Interview with Wai-Mo Suen

0:00 That's no good. So now it's actually recording something, and it's the 2nd of March at 3.30 in the afternoon, and I'm speaking with Waimo soon. As I'm sure you know, we have been pushing the Neutron Star Grand Challenge effort. The code that we are building, the GR code is the so-called Cactus code. I'm sure you have heard our name many times. It is a code that our so-called NCSA, Poster and WashU collaboration is pushing, developing together with many other groups now joining us using that. This code has the Einstein equation part, treated by many different evolution routines, including both ADM and ADM-like formulation and hyperbolic formulation. So we have different modules in it. And we found switching between different modules in evolution particularly useful. what is caused by what. Whether it's the Virgo method, whether it's the formulation, the form of the equation, or whether it's a bug in the code. So in this code we have this capability of using different formulations and of course also each formulation would come with it many different fields and these are fields of the Virgo method. So we have many many combinations. So this is the vacuum part of the code. We have also been working on, as I described in the Texas meeting last time, on putting generatoristic hydrodynamic equation into the code. We are using so-called high-resolution shock capturing schemes, some modern hydrodynamic methods to co-opt the high programming equation, which governs the evolution of the right-hand side of the Einstein equation, the source term.

2:30 So, again, there are many, we have three different high-resolution shot-capturing methods, also. And they then coupled with space-time, giving us really many, many combinations. On one hand, we found that so many combinations give us a lot of work. On the other hand, we found that very useful at this stage of the work, where we are venturing into something that we have absolutely no idea is really a virgin area not much investigation before that has been done and the this area they need a couple Einstein hydro equation are so complicated the one know what the solution space look like so we have to very carefully using many different methods make sure that it's not a numerical artifacts that we are looking So this has been a major direction that we are pushing. So the flexibility of having coded up different approaches within the same infrastructure is useful diagnostically as a process that you can say. And to understand really what is causing what. Yeah, it sounds like a very interesting approach because as I mentioned earlier I've been looking have different people with different codes and trying to compare the results of different approaches. There must be some advantage. So we try to integrate all of this into one code, one infrastructure, and so we daily do comparison between methods. Well, I remember one of the first things we talked about, so it's an album where people were emphasizing how it used to live out packed as well as from sort of the collaboration point of view, I guess, collaboration infrastructure. Right. It's a very kind of interesting. So, how far along in general is the Neutron Star Run Challenge Program? The new Translibe Unchallenge program has been running very smoothly.

5:00 We have three milestones altogether for speed. The first milestone was to achieve 10 giga flow with the webcam paradigm sign equation. that was passed in 1997, in the early part of 1997. Then late in 1997, we passed a second milestone, which is 50 gigaflop for space-time plus hydro that we passed. And then the third speed milestone, which is in fact the last speed milestone It's 100 giga flop for the full set of mutations and we passed that last May 98. So as far as constructing a code that satisfies the speed milestone is concerned, we have satisfied NASA's repiner. So we're happy about that. we of course having a fast code doesn't mean that it's doing the right thing and can really treat the physical part so we are really working on that we do not have any real contract with NASA of what physics we have to achieve unlike the Breckle Grand Challenge the promised waveform template which they end up failing to hand out, which is unfortunate. It was just too difficult a problem. And for the Newtown Star Grand Challenge, we make sure that we didn't make the same mistake. We didn't promise too much. The first promise, of course, is a fast running code that's capable of solving the service equation, which itself, of course, is already very ambitious. but fortunately we succeeded doing that. Now the only deliverable that we need still to hand in the NASA is a code that can handle, that can run the same set equation with adaptive mass refiner, AMR, which is something that we are working on It's very hard now. Hopefully it should produce that in the early part of the summer.

7:30 For the purposes of the milestones, what kind of a system do you have to run the code with? Is it just a flat space or a single neutron star or some sort of thing? The system, physical system should be. Some sort of collision of neutron star. And we have been using head-on collision on the transgressors. So we are planning on that head-on collision is what we are capable of doing as a person. We are in fact doing grazing collision already. Although it is not required by NASA, we do plan on to do in-spiral collisions, hopefully within the coming few years. we should be able to start doing this problem but it's not part of the contract with NASA. This is something that I want to make sure that the community understand. The community very often have the impression that the Black Hole Grand Challenge promised two Black holes doing this, therefore the Luchun-style Grand Challenge also have promised this, which is not true. We didn't promise that we'll do this problem, we didn't promise that we'll get a waveform template problem, we promised NASA a fast code that contains the Einstein equation and the Hydro equation. And we also tell NASA we plan on using this code to do the neutron star co-license study. But no promise of waveform template, although we hopefully will be able to do that In the upcoming APS meeting, we showed Grayson collision that we think contains physics in it. In the last Austin meeting, I showed a movie on the Grayson collision, but I keep saying that don't leave any science in it yet. And so in this last few months, we have been visiting it and making sure that it really stands. So basically, the movie that you saw, now we are commenting to say that there are physics units. Okay. So, yeah, well, that's pretty hopeful that you are already after grazing.

10:00 Yeah, we are making progress, perhaps even slightly faster than we want. So from NASA's point of view, they want to have the code be made available to the community? Yeah, which is our intention. We have already released our second milestone code, which is capable of... it. We put it on the web that people can download. And we are very happy to see many, many leading places in the regular activity have downloaded a code. I don't know whether they use it or not or whatever. At least they are interested enough to download a code and take a look. We've had some hits. Yeah. That's interesting. In fact, most places in the local community, we noticed that have come in, so we're hoping that it would have some impact. It's interesting, did they talk to you much about it or did they sort of go and download it and sort of take a look at it themselves or did they kind of ring you up to them? Yes, I understand. Some sent us emails and asked, especially our main guide doing this kind of thing is Mark Miller. He is responsible for answering questions about what happens. He has been helping various groups to understand what their course is doing. The Cactus Code, I guess, was primarily developed here in Cotsdam, but it's also being used by the other people in the Alliance, such as NCSA and so on. Yes. In fact, I would say that there are perhaps 10 places even are now using it. The code has such a structure that is very friendly that many people can use it.

12:30 Even the Breckel-Gran Challenge Alliance code has been ported into Cactus as a subroutine to make use of the complication infrastructure at this point. So, you mean the down-cage computation stuff, or just the ADM? The ADM code of the alliance, the Penn State Group in particular, have developed that Black Hole Alliance code and connect it to the Cactus infrastructure so that it can make use of the parallelism. So it's basically developed as a subcontinental form that helps them to run it on parallelism. When you mentioned earlier that a time that can be very useful to compare different formulations running under the Cactus infrastructure to try and see if syrinsis is an effect that shows up somewhere that's really there or maybe different events. Yeah. Is it usually, does it happen that different groups within the alliance are perhaps working with one or other of the particular set of forms, that's sort of the way that they work and then… You mean in the Neutron Star Grand Channel collaboration? Yeah. Or do you find that, for instance, just you and your own group will switch quite happily between the different… Yeah, we, in fact, for the Neutron Star project, it's really mainly carried out by our group here, working with our collaborative process. So it's really all of us using all parts instead of separating one on one part. Of course, the Blackful Brown Challenge, when they bring their code into Cactus as a subroutine, they are really using mainly that part. We haven't used their part yet after this one. Hopefully, we will develop more overlap.

15:00 I'm sure there are components in their subroutine can benefit our research also. So is there, well, having been involved in both the Black Hole Alliance and the Neutron-Star Alliance, what do you think are the main differences between the two? I was going to ask, I guess, as an example, if, from what you said, there was sort of less division of labor in the Neutron Star Alliance. I had the impression of Black Hole that different groups were assigned in certain spheres of responsibility. Yeah, it's clear that the Black Hole Grand Challenge offers a bigger collaboration. They're more interested to involve and they have more money, and they have more money. They have, from where I stand, I see that they, all these groups are in neurobehavioral activity. They kind of overlap with one another a lot in terms of expertise. So, this situation is very different from the Neutron Star Grand Challenge project, or alliance, that we put together. In this Neutron Star work, we have all the groups involved really have different expertise. So we have R1 group, which is living in computer science, the Stony Brook group, which is in the equation of state. We have a group at NCSA URUC. In fact, there are two groups there at URUC. One is in numerical algorithm, elliptic equations, and another group on Newtonian hydrodynamics. And then we have at WashU and POSDEM, more numerical activity groups. Our effort at WashU, we step back, spend more effort on hydrodynamic evolution. While at the same time, we co-developed the webcam part with the Potsdam Guide, whereas

17:30 the Potsdam Guide spent more energy in getting the webcam part, and it's also joining us in the hyper dynamic part. So in many sense, our collaboration is more or less complementary of one another, given expertise to put together. make life a lot easier. Right, so there's less overlap of functions. Yes, instead of competing with one another. We surely don't want to compete with the Equational State people at Stonewall, right? They are doing Equational State for 30 years. And is Cactus useful in that context of bringing together the different areas of expertise from people? Yeah, sure. It's a good collaboration tool. So they provide, presumably, a thorn that works on their part and you can just plug that in? Yes, so the Somibro group would produce an equational state table that we can reconnect into a cactus as a form. The Arbon group would be developing metacomputing capabilities. So they developed something called Robus, which enables the code to make use of many machines at the same time. So they develop robust and robust, it's now connected to cactus also. And how user-friendly does it work out in practice in your experience? I mean, does it happen that when somebody produces a thorn based on their work, that they're able to run it with the various elements of cactus and test it sufficiently can they actually make it available to the other groups in the Alliance that it will be large and travel free or does it usually take one or two iterations? Yeah, sure. Iteration would definitely be needed but this is doable and it's definitely our best experience.

20:00 In the past, we were developing, of course, similar code, what we called G-Code, with 10 of us trying to work on the G-Code, and each of us developing something. Very soon, it becomes just too difficult to manage. You change something, immediately break other people's heart, and you fix it, and then break yours, and there's a lot of conflict like that. And a problem like this has been really kept at a minimum with the good software engineering that we put in, we captured. And how necessary it still is face-to-face contact between the different groups and the alliance? I mean, how often do you go? However, in fact, just last Saturday, we had a meeting here in St. Louis, so-called fourth science meeting of the Neutron-Style-Brun Challenge collaboration. We have people from NCSA in Stonebrook and Potsdam coming over. It's not actually necessary as far as research is lots of time, but seeing each other face to face helps a lot in personal feeling. Sometimes communication through email is always very easy, of course, in doing research you and AI, I think, of course, you write is junk. So communication, only through email, lack a lot of human side. And when we see one another, we can make up the human side. Go out to dinner together, have a beer, that have a lot. So it's more necessary from the point of cohesiveness than really to to do with technical issues. Right, right, right. Of course, technically you can also be dealt with faster with a few of us sitting in the same room. Right. Sure. So are you mostly interested in the neutron star binary problem at this point?

22:30 No, no. We are very ambitious. We want to do neutron stars, black hole, gravitational wave, we keep saying. All three projects, all three fronts, we want to push our head on. In the past two years, indeed we have spent more of our effort pushing neutron stars. And we are just lately trying to pick up, again, black hole work, pushing the so-called, a friend called Eisenbahn Commission, the HBC, treatment for studying black holes. And since our code is capable of doing that, there's no reason to drop it. So we are also working a lot on wave, rotational wave, strong rotational wave. One particularly interesting thing that we have succeeded in doing lately with our positive collaborators is the collapsing gravitational, strong gravitational wave. If you put a wave strong enough, then it would collapse. And we see kind of critical phenomena already for 3D gravitational wave collapse. That has never been done, has never been able to look at 3D gravitational wave. Now that is the first opportunity. And we do see that there seems to be some interesting critical development going on. But we're not ready to bypass yet. Okay. We hopefully will be able to write something within a month. Is there a particular reason why you're concentrating on the neutron star phone in recent times? or was there some particular motivation that you're trying to do? Well, of course. A big reason is that we got the grant from NASA and we have to finish the milestone. So that sort of keeps you? Yes, and also I understand the community have been holding us accountable in getting this program done. So, we have to meet the expectations, basically. And you think the community is largely interested in the problem from the point of view of LIGO or other applications?

25:00 Yeah, they are not interested in the Cartesian Institute. People are interested in it from the direction of gamma ray burst. Really? even, than gravitational wave. Since gravitational wave is more like a Mercury relativity circle, which is a lot smaller than astrophysics at large, astrophysics committee, and a lot of astrophysics, I was in the Texas symposium, a lot of people have been doing neutron star Black Hole Coalition or Neutron Star-Neutron Star Coalition with Newtonian. Yes. So, through the Black Hole, they would just cut out and say that this is Black Hole, everything disappears. Hopefully, we'll be able to do something more rigorous along this line. So there's a lot of interest from the astrophysicists. I guess I'm curious in the question, because I haven't talked to many astrophysicists about their interest in numerical solutions, the neutron star thing. But I was interested to some extent because I wasn't mad at the same thing to try and come up with a gamma-ray burst model out of the star-fishing. So I guess I was curious as to what kind of a model they were looking for. I mean, do you find many of the astro-business people asking you about the blizzing magic results? No, no. Most of them, I would think, are not familiar with their piece of work. This is more like in the mercury activity community. Of course, a lot of people are interested in the neutron star core license simulations as the first models. So they're interested in the possibility of a simulation that provides a model, but it's

27:30 not in the case that they're not waiting for every paper that comes out or anything. Yeah, I suppose everybody, while they are very interested in this, but they also know that it's going to be a long process. Since to really get the gamma ray burst understood, there are just too many components to have to go in. You have to do the pre-node transport write, you have to do MHD, you have to put in generativity. It's just too many components. So it's going to be a long process. It won't be. It's impossible to have a breakthrough next week. Yeah, so it's quite a long way down the road. One reason that we are particularly interested in the neutron star coalescence problem is that I think this served as something that would capture the interest of a lot of astrophysicists, more so than black hole correlations. Yeah, that's what I was just going to say. Yeah, in a sense that one advertisement that I very often pull up these days when I go to talk is to say that computational, general, theuristic astrophysics is good for your health. What I'm saying is that we have a lot of observations in high-energy astronomy, gamma-ray satellite, X-ray satellite, a lot of observations, and there's going to be a lot of data coming in from gravitational wave astronomy. So both of these two branches, when you think about it, you need generativity, you need the Einstein equation, the full set of Einstein equation to do the ball of it, to do the simulation.

30:00 So this is what I call general realistic astrophysics. So it's something that you need to do a full set of Einstein equations. And of course, to really simulate, to really understand realistic observational data, you also need computation. So basically, you need to put your Einstein system massive parallel computer. So this is precisely what we're doing now. So we're trying to bring numerical relativity into computational astrophysics as a tool, which is something that hopefully will make a lot of astrophysicists be interested in. So that it's like an important tool that was not traditionally in the tool bag of computational services. So now we are bringing this tool in and seeing that this tool is a crucial tool for the observation in the coming 10 years. Yeah, it's interesting that you say that it potentially would make GR much more relevant. Yeah, GR much more relevant and in a part there has been a small community thing. Now that we're trying to bring it to Salem and the whole, the large community should pay attention to this. I noticed you have appeared at the first international conference on computational development and signatures. just to amuse ourselves. A few years ago, when we just started the Newtran State Grand Challenge project, I was trying to bring some graduate students in Hong Kong to here to join us for a summer, for a kind of get-together. But to bring them out to get a visa, we need to come up with a name for a conference. That was it. So we make up the whole international conference, so we want this visitor, we need to bring these visitors in, so that they get easier to come and stay here. I hold a joint appointment at the Chinese University of Hong Kong, so I have some graduate

32:30 students back there working with me also in this direction, so I want them to get the I remember talking to Joao Nassil in Potsdam that he stressed that one of the things he hoped for Cactus was to make, he spoke of making his general relativity a tool available to the astrophysics community. I think he wanted to make on numerical relativity, he spoke of wanting to make numerical relativity a tool available to physicists who, you know, were outside the wealthiest research centers, that he saw the idea of cactus being available to anyone who wanted it and that it was parallelized so that in principle if you had a bunch of workstations that you could work on it. that he had a super computer and so on, and specifically his idea I think was that he was from, I guess he's from the Balear Giles in Spain, and so I think he was going back there. And the idea was that in places, universities like that, which just didn't have access to these, the sort of same computing facilities that, you know, would have a place like NCSA, that they could compete on more equal terms with code on campus. Do you see that as a possibility that it would be more within the range of the algorithms, this kind of environment, because of things like campus? I do not really see that, say, someone who has no expertise in numerical activity would be able to just use the code and do a piece of reliable research. That probably would be verified in the future, if it's possible. So, although we try to develop cactus so they make it friendly, make it usable, maybe it's someone easy to pick it up, but still, for someone without a background in is still just too complicated a tool. It's not a black box yet for a long time. And about whether using it on a smaller computer, you'll be able to compete with a big computer.

35:00 That probably would also be difficult. Yeah, Cactus can run even on notebook. But on the notebook, you don't have the resolution that you need and you can do some problem but there's a lot of problem that you cannot do. So intrinsically, you still need a big computer because the problem is that the Einstein equation is too complicated. So, we still need big machines, but there are obviously some problems that you can do on a smaller computer. That, I think, probably is not a major advantage of Cactus. Yeah, it's still a computationally intensive problem. But the major advantage, as I see it, is the collaborative infrastructure. And the fact that it enables so many of us to work together in an efficient manner. I think it's the biggest advantage. The biggest advantage over a previous version of general activity codes. Yeah. So how long do you think it will take in terms of making numerical relativity a tool within the astrophysics community? Is that community in a better position to take off the tool in the sense that it has a sufficient familiarity with relativity that it doesn't have to completely blackbox the code? You know, since you said that, well, I think I took you to say that maybe the astrophysicists would be able to use a cactus to attack problems requiring the American relativity. But obviously, there too will face the problem that they won't be as familiar with relativity as the American relativity is. What a mission is really something like when more and more astrophysicists get interest in, say, phenomena involving strong rotational field that need the ion-sign equation to treat, they will join in into this development, like, for example, they have a lot of experience in writing codes for, say, neutrino transport, for example.

37:30 who has been doing the radiation hydrodynamic for astrophysics problem, now found that they want to do a problem in which you have a strong rotational field. Then they would connect to people with expertise in general activity, with relativistic hydrodynamic, and then work together. And this I would foresee to be happening pretty soon, hopefully within the next few years. There will be a lot of such collaboration going on. So it's in the sense that they would then have access to this relative to the tool. So the model again is taking advantage of the collaboration infrastructure, not so that before you make the code of that box which is so far in the future. That people will still be able to take advantage of it because it's easier for them to be in a collaboration. Yeah, I think that probably is the more likely model. Instead of we put this code, we just scrap this code and just run it. Most likely they run in trouble. I agree. It seems a very interesting model because, again, to go back to this paper I've been working on about some Nazis thing, one of the things that I think is probably a lesson, I suppose I'll have to look at it some more, is that if you really want to sort out a disagreement between two theories or two groups of theories, they're probably, to really convince each other, they probably actually have to collaborate on actually solving the problem together. to be absolutely sure that they would finally agree. And, of course, in the case of Wilson and Matthews, for instance, much of the disagreement was between numericists like Wilson and Matthews themselves and more analytic people like the Caltech group who really weren't in a position to collaborate even if they wanted to because their approaches to this problem were so different. So it would be interesting to see if something like Cactus enables

40:00 I guess I was interested to ask what are the main technical difficulties that you face now in the neutron star binary or black hole binary problem, especially in the neutron star binary problem, what are the main technical obstacles in the web? towards, say, evolving the code in Spiral? What are the main difficulties that you have in front of you, right? No matter which direction you look, you have difficulty, right? All of them is probably this is a very highly time-dependent problem. Tell me a quick question. Namely, sometimes is the, say, with the bound condition It's giving us better trouble. We solve that and then we found that there's a stability of the evolution equation is giving us trouble. Then we improve that, then we go back and see, oh, the initial data is not good enough. We have to improve upon initial data. Then we have to do the initial data better, then we say, oh, the boundaries are now giving us trouble again. So each time you move forward and you encounter some people to solve it, you move forward a little bit, then you're fun. time-dependent in that sense. So the way that all the part elements hang together, you, you know, as soon as you push forward one further, you have to go back to the other. Yeah, of course, this is also a problem of putting it running efficiently on large computers. The computer keeps upgrading. The software is not stable. You have to keep up with it. This is another set of problems. Oh, so more or less changing the porting code For example, we have the Cactus going to upgrade to 4.0. Now our biggest problem is to successfully carry out this 4.0 upgrade, which is the biggest problem facing us now. The reason is that we want to bring adaptive mesh refinement capability into the code. And to do that, we need to change quite a bit of the infrastructure. existing subroutines have to be changed now.

42:30 So we are now, this is the biggest headache at this point. But I'm sure we'll get over this headache. So does that, does that require, does a fundamental change like that require changing the core code and most of the forms? This is a major upgrade. A major upgrade that we have been planning for months. And now we are finally going back to do it and once you've done that you've probably been in position to to that's the final part for NASA right to put in that up the next time then we'll be able to really test that fish my refinement and 3d adaptive mesh refinement, forget about Einstein equation, just hydro. 3D adapted mesh refinement for hydro is already a big problem that is a state of the art problem. So now we also have to compound this with the Einstein equation. So it's something that we will probably take months and months before we can really fully keep up the system and claim that it's working. But for the NASA project, we hope at least we can focus on one specific problem, make sure that we get adaptive mesh refinement working at least for that specific problem that hopefully is within reach in a few months. So you were saying that even just having adaptive mesh refinement with hydrodynamics Big code for, useful code for astrophysicists. Yes, even with the Einstein equation. At present they are field 3D Hydro AMR attempt, but most of them are single processor applications. And now we are going for multiprocessor application with dynamic load balancing and fancy computer science stuff, what you at Angel do is a minute of bucks, I'm sure, once we try it.

45:00 Oh yeah, I guess I was going to say that's sort of the last requirement that you have to fulfill for NASA. How long more does the Alliance program have to run? Officially, the Neutron Star Grand Challenge would end at the end of 1999, not summer of 1999. But we are planning on going for the third round. This is the second round of the NASA Grand Challenge project. There will be a third round coming up and we will go for the third round also. Ok, so you'll apply a new round. Right. And how long is each round? Each round is three years. three. So you're coming up to the end of three. Yeah, we're coming up to the end of three. So, it's interesting anyway that you found yourself being motivated more towards what than, say, the needs of the gravitation or the people? I guess because the community is bigger, actually. Yeah, the community is definitely bigger and more funding opportunities. Yes, sure. Of course, we ourselves are really more realtorists than the Swissists. But I think this is something good that we can offer to a larger community. Do you still see the neutron star lines as having a contribution to make to LIGO in the way that the Bifold branch had? Yeah, sure. We are hoping that, of course, we will compute gravitational wave template. I was kind of curious because of course Kip was asking everyone to bet on getting templates from black pole binaries on the same time scale that Mongo would be hopefully detecting what do you think are the prospects for getting templates from neutron star binaries?

47:30 It depends on how accurate you want. If you do not have any accuracy requirement put on us, we can produce a template now. But then if you say you want me to produce a template which is comparable to releasing what we expect to observe from the sky, I would be very nervous. Because I do not know, for example, how, say, for example, the neutrino transport, how would it affect the whole thing, radiation transport, in general. So part of the problem would be that there's no telling how much of the detail of the normal, messy. Yeah, there will be a lot of messy detail. It's probably very difficult to estimate even how accurate one can get. is to say that we solve the Einstein equation, we solve the hydrodynamic equation, we put in this equation of state that people think one of the state-of-the-art equations of state. We carry out this simulation and we obtain this waveform and we can do the best of our in the sense that we can say this waveform is accurate to what percentage level as determined by the resolution of a grid. But we cannot say confidently the accuracy of this template with respect to the true physics you see. I can estimate the error from the numerical resolution set of equation. But maybe the equation of state all along, we can't do much about that. And if the actual process is strongly influenced by neutrino processes, then also beyond the So you have to kind of do the controller and the unknown control. Right. The, uh, do you expect the, at what, uh, at what stage in the inspire or sort of, if

50:00 you like, at what frequency in terms of the gravitational wave emissions, do you expect the results of numerical simulations to start to show considerable deviations from, say, post-Newtonian relations? Typically, Creeves would estimate that the post-Newtonian calculation will break down at about 10m, when two stars are 10m separated. So hopefully we can pick up at that point, say, from 10m. And whether we will be able to pick up from 10M or not is still an unknown. Right, because you might not be able to run for such a long time. You might not have big enough computer to do this. So, I know that this was something that people had spoken about in the back office too, of having an intermediate bit that could be... In fact, say this is the immediate grand challenge or the immediate challenge. So the same could hold true in the immediate . Yeah, the same could happen in the Translac case. So in that case, if it proves that it's difficult to run except for the last few words and so on, Is it possible that it will be an indefinite time before anyone has a code that can, that will be, you know, able to replicate Wilson and Matthew's code to see if they're really right about the star crushing effect in the sense that maybe it won't be able to run over the same area in whatever time. You mean to prove or disprove persons? I think, in fact, probably we have that capability even now. I think so. But I don't want to jump into that. But you think maybe just doing kind of the grazing? Yeah. I was going to ask about that too. Yeah, we can do grazing, of course. We can also do this and try to look for this effect. But first, I think this is not particularly rewarding in the sense that I strongly doubt that there is such a star-crossing effect in the first place.

52:30 doing this, it's probably teaching us less, since I've come already convinced that it's most likely not this kind of effect, so doing this sort of thing will not teach us much. And there are a lot of interesting physics that we can easily get at with a simpler problem like this one, even head-on collision. There are a lot of physics that one can get from head-on collision, which does not require a long-term integration of this sort. This, we will have to string along quite a few calculations before we can really get So it's pretty competition-expansive, so we want to first understand all the physics involved in head-on collision and crazy collision before we jump into it. So given the amount of time and expenditure, what do you do? Well, of course, if someone comes out and says that they now believe who's on the right or you keep come out and say hey it's like it's possible that they are right then we'll jump into it and really put our resources and try to do that just too many interesting things that we can do yes we don't have enough hands to do them it's just like we develop a really light tool that uh much of which direction we fall into new help that we can find. So we're really excited. People in the group often come in at 3 a.m. or 4 a.m. to work. They didn't really kiss me. It's so exciting. I couldn't sleep last night when I came back to work. Yeah, great. So there's a lot of exciting work.

55:00 So I was curious, since you were saying about numerical relativity becoming astrophysics, in the kind of history work that I did for my thesis, one of the things that seemed to emerge was different cultural traditions within general relativity more of astrophysical than some people more astrophysical, than some people more mathematical background and I know talking to so I was curious to know if you see numerical relativity as falling into one of those one or other of those traditions or transcending them or different people have different backgrounds I mean I guess some people in numerical relativity have quite a mathematical background Now, of course, a lot of numerical relativity work is, as you say, becoming more and more relevant to astrophysics. Do you see these kind of cultural differences playing a role within numerical relativity as its practice, or is it really just sort of a new style of numerical relativity that kind of transcends the differences that, well, just in the case of the problem of gravitational in the 70s especially where the people would be saying, well, we've done this problem, you know, and people say, no, you didn't do it rigorously enough, it's not, you didn't even do it properly. Is that kind of a debate that occurs in American relativity or is that it's not really, you know? Yeah, of course, in American relativity also, there are many styles in doing it. And what I'm saying is that I find, I feel that there's a very fruitful direction for us to develop into. Of course, there are also other full directions, like numerical relativity, when it gets better developed, can also give us a handle on mathematical issues in relativity, like structural singularity, for example, using numerical method to get to it. on various, like one interesting numerical application of our study of black hole has been studying the caustic structure of the human horizon, which is highly academic, right? We don't expect to observe anything like that.

57:30 So it's not in that, not interest in that direction, but it has a lot of interest in in the magnetic direction in relativity. So once we develop this very powerful tool, namely the numerical treatment of the four-iron science system, we can just use it in any direction. And a particular four-four direction, I think, is to bring it to the astrophysics committee and so that they can use it for a lot of observational-related problem. It would be interesting to see how it goes. I guess we can leave it there. I won't take up more of your time. It's been fun talking to you.