WEBVTT 1 00:00:04.650 --> 00:00:23.940 Mark Kushner: So it's my pleasure to introduce our speaker today, Dr tammy ma is the advanced photon technologies program elements meter high intensity laser high energy density science within this photon sciences at Lawrence livermore national lab burger pioneers the use of high intensity lasers. 2 00:00:24.960 --> 00:00:35.100 Mark Kushner: To investigate the novel high energy density states of matter generate energetic beings of particles studied laboratory astrophysics and explore fusion physics. 3 00:00:36.330 --> 00:00:45.090 Mark Kushner: Dr graduated from caltech with her bachelor's and her master's and PhD from the University of California San Diego. 4 00:00:46.050 --> 00:00:49.920 Mark Kushner: she's authored or co authored over 185 journal articles. 5 00:00:50.640 --> 00:01:01.350 Mark Kushner: are currently sits on the fusion energy science advisory committee, providing advice to the US Department of Energy office of science, on issues related to fusion energy level research. 6 00:01:01.950 --> 00:01:13.950 Mark Kushner: she's the recipient of the presidential early career Award for science, engineering the APS Thomas six Award for working quantified hydrodynamic instability mix and I still have two oceans. 7 00:01:14.580 --> 00:01:24.300 Mark Kushner: And the do eat early career researchers award and she served currently serves as Lawrence livermore deputy director for a laboratory director of research to develop. 8 00:01:25.950 --> 00:01:34.650 Mark Kushner: we're honored to have you here at Michigan we'd like to present you with symbolic awesome Thank you so much is more rare than a Nobel Prize oh. 9 00:01:36.660 --> 00:01:36.810 Mark Kushner: here. 10 00:01:43.710 --> 00:01:47.790 Mark Kushner: I am honored Thank you my mask off for the first time. 11 00:01:52.890 --> 00:01:53.160 Mark Kushner: Okay. 12 00:01:54.810 --> 00:02:01.500 Mark Kushner: Well, thank you, everybody, and I am very honored to be invited to give them mitzi seminar. 13 00:02:03.480 --> 00:02:04.740 Mark Kushner: Michelle that's up. 14 00:02:26.880 --> 00:02:32.640 Mark Kushner: Oh, it isn't my screen because i'm projecting So if I do that. 15 00:02:34.290 --> 00:02:34.950 Mark Kushner: said okay. 16 00:02:36.060 --> 00:02:41.940 Mark Kushner: awesome okay cool and and carolyn that looks okay to you i'll say. 17 00:02:44.970 --> 00:02:45.690 Carolyn Christine Kuranz: It looks great. 18 00:02:46.350 --> 00:02:47.640 Mark Kushner: awesome Thank you okay. 19 00:02:48.330 --> 00:03:05.010 Mark Kushner: Right, so thank you all for having me and it's been fun meeting many of you today and seeing old friends, so today we're going to be talking about a new regime of higher energy density or hdd physics Howard coupling high REP right lasers with cognitive stimulation. 20 00:03:06.690 --> 00:03:07.320 Mark Kushner: nope. 21 00:03:08.670 --> 00:03:10.560 Mark Kushner: Okay, not going to do that. 22 00:03:16.560 --> 00:03:26.910 Mark Kushner: And so, first let me start by acknowledging, a very large group of people, and most of these names are actually associated with the high REP right work i'm going to be talking about. 23 00:03:27.420 --> 00:03:36.660 Mark Kushner: But before I do that i'm also going to talk a little bit about our recent results, because they are incredibly exciting and they set the stage for why we're moving to high REP rate. 24 00:03:37.050 --> 00:03:46.200 Mark Kushner: And of course shout out to those of us that and those that help us pay our bills and our funding agencies down there at the bottom and you'll see that. 25 00:03:46.650 --> 00:03:58.680 Mark Kushner: We have many, many collaborators at different institutions and and personally I think that's why our field is just so much fun to work in and because we get to meet people from all over. 26 00:03:59.910 --> 00:04:07.350 Mark Kushner: So the outline for today's talk is that higher energy density experimental physics is starting to move towards this high repetition right. 27 00:04:07.860 --> 00:04:17.340 Mark Kushner: And i'll explain what that means in a couple of slides and because these next gen high REP right laser facilities are already coming online around the world. 28 00:04:17.700 --> 00:04:23.640 Mark Kushner: And this is absolutely an opportunity to accelerate the rate of knowledge acquisition in hdd. 29 00:04:24.540 --> 00:04:41.160 Mark Kushner: So what we're working towards is not just doing experiments faster, but really a full integrated hdd capability, which means feedback control loops within the laser experiment and simulations as well to simultaneously optimize the entire system in real time. 30 00:04:42.180 --> 00:04:52.440 Mark Kushner: To make this a reality we are relying on a number of emerging technologies that we get to leverage, and this includes a new diagnostics automated analysis active feedback. 31 00:04:52.890 --> 00:05:06.600 Mark Kushner: And also novel cognitive stimulation techniques and then like I said, these collaborations with scientists, both inside and outside the field and the very exciting thing is this also helps us set the stage for inertial fusion energy. 32 00:05:09.570 --> 00:05:19.440 Mark Kushner: So some of you might have seen some exciting news coming out of the national ignition facility about a month, two months ago now. 33 00:05:20.190 --> 00:05:28.680 Mark Kushner: August, where we demonstrated more than 1.3 mega joules of fusion yield, and this is a significant advance in ice research. 34 00:05:29.580 --> 00:05:41.970 Mark Kushner: The main takeaway was this shot, with eight X higher yield than the previous sniff record and was the first NIF shot to achieve capital gain of over five times and propagating burn. 35 00:05:43.590 --> 00:05:53.160 Mark Kushner: Now if you're not familiar with beneath the national ignition facility sits at Lawrence livermore national lab that's about 50 miles east of San Francisco. 36 00:05:53.730 --> 00:06:07.200 Mark Kushner: It is the world's most energetic laser enabling the study of very extreme hdd conditions and there's many professors here at Michigan and that we get to collaborate with and do some very cool experiments. 37 00:06:08.160 --> 00:06:21.510 Mark Kushner: Now the NIF is 192 beams 1.9 mega joules of energy 500 Terra Watts of power, what that means it's is it's 50 times more power than the entire us electrical grid. 38 00:06:22.110 --> 00:06:28.770 Mark Kushner: But the reason we don't take the grid down every time we take a shot is because we're going to press all of that energy into just nanoseconds. 39 00:06:29.760 --> 00:06:40.830 Mark Kushner: And then, some of them matter temperatures densities pressures are shown here and very, very extreme similar to what we have inside giant planets and the cores of stars. 40 00:06:42.840 --> 00:06:55.290 Mark Kushner: And, and so, if I take the roof off of the facility, you can have a look inside in total this facility is the size of three football fields kind of side by side on lined up there. 41 00:06:56.400 --> 00:07:11.340 Mark Kushner: it's 10s of thousands of optics that we used to amplify the laser energy up focus it all down to the target Chamber down here we're right in the middle sits a little target that's about the size, with the field capsule that sits right in the middle of that whole room. 42 00:07:12.480 --> 00:07:28.500 Mark Kushner: And this talk is not about fusion so we're not going to spend too much time on this, but just one quick slide about the physics here is that Vanessa uses what we call a laser driven whole Rom, which is a highly canister shown here, where we bring the. 43 00:07:29.640 --> 00:07:38.760 Mark Kushner: laser beams in through two holes on the top of the bottom we generate a very energetic X Ray flux that then compresses the field capsule that sits right in the middle. 44 00:07:39.780 --> 00:07:49.500 Mark Kushner: The field capsule and consists of an escalator surrounding DTI some gas and the name of the game is to turn 100 million atmospheres of pressure into about 300 billion. 45 00:07:49.800 --> 00:08:00.540 Mark Kushner: Such that you are so hot and so dense that you get propagating burn and you eat up all of that fusion fuel to get more energy out, then you put in with the lasers that's our definition of ignition. 46 00:08:03.450 --> 00:08:13.140 Mark Kushner: So this program steadily advanced or physics, understanding and the technology over the last decade, to try to improve performance. 47 00:08:13.530 --> 00:08:20.280 Mark Kushner: So the plot on the left is fusion yield and kilojoules over time from 2011 to this year. 48 00:08:21.150 --> 00:08:31.380 Mark Kushner: And then on the right we're showing hotspot pressure versus hotspot energy and that we achieved from our deuterium and tritium shots shots where we actually put fusion fuel in. 49 00:08:32.130 --> 00:08:41.460 Mark Kushner: So you can see that in 2011 2012 when we started our experiments we weren't generating all that much fusion you'll just two and a half kilojoules. 50 00:08:42.600 --> 00:09:05.280 Mark Kushner: And then over time we figured out issues with our implosions change the different pole shapes change different targets types, to make progress and and very slowly made progress towards these pervs up here which are many times yield amplification and to get closer and closer to ignition. 51 00:09:06.810 --> 00:09:26.940 Mark Kushner: And so recently in 2021 this is how that most recent shot compared in terms of fusion yield, and so this 1.35 mega joules that we measured coming back out of our implosion was about 70% of the energy laser energy that we put in so. 52 00:09:28.530 --> 00:09:41.910 Mark Kushner: Why did I walk you through all of this well, the point is that it took us over 10 years on the NIF and 171 dt shots to get to this and already, this was not that. 53 00:09:42.780 --> 00:10:00.780 Mark Kushner: I think a lot of us were we were always very hopeful, but we were still taken by surprise that we got to this point, this is building on decades and decades of research infusion research on the NIF to get to this point years and years now, the point here, though. 54 00:10:03.150 --> 00:10:13.410 Mark Kushner: Is that we currently make use of a number of these different premier facilities around the US and the world conduct this for fun hdd science, some of these may look familiar to you guys. 55 00:10:14.610 --> 00:10:23.130 Mark Kushner: And until recently to do a shady science we focus on these very large energetic drivers and they are mostly single shot. 56 00:10:23.580 --> 00:10:38.850 Mark Kushner: Meeting shot per hour shot every couple hours on the NIF, we can only do one shot every six to eight hours that's set by the cooling of the laser and set by how long it takes to pull your film out. 57 00:10:40.140 --> 00:10:41.730 Mark Kushner: Put in a new target, etc. 58 00:10:42.900 --> 00:10:51.810 Mark Kushner: However, the next generation lasers coming online are already REP rated so greater than 10 hertz 10 shots per second. 59 00:10:54.810 --> 00:11:02.730 Mark Kushner: And some of those laser architectures are shown here, and some have already been built and delivered. 60 00:11:03.270 --> 00:11:08.160 Mark Kushner: Like be helpless laser, this is the high repetition rate advanced head of what laser system. 61 00:11:08.700 --> 00:11:16.350 Mark Kushner: livermore built this it's a pen what laser that can shoot at 10 hertz it was delivered to the eli beam lines in the Czech Republic. 62 00:11:17.310 --> 00:11:34.320 Mark Kushner: we've got some new architectures that we're working on such as the shark laser hundred 50 jewels so higher energy hundred 57 seconds so again a pedal what that should also be able to shoot a 10 hertz and this this laser architecture will be. 63 00:11:35.430 --> 00:11:46.650 Mark Kushner: put on the NBC upgrade at slack and that project recently moved into CD one at under the office size and then new architectures to. 64 00:11:47.190 --> 00:11:52.620 Mark Kushner: Like a boolean meal system that is capable of shooting at 10 kilohertz. 65 00:11:53.520 --> 00:12:04.290 Mark Kushner: And you can see here that, with these new architectures they're also getting more and more efficient so well as apple's only has a wall plug efficiency of about one and a half percent. 66 00:12:04.620 --> 00:12:16.260 Mark Kushner: These newer architectures approach 20% or higher, and so this starts to move us into a very efficient range, should we decide to use lasers for other applications. 67 00:12:16.620 --> 00:12:22.800 Mark Kushner: And it becomes also more attractive to industry as well, for some of the translational science that we eventually would like to do. 68 00:12:24.090 --> 00:12:40.620 Mark Kushner: And then, of course, there is the zoo slicer that you guys are now building here in Michigan, and this is very, very exciting, not only will Zeus be the highest power and US laser system when it's completed, it also can do quite nice shot rates so you can make. 69 00:12:41.670 --> 00:12:43.230 Mark Kushner: A lot of progress on. 70 00:12:44.340 --> 00:13:01.980 Mark Kushner: So the point here is that the US currently beads in laser technology development, even if some of these lasers don't actually sit on our soil, right now, and by pursuing these very new and very innovative laser technologies bell enable additional science opportunities. 71 00:13:03.210 --> 00:13:12.090 Mark Kushner: Because high REP right laser science really represents a very radical paradigm shift in the way we actually do hdd experiments. 72 00:13:13.080 --> 00:13:28.140 Mark Kushner: High REP rate, if you think about it, if you can shoot at 10 times a second versus a shot per hour right it's not only more shots and more data with that you get more statistics you can bring down arrow bars get more precision. 73 00:13:29.310 --> 00:13:39.060 Mark Kushner: Immediately explore more face to face, you have more room for exploration and do more experiments and essentially have a faster response to close mission need checks. 74 00:13:40.080 --> 00:13:53.970 Mark Kushner: So I kind of like this Pippi equation, right here, but your scientific knowledge in hd is the equivalent to your drive and your prob times your repetition rate times time. 75 00:13:57.090 --> 00:14:03.210 Mark Kushner: So for many of us that kind of grew up doing experiments on these very large scale facilities. 76 00:14:04.380 --> 00:14:12.870 Mark Kushner: We sometimes have a little bit of trouble of how moving to high record can actually help us and here are just a couple of examples. 77 00:14:14.280 --> 00:14:27.090 Mark Kushner: of how different types of experiments will benefit so, for example, with high pressure material properties were you trying to fill out a face face clot in this case you're looking at temperature versus pressure. 78 00:14:27.480 --> 00:14:35.820 Mark Kushner: Of magnesium here, out of them mcwilliams paper, if you can do more shots you can more systematically fill out. 79 00:14:36.870 --> 00:14:44.490 Mark Kushner: This area right, I think it took him quite a while to even build up this set of shots to look at that. 80 00:14:45.540 --> 00:14:46.800 Mark Kushner: That shaka goni. 81 00:14:48.090 --> 00:14:51.960 Mark Kushner: In the regime of capacities and radiative properties. 82 00:14:53.400 --> 00:15:03.060 Mark Kushner: What we're interested here is the capacities of materials how radiation transfers and the very. 83 00:15:04.200 --> 00:15:09.000 Mark Kushner: Famous Jim Bailey experiment that he did on the Z machine. 84 00:15:09.780 --> 00:15:22.350 Mark Kushner: found discrepancies in be opacity, is of iron, which actually changed the boundary where the the boundary is between the radiative zone and convective zone of the sun. 85 00:15:22.950 --> 00:15:31.950 Mark Kushner: And we've been trying to replicate that experiment for years and years now on different laser facilities to see if it was an experimental issue or a. 86 00:15:33.270 --> 00:15:35.580 Mark Kushner: issue with our theory and our coats. 87 00:15:36.960 --> 00:15:39.630 Mark Kushner: And so, with. 88 00:15:41.370 --> 00:15:52.290 Mark Kushner: With more shots you can essentially increase your measurement confidence right and and constrain those complex models bring down the error bars that we use. 89 00:15:54.990 --> 00:16:03.990 Mark Kushner: In secondary source particle and radiation beam generation, we use secondary sources such as protons neutrons gamma rays X rays. 90 00:16:04.560 --> 00:16:13.800 Mark Kushner: For a number of different experiments for radiography and probing if you have a high REP right laser you can then generate very high brightness tuned. 91 00:16:14.220 --> 00:16:19.860 Mark Kushner: And these beams are going to be reproducible because, if your laser is shooting at 10 hertz, it has to. 92 00:16:20.430 --> 00:16:27.990 Mark Kushner: be very, very stable you can't have a jumping all over the place, because you'll start damaging things, or you want to understand what your system is doing. 93 00:16:28.860 --> 00:16:42.600 Mark Kushner: So that's important and then in the realm of plasma nuclear physics, you can again, for example, you see secondary sources and have very high average numbers of those particles to pro very subtle or rare nuclear reactions. 94 00:16:44.190 --> 00:16:47.460 Mark Kushner: Okay So what do I mean, though, by a high REP rate system. 95 00:16:48.720 --> 00:16:57.060 Mark Kushner: Well, this is a typical laser hdd experiment first you've got your ginormous laser coming in. 96 00:16:57.450 --> 00:17:07.770 Mark Kushner: hits a target to generate an hdd plasma and then you'll usually be running a number of different diagnostics and some optical some X Ray different particle nuclear diagnostics. 97 00:17:08.430 --> 00:17:20.730 Mark Kushner: Each of these instruments separately collects data and then data analysis is done typically by you guys the graduate students, each one maybe on your own laptop right and going away at it. 98 00:17:22.020 --> 00:17:33.870 Mark Kushner: In a separate system now like I said the laser is going to be shooting at hertz rates are faster, so what you want to do is actually have a laser control system. 99 00:17:34.350 --> 00:17:52.530 Mark Kushner: That is measuring the laser output, with each shot a good set of diagnostics that feeds back into the system and you have control signals to make sure your laser stays in a safe operating regime doesn't blow itself up right doesn't damage itself. 100 00:17:54.690 --> 00:17:56.700 Mark Kushner: This is what we call the laser loop. 101 00:17:58.770 --> 00:18:07.470 Mark Kushner: And then, if your laser is shooting at the super fast rate it's got to shoot something so you bring in your high repetition rate targets rating. 102 00:18:11.130 --> 00:18:16.800 Mark Kushner: Of course you're still going to have your diagnostics, they should be collecting data at a commensurate rate. 103 00:18:17.370 --> 00:18:28.290 Mark Kushner: But now we have to also add a step of data handling and processing it's not quite good enough to have five different laptops just collecting data from the individual diagnostic anymore. 104 00:18:28.800 --> 00:18:38.970 Mark Kushner: You have to be able to archive and be able to track that data to a specific laser shot and track that data to each other. 105 00:18:41.310 --> 00:18:48.780 Mark Kushner: Ideally, you have it, an operational feedback from that step as you're analyzing the data, you can look at the data quality see if it's good enough. 106 00:18:48.990 --> 00:19:05.130 Mark Kushner: See how you can continue optimizing the instrument itself to get better data, and so this might mean changing bias voltages might mean changing filtration but you do have a feedback step there and to make sure that your data stays as high quality as possible. 107 00:19:06.390 --> 00:19:17.250 Mark Kushner: And then, where this starts getting really interested is now, you can do an integrated analysis you're not just analyzing each diagnostic on its own, you can pull it all together. 108 00:19:17.610 --> 00:19:31.140 Mark Kushner: and see from this different multimodal data, some of it might be spectra some of it might be images, some of it might be sailors, but bring that, all together, simultaneously, try to understand what's going on. 109 00:19:33.090 --> 00:19:41.910 Mark Kushner: To do that integrated analysis you'll have to pull in information from exactly what your laser was know the characteristics of the target that you shot. 110 00:19:44.550 --> 00:19:58.440 Mark Kushner: And then, once you have that you can feed that information back into the optimizer to propose input parameters blazer and for the next shots and propose input parameters to the target that you want to shoot for the next shot. 111 00:20:00.180 --> 00:20:02.160 Mark Kushner: And this is what we call the experiments like. 112 00:20:04.710 --> 00:20:11.400 Mark Kushner: Now, the next step that you can also do is add in your hdd codes your simulations. 113 00:20:11.880 --> 00:20:24.600 Mark Kushner: And you can have a direct feedback between your integrated analysis and the simulation codes, where the data that you're collecting can inform and actually improve the physics model in your codes in real time. 114 00:20:25.140 --> 00:20:43.050 Mark Kushner: And also use the codes to inform your analysis your data as it's coming out so that you can understand your experiment in different ways, and then also use what you're learning from these codes and feeding into the optimizer so that is the modeling and simulation. 115 00:20:46.410 --> 00:21:02.640 Mark Kushner: So for us a high REP rate laser suspended hibernate laser science means a fully integrated system that leverages technological capabilities in many domains, and obviously this isn't easy, but these challenges are starting to be addressed. 116 00:21:05.490 --> 00:21:14.820 Mark Kushner: So a number of those emerging technologies and capabilities that we can leverage and to enhance these experiments is shown here. 117 00:21:16.020 --> 00:21:23.700 Mark Kushner: And for us at livermore we're very lucky that we actually have a lot of this expertise in house across the campus that we can pull in. 118 00:21:24.090 --> 00:21:36.990 Mark Kushner: So some of that includes optical technology and not just large optics but also hardened optics if you're going to be shooting at REP right you're sticking a lot of energy through those optics and over time. 119 00:21:37.680 --> 00:21:45.030 Mark Kushner: So they have to be quite damage resistant and you want to incorporate high performance computing in hd codes. 120 00:21:45.630 --> 00:22:03.510 Mark Kushner: target fabrication and then start bringing in things like additive manufacturing 3D printing not only to develop and test novel materials, but also adaptive preparation, what if you could 3D print a different target each time right before you shot. 121 00:22:05.010 --> 00:22:15.300 Mark Kushner: And laser facilities, can you be ties obviously diagnostics and then frontier laser technology, so what we're trying to do now is bring all of this together simultaneously. 122 00:22:18.780 --> 00:22:31.080 Mark Kushner: So currently the lack of appropriate diagnostics, is a major bottleneck to making high REP hdd science a reality, and so, if you take a look at this diagram on the bottom. 123 00:22:31.710 --> 00:22:40.260 Mark Kushner: This is a different version of that complicated loop diagram I was showing you earlier we've marked in different colors kind of the. 124 00:22:40.920 --> 00:22:55.140 Mark Kushner: Development and the technical readiness level of these different subsystems currently we've got the lasers we've got the laser control and we have different forms of target trees, such as. 125 00:22:56.550 --> 00:23:04.500 Mark Kushner: Gas jets liquids, you can do strip targets that you can shoot at pretty high REP rates. 126 00:23:05.550 --> 00:23:17.220 Mark Kushner: We have some forms of active feedback already in development, but what we're missing still in the hdd field is diagnostics and data handling. 127 00:23:18.540 --> 00:23:27.420 Mark Kushner: So we've undertaken a project over the next few years to develop a suite of REP rate capable diagnostics for X Ray and particle beams. 128 00:23:28.230 --> 00:23:45.690 Mark Kushner: That are robust hd environments what I mean by that is when you were shooting at 10 hertz or even much less than that you're generating a huge amount of emp radiation debris So those are all things that you need to take account for and mitigate if you can. 129 00:23:46.860 --> 00:24:01.020 Mark Kushner: And then we want to bring in techniques for high throughput analysis so to do that and we rely on machine learning based algorithms to give us real time analysis of those experimental metrics so that we could eventually feedback in. 130 00:24:01.680 --> 00:24:05.550 Mark Kushner: And finally, active feedback like we've talked about before. 131 00:24:09.300 --> 00:24:19.500 Mark Kushner: And so the plan here is these new high REP diagnostics are currently being deployed and tested at Colorado state university's a left laser. 132 00:24:20.130 --> 00:24:32.700 Mark Kushner: There are 3.3 hertz laser and part of the laser net us facility and and we are using this as a blueprint to not only test and deploy these instruments. 133 00:24:33.180 --> 00:24:47.730 Mark Kushner: But also look at different detector media and test out our analysis routines and then eventually hopefully port that over to other laser net facilities and that might find it useful, and hopefully sinks as well. 134 00:24:50.970 --> 00:25:07.230 Mark Kushner: Okay, and then we are also applying cognitive simulation and to hdd, and this means really an integrated modeling tool that both learns from and guides experiments. 135 00:25:07.980 --> 00:25:20.970 Mark Kushner: And there was actually a report written in 2009 and that the fusion energy sciences division under the science right with asked her, and if you were interested. 136 00:25:21.600 --> 00:25:26.010 Mark Kushner: In machine learning as applied to plasma physics, this is actually a really, really nice report. 137 00:25:26.670 --> 00:25:40.140 Mark Kushner: To kind of pick up and I don't usually tell people to go read reports that's my job, not students jobs but I actually find it super useful because there is a summary. 138 00:25:40.920 --> 00:25:46.860 Mark Kushner: sheet where they actually lay out a lot of new techniques and machine learning that can be applied. 139 00:25:47.580 --> 00:25:53.760 Mark Kushner: For plasma science and for just scientific machine learning in general, and it really helped me to. 140 00:25:54.600 --> 00:26:06.960 Mark Kushner: to learn about machine learning in general, because I don't I still don't know anything about machine learning but anyway, the idea here is to really connect the loop between experimental data. 141 00:26:08.100 --> 00:26:17.850 Mark Kushner: inform your simulated data and to have this loop where your data informs your experimental design and together that gives you better physics understanding. 142 00:26:18.390 --> 00:26:29.550 Mark Kushner: And this is something the simulated data and we don't do a lot of in our field, right now, and it is a lot of work to try to simulate an experiment before you actually do it right. 143 00:26:30.390 --> 00:26:42.600 Mark Kushner: So part of what we're doing with machine learning is generating synthetic data synthetic signals of what we might expect to see on all of our different diagnostics and you use that for training and you use that for. 144 00:26:44.040 --> 00:26:49.680 Mark Kushner: You incorporate that into your overall models so that you have a better understanding of the overall system. 145 00:26:50.340 --> 00:27:01.800 Mark Kushner: Now why cognitive simulation why now it's because there's currently this revolution and computational power and machine learning techniques that makes this all possible. 146 00:27:02.310 --> 00:27:09.210 Mark Kushner: Right now, our group is working on porting over some of our hd codes to gpu machines. 147 00:27:09.600 --> 00:27:22.650 Mark Kushner: And that's so that we can run ensembles of 10s of thousands, hundreds of thousands of simulations much faster than previously possible and only with this confluence of all these very cool technologies is all possible. 148 00:27:25.620 --> 00:27:35.430 Mark Kushner: So he is ripe for the application of novel machine learning and Ai techniques and some of what we are looking at is is listed here. 149 00:27:37.500 --> 00:27:42.210 Mark Kushner: First, what we're doing is multimodal data representations. 150 00:27:44.430 --> 00:27:56.190 Mark Kushner: Like I said, generating synthetic data for all of our different instruments particle diagnostics optical diagnostics X Ray spectra images blah blah all of it together. 151 00:27:56.760 --> 00:28:15.510 Mark Kushner: And to use for training for for representation learning and, in many cases, a more holistic treatment of our diagnostics what I mean by that is very often, we will take a 2d image of something going on on the NIF it's a 2d image of our implosion maybe made with X rays. 152 00:28:16.830 --> 00:28:27.270 Mark Kushner: that's great we look at it, we boil it down to just a few sailors, we might look at the a cemetery of you know, some spiritual implosion. 153 00:28:27.900 --> 00:28:39.060 Mark Kushner: And then you don't do anything else with that we don't know how to but really there's a ton of other information in these 3D images that you can unleash machine learning on to try to. 154 00:28:40.320 --> 00:28:41.940 Mark Kushner: Get deeper into the physics, really. 155 00:28:43.410 --> 00:28:59.790 Mark Kushner: And then there's there's enhanced diagnostics, this is the transfer learning and the virtual diagnostics, that I am talk a little bit about, and this is not just generating synthetic data but generating diagnostics of. 156 00:29:01.950 --> 00:29:10.440 Mark Kushner: Experimental phenomenon that you don't directly measure so, for example in kind of one of the canonical. 157 00:29:11.070 --> 00:29:15.750 Mark Kushner: laser plasma experiments that we like to do you shoot an ultra intense laser to solid. 158 00:29:16.350 --> 00:29:27.000 Mark Kushner: And you generate very energetic electrons those electrons escape out the rear surface and then they pull off the protons off of the rear surface of bed and solid target. 159 00:29:27.540 --> 00:29:36.690 Mark Kushner: And with that you get a beam of protons right and we've been doing these experiments for 20 years and what we do is we say Oh, this is what the laser shot. 160 00:29:37.170 --> 00:29:54.510 Mark Kushner: This is what we measured with the protons and, but what we actually know is the entire interaction physics, is actually governed by the sheath of hot electrons that comes off the rear surface, we have no way of directly measuring that. 161 00:29:55.650 --> 00:30:02.760 Mark Kushner: And we don't measure it as a function of space and time, however, what you can do with machine learning. 162 00:30:03.210 --> 00:30:06.960 Mark Kushner: Is you can build up a huge ensemble of simulations. 163 00:30:07.320 --> 00:30:17.460 Mark Kushner: Where you very the laser you see what results you get and then you can probe the internal physics, so that you can see what's going on in that solid target, you can see what's going on in that sheet fields. 164 00:30:18.000 --> 00:30:30.180 Mark Kushner: And, essentially, you have a virtual diagnostics seeing into the plasma and places that you cannot see before so that once you measure, something you can say ah, now I can backtrack and see what else was going on. 165 00:30:32.190 --> 00:30:38.490 Mark Kushner: we're also doing modeling and optimization under uncertainties meaning uncertainty quantification. 166 00:30:39.510 --> 00:30:48.780 Mark Kushner: Meaning you run a simulation now you actually have no idea what those error bars really are without probably running a bunch more simulations without thinking. 167 00:30:49.890 --> 00:30:54.660 Mark Kushner: So with machine learning, you can get at that uncertainty quantification. 168 00:30:55.830 --> 00:30:59.130 Mark Kushner: right also developing advanced architectures at the edge. 169 00:30:59.670 --> 00:31:11.250 Mark Kushner: Meaning edge computing so for some of the high performance computing that we're talking about you're using very high fidelity 2d 3D codes, I can only run on the big machines. 170 00:31:11.760 --> 00:31:21.930 Mark Kushner: However, for some of the diagnostic analysis that we were talking about you want to do that as close as possible to where the interaction happens. 171 00:31:22.350 --> 00:31:30.690 Mark Kushner: Because once you start running at hertz levels at killer hurts levels, the latency it takes to send that data back and forth becomes an issue. 172 00:31:31.020 --> 00:31:47.190 Mark Kushner: So you really have to start thinking about where you put those computational systems so that you have the most efficient platform that you can and and your don't want your computation to be the long lead item in these loops right because that you can actually control quite well. 173 00:31:48.360 --> 00:31:57.690 Mark Kushner: And then finally on the fly model elevation, meaning that we are actually hoping to improve our codes in real time by feeding in experimental data. 174 00:31:59.880 --> 00:32:08.610 Mark Kushner: So we are combining cutting edge machine learning with operations to demonstrate this real time feedback and integrated experiments on simulation etc. 175 00:32:13.140 --> 00:32:30.660 Mark Kushner: So where are we actually using this now and one area is we are exploring the manipulation of particle acceleration at the Center second and the second timescales for very precise control over these secondary source characteristics. 176 00:32:31.770 --> 00:32:34.440 Mark Kushner: And so, for those of you that. 177 00:32:35.730 --> 00:32:56.700 Mark Kushner: might have worked in the lab with lasers and not even necessarily short pulse lasers and you typically take whatever pulse the laser spits out at you right and you kind of assume that, in time, it is a gaussian and in space that it's a nice well behaved gaussian like. 178 00:33:00.240 --> 00:33:15.690 Mark Kushner: we've now got these new tools which can actually shape the laser pulse as it's coming out on a femtosecond resolution time scale, and so this technology has existed for a nanosecond scale. 179 00:33:16.200 --> 00:33:22.830 Mark Kushner: lasers, for a while now and that's what we do on the NIF before these shorter pulse lasers typically peak the second and less time scales. 180 00:33:23.670 --> 00:33:33.960 Mark Kushner: It has been difficult to do this, but now what we can do is actually shape the laser pulse meaning, you can very precisely modify. 181 00:33:34.320 --> 00:33:41.970 Mark Kushner: The power output as a function of time, and you can make that laser pulse look whatever you want, and so, for example, in this case. 182 00:33:42.570 --> 00:33:50.220 Mark Kushner: We have this very strong picking where we we spit out a very large intensity early on follow them by this longer hump. 183 00:33:51.180 --> 00:34:04.410 Mark Kushner: And when you do this, the early portion of the laser starts interacting with your target in this case let's say it's just a solid foil target it can be whatever material want titanium copper whatever. 184 00:34:05.610 --> 00:34:19.980 Mark Kushner: And this very strong picket fence sets up a particular under dense plasma gradient right and then, when you bring in the pulse later in time. 185 00:34:20.400 --> 00:34:24.630 Mark Kushner: That lower intensity then interacts with the plasma that you've already set up. 186 00:34:25.290 --> 00:34:35.130 Mark Kushner: And what that does is actually modify be interaction physics and modifies the beam of particles that might stream out of the rear surface. 187 00:34:35.940 --> 00:34:49.410 Mark Kushner: cool right okay so that's cool physics and the problem now is there is just an infinite parameter space of different ways and pulses that you can test right, how do you know. 188 00:34:49.950 --> 00:35:05.880 Mark Kushner: You know what's the most interesting, how do you know what will happen as you change this, and this is like one of the most simple shapes pulses that you can have you can start adding more bumps and more digits and and shape that well that's where the the machine learning comes in. 189 00:35:07.140 --> 00:35:08.820 Mark Kushner: To try to. 190 00:35:10.260 --> 00:35:18.630 Mark Kushner: What you do is you then generate a huge ensemble of PIC simulations and where you very the pulse, and you train it. 191 00:35:19.950 --> 00:35:26.040 Mark Kushner: And then you can look for places that would be most interesting to experiment, we test initially. 192 00:35:27.240 --> 00:35:30.210 Mark Kushner: And this is actually. 193 00:35:31.500 --> 00:35:46.620 Mark Kushner: simulation that we did the old style with these very high fidelity PICs simulations, but we can start to see very interesting trends, as you do these shapes pulses, where we get to a much, much higher energies and you would without shaping the pulse. 194 00:35:50.520 --> 00:35:59.100 Mark Kushner: And so that's just one example of how we are using this high REP rate scheme. 195 00:36:00.210 --> 00:36:11.580 Mark Kushner: And the challenges are enormous, and we are looking to collaborate from others, and also learn from other fields and the accelerator community. 196 00:36:12.960 --> 00:36:22.080 Mark Kushner: has been running at kilohertz megahertz for years, and so they've already tackled a lot of these issues and and you know where we just need help. 197 00:36:22.470 --> 00:36:33.690 Mark Kushner: And we need young people, quite honestly, because i'm too old too much um but anyways a commonly voiced concern is that Ai will remove expert judgment right. 198 00:36:34.380 --> 00:36:45.030 Mark Kushner: How do we mitigate this so very often when you do these neural networks, you know you hear people say oh it's a black box, we got an answer, we have no idea what actually happened in a box, that is absolutely true. 199 00:36:45.690 --> 00:36:56.100 Mark Kushner: Right, and so what we actually need is domain aware interpreter interprets hubble and robust scientific machine learning to support this progress. 200 00:36:57.210 --> 00:37:02.580 Mark Kushner: Meaning that we, as a physicist as engineers and scientists. 201 00:37:03.990 --> 00:37:10.500 Mark Kushner: Actively stay involved, at least initially, and while we're bringing all these systems up to make sure everything actually makes sense. 202 00:37:11.640 --> 00:37:21.570 Mark Kushner: We also want a very constant iterative cycle with tight coupling and validation between experiments and modeling you never just want to rely on your modeling alone. 203 00:37:21.990 --> 00:37:34.350 Mark Kushner: Or that machine learning alone to scale to a different parameter space you absolutely need to make sure that you do those experiments and then you use that information to feedback in. 204 00:37:36.000 --> 00:37:48.750 Mark Kushner: We also want to work across facilities and scale between them right and you guys that are experimentalists know you have good days in the lab and bad days in the lab sometimes your laser. 205 00:37:49.080 --> 00:38:01.350 Mark Kushner: starts acting a certain way, and you don't know why right you don't want to go and do now, thousands of shots right when there is some odd systematic that you don't quite understand. 206 00:38:02.430 --> 00:38:13.290 Mark Kushner: But maybe that's Okay, and if you can also go somewhere else to a different facility or change parameters and make sure that things stay consistent between different tool sets. 207 00:38:15.120 --> 00:38:15.510 Mark Kushner: um. 208 00:38:16.740 --> 00:38:23.250 Mark Kushner: Some other big challenges that I, as an hdd physicists had never thought about before embarking on this work. 209 00:38:25.980 --> 00:38:27.690 Mark Kushner: Is are some of them are listed here. 210 00:38:29.040 --> 00:38:34.680 Mark Kushner: machine and data interfaces and communication across those different tools and control systems. 211 00:38:36.210 --> 00:38:41.790 Mark Kushner: Typically, when we shoot a laser shot, you have humans in the control room right but are. 212 00:38:42.870 --> 00:38:51.510 Mark Kushner: You putting in the parameters that you want to shoot at, and then the data comes out you're like me and you yell across the room turn off the laser right. 213 00:38:52.080 --> 00:38:59.550 Mark Kushner: That doesn't work anymore right you want the diagnostic to directly talk to the laser now, so we have to figure out. 214 00:39:00.030 --> 00:39:17.310 Mark Kushner: The control systems, the languages that are common across all of these platforms and, ideally, not just a system that will work for that one single layer system right so that's why we're like to work with the Community to come up with a capability, that is portable and that is standardized. 215 00:39:18.630 --> 00:39:26.220 Mark Kushner: Which is my next point standardization and then also you start generating just volumes and volumes of data. 216 00:39:29.790 --> 00:39:32.790 Mark Kushner: How do you deal with so much data right. 217 00:39:33.810 --> 00:39:44.160 Mark Kushner: In the accelerator community, and very often even on the NIF we throw away our data you take the parts that you think are useful and you throw everything else away. 218 00:39:44.610 --> 00:39:48.930 Mark Kushner: You just don't have the capability to store everything and especially. 219 00:39:49.650 --> 00:40:05.430 Mark Kushner: When you have sensors all over your system right and you're collecting that data there, and you just it doesn't make sense to store it all, so we have to find ways to reliably boil that data down keep what is useful and throw everything else away. 220 00:40:07.020 --> 00:40:10.350 Mark Kushner: Reliability and robustness is incredibly important. 221 00:40:12.270 --> 00:40:19.980 Mark Kushner: Across all your different subsystems you're only as good as your slowest or least reliable system. 222 00:40:20.670 --> 00:40:36.180 Mark Kushner: and distributed debugging becomes an issue when you're talking between different systems, you have to figure out what's going wrong, where some of those systems might be off site, if you are talking to the supercomputer at some other university. 223 00:40:36.780 --> 00:40:51.450 Mark Kushner: And then finally data provenance how often do we forget to date, a page in our lab book right or you miss label a piece of data and we can't have that happen now right, you have to mark. 224 00:40:52.110 --> 00:41:02.070 Mark Kushner: You have to know exactly which flavor laser shot corresponds to which piece of data for each of the different diagnostics and make sure that stays consistent you can't occasionally skip right. 225 00:41:03.330 --> 00:41:13.530 Mark Kushner: But our goal here is to really develop this high throughput capability that can help us to more rapidly respond to those really cool hdd questions across the board. 226 00:41:17.670 --> 00:41:28.470 Mark Kushner: So another way to look at those loops that I laid out previously, is shown here again we've got the laser control loop where we're trying to maximize safe laser delivery. 227 00:41:28.830 --> 00:41:40.680 Mark Kushner: And the experiment performance loop where we want to steer experiments to maximum performance and then finally select experiments using our modeling and simulation loop to minimize the simulation air. 228 00:41:42.120 --> 00:41:52.260 Mark Kushner: And so, some of the key technical opportunities are listed here of areas where we can make a large impact in a very short amount of time. 229 00:41:52.950 --> 00:42:06.840 Mark Kushner: By developing these systems and together by tightly coupling laser technology with state of the art experiments and cogs in the fleet's to what we call intelligent experimentation and faster learning. 230 00:42:11.490 --> 00:42:20.220 Mark Kushner: Now coming back to that niche shot that we talked about earlier essentially why it's so exciting is because. 231 00:42:21.510 --> 00:42:30.570 Mark Kushner: This shot also happened to establish basic scientific feasibility of laser driven inertial fusion energy what we call ISP. 232 00:42:31.170 --> 00:42:46.620 Mark Kushner: And, and in all of these articles and in almost I think every single article that came out about that NIF shot people were really excited to talk about how this lays the foundation for what could potentially be IFC. 233 00:42:50.520 --> 00:43:02.730 Mark Kushner: And this is again a whole nother talk, but realizing if you will require a whole bunch of other advances in science and technology it, it really is one of the grand challenges of humankind. 234 00:43:03.120 --> 00:43:09.900 Mark Kushner: And some of those key components for an IP system, but still needs to be addressed are listed here you'll need. 235 00:43:10.350 --> 00:43:17.310 Mark Kushner: Efficient drivers you'll need target injection Chamber technology tritium process economic cetera so it goes on and on. 236 00:43:18.000 --> 00:43:33.210 Mark Kushner: Now the takeaway, though, is if he is inherently a pulse energy source, meaning you drop a target in you shoot it you get a burst of energy, but then you have to repeat the reaction and it's different from magnetic fusion, in that way. 237 00:43:34.410 --> 00:43:44.880 Mark Kushner: But it's envisioned that a power plant would require reaction on the order of 10 hertz or so so like I said before, the NIF we shoot once every eight hours. 238 00:43:45.480 --> 00:44:04.650 Mark Kushner: For a fusion inertial fusion power plan to be viable you'd have to do it about 10 times a second but we come back to the same scale this hurts 10 hertz scale that we've just touched on, but lasers, are already at and, but we need to bring up the rest of the subsystems. 239 00:44:06.090 --> 00:44:09.900 Mark Kushner: And so there's there's many, many opportunities now. 240 00:44:11.760 --> 00:44:14.130 Mark Kushner: To to push this field forward. 241 00:44:15.180 --> 00:44:16.770 Mark Kushner: By bringing in high referee. 242 00:44:18.660 --> 00:44:30.150 Mark Kushner: So yes, Sir, so i'm just going to wrap up and high repetition rate hdd experimental physics couple to cognitive stimulation will accelerate the rate of knowledge acquisition. 243 00:44:30.570 --> 00:44:38.730 Mark Kushner: And I see if experiments on the NIF are demonstrating very exciting progress, but the scale and complexity of those experiments means progress is slow. 244 00:44:39.450 --> 00:44:45.690 Mark Kushner: The next gen high REP rate laser facilities will provide a very different paradigm hdd experiments. 245 00:44:46.110 --> 00:44:57.660 Mark Kushner: And we're envisioning that this is an integrated hdd capability, with a series of different feedback loops and to make it possible, we of course need to rely on many emerging technologies. 246 00:44:58.230 --> 00:45:04.320 Mark Kushner: and building on those recent great and if results, and we are pushing those technologies for IFA with high REP. 247 00:45:04.620 --> 00:45:17.460 Mark Kushner: playing a major role, so this is a very exciting time for fusion and puzzle research, and I hope all of you, students and stay in the field come help us out, I obviously don't know what i'm doing so. 248 00:45:19.230 --> 00:45:28.650 Mark Kushner: We need all the help we can get, and so I am obligated, of course, to try to do a little bit of advertising of some of the opportunities we have at Lawrence livermore. 249 00:45:29.310 --> 00:45:41.850 Mark Kushner: We do have a postdoc program, and that is quite large we have several hundred and postdocs across the lab and I didn't clean that up, obviously, but you can just Google. 250 00:45:42.570 --> 00:46:02.130 Mark Kushner: And postdocs at livermore and, in fact, in my group we do have a posting up right now that you can find on the livermore website and we are trying to hire and not just on the experimental side, but on the modeling side as well, we need computational physicists we need. 251 00:46:03.300 --> 00:46:06.480 Mark Kushner: target experts, we need diagnostic experts, we need everything. 252 00:46:08.850 --> 00:46:19.260 Mark Kushner: And we also bring in a large number of students every summer and I left when the 2019 numbers on here, because that was the last time we were able to have students on sites. 253 00:46:19.710 --> 00:46:30.780 Mark Kushner: But I think even in 2020 we had over 1000 students so, not just in fusion plasma is not just an add on but all those different areas as well, so. 254 00:46:32.040 --> 00:46:41.220 Mark Kushner: I encourage you to if your advisor allows you to to step out of your group and see what else is out there. 255 00:46:43.980 --> 00:46:59.610 Mark Kushner: And then, finally, we just started a faculty mini sabbatical program where we bring in faculty for somewhere around 123 months and host you at livermore where. 256 00:47:02.250 --> 00:47:06.090 Mark Kushner: You come in and do research with us and we try to. 257 00:47:07.380 --> 00:47:16.110 Mark Kushner: share some of the the numerous resources that we have available, whether it's laser facilities or high performance computing resources, etc. 258 00:47:17.610 --> 00:47:21.900 Mark Kushner: So that's it and i'd be happy to take any question, thank you. 259 00:47:39.870 --> 00:47:40.500 Mark Kushner: Because. 260 00:47:43.980 --> 00:47:46.620 Mark Kushner: You have like a library of data that you're using you're. 261 00:47:48.240 --> 00:47:48.570 Mark Kushner: Like oh. 262 00:47:51.810 --> 00:48:05.370 Mark Kushner: yeah right now we're starting with doing this huge ensembles of simulations, a priori, and some of that is is because of coffee and we haven't actually been able to go do the experiments so we've been focusing on the computational side. 263 00:48:05.910 --> 00:48:10.560 Mark Kushner: um but yeah so we've been doing we've been using epoch actually. 264 00:48:11.760 --> 00:48:14.130 Mark Kushner: To do a huge ensemble of. 265 00:48:15.240 --> 00:48:25.530 Mark Kushner: API laser plasma interaction simulations that we've been training on, and then the idea is to augment that with experimental data as it comes in. 266 00:48:26.670 --> 00:48:39.780 Mark Kushner: But the idea would be we start with this this course setup and samples, and then you can do a different set of ensembles with different fidelity different parameters space. 267 00:48:40.950 --> 00:48:48.030 Mark Kushner: And then, a train across them and then augment with data, in both cases and then continue adding as we go. 268 00:48:48.900 --> 00:49:06.180 Mark Kushner: And these models are even if they don't capture the full physics are never wrong right they are our best understanding that we have right now, so the hope is that we we continue adding and adding and adding one day 2d 3D simulations over time. 269 00:49:08.280 --> 00:49:22.740 Mark Kushner: Reality it'll probably break at some point right but that that's kind of the hope right now yeah, so we are starting with set the simulations done before we start the experiments right now and in and. 270 00:49:23.940 --> 00:49:27.660 Mark Kushner: Ideally actually our next and. 271 00:49:29.100 --> 00:49:38.790 Mark Kushner: The next step beyond this is, we would like to do expect a certain set of simulations before an experiment in form the first experiments we do. 272 00:49:39.330 --> 00:49:55.680 Mark Kushner: collect that data and analyze it in real time but collected over a day and then from that say Okay, we need to run a set of simulations overnight and then use that set of data to run overnight and then inform the next day so that's kind of where we're starting right now. 273 00:50:00.390 --> 00:50:01.470 You know so based on. 274 00:50:04.740 --> 00:50:08.520 Mark Kushner: Your workflow it's also a solution so are you. 275 00:50:09.930 --> 00:50:12.780 Mark Kushner: are just like I was almost cookie. 276 00:50:13.860 --> 00:50:19.080 Mark Kushner: And we are mixing the data and and we do. 277 00:50:20.190 --> 00:50:23.400 Mark Kushner: And I know our computational scientists are actually. 278 00:50:24.540 --> 00:50:36.480 Mark Kushner: And they call it adversarial training, I think, where they they get the different models to compete against each other, and so we're doing we're doing both and we're also using different codes to. 279 00:50:37.770 --> 00:50:45.150 Mark Kushner: So right now we're in the middle of doing a validation from epoch over to webex webex happens to run on gpu machines. 280 00:50:46.230 --> 00:50:53.820 Mark Kushner: So we're validating first of all, to make sure that we're backs it's giving us reasonable answers and it's taken us two years to actually. 281 00:50:54.330 --> 00:51:10.050 Mark Kushner: sort that out, and when it finally seems to be we're now actually using it to do some of this adversarial training as well and it turns out webex runs 40 times faster on the gpu machines than it does on a Poc and if it works, then we can just move over. 282 00:51:14.670 --> 00:51:15.630 Mark Kushner: To this projects. 283 00:51:24.900 --> 00:51:25.650 Mark Kushner: machine learning. 284 00:51:31.110 --> 00:51:40.260 Mark Kushner: As much as we can, we will try to share as much as it makes sense it doesn't always make sense to share right so right now we're using ethics for our control system. 285 00:51:40.860 --> 00:51:53.940 Mark Kushner: Which is, I know what they are using over at eli and and what we are planning to use at the nsc upgrade and so we asked CSU. 286 00:51:54.660 --> 00:52:10.650 Mark Kushner: Colorado state to start using epics as well and, ideally, the tools that we build can be shared and portable because it doesn't make sense to build a diagnostic I can only work at one facility and does a plug and play somewhere else so. 287 00:52:13.470 --> 00:52:20.940 Mark Kushner: yeah and, hopefully, you know next year we'll probably start doing you know more Community sorry guys more workshops. 288 00:52:22.080 --> 00:52:30.270 Mark Kushner: With the Community to try to sort some of these issues out and we would love to work with you and the CS team yeah. 289 00:52:41.730 --> 00:52:42.510 Mark Kushner: it's maybe. 290 00:52:43.860 --> 00:52:44.520 Mark Kushner: Equal pay. 291 00:52:46.350 --> 00:52:47.730 Mark Kushner: Equal Tracy. 292 00:52:49.680 --> 00:52:51.900 Mark Kushner: Yes, yeah it probably will be a. 293 00:52:53.160 --> 00:53:10.860 Mark Kushner: Right now, they already are separate facilities not connected to this at all and we don't think there's really an easy path to add a high REP right laser to the NIF also the NIF is more than 20 years old, now as cool as it is. 294 00:53:11.910 --> 00:53:26.550 Mark Kushner: And so it it's not just the laser that is age it's all of the computational systems that have aged to ominous, and so I I think this would be on four different facility. 295 00:53:30.870 --> 00:53:31.350 So. 296 00:53:33.060 --> 00:53:35.130 Mark Kushner: Interesting song so. 297 00:53:37.980 --> 00:53:38.850 Mark Kushner: You mentioned that he. 298 00:53:41.520 --> 00:53:43.260 Mark Kushner: feel about high REP range. 299 00:53:46.140 --> 00:53:46.920 Mark Kushner: is how are. 300 00:53:48.780 --> 00:54:01.410 Mark Kushner: You can apply that all of these things are fancy the appropriate great so it's a targeted development is slower than things like the laser technology and technology. 301 00:54:02.850 --> 00:54:05.280 Mark Kushner: That can be a bottleneck in. 302 00:54:06.810 --> 00:54:08.490 Mark Kushner: strategy so yes. 303 00:54:11.400 --> 00:54:14.310 Mark Kushner: Can you advance all these things necessarily agree. 304 00:54:15.960 --> 00:54:16.110 Mark Kushner: With. 305 00:54:17.220 --> 00:54:25.590 Mark Kushner: ya know that is a great question, and it is something we struggle with like I said, like, for example with targets, there are. 306 00:54:26.430 --> 00:54:32.070 Mark Kushner: Decent solutions, right now, not eventually what we want to get to but 30% and they'll do. 307 00:54:32.940 --> 00:54:43.380 Mark Kushner: Which is why we decided to focus on diagnostics and the link with, and this development of cognitive stimulation, because those are things that. 308 00:54:44.340 --> 00:54:58.260 Mark Kushner: Just don't really exist right now and and yeah you're right and you again will be limited by whatever is slow, is there hasn't been developed quite yet and and. 309 00:54:58.950 --> 00:55:08.100 Mark Kushner: So you have to do it in a stepped form right with the targets that we have now, you might do a burst mode of let's do 100 shots. 310 00:55:08.610 --> 00:55:16.650 Mark Kushner: And yes, i'm still going to collect the data on my little laptop and sorry there's still going to have to be a lot of manual analysis but. 311 00:55:17.460 --> 00:55:28.950 Mark Kushner: yeah let's a little bits at a time and it turns out it's also very personnel dependent like whatever expertise is in your group at that time so that's why i'm recruiting like crazy. 312 00:55:30.180 --> 00:55:30.720 Mark Kushner: and 313 00:55:32.460 --> 00:55:33.870 Mark Kushner: yeah we'll do what we can. 314 00:55:36.960 --> 00:55:39.510 more powerful waiters fire fire. 315 00:55:40.980 --> 00:55:49.770 Mark Kushner: Safety your words here Is this something good see a sort of a fundamental little big factors. 316 00:55:52.200 --> 00:55:52.890 Mark Kushner: or something. 317 00:55:56.820 --> 00:56:00.720 Mark Kushner: And, and it is definitely a huge consideration. 318 00:56:02.040 --> 00:56:16.470 Mark Kushner: At the REP rates that we're at right now, where okay kind of more or less operating in a conventional mode, that we do now, maybe with a little more shielding but the whole idea is that eventually. 319 00:56:17.670 --> 00:56:23.520 Mark Kushner: I put myself, out of a job right I don't actually need to be at the laser facility to run an experiment. 320 00:56:23.910 --> 00:56:36.330 Mark Kushner: And so that's the whole vision is that we kind of get all of it to be automated and then, and then you can figure out the parts that you really need to shield or the parts, you really need to protect and just kind of. 321 00:56:36.900 --> 00:56:42.570 Mark Kushner: stay away, we might also find, though, that this is just not an efficient way to do it should be in science. 322 00:56:44.130 --> 00:57:00.300 Mark Kushner: We don't know that yet right now, it seems like limitless opportunities but eventually maybe you start collecting junk data because you're not thinking hard enough about and that's something we need to be cognizant of and more is not always the best way to go. 323 00:57:03.870 --> 00:57:04.350 yeah. 324 00:57:24.900 --> 00:57:25.080 Mark Kushner: The. 325 00:57:27.210 --> 00:57:31.890 Mark Kushner: same standard diagnostic a never ending by. 326 00:57:33.180 --> 00:57:35.730 Mark Kushner: Step function, which is it for love. 327 00:57:36.900 --> 00:57:46.920 Mark Kushner: ya know we absolutely need to continue developing better and better diagnostics higher spatial and temporal resolution diagnostics and for our field. 328 00:57:47.550 --> 00:57:57.270 Mark Kushner: For the laser field better diagnosis of the laser itself, because if you don't know what's coming in whatever you're coming at whatever is coming out and. 329 00:57:58.650 --> 00:57:59.640 Mark Kushner: It just gets right. 330 00:58:01.080 --> 00:58:11.100 Mark Kushner: And so there is very much an ongoing effort to build those better diagnose find new detector materials develop new materials. 331 00:58:12.570 --> 00:58:24.780 Mark Kushner: And so that's actually a place where we do a lot of collaboration with universities as well because it's a great PhD projects and to develop a new instrument yeah so. 332 00:58:25.830 --> 00:58:26.340 Mark Kushner: Definitely. 333 00:58:28.350 --> 00:58:28.800 Mark Kushner: Okay. 334 00:58:31.740 --> 00:58:36.330 Mark Kushner: Oh repeat the question okay I don't know what sorry carolyn. 335 00:58:38.670 --> 00:58:42.570 Mark Kushner: wouldn't a question that you mentioned, keep being the expert. 336 00:58:44.790 --> 00:58:49.770 Mark Kushner: I didn't I don't know if I call her Mrs is there any like show progress. 337 00:58:52.050 --> 00:59:09.390 Mark Kushner: Sure, and so the question is how do we continue to incorporate the expertise of the human in the system and so again, as we bring these systems up it's a very systematic step process. 338 00:59:10.350 --> 00:59:31.980 Mark Kushner: don't go off and shoot 1000 targets before stopping and thinking that to see what the data comes out and makes sense right um, but I think that's actually a bigger existential question of how in all of science to we do that, in a way. 339 00:59:33.870 --> 00:59:36.420 Mark Kushner: That actually benefits the science that we're doing. 340 00:59:37.590 --> 00:59:48.330 Mark Kushner: And one of the ways that seems to make sense to us but we're open to other ideas is that direct link between experiment and simulations right. 341 00:59:49.110 --> 01:00:08.010 Mark Kushner: continuously validating the simulation you make a prediction do the experiment see what comes out and see how it matched or did not and then try to understand those discrepancies that is a process that we cannot give up that is so important for field and so. 342 01:00:09.630 --> 01:00:18.390 Mark Kushner: yeah that's that's it and then you know systematically doing different sets of experiments, where you understand the things that you're changing. 343 01:00:18.720 --> 01:00:30.300 Mark Kushner: I think early on a lot of our work will just be replicating what has been done already, in comparing to publish literature and just to make sure we're not making mistakes. 344 01:00:31.830 --> 01:00:35.730 Mark Kushner: But once you do that, then there's there's so much low hanging fruit to go. 345 01:00:41.640 --> 01:00:46.920 Mark Kushner: Real time control your friend to run control and there's batch control. 346 01:00:48.660 --> 01:00:53.640 Mark Kushner: Resources extremely ferocious you do run into real time. 347 01:00:54.870 --> 01:01:02.310 Mark Kushner: When the resource is less precious than you're willing to throw away some resource in order to optimize later. 348 01:01:04.560 --> 01:01:09.120 Mark Kushner: Today, and I see if a particular every shot is precious. 349 01:01:10.260 --> 01:01:25.590 Mark Kushner: But as the REP rate increases that before waste of shots for the batch mode absolutely in order to improve the lives of matches exactly does that change the. 350 01:01:26.760 --> 01:01:34.650 Mark Kushner: It it does so the question is about the scarcity of particular resources and how you can use rapper it to your advantage. 351 01:01:35.640 --> 01:01:43.260 Mark Kushner: Right so right now, when we do a shot on the NIF we characterize the hell out of the target before we shoot it because, once you shoot it it's gone. 352 01:01:43.620 --> 01:01:52.440 Mark Kushner: And you want to know all of the parameters that every dimension that you can every bump every divot on the capsule so that you can simulate it afterwards right. 353 01:01:53.790 --> 01:01:58.800 Mark Kushner: And that is incredibly intensive in work and cost. 354 01:02:00.450 --> 01:02:13.470 Mark Kushner: If you were to be able to shoot a REP rate your targets don't have to be as perfect anymore, you can just say, I made a batch of capsules and they fit within this range and I don't know exactly which capsule I shot on which shot. 355 01:02:14.640 --> 01:02:23.370 Mark Kushner: But over this this batch this was the range of values that I measured afterwards, and that is. 356 01:02:24.090 --> 01:02:40.470 Mark Kushner: Just as valuable because now, you also have good aero bars on your experiment, so that is the trade off that you can do now, and you don't necessarily need everything as perfect, as it was before, because you can repeat shots very quickly and very cheaply. 357 01:02:42.480 --> 01:02:42.750 yeah. 358 01:03:00.750 --> 01:03:02.310 Mark Kushner: Sorry, can you repeat that one more time. 359 01:03:11.310 --> 01:03:15.390 Mark Kushner: Oh, it is it's more general yeah so. 360 01:03:21.240 --> 01:03:27.810 Mark Kushner: they're not every hdd experiment fits REP right um, but there is. 361 01:03:29.160 --> 01:03:35.970 Mark Kushner: A lot of the short pulse experiments that we do today and do fit quite well with rate. 362 01:03:37.980 --> 01:03:38.640 Mark Kushner: So. 363 01:03:42.090 --> 01:03:50.580 Mark Kushner: yeah yeah so so very often, you have a different sets of diagnostics and that's why it was so important for us to try to figure out a standardization. 364 01:03:50.850 --> 01:04:00.060 Mark Kushner: And that's actually an ongoing conversation that we need to have with the Community like what is that standards so that we can build new diagnostics and we can swap them. 365 01:04:00.900 --> 01:04:12.000 Mark Kushner: in and out for different types of experiments, but the idea of that those feedback loop and general techniques we're hoping will be where the applicable. 366 01:04:25.230 --> 01:04:26.070 Mark Kushner: Thank you, everybody. 367 01:04:32.220 --> 01:04:33.420 Mark Kushner: And all the data they generate a. 368 01:04:36.810 --> 01:04:37.830 Mark Kushner: big part of their.