arXiv subjects

Tomas Oppelstrup

Publications and source records attributed to Tomas Oppelstrup.

15 recordsLinked to original sources

arXiv2026

Ductility and Brittle Fracture of Tungsten by Disconnection Pile-up on Twin Boundaries

Refractory body-centered cubic (BCC) metals and alloys are of extraordinary importance in modern technological and structural applications. However, their wider adoption in science and technology is severely restricted by low-temperature brittleness, quantified by an unacceptably high value of the brittle-to ductile transition temperature (DBTT). The DBTT of these alloys is known to depend strongly on the particular microstructure of the material following mechanisms that are not well understood. Here we apply cross-scale molecular dynamics (MD), a simulation approach that preserves full atomic resolution while capturing the collective evolution of dislocations, twins, and cracks in near-micron-scale volumes, to investigate ductility and fracture in single-crystal tungsten pillars as a function of initial defect microstructure, deformation conditions, and temperature. The simulations reveal a sequence of microscopic processes conducive to failure: dislocation starvation, nucleation and growth of twins, pinning of the twin boundaries at surface asperities, resulting in disconnection pile-ups that trigger crack nucleation and propagation at low macroscopic stresses along incoherent boundary segments. By resolving these processes within a single atomistic framework, our simulations connect defect-level dynamics to macroscopic fracture behavior and identify microstructural pathways capable of shifting the DBTT through targeted promotion or suppression of the underlying deformation mechanisms.

cond-mat.mtrl-sci↗

arXiv2025

Scaling Kinetic Monte-Carlo Simulations of Grain Growth with Combined Convolutional and Graph Neural Networks

Graph neural networks (GNN) have emerged as a promising machine learning method for microstructure simulations such as grain growth. However, accurate modeling of realistic grain boundary networks requires large simulation cells, which GNN has difficulty scaling up to. To alleviate the computational costs and memory footprint of GNN, we propose a hybrid architecture combining a convolutional neural network (CNN) based bijective autoencoder to compress the spatial dimensions, and a GNN that evolves the microstructure in the latent space of reduced spatial sizes. Our results demonstrate that the new design significantly reduces computational costs with using fewer message passing layer (from 12 down to 3) compared with GNN alone. The reduction in computational cost becomes more pronounced as the spatial size increases, indicating strong computational scalability. For the largest mesh evaluated (160^3), our method reduces memory usage and runtime in inference by 117x and 115x, respectively, compared with GNN-only baseline. More importantly, it shows higher accuracy and stronger spatiotemporal capability than the GNN-only baseline, especially in long-term testing. Such combination of scalability and accuracy is essential for simulating realistic material microstructures over extended time scales. The improvements can be attributed to the bijective autoencoder's ability to compress information losslessly from spatial domain into a high dimensional feature space, thereby producing more expressive latent features for the GNN to learn from, while also contributing its own spatiotemporal modeling capability. The training was optimized to learn from the stochastic Potts Monte Carlo method. Our findings provide a highly scalable approach for simulating grain growth.

cs.LG↗

arXiv2025

Beyond Exascale: Dataflow Domain Translation on a Cerebras Cluster

Simulation of physical systems is essential across scientific and engineering domains. Commonly used domain decomposition methods are unable to simultaneously deliver both high simulation rate and high utilization in network computing environments. In particular, Exascale systems deliver only a small fraction their peak performance for these workloads. This paper introduces the novel Domain Translation algorithm, designed to overcome these limitations. On a cluster of 64 Cerebras CS-3 systems, we use this method to demonstrate unprecedented cluster performance across a range of metrics: we show simulations running in excess of 1.6 million time steps per second; we also demonstrate perfect weak scaling at 88% of peak performance. At this cluster scale, our implementation provides 112 PFLOP/s in a power-unconstrained environment, and 57 GFLOP/J in a power-limited environment. We illustrate the method by applying the shallow-water equations to model a tsunami following an asteroid impact at 460m-resolution on a planetary scale.

cs.DC↗

arXiv2024

Breaking the mold: overcoming the time constraints of molecular dynamics on general-purpose hardware

The evolution of molecular dynamics (MD) simulations has been intimately linked to that of computing hardware. For decades following the creation of MD, simulations have improved with computing power along the three principal dimensions of accuracy, atom count (spatial scale), and duration (temporal scale). Since the mid-2000s, computer platforms have however failed to provide strong scaling for MD as scale-out CPU and GPU platforms that provide substantial increases to spatial scale do not lead to proportional increases in temporal scale. Important scientific problems therefore remained inaccessible to direct simulation, prompting the development of increasingly sophisticated algorithms that present significant complexity, accuracy, and efficiency challenges. While bespoke MD-only hardware solutions have provided a path to longer timescales for specific physical systems, their impact on the broader community has been mitigated by their limited adaptability to new methods and potentials. In this work, we show that a novel computing architecture, the Cerebras Wafer Scale Engine, completely alters the scaling path by delivering unprecedentedly high simulation rates up to 1.144M steps/second for 200,000 atoms whose interactions are described by an Embedded Atom Method potential. This enables direct simulations of the evolution of materials using general-purpose programmable hardware over millisecond timescales, dramatically increasing the space of direct MD simulations that can be carried out.

cs.DC↗

arXiv2024

Breaking the Molecular Dynamics Timescale Barrier Using a Wafer-Scale System

Molecular dynamics (MD) simulations have transformed our understanding of the nanoscale, driving breakthroughs in materials science, computational chemistry, and several other fields, including biophysics and drug design. Even on exascale supercomputers, however, runtimes are excessive for systems and timescales of scientific interest. Here, we demonstrate strong scaling of MD simulations on the Cerebras Wafer-Scale Engine. By dedicating a processor core for each simulated atom, we demonstrate a 179-fold improvement in timesteps per second versus the Frontier GPU-based Exascale platform, along with a large improvement in timesteps per unit energy. Reducing every year of runtime to two days unlocks currently inaccessible timescales of slow microstructure transformation processes that are critical for understanding material behavior and function. Our dataflow algorithm runs Embedded Atom Method (EAM) simulations at rates over 270,000 timesteps per second for problems with up to 800k atoms. This demonstrated performance is unprecedented for general-purpose processing cores.

physics.comp-ph↗

arXiv2023

Dynamic Density Functional Theory of Multicomponent Cellular Membranes

We present a continuum model trained on molecular dynamics (MD) simulations for cellular membranes composed of an arbitrary number of lipid types. The model is constructed within the formalism of dynamic density functional theory and can be extended to include features such as the presence of proteins and membrane deformations. This framework represents a paradigm shift by enabling simulations that can access length scales on the order of microns and time scales on the order of seconds, all while maintaining near fidelity to the underlying MD models. These length and time scales are significant for accessing biological processes associated with signaling pathways within cells. Membrane interactions with RAS, a protein implicated in roughly 30% of human cancers, are considered as an application. Simulation results are presented and verified with MD simulations, and implications of this new capability are discussed.

physics.bio-ph↗

arXiv2021

Nucleation of Grain Boundary Phases

We derive a theory that describes homogeneous nucleation of grain boundary (GB) phases. Our analysis takes account of the energy resulting from the GB phase junction, the line defect separating two different GB structures, which is necessarily a dislocation as well as an elastic line force due to the jump in GB stresses. The theory provides analytic forms for the elastic interactions and the core energy of the GB phase junction that, along with the change in GB energy, determine the nucleation barrier. We apply the resulting nucleation model to simulations of GB phase transformations in tungsten. Our theory explains why under certain conditions GBs cannot spontaneously change their structure even to a lower energy state.

cond-mat.mtrl-sci↗

arXiv2020

Metal hardening in atomistic detail

Through millennia humans exploited the natural property of metals to get stronger or hardened when mechanically deformed. Ultimately rooted in the motion of dislocations, mechanisms of metal hardening remained in the crosshairs of physical metallurgists for over a century. Here, we performed atomistic simulations at the limits of supercomputing, which are sufficiently large to be statistically representative of macroscopic crystal plasticity yet fully resolved to examine the origins of metal hardening at its most fundamental level of atomic motion. We demonstrate that the notorious staged (inflection) hardening of metals is a direct consequence of crystal rotation under uniaxial straining. At variance with widely divergent and contradictory views in the literature, we observe that basic mechanisms of dislocation behavior are the same across all stages of metal hardening.

cond-mat.mtrl-sci↗

arXiv2018

Structures and transitions in bcc tungsten grain boundaries and their role in the absorption of point defects

We use atomistic simulations to investigate grain boundary (GB) phase transitions in el- emental body-centered cubic (bcc) metal tungsten. Motivated by recent modeling study of grain boundary phase transitions in [100] symmetric tilt boundaries in face-centered cu- bic (fcc) copper, we perform a systematic investigation of [100] and [110] symmetric tilt high-angle and low-angle boundaries in bcc tungsten. The structures of these boundaries have been investigated previously by atomistic simulations in several different bcc metals including tungsten using the the γ-surface method, which has limitations. In this work we use a recently developed computational tool based on the USPEX structure prediction code to perform an evolutionary grand canonical search of GB structure at 0 K. For high-angle [100] tilt boundaries the ground states generated by the evolutionary algorithm agree with the predictions of the γ-surface method. For the [110] tilt boundaries, the search predicts novel high-density low-energy grain boundary structures and multiple grain boundary phases within the entire misorientation range. Molecular dynamics simulation demonstrate that the new structures are more stable at high temperature. We observe first-order grain boundary phase transitions and investigate how the structural multiplicity affects the mechanisms of the point defect absorption. Specifically, we demonstrate a two-step nucleation process, when initially the point defects are absorbed through a formation of a metastable GB structure with higher density, followed by a transformation of this structure into a GB interstitial loop or a different GB phase.

cond-mat.mtrl-sci↗

arXiv2017

Visual, user-interactive generation of bond networks in 3D particle configurations

We present a new program able to perform visual structural analysis on 3D particle systems called PASYVAT (PArticle SYstem Visual Analysis Tool). More specifically, it can select multiple interparticle distance ranges from a radial distribution function (RDF) plot and display them in 3D as bonds between the particles falling within the selected distance range, thus generating a network of bonds. This software can be used with any data set representing a system of points or other objects having a welldefined center of mass or geometric center in 3D space. In this article we describe the program and its internal structure, with emphasis on its applicability in the study of certain particle configurations, obtained from classical molecular dynamics simulation in condensed matter physics.

physics.comp-ph↗

arXiv2016

Probing the ultimate limits of metal plasticity

Along with high strength, plasticity is what makes metals so widely usable in our material world. Both strength and plasticity properties of a metal are defined by the motion of dislocations - line defects in the crystal lattice that divide areas of atomic planes displaced relative to each other by an interatomic distance. Here we present first fully dynamic atomistic simulations of single crystal plasticity in metal tantalum predicting that above certain maximum rate of straining - the ultimate limit - the dislocations can no longer relieve mechanical loads and another mechanism, twinning, comes into play and takes over as the dominant mode of dynamic response. At straining rates below the ultimate limit, the metal attains a path-independent stationary state of plastic flow in which both flow stress and dislocation density remain constant indefinitely for as long as the straining conditions remain unchanged. In this distinct state tantalum flows like a viscous fluid while still remaining a strong and stiff metal.

cond-mat.mtrl-sci↗

arXiv2015

Formation of a Columnar Liquid Crystal in a Simple One-Component System of Particles

We report a molecular dynamics simulation demonstrating that a columnar liquid crystal, commonly formed by disc-shaped molecules, can be formed by identical particles interacting via a spherically symmetric potential. Upon isochoric cooling from a low-density isotropic liquid state the simulated system performed a weak first order phase transition which produced a liquid crystal phase composed of parallel particle columns arranged in a hexagonal pattern in the plane perpendicular to the column axis. The particles within columns formed a liquid structure and demonstrated a significant intracolumn diffusion. Further cooling resulted in another first-order transition whereby the column structure became periodically ordered in three dimensions transforming the liquid-crystal phase into a crystal. This result is the first observation of a liquid crystal formation in a simple one-component system of particles. Its conceptual significance is in that it demonstrated that liquid crystals that have so far only been produced in systems of anisometric molecules, can also be formed by mesoscopic soft-matter and colloidal systems of spherical particles with appropriately tuned interatomic potential.

cond-mat.soft↗

arXiv2010

A simple monatomic ideal glass former: the glass transition by a first-order phase transition above the melting point

A liquid can form under cooling a glassy state either as a result of a continuous slowing down or by a first order polyamorphous phase transition. The second scenario has so far always been observed below the melting point where it interfered with crystalline nucleation. We report the first observation of the liquid-glass transition by a first order phase transition above the melting point. The observation was made in a molecular dynamics simulation of a one-component system with a model metallic pair potential. This is also the first observation of a simple monatomic ideal glass former -- a liquid that avoids crystallization at any cooling rate. Besides its conceptual importance, this result indicates a possibility of existence of metallic ideal glass formers.

cond-mat.soft↗

arXiv2008

Anomalous diffusion in supercooled liquids: a long-range localisation in particle trajectories

A statistical analysis of the geometries of particle trajectories in the supercooled liquid state is reported. We examine two structurally different fragile glass-forming liquids simulated by molecular dynamics. In both liquids, the trajectories are found to exhibit a long-range localisation distinct from the short-range localisation within the cage of nearest neighbours. This novel diffusion anomaly is interpreted as a result of the potential-energy landscape topography of fragile glass-formers where the local energy minima coalesce into metabasins - compact domains with low escape probability.

cond-mat.soft↗

arXiv2005

Universality of liquid dynamics

We investigate the origin of the Stokes-Einstein relation in liquids. The hard-sphere dynamics is analyzed using a new measure of structural relaxation - the minimum Euclidean distance between configurations of particles. It is shown that the universal relation between the structural relaxation and diffusion in liquids is caused by the existence of one dominating length scale imposed by the structural correlations and associated with de Gennes narrowing. We demonstrate that this relation can be described by a model of independent random walkers under the single-occupancy constraint.

cond-mat.soft↗