically, because it's null. It turns out to be quite a complicated formulation. It's done in terms of, rather, we tend to work into the metric variables. They work into the metric first derivatives, which are sort of Christopher symbols, and then second derivative terms as well. So the formulation is much more complicated. I mean, they have loads and loads of variables which they have to carry. But he's done some preliminary work at Freundin, and it looks very promising, and it's a shame that so few people are looking at that, I wouldn't be surprised, because Friedrichs is probably the leading person in the world in terms of the mathematics of the Cauchy problem. He's done the most to sort of set GR into a proper framework where you can use hard theorems to prove existence, uniqueness, stability, properties of the field equations. I mean, the people that do Cauchy evolution, that's kind of like folklore that they use. You have these things called constraints, which when you set your initial data, you can't just choose it in the timeline freely. So you choose sort of like the metric of the surface and its extrinsic curvature, sitting inside the four-dimensional geometry. But you can't specify that freely. You have to specify it subject to four differential equations called the constraint equations, and solving those is a major problem. That's where you get these elliptic equations. And then the question is, what do you do subsequently? So, all right, you solve these equations, you set up initial data, then you let the thing evolve. What do you do with the constraints? Well, there's three things that people do. One is to forget them and just let the thing go in free evolution. drift away from the true solution but then you use the constraints which in a sense have to be satisfied as a check on how good your solution is you know how closely to zero the mathematics tells you that um in terms of analytic solutions that once the constraints are satisfied and they're satisfied for all time by virtue of what's called the contracted bianchi identity some mathematics

47:30 that tells you that you only have to satisfy them once of course in numerical regimes where you've and an infinitude of finite different schemes which will lead to different solutions. You have to know what to do with them. So one of them is to leave them, forget them, for evolution. Another is to impose them periodically, which is called chopped evolution. So every now and then you try and make these zero and bring the solution back to the correct solution. And the other one is to apply them in every time step. And so you get different solutions accordingly. There's a sort of folklore that the constraints more powerful in some ways than the evolution equations. And that Chobtwig, who's like the main figure in numerical relativity in terms of, you know, his stuff that Carsten is working on, critical phenomena, applies them at every stage and almost uses them in preference to the evolution equations and seems to get better stability that way. Now, there's no mathematics that underpins that. just something he's done by trial and error, so to speak. He's tried various schemes. This seems to work best to give him longer-term evolution without stability problems propping up. But I think the feeling is that's what's needed. It really needs a proper mathematical underpinning. I think we have it characteristic. I think it's there for hyperboloidal. It isn't there as I would understand it for Cauchy evolution it's going to crop up at some stage or another because of the probability if it's not well formulated mathematically then you're going to have stability problems and you need these long term evolutions if numerical relativity is going to do its bit in the binary black hole bit as you know you can do linearised approximations for when they're a long way away and you can do For the ring-down phase, you can do, again, approximations, or exact theory for that. But it's the intermediate phase where you need numerical relativity if you want to track that thing for its full decay. And at the moment, none of the simulations go on long enough.

50:00 I mean it's a big effort to try and look at neutron star coalescence same sort of thing in Japan it's probably the biggest effort they have the biggest computers and the fastest computers and the most amount of computing resources and the last time I'd heard how they were getting on I don't think they'd got through one revolution stability problems Now, their attitude is to throw more and more processing power at it, but probably what it needs is some proper mathematical underpinning. Within the numerical relativity field, do many of the people have a very mathematical They tend to be physicists. They come from a physical background. There are a few, I mean, in the States. It's so old now, I just cannot remember any of his names. I can see the guy, I can hear him talking, I can't get his name. anyway there is a mathematician in the states who has some interest in he's done some work recently he is certainly capable of and he has worked on the Cauchy problem with the big names like Jerry Marsden Fisher and people like that Moncrief Vince Moncrief I've said in Germany Helmut Friedrichs, and who's the other guy? Ben Schmidt, probably mathematically inflected. Alan Rendell, who's British, Scottish, who's there, he's also very mathematically inflected, but he's not really looking at that end of the problem. It's more on the matter end of but I think the vast majority of people are physicists by background and interest really because they're doing it because they want to understand what happens to configurations of matter my friend next door is really an astrophysicist he wants to know about nutrient stars

52:30 and makes them cook So, are most mathematical voters not inclined to get involved in the computer, the numerical aspect, or is it in programming? Yeah. I had a tremendous difficulty trying to persuade the community about the efficacy of computer algebra when I started. They just didn't believe the results, you know. They said, how do you know you've got the right answer? To which I would say, well how do you know you've got the right answer, you've done it by hand. And it was really only when there were a lot of different systems in existence with different hardware, different software, different configurations, different strategies for solving problems, and you gave them the same problem and they got the same results, that people were able to gain confidence in it. And when I introduced it, I said it would take 25 years before this was sort of accepted as a hack tool, standard tool, which you wouldn't refer to. And indeed, that's probably about how long it took. of the key figures in the field when I came in, there was still a key figure in the field, was Roger Penrose, to whom doing anything using machine was anathema, and you had to do it using your brain to know that something was right. And then somewhere in the middle, I didn't quite know what happened, but he suddenly got interested in artificial intelligence, and then he's now a guru about computing and what it can and can't do. So there's some irony there, but I think there's now a complete acceptance of computer algebra. It's an everyday tool. People use it if and when it's necessary. I mean, for example, it's used a lot in numerical relativity, because a lot of the equations you have are very complicated. You don't want to do them by hand. You stick them on the machine and mutate the equation. In fact, it's even more than that. You can actually get the equations for your problem and get the computer algebra systems to produce the code and actually do that part and it will be automated. they're accepted and i think numerical relativity grew up probably just in the last conference you know we have a major conference every three years gr whatever the last one was the one that was in puna in india gr15 was it uh where the keynote speaker the main leaner was by Matt Choctaw. In fact, he had won the prestigious medal that the GR community

55:00 gives every three years for his work. And his work on gravitational, his work on critical phenomena, gravitational collapse, is very impressive. And the fact was that he got all Well, this is what is kind of generic, analytic results about families' solutions from numerical techniques. They weren't just saying, here's a solution, this is what it looks like. He was able to talk about generosity and find this particular behaviour and come up with these indices, critical indices and so on. So I think it's accepted, whereas in previous conferences it was looked on with that same sort of scepticism, I think, that computer algebra was when it first came in. People said, how do you trust the results, you know, behind the scenes? And, I mean, with some justification in the sense that it was a small field and people were working in it and, you know, there wasn't enough sort of checking of results. There's rather more of that now because there's been quite a nice marriage between approximate techniques and numerical techniques in recent years where one has been used to check against the other and there's been some very nice work that has shown that the two can be used in harness. so I think it's got respectability now and I think it's growing because it's one of the few fields where people are looking for applicants postdoctoral fellows, you look at them most of them, a lot of them are in new micro-instructions but one of the problems are that there's not enough people coming through who are trained to take up those positions who can go straight into the job and work they have to be trained up themselves and there's a bit of a time delay there so micro-relativity is not going at the speed I think people would want it to because of this training problem and one of the planks of the EU network was that you know if nothing else at the end of our tenure there should be some people who are trained up who are experts in the field who can you know move the discipline on. So there's a specific training aspect? Yeah yeah it's one of the

57:30 main planks of the grant. I mean the money is for postdocs and predocs you You mentioned right at the beginning that in the 60s I guess was when there was an outburst of finding exact solutions. I'm curious to know if you have any idea, any feeling as to why at that point this started to happen when as you say up to there there have been very few exact solutions um i think it's because up until then there have been very anti-hoc approaches to looking for exact solutions people just sort of pulled an ansatz out of the air and tried it and it either floated or it sank, and usually it sank. And in the 60s, though, people had a better understanding of the sort of mathematical structure of Einstein's equations, a better understanding of the sort of geometric significance of it and the algebraic characteristics of it, the Petrov classification, the optical scalars, geometrical optical scalars. and that these invariant techniques in turn were then applied to exact solutions and it turned out to be fruitful in the sense that they worked for generating closer solutions and as I said that's the good news, there's lots of them you look at McCallum and et al's exact solutions book and you look in the back and there's probably the references run into thousands and they're all independent and as I say, how do you count solutions anyway when they're not completely explicit, you know, the modular other equations, so there's families of solutions and infinitude of families in some cases. Nonetheless, you know, there are a huge number. But how many of them are everyday terminology? I mean, I don't want to decry exact solutions. I think most of our understanding of GR stems from exact solutions. They've been important, although probably too much effort went into just of finding them, rather than to other areas, but nonetheless they have led to most of our insight into what the theory says. So yeah, I think it was these invariant techniques

1:00:00 that were coming in that gave a new injection to approaches for solving the field equations And I was curious to know whether, with your work on computer algebra, whether you started off as it were with the interest in general relativity and this was the tool to solve general relativity, or you were interested in computer algebra and general relativity was a useful application. That was the former, definitely. It was just a means to an end. And, I mean, I got interested, obviously, once I did it. And, I mean, I started off wanting to use other people's systems and found out that there weren't really any other systems. And then I wanted to harness the best expertise there was in the UK and I discovered that it was probably me. So I couldn't get any further. But, I mean, I made the decision that Lisp was the right platform on which to base a computer-out system. And still, a lot of systems are Lisp-based today. And I got interested in LISP. It's such a sort of appealing language to a mathematician. It's so different from Fortran and the other scientific languages. And that led me in turn to have a sort of bit of a semi-interest, I suppose, in artificial intelligence. but sufficient of an interest to dream up a course and it was one of the first AI courses in the country and it ran for 20 odd years quite a lot of people came through it and I still have fascination with Lisp although it does seem to be becoming increasingly a historic language things have moved on to other areas since but certainly in the history of AI it was the queen of the languages, nearly every area used LISP in some form or other. In terms of computer algebra, it was the most successful platform in the early days. But now, sort of, you know, Maple and Mathematica and so on are C-based, I guess. You mentioned your work here on G-Tensor. I'm curious as to how much these programs

1:02:30 that are familiar to people like me today, occasionally to use computer algebra like Maple and Mathematica arose out of the early work that you were involved in? Oh yes, definitely, because when we started, there were two communities of people working. There were the computer algebra, the computer scientists who were trying to build algebra systems, and the physicists who wanted to have systems which would apply to their field of interest. And there's no doubt that the former systems never worked in real applications, and the latter did and not only that but that you could extrapolate them to other areas and the key example of that I suppose was Reduce with Tony Hearn who had originally written a system for doing Feynman diagram calculations and quantum electrodynamics and Reduce is now a general purpose system which is it's not got the sort of money behind it and the hype and the finance but if you want to do polynomial factorisation integration probably reduces the best program around to do it I just think mathematical make more fighting to get close but the communities learned from each other I mean in the end all the guys who started with physicists went and became computer scientists and then Tony Haines is a professor of computer science John Fitch who wrote CAMEL, Cambridge Algebra System came out with a physics background he's a professor of computing at Bath and so on So, historically, the people who are interested in applications, I think, made the biggest contribution to moving the systems forward as usable systems and workable systems and working systems, but bit by bit. i mean in the early days processing power was the problem um there's no use having this all singing all dancing program that could you know factorize x squared minus y squared is x plus y x minus y when what you want is to do calculations where you've got a million terms you know order eight and you want to simplify them you can't get beyond cubics or something whereas our systems were designed to do that and you had real problems with processing power in terms of you know how fast the machine was and storage particularly and the trouble with computer algebra is that you get this intermediate terms well where the terms grow as you expand

1:05:00 everything in sight and so to speak and then they collapse down again as you collect things up and make cancellations and so on and the memory was a problem which is why I went for LISP because LISP has this thing called an automatic garbage collector so it means when It can do a sweep to see what's active memory and what's junk garbage, and it can reclaim, throw that garbage away, reclaim the memory and start up. So the system could keep going for a lot longer than those that used the sort of once and for all allocation of storage. It wasn't dynamic storage allocation. Yeah. Yeah. I mean, and then even now, these general systems are not really designed to do relativity calculations, the sort of things that we want to do on a day-to-day basis. You nearly need applications packages, so every system has its applications package. I mean, sheep, the system that I was historically responsible for, is a system in its own right for doing general relativity. But the algebra that underpins it is very limited. So, for example, you can't do polynomial factorization, you can't do integration. But you can interface it to reduce, which does. In fact, there's a version of sheep with our sheep, reduced sheep, which does. The two sit together. that sits on top of Reduce, the general computer algebra system. GR tensor is a package that sits on top of Maple, a general tensor. I forget the name of the one that sits on top of Mathematica, but there is at least one, and so on for other. And the one that sits on top of Reduce, there's another one which is called... Anyway, you carry on asking me questions unless you've finished. I was going to ask if, of course, you mentioned having packages to work with these more general codes. Is that basically the same sort of way that Cactus works, the same general kind of principle?

1:07:30 First of all, I know very little about Cactus, so I'm not the person to ask. But secondly, I don't think it is. It's to do with codes specifically with numerical applications in mind in solving partial differential equations, nonlinear partial differential equations with specific reference to general relativity. And I think the idea is that you have this small flesh which is at the centre which is really a kind of message-passing sort of underpinning where you can get these various thorns to communicate to each other. So it's just a way of saying, right, this is going to be the way in which information has to be passed backwards and forwards. And then the thorns are the more important things, and there's a package for solving partial differential equations of this class. Here's a package for visualisation. Here's a package for parallelization. Here's a package for, sorry, I keep saying package, a module, a thorn, a thorn for grid refinement. And I think the idea is that you make the flesh as small as possible and the thorns as user-friendly and as capable as possible so that people can have a particular problem. They need these facilities and they just pick up the thorns that are necessary for their application. So in that sense, it's not the same as having a general computer which these particular applications sit on top of. There isn't, I think, a general numerical solving capacity at the base of Cactus. Right. And I gather that, as I understand it, one of the ideas of Cactus is that the person riding a new thorn to the work that's already existing on it needs to know relatively

1:10:00 little about how everything else is written. I was wondering, since you were saying that... job. Yeah, I hope it's true. I'm very skeptical, I have to say, I haven't been in this business for a long time. It's a wonderful idea, but I bet it doesn't work in practice. I bet you need to know the detailed workings of something before you can get it to do. Because I suppose you've had some experience of that, so getting things like sheet to work would reduce? Absolutely, yeah. Where you really need to know it. Can I just... Xcalc, E-X-C-A-L-C, is a relativity package. The main relativity package would reduce, I think, for us. Xcalc, because it stands for exterior calculus, that was the approach that was used there. I'd imagine that the physicists who did a lot of this early work on computer algebra would have been from computationally intensive fields like you mentioned QED and general relativity. Yeah, and celestial mechanics. And in fact I'm going to a conference and if my visa comes through, St. Petersburg in Russia in June Let's see if I can find the bookmark for it. Still not going to find it. Doesn't look great. It's a shame. Ah, here we are. It's called IMAX 2000. And this is a joint session as a computer-based organisation on computer outer and celestial mechanics and relativity. So historically, those two have sat quite close to each other. When I first did it, I had a system called LAM, the sponge brain manipulator for relativity. and there was a system at Cambridge called Camel Cambridge Algebra System and then Lamb eventually became called Sheep so they were like these two animals, Sheep and Camel

1:12:30 which were in competition with each other and in the end Sheep won, Sheep was able to do things that Camel couldn't do because of the storage problems that Camel had and it was largely a polynomial manipulator which kind of limited it, you had to hack it to get it to do general functions whereas ours was set up as a general function But, so the guy who was mostly involved at that was John Fitch at the time, and there were some others as well, Barton, who were born. But we worked together, and if you look at his thesis, you will see Camel used to do general relativity calculations, which were originally done in sheep, to check them. and similarly some of our calculations we did some celestial mechanics to get the same answer to check that we could get the same answers as them so historically it was quite important again trying to establish confidence in the systems if my abstract is up here Did the Celestial Mechanics people have to have familiarity with GR to the extent that they used, say, post-Antonian approaches, or was it strictly Newtonian for you? The people who were doing it at the same time as me, oh yeah, they needed to know some general Oh, yeah. All right. So I have an abstract, which is on the web. Good. So I guess that the last thing that I was going to ask you about is touching on this issue of whether, you know, one can successfully make it a new, well, torn in the case of In fact, it's sort of a new part of the overall programming suite without knowing the details of what's going on elsewhere. And in the present case, obviously, in the case of the European network, you have sort of ten different branches. It's part of the idea that there will be, as it were, a sufficient division of labour between the different approaches that the different groups can work independently of

1:15:00 each other so that, for instance, you'll be doing your Kochi characteristic matching here and, say, they'll be doing just the Kochi code in Berlin, or Potsdam, rather, or is it envisaged that there will really have to be quite close collaboration between the different... There's a whole sequence of activities that have been planned amongst the ten institutions. There's a sort of grid where you've got all the topics that we want to look at and all the institutions, and there's various ticks and crosses that go across saying which institutions are involved and what the level of collaboration is and which is the prime institution. So I would guess that in most cases it's a multi-institutional effort and that they will have to be in close contact with each other. And once it gets going, I think there will have to be a lot of communication for it. and it's sort of agreed by everybody but no one else has done much about it is that we want video conferencing facilities up and running in fact you may see there's a little camera there we've been experimenting with video conferencing we ran a sort of joint series with a group in Germany last academic year for a term no it was two terms which we kind of worked right the visuals were right over the web but the sound wasn't so got a handset that's got a speaker in it so that we could talk to the telephone and get back and so we had the sessions here. So I think it's important that we should be able to sort of communicate on as frequent a basis as is necessary. And there's a lot of travel money involved and I think the idea is to have a lot of organisational meetings. So And NEOS is one of the major centres of work connected with neutron stars, and that involves half a dozen of the other institutions, and there's a lot of specific programmes which run the way, some of which we head up, some of which we're only partners in. The CCM is a bit of a sort of thing in its own right, really. It does sit separately on its own, probably just myself and some interest from Philip Papadopoulos at Portsmouth, I think those are the only people in general terms, no, the whole idea of the network is that it is very much

1:17:30 collaborative effort with a lot of internal communication How successful were the trials at the video conference and did you feel that it allowed enough interaction between the people and the different primitive system which was used in sort of non-dedicated communication over the net and it was extremely volatile so sometimes it was really good and no problems at all and if there was heavy traffic on the net it was terrible you know they would be doing a presentation at the blackboard or whiteboard whatever and it would just freeze and it would sit there for ages before typically you know they'd be writing on the board and you get one line and it would freeze for a while and three lines of stuff written you know it's that sort of thing so it was really the verbal stuff which was the most effective way of communicating and that went well I thought it went well and were it not for the fact that we're starved of resources here we have no money for anything we're broke again when they're talking about you know sacking people in order to make the budgets and balance I think I think we would invest more resources into it we had a trial session in our neighbour ECS, Electronics Computer Science the computer science group there I know some of the people in it historically they used to be part of mathematics and they've got kit coming out of their ears they've got kit for everything and we used some of their state of the art video conferencing facilities and it was fantastic they had dedicated ISDM lines and really top dedicated computers and big screens and it worked superbly it has the potential. The question is whether we have the resources, the people involved in the group, and we've allocated enough money to it. I mean, actually, I have, I've put specifically in the grant money for video conferencing, but I don't think any of the others have. So there might be precisely this problem you were saying, the bureaucracy about will they let the others spend money on this if they haven't budgeted for it. but I think that was one of the organizational issues we wanted to get set at this Trieste meeting that's when we were going to start we were going to talk about how often do we need to meet and what fora how are we going to communicate on a daily weekly, monthly basis

1:20:00 so you said right at the beginning that there have been and getting the network started, now that it's been granted, money has been granted and so on. And is that mostly connected with the paucity of suitable candidates for positions? No, completely unrelated. No, this is just the EU being short of personnel. They're just way behind in processing the grants. When we last spoke, they still had 70 grants to process, and they were saying they didn't have enough secretarial support and that they had brought in some temporary support and they were hoping to process this remaining batch of grants ASAP and we said, please, can you have it ours done, prioritise it because we have this conference and everything's set up for this. They said they'd see what they could do and I've heard nothing since. I think they'd seen what they could do and it wasn't anything that was any use to us. So I guess that's particularly interesting. the, it's the you who's delaying the thing but they also won't let you, let you spend any money ahead of time. That's right, absolutely, yeah. No, they're very, very aggressive about that sort of bureaucracy aspect of things. Science is not the most important thing by any means. The, uh, uh, when I, you mentioned that there were, there was an earlier version of the of the grant yes so the but that one was turned down yeah it was turned down and it was ridiculous because it said one of the main criticisms was that it wasn't timely and that seems to me absurd it's all this money that's been invested in kit to look for you know with interferometers to look for gravitational waves and it wasn't the support to say well you know one of the ways we need to detect them is to model them and have templates that we can represent and they said it wasn't particularly timely it wasn't particularly significant or something like that i mean we knew we'd score low on some aspects like you know commercial exploits ability and all that sort of stuff so it was a bit of a surprise and the person who was in charge of it who was at Jena in what used to be East Germany was very dispirited by the fact that it

1:22:30 was unsuccessful and didn't want to continue and it was my suggestion that we reformed and had another go and that we did it under leadership with Ed Seidel Ed Seidel's head of numerical relativity at Potsdam and he's taken on a lot of work and put together a very good application in this time of success now to us having a postdoc and some travel money significant it makes a big difference to our group to him it's peanuts they've got postdocs coming out of their ears and grants of all sorts money's from here there and everywhere it's a huge institution I'm very grateful to him that he's putting a lot of time and effort to make this thing happen but it's not of that great significance to him or his group in the sense of one more postdocs neither here nor there but I mean he's very keen on the collaborative effort have you seen I don't know if it got published, but I got a pre-print of a paper by him, was it Wymo Swain and so on, about captors. Yeah, I don't think I've looked at it yet, although I've... Well, it goes on and on about how we're in a new era because of the fact that it needs such a welter of different expertise, from computing to, you know, microphysics specialities in relativity and all sorts of things that it really has to be a collaborative effort with a lot of personnel involved. There's no one individual, no one group that could really take this on. So that's what the paper says. That's his kind of message. And I guess in that sense, the network probably has more significance than the funds for one postdoc in the collaboration that should flow from it. Right. And you mentioned that your experience suggests that it might be difficult to really achieve some of the collaborative goals and practices that have for itself, where somebody should be able to slot their piece of code in without knowing too much about what the other bits of code are doing. Well, I'm only guessing. I mean, I'm just extrapolating from sort of little bits and pieces of computer algebra, and it may not be the case. Of course, we live in a different environment now where, you know, email makes instantaneous

1:25:00 contact much more easy and where, you know, I have, now I'm a professor, I can phone up my colleague if I need to and not have to sort of go to the head department and get special permission to make a phone call and where in terms of computing I don't have the lowest priority going where my turnaround was four days when I did my computer algebra development put the program in and get it about four days later and if I wanted to do any better than that then I had to go in three o'clock in the morning on a Sunday and run the machine myself you know it was all that sort of thing and I think video conferencing is another step forward in terms of you know we had a go at this seminar weekly seminar between two groups one member from their group would do it one week in Germany, we would do it the next week and it worked really well and we would have continued it were it not for problems with technology I mean what happened is that Niels my colleague is the man who wanted to exploit this put in a bid for some equipment we got two PC cameras and stuff in the meantime his PC has gone wrong, it won't support poor support from, I mean, we've only got one guy that does it for the whole of the faculty. There's 50 people and one guy that looks after that support ordering equipment and the rest of it. It's too much for him. He's had very poor support and, you know, he's had a computer that wasn't functioning and couldn't get on with the job. So the whole thing is stalled. Now, it's absurd that it should be like that because we're talking about you know small hundreds of pounds to have these things fixed but you know that's been the reality of it in the meantime you have to get on with the you know what facilities you've got but it is interesting i'm interested in this idea of collaboration yeah different groups and so on so from your point of view you think that given the with the use of email and uh greater access to talk communications and a particular video conferencing that actually it might make this kind of collaboration i think so especially as we've got i mean we've got a lot of money for travel so i am anticipating a lot more meetings and you know even things like sending postdocs off here there and everywhere for periods to collaborate if they need to i mean i think the idea is that you're supposed to spend like half your time in another institution as opposed to anyway which suits i mean that works particularly well with nils because he has a close collaboration with a Greek partner, and they're building up quite a good group actually, Thessaloniki.

1:27:30 And this guy who applied to us was very attracted by the end of the project that Thessaloniki were doing, so we said, well, you know, come here and we'll send you to Thessaloniki for a sizable period of your contract. Niels is much more of a jet setter than I am, I mean he's always flying off here and everywhere, you know, he comes out of a long post-doc period where he has flitted around the world, has lots of contacts, is used to that working environment. So he has a lot of ongoing collaborative research projects. So he's used to that, whereas that's not been as a sort of an average academic in an average British university. Funds for travel have been pretty scarce. It's pretty hard. Even if you had it, you couldn't get away because you have too much teaching duties anyway. Since he's coming to post, I've tried to protect him as far as I could from duties and allow him to do this, get up and go at short notice if necessary. Because we live in a slightly different culture now, you know, it's much more research orientated and that is given priority whereas it wasn't for the last 25 years, my 30 years here, the first 25, it wasn't like that. So there's certainly a cultural shift. And we are working more with research-inflected institutions rather than teaching ones. I mean, post-time is research only. If you work with standard universities then people have teaching commitments and administrative commitments as well. It makes them flying off at short notice very difficult, whereas these guys can come and go at the drop of a hat we have our closest collaboration here is with the Portsmouth group who we get on with very well they are much more cosmologically inflected, that's mathematical cosmology is their main thing, but they do have interest in this, which is going Philippus Papadopoulos is one, there's a guy called Ungarelli who's involved in computational wave detection and data analysis associated with that about my third year research student he's actually on a kind of week to week basis being supervised largely by Papadopoulos in Portsmouth so he goes there on a Monday he works with him, he comes back and on Tuesday he reports to Niels and I

1:30:00 and James Vickers about what they discussed and we move the work and that was working quite well on a collaborative level and Portsmouth has just last Friday submitted a GIF Anything to join infrastructure from there? It's some money that the government has given, which I think is actually distributed through the research councils for large building projects. And they have put in a bid for a centre for gravitation and cosmology at Portsmouth. So it's a research centre. And Niels and I are co-applicants with them. So we're sort of signatories of support, essentially. but this would be very good if it comes about because I've seen the designs it's a lovely building, it's a really nice part it's got rooms places for people to come, visit, stay it's got good teaching facilities now they're much more research oriented recently they took on three new staff as research lecturers who don't have to teach for the first three years of their appointment and maybe might not even have to teach subsequent to them. And they've got a lot of success with money for post-docs. They've got two Marie Curie-funded post-docs. They've got one EPSRC senior research fellowship, and so on. So they are going much more in a research-orientated direction. And, you know, we're close geographically and sort of culturally, perhaps, to them. And I think that will feed into our approach as well. Is that space aspect another feature that sets the more research-infected places apart, having the space to accommodate? Yeah, absolutely. We're full up here. I don't know how we can cope next year. We've got two new research fellows and more research students. put them. So either we're going to have to double up or they're going to have to go into another floor, which is not the way to go. I mean, you know, when I was made head of the group, I fought long and hard to get some offices near each other. And it's worked fantastically. I mean, you know, there's so much collaboration that goes on now on a day-to-day

1:32:30 basis. You know, there's science going on in every corner of this corridor. It just didn't happen when there was one room here one room there and one room there in the building yeah so i think i will go in the direction of just squeezing everybody into rooms if necessary to keep up the interactivity it will be a slightly impoverished environment in fact if you go into our research students room you see sometimes it's really crowded you just can't get in people talking to the board and so they try i think they they try different rotas of times when they're in there. I originally set it out for six students, but I mean, you can't get six in. Five is crowded. Yeah, which I think, you know, Portsmouth will have a lovely environment if they are successful. It's a long shot, but I think they might actually pull it off. It's a very impressive scientific case if you read it. It's very good. And it's the and I here, Sathya at Cardiff, you know, and three in London, Malcolm Cullen, Bernard Carr, James Lidsey. So, you know, between the four institutions it's a reasonable spread of classical and cosmology. From the point of view of applying for these larger type of research grants like I suppose the network and things like the Centre and Portsmouth, although I don't know if the Portsmouth one ties into that at all, do big projects like say gravitational wave detector projects like Virgo, so do they help matters by raising the profile? Absolutely, yeah. Raising the scale of activity, raising the monies that are involved. Absolutely, yeah. I mean, when you look at the British involvement in these things, in terms of personnel, it's impressive. In terms of their contribution to the project, it's impressive. In terms of money, it's laughable. I can't remember recently I was on a Peep Park rolling grant committee there was a bid to be involved in Virgo off the top of my head it was something like a quarter of a million pounds so that they are part of the organising panel and the annual I mean the project grants like 200 million pounds absolute peanuts

1:35:00 it's all that Peep Park could afford in fact Peep Park could only afford this down support for other areas. That's essentially how it's going to work. So, you know, there's a lot of British involvement, LIGO particularly, and GO600 is a British-German collaboration, but the actual monies involved are minuscule by international standards. So we really, this country punch above our weight given the lack of resources I mean it's absurd that we should still be in a position of thinking about cutting the pencils in half because there's no money to try and make them go there's talk that there's going to be no money for equipment at all next year it's one of the scenarios there's a redundancy committee just being set up at the university to start firing people because Southampton University even though it's the the third largest income from research councils in the country I think it goes Cambridge, Bureau College, Southampton so we've got a huge amount of money coming in we are still projected to, I remember it was two and a half million deficit at the end of this year so and the towers that be have said no it can't happen, you've got to get your house in order, they've allowed us to go so far in the red this year and be back to balancing the books next year, so it means there's real problems about money And the decision is, what do we do? Do we just stop doing everything? You know, no money for equipment, no money for travel, no money for seminars, no money for anything, no money for telephones and so on. Or do we bite the bullet and sack some people, get them to go one way or the other, and, you know, still try and carry on some sort of level of, supposed to be a research-led institution. You've got to have some money to let that happen. So, difficult times. It is. Well, but at the same time it's interesting to hear about the European network and so Yeah. I'll be very interested to see how it progresses. Yeah, it'll be very interesting from the collaborative point of view. Yeah. I'll be interested to learn more as things go on. Yeah. Thanks very much. Okay, well maybe next time you come, I don't know what you did, but set up a meeting with else because I think he's going to be the main player from now on.