arXiv subjects

Jonathan L. Belof

Publications and source records attributed to Jonathan L. Belof.

At least 19 recordsLinked to original sources

arXiv2026

Multi-Agent Collaboration for Automated Design Exploration on High Performance Computing Systems

Today's scientific challenges, from climate modeling to Inertial Confinement Fusion design to novel material design, require exploring huge design spaces. In order to enable high-impact scientific discovery, we need to scale up our ability to test hypotheses, generate results, and learn from them rapidly. We present MADA (Multi-Agent Design Assistant), a Large Language Model (LLM) powered multi-agent framework that coordinates specialized agents for complex design workflows. A Job Management Agent (JMA) launches and manages ensemble simulations on HPC systems, a Geometry Agent (GA) generates meshes, and an Inverse Design Agent (IDA) proposes new designs informed by simulation outcomes. While general purpose, we focus development and validation on Richtmyer--Meshkov Instability (RMI) suppression, a critical challenge in Inertial Confinement Fusion. We evaluate on two complementary settings: running a hydrodynamics simulations on HPC systems, and using a pre-trained machine learning surrogate for rapid design exploration. Our results demonstrate that the MADA system successfully executes iterative design refinement, automatically improving designs toward optimal RMI suppression with minimal manual intervention. Our framework reduces cumbersome manual workflow setup, and enables automated design exploration at scale. More broadly, it demonstrates a reusable pattern for coupling reasoning, simulation, specialized tools, and coordinated workflows to accelerate scientific discovery.

cs.AI↗

arXiv2026

Spatio-temporal, multi-field deep learning of shock propagation in meso-structured media

Predicting the extreme hydrodynamic response of porous and architected lattice materials is a fundamental challenge in high energy density physics, where shock-induced pore collapse, baroclinic vorticity, and anomalous kinetic and thermodynamic states must be resolved across multiple scales. Traditional high-fidelity hydrocodes are computationally prohibitive for large-scale design exploration in applications like planetary defense and inertial confinement fusion. We present a multi-field spatio-temporal model (MSTM) designed to overcome the limitations of standard machine learning surrogates, which often fail to capture the sharp gradients and non-linear field couplings characteristic of shock propagation. By training on high-fidelity, multiscale multiphysics data, MSTM simultaneously evolves seven coupled thermodynamic and kinetic fields - including pressure, temperature, density, and velocity - across complex material architectures. Our framework demonstrates the ability to accurately predict anomalous responses, such as counterintuitive post-shock density reductions and localized hotspot formation, with mean root mean squared errors as low as 1.4%. Crucially, the model's multi-field formulation maintains physical consistency and interface stability over long autoregressive rollouts, outperforming single-field models by 94% in structural fidelity. This framework enables a 1000x reduction in time to solution, providing a practical pathway for the real-time analysis and optimization of energy dissipation and momentum transfer in meso-structured media.

cs.LG↗

arXiv2025

Multi-Agent Design Assistant for the Simulation of Inertial Fusion Energy

Inertial fusion energy promises nearly unlimited, clean power if it can be achieved. However, the design and engineering of fusion systems requires controlling and manipulating matter at extreme energies and timescales; the shock physics and radiation transport governing the physical behavior under these conditions are complex requiring the development, calibration, and use of predictive multiphysics codes to navigate the highly nonlinear and multi-faceted design landscape. We hypothesize that artificial intelligence reasoning models can be combined with physics codes and emulators to autonomously design fusion fuel capsules. In this article, we construct a multi-agent system where natural language is utilized to explore the complex physics regimes around fusion energy. The agentic system is capable of executing a high-order multiphysics inertial fusion computational code. We demonstrate the capacity of the multi-agent design assistant to both collaboratively and autonomously manipulate, navigate, and optimize capsule geometry while accounting for high fidelity physics that ultimately achieve simulated ignition via inverse design.

physics.app-ph↗

arXiv2025

Advancing Material Modeling in Hydrocodes Beyond Equations of State

We present a multiscale simulation framework that couples the Finite Element Method with molecular dynamics. Bypassing traditional equations of state (EOS) by using in-line atomistic simulations, the method offers the advantage of incorporating detailed microscale physics not easily represented with coarse-grained models. Coupling consistency with the continuum code is ensured through the use of lifting and restriction operators, in line with heterogeneous multiscale methods. The concurrent continuum-atomistic framework is validated through comparison with experimental results and conventional EOS models, and demonstrated in a shock-driven hydrodynamic flow simulation under extreme conditions. We further evaluate the framework's usability by comparing it to state-of-the-art EOS models of deuterium. A computational performance study reveals that the atomistic EOS evaluation is a feasible alternative to conventional approaches, and demonstrates a weak scaling of 99% efficiency. These results highlight the framework's potential for large-scale multiscale modeling across a broad range of materials and conditions.

physics.comp-ph↗

arXiv2025

Local Reduced-Order Modeling for Electrostatic Plasmas by Physics-Informed Solution Manifold Decomposition

Despite advancements in high-performance computing and modern numerical algorithms, computational cost remains prohibitive for multi-query kinetic plasma simulations. In this work, we develop data-driven reduced-order models (ROMs) for collisionless electrostatic plasma dynamics, based on the kinetic Vlasov-Poisson equation. Our ROM approach projects the equation onto a linear subspace defined by the proper orthogonal decomposition (POD) modes. We introduce an efficient tensorial method to update the nonlinear term using a precomputed third-order tensor. We capture multiscale behavior with a minimal number of POD modes by decomposing the solution manifold into multiple time windows and creating temporally local ROMs. We consider two strategies for decomposition: one based on the physical time and the other based on the electric field energy. Applied to the 1D1V Vlasov-Poisson simulations, that is, prescribed E-field, Landau damping, and two-stream instability, we demonstrate that our ROMs accurately capture the total energy of the system both for parametric and time extrapolation cases. The temporally local ROMs are more efficient and accurate than the single ROM. In addition, in the two-stream instability case, we show that the energy-windowing reduced-order model (EW-ROM) is more efficient and accurate than the time-windowing reduced-order model (TW-ROM). With the tensorial approach, EW-ROM solves the equation approximately 90 times faster than Eulerian simulations while maintaining a maximum relative error of 7.5% for the training data and 11% for the testing data.

math.NA↗

arXiv2024

GPLaSDI: Gaussian Process-based Interpretable Latent Space Dynamics Identification through Deep Autoencoder

Numerically solving partial differential equations (PDEs) can be challenging and computationally expensive. This has led to the development of reduced-order models (ROMs) that are accurate but faster than full order models (FOMs). Recently, machine learning advances have enabled the creation of non-linear projection methods, such as Latent Space Dynamics Identification (LaSDI). LaSDI maps full-order PDE solutions to a latent space using autoencoders and learns the system of ODEs governing the latent space dynamics. By interpolating and solving the ODE system in the reduced latent space, fast and accurate ROM predictions can be made by feeding the predicted latent space dynamics into the decoder. In this paper, we introduce GPLaSDI, a novel LaSDI-based framework that relies on Gaussian process (GP) for latent space ODE interpolations. Using GPs offers two significant advantages. First, it enables the quantification of uncertainty over the ROM predictions. Second, leveraging this prediction uncertainty allows for efficient adaptive training through a greedy selection of additional training data points. This approach does not require prior knowledge of the underlying PDEs. Consequently, GPLaSDI is inherently non-intrusive and can be applied to problems without a known PDE or its residual. We demonstrate the effectiveness of our approach on the Burgers equation, Vlasov equation for plasma physics, and a rising thermal bubble problem. Our proposed method achieves between 200 and 100,000 times speed-up, with up to 7% relative error.

cs.CE↗

arXiv2024

Explosively driven Richtmyer--Meshkov instability jet suppression and enhancement via coupling machine learning and additive manufacturing

The ability to control the behavior of fluid instabilities at material interfaces, such as the shock-driven Richtmyer--Meshkov instability, is a grand technological challenge with a broad number of applications ranging from inertial confinement fusion experiments to explosively driven shaped charges. In this work, we use a linear-geometry shaped charge as a means of studying methods for controlling material jetting that results from the Richtmyer--Meshkov instability. A shaped charge produces a high-velocity jet by focusing the energy from the detonation of high explosives. The interaction of the resulting detonation wave with a hollowed cavity lined with a thin metal layer produces the unstable jetting effect. By modifying characteristics of the detonation wave prior to striking the lined cavity, the kinetic energy of the jet can be enhanced or reduced. Modifying the geometry of the liner material can also be used to alter jetting properties. We apply optimization methods to investigate several design parameterizations for both enhancing or suppressing the shaped-charge jet. This is accomplished using 2D and 3D hydrodynamic simulations to investigate the design space that we consider. We also apply new additive manufacturing methods for producing the shaped-charge assemblies, which allow for experimental testing of complicated design geometries obtained through computational optimization. We present a direct comparison of our optimized designs with experimental results carried out at the High Explosives Application Facility at Lawrence Livermore National Laboratory.

physics.app-ph↗

arXiv2024

Modification of Jet Velocities in an Explosively Loaded Copper Target with a Conical Defect

In this work, the design and execution of an experiment with the goal of demonstrating control over the evolution of a copper jet is described. Simulations show that when using simple multi-material buffers placed between a copper target with a conical defect and a cylinder of high-explosive, a variety of jetting behaviors occur based on material placement, including both jet velocity augmentation and mitigation. A parameter sweep was performed to determine optimal buffer designs in two configurations. Experiments using the optimal buffer designs verified the effectiveness of the buffer and validated the modeling.

physics.app-ph↗

arXiv2023

Data-Driven Autoencoder Numerical Solver with Uncertainty Quantification for Fast Physical Simulations

Traditional partial differential equation (PDE) solvers can be computationally expensive, which motivates the development of faster methods, such as reduced-order-models (ROMs). We present GPLaSDI, a hybrid deep-learning and Bayesian ROM. GPLaSDI trains an autoencoder on full-order-model (FOM) data and simultaneously learns simpler equations governing the latent space. These equations are interpolated with Gaussian Processes, allowing for uncertainty quantification and active learning, even with limited access to the FOM solver. Our framework is able to achieve up to 100,000 times speed-up and less than 7% relative error on fluid mechanics problems.

cs.CE↗

arXiv2023

Reducing Richtmyer Meshkov Instability Jet Velocity via Inverse Design

In this work, we detail a novel application of inverse design and advanced manufacturing to rapidly develop and experimentally validate modifications to a shaped charge jet analogue. The shaped charge jet analogue comprises a conical copper liner, high explosive (HE), and silicone buffer. We apply a genetic algorithm to determine an optimal buffer design that can be placed between the liner and the HE that results in the largest possible change in jet velocity. The use of a genetic algorithm allows for discoveries of unintuitive, complex, yet optimal buffer designs. Experiments using the optimal design verified the effectiveness of the buffer and validated the modeling.

physics.app-ph↗

arXiv2022

Certified data-driven physics-informed greedy auto-encoder simulator

A parametric adaptive greedy Latent Space Dynamics Identification (gLaSDI) framework is developed for accurate, efficient, and certified data-driven physics-informed greedy auto-encoder simulators of high-dimensional nonlinear dynamical systems. In the proposed framework, an auto-encoder and dynamics identification models are trained interactively to discover intrinsic and simple latent-space dynamics. To effectively explore the parameter space for optimal model performance, an adaptive greedy sampling algorithm integrated with a physics-informed error indicator is introduced to search for optimal training samples on the fly, outperforming the conventional predefined uniform sampling. Further, an efficient k-nearest neighbor convex interpolation scheme is employed to exploit local latent-space dynamics for improved predictability. Numerical results demonstrate that the proposed method achieves 121 to 2,658x speed-up with 1 to 5% relative errors for radial advection and 2D Burgers dynamical problems.

cs.LG↗

arXiv2022

Using Conservation Laws to Infer Deep Learning Model Accuracy of Richtmyer-meshkov Instabilities

Richtmyer-Meshkov Instability (RMI) is a complicated phenomenon that occurs when a shockwave passes through a perturbed interface. Over a thousand hydrodynamic simulations were performed to study the formation of RMI for a parameterized high velocity impact. Deep learning was used to learn the temporal mapping of initial geometric perturbations to the full-field hydrodynamic solutions of density and velocity. The continuity equation was used to include physical information into the loss function, however only resulted in very minor improvements at the cost of additional training complexity. Predictions from the deep learning model appear to accurately capture temporal RMI formations for a variety of geometric conditions within the domain. First principle physical laws were investigated to infer the accuracy of the model's predictive capability. While the continuity equation appeared to show no correlation with the accuracy of the model, conservation of mass and momentum were weakly correlated with accuracy. Since conservation laws can be quickly calculated from the deep learning model, they may be useful in applications where a relative accuracy measure is needed.

physics.flu-dyn↗

arXiv2018

Observation of a metastable intermediate during solid-solid phase transformation in response to rapid compression

In order to probe the mechanism of solid-solid phase transformations, we have applied ultrafast shock wave compression (120 picosecond duration) and ultrashort (130 femtosecond) x-ray diffraction at the Linac Coherent Light Source (LCLS) to probe the compression-induced phase transition pathway in zirconium. Surprisingly, rather than transform from alpha-Zr to the more disordered hex-3 equilibrium omega-Zr phase, in its place we find the formation of a non-equilibrium body-centered cubic (bcc) metastable intermediate. Theoretically hypothesized for several decades, this bcc intermediate state has now been found to be dynamically stabilized under uniaxial loading at sub-nanosecond timescales. Molecular dynamics simulations of shock-wave propagation in zirconium predict this transformation via the dynamical intermediate state. In contrast with longer timescale experiments where the phase diagram alone is an adequate predictor of the crystalline structure of a material, our recent study highlights the importance of metastability and time-dependence in the kinetics of phase transformation at extreme conditions.

cond-mat.mtrl-sci↗

arXiv2018

Anharmonicity-induced isostructural phase transition of Zirconium under pressure

We have performed a detailed x-ray diffraction structural study of Zr under pressure and unambiguously identify the existence of a first-order isostructural bcc-to-bcc phase transition near 58 GPa. First-principles quantum molecular dynamics lattice dynamics calculations support the existence of this phase transition, in excellent agreement with experimental results, triggered by anharmonic effects. Our results highlight the potential ubiquity of anharmonically driven isostructural transitions within the periodic table under pressure and calls for follow-up experimental and theoretical studies.

cond-mat.mtrl-sci↗

arXiv2017

Non-Equilibrium Response of Deuterium-Tritium Fusion Plasma: Molecular Dynamics Study of Ion Tail Replenishment

It has recently been proposed that transport of energetic deuterium-trition ions (Knudsen layer), owing to their long mean free path, may deplete the tail of the velocity distribution in the vicinity of the Gamow peak and thus serve to greatly reduce the yield of an inertial confinement fusion (ICF) plasma [Molvig \emph{et al.}, \emph{Phys. Rev. Lett.}, 109:095001 (2012)]. Molecular dynamics simulations of the non-equilibrium response of a fusion plasma to ion distribution tail loss show that equilibrium is re-established extremely rapidly, with ion tail replenishment complete in less than 1 ps for $ρr > 0.01$. The exceedingly fast relaxation of the fusion plasma dynamics, relative to the Fokker-Planck collisional operator, can be understood in terms of a large entropic driving force that pushes the phase space distribution toward a Maxwellian, and is a characteristic of the Lyupanov instability for the Coulomb system. The thermal reactivity $\langle σv \rangle$ for the D(t,n)$α$ fusion reaction is calculated as a function of progress toward equilibrium for a series of densities relevant to ICF hot spot ignition.

physics.plasm-ph↗

arXiv2017

Microcanonical Effective Partition Function for the Anharmonic Oscillator

The microcanonical effective partition function, constructed from a Feynman-Hibbs potential, is derived using generalized ensemble theory. The form of the effective Hamiltonian is amenable to Monte Carlo simulation techniques and the relevant Metropolis function is presented. Using the derived expression for the microcanonical effective partition function, the low-temperature entropy of a proton in an anharmonic potential is numerically evaluated and compared with the exact quantum mechanical canonical result.

cond-mat.stat-mech↗

arXiv2014

Volume Determination of Globular Proteins by Molecular Dynamics

Molecular dynamics simulations of myoglobin and aspartate aminotransferase, with explicit solvent, are shown to accurately reproduce the experimentally measured molar volumes. Single amino-acid substitution at VAL39 of aspartate aminotransferase is known to produce large volumetric changes in the enzyme, and this effect is demonstrated in simulation as well. This molecular dynamics approach, while more computationally expensive that extant computational methods of determining the apparent volume of biological systems, is quite feasible with modern computer hardware and is shown to yield accurate volumetric data with as little as several nanoseconds of dynamics.

cond-mat.soft↗

arXiv2013

Alternative Derivation of the Partition Function for Generalized Ensembles

A pedagogical approach for deriving the statistical mechanical partition function, in a manner that emphasizes the key role of entropy in connecting the microscopic states to thermodynamics, is introduced. The connections between the combinatoric formula $S= k \ln W$ applied to the Gibbs construction, the Gibbs entropy, $S = -k \sum\limits_i p_i \ln p_i$, and the microcanonical entropy expression $S= k \ln Ω$ are clarified. The condition for microcanonical equilibrium, and the associated role of the entropy in the thermodynamic potential is shown to arise naturally from the postulate of equal {\itshape a priori} states. The derivation of the canonical partition function follows simply by invoking the Gibbs ensemble construction at constant temperature and using the first and second law of thermodynamics (\emph{via} the fundamental equation $dE = TdS - PdV + μdN$) that incorporate the conditions of conservation of energy and composition without the needs for explicit constraints; other ensemble follow easily. The central role of the entropy in establishing equilibrium for a given ensemble emerges naturally from the current approach. Connections to generalized ensemble theory also arise and are presented in this context.

cond-mat.stat-mech↗