arXiv subjects

Timothy C. Germann

Publications and source records attributed to Timothy C. Germann.

18 recordsLinked to original sources

arXiv2026

Stoichiometry Dependent Properties of Cerium Hydride: An Active Learning Developed Interatomic Potential Study

Cerium hydride has a variety of interesting properties, including a known lattice contraction and densification with increasing hydrogen content. However, precise stoichiometric control is not experimentally straightforward and {\it ab initio} approaches are not computationally feasible for many properties such as melting and low temperature diffusion. Therefore, we develop a machine-learned interatomic potential for cerium hydride that is valid for H to Ce ratios from 2.0 to 3.0. A query-by-committee active learning approach is used to develop the training set. Leveraging classical molecular dynamics simulations, we assess a range of properties and provide fundamental mechanisms for the trends with stoichiometry. A majority of the properties follow the trend of lattice contraction, being governed by the stronger lattice binding induced by adding octahedral atoms.

cond-mat.mtrl-sci↗

arXiv2025

Simulating nationwide coupled disease and fear spread in an agent-based model

Human cognitive responses, behavioral responses, and disease dynamics co-evolve over the course of any disease outbreak, and can result in complex feedbacks. We present a dynamic agent-based model that explicitly couples the spread of disease with the spread of fear surrounding the disease, implemented within the EpiCast simulation framework. EpiCast models transmission across a realistic synthetic population, capturing individual-level interactions. In our model, fear propagates through both in-person contact and broadcast media, prompting individuals to adopt protective behaviors that reduce disease spread. In order to better understand these coupled dynamics, we create and compare a range of compartmental surrogate models to analyze the impact of including various disease states. Additionally, we compare a range of behavioral scenarios within EpiCast, varying the level and intensity of fear and behavioral change. Our results show that the addition of asymptomatic, exposed, and pre-symptomatic disease states can impact both the rate at which an outbreak progresses and its overall trajectory. Moreover, the combination of non-local fear spread via broadcasters and strong behavioral responses by fearful individuals generally leads to multiple epidemic waves, an outcome that occurs only within a narrow parameter range when fear spreads purely through local contact. Accounting for the coupled spread of fear and disease is critical for understanding disease dynamics and designing timely, targeted responses to emerging infectious threats.

q-bio.PE↗

arXiv2025

Epicast 2.0: A large-scale, demographically detailed, agent-based model for simulating respiratory pathogen spread in the United States

The recent history of respiratory pathogen epidemics, including those caused by influenza and SARS-CoV-2, has highlighted the urgent need for advanced modeling approaches that can accurately capture heterogeneous disease dynamics and outcomes at the national scale, thereby enhancing the effectiveness of resource allocation and decision-making. In this paper, we describe Epicast 2.0, an agent-based model that utilizes a highly detailed, synthetic population and high-performance computing techniques to simulate respiratory pathogen transmission across the entire United States. This model replicates the contact patterns of over 320 million agents as they engage in daily activities at school, work, and within their communities. Epicast 2.0 supports vaccination and an array of non-pharmaceutical interventions that can be promoted or relaxed via highly granular, user specified policies. We illustrate the model's capabilities using a wide range of outbreak scenarios, highlighting the model's varied dynamics as well as its extensive support for policy exploration. This model provides a robust platform for conducting what if scenario analysis and providing insights into potential strategies for mitigating the impacts of infectious diseases.

q-bio.PE↗

arXiv2024

High Pressure Suppression of Plasticity due to Over-Nucleation of Shear Strain

High pressure shear band formation is a critical phenomenon in energetic materials due to its influence on both mechanical strength and mechanochemical activation. While shear banding is know to occur in a variety of these materials, the governing dynamics of the mechanisms is not well defined for molecular crystals. We conduct molecular dynamics simulations of shock wave induced shear band formation in the energetic material 1,3,5-trinitroperhydro-1,3,5-triazine (RDX) to assess shear band nucleation processes. We find, that at high pressures, the initial formation sites for shear bands 'over-nucleate' and rapidly lower deviatoric stresses prior to shear band formation and growth. This results in the suppression of plastic deformation. A local cluster analysis is used to quantify and contrast this mechanism with a more typical shear banding seen at lower pressures. These results demonstrate a mechanism that is reversible in nature and that supersedes shear band formation at increased pressures. We anticipate that these results will have a broad impact on the modeling and development of high strain rate application materials such as those for high explosives and hypersonic systems.

cond-mat.mtrl-sci↗

arXiv2023

Using Limited Neural Networks to Assess Relative Mechanistic Influence on Shock Heating in Granular Solids

The rapid compaction of granular media results in localized heating that can induce chemical reactions, phase transformations, and melting. However, there are numerous mechanisms in play that can be dependent on a variety of microstructural features. Machine learning techniques such as neural networks offer a ubiquitous method to develop models for physical processes. Limiting what kinds of microstructural information is used as input and assessing normalized changes in network error, the relative importance of different mechanisms can be inferred. Here we utilize binned, initial density information as network inputs to predict local shock heating in a granular high explosive trained from large scale, molecular dynamics simulations. The spatial extend of the density field used in the network is altered to assess the importance and relevant length scales of the physical mechanisms in play, where different microstructural features result in different predictive capability.

cond-mat.soft↗

arXiv2023

Interplay of Mechanochemistry and Material Processes in the Graphite to Diamond Phase Transformation

he manifestation of intra-molecular strains in covalent systems is widely known to accelerate chemical reactions and open alternative reaction paths. This process is moderately understood for isolated molecules and uni-molecular processes. However, in condensed matter processes such as phase transformations, material properties and structure may influence typical mechanochemical effects. Therefore, we utilize steered molecular dynamics to induce out of plane strains in graphite and compress the system under a constant strain rate to induce phase transformation. We show that the out of plane strain allows for phase transformations to initiate at lower amounts of compressive strain. Yet, in contrast to typical mechanochemical results, the sum of compressive and out of plane work needed to form diamond has a local minimum due to altered defect formation processes during phase transformation. Additionally, these altered processes slow the kinetics of the phase transformation, taking longer from initiation to total material transformation.

cond-mat.mtrl-sci↗

arXiv2022

Energy Localization Efficiency in TATB Pore Collapse Mechanisms

Atomistic and continuum scale modeling efforts have shown that shock induced collapse of porosity can occur via a wide range of mechanisms dependent on pore morphology, shockwave pressure, and material properties. The mechanisms that occur under weaker shocks tend to be more efficient at localizing thermal energy, but do not result in high, absolute temperatures or spatially large localizations compared to mechanisms found under strong shock conditions. However, the energetic material TATB undergoes a wide range of collapse mechanisms that are not typical of similar materials, leaving the collapse mechanisms and the resultant energy localization from collapse, i.e., hotspots, relatively uncharacterized. Therefore, we present pore collapse simulations of cylindrical pores in TATB for a wide range of pore sizes and shock strengths that trigger viscoplastic collapses that occur almost entirely perpendicular to the shock direction for weak shocks and hydrodynamic-like collapses for strong shocks that do not break the strong hydrogen bonds of the TATB basal planes. The resulting hotspot temperature fields from these mechanisms follow trends that differ considerably from other energetic materials; hence we compare them under normalized temperature values to assess the relative efficiency of each mechanism to localize energy. The local intra-molecular strain energy of the hotspots is also assessed to better understand the physical mechanisms behind the phenomena that leads to a latent potential energy.

cond-mat.mtrl-sci↗

arXiv2022

Predictive Scale-Bridging Simulations through Active Learning

Throughout computational science, there is a growing need to utilize the continual improvements in raw computational horsepower to achieve greater physical fidelity through scale-bridging over brute-force increases in the number of mesh elements. For instance, quantitative predictions of transport in nanoporous media, critical to hydrocarbon extraction from tight shale formations, are impossible without accounting for molecular-level interactions. Similarly, inertial confinement fusion simulations rely on numerical diffusion to simulate molecular effects such as non-local transport and mixing without truly accounting for molecular interactions. With these two disparate applications in mind, we develop a novel capability which uses an active learning approach to optimize the use of local fine-scale simulations for informing coarse-scale hydrodynamics. Our approach addresses three challenges: forecasting continuum coarse-scale trajectory to speculatively execute new fine-scale molecular dynamics calculations, dynamically updating coarse-scale from fine-scale calculations, and quantifying uncertainty in neural network models.

cs.LG↗

arXiv2022

Machine Learning of consistent thermodynamic models using automatic differentiation

We propose a data-driven method to describe consistent equations of state (EOS) for arbitrary systems. Complex EOS are traditionally obtained by fitting suitable analytical expressions to thermophysical data. A key aspect of EOS are that the relationships between state variables are given by derivatives of the system free energy. In this work, we model the free energy with an artificial neural network, and utilize automatic differentiation to directly learn the derivatives of the free energy. We demonstrate this approach on two different systems, the analytic van der Waals EOS, and published data for the Lennard-Jones fluid, and show that it is advantageous over direct learning of thermodynamic properties (i.e. not as derivatives of the free energy, but as independent properties), in terms of both accuracy and the exact preservation of the Maxwell relations. Furthermore, the method implicitly provides the free energy of a system without explicit integration.

physics.comp-ph↗

arXiv2020

Machine Learning-Powered Mitigation Policy Optimization in Epidemiological Models

A crucial aspect of managing a public health crisis is to effectively balance prevention and mitigation strategies, while taking their socio-economic impact into account. In particular, determining the influence of different non-pharmaceutical interventions (NPIs) on the effective use of public resources is an important problem, given the uncertainties on when a vaccine will be made available. In this paper, we propose a new approach for obtaining optimal policy recommendations based on epidemiological models, which can characterize the disease progression under different interventions, and a look-ahead reward optimization strategy to choose the suitable NPI at different stages of an epidemic. Given the time delay inherent in any epidemiological model and the exponential nature especially of an unmanaged epidemic, we find that such a look-ahead strategy infers non-trivial policies that adhere well to the constraints specified. Using two different epidemiological models, namely SEIR and EpiCast, we evaluate the proposed algorithm to determine the optimal NPI policy, under a constraint on the number of daily new cases and the primary reward being the absence of restrictions.

cs.LG↗

arXiv2020

Accurate Calibration of Agent-based Epidemiological Models with Neural Network Surrogates

Calibrating complex epidemiological models to observed data is a crucial step to provide both insights into the current disease dynamics, i.e.\ by estimating a reproductive number, as well as to provide reliable forecasts and scenario explorations. Here we present a new approach to calibrate an agent-based model -- EpiCast -- using a large set of simulation ensembles for different major metropolitan areas of the United States. In particular, we propose: a new neural network based surrogate model able to simultaneously emulate all different locations; and a novel posterior estimation that provides not only more accurate posterior estimates of all parameters but enables the joint fitting of global parameters across regions.

cs.LG↗

arXiv2016

Graph-based linear scaling electronic structure theory

We show how graph theory can be combined with quantum theory to calculate the electronic structure of large complex systems. The graph formalism is general and applicable to a broad range of electronic structure methods and materials, including challenging systems such as biomolecules. The methodology combines well-controlled accuracy, low computational cost, and natural low-communication parallelism. This combination addresses substantial shortcomings of linear scaling electronic structure theory, in particular with respect to quantum-based molecular dynamics simulations.

physics.comp-ph↗

arXiv2014

Mesodynamics with implicit degrees for freedom

Mesoscale phenomena -- involving a level of description between the finest atomistic scale and the macroscopic continuum -- can be studied by a variation on the usual atomistic-level molecular dynamics (MD) simulation technique. In mesodynamics, the mass points, rather than being atoms, are mesoscopic in size, for instance representing the centers of mass of polycrystalline grains or molecules. In order to reproduce many of the overall features of fully atomistic MD, which is inherently more expensive, the equations of motion in mesodynamics must be derivable from an interaction potential that is faithful to the compressive equation of state, as well as to tensile de-cohesion that occurs along the boundaries of the mesoscale units. Moreover, mesodynamics diffrers from Newton's equations of motion in that dissipation -- the exchange of energy between mesoparticles and their internal degrees of freedom (DoFs) -- must be described, and so should the transfer of energy between the internal modes of neighboring mesoparticles. We present a formulation where energy transfer between the internal modes of a mesoparticle and its external center-of-mass DoFs occurs in the phase space of mesoparticle coordinates, rather than momenta, resulting in a Galilean invariant formulation that conserves total linear momentum and energy (including the energy internal to the mesoparticles). We show that this approach can be used to describe, in addition to mesoscale problems, conduction electrons in atomic-level simulations of metals, and we demonstrate applications of mesodynamics to shockwave propagation and thermal transport.

physics.chem-ph↗

arXiv2010

Critical velocities for deflagration and detonation triggered by voids in a REBO high explosive

The effects of circular voids on the shock sensitivity of a two-dimensional model high explosive crystal are considered. We simulate a piston impact using molecular dynamics simulations with a Reactive Empirical Bond Order (REBO) model potential for a sub-micron, sub-ns exothermic reaction in a diatomic molecular solid. The probability of initiating chemical reactions is found to rise more suddenly with increasing piston velocity for larger voids that collapse more deterministically. A void with radius as small as 10 nm reduces the minimum initiating velocity by a factor of 4.

cond-mat.mtrl-sci↗

arXiv2008

Molecular Dynamics Simulations of Detonation Instability

After making modifications to the Reactive Empirical Bond Order potential for Molecular Dynamics (MD) of Brenner et al. in order to make the model behave in a more conventional manner, we discover that the new model exhibits detonation instability, a first for MD. The instability is analyzed in terms of the accepted theory.

cond-mat.mtrl-sci↗

arXiv2008

An Interaction Potential for Atomic Simulations of Conventional High Explosives

In an effort to develop a chemically reactive interaction potential suitable for application to the study of conventional, organic explosives, we have modified the diatomic AB potential of Brenner et al. such that it exhibits improved detonation characteristics. In particular, equilibrium molecular dynamics (MD) calculations of the modified potential demonstrate that the detonation products have an essentially diatomic, rather than polymeric, composition and that the detonation Hugoniot has the classic, concave-upward form. Nonequilibrium MD calculations reveal the separation of scales between chemical and hydrodynamic effects essential to the Zeldovitch, von Neumann, and Doering theory.

cond-mat.mtrl-sci↗

arXiv2006

The Influence of Interatomic Bonding Potentials on Detonation Properties

The dependence of macroscopic detonation properties of a two-dimensional diatomic (AB) molecular system on the fundamental properties of the molecule were investigated. This includes examining the detonation velocity, reaction zone thickness, and critical width as a function of the exothermicity of the gas-phase reaction and the gas-phase dissociation energy for. Following previous work, molecular dynamics (MD) simulations with a reactive empirical bond-order potential were used to characterize the shock-induced response of a diatomic AB molecular solid, which exothermically reacts to produce gaseous products. MD simulations reveal that there is a linear dependence between the square of the detonation velocity and each of these molecular parameters. The detonation velocities were shown to be consistent with the Chapman-Jouguet model, demonstrating that these dependencies arise from how the Equation of State of the products and reactants are affected.

cond-mat.mtrl-sci↗

arXiv1994

Dimensional Perturbation Theory on the Connection Machine

A recently developed linear algebraic method for the computation of perturbation expansion coefficients to large order is applied to the problem of a hydrogenic atom in a magnetic field. We take as the zeroth order approximation the $D \rightarrow \infty$ limit, where $D$ is the number of spatial dimensions. In this pseudoclassical limit, the wavefunction is localized at the minimum of an effective potential surface. A perturbation expansion, corresponding to harmonic oscillations about this minimum and higher order anharmonic correction terms, is then developed in inverse powers of $(D-1)$ about this limit, to 30th order. To demonstrate the implicit parallelism of this method, which is crucial if it is to be successfully applied to problems with many degrees of freedom, we describe and analyze a particular implementation on massively parallel Connection Machine systems (CM-2 and CM-5). After presenting performance results, we conclude with a discussion of the prospects for extending this method to larger systems.

chem-ph↗