Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “accelerator modeling”

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 1,207 records · Page 67
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Nitromethane Decomposition via Automated Reaction Discovery and an Ab Initio Corrected Kinetic Model

In the explore the systematic construction of kinetic models from in silico reaction data for the decomposition of nitromethane. Our models are constructed in a computationally affordable manner by using reactions discovered through accelerated molecular dynamics simulations using the ReaxFF reactive force field. The reaction paths are then optimized to determine reaction rate parameters. We introduce a reaction barrier correction scheme that combines accurate thermochemical data from density functional theory with ReaxFF minimal energy paths. We validate our models across different thermodynamic regimes, showing predictions of gas phase CO and NO concentrations and high-pressure induction times that are similar to experimental data. The kinetic models are analyzed to find fundamental decomposition reactions in different thermodynamic regimes.

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

Toward performance-portable PETSc for GPU-based exascale systems

The Portable Extensible Toolkit for Scientific computation (PETSc) library delivers scalable solvers for nonlinear time-dependent differential and algebraic equations and for numerical optimization. The PETSc design for performance portability addresses fundamental GPU accelerator challenges and stresses flexibility and extensibility by separating the programming model used by the application from that used by the library, and it enables application developers to use their preferred programming model, such as Kokkos, RAJA, SYCL, HIP, CUDA, or OpenCL, on upcoming exascale systems. Furthermore, a blueprint for using GPUs from PETSc-based codes is provided, and case studies emphasize the flexibility and high performance achieved on current GPU-based systems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

First-principles-integrated Study of Blazar Synchrotron Radiation and Polarization Signatures from Magnetic Turbulence

Blazar emission is dominated by nonthermal radiation processes that are highly variable across the entire electromagnetic spectrum. Turbulence, which can be a major source of nonthermal particle acceleration, can widely exist in the blazar emission region. The Turbulent Extreme Multi-Zone (TEMZ) model has been used to describe turbulent radiation signatures. Recent particle-in-cell (PIC) simulations have also revealed the stochastic nature of the turbulent emission region and particle acceleration therein. However, radiation signatures have not been systematically studied via first-principles-integrated simulations. In this paper, we perform combined PIC and polarized radiative transfer simulations to study synchrotron emission from magnetic turbulence in the blazar emission region. We find that the multiwavelength flux and polarization are generally characterized by stochastic patterns. Specifically, the variability timescale and average polarization degree (PD) are governed by the correlation length of the turbulence. Interestingly, magnetic turbulence can result in polarization angle swings with arbitrary amplitudes and duration, in either direction, that are not associated with changes in flux or PD. Surprisingly, these swings, which are stochastic in nature, can appear either bumpy or smooth, although large-amplitude swings (>180°) are very rare, as expected. Our radiation and polarization signatures from first-principles-integrated simulations are consistent with the TEMZ model, except that in the latter, there is a weak correlation, with zero lag, between flux and degree of polarization.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2025

Investigating the FLASH Effect in a Rat Brain Organotypic Model With a Novel High-Energy Electron Beam

Ultrahigh dose rate (FLASH) radiation therapy is reported to reduce normal tissue toxicity while maintaining tumor control; however, mechanism(s) remain obscure. To study FLASH mechanisms in brain tissue, we developed a novel experimental platform featuring a specialized high-energy electron linear accelerator, High Intensity Gamma Ray Source (HIGS), paired with an organotypic ex vivo brain metastasis model. We varied interpulse spacing to modulate the mean dose rate (MDR) of our unique 35 MeV electron beam, while maintaining extremely high instantaneous dose rate (IDR). We characterized dosimetry and targeting accuracy of the FLASH beam with film dosimetry. We combined this FLASH beam with an organotypic rat brain slice/breast carcinoma coculture model of brain metastasis to assess effects on normal and neoplastic tissues. Live-cell and bioluminescence imaging demonstrated cancer cell growth effects, whereas normal tissue responses and immune activation were assessed using live-cell imaging, cytokine profiles, and confocal microscopy. Here, we performed comparison experiments with 20 MeV electrons from a Varian clinical linear accelerator (VCLA) using conventional dose rates. The highest IDR of the FLASH beam to date was 20.7 ± 0.6 MGy/s, with maximum MDR of 20.7 MGy/s delivered in 1 pulse of 1 µs duration. Beam targeting was accurate to <1 mm and reproducible. HIGS-FLASH and VCLA dose rates equivalently decreased cancer cell growth. HIGS-FLASH irradiation significantly increased tumor necrosis factor α and fractalkine levels and confocal microscopy revealed distinct changes in microglial morphology slices suggesting microglia activation. Our novel experimental platform produces extremely high dose rates and rapid normal/neoplastic tissue readouts for mechanistic research into the effects of FLASH radiation in the brain. HIGS-FLASH irradiation induces comparable cancer cell growth inhibition but differential effects on cytokines and microglial morphology, suggesting that acute innate immune responses may be involved in FLASH normal tissue effects in the brain.

Kay, Tyler V. [Duke University, Durham, NC (United↗
DOE OSTI2025

Twinac: initiation of a community-driven accelerator digital twin framework

We present the initiation of a community-driven framework for the integration of accelerator digital twins into control systems: Twinac. Few facilities have fully integrated accelerator digital twins like at Cornell’s CHESS. Many facilities have active research to employ surrogate models to aid in operational decisions like at Argonne’s ALS, MSU’s FRIB, SLAC’s LCLS-II, and Fermilab’s FAST/IOTA, PIP-II, and main complex. To lower the barrier to entry for all accelerator facilities to build and benefit from a digital twin of their own accelerators, we propose the following software framework. Twinac will provide the capability to compose one’s own digital twin using reusable components engineered at other facilities. With this model in place, Twinac will also support tools for (1) predictive maintenance systems; (2) discovery of correlated but uncontrolled environmental factors, like seasonal temperature variations causing performance changes on power supplies, magnets, etc.; and (3) prototyping and updating sophisticated optimization and controls algorithms. The Twinac framework will enable sharing and simplified deployment of modeled components and control algorithms at all facilities. With an inter-facility team to build and support the Twinac framework, it will be easy to publish and try out the latest advancements at one’s own facility.

Miceli, Tia [Fermilab]↗
DOE OSTI2023

AI-accelerated protein-ligand docking for SARS-CoV-2 is 100-fold faster with no significant change in detection

Protein-ligand docking is a computational method for identifying drug leads. The method is capable of narrowing a vast library of compounds down to a tractable size for downstream simulation or experimental testing and is widely used in drug discovery. While there has been progress in accelerating scoring of compounds with artificial intelligence, few works have bridged these successes back to the virtual screening community in terms of utility and forward-looking development. We demonstrate the power of high-speed ML models by scoring 1 billion molecules in under a day (50 k predictions per GPU seconds). We showcase a workflow for docking utilizing surrogate AI-based models as a pre-filter to a standard docking workflow. Our workflow is ten times faster at screening a library of compounds than the standard technique, with an error rate less than 0.01% of detecting the underlying best scoring 0.1% of compounds. Our analysis of the speedup explains that another order of magnitude speedup must come from model accuracy rather than computing speed. In order to drive another order of magnitude of acceleration, we share a benchmark dataset consisting of 200 million 3D complex structures and 2D structure scores across a consistent set of 13 million “in-stock” molecules over 15 receptors, or binding sites, across the SARS-CoV-2 proteome. We believe this is strong evidence for the community to begin focusing on improving the accuracy of surrogate models to improve the ability to screen massive compound libraries 100 × or even 1000 × faster than current techniques and reduce missing top hits. The technique outlined aims to be a fast drop-in replacement for docking for screening billion-scale molecular libraries.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Accelerating magnonic simulations with the pseudospectral Landau-Lifshitz equation

The pseudospectral Landau-Lifshitz (PS-LL) model can describe atomic-scale magnetic exchange interactions within a continuum framework. This is achieved by employing a convolution kernel that models the nonlocal interaction in a grid-independent manner. Even though the PS-LL was originally introduced to address atomic exchange, any nonlocal kernel can be modeled. In the field of magnonics, the dipole field is fundamental to describe the dispersion relation of magnons, the quasiparticle representation of angular momentum. Because dipole-dipole interactions are long-range, numerical approaches typically rely on convolutions. Here, we demonstrate that the PS-LL model can be used to perform magnonic simulations with a single convolution kernel derived from analytical solutions. We demonstrate a twofold increase in computational speed compared with the full dipole calculation. This approach is valid insofar as the excitations are linear, which is typically the case for magnons. Our results have the potential to accelerate magnonic research, particularly for the inverse design method, where several simulations must be performed to achieve the desired outcome.

Mathematics and computing↗
DOE OSTI2020

Lattice sine-Gordon model

We obtain nonperturbative results on the sine-Gordon model using the lattice field technique. In particular, we employ the Fourier accelerated hybrid Monte Carlo algorithm for our studies. We find the critical temperature of the theory based on autocorrelation time, as well as the finite size scaling of the "thickness" observable used in an earlier lattice study by Hasenbusch et al. We study the entropy, which is smooth across all temperatures, supportive of an infinite order transition. This system has a well-known duality with the massive Thirring model, which can play the role of a toy models for Montonen-Olive duality in $\mathcal {N}$ = 4 super-Yang-Mills theory, since it relates solitons to elementary field excitations. Our research lays a groundwork for such study on the lattice.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Correlated solvent coordinates accelerate multi-donor proton-coupled electron transfer

The rate of charge transfer within a discrete donor–acceptor (D/A) pair is well-described by semi-classical electron transfer theory, but the effects of multiple equivalent redox sites remain less understood. We report a series of ground-state intramolecular proton-coupled electron transfer (PCET) complexes designed to isolate the effects of donor number, N, while holding geometry, coupling, and driving forces constant. The [Ru(L) 3−N (OH) N ] 2+ complexes incorporate one, two, or three identical phenolic electron donors linked to Ru through rigid phenanthroline bridges (OH = 2,4-di-tert-butyl-6-(1-methyl-1H-imidazo[4,5-f][1,10]phenanthrolin-2-yl)phenol). Upon flash photolysis and oxidative quenching with methyl viologen (MV 2+ ), the transient Ru(III) oxidizes an appended phenol by PCET with the hydrogen-bonded imidazole nitrogen atom functioning as the base. The rate increased by 3.4-fold and 5.7-fold (1.7-fold and 1.9-fold after statistical correction) for two- and three-donor complexes compared to the single-donor system. The supra-statistical acceleration is attributed to a reduced effective outer sphere reorganization energy (λ m ) modeled by a partially shared solvent reaction coordinate, in which a subset of solvent dipoles is already oriented to stabilize charge from any donor. The final phenoxyl radical state is localized due to the transfer of a proton, and the recombination reaction with the viologen radical is not accelerated. These results demonstrate the effects of solvent dynamics on intramolecular PCET rates, offering a new strategy for the design of synthetic charge transfer systems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Accelerating defect predictions in semiconductors using graph neural networks

First-principles computations reliably predict the energetics of point defects in semiconductors but are constrained by the expense of using large supercells and advanced levels of theory. Machine learning models trained on computational data, especially ones that sufficiently encode defect coordination environments, can be used to accelerate defect predictions. Here, we develop a framework for the prediction and screening of native defects and functional impurities in a chemical space of group IV, III–V, and II–VI zinc blende semiconductors, powered by crystal Graph-based Neural Networks (GNNs) trained on high-throughput density functional theory (DFT) data. Using an innovative approach of sampling partially optimized defect configurations from DFT calculations, we generate one of the largest computational defect datasets to date, containing many types of vacancies, self-interstitials, anti-site substitutions, impurity interstitials and substitutions, as well as some defect complexes. We applied three types of established GNN techniques, namely crystal graph convolutional neural network, materials graph network, and Atomistic Line Graph Neural Network (ALIGNN), to rigorously train models for predicting defect formation energy (DFE) in multiple charge states and chemical potential conditions. We find that ALIGNN yields the best DFE predictions with root mean square errors around 0.3 eV, which represents a prediction accuracy of 98% given the range of values within the dataset, improving significantly on the state-of-the-art. We further show that GNN-based defective structure optimization can take us close to DFT-optimized geometries at a fraction of the cost of full DFT. The current models are based on the semi-local generalized gradient approximation-Perdew–Burke–Ernzerhof (PBE) functional but are highly promising because of the correlation of computed energetics and defect levels with higher levels of theory and experimental data, the accuracy and necessity of discovering novel metastable and low energy defect structures at the PBE level of theory before advanced methods could be applied, and the ability to train multi-fidelity models in the future with new data from non-local functionals. The DFT-GNN models enable prediction and screening across thousands of hypothetical defects based on both unoptimized and partially optimized defective structures, helping identify electronically active defects in technologically important semiconductors.

Rahman, Md Habibur (ORCID:000000027705984X)↗
DOE OSTI2020

Toward Large-Scale Image Segmentation on Summit

Semantic segmentation of images is an important computer vision task that emerges in a variety of application domains such as medical imaging, robotic vision and autonomous vehicles to name a few. While these domain-specific image analysis tasks involve relatively small image sizes (~ 10 2 × 10 2 ), there are many applications that need to train machine learning models on image data with extents that are orders of magnitude larger (~10 4 × 10 4 ). Training deep neural network (DNN) models on large extent images is extremely memory-intensive and often exceeds the memory limitations of a single graphical processing unit, a hardware accelerator of choice for computer vision workloads. Here, an efficient, sample parallel approach to train U-Net models on large extent image data sets is presented. Its advantages and limitations are analyzed and near-linear strong-scaling speedup demonstrated on 256 nodes (1536 GPUs) of the Summit supercomputer. Using a single node of the Summit supercomputer, an early evaluation of a recently released model parallel framework called GPipe is demonstrated to deliver ~ 2X speedup in executing a U-Net model with an order of magnitude larger number of trainable parameters than reported before. Performance bottlenecks for pipelined training of U-Net models are identified and mitigation strategies to improve the speedups are discussed. Together, these results open up the possibility of combining both approaches into a unified scalable pipelined and data parallel algorithm to efficiently train U-Net models with very large receptive fields on data sets of ultra-large extent images.

Seal, Sudip↗
DOE OSTI2021

Modeling Transverse Space Charge effects in IOTA with pyORBIT

The role and mitigation of space charge effects are important aspects of the beam physics research to be performed at the Integrable Optics Test Accelerator (IOTA) at Fermilab. The impact of nonlinear integrability (partial and complete) on space charge driven incoherent and coherent resonances will be a new feature of this accelerator and has the potential to influence the design of future low energy proton synchrotrons. In this report we model transverse space charge effects using the PIC code pyORBIT. First we benchmark the single particle tracking results against MADX with checks on symplecticity, tune footprints, and dynamic aperture in a partially integrable lattice realized with a special octupole string insert. Our space charge calculations begin with an examination of the 4D symplecticity. Short term tracking is done first with the initial bare lattice and then with a matching of the rms values with space charge. Next, we explore slow initialization of charge as a technique to establish steady state and reduce emittance growth and beam loss following injection into a mismatched lattice. We establish values of space charge simulation parameters so as to ensure numerical convergence. Finally, we compare the simulated space charge induced tune shifts and footprints against theory.

43 PARTICLE ACCELERATORS↗
DOE OSTI2022

Fourier analysis of continuous fractional diffusion synthetic acceleration schemes in slab geometry

We propose two fractional extensions of continuous diffusion synthetic acceleration (DSA) with fractional derivative order α varying over the interval 2 ≥ α ≥ 1 . We investigate the spectral properties of the corresponding continuous families of fractional preconditioners by performing Fourier analysis for a model infinite homogeneous medium problem in slab geometry. The first family results in a fractional acceleration scheme, FrDSAo, that reduces to traditional DSA for .α = 2 and scattering ratio c limiting to a unit value (c → 1) but is otherwise optimized via the Fourier analysis, to obtain the smallest possible spectral radius, for c < 1 and 2 ≥ α ≥ 1. The second family corresponds to a fractional acceleration scheme, FrDSAs, that reduces to traditional DSA for α = 2 for all values of c. The latter scheme is not optimized but has the advantage of lending itself to a more straightforward implementation. For high values of c, the results of the Fourier analysis point to the existence of an interval 2 > α > ∼1.8 where both FrDSAo and FrDSAs can achieve a lower spectral radius than DSA. For example, DSA has a spectral radius of ∼0.2246 for c = 0.9999 while FrDSAo produces a value of ∼0.1616 at α = 1.92 and FrDSAs results in ∼0.2116 at α =1.93. (author)

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Navigating the Path to Autonomy: Real-World Lessons from an Air-Free Self-Driving Laboratory

While autonomous experimentation has promise to accelerate discovery in physcial sciences, the real-world integration of predictive models and experimentation is non-trivial. Here we describe the genesis of a self-driving laboratory (SDL) for air-sensitive chemistry at Argonne National Laboratory and demonstrate the experimental design considerations needed for high-throughput experiments before predictive models can lead to scientific discovery. Our SDL was designed to explore battery electrolyte stability. Our final SDL utilized plate readers in a glovebox with a nitrogen atmosphere to perform kinetic assays and screen hundreds of battery-relevant solvents. However, the roadmap to autonomy and airfree-friendly experimentation required the complex evaluation of several spectroscopic and chromatographic methods. The greatest experimental challenges were (a) developing long-term sampling methods that remained air-free; (b) accelerating kinetics to advance reactivity projections; and (c) ensuring labware compatibility with nonaqueous solvents used in battery chemistry. Our experiences highlight the practical gap between closed-loop aspirations and the realities of chemical discovery, offering lessons on the challenges of transferring every day laboratory workflows to autonomy. These results suggest a more realistic blueprint for autonomy in chemistry—one that balances thoughtful and realistic experimental formulation.

Robertson, Lily A.↗
DOE OSTI2024

Photonuclear Physics in SCALE

Radiation shielding modeling involves a wide range of applications, such as power reactors, critical experiments, portable sources, linear accelerators, and fusion systems. Particle energy ranges vary from thermal energies to fast energy spectra. MAVRIC, as the radiation shielding sequence of the SCALE code system [1] developed and maintained by Oak Ridge National Laboratory, is capable of modeling both neutron and photon interactions in matter and is widely used by the US Nuclear Regulatory Commission, Department of Energy, the National Nuclear Security Administration, and radiation shielding practitioners all around the world. Although most applications don’t involve highly energetic photons that can cause photonuclear reactions and create coupled chains of high energy, ionized electrons and bremsstrahlung radiation within a material medium, when present, these reactions can have a significant impact upon the downstream particle fluxes from incident photons above 20 MeV. For example, linear accelerators used for medical isotope production are examples of applications in which high-energetic photons are generated and used for driving or controlling other components of a complex facility. Radiation shielding analyses of such a facility would require accounting for any and all significant sources of radiation. Unfortunately, MAVRIC included no data and methods to model either photonuclear reactions or the bremsstrahlung photons until this work, which enabled accounting for both. Photonuclear reaction cross sections were imported from the Evaluated Nuclear Data File libraries, and corresponding methods were implemented in the SCALE code system; see Section 2 for how the nuclear data are imported and used. Bremsstrahlung photons were modeled by assuming thick-target approximation whereby all generated electrons and positrons due to photon interactions are assumed to be absorbed within the same material in which they were born, yielding a quick particle transport simulation method by using fast lookup tables for generating subsequent gammas; see Section 4 for details. A suite of verification models was used to test implemented methods and associated data. Tests yielded satisfactory results for the new features. Enhancements will be released as part of the next major SCALE release (v7.0), which is anticipated to be deployed in 2025, with beta releases available in the interim.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2023

Rapid discovery of high hardness multi-principal-element alloys using a generative adversarial network model

Multi-principal element alloys (MPEAs) continue to gain research prominence due to their promising high-temperature microstructural and mechanical properties. Recently, machine learning (ML) and materials informatics have been used extensively for screening MPEAs, however, most of these efforts were focused on constructing classification and regression models for predicting phase stability and mechanical properties of known compositions. These approaches may accelerate the screening process but optimizing new compositions with desirable properties within a practical time frame from an infinitely large design space of MPEA systems remains a grand challenge. To tackle this composition optimization challenge, a generative adversarial network coupled with a neural-network ML model was utilized to design MPEAs by filtering compositions that have high hardness. Even in a high-dimensional space with 18 elements as descriptors, the ML model was able to generate optimized compositions from which one composition was found to have 10% higher hardness (941 HV) than the maximum in the training data (857 HV). Density-functional theory was used to provide thermodynamic and electronic insights to higher hardness of the new MPEA found. The present work can optimize compositions from a wide design space of 18 elements (including W, Ta and Nb) that presents an opportunity to synthesize new compositions for applications ranging from corrosion-resistant alloys to nuclear materials. Here the findings suggest that generative ML can greatly accelerate materials discovery by identifying novel compositions, which can serve as a data-informed tool to guide experiments.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Closure models for the feedback of energetic particles on plasma turbulence

Energetic particles interact with the plasma surrounding them, resonating with certain types of plasma waves to stabilize them while destabilizing others, and changing the character of the background turbulence in ways that have not been fully quantified or understood. Interaction with the turbulent background plasma is key to the acceleration of many types of energetic particles including high-energy cosmic rays, solar energetic particles, and pick-up ions. The acceleration of particles is a process that would ideally be described by a kinetic model, a type of model that follows a probability distribution function (PDF) for all particles in 7-dimensional (x, y, z, v x , v y , v z , t) space. Because of the high dimensionality of a kinetic model, simulations that solve kinetic equations use the largest computational resources currently available, and are yet unable to simulate a realistic number of particles, reach the large scales necessary for astrophysical problems, and use high-precision numerical methods. Two available alternatives to kinetic plasma models have been explored for this problem, with limited success. One is a multi-fluid model produced by a cumulant discarding closure, which evolves coupled equations for the velocity, magnetic field, and internal energy for both the background plasma and the fluid of energetic particles. However, simulations that solve multi-fluid magnetohydrodynamic (MHD) equations are able to include the interaction with energetic particles only in crude ways, typically as an add-on pressure term. The second alternative is to use a hybrid method to couple a fluid description of the background plasma to a kinetic model or a Fokker–Planck model for the energetic particles. These methods are hampered by the physical modeling of the coupling. In this work, we develop a new model, which follows the PDF for all particles; this can be viewed as a step toward physical realism above a multi-fluid MHD model, while also being more computationally efficient than a kinetic model. The equations we develop model both the background plasma and the energetic particles self-consistently. Over the last decade, similar PDF methods have been developed to a high level of sophistication to model reactive flows and turbulent combustion for engineering applications. For treatment of the feedback of the energetic particles on a background plasma, a PDF closure approach should evaluate the mean characteristics, including the density, with better statistical quality than will particle-sampling procedures.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

The Determinants of Offshore Wind's Role in a Future U.S. Energy System: A Preliminary Modeling Sensitivity Analysis [Slides]

Offshore wind is an emerging industry in the with rapidly changing technology advancements and an accelerating global deployment. Policy demands and interest in the United States has grown, but the long-term future for offshore wind remains uncertain. This study is designed to evaluate offshore wind's potential role in the future U.S. energy system. The analysis uses long-term power system models to assess a wide-range of power system possibilities. Specific objectives of the analysis are to: (1) identify the determinants of offshore wind deployment under different scenarios and (2) understand the impacts of substantial levels of offshore wind deployment on the power system. We also highlight future research needs for more robust understanding the opportunities and challenges to U.S. offshore wind development.

17 WIND ENERGY↗