arXiv subjects

Albert P. Bartók

Publications and source records attributed to Albert P. Bartók.

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

Kink-Helium Interactions in Tungsten: Increased Dislocation Mobility in the Infinitely Dilute Regime

Point defects such as interstitial atoms are known to be attracted to screw dislocations. Understanding these interaction mechanisms is key to predicting the plasticity of real materials. Using a new machine learning interatomic potential derived from ab initio calculations of helium in tungsten, we investigate the binding of small helium clusters ($He_{n}$, ${n}=1-3$) to screw dislocation kinks. We find that helium binds significantly more strongly to kinks than to straight dislocation segments. For a single helium atom, the preference reduces the kink pair nucleation energy from 1.58 eV in pure tungsten to 0.48 eV when helium binds to the vacancy kink, indicative of increased dislocation mobility (material softening). In the case of ${n}=2$, kink binding stabilises the kink pair configuration as the ground state, while the straight dislocation is metastable; the two are separated by a 0.60 eV barrier that is the rate determining step for dislocation motion. The energy difference between the ground state kink pair and the metastable straight dislocation increases for ${n}=3$, connected by a 1.00 eV barrier, indicating a progressive loss of the softening effect with increasing cluster size. Molecular dynamics simulations at 900 K support the proposed existence of helium-induced softening in this extremely dilute regime.

cond-mat.mtrl-sci↗

arXiv2026

Uncertainty quantification design principles for machine learning interatomic potentials: lessons learned from hierarchical Bayesian inference

Reliable uncertainty quantification (UQ) remains a major challenge for machine learning interatomic potentials (MLIPs). Here, we benchmark UQ strategies for message passing neural networks such as MACE and Gaussian process (GP)-based interatomic potentials using coupled-cluster reference data for argon as a test case. We develop a hierarchical GP potential combining physics-informed priors with Bayesian propagation of hyperparameter uncertainty that retains useful discrimination of prediction errors while remaining conservative rather than overconfident, with a probabilistic structure that allows the sources of predictive uncertainty to be explicitly dissected. These results expose important distinctions between accuracy, calibration, sharpness, and error discrimination and provide practical design principles for MLIP UQ in uncertainty-critical applications.

cond-mat.mtrl-sci↗

arXiv2026

Data-Driven Equation Discovery for Nonlinear Liquid Film Flows

Over the past decade data-driven equation discovery emerged as a powerful alternative to first principles-based methodologies traditionally used in mathematical modelling cycles. The approach provides a promising path towards deep, physical insight into systems that have previously evaded rigorous mathematical derivation procedures, often due to intractable complexity. The strengths of such techniques have been successfully established for many problem classes described by systems of ordinary differential equations and continue to be extended, with their reach into partial differential equation systems gaining momentum, though comparatively nascent. Herein we tackle such a frontier: elucidating the dynamics of liquid film flows, a problem space providing a rich backdrop in terms of asymptotic analytical building blocks. By leveraging expert knowledge and the ability to carefully curate data, we establish a best-case scenario for identifying the underlying governing equations. Even here, multi-collinearity, stemming from the choice of monomial basis functions in our multi-scale flow configuration, introduces complex identifiability issues. Early-time transients compound this further, as the most dynamically rich behaviour carries the largest residuals in training data. Pinpointing such vulnerabilities allows us to better define the boundaries of current discovery techniques and paves the way for the next generation of more resilient, numerically stable algorithms.

physics.flu-dyn↗

arXiv2026

Vibrational infrared and Raman spectra of the methanol molecule with equivariant neural-network property surfaces

Electric dipole and polarizability surfaces are developed for the methanol (CH$_3$OH) molecule using ab initio electronic structure data, computed at the CCSD/aug-cc-pVTZ level of theory, and equivariant neural networks. These property surfaces are used to compute vibrational infrared and Raman intensities with variational vibrational energies and wave functions. The energies and wave functions, fully accounting for the large-amplitude motion and tunneling splitting states, are from continued variational vibrational computations, based on earlier work [Sunaga et al., J. Chem. Phys., 2025, 163, 064101], up to 3700 cm$^{-1}$ beyond the zero-point vibration, now reaching the O-H stretching fundamental. All vibrational fundamentals, combination and overtone bands are in excellent agreement with available (gas-phase) experimental data, with a 2.2 cm$^{-1}$ root-mean-squared deviation of the fundamentals from experiment. These developments constitute an important step towards a quantitative and comprehensive exact quantum dynamics model of the methanol molecule, and a linelist for astrophysical applications.

physics.chem-ph↗

arXiv2026

Regularity Priors for the Linear Atomic Cluster Expansion

Machine-learned interatomic potentials enable large systems to be simulated for long time scales at near ab-initio accuracy. This accuracy is achieved by fitting extremely flexible model architectures to high quality reference data. In practice, this flexibility can cause unwanted behavior such as jagged predicted potential energy surfaces and generally poor out-of-distribution behavior. We investigate a general strategy for incorporating prior beliefs on the regularity of the target energy into linear Atomic Cluster Expansion (ACE) models and explore to what extent this approach improves the quality of the fitted models. Our main focus is an over-regularisation that replicates the Gaussian broadening used in Smooth Overlap of Atomic Positions (SOAP) descriptors within the ACE framework. Numerical tests indicate that the exact form of the prior is non-critical but that including such a prior leads to significant improvement in test errors, consistent repulsion at close-approach, eliminates spurious false minima in the potential energy and enhances stability during molecular dynamics simulations.

physics.chem-ph↗

arXiv2025

Autonomous thermodynamically informed database generation for machine-learned interatomic potentials and application to magnesium

We propose a novel approach for constructing training databases for Machine-Learned Interatomic Potential (MLIP) models, specifically designed to capture phase properties across a wide range of conditions. The framework is uniquely appealing due to its ease of automation, its suitability for iterative learning, and its independence from prior knowledge of stable phases, avoiding bias towards pre-existing structural data. The approach uses Nested Sampling (NS) to explore the configuration space and generate thermodynamically relevant configurations, forming the database which undergoes ab initio Density Functional Theory (DFT) evaluation. We use the Atomic Cluster Expansion (ACE) architecture to fit a model on the resulting database. To demonstrate the efficiency of the framework, we apply it to magnesium, developing a model capable of accurately describing behaviour across pressure and temperature ranges of 0-600 GPa and 0-8000 K, respectively. We benchmark the model's performance by calculating phonon spectra and elastic constants, as well as the pressure-temperature phase diagram within this region. The results showcase the power of the framework to produce robust MLIPs while maintaining transferability and generality, for reduced computational cost.

cond-mat.mtrl-sci↗

arXiv2025

Fine-tuning foundation models of materials interatomic potentials with frozen transfer learning

Machine-learned interatomic potentials are revolutionising atomistic materials simulations by providing accurate and scalable predictions within the scope covered by the training data. However, generation of an accurate and robust training data set remains a challenge, often requiring thousands of first-principles calculations to achieve high accuracy. Foundation models have started to emerge with the ambition to create universally applicable potentials across a wide range of materials. While foundation models can be robust and transferable, they do not yet achieve the accuracy required to predict reaction barriers, phase transitions, and material stability. This work demonstrates that foundation model potentials can reach chemical accuracy when fine-tuned using transfer learning with partially frozen weights and biases. For two challenging datasets on reactive chemistry at surfaces and stability and elastic properties of tertiary alloys, we show that frozen transfer learning with 10-20% of the data (hundreds of datapoints) achieves similar accuracies to models trained from scratch (on thousands of datapoints). Moreover, we show that an equally accurate, but significantly more efficient surrogate model can be built using the transfer learned potential as the ground truth. In combination, we present a simulation workflow for machine learning potentials that improves data efficiency and computational efficiency.

cond-mat.mtrl-sci↗

arXiv2025

Multi-Phase Dataset for Ti and Ti-6Al-4V

Titanium and its alloys are technologically important materials that display a rich phase behaviour. In order to enable large-scale, realistic modelling of Ti and its alloys on the atomistic scale, Machine Learning Interatomic Potentials (MLIPs) are crucial, but rely on databases of atomic configurations. We report databases of such configurations that represent the α, \b{eta}, ω and liquid phases of Ti and the Ti-6Al-4V alloy, where we provide total energy, force and stress values evaluated by Density Functional Theory (DFT) using the PBE exchange-correlation functional. We have also leveraged and extended a data reduction strategy, via non-diagonal supercells, for the vibrational properties of Ti and sampling of atomic species within bulk crystalline data for Ti-6Al-4V. These configurations may be used to fit MLIP models that can accurately model the phase behaviour of Ti and Ti-6Al-4V across a broad range of thermodynamic conditions. To validate models, we assembled a set of benchmark protocols, which can be used to rapidly develop and evaluate MLIP models. We demonstrated the utility of our databases and validation tools by fitting models based on the Gaussian Approximation Potential (GAP) and Atomic Cluster Expansion (ACE) frameworks.

cond-mat.mtrl-sci↗

arXiv2024

Experiment-driven atomistic materials modeling: A case study combining X-ray photoelectron spectroscopy and machine learning potentials to infer the structure of oxygen-rich amorphous carbon

An important yet challenging aspect of atomistic materials modeling is reconciling experimental and computational results. Conventional approaches involve generating numerous configurations through molecular dynamics or Monte Carlo structure optimization and selecting the one with the closest match to experiment. However, this inefficient process is not guaranteed to succeed. We introduce a general method to combine atomistic machine learning (ML) with experimental observables that produces atomistic structures compatible with experiment by design. We use this approach in combination with grand-canonical Monte Carlo within a modified Hamiltonian formalism, to generate configurations that agree with experimental data and are chemically sound (low in energy). We apply our approach to understand the atomistic structure of oxygenated amorphous carbon (a-CO$_{x}$), an intriguing carbon-based material, to answer the question of how much oxygen can be added to carbon before it fully decomposes into CO and CO$_2$. Utilizing an ML-based X-ray photoelectron spectroscopy (XPS) model trained from $GW$ and density functional theory (DFT) data, in conjunction with an ML interatomic potential, we identify a-CO$_{x}$ structures compliant with experimental XPS predictions that are also energetically favorable with respect to DFT. Employing a network analysis, we accurately deconvolve the XPS spectrum into motif contributions, both revealing the inaccuracies inherent to experimental XPS interpretation and granting us atomistic insight into the structure of a-CO$_{x}$. This method generalizes to multiple experimental observables and allows for the elucidation of the atomistic structure of materials directly from experimental data, thereby enabling experiment-driven materials modeling with a degree of realism previously out of reach.

cond-mat.mtrl-sci↗

arXiv2024

A collinear-spin machine learned interatomic potential for Fe$_{7}$Cr$_{2}$Ni alloy

We have developed a new machine learned interatomic potential for the prototypical austenitic steel Fe$_{7}$Cr$_{2}$Ni, using the Gaussian approximation potential (GAP) framework. This new GAP can model the alloy's properties with close to density functional theory (DFT) accuracy, while at the same time allowing us to access larger length and time scales than expensive first-principles methods. We also extended the GAP input descriptors to approximate the effects of collinear spins (Spin GAP), and demonstrate how this extended model successfully predicts structural distortions due to antiferromagnetic and paramagnetic spin states. We demonstrate the application of the Spin GAP model for bulk properties and vacancies and validate against DFT. These results are a step towards modelling the atomistic origins of ageing in austenitic steels with higher accuracy.

cond-mat.mtrl-sci↗

arXiv2023

Gaussian Approximation Potentials: theory, software implementation and application examples

Gaussian Approximation Potentials are a class of Machine Learned Interatomic Potentials routinely used to model materials and molecular systems on the atomic scale. The software implementation provides the means for both fitting models using ab initio data and using the resulting potentials in atomic simulations. Details of the GAP theory, algorithms and software are presented, together with detailed usage examples to help new and existing users. We review some recent developments to the GAP framework, including MPI parallelisation of the fitting code enabling its use on thousands of CPU cores and compression of descriptors to eliminate the poor scaling with the number of different chemical elements.

cond-mat.mtrl-sci↗

arXiv2022

Massively Parallel Fitting of Gaussian Approximation Potentials

We present a data-parallel software package for fitting Gaussian Approximation Potentials (GAPs) on multiple nodes using the ScaLAPACK library with MPI and OpenMP. Until now the maximum training set size for GAP models has been limited by the available memory on a single compute node. In our new implementation, descriptor evaluation is carried out in parallel with no communication requirement. The subsequent linear solve required to determine the model coefficients is parallelised with ScaLAPACK. Our approach scales to thousands of cores, lifting the memory limitation and also delivering substantial speedups. This development expands the applicability of the GAP approach to more complex systems as well as opening up opportunities for efficiently embedding GAP model fitting within higher-level workflows such as committee models or hyperparameter optimisation.

cond-mat.mtrl-sci↗

arXiv2022

Optimal data generation for machine learned interatomic potentials

Machine learning interatomic potentials (MLIPs) are routinely used atomic simulations, but generating databases of atomic configurations used in fitting these models is a laborious process, requiring significant computational and human effort. A computationally efficient method is presented to generate databases of atomic configurations that contain optimal information on the small-displacement regime of the potential energy surface of bulk crystalline matter. Utilising non-diagonal supercell (NDSC), an automatic process is suggested for ab initio data generation. MLIPs were fitted for Al, W, Mg and Si, which very closely reproduce the ab initio phonon and elastic properties. The protocol can be easily adapted to other materials and can be inserted in the workflow of any flavour of MLIP generation.

cond-mat.mtrl-sci↗

arXiv2022

Comment on "Manifolds of quasi-constant SOAP and ACSF fingerprints and the resulting failure to machine learn four body interactions"

The "quasi-constant" SOAP and ACSF fingerprint manifolds recently discovered by Parsaeifard and Goedecker are a direct consequence of the presence of degenerate pairs of configurations, a known shortcoming of all low-body-order atom-density correlation representations of molecular structures. Contrary to the configurations that are rigorously singular -- that we demonstrate can only occur in finite, discrete sets -- the continuous "quasi-constant" manifolds exhibit low, but non-zero, sensitivity to atomic displacements. Thus, it is possible to build interpolative machine-learning models of high-order interactions along the manifold, even though the numerical instabilities associated with proximity to the exact singularities affect the accuracy and transferability of such models, to an extent that depends on numerical details of the implementation.

physics.chem-ph↗

arXiv2021

Machine learning force fields based on local parametrization of dispersion interactions: Application to the phase diagram of C$_{60}$

We present a comprehensive methodology to enable addition of van der Waals (vdW) corrections to machine learning (ML) atomistic force fields. Using a Gaussian approximation potential (GAP) [Bartók et al., Phys. Rev. Lett. 104, 136403 (2010)] as baseline, we accurately machine learn a local model of atomic polarizabilities based on Hirshfeld volume partitioning of the charge density [Tkatchenko and Scheffler, Phys. Rev. Lett. 102, 073005 (2009)]. These environment-dependent polarizabilities are then used to parametrize a screened London-dispersion approximation to the vdW interactions. Our ML vdW model only needs to learn the charge density partitioning implicitly, by learning the reference Hirshfeld volumes from density functional theory (DFT). In practice, we can predict accurate Hirshfeld volumes from the knowledge of the local atomic environment (atomic positions) alone, making the model highly computationally efficient. For additional efficiency, our ML model of atomic polarizabilities reuses the same many-body atomic descriptors used for the underlying GAP learning of bonded interatomic interactions. We also show how the method enables straightforward computation of gradients of the observables, even when these remain challenging for the reference method (e.g., calculating gradients of the Hirshfeld volumes in DFT). Finally, we demonstrate the approach by studying the phase diagram of C$_{60}$, where vdW effects are important. The need for a highly accurate vdW-inclusive reactive force field is highlighted by modeling the decomposition of the C$_{60}$ molecules taking place at high pressures and temperatures.

cond-mat.mtrl-sci↗

arXiv2021

Physics-inspired structural representations for molecules and materials

The first step in the construction of a regression model or a data-driven analysis, aiming to predict or elucidate the relationship between the atomic scale structure of matter and its properties, involves transforming the Cartesian coordinates of the atoms into a suitable representation. The development of atomic-scale representations has played, and continues to play, a central role in the success of machine-learning methods for chemistry and materials science. This review summarizes the current understanding of the nature and characteristics of the most commonly used structural and chemical descriptions of atomistic structures, highlighting the deep underlying connections between different frameworks, and the ideas that lead to computationally efficient and universally applicable models. It emphasizes the link between properties, structures, their physical chemistry and their mathematical description, provides examples of recent applications to a diverse set of chemical and materials science problems, and outlines the open questions and the most promising research directions in the field.

physics.chem-ph↗

arXiv2021

Insight into liquid polymorphism from the complex phase behaviour of a simple model

We systematically explored the phase behavior of the hard-core two-scale ramp model suggested by Jagla[E. A. Jagla, Phys. Rev. E 63, 061501 (2001)] using a combination of the nested sampling and free energy methods. The sampling revealed that the phase diagram of the Jagla potential is significantly richer than previously anticipated, and we identified a family of new crystalline structures, which is stable over vast regions in the phase diagram. We showed that the new melting line is located at considerably higher temperature than the boundary between the low- and high-density liquid phases, which was previously suggested to lie in a thermodynamically stable region. The newly identified crystalline phases show unexpectedly complex structural features, some of which are shared with the high-pressure ice VI phase.

cond-mat.mtrl-sci↗