-
Uncertainty in AI-driven Monte Carlo simulations
Authors:
Dimitrios Tzivrailis,
Alberto Rosso,
Eiji Kawasaki
Abstract:
In the study of complex systems, evaluating physical observables often requires sampling representative configurations via Monte Carlo techniques. These methods rely on repeated evaluations of the system's energy and force fields, which can become computationally expensive, particularly in the presence of long-range interactions. To accelerate these simulations, deep learning models are increasing…
▽ More
In the study of complex systems, evaluating physical observables often requires sampling representative configurations via Monte Carlo techniques. These methods rely on repeated evaluations of the system's energy and force fields, which can become computationally expensive, particularly in the presence of long-range interactions. To accelerate these simulations, deep learning models are increasingly employed as surrogate functions to approximate the energy landscape or force fields. However, such models introduce epistemic uncertainty in their predictions, which may propagate through the sampling process and affect the system's macroscopic behavior. In this work, we present the Penalty Ensemble Method (PEM) to quantify epistemic uncertainty and mitigate its impact on Monte Carlo sampling. Our approach introduces an uncertainty-aware modification of the Metropolis acceptance rule, which increases the rejection probability in regions of high uncertainty, thereby enhancing the reliability of the simulation outcomes.
△ Less
Submitted 17 June, 2025;
originally announced June 2025.
-
Targeting the partition function of chemically disordered materials with a generative approach based on inverse variational autoencoders
Authors:
Maciej J. Karcz,
Luca Messina,
Eiji Kawasaki,
Emeric Bourasseau
Abstract:
Computing atomic-scale properties of chemically disordered materials requires an efficient exploration of their vast configuration space. Traditional approaches such as Monte Carlo or Special Quasirandom Structures either entail sampling an excessive amount of configurations or do not ensure that the configuration space has been properly covered. In this work, we propose a novel approach where gen…
▽ More
Computing atomic-scale properties of chemically disordered materials requires an efficient exploration of their vast configuration space. Traditional approaches such as Monte Carlo or Special Quasirandom Structures either entail sampling an excessive amount of configurations or do not ensure that the configuration space has been properly covered. In this work, we propose a novel approach where generative machine learning is used to yield a representative set of configurations for accurate property evaluation and provide accurate estimations of atomic-scale properties with minimal computational cost. Our method employs a specific type of variational autoencoder with inverse roles for the encoder and decoder, enabling the application of an unsupervised active learning scheme that does not require any initial training database. The model iteratively generates configuration batches, whose properties are computed with conventional atomic-scale methods. These results are then fed back into the model to estimate the partition function, repeating the process until convergence. We illustrate our approach by computing point-defect formation energies and concentrations in (U, Pu)O2 mixed-oxide fuels. In addition, the ML model provides valuable insights into the physical factors influencing the target property. Our method is generally applicable to explore other properties, such as atomic-scale diffusion coefficients, in ideally or non-ideally disordered materials like high-entropy alloys.
△ Less
Submitted 10 September, 2024; v1 submitted 27 August, 2024;
originally announced August 2024.
-
Semi-supervised generative approach to point-defect formation in chemically disordered compounds: application to uranium-plutonium mixed oxides
Authors:
Maciej J. Karcz,
Luca Messina,
Eiji Kawasaki,
Serenah Rajaonson,
Didier Bathellier,
Emeric Bourasseau
Abstract:
Machine-learning methods are nowadays of common use in the field of material science. For example, they can aid in optimizing the physicochemical properties of new materials, or help in the characterization of highly complex chemical compounds. An especially challenging problem arises in the modeling of chemically disordered solid solutions, for which some properties depend on the distribution of…
▽ More
Machine-learning methods are nowadays of common use in the field of material science. For example, they can aid in optimizing the physicochemical properties of new materials, or help in the characterization of highly complex chemical compounds. An especially challenging problem arises in the modeling of chemically disordered solid solutions, for which some properties depend on the distribution of chemical species in the crystal lattice. This is the case of defect properties of uranium-plutonium mixed oxides nuclear fuels. The number of possible configurations is so high that the problem becomes intractable if treated with direct sampling. We thus propose a machine learning approach, based on generative modeling, to optimize the exploration of this large configuration space. A probabilistic, semi-supervised approach using Mixture Density Network is applied to estimate the concentration of thermal defects in (U, Pu)O2. We show that this method, based on the prediction of the density of states of formation energy of a defect, is computationally much more cost-efficient compared to other approaches available in the literature.
△ Less
Submitted 28 November, 2022; v1 submitted 22 November, 2022;
originally announced November 2022.
-
Finite Temperature Phases of Two Dimensional Spin-Orbit Coupled Bosons
Authors:
Eiji Kawasaki,
Markus Holzmann
Abstract:
We determine the finite temperature phase diagram of two dimensional bosons with two hyperfine (pseudo-spin) states coupled via Rashba-Dresselhaus spin-orbit interaction using classical field Monte Carlo calculations. For anisotropic spin-orbit coupling, we find a transition to a Berenzinskii-Kosterlitz-Thouless superfluid phase with quasi-long range order. We show that the spin-order of the quasi…
▽ More
We determine the finite temperature phase diagram of two dimensional bosons with two hyperfine (pseudo-spin) states coupled via Rashba-Dresselhaus spin-orbit interaction using classical field Monte Carlo calculations. For anisotropic spin-orbit coupling, we find a transition to a Berenzinskii-Kosterlitz-Thouless superfluid phase with quasi-long range order. We show that the spin-order of the quasi-condensate is driven by the anisotropy of interparticle interaction, favoring either a homogeneous plane wave state or stripe phase with broken translational symmetry. Both phases show characteristic behavior in the algebraically decaying spin density correlation function. For fully isotropic interparticle interaction, our calculations indicate a fractionalized quasi-condensate where the mean-field degeneracy of plane wave and stripe phase remains robust against critical fluctuations. In the case of fully isotropic spin-orbit coupling, the circular degeneracy of the single particle ground state destroys the algebraic ordered phase in the thermodynamic limit, but a cross-over remains for finite size systems.
△ Less
Submitted 18 January, 2017;
originally announced January 2017.
-
Dynamical depinning of a Tonks Girardeau gas
Authors:
Florian Cartarius,
Eiji Kawasaki,
Anna Minguzzi
Abstract:
We study the dynamical depinning following a sudden turn off of an optical lattice for a gas of impenetrable bosons in a tight atomic waveguide. We use a Bose-Fermi mapping to infer the exact quantum dynamical evolution. At long times, in the thermodynamic limit, we observe the approach to a non-equilibrium steady state, characterized by the absence of quasi-long-range order and a reduced visibili…
▽ More
We study the dynamical depinning following a sudden turn off of an optical lattice for a gas of impenetrable bosons in a tight atomic waveguide. We use a Bose-Fermi mapping to infer the exact quantum dynamical evolution. At long times, in the thermodynamic limit, we observe the approach to a non-equilibrium steady state, characterized by the absence of quasi-long-range order and a reduced visibility in the momentum distribution. Similar features are found in a finite-size system at times corresponding to half the revival time, where we find that the system approaches a quasi-steady state with a power-law behaviour.
△ Less
Submitted 26 October, 2015; v1 submitted 5 May, 2015;
originally announced May 2015.
