Search NASA⌕ Search

SEARCH · Search NASA

Results for “parallel coordinates”

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 199 records · Page 11

Synthetic Foveal Imaging Technology

Apparatuses and methods are disclosed that create a synthetic fovea in order to identify and highlight interesting portions of an image for further processing and rapid response. Synthetic foveal imaging implements a parallel processing architecture that uses reprogrammable logic to implement embedded, distributed, real-time foveal image processing from different sensor types while simultaneously allowing for lossless storage and retrieval of raw image data. Real-time, distributed, adaptive processing of multi-tap image sensors with coordinated processing hardware used for each output tap is enabled. In mosaic focal planes, a parallel-processing network can be implemented that treats the mosaic focal plane as a single ensemble rather than a set of isolated sensors. Various applications are enabled for imaging and robotic vision where processing and responding to enormous amounts of data quickly and efficiently is important.

Monacos, Steve P.↗

A step towards the final frontier: Lessons learned from acceptance testing of the first HPE/Cray EX 3000 system at ORNL

Summary In this article, we summarize the deployment of the Air Force Weather (AFW) HPC11 system at Oak Ridge National Laboratory (ORNL) including the process followed to successfully complete acceptance testing of the system. HPC11 is the first HPE/Cray EX 3000 system that has been successfully released to its user community in a federal facility. HPC11 consists of two identical 800‐node supercomputers, Fawbush and Miller, with access to two independent and identical lustre parallel file systems. HPC11 is equipped with Slingshot 10 interconnect technology and relies on the HPE Performance Cluster Manager software for system configuration. ORNL has a clearly defined acceptance testing process used to ensure that every new system deployed can provide the necessary capabilities to support user workloads. We worked closely with HPE and AFW to develop a set of tests that used the United Kingdom's Meteorological Office's Unified Model and 4‐dimensional variational data assimilation. We also included benchmarks and applications from the Oak Ridge Leadership Computing Facility portfolio to fully exercise the HPE/Cray programming environment and evaluate the functionality and performance of the system. Acceptance testing of HPC11 required parallel execution of each element on Fawbush and Miller. In addition, careful coordination was needed to ensure successful acceptance of the newly deployed lustre file systems alongside the compute resources. In this work, we present test results from specific system components and provide an overview of the issues identified, challenges encountered, and the lessons learned along the way.

Melesse Vergara, Verónica G.↗

Dynamics of threading dislocations in porous heteroepitaxial GaN films

Behavior of threading dislocations in porous heteroepitaxial gallium nitride (GaN) films has been studied using computer simulation by the two-dimensional discrete dislocation dynamics approach. A computational scheme, where pores are modeled as cross sections of cylindrical cavities, elastically interacting with unidirectional parallel edge dislocations, which imitate threading dislocations, is used. Time dependences of coordinates and velocities of each dislocation from dislocation ensembles under investigation are obtained. Visualization of current structure of dislocation ensemble is performed in the form of a location map of dislocations at any time. It has been shown that the density of appearing dislocation structures significantly depends on the ratio of area of a pore cross section to area of the simulation region. In particular, increasing the portion of pores surface on the layer surface up to 2% should lead to about a 1.5-times decrease of the final density of threading dislocations, and increase of this portion up to 15% should lead to approximately a 4.5-times decrease of it.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

A COMPARISON OF TRANSIENT INFINITE ELEMENTS AND TRANSIENT KIRCHHOFF INTEGRAL METHODS FOR FAR FIELD ACOUSTIC ANALYSIS

Finite element analysis of transient acoustic phenomena on unbounded exterior domains is very common in engineering analysis. In these problems there is a common need to compute the acoustic pressure at points outside of the acoustic mesh, since meshing to points of interest is impractical in many scenarios. In aeroacoustic calculations, for example, the acoustic pressure may be required at tens or hundreds of meters from the structure. In these cases, a method is needed for post-processing the acoustic results to compute the response at far-field points. In this paper, we compare two methods for computing far-field acoustic pressures, one derived directly from the infinite element solution, and the other from the transient version of the Kirchhoff integral. Here, we show that the infinite element approach alleviates the large storage requirements that are typical of Kirchhoff integral and related procedures, and also does not suffer from loss of accuracy that is an inherent part of computing numerical derivatives in the Kirchhoff integral. In order to further speed up and streamline the process of computing the acoustic response at points outside of the mesh, we also address the nonlinear iterative procedure needed for locating parametric coordinates within the host infinite element of far-field points, the parallelization of the overall process, linear solver requirements, and system stability considerations.

42 ENGINEERING↗

Axom

The Axom library consists of n collection of software components that provide core computer science infrastructure capabilities that can be shared by diverse high performance computing (HPC) applications. The current set of capabilities that Axom provides includes: Customizable support for error/warning and diagnostic message reporting and coordination among components of integrated applications (e.g., physics packages, libraries, etc.); Scalable parallel aggregation and filtering of message logging; In memory datastore for hierarchical, mesh-aware simulation data. The datastore supports data description, allocation, deallocation. parallel l/0, etc; Mesh data model abstractions that enable the development of mesh-based computational algorithms that work with many diverse, application· specific mesh types; and Computational geometry primitives for developing high performance parallel algorithms based on spatial geometry and meshes, e.g., spatial acceleration data structures, sets, maps, relations, etc.

Hornung, RichardD.↗

Numerical solutions of Navier-Stokes equations for a Butler wing

The flow field is simulated on the surface of a given delta wing (Butler wing) at zero incident in a uniform stream. The simulation is done by integrating a set of flow field equations. This set of equations governs the unsteady, viscous, compressible, heat conducting flow of an ideal gas. The equations are written in curvilinear coordinates so that the wing surface is represented accurately. These equations are solved by the finite difference method, and results obtained for high-speed freestream conditions are compared with theoretical and experimental results. In this study, the Navier-Stokes equations are solved numerically. These equations are unsteady, compressible, viscous, and three-dimensional without neglecting any terms. The time dependency of the governing equations allows the solution to progress naturally for an arbitrary initial initial guess to an asymptotic steady state, if one exists. The equations are transformed from physical coordinates to the computational coordinates, allowing the solution of the governing equations in a rectangular parallel-piped domain. The equations are solved by the MacCormack time-split technique which is vectorized and programmed to run on the CDC VPS 32 computer.

Abolhassani, J. S.↗

On the Theory of Thin Shallow Shells

This report is concerned with the theory of thin shallow shells. It does not employ the lines of curvature as the coordinate system, but employs "almost cartesian coordinates" or the coordinates obtained by cutting the surface into two mutually orthogonal systems of parallel planes.

Nazarov, A. A.↗

Processing EOS MLS Level-2 Data

A computer program performs level-2 processing of thermal-microwave-radiance data from observations of the limb of the Earth by the Earth Observing System (EOS) Microwave Limb Sounder (MLS). The purpose of the processing is to estimate the composition and temperature of the atmosphere versus altitude from .8 to .90 km. "Level-2" as used here is a specialists f term signifying both vertical profiles of geophysical parameters along the measurement track of the instrument and processing performed by this or other software to generate such profiles. Designed to be flexible, the program is controlled via a configuration file that defines all aspects of processing, including contents of state and measurement vectors, configurations of forward models, measurement and calibration data to be read, and the manner of inverting the models to obtain the desired estimates. The program can operate in a parallel form in which one instance of the program acts a master, coordinating the work of multiple slave instances on a cluster of computers, each slave operating on a portion of the data. Optionally, the configuration file can be made to instruct the software to produce files of simulated radiances based on state vectors formed from sets of geophysical data-product files taken as input.

Snyder, W. Van↗

EOS MLS Level 2 Data Processing Software Version 3

This software accepts the EOS MLS calibrated measurements of microwave radiances products and operational meteorological data, and produces a set of estimates of atmospheric temperature and composition. This version has been designed to be as flexible as possible. The software is controlled by a Level 2 Configuration File that controls all aspects of the software: defining the contents of state and measurement vectors, defining the configurations of the various forward models available, reading appropriate a priori spectroscopic and calibration data, performing retrievals, post-processing results, computing diagnostics, and outputting results in appropriate files. In production mode, the software operates in a parallel form, with one instance of the program acting as a master, coordinating the work of multiple slave instances on a cluster of computers, each computing the results for individual chunks of data. In addition, to do conventional retrieval calculations and producing geophysical products, the Level 2 Configuration File can instruct the software to produce files of simulated radiances based on a state vector formed from a set of geophysical product files taken as input. Combining both the retrieval and simulation tasks in a single piece of software makes it far easier to ensure that identical forward model algorithms and parameters are used in both tasks. This also dramatically reduces the complexity of the code maintenance effort.

Livesey, Nathaniel J.↗

A Step Towards the Final Frontier: Lessons Learned from Acceptance Testing of the First HPE/Cray EX 3000 System at ORNL

In this paper, we summarize the deployment of the Air Force Weather (AFW) HPC11 system at Oak Ridge National Laboratory (ORNL) including the process followed to successfully complete acceptance testing of the system. HPC11 is the first HPE/Cray EX 3000 system that has been successfully released to its user community in a federal facility. HPC11 consists of two identical 800-node supercomputers, Fawbush and Miller, with access to two independent and identical Lustre parallel file systems. HPC11 is equipped with Slingshot 10 interconnect technology and relies on the HPE Performance Cluster Manager (HPCM) software for system configuration. ORNL has a clearly defined acceptance testing process used to ensure that every new system deployed can provide the necessary capabilities to support user workloads. We worked closely with HPE and AFW to develop a set of tests that used the United Kingdom’s Meteorological Office’s Unified Model (UM) and 4DVAR. We also included benchmarks and applications from the Oak Ridge Leadership Computing Facility (OLCF) portfolio to fully exercise the HPE/Cray programming environment and evaluate the functionality and performance of the system. Acceptance testing of HPC11 required parallel execution of each element on Fawbush and Miller. In addition, careful coordination was needed to ensure successful acceptance of the newly deployed Lustre file systems alongside the compute resources. In this work, we present test results from specific system components and provide an overview of the issues identified, challenges encountered, and the lessons learned along the way.

Melesse Vergara, Veronica↗

IAEA Coordinated Research Project on HTGR Physics, Thermal Hydraulics, and Depletion Uncertainty Analysis: PHISICS/RELAP5-3D Results for the Phase III Coupled Core Exercises

This report details the Parallel and Highly Innovative Simulation for INL Code System (PHISICS)/Reactor Excursions and Leak Analysis Program (RELAP5)-3D results obtained for the neutronics stand-alone core exercises defined for Phase II of the International Atomic Energy Agency (IAEA) Coordinated Research Project (CRP) on high-temperature gas cooled reactor (HTGR) uncertainty analysis in modeling (UAM). The Phase II models and results are linked to the earlier Standardized Computer Analyses for Licensing Evaluation (SCALE)/Sampler/New ESC-based Weighting Transport (NEWT) data generated for the lattice physics (lattice) stage Phase I of the CRP. The focus of this report is the propagation of cross-section uncertainties from the lattice to the core calculation (core) phase of the prismatic 350 megawatt (MW) General Atomics (GA) Modular HTGR (MHTGR). The Phase II neutronics exercises consist of two sub-cases (II-2a and II-2b) specified for the fresh and mixed core loading of the MHTGR-350 prismatic high-temperature reactor design. Exercises II-2a and II-2b are performed without temperature feedback using isothermal temperature profiles. For the nominal lattice physics (i.e., best-estimate) model, a series of fresh and depleted single block and super cell SCALE/KENO and SCALE/NEWT models were developed at Idaho National Laboratory (INL) to assess the impact of spectral effects on the few-group cross sections. A combination of these lattice cells were used in a super cell calculation to construct 8-group cross-section libraries for use in the INL-developed PHISICS/RELAP5-3D code suite. The block-level homogenized libraries were assigned to the 220 fuel zones utilized in the MHTGR-350 PHISICS core model, according to the fresh and mixed fresh/depleted core loading pattern. In order to evaluate the effect of uncertainty propagation, several sets consisting of 1,000 perturbed NEWT lattice and PHISICS/RELAP5-3D core calculations were performed using the SCALE 6.2 ENDF/B-VII.1 cross-section co-variance data and the stochastic sampling module Sampler in SCALE 6.2.0. The Figures of Merit (FOM) investigated include the mean and standard deviation uncertainty values obtained for the eigenvalues, control rod worth values, block-averaged power profiles, power axial offsets, and isothermal temperature feedback coefficients. A sensitivity study on the main nuclide reaction contributors to the FOM uncertainties has not yet been performed since the SCALE/PHISICS/RAVEN coupling with ENDF/B-VII co-variance data is still being developed. It was found that uncertainties in the cross sections lead to uncertainties in the local power values of up to 1.5%. The variances between the fresh and mixed/depleted core models are not as large as the axial variations between the top and bottom within each core model. The core eigenvalue uncertainty is approximately 0.5%, which is significant in comparison to typical eigenvalue difference targets (less than ~0.1% for LWRs). The power peaking and axial offset uncertainties are smaller (0.10% – 0.15%) and probably below the level of significance in typical core design metrics. The use of an 8-group energy structure slightly under-estimates some of the uncertainties, but in absolute terms, the uncertainty results are reasonably close to the 26-group results. The 26- and 8-group mean and nominal results differ by a larger degree, and if it can be afforded in terms of calculation resources, the 26-group structure should be used for HTGR calculations. The 8-group structure can, however, still be useful for reasonably accurate uncertainty assessments of transient cases where computational time is important. It was furthermore found that the use of a super cell during the cross-section generation process for better representation of the spectral environment at the core periphery did not lead to significant changes in the core eigenvalues, axial offset, and power peaking uncertainties. The impact of the super cell cross sections on the nominal and mean eigenvalues is, however, important enough (0.6% – 1.1% lower) to justify the use of these super cells for HTGR core simulation to represent the softer spectral environment in the core peripheral regions for nominal best-estimate calculations. The use of super cells for HTGR uncertainty studies is, therefore, not strictly necessary, but the improvements in nominal results could justify their use.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Avoiding Communication in Primal and Dual Block Coordinate Descent Methods

Primal and dual block coordinate descent methods are iterative methods for solving regularized and unregularized optimization problems. Distributed-memory parallel implementations of these methods have become popular in analyzing large machine learning datasets. However, existing implementations communicate at every iteration, which, on modern data center and supercomputing architectures, often dominates the cost of floating-point computation. Recent results on communication-avoiding Krylov subspace methods suggest that large speedups are possible by re-organizing iterative algorithms to avoid communication. We show how applying similar algorithmic transformations can lead to primal and dual block coordinate descent methods that only communicate every $s$ iterations---where $s$ is a tuning parameter---instead of every iteration for the regularized least-squares problem. We show that the communication-avoiding variants reduce the number of synchronizations by a factor of $s$ on distributed-memory parallel machines without altering the convergence rate and attain strong scaling speedups of up to $6.1\times$ over the “standard algorithm" on a Cray XC30 supercomputer.

97 MATHEMATICS AND COMPUTING↗

Material Model Testing in the DYNA3D Code: Material Models 1 through 5

This report is the first volume in an effort to study the material model implementations in DYNA3D, using the latest developmental version of the code. Input files were derived from a template supplied by Jerry Lin to perform an equation-of-state study. Most of the simulations use 10 cubes, each 1 cm on a side with a density of 8.03 (steel). Each of the cubes tests a different aspect of the material model. The first three cubes are testing normal stresses and normal strains (uniaxial stresses and strains) placed on the cube parallel to the x-axis, y-axis and z-axis, respectively. The x (or y or z) = 1 face of the cube is moved along the x-axis (or y-axis or z-axis) to test the model, while the opposite side remains in the x (or y or z) = 0 plane. For some of the tests, the nodes are free to move perpendicular to the applied strain to allow the cross-sectional area of the cube to change in response to being placed under pressure or tension (to test Poisson’s Ratio), while in others the nodes are frozen to eliminate dynamic effects resulting from the code. The cube is placed in tension along the axis being tested and then released, followed by a cycle of compression and release. This tension/compression cycle is repeated 3 more times. The additional cycles allow for the model to be tested under different loads. For these calculations, the amount by which the cube is stretched and compressed is varied from cycle to cycle to allow testing of model features. Cubes number 4 and 5 are intended to test xy shear stresses/strains. This is done by sliding one face of the cube in its plane parallel to the axis of interest. For example, for cube 4, the upper face of the cube is slid in the y=1 plane in a direction parallel to the x-axis (so there are no changes in the y and z coordinates of the points in the plane) and then returned to its starting position. This preserves the volume of the cube, so there is no change in density or pressure. The cycle is repeated 8 times. For cube 5, the plane of the cube satisfying x=1 is slid in this plane in a direction parallel to the y-axis and then returned to its starting position. This preserves the volume of the cube, so there is no change in density or pressure. The cycle is repeated 8 times. Cubes number 6 through 9 repeat these tests for yz shear stresses/strains (Cubes 6 and 7) and zx shear stresses and strains (Cubes 8 and 9). For reasons discussed in the next paragraph, it is unclear if these tests represent a state of pure shear stress and shear strain in the code. Application of shear stresses and strains should not, in general, result in the appearance of normal stresses and strains in Cubes 4 - 9. In the infinitesimal limit, this result might be obtained. However, DYNA3D is a finite deformation code and in the real deformations used in the code, this result is not achieved and the individual components of the normal stresses and strains are non-zero during the cycling. However, the sums of the normal strains are zero in these tests and, except in the case of thermally dependent models, so are the sums of the normal stresses. Cube 10 tests the response of the material model to a hydrostatic loading of the cube. The load curve used for the first three tests is used again and applied to each of the 8 nodes in a manner causing the cube to be either compressed or expanded. This compression/tension cycle is repeated 4 times. The additional cycles allow for the model to be tested under different loads. The amount by which the cube is stretched and compressed is varied from cycle to cycle to allow testing of model features. This document will discuss the testing of Material Models 1 through 5. Relevant results from the test cubes will be discussed. A discussion of the results from all 10 test cubes will be presented only if necessary.

36 MATERIALS SCIENCE↗

A nonvoxel-based dose convolution/superposition algorithm optimized for scalable GPU architectures

Purpose: Real-time adaptive planning and treatment has been infeasible due in part to its high computational complexity. There have been many recent efforts to utilize graphics processing units (GPUs) to accelerate the computational performance and dose accuracy in radiation therapy. Data structure and memory access patterns are the key GPU factors that determine the computational performance and accuracy. In this paper, the authors present a nonvoxel-based (NVB) approach to maximize computational and memory access efficiency and throughput on the GPU. Methods: The proposed algorithm employs a ray-tracing mechanism to restructure the 3D data sets computed from the CT anatomy into a nonvoxel-based framework. In a process that takes only a few milliseconds of computing time, the algorithm restructured the data sets by ray-tracing through precalculated CT volumes to realign the coordinate system along the convolution direction, as defined by zenithal and azimuthal angles. During the ray-tracing step, the data were resampled according to radial sampling and parallel ray-spacing parameters making the algorithm independent of the original CT resolution. The nonvoxel-based algorithm presented in this paper also demonstrated a trade-off in computational performance and dose accuracy for different coordinate system configurations. In order to find the best balance between the computed speedup and the accuracy, the authors employed an exhaustive parameter search on all sampling parameters that defined the coordinate system configuration: zenithal, azimuthal, and radial sampling of the convolution algorithm, as well as the parallel ray spacing during ray tracing. The angular sampling parameters were varied between 4 and 48 discrete angles, while both radial sampling and parallel ray spacing were varied from 0.5 to 10 mm. The gamma distribution analysis method (γ) was used to compare the dose distributions using 2% and 2 mm dose difference and distance-to-agreement criteria, respectively. Accuracy was investigated using three distinct phantoms with varied geometries and heterogeneities and on a series of 14 segmented lung CT data sets. Performance gains were calculated using three 256 mm cube homogenous water phantoms, with isotropic voxel dimensions of 1, 2, and 4 mm. Results: The nonvoxel-based GPU algorithm was independent of the data size and provided significant computational gains over the CPU algorithm for large CT data sizes. The parameter search analysis also showed that the ray combination of 8 zenithal and 8 azimuthal angles along with 1 mm radial sampling and 2 mm parallel ray spacing maintained dose accuracy with greater than 99% of voxels passing the γ test. Combining the acceleration obtained from GPU parallelization with the sampling optimization, the authors achieved a total performance improvement factor of >175 000 when compared to our voxel-based ground truth CPU benchmark and a factor of 20 compared with a voxel-based GPU dose convolution method. Conclusions: The nonvoxel-based convolution method yielded substantial performance improvements over a generic GPU implementation, while maintaining accuracy as compared to a CPU computed ground truth dose distribution. Such an algorithm can be a key contribution toward developing tools for adaptive radiation therapy systems.

60 APPLIED LIFE SCIENCES↗

Modelling of nanometer scale dust grains in tokamak

Dust poses a serious threat to tokamak operation and safety. It is important to study the behaviour of dust grains under tokamak's discharge conditions, which depends heavily on their size and charge. Existing simulations mainly address issues on dust grains with radii larger than 1 μm, in which case, the drift effect due to electromagnetic fields can be safely ignored. For nanometer scale dust grains, however, the drift effect becomes significant and a new model based on guiding-centre system needs to be established. In this work, the NDS has been done under BOUT++ framework. The simulation contains two parts. Part one, NDS evaluates the charging and ablation processes of the dust grains. In the second part, the guiding-centre orbits of dust particles are tracked in tokamak plasmas, whose parameters are obtained from BOUT++, a highly desirable C++ code package for performing parallel plasma fluid simulations with an arbitrary number of equations in 3D curvilinear coordinates. The orbit of nanodust dynamics is described by guiding centre equations for simplicity, and these equations are numerically solved by conventional fourth-order Runge Kutta method. Simulations provide results such as trajectories and evolutions of dust particles with different sizes and velocities for different tokamak geometries. Results show tungsten dust grains with a radius of a few nanometers launched from outer midplane will oscillate before totally ablated in C-Mod. The oscillation in this case is driven by the ion drag force. Larger Nanodust with a radius of 100 nm, on the contrary, cannot be completely constrained by the electromagnetic field. In conclusion, the high plasma temperature and density in the seperatrix region causes severe dust ablation, resulting in total ablation within several ms.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Finite-difference schemes for anisotropic diffusion

In fusion plasmas diffusion tensors are extremely anisotropic due to the high temperature and large magnetic field strength. This causes diffusion, heat conduction, and viscous momentum loss, to effectively be aligned with the magnetic field lines. This alignment leads to different values for the respective diffusive coefficients in the magnetic field direction and in the perpendicular direction, to the extent that heat diffusion coefficients can be up to 10{sup 12} times larger in the parallel direction than in the perpendicular direction. This anisotropy puts stringent requirements on the numerical methods used to approximate the MHD-equations since any misalignment of the grid may cause the perpendicular diffusion to be polluted by the numerical error in approximating the parallel diffusion. Currently the common approach is to apply magnetic field-aligned coordinates, an approach that automatically takes care of the directionality of the diffusive coefficients. This approach runs into problems at x-points and at points where there is magnetic re-connection, since this causes local non-alignment. It is therefore useful to consider numerical schemes that are tolerant to the misalignment of the grid with the magnetic field lines, both to improve existing methods and to help open the possibility of applying regular non-aligned grids. To investigate this, in this paper several discretization schemes are developed and applied to the anisotropic heat diffusion equation on a non-aligned grid.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Interaction of carbon dioxide laser radiation with a nanotube array in the presence of a constant electric field

The dependence of the current density on the leading edge width of the alternating (high-frequency) field amplitude is studied at various constant (or unsteady) fields. The dependence of amplified microwaves in the two-millimeter range on a longitudinal coordinate is determined. The problem of submillimeter radiation generation in a system of parallel carbon nanotubes exposed to two-frequency carbon dioxide (CO{sub 2} laser) laser radiation in the presence of a constant (or unsteady) field is studied. The possibility of using freely oriented carbon nanotubes parallel to each other is shown.

36 MATERIALS SCIENCE↗

IMS Rapid Response 2024 Summary Report: A Machine Learning Potential for the Periodic Table

Stockpile stewardship and nuclear waste remediation are inherently chemically complex, involving practically the full diversity of the periodic table, but existing methods are too expensive or not functional for a large diversity of atom types. Overall, the field of machine learning interatomic potentials (MLIPs) has advanced dramatically in 2024 with large high-accuracy datasets existing for bulk, surface, and organic chemical systems and new online leaderboards for diverse chemistry. To participate in, and bring LANL interests into this ecosystem, here, we have built upon existing technologies created by LANL to create a framework capable of creating machine learning interatomic potentials (MLIPs) for over 90 atom types. Our results have created a massively diverse coordination complex training dataset more than 3 times the size of existing datasets, parallelized MLIP training over multiple GPUs, enabling the training of an MLIP spanning the periodic table at 20 times the speed of prior training on 32 GPUs. These advances are substantial towards creation on foundational MLIPs for LANL-specific application areas.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗