arXiv subjects

Mihai-Cosmin Marinica

Publications and source records attributed to Mihai-Cosmin Marinica.

At least 19 recordsLinked to original sources

arXiv2026

A strategic roadmap for an atomistic machine-learning ecosystem

Data-driven machine learning (ML) techniques have become an essential tool in many domains of science. Their application to atomistic simulations of matter is particularly widespread and impactful. This success is due largely to the existence of a well-developed and established physics-based modeling framework, ranging from first-principles electronic-structure calculations to molecular dynamics and statistical sampling, into which ML was integrated naturally to reshape long-standing trade-offs between accuracy, efficiency, and scale. Nevertheless, this integration raises both conceptual and practical challenges, from choosing between data-centric and physics-based modeling approaches to adapting established software stacks to modern hardware accelerators and ML libraries. As the field evolves rapidly, fueled in part by widespread enthusiasm but also by tangible impact, it seems appropriate to take a moment to consider the current state of the art and open challenges, and reflect on what can be done to better coordinate efforts across the community. With this goal in mind, several members of this community met in Lausanne in January 2026 at CECAM to discuss algorithms, models, software and hardware infrastructure, and the most promising scientific applications that have become possible thanks to the use of artificial intelligence in atomic-scale simulations. This strategic roadmap paper summarizes the outcomes of these discussions, suggesting some long-term goals, and some concrete actions, to establish a healthy, sustainable and impactful atomistic ML ecosystem.

physics.chem-ph↗

arXiv2026

Revisiting quantum effects on dislocation glide in bcc metals from DFT calculations and machine-learning potentials

Quantum zero-point effects have been long proposed to explain a well-known discrepancy between the low-temperature flow stresses of body-centered cubic metals and corresponding atomistic models of plastic flow. Previous investigations on quantum effects relied on empirical interatomic potentials, which poorly reproduce dislocation energy landscapes compared to density functional theory (DFT) calculations. Here, we revisit this problem using DFT and machine-learning interatomic potentials (MLIPs). We show that while quantum effects do contribute to dislocation glide at low temperature, their magnitude is much lower than previously reported, and insufficient to reconcile atomistic predictions with experiments. Our results thus reopen a long-standing question and challenge for predictive atomistic modeling, on a fundamental property of crystals.

cond-mat.mtrl-sci↗

arXiv2026

Systematic and accurate anharmonic formation free energies of metastable defects via a constrained Bayesian Adaptive Biasing Force framework

Computing the formation free energies of metastable defects at finite temperature remains challenging due to anharmonicity, the multiplicity of basins and the frequent occurrence of migration events. Using the previously introduced constrained Bayesian Adaptive Biasing Force method (BABFc), we establish a practical thermodynamic integration framework for computing restricted anharmonic formation free energies associated with individual metastable defect basins. We demonstrate that BABFc can be exploited as a robust, systematic and numerically efficient tool for addressing a long standing bottleneck in atomistic materials science: the finite temperature thermodynamic characterization of metastable defects in complex, highly anharmonic energy landscapes. The proposed workflow requires only a reference local minimum and a confinement strategy and does not rely on defect specific collective variables. It therefore provides a general route to assign well-defined free energies to individual metastable basins, even when these basins are separated by low barriers and embedded in a dense landscape of competing configurations. In this setting, BABFc enables stable, bias corrected free energy estimates with statistical accuracies at the meV/atom level and a very low failure rate, including for systems containing approximately one thousand atoms. We demonstrate this capability through thousands of independent free energy calculations covering hundreds of four-interstitial and four-vacancy configurations in bcc $α$-Fe over a broad temperature range, using both a traditional empirical potential and a data driven force field.

cond-mat.mtrl-sci↗

arXiv2026

Girsanov Reweighting for Uncertainty Propagation in Rare-Event Kinetics

Machine-learning interatomic potentials (MLIPs) have become a powerful tool for rare event sampling in molecular dynamics, offering near ab initio accuracy at a fraction of the computational cost. However, the uncertainty associated with these models remains a major challenge. Existing uncertainty quantification approaches have largely focused on point-wise quantities, such as energies and forces, or on equilibrium thermodynamic observables. In this work, we introduce a framework for propagating MLIP uncertainty to the averaged committor probability, a kinetic observable that enables reaction-rate calculations. Our approach combines rare event sampling methods such as Adaptive Multilevel Splitting with Girsanov reweighting to estimate the sensitivity of committor probabilities to variations in MLIP parameters, without requiring the costly resampling of reactive trajectories for each parameter realization. We derive exact and approximate Girsanov-based estimators for uncertainty propagation and validate them on several benchmark systems, including a rugged Muller-Brown potential, a dimer in a solvent, and the conformational transition of butane. The proposed framework enables the construction of uncertainty-aware probability distributions for rare event observables and successfully recovers reference rare event probabilities from uncertain surrogate models. Under mild assumptions on the accuracy of the MLIP within metastable basins, the framework can also provide uncertainty bounds on reaction rates through Hill's relation. These results demonstrate that path-space reweighting provides an efficient route for propagating MLIP uncertainty to rare event kinetics.

physics.chem-ph↗

arXiv2026

Fully anharmonic calculations of the free energy of migration of point defects in UO2 and PuO2

Calculating diffusion rates of point defects in materials typically relies on the harmonic approximation to estimate migration free energies. However, anharmonic effects can have a large impact on diffusion properties, and explicitly accounting for them is usually computationally demanding and difficult to achieve in practice. In this work, we investigate the role of anharmonic effects on defect migration in UO2 and PuO2 using the potential of average force integration (PAFI) method. Fully anharmonic migration free energies are computed for several cation and anion defect types, using the Cooper-Rushton-Grimes (CRG) potential and a recently developed machine learning spectral neighbour analysis potential (SNAP) for UO2. Results are systematically compared to harmonic estimates based on attempt frequencies and the Debye approximation. Our results reveal that the validity of the harmonic approximation strongly depends on the defect type and the underlying potential, with significant deviations observed in several cases. In particular, defect migration barriers are found to decrease strongly with increasing temperature (up to 1 eV between 0 and 1200 K), and anharmonic contributions can substantially modify migration entropies and, consequently, diffusion coefficients. Comparing defect migration in UO2 and PuO2 using the CRG potential reveals that PuO2 has lower migration enthalpies at 0~K for all considered defects, but this is compensated by higher attempt frequencies, resulting in similar overall jump frequencies in UO2 and PuO2. These findings provide insight into the limitations of commonly used approximations and highlight the importance of anharmonic effects for predictive modeling of diffusion in nuclear fuels as well as in other classes of materials.

cond-mat.mtrl-sci↗

arXiv2025

Activation entropy of dislocation glide in body-centered cubic metals from atomistic simulations

The activation entropy of dislocation glide, a key process controlling the strength of many metals, is often assumed to be constant or linked to enthalpy through the empirical Meyer-Neldel law-both of which are simplified approximations. In this study, we take a more direct approach by calculating the activation Gibbs energy for kink-pair nucleation on screw dislocations of two body-centered cubic metals, iron and tungsten. To ensure reliability, we develop machine learning interatomic potentials for both metals, carefully trained on dislocation data from density functional theory. Our findings reveal that dislocations undergo harmonic transitions between Peierls valleys, with an activation entropy that remains largely constant, regardless of temperature or applied stress. We use these results to parameterize a thermally-activated model of yield stress, which consistently matches experimental data in both iron and tungsten. Our work challenges recent studies using classical potentials, which report highly varying activation entropies, and suggests that simulations relying on classical potentials-widely used in materials modeling-could be significantly influenced by overestimated entropic effects.

cond-mat.mtrl-sci↗

arXiv2025

Agnostic calculation of atomic free energies with the descriptor density of states

We present a new method to evaluate vibrational free energies of atomic systems without a priori specification of an interatomic potential. Our model-agnostic approach leverages descriptors, high-dimensional feature vectors of atomic structure. The entropy of a high-dimensional density, the descriptor density of states, is accurately estimated with conditional score matching. Casting interatomic potentials into a form extensive in descriptor features, we show free energies emerge as the Legendre-Fenchel conjugate of the descriptor entropy, avoiding all high-dimensional integration. The score matching campaign requires less resources than fixed-model sampling and is highly parallel, reducing wall time to a few minutes, with tensor compression schemes allowing lightweight storage. Our model-agnostic estimator returns differentiable free energy predictions over a broad range of potential parameters in microseconds of CPU effort, allowing rapid forward and back propagation of potential variations through finite temperature simulations, long desired for uncertainty quantification and inverse design. We test predictions against thermodynamic integration calculations over a broad range of models for BCC, FCC and A15 phases of W, Mo and Fe at high homologous temperatures. Predictions pass the stringent accuracy threshold of 1-2 meV/atom (1/40-1/20 kcal/mol) for phase prediction with propagated score uncertainties robustly bounding errors. We also demonstrate targeted fine-tuning, reducing the alpha-gamma transition temperature in a non-magnetic machine learning model of Fe from 2030 K to 1063 K through back-propagation, with no additional sampling. Applications to liquids and fine-tuning foundational models are discussed along with the many problems in computational science which estimate high-dimensional integrals.

cond-mat.mtrl-sci↗

arXiv2024

Calculation of crystal defects induced in CaWO$_{4}$ by 100 eV displacement cascades using a linear Machine Learning interatomic potential

We determine the energy stored in the crystal defects induced by $\mathcal{O}(10-100)$\,eV nuclear recoils in low-threshold CaWO$_{4}$ cryogenic detectors. A Machine Learning interatomic potential is developed to perform molecular dynamics simulations. We show that the energy spectra expected from Dark Matter and neutrino coherent scattering are affected by the crystal defects and we provide reference predictions. We discuss the special case of the spectrum of nuclear recoils induced by neutron capture, which could offer a unique sensitivity to the calculated stored energies.

physics.ins-det↗

arXiv2023

Neighbors Map: an Efficient Atomic Descriptor for Structural Analysis

Accurate structural analysis is essential to gain physical knowledge and understanding of atomic-scale processes in materials from atomistic simulations. However, traditional analysis methods often reach their limits when applied to crystalline systems with thermal fluctuations, defect-induced distortions, partial vitrification, etc. In order to enhance the means of structural analysis, we present a novel descriptor for encoding atomic environments into 2D images, based on a pixelated representation of graph-like architecture with weighted edge connections of neighboring atoms. This descriptor is well adapted for Convolutional Neural Networks and enables accurate structural analysis at a low computational cost. In this paper, we showcase a series of applications, including the classification of crystalline structures in distorted systems, tracking phase transformations up to the melting temperature, and analyzing liquid-to-amorphous transitions in pure metals and alloys. This work provides the foundation for robust and efficient structural analysis in materials science, opening up new possibilities for studying complex structural processes, which can not be described with traditional approaches.

cond-mat.mtrl-sci↗

arXiv2023

Robust crystal structure identification at extreme conditions using a density-independent spectral descriptor and supervised learning

The increased time- and length-scale of classical molecular dynamics simulations have led to raw data flows surpassing storage capacities, necessitating on-the-fly integration of structural analysis algorithms. As a result, algorithms must be computationally efficient, accurate, and stable at finite temperature to reliably extract the relevant features of the data at simulation time. In this work, we leverage spectral descriptors to encode local atomic environments and build crystal structure classification models. In addition to the classical way spectral descriptors are computed, i.e. over a fixed radius neighborhood sphere around a central atom, we propose an extension to make them independent from the material's density. Models are trained on defect-free crystal structures with moderate thermal noise and elastic deformation, using the linear discriminant analysis (LDA) method for dimensionality reduction and logistic regression (LR) for subsequent classification. The proposed classification model is intentionally designed to be simple, incorporating only a limited number of parameters. This deliberate simplicity enables the model to be trained effectively even when working with small databases. Despite the limited training data, the model still demonstrates inherent transferability, making it applicable to a broader range of scenarios and datasets. The accuracy of our models in extreme conditions is compared to traditional algorithms from the literature, namely adaptive common neighbor analysis (a-CNA), polyhedral template matching (PTM) and diamond structure identification (IDS). Finally, we showcase two applications of our method: tracking a solid-solid BCC-to-HCP phase transformation in Zirconium at high pressure up to high temperature, and visualizing stress-induced dislocation loop expansion in single crystal FCC Aluminum containing a Frank-Read source, at high temperature.

cond-mat.mtrl-sci↗

arXiv2021

Synergistic coupling in ab initio-machine learning simulations of dislocations

Ab initio simulations of dislocations are essential to build quantitative models of material strength, but the required system sizes are often at or beyond the limit of existing methods. Many important structures are thus missing in the training or validation of interatomic potentials, whilst studies of dislocation-defect interactions must mitigate the effect of strong periodic image interactions along the line direction. We show how these restrictions can be lifted through the use of linear machine learning potentials in hybrid simulations, where only a subset of atoms are governed by ab initio forces. The linear form is exploited in a constrained retraining procedure, qualitatively expanding the range of training structures for learning and giving precise matching of dislocation core structures, such that lines can cross the quantum/classical boundary. We apply our method to fully three dimensional studies of impurity segregation to edge and screw dislocations in tungsten. Our retrained potentials give systematically improved accuracy to QM/ML reference data and the three dimensional geometry allows for long-range relaxations that qualitatively change impurity-induced core reconstructions compared to simulations using short periodic supercells. More generally, the ability to treat arbitrary sub-regions of large scale simulations with ab initio accuracy opens a vast range of previously inaccessible extended defects to quantitative investigation.

cond-mat.mtrl-sci↗

arXiv2021

Quantum-accurate magneto-elastic predictions with classical spin-lattice dynamics

A data-driven framework is presented for building magneto-elastic machine-learning interatomic potentials (ML-IAPs) for large-scale spin-lattice dynamics simulations. The magneto-elastic ML-IAPs are constructed by coupling a collective atomic spin model with an ML-IAP. Together they represent a potential energy surface from which the mechanical forces on the atoms and the precession dynamics of the atomic spins are computed. Both the atomic spin model and the ML-IAP are parametrized on data from first-principles calculations. We demonstrate the efficacy of our data-driven framework across magneto-structural phase transitions by generating a magneto-elastic ML-IAP for α-iron. The combined potential energy surface yields excellent agreement with first-principles magneto-elastic calculations and quantitative predictions of diverse materials properties including bulk modulus, magnetization, and specific heat across the ferromagnetic-paramagnetic phase transition.

cond-mat.mtrl-sci↗

arXiv2020

Solute diffusion by self-interstitial defects and radiation-induced segregation in ferritic Fe-X (X=Cr, Cu, Mn, Ni, P, Si) dilute alloys

This work investigates solute transport due to self-interstitial defects and radiation induced segregation tendencies in dilute ferritic alloys, by computing the transport coefficients of each system based on ab initio calculations of binding energies and migration rates. The implementation of the self-consistent mean field method in the KineCluE code allows to extend the calculation of transport coefficients to arbitrary interaction ranges, crystal structures, and diffusion mechanisms. The results show that the diffusivity of P, Mn, and Cr solute atoms is dominated by the dumbbell mechanism, that of Cu by vacancies, while the two mechanisms might be in competition for Ni and Si, despite the fact that the corresponding mixed dumbbells are not stable. Systematic enrichment at defect sinks is expected for P and Mn solutes due to dumbbell diffusion, and for Si due mainly to vacancy drag. Vacancy drag is also responsible for Cu and Ni enrichment below 1085 K. The RIS behavior of Cr is the outcome of a fine balance between dumbbell enrichment and vacancy depletion. Therefore, for dilute Cr concentrations global enrichment occurs below 540 K, and depletion above. This threshold temperature grows with solute concentration. The findings are in agreement with experimental observations of RIS and clustering phenomena, and confirm that solute-defect kinetic coupling plays an important role in the formation of solute clusters in reactor pressure vessel steels and other alloys.

cond-mat.mtrl-sci↗

arXiv2018

Relaxation volumes of microscopic and mesoscopic irradiation-induced defects in tungsten

The low energy structures of irradiation-induced defects have been studied in detail, as these determine the available modes by which a defect can diffuse or relax. As a result, there are many studies concerning the relative energies of possible defect structures, and empirical potentials are commonly fitted to or evaluated with respect to these energies. But recently [Dudarev et al Nuclear Fusion 2018], we have shown how to determine the stresses, strains and swelling of reactor components under irradiation from the elastic properties of ensembles of irradiation-induced defects. These elastic properties have received comparatively little attention. Here we evaluate relaxation volumes of irradiation-induced defects in tungsten computed with empirical potentials, and compare to density functional theory results where available. Different empirical potentials give different results, but some potential-independent trends in relaxation volumes can be identified. We show that the relaxation volume of small defects can be predicted to within 10% from their point-defect count. For larger defects we provide empirical fits for the relaxation volume of as a function of size. We demonstrate that the relaxation volume associated with a single primary-damage cascade can be estimated from the primary knock-on atom (PKA) energy. We conclude that while annihilation of vacancy- and interstitial- character defects will invariably reduce the total relaxation volume of the cascade debris, empirical potentials disagree whether coalescence of defects will reduce or increase the total relaxation volume.

cond-mat.mtrl-sci↗

arXiv2018

Smart energy models for atomistic simulations using a DFT-driven multifidelity approach

The reliability of atomistic simulations depends on the quality of the underlying energy models providing the source of physical information, for instance for the calculation of migration barriers in atomistic Kinetic Monte Carlo simulations. Accurate (high-fidelity) methods are often available, but since they are usually computationally expensive, they must be replaced by less accurate (low-fidelity) models that introduce some degrees of approximation. Machine-learning techniques such as artificial neural networks are usually employed to work around this limitation and extract the needed parameters from large databases of high-fidelity data, but the latter are often computationally expensive to produce. This work introduces an alternative method based on the multifidelity approach, where correlations between high-fidelity and low-fidelity outputs are exploited to make an educated guess of the high-fidelity outcome based only on quick low-fidelity estimations, hence without the need of running full expensive high-fidelity calculations. With respect to neural networks, this approach is expected to require less training data because of the lower amount of fitting parameters involved. The method is tested on the prediction of ab initio formation and migration energies of vacancy diffusion in iron-copper alloys, and compared with the neural networks trained on the same database.

physics.comp-ph↗

arXiv2015

Using Bayes formula to estimate rates of rare events in transition path sampling simulations

Transition path sampling is a method for estimating the rates of rare events in molecular systems based on the gradual transformation of a path distribution containing a small fraction of reactive trajectories into a biased distribution in which these rare trajectories have become frequent. Then, a multistate reweighting scheme is implemented to postprocess data collected from the staged simulations. Herein, we show how Bayes formula allows to directly construct a biased sample containing an enhanced fraction of reactive trajectories and to concomitantly estimate the transition rate from this sample. The approach can remediate the convergence issues encountered in free energy perturbation or umbrella sampling simulations when the transformed distribution insufficiently overlaps with the reference distribution.

cond-mat.stat-mech↗

arXiv2014

Free energy calculations from adaptive molecular dynamics simulations with adiabatic reweighting

We propose an adiabatic reweighting algorithm for computing the free energy along an external parameter from adaptive molecular dynamics simulations. The adaptive bias is estimated using Bayes identity and information from all the sampled configurations. We apply the algorithm to a structural transition in a cluster and to the migration of a crystalline defect along a reaction coordinate. Compared to standard adaptive molecular dynamics, we observe an acceleration of convergence. With the aid of the algorithm, it is also possible to iteratively construct the free energy along the reaction coordinate without having to differentiate the gradient of the reaction coordinate or any biasing potential.

cond-mat.stat-mech↗

arXiv2013

Point defect modeling in materials: coupling ab initio and elasticity approaches

Modeling point defects at an atomic scale requires careful treatment of the long-range atomic relaxations. This elastic field can strongly affect point defect properties calculated in atomistic simulations because of the finite size of the system under study. This is an important restriction for ab initio methods which are limited to a few hundred atoms. We propose an original approach coupling ab initio calculations and linear elasticity theory to obtain the properties of an isolated point defect for reduced supercell sizes. The reliability and benefit of our approach are demonstrated for three problematic cases: the self-interstitial in zirconium, clusters of self-interstitials in iron, and the neutral vacancy in silicon.

cond-mat.mtrl-sci↗