HPC Modeling to Aid Accelerator Component Manufacturing
Explore the source record for details and available documents.
SEARCH · Search NASA
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.
Explore the source record for details and available documents.
Proton accumulator rings and other circular hadron accelerators are susceptible to intensity-driven parametric instabilities because the zero-current charged particle dynamics are characterized by a single tune. Landau damping can suppress these instabilities, which requires energy spread in the beam or introducing nonlinear magnets such as octupoles. However, this approach reduces dynamic aperture. Nonlinear integrable optics can suppress parametric instabilities independent of energy spread in the distribution, while preserving the dynamic aperture. This novel approach promises to reduce particle losses and enable order-of-magnitude increases in beam intensity. In this paper we present results, obtained using the Lie operator formalism, on how chromaticity and dispersion affect particle orbits in integrable optics. We conclude that chromaticity in general breaks the integrability, unless the vertical and horizontal chromaticities are equal. Because of this, the chromaticity correcting magnets can be weaker and fewer correcting magnet families are required, thus minimizing the impact on dynamic aperture.
Explore the source record for details and available documents.
In this article we consider two dynamical systems: the McMillan sextupole and octupole integrable mappings originally introduced by Edwin McMillan; the second one is also known as canonical McMillan map. Both of them are simplest symmetric McMillan maps with only one intrinsic parameter, the trace of the Jacobian at the fixed point. While these dynamical systems have numerous of applications and are used in many areas of math and physics, some of their dynamical properties have not been described yet. We fulfill the gap and provide complete description of all stable trajectories including parametrization of invariant curves, Pioncaré rotation numbers and canonical action-angle variables. In the second part we relate these maps with general chaotic map in McMillan-Turaev form. We show that McMillan sextupole and octupole mappings are first order approximations of dynamics around the fixed point, in a similar way as linear map and quadratic invariant (Courant-Snyder invariant in accelerator physics) is the zeroth order approximation (known as linearization). Finally we suggest the new formalism of nonlinear Twiss parameters which incorporate dependence of rotation number as a function of amplitude, in contrast to e.g. betatron phase advance used in accelerator physics which is independent of amplitude. Specifically in application to accelerator physics this new formalism is capable of predicting dynamical aperture around 1-st, 2-nd, 3-rd and 4-th order resonances for flat beams, which is critical for beam injection/extraction.
The discovery of new catalysts is essential for the design of new and more efficient chemical processes in order to transition to a sustainable future. We introduce an AI-guided computational screening framework unifying linguistic reasoning with quantum-chemistry based feedback from 3D atomistic representations. Our approach formulates catalyst discovery as an uncertain environment where an agent actively searches for highly effective catalysts via the iterative combination of large language model (LLM)-derived hypotheses and atomistic graph neural network (GNN)-derived feedback. Identified catalysts in intermediate search steps undergo structural evaluation based on spatial orientation, reaction pathways, and stability. Scoring functions based on adsorption energies and barriers steer the exploration in the LLM's knowledge space toward energetically favorable, high-efficiency catalysts. We introduce planning methods that automatically guide the exploration without human input, providing competitive performance against expert-enumerated chemical descriptor-based implementations. By integrating language-guided reasoning with computational chemistry feedback, our work pioneers AI-accelerated, trustworthy catalyst discovery.
Explore the source record for details and available documents.
We demonstrate an implementation of the Boris Particle Pusher Algorithm using CUDA-enabled processors. A simplified PIC code is used as a reference for the pusher and several tests are shown. The deployment time from a developer point of view is also evaluated. Results show that CUDA is a good tool for significant code acceleration with minimal code impact and minimal effort for the developer.
The γ -ray flares from the Crab Nebula observed by AGILE and Fermi -LAT reaching GeV energies and lasting several days challenge the standard models for particle acceleration in pulsar-wind nebulae because the radiating electrons have energies exceeding the classical radiation-reaction limit for synchrotron. Previous modeling has suggested that the synchrotron limit can be exceeded if the electrons experience electrostatic acceleration, but the resulting spectra do not agree very well with the data. As a result, there are still some important unanswered questions about the detailed particle acceleration and emission processes occurring during the flares. We revisit the problem using a new analytical approach based on an electron transport equation that includes terms describing electrostatic acceleration, stochastic wave-particle acceleration, shock acceleration, synchrotron losses, and particle escape. An exact solution is obtained for the electron distribution, which is used to compute the associated γ -ray synchrotron spectrum. We find that in our model the γ -ray flares are mainly powered by electrostatic acceleration, but the contributions from stochastic and shock acceleration play an important role in producing the observed spectral shapes. Our model can reproduce the spectra of all the Fermi -LAT and AGILE flares from the Crab Nebula, using magnetic field strengths in agreement with the multi-wavelength observational constraints. We also compute the spectrum and duration of the synchrotron afterglow created by the accelerated electrons, after they escape into the region on the downstream side of the pulsar-wind termination shock. The afterglow is expected to fade over a maximum period of about three weeks after the γ -ray flare.
Detector simulation is critical to experimental HEP; however this simulation (commonly done through toolkits such as Geant4) is computationally intensive. Performance can be improved somewhat through technical optimization, but more is needed. Using machine learning (ML) to accelerate simulation is a promising field, however efforts to use generative adversarial networks (GANs) or optimized autoencoders have faced issues. Using convolutional neural networks (CNNs) for denoising has been successful in non-HEP applications such as image processing. This poster investigates the efficacy of using CNNs to denoise Geant4 simulations. This could increase the accuracy of simulations performed under settings designed to increase computational efficiency.
The physical and chemical phenomena occurring in a battery are many and complex and in many different scales. Without a better knowledge of the interplay among the multi-physics occurring across the varied scales, it is very challenging and time consuming to design long-lasting, high-performing, safe, affordable large battery systems, enabling electrification of the vehicles and modernization of the grid. The National Renewable Energy Laboratory, a U.S. Department of Energy laboratory, has been developing thermal and electrochemical models for cells and battery packs. Working with software producers, carmakers, and battery developers, computer-aided engineering tools have been developed that can accelerate the electrochemical and thermal design of batteries, reducing time to develop and optimize them and thus reducing the cost of the system. In the past couple of years, we initiated a project to model the mechanical response of batteries to stress, strain, fracture, deformation, puncture, and crush and then link them to electrochemical and thermal models to predict the response of a battery. This modeling is particularly important for understanding the physics and processes that happen in a battery during a crush-inducing vehicle crash. In this paper, we provide an overview of electrochemical-thermal-mechanical models for battery system understanding and designing.
Scientific productivity can be enhanced through workflow management tools, relieving large High Performance Computing (HPC) system users from the tedious tasks of scheduling and designing the complex computational execution of scientific applications. This paper presents a study on the usage of ensemble workflow tools to accelerate science using the Summit and Frontier supercomputing systems. The research aims to connect science domain simulations using Oak Ridge Leadership Computing Facility (OLCF) supercomputing platforms with ensemble workflow methods in order to accelerate HPC-enabled discovery and boost scientific impact. We present the coupling, porting and optimization of Radical-Cybertools on three applications: Chroma, NAMD and LAMMPS. The tools augment traditional HPC monolithic runs with a pilot scheduler. Lessons-learned are discussed for physics, biology and materials science applications. We discuss intrinsic limitations of coupling and porting ensemble workflow tools to applications that run on large HPC systems. The origins of technical challenges and their solutions developed during the implementation process are discussed. Data management strategies, OLCF’s policies for ensembles, and natively supported workflow tools are also summarized.
The E166 experiment at the Stanford Linear Accelerator Center (SLAC) has demonstrated a scheme for the production of polarized positrons which is suitable for implementation in a future Linear Collider. A multi-GeV electron beam passed through a helical undulator to generate multi-MeV, circularly polarized photons which were then converted in a thin target to produce positrons (and electrons) with longitudinal polarization above 80% at 6 MeV. The results are in agreement with GEANT4 simulations that include the dominant polarization-dependent interactions of electrons, positrons and photons in matter.
In the near future, the accelerator and experimental facilities at GANIL (Caen/France) will be extended by a new facility called SPIRAL2. It will be based on a superconducting linear accelerator which will deliver a 40-MeV deuteron beam (5 mA; 200 kW). These accelerated particles will produce a very high neutron flux with a fast spectrum (most probable neutron energy is about 14 MeV with an extension to 40 MeV). With the neutron-induced fission of the natural uranium, intense beams of neutron-rich nuclei will be created and will become available at SPIRAL2. In this work, done with the support of SAFERIB and EURONS, we present the results of the modelling of the uranium carbide target. The modelling was mainly done with the code MCNPX26B. The target is theoretically optimised to produce a global fission rate greater than 5x10{sup 13} fissions.s{sup -1}. In fact, the optimised target, of 2.3 kg of carbide uranium, has a calculated fission rate closer to about 6x10{sup 13} fissions.s{sup -1}.
In this study, a computer-based system for routine quality assurance (QA) of a linear accelerator (linac) was developed by using the dosimetric properties of an amorphous silicon electronic portal imaging device (EPID). An acrylic template phantom was designed such that it could be placed on the EPID and be aligned with the light field of the collimator. After irradiation, portal images obtained from the EPID were transferred in DICOM format to a computer and analyzed using a program we developed. The symmetry, flatness, field size, and congruence of the light and radiation fields of the photon beams from the linac were verified simultaneously. To validate the QA system, the ion chamber and film (X-Omat V2; Kodak, New York, NY) measurements were compared with the EPID measurements obtained in this study. The EPID measurements agreed with the film measurements. Parameters for beams with energies of 6 MV and 15 MV were obtained daily for 1 month using this system. It was found that our QA tool using EPID could substitute for the film test, which is a time-consuming method for routine QA assessment.
Purpose: To create an accurate 6 MV Monte Carlo simulation phase space for the Varian TrueBeam treatment head geometry imported from cad (computer aided design) without adjusting the input electron phase space parameters. Methods: geant4 v4.9.2.p01 was employed to simulate the 6 MV beam treatment head geometry of the Varian TrueBeam linac. The electron tracks in the linear accelerator were simulated with Parmela, and the obtained electron phase space was used as an input to the Monte Carlo beam transport and dose calculations. The geometry components are tessellated solids included in geant4 as gdml (generalized dynamic markup language) files obtained via STEP (standard for the exchange of product) export from Pro/Engineering, followed by STEP import in Fastrad, a STEP-gdml converter. The linac has a compact treatment head and the small space between the shielding collimator and the divergent arc of the upper jaws forbids the implementation of a plane for storing the phase space. Instead, an IAEA (International Atomic Energy Agency) compliant phase space writer was implemented on a cylindrical surface. The simulation was run in parallel on a 1200 node Linux cluster. The 6 MV dose calculations were performed for field sizes varying from 4 x 4 to 40 x 40 cm{sup 2}. The voxel size for the 60x60x40 cm{sup 3} water phantom was 4x4x4 mm{sup 3}. For the 10x10 cm{sup 2} field, surface buildup calculations were performed using 4x4x2 mm{sup 3} voxels within 20 mm of the surface. Results: For the depth dose curves, 98% of the calculated data points agree within 2% with the experimental measurements for depths between 2 and 40 cm. For depths between 5 and 30 cm, agreement within 1% is obtained for 99% (4x4), 95% (10x10), 94% (20x20 and 30x30), and 89% (40x40) of the data points, respectively. In the buildup region, the agreement is within 2%, except at 1 mm depth where the deviation is 5% for the 10x10 cm{sup 2} open field. For the lateral dose profiles, within the field size for fields up to 30x30 cm{sup 2}, the agreement is within 2% for depths up to 10 cm. At 20 cm depth, the in-field maximum dose difference for the 30x30 cm{sup 2} open field is within 4%, while the smaller field sizes agree within 2%. Outside the field size, agreement within 1% of the maximum dose difference is obtained for all fields. The calculated output factors varied from 0.938{+-}0.015 for the 4x4 cm{sup 2} field to 1.088{+-}0.024 for the 40x40 cm{sup 2} field. Their agreement with the experimental output factors is within 1%. Conclusions: The authors have validated a geant4 simulated IAEA-compliant phase space of the TrueBeam linac for the 6 MV beam obtained using a high accuracy geometry implementation from cad. These files are publicly available and can be used for further research.
In the rapidly evolving field of high-performance computing (HPC), effectively leveraging heterogeneous devices through asynchronous task programming is paramount. This paper presents a robust asynchronous task programming model tailored for a multi-device, multi-stream execution environment that incorporates a diverse array of heterogeneous computing units, including GPUs from various vendors and other accelerators. Current state-of-the-art task programming models provide methodologies to support asynchronous task executions, but they typically handle homogeneous devices using native programming languages, while support for heterogeneous devices is limited to frameworks like OpenCL. This gap presents significant challenges in abstracting heterogeneous devices to harness their true asynchronous capabilities effectively using their native programming languages. By implementing asynchronous task execution, our model significantly boosts the performance of tiled algorithm task graphs through overlapping data transfers with computation and enabling the simultaneous execution of multiple kernels. We integrate this approach into a heterogeneous Intelligent Runtime System (IRIS) and assess its performance using a suite of tiled algorithm benchmarks from the heterogeneous math kernels library (MatRIS) based on IRIS. Experimental results demonstrate a performance improvement ranging from 1.6 × to 2 × over IRIS without asynchronous support, and a notable 22% performance enhancement compared to established runtime systems such as StarPU and PaRSEC. This approach significantly improves computation efficiency of HPC workflows and provides a solid base for future exploration and development in the area of asynchronous task programming in heterogeneous systems.
Supercomputing is evolving towards hybrid and accelerator-based architectures with millions of cores. The HACC (Hardware/Hybrid Accelerated Cosmology Code) framework exploits this diverse landscape at the largest scales of problem size, obtaining high scalability and sustained performance. Developed to satisfy the science requirements of cosmological surveys, HACC melds particle and grid methods using a novel algorithmic structure that flexibly maps across architectures, including CPU/GPU, multi/many-core, and Blue Gene systems. We demonstrate the success of HACC on two very different machines, the CPU/GPU system Titan and the BG/Q systems Sequoia and Mira, attaining unprecedented levels of scalable performance. We demonstrate strong and weak scaling on Titan, obtaining up to 99.2% parallel efficiency, evolving 1.1 trillion particles. On Sequoia, we reach 13.94 PFlops (69.2% of peak) and 90% parallel efficiency on 1,572,864 cores, with 3.6 trillion particles, the largest cosmological benchmark yet performed. HACC design concepts are applicable to several other supercomputer applications.
In this work, we present JACC.shared, a new feature of Julia for ACCelerators (JACC), which is the performanceportable and metaprogramming model of the just-in-time and LLVM-based Julia language. This new feature allows JACC applications to leverage the high-performance computing (HPC) capabilities of high-bandwidth, on-chip GPU memory. Historically, exploiting high-bandwidth, shared-memory GPUs has not been a priority for high-level programming solutions. JACC.shared covers that gap for the first time, thereby providing a highlevel, portable, and easy-to-use solution for programmers to exploit this memory and supporting all current major accelerator architectures. Well-known HPC and AI workloads, such as multi/hyperspectral imaging and AI convolutions, have been used to evaluate JACC.shared on two exascale GPU architectures hosted by some of the most powerful US Department of Energy supercomputers: Perlmutter (NVIDIA A100) and Frontier (AMD MI250X). The performance evaluation reports speedup of up to 3.5× by adding only one line of code to the base codes, thus providing important accelerators in a simple, portable, and transparent way and elevating the programming productivity and performance-portability capabilities for Julia/JACC HPC, AI, and scientific applications.