Search NASA⌕ Search

SEARCH · Search NASA

Results for “Trajectory optimization”

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 73 records · Page 4
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Nonadiabatic Force Matching for Alchemical Free-Energy Estimation

We propose a method to compute free-energy differences from nonadiabatic alchemical transformations by using flow-based generative models. The method, nonadiabatic force matching, hinges on estimating the dissipation along an alchemical switching process in terms of a nonadiabatic force field that can be learned through stochastic flow matching. The learned field can be used in conjunction with short-time trajectory data to evaluate upper and lower bounds on the alchemical free energy that variationally converge to the exact value if the field is optimal. Applying the method to evaluate the alchemical free energy of atomistic models shows that it can substantially reduce the simulation cost of a free-energy estimate at a negligible loss of accuracy when compared with thermodynamic integration.

Computational chemistry↗
DOE OSTI2026

From atomistic models to machine learning: Predictive design of nanocarbons under extreme conditions

The formation of technologically valuable nanocarbon structures under extreme conditions, such as those produced during high-explosive detonations, remains poorly understood but holds significant potential for the development of controlled synthesis pathways. While detonation shockwaves provide the high-pressure, high-temperature environment required for nanodiamond formation, subsequent cooling and decompression dictate whether the diamond phase is preserved or transformed into other nanocarbon structures. Here, in this study, we employ GPU-accelerated reactive molecular dynamics (ReaxFF) simulations to investigate the graphitization and structural remodeling of detonation nanodiamond under nonlinear quench and pressure-release trajectories. We further investigate how the initial nanodiamond morphology; cuboctahedral, octahedral, or hexagonal prism influences the resulting transformation products. Evolution of nanostructure, allotrope (via simulated x-ray diffraction), carbon hybridization, and ring statistics are tracked during a two-stage quench from 5000 K to 60 GPa. Rapid cooling combined with slow decompression optimizes cubic diamond retention, whereas slow cooling with rapid pressure release promotes surface-to-core graphitization, producing concentric sp 2 -hybridized layers and hollowed inner shells. Octahedral nanodiamonds evolve into carbon nano-onions, initially forming bucky diamonds that progressively transform into fully sp 2 -hybridized structures, while hexagonal prisms preferentially form parallel-stacked graphite layers resembling carbon dots. Transient hexagonal diamond (lonsdaleite) emerges as an interfacial phase, suggesting potential reversibility in the shock-induced graphite-to-diamond transformation pathway transformation route. To extend predictive capabilities, we trained machine learning (ML) regressors on over 10 5 node-hours of molecular dynamics (MD) trajectories. A multilayer perceptron (MLP) model reliably predicts the number of graphitized layers from temperature–pressure trajectories with a coefficient of determination (R 2 ) exceeding 0.90. This high predictive fidelity enables efficient, high-throughput mapping of the synthesis parameter space for optimized graphitization outcomes. Collectively, morphological control combined with optimized quench–decompression conditions promote the selective synthesis of nanocarbon allotropes. This work establishes a data-driven framework for the rational, a priori design of carbon nanomaterials for applications in energy storage, sensing, and biomedicine.

Detonation nanodiamond remodeling↗
DOE OSTI2026

Gearbox bearing crack growth prognostics and uncertainty quantification with physics-informed machine learning

This paper introduces the extreme theory of functional connections (X-TFC), a physics-informed machine learning algorithm, and tailors it to estimate the remaining useful life (RUL) of wind turbine gearbox bearings experiencing fatigue crack growth. Unlike purely data-driven methods, X-TFC embeds a physics model, based on Head's theory in this work, into its training objective. The core of X-TFC is a random-projection single-layer neural network trained via an extreme learning machine, which requires only limited damage progression data and solves for output weights with a least-squares optimization algorithm. A composite loss function balances the network's fit to observed degradation data against the residuals of the governing crack growth differential equation, ensuring the learned damage trajectory remains physically plausible. When applied to a vibration-based health-index (HI) dataset measured during the growth of a crack on the inner ring of a high-speed bearing in a wind turbine gearbox (Bechhoefer and Dubé, 2020), X-TFC achieves near-zero prediction bias. Even when trained on only the first 10 %–20 % of the damage progression data, with sufficient physics weighting its predictions remain monotonic and smooth, delivering high prognosability and trendability. To quantify the epistemic uncertainty, we employ a Monte Carlo ensemble of independently initialized X-TFC models trained on noise-perturbed data, which yields confidence intervals around each RUL estimate and captures both model-parameter and epistemic uncertainty. In addition to a vibration-based HI, we demonstrate that the proposed framework can be directly applied to a supervisory control and data acquisition (SCADA) data-based HI (Eftekhari Milani et al., 2026) measured during similar wind turbine gearbox bearing crack faults, preserving its accuracy and interpretability. This extension shows the versatility of our approach, which is applicable to bearings of multiple gearbox manufacturers, models, and ratings using only SCADA data. By integrating domain knowledge with machine learning, X-TFC offers a rapid, reliable tool for crack prognostics. Its adaptability to other bearing failure modes, such as pitch bearing ring cracks, positions X-TFC as a powerful enabler of data-driven, physics-informed asset management in the wind energy sector and beyond.

17 WIND ENERGY↗
DOE OSTI2025

Hydrogen and Electric Charging Infrastructure for Heavy-Duty Trucks: A Nationally Scalable Megaregion Assessment

Decarbonizing regional and long-haul freight is challenging due to the limitations of battery-electric commercial vehicles and infrastructure constraints. Hydrogen fuel cell medium- and heavy-duty vehicles (MHDVs) offer a viable alternative, aligning with the decarbonization goals of the Department of Energy and commercial entities. Historically, alternative fuels like compressed natural gas and liquefied propane gas have faced slow adoption due to barriers like infrastructure availability. To avoid similar issues, effective planning and deploying zero-emission hydrogen fueling infrastructure is crucial. This research develops deployment plans for affordable, accessible, and sustainable hydrogen refueling stations, supporting stakeholders in the decarbonized commercial vehicle freight system. It aims to benefit underserved and rural energy-stressed communities by improving air quality, reducing noise pollution, and enhancing energy resiliency. This research also provides a blueprint for replacing diesel in over-the-road Class 8 freight truck applications with hydrogen fueling solutions. The study focuses on the Texas Triangle Megaregion (I-45, I-35, and I-10), the I-10 corridor between San Antonio, TX, and Los Angeles, CA, and the I-5/CA-99 corridors between Los Angeles, CA, and San Francisco, CA. This area represents a significant portion of U.S. heavy-duty freight movement, carrying ~8.5% of the national freight volume. Using the OR-AGENT (Optimal Regional Architecture Generation for Efficient National Transport) modeling framework, the study conducts an advanced assessment of commercial vehicles, road and freight networks, and energy systems. The framework integrates data on freight mobility, traffic, weather, and energy pathways to deliver a region-specific, optimized vehicles powertrain architectures, infrastructure deployment solutions, operational logistics, and energy pathways. By considering all vehicle origin-destination pairs utilizing these corridors and all feasible fueling station location options, the framework's genetic algorithm identifies the minimum number and optimal locations of hydrogen refueling stations, ensuring no vehicle is stranded. It also determines fuel schedules and quantities at each station. A roadmap for station deployment based on multiple adoption trajectories ensures a strategic rollout of hydrogen refueling infrastructure.

Sujan, Vivek [ORNL] (ORCID:0000000269882342)↗
DOE OSTI2025

Low Peripheral Blood Counts and Elevated Proinflammatory Cytokines Signal a Poor CD19 Chimeric Antigen Receptor T-cell Response in Acute Lymphoblastic Leukemia

CD19 chimeric antigen receptor T-cell (CAR-T) therapy has significantly improved outcomes for patients with relapsed/refractory B-cell acute lymphoblastic leukemia (R/R B-ALL). However, approximately 20% of patients fail to achieve a complete remission (CR), and some develop severe, life-threatening toxicities. Understanding the biological mechanisms underlying both dysfunctional responses and severe toxicity is essential for optimizing patient management and improving therapeutic efficacy. This study aimed to (1) characterize cytokine profiles associated with dysfunctional responses and severe toxicity following CAR-T infusion, (2) examine the timing and trajectory of cytokine changes in relation to treatment outcomes, and evaluate potential strategies for mitigating toxicity and treatment failure. We conducted a comprehensive analysis of serum cytokine profiles in 86 adult and pediatric patients undergoing autologous CD19 CAR-T therapy for B-ALL. Patients were categorized into three groups: (1) Dysfunctional response—Patients who failed to achieve a minimal residual disease-negative CR (MRD-CR) by Day 63 or who experienced recurrence of CD19+ disease in the setting ongoing CAR-T cell detection before Day 63. (2) Functional response with severe cytokine release syndrome (CRS) and/or neurotoxicity (NTX)—Patients with best response of MRD-CR by Day 63 who experienced grade 3 or higher CRS or NTX. (3) Functional response without severe CRS or NTX—Patients with best response of MRD-CR by Day 63 who did not experience grade =3 CRS or NTX. Cytokine levels were measured during the first-week postinfusion and correlated with treatment efficacy, toxicity outcomes, complete blood counts, and CAR-T expansion dynamics. This analysis aimed to better understand how cytokine profiles relate to patient outcomes and immune responses in CAR-T therapy. Patients with dysfunctional response exhibited decreased neutrophils, platelets, and levels of granulocytic cytokines (suggestive of low bone marrow reserve) alongside elevated pro-inflammatory cytokines by Day 1. Functional response with severe toxicity patients showed a progressive rise in proinflammatory cytokines, reaching similar levels to dysfunctional response patients by Day 7. We observed that high cytokines at both the Day 1 and Day 7 time points were associated with poor survival. These findings remained significant when adjusting for high disease burden, a known predictor of severe inflammatory toxicity and lack of response. Early post-CAR-T infusion inflammation is associated with both dysfunctional response and severe toxicity—even after adjusting for disease burden. This suggests that inflammation, in addition to disease burden, plays a role in determining patient outcome. Therefore, strategies aimed at reducing the pro-inflammatory state prior to or early after CAR-T cell infusion may improve outcomes for R/R B-ALL patients.

Serum cytokines↗
DOE OSTI2025

Selective capture and recovery of uranium oxide colloids from aqueous soil suspensions using high gradient magnetic filtration

High Gradient Magnetic Filtration (HGMF) is a promising method for the selective capture and recovery of uranium oxide from surface soils. To date, however, magnetic filtration of uranium oxide has only been demonstrated at a proof-of-principle scale using relatively small filters (<5 cm 3 ) at low flowrates (<60 mL/min). Here, to explore the efficacy of magnetic filtration of uranium oxide at a larger scale, a newly designed HGMF apparatus that is more than an order of magnitude larger than our earlier filters (106 cm 3 ) was designed, fabricated, and tested at relatively high flowrates. Filtration experiments were performed using aqueous uranium oxide particle suspensions with Arizona Road Dust (ARD) as a soil simulant. At a flowrate of 125 mL/min, the apparatus’ uranium capture rate was exceptionally high (96 %), but selectivity was poor due to the high rate of capture for diamagnetic soil constituents (e.g., 77 % for silicon). All particles were captured at a lower rate when the flowrate was increased to 250 mL/min, but uranium selectivity was significantly increased due to the more substantial reduction in diamagnetic particle capture (i.e., capture rate of 77 % and 15 % for uranium and silicon, respectively). When backwashing the apparatus at the same flowrates used during filtration experiments, the rate of uranium recovery tended to be fairly low. Nevertheless, higher flowrates (1 L/min) and sonication were both shown to be highly effective methods of increasing uranium recovery. Magnetic field simulations were also performed to investigate potential optimizations to the design of the apparatus. These simulations showed that the intensity of the applied magnetic field could be increased by increasing the thickness of the steel magnetic housing. Additionally, stochastic trajectory simulations were performed to investigate the potential mechanisms of particle capture.

HGMF↗
DOE OSTI2025

CRCNS22 Learning Rules in the Hippocampus and their Mapping to Neuromorphic Systems (Final Technical Report)

Large scale biologically-realistic computational models are key to investigating the interplay between structure and function in nervous systems, thus paving the way to new clinical methods and neuro-inspired computing solutions. This project focuses on the hippocampus, in particular the CA3-CA1 regions, due to their role in associative learning and memory, pattern separation and completion, and spatial navigation. Investigations into the neuronal organization and learning rule(s) of this circuit can shed light into how declarative memories are formed, stored, recalled and forgotten and inform computational, experimental and clinical neuroscience work. Our project aims at developing a novel data-driven methodology supported by a broad heterogeneous base of neuroscience experimental knowledge and inspired from advances in computer science and engineering. Specifically, this work will benchmark existing and new learning rules within a full-scale spiking neural network simulation of the CA3-CA1 region. The model will be based on an open-source repository, called the Hippocampome, which contains neuronal morphologies, firing patterns, synapse probabilities, and most other required parameters for all known neuron types in the rodent hippocampal formation. The model will be first trained in a supervised fashion for associative memory tasks using backpropagation through time traditionally used in computer science, enhanced with a new technique called the surrogate gradient method. This optimization method will be used to obtain a global loss minimization, but it is not biologically inspired as it assumes the use of data not locally available to the synapses. However, we propose its use as a benchmarking tool, to compare the training performance of local biologically plausible and hardware-mappable learning rules at scale. New rules or combinations will be proposed and tested as needed, based on the obtained results. Progress in this area will also drive the development of novel hardware-mappable algorithms for continual lifelong learning and categorization of new events from few presented examples. This project goes beyond the existing state-of-the-art by looking at large scale realistic neuronal circuits as networks trainable via global optimization methods such as surrogate gradient descent. The objective function of the brain that supports learning is largely unknown, but it is likely that it operates through local learning rules. Studying network trajectories around local minima as proposed in this work represents a useful strategy for understanding whether a network is training by using a specific (set of) learning rule(s). Starting from a completely untrained network is a challenging test since it is difficult to determine how the learning rule affects the trajectory of the network. This interdisciplinary project will help understand what rule governs learning in these regions or if multiple learning rules are involved. The work will develop a robust methodology to measure if the network is converging to the target solution, oscillating around it, or diverging away.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Design update on the transition beamline for the CEBAF Energy Upgrade

For Jefferson Lab’s 22GeV upgrade, two new permanent-magnet Fixed-Field Alternating Gradient (FFA) arcs will be integrated to serve the accelerator’s six highest-energy recirculation passes. Connecting these FFA arcs to the existing linear accelerator (linac) requires a carefully engineered transition section. The current design has two parts where the first part adiabatically matches the beam dispersion and orbit trajectories, while the second part aligns the Twiss parameters (alpha and beta functions) with those at the linac entrance. Given the tight spatial constraints and multiple matching requirements, a genetic algorithm is being explored to optimize the beam optics matching. This paper presents the current progress in developing and optimizing this transition.

Accelerator Physics↗
DOE OSTI2024

Precise Motion Control of Hybrid Hydraulic Electric Architecture (HHEA)

Off-highway heavy-duty vehicles have been long-standing users of hydraulic systems for power transmission and control. However, traditional hydraulic systems suffer from significant energy losses which lead to increased operating costs and a larger carbon footprint due to higher CO2 emissions. Improving the efficiency of these mobile machines is crucial not only for reducing their environmental impact but also for saving billions of dollars in operating costs. Currently, the state-of-the-art Load Sensing Architecture uses throttling valves for control, which significantly reduces its efficiency and does not recuperate energy from over-running loads. Researchers have developed several architectures such as Common Pressure Rail systems, Displacement Control, STEAM, and Electrohydraulic Architecture to improve the efficiency of off-road mobile machines. However, each of these architectures has its drawbacks. To increase system efficiency and take advantage of electrification benefits, our research group has developed a novel Hybrid Hydraulic-Electric Architecture (HHEA). The HHEA can significantly improve efficiency, decrease the size of electrical components, and maintain control performance. This new architecture has the potential to revolutionize the off-highway mobile machine industry and lead to a more sustainable future. The HHEA uses a set of common pressure rails to provide the majority of power to the actuators via power-dense hydraulics and uses electric motors for precise control and power modulation. In the context of off-road mobile machines, energy savings are undoubtedly important but it is equally important to consider the machines’ ability to perform tasks with precision and accuracy according to given commands. Therefore, precise motion control is of utmost importance to maintain the utility of Hybrid Hydraulic-Electric Architecture (HHEA). The HHEA presents a unique challenge to motion control due to the discrete pressure changes that occur when the system switches between selected pressure rails. These changes are made to minimize system inefficiencies or to keep the system within the torque capability of the electric motor. Hence, it is important to solve the motion control challenges for HHEA. This thesis aims at developing an effective motion control strategy for HHEA. The dissertation presents a two-tiered control strategy for HHEA, comprising a high-level and a low-level controller. The primary responsibility of the high- level controller is to optimize energy efficiency by making informed pressure rail selections. On the other hand, the low-level controller is focused on achieving precise motion control of the HHEA, which is crucial for realizing the desired reference trajectories. To achieve this, the low-level controller utilizes a passivity-based backstepping integral controller as the nominal control, which handles the motion control between two pressure rail switches. Additionally, a separate least norm controller is utilized as a transition controller to manage motion control during pressure rail transitions. The effectiveness of the combined control strategy is demonstrated through experiments conducted on two hardware-in-the-loop testbeds. Furthermore, the HHEA is installed on the boom and stick actuators of a backhoe arm to build a Human-in-the-Loop system that a human operator can control. A real-time rail switching algorithm is developed to determine pressure rail switching based on present duty cycle information from the operator. Modifications have been made to the human-machine interface to achieve more intuitive control. Modifications include performing control in the task-oriented coordinates, incorporating pressure feedback to enhance control with physical interaction, and using velocity field control to simplify multi-degree-of-freedom tasks and to enable novice operators to perform them with reduced risk, improved efficiency, and productivity. The research in this dissertation makes significant contributions to the field of off-road mobile machine control, providing a novel and effective control strategy for the HHEA, and demonstrating the potential for simplified machine operation.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2026

Optimizing infrasound observations for sample return capsule re-entry: Insights from OSIRIS-REx and Hayabusa2

The atmospheric entry of meteoroids presents a rare and unpredictable phenomenon, posing challenges for systematic observation and detailed characterization. Such events are nonetheless critical for advancing understanding of acoustic wave propagation, atmospheric structure, and entry dynamics. In contrast, sample return capsules (SRCs) from space missions follow well constrained re-entry trajectories, enabling planned observations of shock wave generation and propagation under controlled conditions. This study compares two SRC atmospheric entries, Hayabusa2 in 2020 and OSIRIS-REx in 2023, to assess how different infrasound array configurations influence shock wave detection and trajectory validation. Hayabusa2’s re-entry was monitored using a distributed network of 28 portable infrasound sensors across seven arrays in Woomera, Australia, permitting three-dimensional reconstruction of the trajectory and analysis of wave characteristics. For OSIRIS-REx, a compact four-sensor array deployed near Eureka Airport provided trajectory confirmation through arrival-time differences and back-azimuth estimates. Spectral and waveform analyses revealed differences in signal properties associated with variations in entry angle and velocity. The results illustrate both the strengths and the limitations of the deployed array configurations. The dense and distributed arrays during Hayabusa2’s re-entry enabled detailed trajectory reconstruction, whereas the compact array at Eureka primarily provided confirmation of signal coherence and back-azimuth consistency. These case studies highlight that even relatively small arrays, if located close to the predicted ground track, can still capture useful information on arrival direction and timing. Such insights provide practical guidance for planning future observational campaigns of SRC returns and other controlled atmospheric entries, and may also inform approaches to opportunistic observations of meteoroid events and other atmospheric acoustic phenomena.

celestial mechanics: orbit determination↗
DOE OSTI2024

Community Public Mobility Using On-Demand, Low-Speed Electric Vehicles: A Case Study in Downtown St. Louis, Missouri

Legacy fixed route transit systems designed to serve commuters struggle to provide efficient and effective service for short neighborhood trips and for population groups unable to access and egress transit stops using active modes (e.g., elderly, disabled). Neighborhood on-demand transit (ODT) services using low-speed electric vehicles (LSEV) are an innovative technological solution that can help fill this gap in service (e.g., short, high-frequency trips) for diverse populations and trip types. This study evaluated user characteristics and travel behavior for a neighborhood ODT service (using LSEVs) in downtown St. Louis, Missouri using responses from a community survey (n=244), ridership data, and vehicle trajectory information. A comparative analysis between neighborhood ODT, fixed route transit, and transportation network companies (TNC) was also conducted from the perspectives of total travel time, cost, and greenhouse gas emissions. Ultimately, the goal of the analysis was to motivate and inform holistic public mobility systems where different services are optimized to meet specific community needs. Findings indicate that the neighborhood ODT was effective at reaching diverse populations (elderly (20%), lower income (27%), and households with limited access to private vehicles (34%)). ODT reduced total travel time by 32% compared to fixed route transit, produced 2.4 - 4.3 times less greenhouse gas emissions per passenger mile (compared to transit and TNCs), and was more affordable (free to users) than alternative options ($1 for transit, $10-12 for TNCs). Overall satisfaction rates were high, with 80% of respondents rating the service a 4 or 5 out of 5.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2026

Performance of the Particle-Identification Silicon-Telescope Array coupled with the VAMOS++ magnetic spectrometer

The Particle-Identification Silicon-Telescope Array (PISTA) is a new detection system designed for high-resolution studies of fission process induced by multi-nucleon transfer in inverse kinematics. It is specifically optimized for experiments with the VAMOS++ magnetic spectrometer at GANIL (Grand Accélérateur National d’Ions Lourds). The array comprises eight trapezoidal ΔE-E silicon telescopes arranged in a lamp shade configuration. Each telescope integrates two single-sided stripped silicon detectors, enabling target-like recoil identification, energy loss measurements, and trajectory reconstruction. Positioned in close proximity to the target, PISTA’s compact geometry achieves high-efficiency tracking of target-like recoils produced in multi-nucleon transfer reactions at Coulomb barrier energies. The spatial segmentation of the array allows precise determination of the mass and charge of the target-like nucleus, and excitation energy of fissioning systems. This work presents the particle identification and excitation energy reconstruction performances for the interactions of 238 U beam with 12 C target. An excitation energy resolution of 800 keV (FWHM) was determined together with mass resolution of 1.1% (FWHM). The combination of PISTA and VAMOS++ magnetic spectrometer enables unprecedented investigations of the fission process as a function of the excitation energy of the fissioning nucleus, particularly for exotic systems produced in transfer-induced reactions.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Numerical analysis of air dehumidification through electrospray-enhanced vortical flow cyclone separator

The dehumidification of air is a highly energy-intensive process, typically requiring vapor compression systems to cool the air significantly below its dew point temperature. Here, this study numerically investigates air dehumidification achieved by electrostatic and dielectrophoresis forces in a converging duct and centrifugal principles in cyclone separators. An electrospray was employed to energize water droplets, which then captured water vapor via dielectrophoresis. A novelty of the present study is the inclusion of a cyclone separator positioned downstream of the electrospray. This combination increased the dehumidification efficiency and controlled the flow of the condensate. By integrating electrospray technology and applying an electric field, this study examines the impact of particle separation efficiency, pressure drops, and cyclone performance on air dehumidification. The control parameters for air dehumidification were investigated. Three cyclones were modelled for airflow rates of 5 cubic feet per minute [cfm] (0.0024 m 3 /s), 38 cfm (0.0179 m 3 /s), and 200 cfm (0.094 m 3 /s). The results indicated air dehumidification was up to 5.7 % under environmental conditions, similar to those used to validate the CFD model. The degree of air dehumidification was primarily influenced by the electric charge acquired by the water droplets, their flight trajectory, and the vorticity intensity within the cyclone. Air dehumidification could increase by up to 13 % by increasing the number of droplets injected into the air stream, the flight time, and the electrical charge. The breakdown electric field strength threshold near the electrospray was a limiting factor for further enhancing dehumidification performance. This study examines the optimization of an electro-assisted dehumidification system that includes cyclone-based droplet separation through numerical analysis. The results provide insights into the design of energy-efficient air treatment technologies.

42 ENGINEERING↗
DOE OSTI2025

Advancing Multiscale Simulation of Plasma-Surface Interfaces

We report the development of an atomistic-informed, surface-state-dependent predictive model for particle exchange in a carbon-tungsten plasma-surface interface. The predictive model uses machine learning (ML) techniques to learn the energy and angular distributions for particle exchange and rate functions for surface state evolution from molecular dynamics simulations of cumulative bombardment of tungsten by energetic carbon ions. Each predictive component is sensitive to the energy and trajectory of incident plasma species and the surface state. The surface state is represented by a set of surface state descriptors, which were derived from the atomistic surface state for each independent carbon bombardment event. These descriptors are representative of the composition and degree of amorphization of the outermost angstrom of surface material and were chosen to optimize predictive performance for particle exchange at the interface. The distributions for particle exchange (reflection/sputtering) are demonstrated to vary with each surface state descriptor, motivating the development of surface-state-dependent particle exchange models for plasma simulations. The performance of various ML methods was compared, including polynomial quantile regression, artificial neural networks, k-nearest neighbors, and random forest algorithms, with polynomial regression performing the best for interpolation and extrapolation of learned relationships. In addition to the particle exchange model, a neutral network was developed and used to identify data sufficiency throughout surface descriptor space, which will enable real-time feedback during future data production to ensure data is produced where it is most needed, and we provide commentary on improvements to the data production workflow for future endeavors.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2025

Using "AI Poincare" to analyze non-linear integrable optics

This study dives into the applicability of using automated discovery of conserved quantities in dynamical systems relevant to accelerator physics. Specifically, we explore the performance of AI Poincaré in analyzing numerical trajectory data obtained using the McMillan system of non-linear integrable optics. A comprehensive evaluation of the algorithm's performance is conducted through diverse methodologies. These include the analysis of the estimated number of conserved quantities embedded in a dataset and the deviation of interpolated points on the inferred manifold with respect to points in actually in the dataset. the investigation identifies an optimal range of perturbation distances where the underlying manifold extraction algorithm inside AI Poincaré exhibits optimal performance. Additionally, an improved neural network architecture is proposed based on the observed results. Finally, we apply the algorithm to preliminary experimental data from the Integrable Optics Test Accelerator at Fermilab to successfully infer the number of conserved quantities even in the presence of fast decoherence of the measured signal.

Osmanov, Lazare [Free U. Tbilisi]↗
DOE OSTI2026

HARMONY: Large-Scale Architecture Search for Efficient Hybrid Language Models

As large language models scale to trillions of parameters, their computational and memory requirements present critical challenges for efficient training and deployment. While Mixture of Experts (MoE) architectures enable efficient scaling through sparse parameter activation, and state-space models like Mamba offer linear-time complexity, principled methods for combining these paradigms remain undeveloped. We introduce HARMONY (Hybrid Architecture Research for Mamba, Optimized with Neural efficiencY), a multi-objective evolutionary neural architecture search framework for discovering efficient hybrid language models that integrate Transformer attention mechanisms, Mixture-of-Experts routing, and Mamba state-space components. Through large-scale distributed search using 16,384 MI250X GPUs on the Frontier supercomputer, HARMONY explores a comprehensive design space encompassing six attention variants (MHA, MQA, GQA, MLA, SWA, and Mamba-2), variable MoE configurations with both routed and shared experts, and extensive Mamba hyperparameters. Our framework discovers heterogeneous architectures that balance training performance with computational efficiency through multi-objective optimization incorporating latency penalties and fitness-based selection. Analysis of discovered architectures reveals that optimal hybrid designs favor heterogeneous component mixing rather than homogeneous patterns, with Mamba-2 and Multi-Head Latent Attention (MLA) emerging as preferred mechanisms. Discovered architectures demonstrate superior training efficiency: our best configuration achieves a final perplexity of 1.0874 with 2.38B parameters while processing 4,320 tokens/second, outperforming significantly larger manually designed models. Full-scale evaluation shows HARMONY's top architectures achieve better loss trajectories than equivalently-sized models using state-of-the-art configurations including Mixtral, Jamba, and Samba. Additionally, we demonstrate 91% weak scaling efficiency when training discovered 36B-parameter models across 1,024 GPUs. HARMONY is released as an open framework with comprehensive tools for building and training hybrid models using expert-data-pipeline parallelism, democratizing access to automated architecture design for next-generation language models.

Herron, Emily [ORNL] (ORCID:0000000273008172)↗
DOE OSTI2024

Underlying mechanism of structural transformation between GaSb and GaAs response to intense electronic excitation

Ion irradiation of semiconductors has emerged as a promising approach for fabricating self-organized nanosystems with high atomic precision, despite often being accompanied by undesirable phenomena. Exploring the mechanisms underlying structural transformations is crucial for assessing nanostructure array types under complex irradiation environments. By quantitatively calculating the thermodynamically driven processes and analyzing the impact of intrinsic structural parameters, distinct structural transformations in response to intense electronic excitation are systematically investigated in gallium antimonide (GaSb) and gallium arsenide (GaAs) systems. In high-energy regimes, the nanofibers layer of GaSb exhibits intriguing structural discrepancy, characterized by partial nanofibers with coherent boundaries, interspersed nanopores accompanied by antisite defects and Ga precipitates, distinguishing to a series of discontinuous latent tracks that emerged within cylindrical trajectories in GaAs. Furthermore, significant diffusion behaviors of the nanohillocks are discovered in GaAs, with higher average roughness than GaSb, driven by the gradient stress distribution influenced by the free-surface effects. The deposition energy for melting phase formation, Gibbs free energy, and Ga diffusion coefficients contribute to the distinctive structural features, evidencing relatively stable morphological configurations and higher irradiation resistance in GaAs. Consequently, special optoelectronic properties associated with structural discrepancies facilitate the design and optimization of material functionalities by irradiation technologies.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Single-Molecule Fluorescence Microscopy Reveals Energy Transfer Active versus Inactive Nanocrystal/Dye Conjugate Pairs

Defect-mediated energy transfer (EnT) is a radiative process that occurs between donor defect states in the forbidden bandgap of semiconductor nanocrystals (NCs) and dye molecules bound to their surfaces. The EnT efficiency depends on the number of dye molecules attached to each NC, the donor–acceptor distance, and the dipole orientation factor between the donor and acceptor, all of which vary across all individual NCs in a sample. While ensemble-level fluorescence spectroscopy measurements have provided average values for donor–acceptor distances, dye-to-NC ratios, and EnT rate constants, questions remain about the impact of donor/acceptor heterogeneity on observed EnT efficiencies. Notably, ensemble-level measurements cannot distinguish between bare NCs and EnT-active versus inactive NC/dye pairs in the same sample batch, limiting the ability to design systems with 100% EnT efficiency. To address this, we studied defect-mediated EnT between AlexaFluor 555 dye acceptors chemically bound to ZnO NC donors at the level of single molecules and single NCs. Interestingly, 20% of bound NC/dye pairs are EnT-inactive, likely contributing to residual defect photoluminescence (PL) observed in ensemble-level measurements and reducing overall EnT efficiency. Single particle-level ZnO defect PL and acceptor fluorescence trajectories exhibited distinct microfluctuations, which are absent in bare ZnO NCs. We hypothesized that our observations can be explained with a competitive dye fluorescence quenching pathway, possibly due to charge transfer between the excited state dye and the ZnO NC. Numerical simulations of single-molecule PL traces for this scenario produced microfluctuations consistent with the experimental results. These findings highlight the impact of sample heterogeneity on EnT processes and provide insights for designing light-harvesting systems with optimized EnT efficiency.

25 ENERGY STORAGE↗