SEARCH · Engineering Papers
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.
We introduce HANNA, the first hybrid neural network model that strictly complies with all thermodynamic consistency criteria for predicting activity coefficients and outperforms current benchmark methods in terms of accuracy and applicability.
NEML2 vectorizes constitutive-model evaluation for large-scale multiphysics simulation, using PyTorch as its tensor backend so that a batch of material-point updates runs on CPU or GPU through a single implementation. In the two prior reports in this series it was a C++-native library, deployed through TorchScript tracing and just-in-time (JIT) compilation; it has since been rewritten from the ground up into a Python-native library deployed through Ahead-of-Time Inductor (AOTInductor), a modern PyTorch graph-compilation backend. The rewrite is driven by a persistent tension, not a language preference: NEML2 composes constitutive models at runtime from a registry of small, independently-authored pieces, and that flexibility is difficult to reconcile with the compile-time knowledge an efficient GPU kernel needs. This report documents the rewrite and the investment that accompanied it: the AOTInductor export pipeline that turns a Python-authored model into a portable, Python-free compiled artifact loadable from pure C++; the eager and compiled runtimes and the new implicit solver layer built on them; a head-to-head benchmark of legacy JIT against AOTInductor; the physics-model catalog and its worked examples; the developer tooling; and the corresponding overhaul of MOOSE’s NEML2 integration that lets MOOSE consume it. A central objective is to examine whether modern PyTorch graph-compilation backends are effective for MOOSE GPU integration. The benchmark answers directly: AOTInductor outperforms legacy JIT on every GPU scenario measured, by 1.0–4.5×. Modern graph-compilation backends are effective for MOOSE GPU integration, and AOTInductor specifically – not compilation in the abstract – is why.
Hybrid manufacturing is a promising route for producing complex aerospace components, yet systematic cost benchmarking across multiple additive-subtractive pathways remains limited. This study presents a comprehensive process-based cost analysis of seven hybrid manufacturing routes, including laser powder bed fusion (L-PBF), powder- and wire-directed energy deposition (DED), wire arc additive manufacturing (WAAM), additive friction stir deposition (AFSD), metal binder jetting (MBJ), and agility forging, followed by scanning and finish machining. Parametric cost models incorporating direct material, labor, and energy costs were developed. L-PBF results are discussed in detail for a pickle fork component and directly compared with commercial pricing. Across all hybrid routes, labor emerged as the dominant cost driver, contributing more than 70% of total manufacturing cost in some cases. AFSD exhibited the lowest cost for aluminum components, with MBJ being its 316 L stainless steel counterpart, after accounting for geometric scaling. Benchmarking against industrial quotes suggests that hybrid manufacturing can achieve cost levels comparable to those of commercial services, although labor-intensive processes exhibit greater deviation. The analysis highlights automation of material handling, setup, and supervision as key opportunities for improving economic competitiveness. Overall, the proposed framework provides a quantitative basis for evaluating and optimizing hybrid manufacturing pathways for aerospace applications.
Multipactor discharge is a persistent challenge in high-power microwave (HPM) and accelerator systems, where secondary electron avalanches can cause heating, vacuum degradation, and failure. This work presents the first supervised machine learning (ML) framework for multipactor prediction, trained on high-fidelity 3D Particle-in-Cell (PIC) simulation data in planar geometries. The model maps operational, geometric, and material-dependent secondary electron yield (SEY) parameters to the time-averaged electron growth rate, enabling rapid reconstruction of susceptibility charts. Among the models evaluated, tree-based ensemble methods such as Random Forest and Extra Trees demonstrate superior generalization to unseen materials compared to neural networks such as multilayer perceptron (MLP). Performance metrics, including Intersection over Union (IoU), Structural Similarity Index Measure (SSIM), and Pearson correlation, show close agreement with simulation benchmarks. Principal Component Analysis attributes generalization limits to material feature-space disjointedness.
Scientific hypothesis generation represents a fundamental challenge in contemporary research due to exponentially expanding literature volumes and increasing disciplinary specialization. Large language models (LLMs) have emerged as transformative tools for automated scientific discovery, moving beyond traditional rule-based and literature-mining approaches. Four paradigmatic approaches define current LLM-driven hypothesis generation: direct prompting and fine-tuning methods, knowledge-enhanced frameworks integrating retrieval-augmented generation (RAG), multi-agent collaborative systems simulating research teams, and reasoning-focused approaches implementing cognitive architectures. Domain-specific applications demonstrate statistical equivalence to human expert performance in social psychology, experimental validation in biomedical research, and near-expert quality in astronomy. Evaluation methodologies encompass human expert assessment, LLM-as-judge frameworks, and comprehensive benchmarking systems. Technical challenges include hallucination management, knowledge integration limitations, and balancing novelty with feasibility. Future directions emphasize hybrid neural-symbolic architectures and sophisticated human-AI collaboration models for responsible scientific discovery acceleration.
By virtue of being atomically thin, the electronic properties of heterostructures built from two-dimensional materials are strongly influenced by atomic relaxation. The atomic layers behave as flexible membranes rather than rigid crystals. Here we develop an analytical theory of lattice relaxation in twisted moiré materials. We obtain analytical results for the lattice displacements and corresponding pseudo gauge fields, as a function of twist angle. We benchmark our results for twisted bilayer graphene and twisted WSe 2 bilayers using large-scale molecular dynamics simulations. Our single-parameter theory is valid in graphene bilayers for twist angles 𝜃 ≳ 0.7°, and in twisted WSe 2 for 𝜃 ≳ 1.6°. Furthermore, we also investigate how relaxation alters the electronic structure in twisted bilayer graphene, providing a simple extension to the continuum model to account for lattice relaxation.
A generative modeling framework is proposed that combines diffusion models and manifold learning to efficiently sample data densities on manifolds. The approach utilizes Diffusion Maps to uncover possible low-dimensional underlying (latent) spaces in the high-dimensional data (ambient) space. Two approaches for sampling from the latent data density are described. The first is a score-based diffusion model, which is trained to map a standard normal distribution to the latent data distribution using a neural network. The second one involves solving an Itô stochastic differential equation in the latent space. Additional realizations of the data are generated by lifting the samples back to the ambient space using Double Diffusion Maps , a recently introduced technique typically employed in studying dynamical system reduction; here the focus lies in sampling densities rather than system dynamics. The proposed approaches enable sampling high dimensional data densities restricted to low-dimensional, a priori unknown manifolds. The efficacy of the proposed framework is demonstrated through a benchmark problem and a material with multiscale structure.
Superdiffusive spin transport in the one-dimensional (1D) Heisenberg model is a key theoretical discovery in nonequilibrium quantum many-body physics. Although extensively studied in 1D systems, the breakdown and sustenance of superdiffusion in two-dimensional (2D) lattices with integrability-breaking terms, as found in real materials, remains an open question. To address this, we develop a toy model that extends the 1D Heisenberg model with a representative set of 2D interaction types and tunable strengths. Our model exhibits varying degrees of superdiffusion breakdown depending on the interaction type, spanning ballistic to diffusive regimes. We establish and justify a hierarchy of 2D interactions based on their resilience against superdiffusion breakdown: Heisenberg >𝑋𝑋 > Ising. This precise control over the superdiffusive behavior also enables rigorous benchmarking of quantum hardware, and our simulations on IBM's Heron devices confirm the hardware's ability to accurately capture these many-body nonequilibrium phenomena. Overall, our results are relevant not only to simulating superdiffusion in real materials, such as the 1D Heisenberg compound KCuF3, which contains modest nonintegrable 2D terms, but also to extending superdiffusive behavior to larger 2D qubit lattices and other 2D materials.
Here, this work develops a predictive density tool in Python, named Plutonium Nitrate Solutions (PuNS), to reduce bias and uncertainty in nuclear criticality safety calculations for plutonium nitrate systems. The Pitzer method and an empirical method were implemented into the PuNS tool to generate atom densities for use in MCNP6 material cards. These material cards are directly prepared into an MCNP6 input text file and are calculated based on customizable user inputs of plutonium content, nitric acid content, temperature, and plutonium isotope weight percentages. The PuNS tool is validated and verified against the International Criticality Safety Benchmark Evaluation Project Handbook experiments and is observed to predict densities within a root mean square error of 0.89% for the Pitzer method and 1.82% for the empirical method. These errors in density lead to up to 1569 pcm difference in MCNP6 calculated k eff for the Pitzer method and up to a 1751 pcm difference for the empirical method when compared to experimental benchmarks. Simultaneous work is also being performed at Los Alamos National Laboratory and the University of New Mexico to create a similar tool for plutonium chloride solutions, named Plutonium Chloride Solution, which aims to provide the accreditation of the chlorine absorption. These capabilities will not only provide more accurate models but also facilitate an improved understanding of solution systems and a potential relaxation in the conservatism of current aqueous plutonium processing criticality safety limits.
Atomistic simulations have become indispensable tools for understanding glass structure, dynamics, and properties, yet persistent challenges limit their predictive power. This perspective examines three interconnected issues, namely glass formation procedures, interatomic potential development, and machine learning applications, which emerged from the 5th International Workshop on Challenges of Atomistic Simulations of Glasses and Amorphous Materials. We identify convergent community priorities for (i) standardized validation protocols, (ii) curated benchmark datasets with complete metadata, and (iii) open repositories for glasses. A systematic was forward is provided by a hierarchical validation framework for assessing the structural fidelity, property prediction, and behavioral realism of simulation techniques. Looking ahead, transformative advances are promised by the fusion of classical techniques with machine learning based approaches, for instance, by integrating swap Monte Carlo with machine-learning (ML) potentials, leveraging foundation models through transfer learning, and finetuning ML potentials with experimental data. Progress depends on the community committing to validated models, reproducible protocols, and sustained data sharing.
IceCube has recently detected a diffuse population of high-energy neutrinos arising from the Milky Way. We use this high-significance detection to place new limits on dark matter (DM) annihilation to neutrinos with two complementary approaches. The first method uses the background-subtracted Galactic longitude distribution of shower events to place a conservative bound on the DM annihilation cross section that does not rely on any assumed Galactic cosmic ray emission model; the resulting limits on the velocity-averaged annihilation cross section improve upon existing bounds by factors of a few. The second method uses the template-dependent neutrino energy spectra from the Inner Galaxy, inferred under different Galactic cosmic ray emission models. This complementary approach shows that the inferred Galactic neutrino intensities are already sensitive to DM contributions near the thermal-relic benchmark for a range of TeV-scale DM masses, though this comparison is more model-dependent. Our results demonstrate that measurements of diffuse Galactic neutrino emission can be used as a powerful probe of DM annihilation into neutrinos. Future observations with IceCube-Gen2 and KM3NeT will substantially extend this sensitivity, potentially allowing a decisive test of the thermal freeze-out mechanism with Galactic neutrino observations.
In this work, we study a two-mode bosonic encoding of a quoctit inside a non-Abelian group-structured constellation of coherent states. This work is motivated by the importance of non-Abelian symmetry in particle physics and the desire to have transversal non-Abelian logical gates. We use the previously developed 2 T constellation of states used to encode a so-called 2 T qutrit. The fidelity of the 2 T quoctit is benchmarked against other bosonic qudits for different noise models and find it compare favorably when power constraints are considered. This paves the way for the construction of higher-dimensional qudits (e.g., a quicosotetrit with 2 T group structure) in bosonic systems with practical applications in quantum simulations of particle physics.
The emergence of AI-driven Earth observation systems promises to broaden access to petabyte-scale geospatial data beyond domain specialists. However, translating this vision into operational scientific infrastructure requires addressing fundamental challenges in data virtualization, code transparency, and domain-specific reasoning. We present Hydrology Copilot, a cloud-native AI framework for natural-language-driven analysis of Earth observation data. To demonstrate operational capabilities at scale, we implement the system using NASA's North American Land Data Assimilation System version 3 (NLDAS-3), which provides surface meteorological forcing and land-surface model output across North and Central America at 1-km resolution, from which drought diagnostics are derived. The system integrates five core contributions: (1) scalable data virtualization using Kerchunk-based cloud optimized access, achieving a 1.5 to 4.6 times improvement in I/O latency across benchmark queries spanning regional single-day extractions (4.6 times speedup) to continental monthly aggregations (1.5 times speedup); (2) transparent code generation through Microsoft Azure AI Foundry agents that expose executable Python workflows for scientific verification; (3) persistent conversational memory enabling multi-turn analytical discourse across sessions; (4) intelligent query validation that enforces dataset boundaries and resolves ambiguous requests before execution; and (5) a multi-agent architecture coordinating query parsing, code generation, and visualization. We evaluate the system through drought-monitoring workflows, demonstrating reliable code generation, accurate results validated against reference computations and the operational U.S. Drought Monitor, and efficient operation across increasingly complex tasks. By bridging natural-language interfaces with rigorous hydrological analysis, Hydrology Copilot advances beyond proof-of-concept demonstrations to provide a deployable framework for operational Earth science applications.
Abstract We introduce the Vector-QM24 (VQM24) dataset comprehensively covering all possible neutral closed-shell small organic and inorganic molecules with up to five heavy (p-block) atoms: C, N, O, F, Si, P, S, Cl, Br. All valid stoichiometries, Lewis-rule-consistent graphs, and stable conformers (identified via GFN2-xTB) were enumerated combinatorially, yielding 577k conformational isomers spanning 258k constitutional isomers and 5,599 unique stoichiometries. DFT (ωB97X-D3/cc-pVDZ) optimizations were performed for all, and diffusion quantum Monte Carlo (DMC@PBE0(ccECP/cc-pVQZ)) energies are provided for 10,793 lowest-energy conformers with up to 4 heavy atoms. VQM24 includes structures, vibrational modes, rotational constants, thermodynamic properties (Gibbs free energies, enthalpies, ZPVEs, entropies, heat capacities), and electronic properties such as atomization, electron interaction, exchange-correlation, dispersion energies, multipole moments (dipole to hexadecapole), alchemical potentials, Mulliken charges, and wavefunctions. Machine learning models of atomization energies on this dataset reveal significantly higher complexity than QM9, with none achieving chemical accuracy. VQM24 offers a rigorous, high-fidelity benchmark for evaluating quantum machine learning models.
Multiphysics modeling of liquid metal fusion blankets, which produce tritium and convert energy of neutrons created via fusion reactions into heat, is crucial for predicting performance, ensuring structural integrity, and optimizing energy production. While traditional blanket modeling of liquid metal flows during normal steady operating conditions commonly employs the inductionless approximation of the magnetohydrodynamics (MHD) equations, transient scenarios, when the plasma-confining magnetic field varies on millisecond time scales, require a full-induction MHD approach that dynamically evolves the magnetic field via the time-dependent induction equation. This paper presents the formulation, implementation, and initial verification of a full-induction MHD solver integrated within the open-source Vertex-CFD framework, which aims to achieve tight multiphysics coupling, a flexible software design enabling easy extension and addition of physics models, and performance portability across computing platforms. The solver utilizes finite element spatial discretization, implicit Runge–Kutta time integration, and an inexact Newton method to solve the resulting discrete nonlinear system, leveraging Trilinos packages for efficient computation. Verification against selected benchmark problems demonstrates accuracy and robustness of the solver. Furthermore, when the solver is applied to an idealized blanket model in 2.5D and full 3D, results obtained with Vertex-CFD are in good agreement with recently published quasi-2D simulations. These findings establish a computational foundation for future simulations of transient MHD phenomena in liquid metal blankets with Vertex-CFD, and open avenues for future extensions and performance optimizations.
Using global interpolation functions (GIF's) boundary element solutions are obtained for two-dimensional laminar flows. Two schemes are proposed for handling the convective terms. The first treats convection as a forcing function, and converts the flow equations to pseudo-Poisson equations. In the second scheme, some convective effect is incorporated into the fundamental solution used in constructing the pertinent integral equations. The lid-driven cavity flow is selected as the benchmark problem.
The Biological and Physical Sciences (BPS) Division of NASA’s Science Mission Directorate (SMD), and its predecessors, has sponsored extensive flight and ground experiments yielding benchmark datasets in many materials science research areas including thermophysical properties and solidification microstructure formation and evolution. The subject study sought to motivate and focus BPS’s engagement within the broader Integrated ICME community to understand the phenomena underlying material processing, structure, and properties in the microgravity environment of space and to support future space exploration efforts.
Solving the alternating current power flow equations in real time is essential for secure grid operation, yet classical Newton–Raphson solvers can be slow under stressed conditions. Existing graph neural networks for power flow are typically trained on a single system and often degrade on different systems. We present PowerModelsGAT-AI, a physics-informed graph attention network that predicts bus voltages and generator injections. The model uses bus-type-aware masking to handle different bus types and balances multiple loss terms, including a power-mismatch penalty, using learned weights. We evaluate the model on 14 benchmark systems (4 to 6,470 buses) and train a unified model on 13 of these under contingency conditions with up to two branch outages, achieving an average normalized mean absolute error of 0.89% for voltage magnitudes and R 2 >0.99 for voltage angles. We also show continual learning: when adapting a base model to a new 1,354-bus system, standard fine-tuning causes severe forgetting with error increases exceeding 1000% on base systems, while our experience replay and elastic weight consolidation strategy keeps error increases below 2% and in some cases improves base-system performance. Interpretability analysis shows that learned attention weights correlate with physical branch parameters (susceptance: r=0.38 ; thermal limits: r=0.22 ), and feature importance analysis supports that the model captures established power flow relationships.