-
Exploring the parameter dependence of atomic minima with implicit differentiation
Authors:
Ivan Maliyov,
Petr Grigorev,
Thomas D Swinburne
Abstract:
Interatomic potentials are essential to go beyond ab initio size limitations, but simulation results depend sensitively on potential parameters. Forward propagation of parameter variation is key for uncertainty quantification, whilst backpropagation has found application for emerging inverse problems such as fine-tuning or targeted design. Here, the implicit derivative of functions defined as a fi…
▽ More
Interatomic potentials are essential to go beyond ab initio size limitations, but simulation results depend sensitively on potential parameters. Forward propagation of parameter variation is key for uncertainty quantification, whilst backpropagation has found application for emerging inverse problems such as fine-tuning or targeted design. Here, the implicit derivative of functions defined as a fixed point is used to Taylor expand the energy and structure of atomic minima in potential parameters, evaluating terms via automatic differentiation, dense linear algebra or a novel sparse operator approach. The latter allows efficient forward and backpropagation through relaxed structures of arbitrarily large systems. The implicit expansion accurately predicts lattice distortion and defect formation energies and volumes with classical and machine-learning potentials, enabling high-dimensional uncertainty propagation without prohibitive overhead. We then show how the implicit derivative can be used to solve challenging inverse problems, minimizing an implicit loss to fine-tune potentials and stabilize solute-induced structural rearrangements at dislocations in tungsten.
△ Less
Submitted 2 July, 2024;
originally announced July 2024.
-
ParSplice: strong exa-scaling of molecular dynamics
Authors:
Thomas D Swinburne
Abstract:
ParSplice D. Perez, E. D. Cubuk, A. Waterland, E. Kaxiras, and A. F. Voter, Long-Time Dynamics through Parallel Trajectory Splicing, Journal of Chemical Theory and Computation, 2016 is a molecular dynamics method for parallel-in-time trajectory generation, allowing this workhorse of in silico science to strong scale on massively parallel computers. Trajectories generated by ParSplice always have r…
▽ More
ParSplice D. Perez, E. D. Cubuk, A. Waterland, E. Kaxiras, and A. F. Voter, Long-Time Dynamics through Parallel Trajectory Splicing, Journal of Chemical Theory and Computation, 2016 is a molecular dynamics method for parallel-in-time trajectory generation, allowing this workhorse of in silico science to strong scale on massively parallel computers. Trajectories generated by ParSplice always have robust theoretical guarantees on their validity, with parameter choices only affecting the parallel efficiency. This commentary summarizes the ParSplice approach with minimal mathematical development, emphasizing how the theoretical underpinning is essential for deployment at the exascale.
△ Less
Submitted 5 February, 2024;
originally announced February 2024.
-
Parameter uncertainties for imperfect surrogate models in the low-noise regime
Authors:
Thomas D Swinburne,
Danny Perez
Abstract:
Bayesian regression determines model parameters by minimizing the expected loss, an upper bound to the true generalization error. However, the loss ignores misspecification, where models are imperfect. Parameter uncertainties from Bayesian regression are thus significantly underestimated and vanish in the large data limit. This is particularly problematic when building models of low-noise, or near…
▽ More
Bayesian regression determines model parameters by minimizing the expected loss, an upper bound to the true generalization error. However, the loss ignores misspecification, where models are imperfect. Parameter uncertainties from Bayesian regression are thus significantly underestimated and vanish in the large data limit. This is particularly problematic when building models of low-noise, or near-deterministic, calculations, as the main source of uncertainty is neglected. We analyze the generalization error of misspecified, near-deterministic surrogate models, a regime of broad relevance in science and engineering. We show posterior distributions must cover every training point to avoid a divergent generalization error and design an ansatz that respects this constraint, which for linear models incurs minimal overhead. This is demonstrated on model problems before application to thousand dimensional datasets in atomistic machine learning. Our efficient misspecification-aware scheme gives accurate prediction and bounding of test errors where existing schemes fail, allowing this important source of uncertainty to be incorporated in computational workflows.
△ Less
Submitted 7 May, 2024; v1 submitted 2 February, 2024;
originally announced February 2024.
-
A foundation model for atomistic materials chemistry
Authors:
Ilyes Batatia,
Philipp Benner,
Yuan Chiang,
Alin M. Elena,
Dávid P. Kovács,
Janosh Riebesell,
Xavier R. Advincula,
Mark Asta,
Matthew Avaylon,
William J. Baldwin,
Fabian Berger,
Noam Bernstein,
Arghya Bhowmik,
Samuel M. Blau,
Vlad Cărare,
James P. Darby,
Sandip De,
Flaviano Della Pia,
Volker L. Deringer,
Rokas Elijošius,
Zakariya El-Machachi,
Fabio Falcioni,
Edvin Fako,
Andrea C. Ferrari,
Annalena Genreith-Schriever
, et al. (51 additional authors not shown)
Abstract:
Machine-learned force fields have transformed the atomistic modelling of materials by enabling simulations of ab initio quality on unprecedented time and length scales. However, they are currently limited by: (i) the significant computational and human effort that must go into development and validation of potentials for each particular system of interest; and (ii) a general lack of transferabilit…
▽ More
Machine-learned force fields have transformed the atomistic modelling of materials by enabling simulations of ab initio quality on unprecedented time and length scales. However, they are currently limited by: (i) the significant computational and human effort that must go into development and validation of potentials for each particular system of interest; and (ii) a general lack of transferability from one chemical system to the next. Here, using the state-of-the-art MACE architecture we introduce a single general-purpose ML model, trained on a public database of 150k inorganic crystals, that is capable of running stable molecular dynamics on molecules and materials. We demonstrate the power of the MACE-MP-0 model - and its qualitative and at times quantitative accuracy - on a diverse set problems in the physical sciences, including the properties of solids, liquids, gases, chemical reactions, interfaces and even the dynamics of a small protein. The model can be applied out of the box and as a starting or "foundation model" for any atomistic system of interest and is thus a step towards democratising the revolution of ML force fields by lowering the barriers to entry.
△ Less
Submitted 1 March, 2024; v1 submitted 29 December, 2023;
originally announced January 2024.
-
Compressing and forecasting atomic material simulations with descriptors
Authors:
Thomas D Swinburne
Abstract:
Atomic simulations of material microstructure require significant resources to generate, store and analyze. Here, atomic descriptor functions are proposed as a general latent space to compress atomic microstructure, ideal for use in large-scale simulations. Descriptors can regress a broad range of properties, including character-dependent dislocation densities, stress states or radial distribution…
▽ More
Atomic simulations of material microstructure require significant resources to generate, store and analyze. Here, atomic descriptor functions are proposed as a general latent space to compress atomic microstructure, ideal for use in large-scale simulations. Descriptors can regress a broad range of properties, including character-dependent dislocation densities, stress states or radial distribution functions. A vector autoregressive model can generate trajectories over yield points, resample from new initial conditions and forecast trajectory futures. A forecast confidence, essential for practical application, is derived by propagating forecasts through the Mahalanobis outlier distance, providing a powerful tool to assess coarse-grained models. Application to nanoparticles and yielding of dislocation networks confirms low uncertainty forecasts are accurate and resampling allows for the propagation of smooth microstructure distributions. Yielding is associated with a collapse in the intrinsic dimension of the descriptor manifold, which is discussed in relation to the yield surface.
△ Less
Submitted 5 September, 2023;
originally announced September 2023.
-
Atomistic simulations of nanoindentation in single crystalline tungsten: The role of interatomic potentials
Authors:
F. J. Dominguez-Gutierrez,
P. Grigorev,
A. Naghdi,
Q. Q. Xu,
J. Byggmastar,
G. Y. Wei,
T. D. Swinburne,
S. Papanikolaou,
M. J. Alava
Abstract:
Computational modeling is usually applied to aid experimental exploration of advanced materials to better understand the fundamental plasticity mechanisms during mechanical testing. In this work, we perform Molecular dynamics (MD) simulations to emulate experimental room temperature spherical-nanoindentation of crystalline W matrices by different interatomic potentials: EAM, modified EAM, and a re…
▽ More
Computational modeling is usually applied to aid experimental exploration of advanced materials to better understand the fundamental plasticity mechanisms during mechanical testing. In this work, we perform Molecular dynamics (MD) simulations to emulate experimental room temperature spherical-nanoindentation of crystalline W matrices by different interatomic potentials: EAM, modified EAM, and a recently developed machine learned based tabulated Gaussian approximation potential (tabGAP) for describing the interaction of W-W. Results show similarities between load displacements and stress-strain curves, regardless of the numerical model. However, a discrepancy is observed at early stages of the elastic to plastic deformation transition showing different mechanisms for dislocation nucleation and evolution, that is attributed to the difference of Burgers vector magnitudes, stacking fault and dislocation glide energies. Besides, contact pressure is investigated by considering large indenters sizes that provides a detailed analysis of screw and edge dislocations during loading process. Furthermore, the glide barrier of this kind of dislocations are reported for all the interatomic potentials showing that tabGAP model presents the most accurate results with respect to density functional theory calculations and a good qualitative agreement with reported experimental data
△ Less
Submitted 11 December, 2022; v1 submitted 18 May, 2022;
originally announced May 2022.
-
Automated calculation and convergence of defect transport tensors
Authors:
Thomas D Swinburne,
Danny Perez
Abstract:
Defect transport is a key process in materials science and catalysis, but as migration mechanisms are often too complex to enumerate a priori, calculation of transport tensors typically have no measure of convergence and require significant end user intervention. These two bottlenecks prevent high-throughput implementations essential to propagate model-form uncertainty from interatomic interaction…
▽ More
Defect transport is a key process in materials science and catalysis, but as migration mechanisms are often too complex to enumerate a priori, calculation of transport tensors typically have no measure of convergence and require significant end user intervention. These two bottlenecks prevent high-throughput implementations essential to propagate model-form uncertainty from interatomic interactions to predictive simulations. In order to address these issues, we extend a massively parallel accelerated sampling scheme, autonomously controlled by Bayesian estimators of statewise sampling completeness, to build atomistic kinetic Monte Carlo models on a state space irreducible under exchange and space group symmetries. Focusing on isolated defects, we derive analytic expressions for defect transport tensors and provide a convergence metric by calculating the Kullback-Leiber divergence across the ensemble of diffusion processes consistent with the sampling uncertainty. The autonomy and efficacy of the method is demonstrated on surface trimers in tungsten and hexa-interstitials in magnesium oxide, both of which exhibit complex, correlated migration mechanisms.
△ Less
Submitted 3 September, 2020; v1 submitted 17 March, 2020;
originally announced March 2020.
-
Self-optimized construction of transition rate matrices from accelerated atomistic simulations with Bayesian uncertainty quantification
Authors:
Thomas D Swinburne,
Danny Perez
Abstract:
A massively parallel method to build large transition rate matrices from temperature accelerated molecular dynamics trajectories is presented. Bayesian Markov model analysis is used to estimate the expected residence time in the known state space, providing crucial uncertainty quantification for higher scale simulation schemes such as kinetic Monte Carlo or cluster dynamics. The estimators are add…
▽ More
A massively parallel method to build large transition rate matrices from temperature accelerated molecular dynamics trajectories is presented. Bayesian Markov model analysis is used to estimate the expected residence time in the known state space, providing crucial uncertainty quantification for higher scale simulation schemes such as kinetic Monte Carlo or cluster dynamics. The estimators are additionally used to optimize where exploration is performed and the degree of temperature ac- celeration on the fly, giving an autonomous, optimal procedure to explore the state space of complex systems. The method is tested against exactly solvable models and used to explore the dynamics of C15 interstitial defects in iron. Our uncertainty quantification scheme allows for accurate modeling of the evolution of these defects over timescales of several seconds.
△ Less
Submitted 7 March, 2018;
originally announced March 2018.
