Bulletin of the American Physical Society
64th Annual Meeting of the APS Division of Plasma Physics
Volume 67, Number 15
Monday–Friday, October 17–21, 2022; Spokane, Washington
Session CO05: ICF: Analytical and Computational TechniquesLive Streamed
|
Hide Abstracts |
Chair: Joshua Sauppe, LANL Room: Ballroom 111 B |
Monday, October 17, 2022 2:00PM - 2:12PM |
CO05.00001: Understanding Cryogenic Target Performance on OMEGA Using Statistics-Based Analysis with a 2-D DRACO Simulation Database Duc Cao, Rahul C Shah, Varchas Gopalaswamy, Aarne Lees, Cliff A Thomas, Dhrumir P Patel, Riccardo Betti, Wolfgang R Theobald, James P Knauer, P. B Radha, Christian Stoeckl, Sean P Regan, William Scullin, Timothy J Collins, Valeri N Goncharov We present a statistical model based on advanced 2-D simulations that allow us to understand the origins behind observed fusion yield dependencies coupled with other observables. This model shows that these dependencies can be understood with physics models included in DRACO 2-D (Ref. [1]) and can systemically describe the impact of known perturbations (e.g., beam-port geometry, ice roughness, full smoothing by spectral dispersion imprint up to mode 50, etc.) on a measured yield. Using this method, the impact of unmodeled/unknown degradation sources can also be isolated. The latter is valuable for determining code improvements that will lead to accurate extrapolations of hydro-equivalent designs at multi-MJ laser facilities. These code improvements can then be better refined when including statistical modeling with other observables such as hot-spot size and areal density. This material is based upon work supported by the Department of Energy National Nuclear Security Administration under Award Number DE-NA0003856. |
Monday, October 17, 2022 2:12PM - 2:24PM |
CO05.00002: Automated design optimization for robust, high yield implosions Kelli D Humbird, Luc Peterson, Brian K Spears As we learn more about the sensitivity of recent megajoule (MJ) class implosions on the National Ignition Facility (NIF) to engineering defects and drive perturbations, there is increasing interest in exploring “robust” MJ yield designs. These implosions would be designed to explicitly maximize performance when subject to uncertain or variable conditions. |
Monday, October 17, 2022 2:24PM - 2:36PM |
CO05.00003: Yield variability and robustness of a hydrodynamically scaled-up N210808 Michael K Kruse, Ryan C Nora, Chris Weber, Jim A Gaffney, Kelli D Humbird, Bogdan Kustowski, Luc Peterson, Brian K Spears 1-d radiation hydrodynamics simulations were performed with HYDRA to assess the yield variability and robustness of N210808 as a function of hydrodynamic scale. It is expected that the yield variability initially increases with scale as one moves up along the ignition cliff, followed by a robust burning phase characterized by low yield variability and high burn-up fraction. We map out the variability as a function of scale while taking into account observed variations in the laser pulse delivery of N210808 as well as target fabrication limits on the capsule (e.g., variations in the ice layer thickness). |
Monday, October 17, 2022 2:36PM - 2:48PM |
CO05.00004: Towards Digital Design at the Exascale: Advances in Bayesian Optimization with Neural Networks Luc Peterson, Jayaraman J Thiagarajan, Rushil Anirudh, Yamen Mubarka, Irene Kim, Peer-Timo Bremer, Brian K Spears, Vivek Narayanaswamy The expense and consequence of many scientific and engineering applications, such as inertial confinement fusion (ICF), necessitate the use of digital design, whereby promising configurations are first explored via numerical simulation before being realized in experiments. But many such computer models are themselves significantly expensive, making manual or brute force search intractable. In these situations, surrogate-based optimization techniques, such as Gaussian-Process-based Bayesian Optimization, are commonly used. However, these methods could have their limitations on emerging exascale compute platforms, which could generate sufficiently large enough quantities of high-dimensional design data as to make Gaussian Processes inefficient. As an alternative approach, we are developing novel algorithms and scalable uncertainty metrics that enable Bayesian Optimization with neural networks. These networks can efficiently model response surfaces in high dimensions and with sparse data, making them attractive for digital design at the exascale. In this work, we explain our new models and show how they outperform standard optimization techniques on both analytic and ICF design examples. |
Monday, October 17, 2022 2:48PM - 3:00PM |
CO05.00005: Linearized neural network variability model of megajoule yield shots at the National Ignition Facility Eugene Kur, Jim A Gaffney, Kelli D Humbird, Michael K Kruse, Bogdan Kustowski, Ryan C Nora, Brian K Spears The inertial confinement fusion (ICF) program at the National Ignition Facility (NIF) achieved a record-breaking 1.3 MJ yield from its N210808 experiment. Efforts have now begun on developing a robust, reproducible platform for delivering MJ-class yields. As part of that effort, several repeat experiments of N210808 were made to assess shot-to-shot variability of the design. In this talk, we analyze the set of repeat shots to characterize the variability under various design perturbations. We do this by combining a neural network surrogate trained on simulated (via HYDRA) capsule implosions with a linearized variability model inferred from the repeat shots. We apply the model to an ensemble of perturbed designs centered on N210808, demonstrating the existence of a high-variability region that needs to be avoided for successful implementation of a robust MJ platform. |
Monday, October 17, 2022 3:00PM - 3:12PM |
CO05.00006: How predictive are THD experiments? Bayesian analysis of the ICF data from the variability campaign Bogdan Kustowski, Jim A Gaffney, Brian K Spears, Ryan C Nora, Chris R Weber, David J Schlossberg, Alison R Christopherson, Stephan A MacLaren The 1.3 MJ inertial confinement fusion (ICF) experiment at the National Ignition |
Monday, October 17, 2022 3:12PM - 3:24PM |
CO05.00007: Role of self-generated magnetic fields in the inertial fusion ignition threshold James D Sadler, Christopher A Walsh, Ye Zhou, Hui Li Magnetic fields spontaneously grow at unstable interfaces around hot-spot asymmetries during inertial confinement fusion implosions. The predicted field strengths exceed 5 kT (50 MG) in current National Ignition Facility experiments. We demonstrate this via magnetic post-processing of two-dimensional xRAGE radiation-hydrodynamic simulation data for HDC shot N170601. Although the resulting magnetic pressure is too small to significantly affect hydrodynamics, magnetic fields can also have an indirect effect by reducing and deflecting the heat flux. The rate of hot-spot heat loss is reduced by > 5% in the post-processed data. We derive a model for the self-magnetization, finding that it varies with the square of the hot-spot temperature and inversely with the hot-spot areal density. The self-magnetized Lawson analysis then gives a slightly reduced static ignition threshold, perhaps by ~200eV for typical hot-spot parameters. Time dependent self-magnetized hot-spot energy balance models corroborate this finding, with the magnetic field quadrupling the fusion yield for near threshold parameters. The model shows that self-magnetization effects should increase with yield, becoming a key consideration in recent burning plasma experiments. We also discuss the implications and caveats with these findings. A key caveat is the unknown self-consistent evolution of asymmetries with magnetized heat flux. A future ignition parameter scan with extended-magnetohydrodynamics simulations should address this. |
Monday, October 17, 2022 3:24PM - 3:36PM Author not Attending |
CO05.00008: Robust MJ-class implosions at NIF: a framework and metrics for understanding ICF variability Brian K Spears Inertial confinement fusion (ICF) experiments at the National Ignition Facility (NIF) have culminated in a milestone experiment producing greater than 1 MJ of yield. The next at NIF is to develop implosions that robustly deliver this kind of MJ-class performance so that it can be harnessed routinely for a range of science and security experiments. We are developing new design and analysis tools to help us quantify the robustness of an implosion. We will report on two major thrusts. Thrust one focuses on efforts to quantify variability of existing implosion platforms. The goal is to deliver quantitative confidence limits on floors of neutron yield based on combinations of modeling and experimental observation. The second thrust focuses on developing design improvement paths that increase robustness (reduced variability). We will discuss versions of a new robustness metric that complements traditional ignition threshold metrics. This robustness metric aims to use observed stagnation conditions on a single shot to help infer the likely sensitivity of the implosion to uncontrolled sources of variation. We will discuss tools drawn from simulation, experiment, and AI-driven ensemble methods that enable these types of variability quantification and robustness estimation for the first time. |
Monday, October 17, 2022 3:36PM - 3:48PM |
CO05.00009: Porting the particle-in-cell code OSIRIS to GPU-accelerated achritectures Roman Lee, Jacob R Pierce, Kyle G Miller, Adam R Tableman, Viktor K Decyk, Ricardo A Fonseca, Warren B Mori Furthering our understanding of many of today's interesting problems in plasma physics requires large-scale kinetic simulations using particle-in-cell (PIC) codes. However, these simulations are extremely demanding, requiring that contemporary PIC codes be designed to efficiently use a new fleet of exascale computing architectures, which are increasingly GPU based. We discuss a GPU algorithm for PIC codes which we implemented on the code OSIRIS [1]. Development is currently ongoing, but a limited feature production code based on CUDA C is complete. Performance on GPUs is especially dependent on memory utilization, and maximizing utilization without exceeding capacity is challenging for many of the real-world problems where computational load can fluctuate dramatically in space and time. Our algorithm is unique compared to other PIC codes which have been ported to GPUs in that it includes two important features to overcome this challenge. First, it is being built on the existing OSIRIS data structures and will thus include dynamic load balancing. Second, we make use of a novel custom memory management strategy to avoid unnecessary—and costly—buffer reallocation in simulations where load imbalance is prevalent. We will also discuss performance and strategies for extending the software to run on non Nvidia GPUs. |
Monday, October 17, 2022 3:48PM - 4:00PM |
CO05.00010: The FLASH Code for Computational High-Energy-Density Physics: Recent Additions and Improvements Petros Tzeferacos, Adam Reyes, Edward C Hansen, Fernando Garcia Rubio, Yingchao Lu, David Michta, Richard Sarkis, Marissa B Adams, Abigail Armstrong, Kasper Moczulski, Pericles S Farmakis, Ananya Mohapatra, Mary McMullan, Victor Chang, Niels Vanderloo, Joshua P Sauppe, Anthony Scopatz, Milad Fatenejad FLASH is a publicly available, finite-volume Eulerian, spatially adaptive radiation-magnetohydrodynamic (MHD) code that has the capabilities to treat a broad range of physical processes. FLASH is being developed by the Flash Center for Computational Science to perform well on a wide range of computer architectures and serve a broad user base spanning numerous research communities. Extensive high-energy-density–physics (HEDP) capabilities exist in FLASH, making it a powerful open toolset. We summarize these capabilities, emphasizing on recent additions. We describe several algorithmic improvements and several extended-MHD capabilities that were added to FLASH, allowing modeling of Z pinch, fusion, and magnetized HEDP experiments. We showcase FLASH’s ability to simulate ab initio complex laboratory astrophysics experiments performed by the Turbulent Dynamo collaboration and describe several collaborative efforts with the academic HEDP community, the national laboratories, and industry in which FLASH simulations were used to design and interpret HEDP experiments. This material is based upon work supported by the Department of Energy National Nuclear Security Administration under Award Numbers DE-NA0003856, DE-NA0002724, DE-NA0003605, DE-NA0003842, DE-NA0003934, and Subcontracts 536203 and 630138 with LANL and B632670 with LLNL; the NSF under Award PHY-2033925; the U.S. DOE Office of Science Fusion Energy Sciences under Award DE-SC0021990; and U.S. DOE ARPA-E under Award DE-AR0001272. |
Monday, October 17, 2022 4:00PM - 4:12PM |
CO05.00011: An autoencoder based reduced order model of low density plasma for optimal experimental design Ravi G Patel, William E Lewis, Patrick F Knapp Parasitic losses at Sandia's Z Machine may occur through near vacuum conditions and limit the current supplied to the load. A low density plasma experiment, consisting of a cylindrical target with an evacuated interior, has been developed to study these losses. However, due to the expense of shots at Z, experiments must be selected carefully to optimize information gain. Simulations aid experimental design, but due to high computational costs, reduced order models (ROM) are necessary. We introduce an autoencoder based ROM for this low density plasma experiment and apply it towards experimental design. Our ROM consists of two parts, an autoencoder that encodes the MHD fields in a latent space, and a ResNet that evolves the latent variables in time. This architecture is trained on 2D GORGON simulations of the low density plasma experiment for a variety of vacuum floors and geometries. Using this ROM, we select geometric parameters with different fixed vacuum floor settings that maximally distinguish quantities of interest, the dynamics of the magnetic fields and densities at various locations. |
Monday, October 17, 2022 4:12PM - 4:24PM |
CO05.00012: Interface Reconstruction Using Gaussian Processes for Volume of Fluid Methods Adam Reyes, Marissa B Adams, Abigail Armstrong, Kasper Moczulski, Pericles S Farmakis, Edward C Hansen, Yingchao Lu, David Michta, Petros Tzeferacos We present a novel framework for reconstructing material interfaces on a local stencil |
Monday, October 17, 2022 4:24PM - 4:36PM |
CO05.00013: Mixed Stochastic-Deterministic Density Functional Theory Within The PAW Formalism: An Efficient Approach To Warm Dense Matter Vidushi Sharma, Alexander J White, Lee A Collins Warm dense matter (WDM), believed to form the cores of exoplanets and dwarf stars, has gathered considerable interest recently due to its realization in inertial confinement fusion. Monumental efforts in experiment as well as theoretical/computational modeling have inched toward elucidating the properties and nonequilibrium dynamics of WDM. In computational materials modeling, density functional theory (DFT) is a powerful tool employed for studying systems ranging from just a few molecules to much more condensed phases. However, the cubic scaling of DFT with system size and temperature renders much of the WDM regime computationally intractable. White and Collins have developed a formalism [1] that mixes the stochastic and deterministic algorithms of DFT to study matter at any temperature. The idea behind mixed DFT is to improve efficiency by introducing stochastic orbitals while at the same time maintaining the accuracy of a deterministic Kohn-Sham DFT computation. In this work, we implement projector augmented wave (PAW)-based pseudopotentials in the mixed DFT formalism. PAW potentials are more accurate and computationally inexpensive compared to most other pseudopotentials, and therefore enhance our ability to investigate a wide area in the (temperature, pressure)-phase space of matter. In this talk, we present results obtained with semicore- and core electrons- PAW for carbon systems in the WDM domain, and compare them with simpler, norm-conserving Goedecker pseudopotentials. |
Monday, October 17, 2022 4:36PM - 4:48PM |
CO05.00014: Kinetic modeling for strongly coupled plasma mixtures Lucas J Stanek, Michael S Murillo, Jeff R Haack Kinetic modeling for strongly coupled plasma mixtures Numerical simulations of strongly coupled plasmas play a vital role in high energy density experiments. A common method for simulating strongly coupled plasmas is molecular dynamics (MD) which explicitly integrates the equations of motion for individual particles including correlations between them. Instead of MD, we propose a model based on kinetic theory which includes strong-coupling through a self-consistent electric field and collision operator. Our kinetic model allows us to simulate strongly coupled plasma mixtures at much larger space and time scales than MD. We apply our numerical method to simulate ongoing ultracold neutral plasma experiments where we discuss the impact of different collision operators in our kinetic model. Namely we show numerical results for the Bhatnagar-Gross-Krook and Lenard-Bernstein-Dougherty operators. We also discuss the various treatments for the underlying electron species by evolving the electrons as their own species or via a Boltzmann distribution. |
Monday, October 17, 2022 4:48PM - 5:00PM |
CO05.00015: Hydrodynamic Density Functional Theory Description of Strongly-Coupled Plasmas Christopher M Gerlach, Michael S Murillo, Liam G Stanton Strongly-coupled plasmas, such as ultracold neutral plasmas, dusty plasmas and warm dense matter, can be difficult model, as a complete understanding of the physics relies on both the dynamics and the underlying particle correlations. Density functional theory (DFT) is a natural formalism for describing such correlations but is limited to equilibrium systems. For non-equilibrium systems, hydrodynamic DFT (HDFT) provides a dynamic generalization of DFT that has recently been applied to plasmas [1, 2]. One of the primary advantages of HDFT is that it establishes a direct connection to atomic-scale correlations self-consistently and without the need for an ad hoc equation of state. Here, we explore various choices of correlation functionals in the HDFT model and address some of the computational challenges that arise from the nonlocal correlation effects. Furthermore, we explore the role that correlations play in plasma waves and coupled modes. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700