Search NASA⌕ Search

SEARCH · Search NASA

Results for “Single atom catalysis”

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.

229 records · Page 13
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Infrared plasmonics: STEM-EELS characterization of Fabry-Pérot resonance damping in gold nanowires

Materials possessing strong midinfrared responses are of current interest because of their potential application to long-wavelength metamaterials, photonic devices, molecular detection, and catalysis. In this work, we utilize high-energy resolution (80 cm -1 , 10 meV) electron-energy-loss spectroscopy (EELS) in a monochromated and aberration-corrected scanning transmission electron microscope (STEM) to resolve multipolar surface plasmon resonances (SPRs), sometimes called Fabry-Pérot (FP) resonances, in gold nanowires with mode energies spanning from ~1000 to 8000 cm -1 . STEM-EELS provides access to these mid- to near-IR responses in a single acquisition, avoiding the difficulties inherent in obtaining the same data using near-field optical techniques. The experimentally measured FP resonance energies and linewidths, together with analytical modeling and full-wave numerical electrodynamics simulations, provide a comprehensive picture of the radiative and intrinsic contributions to the total damping rates. We find some FP modes with dephasing times >60fs, which is almost twice the longest previously reported plasmon dephasing time for individual Au nanoparticles in the infrared. The long dephasing times and the broad tunability of the FP resonance energies throughout the infrared region suggest additional opportunities for harnessing infrared plasmonic energy before dephasing occurs.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Semi-grand canonical Monte Carlo simulation of the acrolein induced surface segregation and aggregation of AgPd with machine learning surrogate models

The single atom alloy of AgPd has been found to be a promising catalyst for the selective hydrogenation of acrolein. It is also known that the formation of Pd islands on the surface will greatly reduce the selectivity of the reaction. As a result, the surface segregation and aggregation of Pd on the AgPd surface under reaction conditions of selective hydrogenation of acrolein are of great interest. In this work, we lay out a workflow that can predict the surface segregation and aggregation of Pd on a FCC(111) AgPd surface with and without the presence of acrolein. We use machine learning surrogate models to predict the AgPd bulk energy, AgPd slab energy, and acrolein adsorption energy on AgPd slabs. Then, we use the semi-grand canonical Monte Carlo simulation to predict the surface segregation and aggregation under different bulk Pd concentrations. Under vacuum conditions, our method predicts that only trace amount of Pd will exist on the surface at Pd bulk concentrations less than 20%. However, with the presence of acrolein, Pd will start to aggregate as dimers on the surface at Pd bulk concentrations as low as 6.5%.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Direct methane activation by atomically thin platinum nanolayers on two-dimensional metal carbides

We report efficient and direct conversion of methane to value-added products has been a long-term challenge in shale gas applications. Here, we show that atomically thin nanolayers of Pt with a single or double atomic layer thickness, supported on a two-dimensional molybdenum titanium carbide (MXene), catalyse non-oxidative coupling of methane to ethane/ethylene (C 2 ). Kinetic and theoretical studies, combined with in-situ spectroscopic and microscopic characterizations, demonstrate that Pt nanolayers anchored at the hexagonal close-packed sites of the MXene support can activate the first C–H bond of methane to form methyl radicals that favour desorption over further dehydrogenation and thus suppress coke deposition. At 750 °C and 7% methane conversion, the catalyst runs for 72 hours of continuous operation without deactivation and exhibits >98% selectivity towards C 2 products, with a turnover frequency of 0.2–0.6 s -1 . Our findings provide insights into the design of highly active and stable catalysts for methane activation and create a platform for developing atomically thin supported metal catalysts.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Evaluating the stability and activity of dilute Cu-based alloys for electrochemical CO 2 reduction

Cu-based catalysts currently offer the most promising route to actively and selectively produce value-added chemicals via electrochemical reduction of CO 2 (eCO 2 R); yet further improvements are required for their wide-scale deployment in carbon mitigation efforts. Here, we systematically investigate a family of dilute Cu-based alloys to explore their viability as active and selective catalysts for eCO 2 R through a combined theoretical–experimental approach. Using a quantum–classical modeling approach that accounts for dynamic solvation effects, we assess the stability and activity of model single-atom catalysts under eCO 2 R conditions. Our calculations identify that the presence of eCO2R intermediates, such as CO*, H*, and OH*, may dynamically influence the local catalyst surface composition. Additionally, we identify through binding energy descriptors of the CO*, CHO*, and OCCO* dimer intermediates that certain elements, such as group 13 elements (B, Al, and Ga), enhance the selectivity of C 2+ species relative to pure Cu by facilitating CO dimerization. The theoretical work is corroborated by preliminary testing of eCO 2 R activity and selectivity of candidate dilute Cu-based alloy catalyst films prepared by electron beam evaporation in a zero-gap gas diffusion electrode-based reactor. Of all studied alloys, dilute CuAl was found to be the most active and selective toward C 2+ products like ethylene, consistent with the theoretical predictions. Finally, we attribute the improved performance of dilute CuAl alloys to more favorable dimerization reaction energetics of bound CO species relative to that on pure Cu. In a broader context, the results presented here demonstrate the power of our simulation framework in terms of rational catalyst design.

30 DIRECT ENERGY CONVERSION↗
DOE OSTI2021

Toward alcohol synthesis from CO hydrogenation on Cu(111)-supported MoS 2 – predictions from DFT+KMC

In the quest for cheap and efficient catalysts for alcohol synthesis from syngas, a material of interest is single-layer MoS2 owing to its low cost, abundancy, and flexible structure. Because of the inertness of its basal plane, however, it is essential to find ways that make it catalytically active. Herein, by means of density functional theory based calculations of reaction pathways and activation energy barriers and accompanying kinetic Monte Carlo simulations, we show that while S vacancy row structures activate the MoS 2 basal plane, further enhancement of chemical activity and selectivity can be achieved by interfacing the MoS 2 layer with a metallic support. When defect-laden MoS2 is grown on Cu(111), there is not only an increase in the active region (surface area of active sites) but also charge transfer from Cu to MoS 2 , resulting in a shift of the Fermi level such that the frontier states (d orbitals of the exposed Mo atoms) appear close to it, making the MoS 2 /Cu(111) system ready for catalytic activity. Finally, our calculated thermodynamics of reaction pathways lead to the conclusion that the Cu(111) substrate promotes both methanol and ethanol as the products, while kinetic Monte Carlo simulations suggest a high selectivity toward the formation of ethanol.

2D materials↗
DOE OSTI2025

Python Library for Monte Carlo Simulations with Ab Initio and Machine-Learned Interatomic Potentials

There is a growing need in the simulation community for software that provides a transparent, reproducible, usable, and extensible (TRUE) Monte Carlo (MC) simulation framework employing energies from ab initio methods and machine-learning interatomic potentials (MLIPs). We introduce a Python library (ASE-MC) that adds Monte Carlo functionality to the Atomic Simulation Environment (ASE) package. Now, we can combine the powerful tools used to build systems and perform ab initio and MLIP in ASE with MC simulation algorithms to sample the configurational space with a concise Python script. After presenting the design philosophy, we demonstrate the flexibility of our approach using selected examples. These example simulations include liquid water described with a message-passing MLIP in the canonical and isothermal–isobaric ensembles, sampling the characteristic dihedral angle of biphenyl and comparing an MLIP to first-principles calculations, and a grand canonical Monte Carlo simulation of ammonia adsorption on Pt(111). These examples showcase the main features of the software, which include flexibility in the choice of ab initio or MLIP engine, ab initio or MLIP grand canonical MC with cavity bias insertions and deletions, the ability to add custom MC moves to the move set, and how users can condense complex MC workflows into a single Python script. Finally, this library serves as a framework for reproducible Monte Carlo simulations, facilitating easy reproduction of the work and application to new systems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Modeling plasma-induced surface charge effects on CO 2 activation by single atom catalysts supported on reducible and irreducible metal oxides

Abstract The accumulation of negative surface charge on catalytic surfaces in the presence of low-temperature plasma (LTP) could influence catalytic performance. However, it is difficult to disentangle the role of surface charging and other LTP catalytic effects in experiment. Herein, we use density functional theory (DFT) modeling to understand the effect of plasma-induced surface charging on CO 2 activation by atomically dispersed single atom (SA) catalysts on both reducible and irreducible metal oxide supports. We model CO 2 adsorption strength and CO 2 dissociation barriers for Co 1 , Ni 1 , Cu 1 , Rh 1 , Pd 1 , and Ag 1 SAs on both reducible and irreducible supports, namely, CeO 2 (100), TiO 2 (101), and γ -Al 2 O 3 (110), to elucidate trends. We find that accumulated surface charge on the SA increases the CO 2 adsorption strength and decreases the CO 2 dissociation barrier for all studied SA/support combinations. For both charged and uncharged (neutral) systems, SAs on the reducible CeO 2 (100) support generally adsorb CO 2 more weakly compared to when on irreducible supports like γ -Al 2 O 3 (110). SAs on γ -Al 2 O 3 (110) typically have larger barriers for CO 2 dissociation for both charged and uncharged systems compared to TiO 2 (101) and CeO 2 (100). The magnitude of surface charging effects on CO 2 binding energies and dissociation barriers depends sensitively on both the SA and the support. In some cases, the CO 2 activation trends qualitatively change between neutral and charged systems for a fixed SA across different supports. This DFT modeling study demonstrates that surface charging should be considered in strong electric fields because it can have a large effect on molecule adsorption and bond-breaking on catalytic surfaces.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Modeling Catalyzed Growth of Single-Wall Carbon Nanotubes (Final Report)

Single-walled carbon nanotubes (SWCNTs) are among the most promising nanostructures because of their special and unusual physico-chemical properties that make them strong candidates for a variety of applications. Since their discovery, chemical vapor deposition (CVD) methods have become the preferred synthesis method. CVD uses a carbon-containing precursor species, which is decomposed over a catalyst yielding C atoms that eventually evolve in carbon nanostructures. CVD temperatures are in the order of 1000K or lower, and pressures in the range of 1-5 atm, although some processes use higher pressures. Depending on the nature of the catalyst, and on the conditions of pressure, temperature, and presence of other inert compounds, the resultant nanostructures may include nanofibers, nano-onions, multiple-, or single-wall carbon nanotubes, among others. When the CVD synthesis process is appropriately tuned, however, it is possible to obtain only SWCNTs, which are geometrically characterized by their diameter and by their chiral angle. The chiral indexes (n,m) determine the degree of helical twist of the graphite lattice along the nanotube axis, where the term chirality is derived from the left and right-handed helicity of the tubes defined by their chiral indexes (n,m) and (m,n) respectively. The main goal of this research is to build fundamental understanding that will help elucidating the nature and structure of the actual catalyst in a combined catalyst/support system, which we claim is essential especially when the catalyst is of nanodimensions. Given that we are considerably advanced in the understanding of selective single-walled carbon nanotube catalyst synthesis, we take this process as a case study. Thus, we focus on the understanding of the selective growth of SWCNTs on supported nanocatalysts in systems where the catalyst synthesis takes place in the same reactor preceding the nanotube synthesis. Once we understand what type of catalyst structure would be necessary for this specific growth, we will try to elucidate the catalytic synthesis process that leads to such catalytic structure. Thus, our aim of controlling nanostructure synthesis may start to be delineated.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Intermetallic Nanocatalysts from Heterobimetallic Group 10–14 Pyridine-2-thiolate Precursors

Intermetallic compounds are atomically ordered inorganic materials containing two or more transition metals and main-group elements in unique crystal structures. Intermetallics based on group 10 and group 14 metals have shown enhanced activity, selectivity, and durability in comparison to simple metals and alloys in many catalytic reactions. While high-temperature solid-state methods to prepare intermetallic compounds exist, softer synthetic methods can provide key advantages, such as enabling the preparation of metastable phases or of smaller particles with increased surface areas for catalysis. In this work, we study a generalized family of heterobimetallic precursors to binary intermetallics, each containing a group 10 metal and a group 14 tetrel bonded together and supported by pincer-like pyridine-2-thiolate ligands. Upon thermal decomposition, these heterobimetallic complexes form 10–14 binary intermetallic nanocrystals. Experiments and density functional theory (DFT) computations help in better understanding the reactivity of these precursors toward the synthesis of specific intermetallic binary phases. Using Pd 2 Sn as an example, we demonstrate that nanoparticles made in this way can act as uniquely selective catalysts for the reduction of nitroarenes to azoxyarenes, which highlights the utility of the intermetallics made by our method. Employing heterobimetallic pincer complexes as precursors toward binary nanocrystals and other metal-rich intermetallics provides opportunities to explore the fundamental chemistry and applications of these materials.

10-14 intermetallics↗
DOE OSTI2026

Electro-chemo-mechanically Driven Ni Exsolution from (Pr,Ce,Ni)O 2−δ : Controlled Nucleation Density and Enhanced Electrode Kinetics

In situ exsolution of metal nanoparticles is a promising strategy to prepare electrocatalysts with enhanced activity and resistance to agglomeration for efficient chemical transformations and energy conversion. Achieving a high nucleation density of nanoparticles under mild conditions and understanding how to tailor the process is important for performance of these electrodes in electrochemical cells. In this work, we demonstrate facile exsolution of Ni nanoparticles using fluorite-structured (Pr,Ce)O 2−δ as the support oxide, driven by electrochemical potential and aided by the metastability of Ni in the solid solution (elastic driving force). We prepare single-phase oriented thin films of (Pr,Ce,Ni)O 2−δ (NPCO) on (Zr,Y)O 2−δ (YSZ) substrates by pulsed laser deposition. With the aid of a high-throughput electrochemical cell that provides a lateral gradient in Nernst voltage, we apply in situ near-ambient pressure synchrotron X-ray photoelectron spectroscopy and ex situ atomic force microscopy to investigate the impact of electrochemical potential on Ni nucleation density. We find that metallic Ni can be successfully exsolved at 550 °C upon cathodic biasing in 20 mTorr O 2 , and its nucleation density increases with increasing electrochemical driving force/decreasing oxygen chemical potential. We further evaluate the electrochemical performance under highly reducing (fuel electrode) conditions by electrochemical impedance spectroscopy. With the exsolved Ni nanoparticles, the surface exchange coefficient of the NPCO is found to be ∼4× higher than for PCO without exsolution. This work confirms mixed conducting fluorites as beneficial host lattices for facile transition-metal exsolution and suggests the possibility for constructing an all ceria-based electrochemical cell with PCO serving as both the cathode and the anode.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Elucidating the Role of Electric Fields in Fe Oxidation via an Environmental Atom Probe

Abstract We quantify the effects of intensely applied electric fields on the Fe oxidation mechanism. The specimen are pristine Fe single crystals exposing a variety of surface structures identified by field ion microscopy. These crystals are simultaneously exposed to low pressures of pure oxygen gas, on the order of 10 −7 mbar, while applying intense electric fields on their surface of several tens of volts per nanometer. The local composition of the different surface structures is probed directly and in real time using an Environmental Atom Probe and successfully compared with first principles‐based models. We found that rough Fe{244} and Fe{112} facets are more reactive toward oxygen than compact Fe{024} and Fe{011} facets. Results demonstrate that the influence of an electric field on the oxidation kinetics depends on the timescales that are involved as the system evolves toward equilibrium. The initial oxidation kinetics show that strong increases in electric fields facilitate the formation of an oxide. However, as one approaches equilibrium, high field values mitigate this formation. Ultimately, this study elucidates how high externally applied electric fields can be used to dynamically exploit reaction dynamics at the nanoscale towards desired products in a catalytic reaction at mild reaction conditions.

Lambeets, Sten V. [Physical and Computational Scie↗
DOE OSTI2025

Elucidating the Role of Electric Fields in Fe Oxidation via an Environmental Atom Probe

We quantify the effects of intensely applied electric fields on the Fe oxidation mechanism. The specimen are pristine Fe single crystals exposing a variety of surface structures identified by field ion microscopy. These crystals are simultaneously exposed to low pressures of pure oxygen gas, on the order of 10 −7 mbar, while applying intense electric fields on their surface of several tens of volts per nanometer. The local composition of the different surface structures is probed directly and in real time using an Environmental Atom Probe and successfully compared with first principles-based models. We found that rough Fe{244} and Fe{112} facets are more reactive toward oxygen than compact Fe{024} and Fe{011} facets. Results demonstrate that the influence of an electric field on the oxidation kinetics depends on the timescales that are involved as the system evolves toward equilibrium. The initial oxidation kinetics show that strong increases in electric fields facilitate the formation of an oxide. However, as one approaches equilibrium, high field values mitigate this formation. Ultimately, this study elucidates how high externally applied electric fields can be used to dynamically exploit reaction dynamics at the nanoscale towards desired products in a catalytic reaction at mild reaction conditions.

09 BIOMASS FUELS↗
DOE OSTI2026

HydraGNN_Predictive_GFM_2026 - Ensemble of predictive graph foundation models for atomistic materials modeling

This release contains data and parameters of HydraGNN-based graph foundation models trained as a result of the work published in the pre-print "Exascale Multi-Task Graph Foundation Models for Imbalanced, Multi-Fidelity Atomistic Data" by M. Lupo Pasini et al. (https://arxiv.org/abs/2604.15380). We jointly train on 16 open first-principles datasets (544+ million structures covering 85+ elements) using a multi-task architecture with per-dataset heads and a scalable ADIOS2/DDStore data pipeline. On Frontier, we execute six large-scale DeepHyper hyperparameter optimization campaigns in FP64 and promote the top-performing message-passing models to sustained 2,048-node training, yielding a PaiNN-based lead model. The version of HydraGNN used to generate the outputs provided in this release is HydraGNN v5.0 (https://github.com/ORNL/HydraGNN/releases/tag/v5.0) The list of datasets used for the training of the graph foundation model is the following: 1) Alexandria [1] 2) ANI1x [2] 3) MPTrj [3] 4) Open Catalyst 2020 (OC20) [4] 5) Open Catalyst 2022 (OC22) [5] 6) Open Catalyst 2025 (OC25) [6] 7) Open Direct ir Capture 2023 (ODAC23) [7] 8) Open Materials 2024 (OMat24) [8] 9) Open Molecules 2025 (OMol25) [9] 10) OMol25-neutral (subset of OMol25 that contains only molecules with zero total charge) 11) OMol25-non-neutral (subset of OMol25 that contains only molecules with non-zero total charge) 12) Open Polymers 2026 (OPoly2026) [10] 13) Nabla2DFT [11] 14) QCML [12] 15) QM7X [reference 13] 16) transition1x [14] Dataset references: [1] J. Schmidt et al., “A dataset of 175k stable and metastable materials calculated with the PBEsol and SCAN functionals,” Scientific Data, vol. 9, p. 64, 2022. [2] J. S. Smith et al., “The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules,” Scientific Data, vol. 7, p. 134, 2020. [Online]. Available: https: //www.nature.com/articles/s41597-020-0473-z [3] A. Jain et al., “Commentary: The Materials Project: A materials genome approach to accelerating materials innovation,” APL Materials, vol. 1, no. 1, p. 011002, 07 2013. [Online]. Available: https://doi.org/10.1063/1.4812323 [4] L. Chanussot et al., “Open catalyst 2020 (oc20) dataset and community challenges,” ACS Catalysis, vol. 11, no. 10, pp. 6059–6072, 2021. [Online]. Available: https://doi.org/10.1021/acscatal.0c04525 [5] K. Tran et al., “Open catalyst 2022 (oc22) dataset and challenges for oxidation electrocatalysts,” ACS Catalysis, vol. 13, no. 5, pp. 3066–3084, 2023. [Online]. Available: https://doi.org/10.1021/acscatal.2c05426 [6] S. J. Sahoo et al., “The open catalyst 2025 (oc25) dataset and models for solid-liquid interfaces,” arXiv preprint arXiv:2509.17862, 2025. [Online]. Available: https://arxiv.org/abs/2509.17862 [7] A. Sriram et al., “The open DAC 2023 dataset and challenges for sorbent discovery in direct air capture,” ACS Central Science, vol. 10, no. 5, pp. 923–941, 2024. [8] L. Barroso-Luque et al., “Open materials 2024 (omat24) inorganic materials dataset and models,” 2024. [Online]. Available: https://arxiv.org/abs/2410.12771 [9] D. S. Levine et al., “The open molecules 2025 (OMol25) dataset, evaluations, and models,” 2025. [Online]. Available: https://arxiv.org/abs/2505.08762 [10] D. S. Levine et al., The open polymers 2026 (OPoly26) dataset and evaluations,” arXiv preprint arXiv:2512.23117, 2025. [Online]. Available: https://arxiv.org/abs/2512.23117 [11] K. Khrabrov et al., “Nabla2dft: A universal quantum chemistry dataset of drug-like molecules and a benchmark for neural network potentials,” in NeurIPS 2024 Datasets and Benchmarks Track, 2024. [Online]. Available: https://openreview.net/forum?id=ElUrNM9U8c [12] S. Ganscha et al., “The QCML dataset, quantum chemistry reference data from 33.5M DFT and 14.7B semi-empirical calculations,” Scientific Data, vol. 12, p. 406, 2025. [13] J. Hoja et al., “QM7-X, a comprehensive dataset of quantum-mechanical properties spanning the chemical space of small organic molecules,” Scientific Data, vol. 8, p. 43, 2021. [Online]. Available: https://www.nature.com/articles/s41597-021-00812-2 [14] M. Schreiner et al., “Transition1x - a dataset for building generalizable reactive machine learning potentials,” Scientific Data, vol. 9, p. 779, 2022. The folder "datasets_ADIOS2_format" contains the set of pre-processed datasets in Adaptable I/O System (ADIOS) format (https://www.exascaleproject.org/research-project/adios/) that have been used for the development and training of GFMs in this work. The "datasets_ADIOS2_format" directory contains 2 sub-directories, one for the version "v1" of the datasets and one for the version "v2" of the datasets. The version "v1" of the datasets provides values of the total energy as they are extracted from the original data as it was released by the respective institutions. The version "v2" of the datasets provides values of the energy that have been realigned. The realignment was performed by training a linear regression model that predicts the total energy as a function of the chemical composition of the atomistic structure, and then subtract such prediction from the original value of the total energy. Both folders "v1" and "v2" contain 16 sub-directories, each corresponding to an ADIOS2-formatted dataset The folder "DeepHyper-results" contains the configurational files and model's parameters for all the 186 HPO trials that were successfully completed by the scalable hyperparameter optimization (HPO) runs on Frontier. The content of the folder "DeepHyper-results" I structured as follows: 1) task-list.txt: list of mpnn name, jobid, and deephyper task id 2) gfm_${MPNN}_${JOBID}_0.${TASKID}: run directory with checkpoint files 3) gfm_${MPNN}: deephyper summary directory (*.csv) for each specific MPNN type 4) deephyper-experiment-${JOBID}: output and error logs for each job The file "deephyper-sorted.csv" contains the details of each HydraGNN model built and tested by HPO, obtained by merging the (*.csv) filed from each HPO run executed. Out of all the HPO trials, we selected 10 to continue the training of the respective HydraGNN models. Due to limited computational budget available in the LRN070 allocation we could not complete the training till convergence for all these 10 selected models. The folder "models" contains multiple sub-folders, one per each HydraGNN model trained. Each model sub-folder contains the parameters of each HydraGNN model, with multiple checkpoint-restarts. The list of sub-folders are as follows: 1) multidataset_hpo-BEST1-fp64 2) multidataset_hpo-BEST2-fp64 3) multidataset_hpo-BEST3-fp64 4) multidataset_hpo-BEST4-fp64 5) multidataset_hpo-BEST5-fp64 6) multidataset_hpo-BEST6-fp64 7) multidataset_hpo-BEST7-fp64 8) multidataset_hpo-BEST8-fp64 9) multidataset_hpo-BEST9-fp64 10) multidataset_hpo-BEST10-fp64 Within each one of these folders, additional auxiliary log files are provided with descriptions about how the training proceeded. The lead PaiNN-model is contained inside "multidataset_hpo-BEST6-fp64". The file "mlp_branch_weights" contains the parameters of the multi-layer perceptron (MLP) used to reconcile the predictions of the 16 output decoding heads of the HydragNN architectures. The MLP takes in input the chemical composition of the atomistic structure and predicts averaging weights to linearly mix the predictions of each output decoding head toward consolidating them into a single one. The folder "1.1billion-structure-inference" contains 1.1 billion atomistic structures randomly generated. Each structures is associated with energy and forces predicted with the lead-PaiNN model combined with the MLP model for reconciliation of the multi-branch predictions generated by the 16 output decoding heads. The folder "1.1billion-structure-inference" contains 9,300 (*.tar.gz) subdirectories, one per Frontier compute node used to execute the inference at exascale. Once uncompressed, each (*.tar.gz) subdirectory contains an ADIOS2 (*.bp) file container, where each atomistic structure is stored as a PyTorch-Geometric Data object. The file "export_dataset_environment_variables.sh" contains the environment variables that need to be set before running the HydraGNN code to reproduce the results provided in this dataset release. The code that can be used to load the ADIOS2 files, load HydraGNN models, and run inference is available at: https://github.com/ORNL/HydraGNN/releases/tag/v5.0

36 MATERIALS SCIENCE↗