Search NASA⌕ Search

SEARCH · Search NASA

Results for “Non-equilibrium properties”

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 55 records · Page 3

Mesoscale Modeling of Dislocation Cell Structure Evolution and Radiation-Induced Segregation in Additively Manufactured Austenitic Stainless Steel

Structural alloys under irradiation develop radiation-induced segregation (RIS) at point defect sinks, which lead to undesired changes in the alloy's properties. Additively manufactured austenitic stainless steels are expected to yield a distinct response to irradiation damage owing to their unique as-printed dislocation cell structure. In this talk, we present the development of a mesoscale model for dislocation cell structure evolution coupled with RIS using the MOOSE framework. The evolution of network dislocations within spatially distinct cells and cell walls are modeled using climb-mediated edge annihilation and generation processes. Starting with initial microstructures of as-printed non-equilibrium segregations at dislocation cell walls, our 2D and 3D simulations explore the effects of irradiation temperature, dose rate, dislocation sink bias, and specimen thickness on microstructure evolution. The results will be compared against experimental characterizations of in-situ and ex-situ ion-irradiated samples, and the implications of evolving cell structure on irradiation damage response will be discussed.

additive manufacturing↗

Simulating topological quantum gates in two-dimensional magnet-superconductor hybrid structures

The creation of topological quantum gates using Majorana zero modes—an outstanding problem in the field of topological quantum computing—relies on our ability to control the braiding process in time and space. Here, we propose two-dimensional magnet-superconductor hybrid structures as a new platformfor the successful implementation of topologically protected √σ z -, σ z - and σ x -quantum gates using Majorana zero modes. Employing a novel theoretical formalism to compute the full timedependent many-body wave-function and utilizing a braiding protocol motivated by recent advances in electron-spin-resonance techniques we simulate quantum gates in 2D systems up to 600 sites, on timescales from a few femto- to nanoseconds. We demonstrate that the braiding process can be visualized in time and space by computing the non-equilibrium local density of states, which is proportional to the time-dependent differential conductance measured in scanning tunneling spectroscopy experiments, allowing us to directly image Majorana world lines.

Superconducting properties and materials↗

Photoionization of seeded combustion products as a method of enhancing the efficiency of magnetohydrodynamic power generators

Here, in this study, we performed an experimental and computational investigation into the feasibility of utilizing photoionization to enhance the electrical conductivity of seeded oxy-fuel combustion products and improve the performance of magnetohydrodynamic (MHD) power generators. We applied a variety of optical and microwave diagnostics to study the ionization and recombination processes of potassium excited by an excimer laser in a high-velocity oxy-fuel free jet. Computational fluid dynamic (CFD) simulations were performed to model the thermophysical properties and species densities of the free jet. The CFD results were validated with position-dependent potassium concentration measurements. Electron recombination exponential lifetimes were measured through time-resolved microwave transmission. The experimental electron lifetimes were compared with lifetimes calculated from CFD-predicted species densities and literature recombination rates. It was determined that K + or O 2 are the most likely recombination partners for photoionized electrons. Time-resolved fluorescence measurements provided evidence of an ionization pathway involving a two-photon ionization of KOH . Finally, a zero-dimensional chemical kinetic model was developed to assess the fundamental viability of inducing a non-equilibrium electron population to provide a net energy return in combustion-driven MHD power generators. We determined that a high energy return is feasible for targeting electrode boundary layers with ultraviolet photoionization. We also found that photoionization could potentially lower the required temperature of the bulk gas flow.

20 FOSSIL-FUELED POWER PLANTS↗

Crossover from quantum to classical transport

Understanding the crossover from quantum to classical transport phenomena has become of fundamental importance not only for technological applications due to the creation of sub-10nm transistors – an important building block of our modern life – but also for elucidating the role played by quantum mechanics in the evolutionary fitness of biological complexes. This article provides a basic introduction into the nature of charge and energy transport in the quantum and classical regimes. Here, it will identify the characteristic transport properties in both limits, and demonstrate how they can be connected through the loss of quantum mechanical coherence. I will identify the salient features of the crossover physics, and demonstrate their importance in opening new transport regimes and for understanding efficient and robust energy transport in biological complexes.

charge↗

Dissipation in open holography

We exploit the holographic realization of a conformal theory coupled to an external bath realized via a double trace deformation and its gravity dual in terms of transparent boundary conditions in order to map out some basic dissipative properties of this simple open holographic system. In particular, we determine the energy transmission coefficient across the boundary, discover a novel duality relating weak and strong coupling to the external bath, and quantify the dissipation in the system by working out the quasi normal modes.

AdS-CFT correspondence↗

Quantum kinetic modeling of KEEN waves in a warm-dense regime

We report the first fully kinetic, quantum study of kinetic electrostatic electron nonlinear (KEEN) waves, showing that quantum diffraction systematically erodes the classical trapping mechanism, narrows harmonic locking to the fundamental, and hastens post-drive decay. Electrons are evolved with a second-order Strang-split 1D1V Wigner–Poisson solver that couples conservative semi-Lagrangian WENO advection to an analytic Fourier space update for the non-local Wigner term, while ions remain classical. We focus on collisionless dynamics in a weakly coupled regime, providing a controlled baseline before collisional extensions. Short, frequency-tuned ponderomotive pulses drive KEEN formation in a uniform Maxwellian plasma; as the dimensionless quantum parameter H rises from the classical limit to values relevant to warm-dense matter, doped semiconductors, and 2D electron systems, the drive threshold increases, higher harmonics are damped, trapped electron vortices diffuse, and the subplasma electrostatic energy relaxes to a lower stationary level, as confirmed by continuous wavelet analysis. These microscopic changes carry macroscopic weight. Ignition-scale capsules now compress matter to regimes where the electron de Broglie wavelength rivals the Debye length, making classical kinetic descriptions insufficient. By extending KEEN physics into this quantum domain, our results offer a potential diagnostic of non-equilibrium electron dynamics for next-generation inertial-confinement designs and high-energy-density platforms, indicating that predictive fusion modeling may benefit from the integration of kinetic fidelity with quantum effects.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Oxygen Effects and Selective Oxidation of Group IV Alloying Additions in Group V-Based Refractory Multi-principal Element Alloys

Refractory multi-principal element alloys (RMPEAs) hold significant potential for advancing ultra-high-temperature technologies due to their remarkable strength retention under these conditions. However, the beneficial effects of alloying on microstructural evolution and mechanical properties are often obscured by the presence of interstitial alloying elements. The intricate interactions between substitutional alloying elements of Groups IV, V, and VI and interstitial content, whether intentional or incidental, remain poorly understood. In this study, we investigate the impact of dissolved oxygen on the microstructural evolution and stability of RMPEAs primarily composed of Group V elements, Nb and V, along with Group IV elements Ti, Zr, and Hf. We show that the strong affinity of oxygen for Group IV elements Zr and Hf leads to the internal precipitation of oxides during heat treatment, which competes with the homogenization of the as-cast dendritic microstructure. Here, we demonstrate that careful control and design of oxygen content and Group IV alloying additions are essential to minimize micro-segregation and prevent the detrimental precipitation of internal oxides. To address these challenges, we propose employing non-equilibrium solidification calculations to predict and optimize as-cast solidification structures, thereby reducing the reliance on homogenization processes.

Alloy design↗

Suppression of ferromagnetic correlations in hole-doped SrCo 2 As 2

We report the growth of high-quality single crystals of hole-doped composition Sr 0.95 K 0.05 Co 2 As 2 and present an investigation of its structural, electronic, thermal, and magnetic properties. Our results show that hole doping leads to a substantial suppression of ferromagnetic (FM) correlations present in SrCo 2 As 2 leading to nearly isotropic magnetic susceptibilities χ ab,c (T) of Sr 0.95 K 0.05 Co 2 As 2 . The presence of FM fluctuations is considered one of the detrimental reasons why SrCo 2 As 2 does not achieve a superconducting ground state despite having a precursor for it in the form of stripe-like antiferromagnetic fluctuations. The suppression of FM fluctuations leads to qualitative similarities between the properties of hole-doped SrCo 2 As 2 and the end member of superconducting 122 iron-arsenide family KFe 2 As 2 . Our results also infer the presence of strongly correlated electronic states in Sr 0.95 K 0.05 Co 2 As 2 . Further, they show that the electronic structure of SrCo 2 As 2 as well as this hole-doped composition is exceptionally unstable in the vicinity of Fermi level EF for small alterations in the structural parameters and sensitively depends upon the details of simulation.

36 MATERIALS SCIENCE↗

Nanopores with dynamic pore opening diameter

Solid state nanopores have emerged as model systems for understanding transport properties on the nanoscale. They serve as templates for both preparing mimics of biological channels and designing biological sensors. Unlike their biological inspirations, the majority of nanopores prepared thus far, however, have been structurally static devices such that the pore opening diameter is fixed. If we could prepare nanopores whose opening diameter fluctuated in time with controlled amplitude at known locations in the pore, we could create ionic memristors as well as achieve new transport modes. Here we present ∼10 nm diameter single nanopores drilled through a 10 nm thick gold layer positioned on top of a 30 nm thick silicon nitride film. Two types of devices were prepared; one containing single stranded DNA and the other containing hairpin DNA attached to the discrete layer of gold using thiol chemistry. When an external electric field was applied across a nanopore with single stranded DNA, the nanoconfined DNA molecules exhibited steric and electrical constraints that led to memristor-like behavior in the current–voltage curves. The degree of hysteresis was controlled by salt concentration, magnitude of voltage and pore diameter. In contrast, nanopores containing DNA hairpins conducted similar currents in forward and reverse bias in agreement with the rigidity of the hairpin molecule. The experiments are explained by Brownian dynamics simulations that reveal voltage and salt concentration induced changes in DNA extension. The degree of DNA extension was also found to be dependent on the location of the molecules along the pore axis. The nanopores presented here provide the first steps towards preparation of non-equilibrium nanopore systems.

Vlassiouk, Ivan [ORNL] (ORCID:0000000254940386)↗

Trajectory Engineering with Modular Patched Conics for Entry Systems and TPS (TEMPEST)

Brief Presenter Biography (35 word limit): Bohdan Wesely is an Aerospace Engineer in the Entry Systems and Technology Division at Ames. He has worked on a variety of projects for NASA including integrated TPS (thermal protection system) flight hardware deliveries and testing services for commercial partners. Introduction: TEMPEST is a new trajectory analysis framework that is designed to fill the gap between dedicated flight mechanics tools and aerothermal and TPS sizing tools. The project started as an SJSU master’s thesis and has since evolved into a general conceptual design tool capable of studying a wide variety of entry problems. Development is ongoing in the Entry Systems and Technology Division at NASA ARC. Why TEMPEST: Space missions involving entry into a planetary atmosphere involve a series of unique requirements across multiple disciplines. Whether it is traditional entry descent and landing (EDL), or aerocapture, the vehicle must navigate to its target landing location or orbit state, and the TPS must protect the payload during entry. The design process typically involves iterative handoffs between various flight mechanics, flow solver, and material response level tools. During the early conceptual phase, a wide variety of feasible trajectories are simulated in a Monte Carlo scenario which broadly satisfy the mission or landing requirements. Next, computational fluid dynamics (CFD), direct simulation Monte Carlo (DSMC), and other flow solver analyses are performed at various key trajectory points to generate an aero-database, heating and TPS design requirements also emerge at this stage. At this point, with updated aerodynamics from the various flow solvers, trajectories can be re-run, this in turn can change the required freestream conditions for the CFD tools, and as a project progresses, these analyses converge, and uncertainty is reduced. However, there is always a “hand-off” occurring between two inherently coupled phenomena. Analysis Description: One of the goals with TEMPEST is to use a variety of first principles estimation methods coupled with an atmosphere model to predict vehicle aerothermodynamics across the entire flight regime while propagating a 3 or 6 degree of freedom (DoF) trajectory. Aerodynamics methods include modified Newtonian, Maxwell and Cercignani- Lampis-Lord (CLL) for continuum, transitional, and free molecular flow regimes. Aerothermodynamics include boundary layer and reference enthalpy methods, and Mutation++ for non-equilibrium chemistry modeling. TEMPEST is also capable of stitching multiple trajectory segments together to study mission scenarios like multi-pass aerocapture and aero-gravity assists. Most of the program is implemented in MATLAB using modern system objects, it relies on several C++ shared libraries for supporting tools like Gmsh, the Global Reference Atmospheric Model (GRAM), and Mutation++. The various first principles aerothermal estimation methods are discretized across either a structured axisymmetric panel mesh or an unstructured tri-mesh generated from an open-source tool such as Gmsh, this allows solutions on the same mesh to be compared across tools such as CB-Aero. CFD Coupling. A physics-aware, gaussian process CFD anchoring scheme is proposed to adjust the various first principles methods as a CFD database is populated. One goal for this anchoring module is to inform the project where CFD should be run. Full knowledge of the entire trajectory, atmosphere, and aerothermodynamics allows for easier identification of high sensitivity areas and uncertainty quantification. While the first principles effects are well known and proven accurate in existing tools such as CB- Aero and Cart3D, a physics aware CFD anchoring scheme increases tool credibility across a project lifecycle. Material Response Modeling. Correct TPS sizing is critical for optimizing mass for science payloads and ensuring mission success. The process typically involves a thermal analysis along the trajectory with surface heating environments as a boundary condition. Several design constraints are maximum bondline temperature and maximum recession with various margining techniques. The material response tool FIAT, developed out of NASA Ames, is currently being integrated into the TEMPEST environment. TPS recession, shape change, mass loss, and mass property alteration are all factors that can perturb an entry trajectory. For missions like Mars 2020, recession was minimal and was safely handled separately as a post process. For missions such as Jupiter Galileo with a high TPS mass fraction or asteroid entries, recession plays a major role. The proposed fully coupled scheme is to use an epoch-based approach where the trajectory integration is halted after a recession threshold, the energy balance and FIAT are solved at each panel, the mesh, aerodynamics, and mass properties are updated, and the trajectory continues. Several computational tradeoffs have been made during the development of TEMPEST to limit the cost of a single trajectory and preserve its utility as a conceptual, rapid iteration tool. Conclusion: Development of TEMPEST is ongoing and the project is still in its infancy. This talk aims to showcase its unique capabilities to support future NASA entry systems missions.

Bohdan O Wesely↗

Observing differential spin currents by resonant inelastic X-ray scattering

Controlling spin currents, that is, the flow of spin angular momentum, in small magnetic devices, is the principal objective of spin electronics, a main contender for future energy-efficient information technologies. A pure spin current has never been measured directly because the associated electric stray fields and/or shifts in the non-equilibrium spin-dependent distribution functions are too small for conventional experimental detection methods optimized for charge transport. Here we report that resonant inelastic X-ray scattering (RIXS) can bridge this gap by measuring the spin current carried by magnons—the quanta of the spin wave excitations of the magnetic order—in the presence of temperature gradients across a magnetic insulator. This is possible due to the sensitivity of the momentum- and energy-resolved RIXS intensity to minute changes in the magnon distribution under non-equilibrium conditions. Furthermore, we use the Boltzmann equation in the relaxation time approximation to extract transport parameters, such as the magnon lifetime at finite momentum, essential for the realization of magnon spintronics.

36 MATERIALS SCIENCE↗

Quantum graph models for transport in filamentary switching

The formation of metallic nanofilaments bridging two electrodes across an insulator is a mechanism for resistive switching. Examples of such phenomena include atomic synapses, which constitute a distinct class of memristive devices the behavior of which is closely tied to the properties of the filament. Until recently, experimental investigation of the low-temperature regime and quantum transport effects has been limited. However, with growing interest in understanding the true impacts of the filament on device conductance, comprehending quantum effects has become crucial for quantum neuromorphic hardware. Here, we discuss quantum transport resulting from filamentary switching in a narrow region where the continuous approximation of the contact is not valid, and only a few atoms are involved. In this scenario, the filament can be represented by a graph depicting the adjacency of atoms and the overlap between atomic orbitals. Using the theory of quantum graphs with locally diffusive node scattering, we calculate the scattering amplitude of charge carriers on this graph and explore the interplay between filamentary formation and quantum transport effects.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Curvature Memory in Electrically Stimulated Lipid Membranes

Here, we demonstrate, using non-equilibrium molecular dynamics simulations, that lipid membrane capacitance varies with surface charge accumulation linked to membrane shape and curvature changes. Specifically, we show that lipid membranes exhibit a hysteretic response when exposed to oscillatory electric fields. The electromechanical coupling in these membranes leads to hysteretic buckling, in which the membrane can spontaneously buckle in one of two distinct directions along the electric field, even for the same ionic charge accumulation at the water–membrane interface. In this regard, these binary buckled membrane states suggest potential applications in neuromorphic computing. Their bistable nature, characterized by two distinct and stable configurations, could serve as a foundation for implementing memory storage systems and logic operations. Furthermore, we introduce a circuit model that captures these dynamic effects, offering insights into emergent memory effects in electrically stimulated lipid membranes. Finally, this work presents lipid bilayers as dynamic, adaptable elements and suggests a new platform for exploring energy storage, information processing, and memory encoding at the lipid membrane level.

59 BASIC BIOLOGICAL SCIENCES↗

Design of novel refractory equiatomic multi-principal elemental alloys based on Mo-Nb-Ti system for Gen IV reactor applications

Excellent irradiation damage resistance demonstrated by multi-principal elemental alloys (MPEAs) has sparked significant interest among researchers, prompting exploration into their vast compositional space, to validate their suitability for nuclear applications. A combined approach of thermodynamic and empirical parameters calculations alongside CALPHAD (CALculation of PHAse Diagrams) for phase formation predictions enable high-throughput material selection for sophisticated applications like nuclear, overcoming laborious and time-consuming experiments. Key thermodynamic and empirical parameters for eight novel equiatomic MPEAs, based on seven low thermal neutron cross section refractory elements, for predicting phase formation were calculated, and equilibrium and non-equilibrium simulations in CALPHAD were employed to comprehensively model the systems. Pseudo binary phase diagram simulations showed that Zr, V or equiatomic CrV additions to the base MoNbTi alloy (MoNbTi-Zr, MoNbTi-V and MoNbTi-CrV alloys) favor the formation of isomorphous body-centered cubic (BCC) phase at high temperatures, while Cr, Al, equiatomic ZrV, or equiatomic CrAl additions (MoNbTi-Cr, MoNbTi-Al, MoNbTi-ZrV or MoNbTi-CrAl alloys) limit the solubility of them. Equilibrium CALPHAD simulations at 750 oC were consistent with XRD results on MoNbTi, MoNbTiZr and MoNbTiCr alloys, and partially for others. Notably, elemental segregation observed in the backscattered electron (BSE) scanning electron microscopy (SEM) images of the alloys was accurately simulated through non-equilibrium Scheil solidification calculations in CALPHAD, further verified by experiments. The precipitation of TiCr2 Laves phase in Cr containing MoNbTiCr and MoNbTiCrAl was accurately predicted while discrepancies were noted in MoNbTiCrV. The equilibrium simulations also provided insights into phase compositions at specific temperatures offering a pathway for tailoring the desired microstructure and properties of these systems. Empirical parameters calculations successfully predicted random solid solution in the base MoNbTi alloy, and with an exception in MoNbTiV and MoNbTiAl, predicted intermetallic precipitation in the rest, especially, Laves phase precipitation in Cr containing alloys.

36 - MATERIALS SCIENCE↗

Thermodynamic consistency and fluctuations in mesoscopic stochastic simulations of reactive gas mixtures

It is essential that mesoscopic simulations of reactive systems reproduce the correct statistical distributions at thermodynamic equilibrium. By considering a compressible fluctuating hydrodynamics (FHD) simulation method of ideal gas mixtures undergoing reversible reactions described by the chemical Langevin equations, we show that thermodynamic consistency in reaction rates and the use of instantaneous temperatures for the evaluation of reaction rates is required for fluctuations for the overall system to be correct. We then formulate the required properties of a thermodynamically consistent reaction (TCR) model. As noted in the literature, while reactions are often discussed in terms of forward and reverse rates, these rates should not be modeled independently because they must be compatible with thermodynamic equilibrium for the system. Using a simple TCR model where each chemical species has constant heat capacity, we derive the explicit condition that the forward and reverse reaction rate constants must satisfy in order for the system to be thermodynamically consistent. We perform equilibrium and non-equilibrium simulations of ideal gas mixtures undergoing a reversible dimerization reaction to measure the fluctuational behavior of the system numerically. We confirm that FHD simulations with the TCR model give the correct static structure factor of equilibrium fluctuations. For the statistically steady simulation of a gas mixture between two isothermal walls with different temperatures, we show using the TCR model that the temperature variance agrees with the corresponding thermodynamic-equilibrium temperature variance in the interior of the system, whereas noticeable deviations are present in regions near walls, where chemistry is far from equilibrium.

Polimeno, Matteo [University of California, Merced↗

Spectral deconvolution without the deconvolution: Extracting temperature from x-ray Thomson scattering spectra without the source-and-instrument function

X-ray Thomson scattering (XRTS) probes the dynamic structure factor of the system, but the measured spectrum is broadened by the combined source-and-instrument function (SIF) of the setup. In order to extract properties such as temperature from an XRTS spectrum, the broadening by the SIF needs to be removed. Recent work [Dornheim et al. Nat. Commun. 13 , 7911 (2022)] has suggested that the SIF may be deconvolved using the two-sided Laplace transform. However, the extracted information can depend strongly on the shape of the input SIF, and the SIF is in practice challenging to measure accurately. Here, we propose an alternative approach: we demonstrate that considering ratios of Laplace-transformed XRTS spectra collected at different scattering angles is equivalent to performing the deconvolution, but without the need for explicit knowledge of the SIF. From these ratios, it is possible to directly extract the temperature from the scattering spectra, when the system is in thermal equilibrium. We find the method to be generally robust to spectral noise and physical differences between the spectrometers, and we explore situations in which the method breaks down. Furthermore, the fact that consistent temperatures can be extracted for systems in thermal equilibrium indicates that non-equilibrium effects could be identified by inconsistent temperatures of a few eV between the ratios of three or more scattering angles.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Insights into distorted lamellar phases with small-angle scattering and machine learning

Lamellar phases are essential in various soft matter systems, with topological defects significantly influencing their mechanical properties. In this report, we present a machine-learning approach for quantitatively analyzing the structure and dynamics of distorted lamellar phases using scattering techniques. By leveraging the mathematical framework of Kolmogorov–Arnold networks, we demonstrate that the conformations of these distorted phases – expressed as superpositions of complex waves – can be reconstructed from small-angle scattering intensities. Through the contour analysis of wave field phase singularities, we obtain the statistics of the spatial distribution of topological defects. Furthermore, we establish that the temporal evolution of these defects can be derived from the time-dependent traveling wave field, informed by the dispersion relation of spectral components. This method opens new avenues for investigating the dynamics of distorted lamellar phases using various dynamic scattering techniques such as neutron spin echo and X-ray photon correlation spectroscopy. These findings enhance our microscopic understanding of how defects influence the physical properties of lamellar materials, with implications for both equilibrium and non-equilibrium states in general lamellar systems.

36 MATERIALS SCIENCE↗

HydraGNN_Predictive_GFM_2024 - Ensemble of predictive graph foundation models for ground state atomistic materials modeling

We provide the ensemble of fifteen pre-trained graph foundation models (GFMs) for atomistic materials modeling applications. Each one of the fifteen GFMs has been trained on five open-source datasets that (once aggregated) amount to over 154 million atomistic structures, which cover over two-thirds of the natural elements of the periodic table and that comprises a broad set of organic and inorganic compounds. This vast set of atomistic structures comprises ground state configurations that are dynamically stable (i.e., equilibrated structures with atomic forces approximately close to zero values) as well as dynamically unstable structures (i.e., non-equilibrium structures with non-negligible non-zero values of atomic forces). The ensemble of datasets aggregated does NOT include excited states. The datasets have been curated to remove atomistic structures with spectral norm of the force tensor above 100 eV/angstrom. Moreover, a linear term of the energy was computed for each dataset using a linear regression model that uses the chemical concentration of each natural element as regressor. The linear term predicted by the linear regression model has been subtracted from each original energy value to perform a re-alignment of the energy values across different electronic structures approximation theories performed to generate the diverse multi-source, multi-fidelity datasets. The folder "ADIOS_files" 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 "ADIOS_files" directory contains 6 sub-directories named as follows: - ANI1x-v3.bp - MPTrj-v3.bp - OC2020-20M-v3.bp - OC2020-v3.bp - OC2022-v3.bp - qm7x-v3.bp Each sub-directory contains the pre-processed datasets converted in Adaptable I/O System (ADIOS) format (https://www.exascaleproject.org/research-project/adios/) that have been used to the development, training, and performance testing of the ensemble go predictive graph foundation models. Each GFM was developed using HydraGNN (https://github.com/ORNL/HydraGNN) as underlying graph neural network (GNN) architecture. The multi-task learning (MTL) capability of HydraGNN was used to simultaneously train the GFMs on labeled values for direct predictions of energy (a total system property of an atomistic structure that measures the chemical stability) and atomic forces (an atomic level property of an atomistic structure that measures the dynamical stability). The hyper parameters of the GFM have been tuned using scalable hyperparameter optimization (HPO) algorithms implemented in the software DeepHyper (https://github.com/deephyper/deephyper). The pre-training of each HPO trial was performed using distributed data parallelism (DDP) to scale the training across 128 compute nodes of the exascale OLCF supercomputer Frontier. Each HPO trial was trained only for 10 epochs and an early stopping was performed to avoid wasting significant computational resources on GNN architectures that were clearly underperforming. For each HPO trial, the 'omnistat' tool developed by (AMD Research - Advanced Micro Device) was used to measure the total energy consumption in kWh. The ensemble of GFMs was obtained by selecting the fifteen best performing HPO trials. Four models have been selected for their clear advantage in accuracy, and these are the GFMs with IDs 229, 156, 147, 260. Additional eleven models have been selected based on judicious balance between accuracy and energy consumption needed for training, and these are the GFMs with IDs 165, 78, 137, 1, 175, 171, 181, 67, 179, 167, 351. Each selected GFM of the ensemble was continued to cumulate a total of at most 30 epochs. In some cases, the total number of epochs actually performed was les than 30 due to two combined factors: (1) the size of the GFM (i.e., the number of model parameters to train) and (2) the total wall-clock time for which the computational resources could be allocated on OLCF-Frontier. The "Ensemble_of_models" directory contains 15 sub-directories named as follows: - gfm_0.229 - gfm_0.156 - gfm_0.147 - gfm_0.260 - gfm_0.165 - gfm_0.78 - gfm_0.137 - gfm_0.1 - gfm_0.175 - gfm_0.171 - gfm_0.181 - gfm_0.67 - gfm_0.179 - gfm_0.167 - gfm_0.351 Each one of these sub-directories refers to one of the fifteen HPO trials that have been selected to continue the pre-training with at most 30 epochs. With each sub-directory associated with a specific HPO trial, the following files can be found: - config.json: file for argument parsing to develop and train an HydraGNN architecture - gfm_0.ID_epoch_N.pk: file with model parameters for HPO ID trial after N epochs of training The ensemble of fifteen GFM architectures was used for (1) ensemble averaging to stabilize the predictions of energy and atomic forces after pre-training for post-processing analysis and (2) ensemble uncertainty quantification (UQ). The code used to develop, pre-train, and load the pre-trained models for post-processing analysis is available on the ORNL-GitHub at the following link: https://github.com/ORNL/HydraGNN/tree/Predictive_GFM_2024

36 MATERIALS SCIENCE↗