Search NASA⌕ Search

SEARCH · Search NASA

Results for “parallel cluster”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 289 records · Page 16
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

The Generalized Green’s function Cluster Expansion: A Python package for simulating polarons

We present an efficient implementation of the Generalized Green’s function Cluster Expansion (GGCE), which is a new method for computing the ground-state properties and dynamics of polarons (single electrons coupled to lattice vibrations) in model electron-phonon systems. The GGCE works at arbitrary temperature and is well suited for a variety of electron-phonon couplings, including, but not limited to, site and bond Holstein and Peierls (Su-SchriefferHeeger) couplings, and couplings to multiple phonon modes with different energy scales and coupling strengths. Quick calculations can be performed efficiently on a laptop using solvers from NumPy and SciPy, or in parallel at scale using the PETSc sparse linear solver engine.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Evaluation of a pair-based, joint-likelihood association approach for regional infrasound event identification

SUMMARY A Bayesian framework for the association of infrasonic detections is presented and evaluated for analysis at regional propagation scales. A pair-based, joint-likelihood association approach is developed that identifies events by computing the probability that individual detection pairs are attributable to a hypothetical common source and applying hierarchical clustering to identify events from the pair-based analysis. The framework is based on a Bayesian formulation introduced for infrasonic source localization and utilizes the propagation models developed for that application with modifications to improve the numerical efficiency of the analysis. Clustering analysis is completed using hierarchical analysis via weighted linkage for a non-Euclidean distance matrix defined by the negative log-joint-likelihood values. The method is evaluated using regional synthetic data with propagation distances of hundreds of kilometres in order to study the sensitivity of the method to uncertainties and errors in backazimuth and time of arrival. The method is found to be robust and stable for typical uncertainties, able to effectively distinguish noise detections within the data set from those in events, and can be made numerically efficient due to its ease of parallelization.

58 GEOSCIENCES↗
DOE OSTI2025

Flexible User-Defined Domain Decomposition in Kilometer-Scale E3SM Land Model Simulation

The Energy Exascale Earth System Model (E3SM) Land Model (ELM) has been extended to kilometer-scale (km-ELM) resolutions, enabling high-fidelity simulations of terrestrial processes at 1 km x 1 km grid spacing. In ELM, domain decomposition partitions the computational domain across processors, ensuring efficient parallel execution. Currently, round-robin decomposition is applied, providing a straightforward way to distribute computational workload. As ELM continues evolving at the kilometer-scale (km-scale), particularly with integrating lateral flow modeling, decomposition strategies must also account for the increased workload and data movement. This paper introduces a flexible user-defined domain decomposition framework, allowing users to customize domain partitioning based on application requirements. The impact of different decomposition strategies is evaluated across various applications concerning computation, communication, and I/O. Results demonstrate that while 1D partitioning yields superior I/O performance, k-nearest neighbors (KNN) clustering effectively reduces inter-process communication overhead. This study lays the groundwork for scalable partitioning in large-scale land surface simulations, enhancing next-generation Earth system modeling.

Wang, Dali [ORNL] (ORCID:0000000168065108)↗
DOE OSTI2026

Comparison of DeePMD, MTP, GAP, ACE and MACE Machine‐Learned Potentials for Radiation‐Damage Simulations: A User Perspective

Accurate and efficient interatomic potentials are essential for molecular dynamics (MD) simulations of radiation damage, gas diffusion, and phase stability in complex ceramics such as LiAlO 2 , especially under extreme conditions relevant to tritium production. Here, we evaluate the performance of six machine-learned interatomic potentials (MLIPs), moment tensor potential (MTP), Gaussian approximation potential, deep potential (DeePMD), atomic cluster expansion (ACE), message-passing ACE (multilayer atomic cluster expansion (MACE) pretrained) and MACE (trained from-scratch), all trained on the same density functional theory dataset with inclusion of tritium. The MLIPs are benchmarked against traditional Buckingham and ReaxFF potentials in terms of energy accuracy, density predictions, thermal equilibration behavior, threshold displacement energy (E d ), tritium diffusivity, and computational cost. Among the models, MTP shows the best overall balance between efficiency and accuracy, with low force and energy errors and realistic E d values for Li and Al. The ACE and MACE (pretrained and trained from scratch) models exhibit high E d (>200 eV) and unphysical pair interactions. DeePMD underestimates Ed due to overly repulsive behavior even at equilibrium distances. All models over-estimate tritium diffusion but the pretrained MACE model behaves well during tritium-diffusion simulations up to 500 K, maintaining diffusivities in the physically consistent 10 −11 m 2 /s range. Finally, we quantify the computational cost of each potential in large-scale atomic/molecular massively parallel simulator, finding that only MTP is more efficient than traditional empirical potentials, while others are significantly more expensive. These findings explain the trade-offs between accuracy and computational cost in MLIP development and provide essential guidance for use in high-throughput radiation damage and gas diffusion simulations in nuclear ceramics.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2025

HydraGNN v4.0

The new version of HydraGNN v4.0 provides additional core capabilities, such as: Inclusion of multi-body atomistic cluster expansion MACE, polarizable atom interaction neural network PAINN, and equivariant principal neighborhood aggregation (PNAEq) among the message passing layers supported -Inclusion of graph transformers to directly model long-range interactions between nodes that are distant in the graph topology Integration of graph transformers with message passing layers by combining the graph embedding generated by the two mechanisms, which allows for an improved expressivity of the HydraGNN architecture Improved re-implementation of multi-task learning (MTL) to allow its use for stabilized training across imbalanced, multi-source, multi-fidelity data Introduction of multi-task parallelism, a newly proposed type of model parallelism specifically for MTL architectures, which allows to dispatch different output decoding heads to different GPU devices Integration of multi-task parallelism with pre-existing distributed data parallelism to enable a 2D parallelization for distributed training Improved portability of the distributed training across Intel GPUs, which has been testes on ALCF exascale supercomputer Aurora Inclusion of 2-level fine-grained energy profilers portable across NVIDIA, AMD, and Intel GPUs to monitor the power and energy consumption associated with different functions executed by the HydraGNN code during data pre-load and training Restructuring of previous examples and inclusion of new sets of examples to illustrate the download, preprocess, and training of HydraGNN models on new large-scale open-source datasets for atomistic materials modeling (e.g., Alexandria, Transition1x, OMat24, OMol25)

Lupo Pasini, Massimiliano [Oak Ridge National Labo↗
DOE OSTI2025

ROOT RNTuple and EOS: The Next Generation of Event Data I/O

For several years, the ROOT team is developing the new RNTuple I/O subsystem in preparation of the next generation of collider experiments. Both HL-LHC and DUNE are expected to start data taking by the end of this decade. They pose unprecedented challenges to event data I/O in terms of data rates, event sizes, and event complexity. At the same time, the I/O landscape is becoming more diverse. HPC cluster file systems and object stores, NVMe disk cache layers in analysis facilities, and S3 storage on cloud resources are mixing with traditional XRootD-managed spinning disk pools.The ROOT team will finalize a first production version of the RNTuple binary format by the end of 2024. After this point, ROOT will provide backward compatibility for RNTuple data. This contribution provides an overview of the RNTuple feature set, the related R&D activities and the long-term vision for RNTuple. We report on performance, interface design, tooling, robustness, integration with experiment frameworks, and validation results, as well as recent R&D on parallel reading and writing and exploitation of modern hardware and storage systems. We will give an outlook on possible future features after a first production release.Collaboratively, the IT and EP departments at CERN have launched a formal project within the Research and Computing sector to evaluate the novel data format for physics analysis data utilized in LHC experiments and other fields. This part of the project focuses on validating the scalability of the EOS storage backend during the transition from the over 25 years old TTree production format to the newly developed RNTuple format, using both replicated and erasure-coded storage profiles.

Blomer, Jakob [CERN]↗
DOE OSTI2020

The impact of the fiducial cosmology assumption on BAO distance scale measurements

ABSTRACT Standard analysis pipelines for measurements of Baryon Acoustic Oscillations (BAO) in galaxy surveys make use of a fiducial cosmological model to guide the data compression required to transform from observed redshifts and angles to the measured angular and radial BAO peak positions. In order to remove any dependence on the fiducial cosmology from the results, all models compared to the data should mimic the compression and its dependence on the fiducial model. In practice, approximations are made when testing models: (1) There is assumed to be no residual dependence on the fiducial cosmology after reconstruction, (2) differences in the distance–redshift relationship are assumed to match a linear scaling, and (3) differences in clustering between true and fiducial models are assumed to be removed by the free parameters used to null the non-BAO signal. We test these approximations using the current standard measurement procedure with a set of halo catalogues from the aemulus suite of N-body simulations, which span a range of wCDM cosmological models. We focus on reconstruction of the primordial BAO and locating the BAO. For the range of wCDM cosmologies covered by the aemulus suite, we find no evidence for systematic errors in the measured BAO shift parameters α∥ and α⊥ to $\lt 0.1\%$. However, the measured errors $\sigma _{\alpha _{\parallel }}$ and $\sigma _{\alpha _{\bot }}$ show a notable absolute increase by up to +0.001 and +0.002, respectively, in the case that the fiducial cosmology does not match the truth. These effects on the inferred BAO scale will be important, given the precision of measurements expected from future surveys including DESI, Euclid, and WFIRST.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Towards Superior Software Portability with SHAD and HPX C++ Libraries

As hardware architectures and software stacks complexity grows, development productivity, performance and software portability, quickly evolve from desirable features to actual needs. SHAD, the Scalable High-performance Algorithms and Data-structures C++ library is designed to mitigate these issues: it provides general purpose building blocks as well as high-level custom utilities, and offers a shared-memory programming abstraction which facilitates the programming of complex systems, scaling up to High Performance Computing clusters. SHAD’s portability is achieved through an abstract runtime interface, which decouples the upper layers of the library and hides the low level details of the underlying architecture. This layer enables SHAD to interface with different runtime/threading systems, e.g. Intel TBB and Global Memory and Threading (GMT). However, current backends targeting distributed systems, rely on a centralized controller which may possibly limit scalability up to hundreds of nodes and creates a network hot spot due to all to one communication for synchronization, and possibly resulting in degraded performance at high process counts. In this research, we explore HPX, the C++ standard library for parallelism and concurrency, as an additional backend in support of the SHAD library, and present the methodologies in support of local and remote task executions in SHAD with respect to HPX. Finally, we evaluate the proposed system by comparing against existing backends of SHAD and analyzing their performance on C++ Standard Template Library algorithms.

Wu, Nanmiao↗
DOE OSTI2018

Time-dependent behavioral data from zebrafish reveals novel signatures of chemical toxicity using point of departure analysis

High-content imaging of embryonic zebrafish behavior can be used as a screening approach to rapidly evaluate the relative potential for chemicals to cause toxicity. However, most statistical methods applied to these data so far transform movement values to incidence-based “hits” and calculate lowest effect levels (LELs), which loses individual fish resolution of behavior and defies hazard ranking due to reliance on applied dose levels. Here, we have developed and parallelized a workflow to calculate benchmark dose (BMD) values from dynamic, high-content zebrafish behavior data that scales for high-throughput chemical screening. To appropriately capture the dynamics of zebrafish response from light to dark stimulus, we summarized time-dependent data using both area under the curve and immediate change at the transition point into two novel metrics that characterized abnormal behavior as a function of chemical concentration. The BMD workflow was applied to calculate BMD 10 values of 1,060 ToxCast chemicals for 24 zebrafish endpoints, including behavior, mortality and morphology. The BMD 10 values provided better precision and separation than LELs for clustering chemicals since they were derived from models that best-fit their concentration-response curves. Clustering analysis revealed behavioral signatures as the most sensitive endpoints. High concordance in chemical activity was observed between ToxCast assays and zebrafish behavioral endpoints, however ToxPi analysis indicated that rankings based on in vitro data was not a reliable predictor of in vivo rankings for lower potency chemicals. The methods developed in this work will enable the use of high-content zebrafish behavioral screening data for BMD analysis in toxicological hazard assessment.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2021

GPU acceleration of rank-reduced coupled-cluster singles and doubles

Here, we have developed a graphical processing unit (GPU) accelerated implementation of our recently introduced rank-reduced coupled-cluster singles and doubles (RR-CCSD) method. RR-CCSD introduces a low-rank approximation of the doubles amplitudes. This is combined with a low-rank approximation of the electron repulsion integrals via Cholesky decomposition. The result of these two low-rank approximations is the replacement of the usual fourth-order CCSD tensors with products of second- and third-order tensors. In our implementation, only a single fourth-order tensor must be constructed as an intermediate during the solution of the amplitude equations. Owing in large part to the compression of the doubles amplitudes, the GPU-accelerated implementation shows excellent parallel efficiency (95% on eight GPUs). Our implementation can solve the RR-CCSD equations for up to 400 electrons and 1550 basis functions—roughly 50% larger than the largest canonical CCSD computations that have been performed on any hardware. In addition to increased scalability, the RR-CCSD computations are faster than the corresponding CCSD computations for all but the smallest molecules. We test the accuracy of RR-CCSD for a variety of chemical systems including up to 1000 basis functions and determine that accuracy to better than 0.1% error in the correlation energy can be achieved with roughly 95% compression of the ov space for the largest systems considered. We also demonstrate that conformational energies can be predicted to be within 0.1 kcal mol -1 with efficient compression applied to the wavefunction. Finally, we find that low-rank approximations of the CCSD doubles amplitudes used in the similarity transformation of the Hamiltonian prior to a conventional equation-of-motion CCSD computation will not introduce significant errors (on the order of a few hundredths of an electronvolt) into the resulting excitation energies.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Measuring and unbiasing the BAO shift in the Ly α forest with AbacusSummit

ABSTRACT The Dark Energy Spectroscopic Instrument (DESI) places sub- per cent constraints on measurements of the Baryon Acoustic Oscillation (BAO) scaling parameters from the Ly $\alpha$ forest. However, no systematic error budget stemming from non-linearities in the three-dimensional clustering of the Ly $\alpha$ forest is included in the DESI-Ly $\alpha$ analysis. In this work, we measure the size of the shift of the BAO peak using large Ly $\alpha$ forest mocks produced on the N-body simulation suite AbacusSummit, which adopt the Fluctuating–Gunn–Peterson Approximation (FGPA). Specifically, we measure the Ly $\alpha$ autocorrelation and the Ly $\alpha$-quasar cross-correlation functions. To mitigate the noise, we adopt a linear control variates technique, reducing the error bars by a factor of up to $\sim \sqrt{50}$ on large scales. From the autocorrelation, we detect a small positive shift in radial direction of $\Delta \alpha _{\parallel }= 0.35~{{\ \rm per\ cent}}$ at the 3$\sigma$ level and virtually no shift in the transverse direction, $\alpha _\perp$. From the cross-correlation, we see a similar shift to $\Delta \alpha _\parallel$, albeit with larger error bars, and a small negative shift, $\Delta \alpha _{\perp }=\sim$0.25 per cent, at the 2$\sigma$ level. We also make a connection with the Ly $\alpha$ forest effective field theory (EFT) framework and find that the one-loop EFT power spectrum yields unbiased measurements of the BAO shift parameters in radial and transverse direction for Ly $\alpha$ auto- and the Ly $\alpha$-quasar cross-correlation measurements. When using the one-loop EFT framework, we find that we can recover the BAO parameters without a shift, which has important implications for future Ly $\alpha$ forest analyses based on EFT. This work paves the way for novel full-shape analyses of the currently observing DESI and future surveys such as the PFS, WEAVE-QSO, and 4MOST.

Hadzhiyska, Boryana↗
DOE OSTI2019

Energy-Screened Many-Body Expansion: A Practical Yet Accurate Fragmentation Method for Quantum Chemistry

We introduce an implementation of the truncated many-body expansion, MBE(n), in which the n-body corrections are screened using the effective fragment potential force field, and only those that exceed a specified energy threshold are computed at a quantum-mechanical level of theory. This energy-screened MBE(n) approach is tested at the n = 3 level for a sequence of water clusters, (H2O) N=6–34 . A threshold of 0.25 kJ/mol eliminates more than 80% of the subsystem electronic structure calculations and is even more efficacious in that respect than is distance-based screening. Even so, the energy-screened MBE(3) method is faithful to a full-system quantum chemistry calculation to within 1–2 kJ/mol/monomer, even in good quality basis sets such as aug-cc-pVTZ. These errors can be reduced by means of a two-layer approach that involves a Hartree–Fock calculation for the entire cluster. Such a correction proves to be necessary in order to obtain accurate relative energies for conformational isomers of (H2O)20, but the cost of a full-system Hartree–Fock calculation remains smaller than the cost of three-body subsystem calculations at correlated levels of theory. At the level of second-order Møller–Plesset perturbation theory (MP2), a screened MBE(3) calculation plus a full-system Hartree–Fock calculation is less expensive than a full-system MP2 calculation starting at N = 12 water molecules. This is true even if all MBE(3) subsystem calculations are performed on a single 40-core compute node, i.e., without significant parallelization. Energy-screened MBE(n) thus provides a fragment-based method that is accurate, stable in large basis sets, and low in cost, even when the latter is measured in aggregate computer time.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2023

Membrane-mediated dimerization of spherocylindrical nanoparticles

Here we present a numerical investigation of the modes of adhesion and endocytosis of two spherocylindrical nanoparticles (SCNPs) on planar and tensionless lipid membranes, using systematic molecular dynamics simulations of an implicit-solvent model, with varying values of the SCNPs' adhesion strength and dimensions. We found that at weak values of the adhesion energy per unit of area, ξ, the SCNPs are monomeric and adhere to the membrane in the parallel mode. As ξ is slightly increased, the SCNPs dimerize into wedged dimers, with an obtuse angle between their major axes that decreases with increasing ξ. However, as ξ is further increased, we found that the final adhesion state of the two SCNPs is strongly affected by the initial distance, d 0 , between their centers of mass, upon their adhesion. Namely, the SCNPs dimerize into wedged dimers, with an acute angle between their major axes, if d 0 is relatively small. However, for relatively high d 0 , they adhere individually to the membrane in the monomeric normal mode. For even higher values of ξ and small values of d 0 , the SCNPs cluster into tubular dimers. However, they remain monomeric if d 0 is high. Finally, the SCNPs endocytose either as a tubular dimer, if d 0 is low or as monomers for large d 0 , with the onset value of ξ of dimeric endocytosis being lower than that of monomeric endocytosis. Dimeric endocytosis requires that the SCNPs adhere simultaneously at nearby locations.

36 MATERIALS SCIENCE↗
DOE OSTI2018

Distributed Louvain Algorithm for Graph Community Detection

In most real-world networks, the nodes/vertices tend to be organized into tightly-knit modules known as communities or clusters, such that nodes within a community are more likely to be “related” to one another than they are to the rest of the network. The goodness of partitioning into communities is typically measured using a well known measure called modularity. However, modularity optimization is an NP-complete problem. In 2008, Blondel, et al. introduced a multi-phase, iterative heuristic for modularity optimization, called the Louvain method. Owing to its speed and ability to yield high quality communities, the Louvain method continues to be one of the most widely used tools for serial community detection. In this paper, we present the design of a distributed memory implementation of the Louvain algorithm for parallel community detection. Our approach begins with an arbitrarily partitioned distributed graph input, and employs several heuristics to speedup the computation of the different steps of the Louvain algorithm. We evaluate our implementation and its different variants using real-world networks from various application domains (including internet, biology, social networks). Our MPI+OpenMP implementation yields about 7x speedup (on 4K processes) for soc-friendster network (1.8B edges) over a state-of-the-art shared-memory multicore implementation (on 64 threads), without compromising output quality. Furthermore, our distributed implementation was able to process a larger graph (uk-2007; 3.3B edges) in 32 seconds on 1K cores (64 nodes) of NERSC Cori, when the state-of-the-art shared-memory implementation failed to run due to insufficient memory on a single Cori node containing 128 GB.

High performance computing, graph algorithms, Dist↗
DOE OSTI2021

The inheritance of anthracnose (Colletotrichum sublineola) resistance in sorghum differential lines QL3 and IS18760

Anthracnose caused by the fungal pathogen C. sublineola is an economically important constraint on worldwide sorghum production. The most effective strategy to safeguard yield is through the introgression of resistance alleles. This requires elucidation of the genetic basis of the different resistance sources that have been identified. In this study, 223 recombinant inbred lines (RILs) derived from crossing anthracnose-differentials QL3 (96 RILs) and IS18760 (127 RILs) with the common susceptible parent PI609251 were evaluated at four field locations in the United States (Florida, Georgia, Texas, and Puerto Rico) for their anthracnose resistance response. Both RIL populations were highly susceptible to anthracnose in Florida and Georgia, while in Puerto Rico and Texas they were segregating for anthracnose resistance response. A genome scan using a composite linkage map of 982 single nucleotide polymorphisms (SNPs) detected two genomic regions of 4.31 and 0.85 Mb on chromosomes 4 and 8, respectively, that explained 10–27% of the phenotypic variation in Texas and Puerto Rico. In parallel, a subset of 43 RILs that contained 67% of the recombination events were evaluated against anthracnose pathotypes from Arkansas (2), Puerto Rico (2) and Texas (4) in the greenhouse. A genome scan showed that the 7.57 Mb region at the distal end of the short arm of chromosome 5 is associated with the resistance response against the pathotype AMP-048 from Arkansas. Comparative analysis identified the genomic region on chromosome 4 overlaps with an anthracnose resistance locus identified in another anthracnose-differential line, SC414-12E, indicating this genomic region is of interest for introgression in susceptible sorghum germplasm. Candidate gene analysis for the resistance locus on chromosome 5 identified an R -gene cluster that has high similarity to another R -gene cluster associated with anthracnose resistance on chromosome 9.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Wide-angle effects for peculiar velocities

ABSTRACT The line-of-sight peculiar velocities of galaxies contribute to their observed redshifts, breaking the translational invariance of galaxy clustering down to a rotational invariance around the observer. This becomes important when the line-of-sight direction varies significantly across a survey, leading to what are known as ‘wide-angle’ effects in redshift-space distortions. Wide-angle effects will also be present in measurements of the momentum field, i.e. the galaxy density-weighted velocity field, in upcoming peculiar velocity surveys. In this work, we study how wide-angle effects modify the predicted correlation function and power spectrum for momentum statistics, both in autocorrelation and in cross-correlation with the density field. Using both linear theory and the Zel'dovich approximation, we find that deviations from the plane-parallel limit are large and could become important in data analysis for low-redshift surveys. We point out that even multipoles in the cross-correlation between density and momentum are non-zero regardless of the choice of line of sight, and therefore contain new cosmological information that could be exploited. We discuss configuration space, Fourier space, and spherical analyses; providing exact expressions in each case rather than relying on an expansion in small angles. We hope these expressions will be of use in the analysis of upcoming surveys for redshift-space distortions and peculiar velocities.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2026

FitCache: A Transparent Drop-In Framework for Multi-Tier Caching to Accelerate Distributed Deep Learning Workloads

Training in Deep learning (DL) remains highly compute- and data-intensive, with I/O becoming a critical bottleneck as models and datasets scale. Recent studies report that data loading can dominate training time, especially on large-scale HPC systems with shared parallel file systems (PFS). Existing caching approaches either rely on single-tier designs or require intrusive modifications to training pipelines, limiting their portability and effectiveness. In this work, we present FitCache, a transparent drop-in framework for multi-tier caching to accelerate distributed DL training by coordinating fast local memory (e.g., DRAM, Persistent Memory (PMem)) and NVMe as hierarchical caches atop PFS. Our design adapts to hardware diversity, i.e., if NVMe is missing, memory transparently acts as a caching tier, ensuring stable performance. FitCache transparently intercepts I/O requests and issues concurrent fetches across all tiers, returning data from the fastest responder without centralized metadata or static redirection paths. FitCache adapts to dynamic workloads and heterogeneous clusters while maintaining POSIX compatibility. Experiments on Frontier (2048 GPUs) and smaller research clusters show that FitCache reduces training time by up to 40% and per-batch I/O latency by up to 71.6% compared to Lustre Orion PFS, offering a drop-in solution for scalable DL training.

Hu, Guangxing [ORNL] (ORCID:0009000283203614)↗
DOE OSTI2020

Self-sustaining sound in collisionless, high- β plasma

Using analytical theory and hybrid-kinetic numerical simulations, we demonstrate that, in a collisionless plasma, long-wavelength ion-acoustic waves (IAWs) with amplitudes $δn/n_0≳2/β$ (where $β\gg1$ is the ratio of thermal to magnetic pressure) generate sufficient pressure anisotropy to destabilize the plasma to firehose and mirror instabilities. These kinetic instabilities grow rapidly to reduce the pressure anisotropy by pitch-angle scattering and trapping particles, respectively, thereby impeding the maintenance of Landau resonances that enable such waves’ otherwise potent collisionless damping. The result is wave dynamics that evince a weakly collisional plasma: the ion distribution function is near-Maxwellian, the field-parallel flow of heat resembles its Braginskii form (except in regions where large-amplitude magnetic mirrors strongly suppress particle transport), and the relations between various thermodynamic quantities are more ‘fluid-like’ than kinetic. A nonlinear fluctuation–dissipation relation for self-sustaining IAWs is obtained by solving a plasma-kinetic Langevin problem, which demonstrates suppressed damping, enhanced fluctuation levels and weakly collisional thermodynamics when IAWs with $δn/n_0≳2/β$ are stochastically driven. We investigate how our results depend upon the scale separation between the wavelength of the IAW and the Larmor radius of the ions, and discuss briefly their implications for our understanding of turbulence and transport in the solar wind and the intracluster medium of galaxy clusters.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗