Searcharxiv⌕ Search

arXiv subjects

Gábor Csányi

Publications and source records attributed to Gábor Csányi.

At least 37 records · Page 2Linked to original sources
arXiv2025

A foundation model for atomistic materials chemistry

Atomistic simulations of matter, especially those that leverage first-principles (ab initio) electronic structure theory, provide a microscopic view of the world, underpinning much of our understanding of chemistry and materials science. Over the last decade or so, machine-learned force fields have transformed atomistic modeling by enabling simulations of ab initio quality over unprecedented time and length scales. However, early ML force fields have largely been limited by: (i) the substantial computational and human effort of developing and validating potentials for each particular system of interest; and (ii) a general lack of transferability from one chemical system to the next. Here we show that it is possible to create a general-purpose atomistic ML model, trained on a public dataset of moderate size, that is capable of running stable molecular dynamics for a wide range of 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 of problems in the physical sciences, including 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 as a starting or "foundation" model for any atomistic system of interest and, when desired, can be fine-tuned on just a handful of application-specific data points to reach ab initio accuracy. Establishing that a stable force-field model can cover almost all materials changes atomistic modeling in a fundamental way: experienced users get reliable results much faster, and beginners face a lower barrier to entry. Foundation models thus represent a step towards democratising the revolution in atomic-scale modeling that has been brought about by ML force fields.

physics.chem-ph↗
arXiv2025

MACE-OFF: Transferable Short Range Machine Learning Force Fields for Organic Molecules

Classical empirical force fields have dominated biomolecular simulation for over 50 years. Although widely used in drug discovery, crystal structure prediction, and biomolecular dynamics, they generally lack the accuracy and transferability required for first-principles predictive modeling. In this paper, we introduce MACE-OFF, a series of short range transferable force fields for organic molecules created using state-of-the-art machine learning technology and first-principles reference data computed with a high level of quantum mechanical theory. MACE-OFF demonstrates the remarkable capabilities of short range models by accurately predicting a wide variety of gas and condensed phase properties of molecular systems. It produces accurate, easy-to-converge dihedral torsion scans of unseen molecules, as well as reliable descriptions of molecular crystals and liquids, including quantum nuclear effects. We further demonstrate the capabilities of MACE-OFF by determining free energy surfaces in explicit solvent, as well as the folding dynamics of peptides.Finally, we simulate a fully solvated small protein, observing accurate secondary structure and vibrational spectrum. These developments enable first-principles simulations of molecular systems for the broader chemistry community at high accuracy and relatively low computational cost.

physics.chem-ph↗
arXiv2025

Towards Routine Condensed Phase Simulations with Delta-Learned Coupled Cluster Accuracy: Application to Liquid Water

Simulating liquid water to an accuracy that matches its wealth of available experimental data requires both precise electronic structure methods and reliable sampling of nuclear (quantum) motion. This is challenging because applying the electronic structure method of choice - coupled cluster theory with single, double and perturbative triple excitations [CCSD(T)] - to condensed phase systems is currently limited by its computational cost and complexity. Recent tour-de-force efforts have demonstrated that this accuracy can indeed bring simulated liquid water into close agreement with experiment using machine learning potentials (MLPs). However, achieving this remains far from routine, requiring large datasets and significant computational cost. In this work, we introduce a practical approach that combines developments in MLPs with local correlation approximations to enable routine CCSD(T)-level simulations of liquid water. When combined with nuclear quantum effects, we achieve agreement to experiments for structural and transport properties. Crucially, this approach extends beyond constant volume to constant pressure simulations, allowing fundamental properties such as the density to now be predicted by MLP-based CCSD(T) models. Importantly, the approach also handles constant pressure simulations, enabling MLP-based CCSD(T) models to predict isothermal-isobaric bulk properties, such as water's density maximum in close agreement with experiment. Encompassing tests across electronic structure, datasets and MLP architecture, this work provides a practical blueprint towards routinely developing CCSD(T)-based MLPs for the condensed phase.

physics.chem-ph↗
arXiv2025

Thermal Conductivity Predictions with Foundation Atomistic Models

Advances in machine learning have led to the development of foundation models for atomistic materials chemistry, enabling quantum-accurate descriptions of interatomic forces across chemically diverse compounds at reduced computational cost. Hitherto, the accuracy and utility of these models have been assessed relying on descriptors based on formation energies or idealized harmonic atomic vibrations. Yet, the rigorous and physically interpretable quantification of their capability to describe both realistic anharmonic atomic dynamics and technologically relevant observables remains a pressing problem. Here, we address this problem, leveraging the Wigner formulation of heat transport and the Grüneisen approach to thermal expansion to connect the atomic-physics awareness of foundation models to their utility in predicting experimentally observable thermomechanical properties, presenting standards and fine-tuning protocols needed to achieve first-principles accuracy. We apply our framework to a database of 103 solids with diverse compositions and structures, demonstrating that it overcomes the major bottlenecks of current methods for designing heat-management materials -- high cost, limited transferability, or lack of physics awareness -- and its potential to discover materials for next-gen technologies ranging from thermal insulation to neuromorphic computing.

cond-mat.mtrl-sci↗
arXiv2025

Resonances in reflective Hamiltonian Monte Carlo

In high dimensions, reflective Hamiltonian Monte Carlo with inexact reflections exhibits slow mixing when the particle ensemble is initialised from a Dirac delta distribution and the uniform distribution is targeted. By quantifying the instantaneous non-uniformity of the distribution with the Sinkhorn divergence, we elucidate the principal mechanisms underlying the mixing problems. In spheres and cubes, we show that the collective motion transitions between fluid-like and discretisation-dominated behaviour, with the critical step size scaling as a power law in the dimension. In both regimes, the particles can spontaneously unmix, leading to resonances in the particle density and the aforementioned problems. Additionally, low-dimensional toy models of the dynamics are constructed which reproduce the dominant features of the high-dimensional problem. Finally, the dynamics is contrasted with the exact Hamiltonian particle flow and tuning practices are discussed.

stat.ML↗
arXiv2024

Random Spin Committee Approach For Smooth Interatomic Potentials

Training interatomic potentials for spin-polarized systems continues to be a difficult task for the molecular modeling community. In this note, a proof-of-concept, random initial spin committee approach is proposed for obtaining the ground state of spin-polarized systems with a controllable degree of accuracy. The approach is tested on two toy models of elemental sulfur where the exact optimal spin configuration can be known. Machine-learning potentials are trained on the resulting data, and increasingly accurate fits with respect to the ground state are achieved, marking a step towards machine-learning force fields for general bulk spin-polarized systems.

physics.chem-ph↗
arXiv2024

Symmetry Breaking in the Superionic Phase of Silver-Iodide

In the superionic phase of silver iodide, we observe a distorted tetragonal structure characterized by symmetry breaking in the cation distribution. This phase competes with the well known bcc phase, with a symmetric cation distribution at an energetic cost of only a few meV/atom. The small energy difference suggests that these competing structures may both be thermally accessible near the superionic transition temperature. We also find that the distribution of silver ions depends on the low-temperature parent polymorph, with memory persisting in the superionic phase on the nanosecond time scales accessible in our simulations. Furthermore, simulations on the order 100ns reveal that even at temperatures where the bcc phase is stable, significant fluctuations toward the tetragonal lattice structure remain. Our results are consistent with many "anomalous" experimental observations and offer a molecular mechanism for the "memory effect" in silver iodide.

cond-mat.mtrl-sci↗
arXiv2024

Transferability of datasets between Machine-Learning Interaction Potentials

With the emergence of Foundational Machine Learning Interatomic Potential (FMLIP) models trained on extensive datasets, transferring data between different ML architectures has become increasingly important. In this work, we examine the extent to which training data optimised for one machine-learning forcefield algorithm may be re-used to train different models, aiming to accelerate FMLIP fine-tuning and to reduce the need for costly iterative training. As a test case, we train models of an organic liquid mixture that is commonly used as a solvent in rechargeable battery electrolytes, making it an important target for reactive MLIP development. We assess model performance by analysing the properties of molecular dynamics trajectories, showing that this is a more stringent test than comparing prediction errors for fixed datasets. We consider several types of training data, and several popular MLIPs - notably the recent MACE architecture, a message-passing neural network designed for high efficiency and smoothness. We demonstrate that simple training sets constructed without any ab initio dynamics are sufficient to produce stable models of molecular liquids. For simple neural-network architectures, further iterative training is required to capture thermodynamic and kinetic properties correctly, but MACE performs well with extremely limited datsets. We find that configurations designed by human intuition to correct systematic model deficiencies transfer effectively between algorithms, but active-learned data that are generated by one MLIP do not typically benefit a different algorithm. Finally, we show that any training data which improve model performance also improve its ability to generalise to similar unseen molecules. This suggests that trajectory failure modes are connected with chemical structure rather than being entirely system-specific.

physics.chem-ph↗
arXiv2024

Efficient Composite Infrared Spectroscopy: Combining the Doubly-Harmonic Approximation with Machine Learning Potentials

Vibrational spectroscopy is a cornerstone technique for molecular characterization and offers an ideal target for the computational investigation of molecular materials. Building on previous comprehensive assessments of efficient methods for infrared (IR) spectroscopy, this study investigates the predictive accuracy and computational efficiency of gas-phase IR spectra calculations, accessible through a combination of modern semiempirical quantum mechanical and transferable machine learning potentials. A composite approach for IR spectra prediction based on the doubly-harmonic approximation, utilizing harmonic vibrational frequencies in combination squared derivatives of the molecular dipole moment, is employed. This approach allows for methodical flexibility in the calculation of IR intensities from molecular dipoles and the corresponding vibrational modes. Various methods are systematically tested to suggest a suitable protocol with an emphasis on computational efficiency. Among these methods, semiempirical extended tight-binding (xTB) models, classical charge equilibrium models, and machine learning potentials trained for dipole moment prediction are assessed across a diverse dataset of organic molecules. We particularly focus on the recently reported machine learning potential MACE-OFF23 to address the accuracy limitations of conventional low-cost quantum mechanical and force-field methods. This study aims to establish a standard for the efficient computational prediction of IR spectra, facilitating the rapid and reliable identification of unknown compounds and advancing automated analytical workflows in chemistry.

physics.chem-ph↗
arXiv2024

Self-consistent Coulomb interactions for machine learning interatomic potentials

A ubiquitous approach to obtain transferable machine learning-based models of potential energy surfaces for atomistic systems is to decompose the total energy into a sum of local atom-centred contributions. However, in many systems non-negligible long-range electrostatic effects must be taken into account as well. We introduce a general mathematical framework to study how such long-range effects can be included in a way that (i) allows charge equilibration and (ii) retains the locality of the learnable atom-centred contributions to ensure transferability. Our results give partial explanations for the success of existing machine learned potentials that include equilibriation and provide perspectives how to design such schemes in a systematic way. To complement the rigorous theoretical results, we describe a practical scheme for fitting the energy and electron density of water clusters.

physics.comp-ph↗
arXiv2024

Data-efficient fine-tuning of foundational models for first-principles quality sublimation enthalpies

Calculating sublimation enthalpies of molecular crystal polymorphs is relevant to a wide range of technological applications. However, predicting these quantities at first-principles accuracy -- even with the aid of machine learning potentials -- is a challenge that requires sub-kJ/mol accuracy in the potential energy surface and finite-temperature sampling. We present an accurate and data-efficient protocol based on fine-tuning of the foundational MACE-MP-0 model and showcase its capabilities on sublimation enthalpies and physical properties of ice polymorphs. Our approach requires only a few tens of training structures to achieve sub-kJ/mol accuracy in the sublimation enthalpies and sub 1 % error in densities for polymorphs at finite temperature and pressure. Exploiting this data efficiency, we explore simulations of hexagonal ice at the random phase approximation level of theory at experimental temperatures and pressures, calculating its physical properties, like pair correlation function and density, with good agreement with experiments. Our approach provides a way forward for predicting the stability of molecular crystals at finite thermodynamic conditions with the accuracy of correlated electronic structure theory.

cond-mat.mtrl-sci↗
arXiv2024

Benchmarking of machine learning interatomic potentials for reactive hydrogen dynamics at metal surfaces

Simulations of chemical reaction probabilities in gas surface dynamics require the calculation of ensemble averages over many tens of thousands of reaction events to predict dynamical observables that can be compared to experiments. At the same time, the energy landscapes need to be accurately mapped, as small errors in barriers can lead to large deviations in reaction probabilities. This brings a particularly interesting challenge for machine learning interatomic potentials, which are becoming well-established tools to accelerate molecular dynamics simulations. We compare state-of-the-art machine learning interatomic potentials with a particular focus on their inference performance on CPUs and suitability for high throughput simulation of reactive chemistry at surfaces. The considered models include polarizable atom interaction neural networks (PaiNN), recursively embedded atom neural networks (REANN), the MACE equivariant graph neural network, and atomic cluster expansion potentials (ACE). The models are applied to a dataset on reactive molecular hydrogen scattering on low-index surface facets of copper. All models are assessed for their accuracy, time-to-solution, and ability to simulate reactive sticking probabilities as a function of the rovibrational initial state and kinetic incidence energy of the molecule. REANN and MACE models provide the best balance between accuracy and time-to-solution and can be considered the current state-of-the-art in gas-surface dynamics. PaiNN models require many features for the best accuracy, which causes significant losses in computational efficiency. ACE models provide the fastest time-to-solution, however, models trained on the existing dataset were not able to achieve sufficiently accurate predictions in all cases.

physics.chem-ph↗
arXiv2024

Energy-conserving equivariant GNN for elasticity of lattice architected metamaterials

Lattices are architected metamaterials whose properties strongly depend on their geometrical design. The analogy between lattices and graphs enables the use of graph neural networks (GNNs) as a faster surrogate model compared to traditional methods such as finite element modelling. In this work, we generate a big dataset of structure-property relationships for strut-based lattices. The dataset is made available to the community which can fuel the development of methods anchored in physical principles for the fitting of fourth-order tensors. In addition, we present a higher-order GNN model trained on this dataset. The key features of the model are (i) SE(3) equivariance, and (ii) consistency with the thermodynamic law of conservation of energy. We compare the model to non-equivariant models based on a number of error metrics and demonstrate its benefits in terms of predictive performance and reduced training requirements. Finally, we demonstrate an example application of the model to an architected material design task. The methods which we developed are applicable to fourth-order tensors beyond elasticity such as piezo-optical tensor etc.

cs.LG↗
arXiv2024

Accurate Crystal Structure Prediction of New 2D Hybrid Organic Inorganic Perovskites

Low dimensional hybrid organic-inorganic perovskites (HOIPs) represent a promising class of electronically active materials for both light absorption and emission. The design space of HOIPs is extremely large, since a diverse space of organic cations can be combined with different inorganic frameworks. This immense design space allows for tunable electronic and mechanical properties, but also necessitates the development of new tools for in silico high throughput analysis of candidate structures. In this work, we present an accurate, efficient, transferable and widely applicable machine learning interatomic potential (MLIP) for predicting the structure of new 2D HOIPs. Using the MACE architecture, an MLIP is trained on 86 diverse experimentally reported HOIP structures. The model is tested on 73 unseen perovskite compositions, and achieves chemical accuracy with respect to the reference electronic structure method. Our model is then combined with a simple random structure search algorithm to predict the structure of hypothetical HOIPs given only the proposed composition. Success is demonstrated by correctly and reliably recovering the crystal structure of a set of experimentally known 2D perovskites. Such a random structure search is impossible with ab initio methods due to the associated computational cost, but is relatively inexpensive with the MACE potential. Finally, the procedure is used to predict the structure formed by a new organic cation with no previously known corresponding perovskite. Laboratory synthesis of the new hybrid perovskite confirms the accuracy of our prediction. This capability, applied at scale, enables efficient screening of thousands of combinations of organic cations and inorganic layers.

cond-mat.mtrl-sci↗
arXiv2024

Zero Shot Molecular Generation via Similarity Kernels

Generative modelling aims to accelerate the discovery of novel chemicals by directly proposing structures with desirable properties. Recently, score-based, or diffusion, generative models have significantly outperformed previous approaches. Key to their success is the close relationship between the score and physical force, allowing the use of powerful equivariant neural networks. However, the behaviour of the learnt score is not yet well understood. Here, we analyse the score by training an energy-based diffusion model for molecular generation. We find that during the generation the score resembles a restorative potential initially and a quantum-mechanical force at the end. In between the two endpoints, it exhibits special properties that enable the building of large molecules. Using insights from the trained model, we present Similarity-based Molecular Generation (SiMGen), a new method for zero shot molecular generation. SiMGen combines a time-dependent similarity kernel with descriptors from a pretrained machine learning force field to generate molecules without any further training. Our approach allows full control over the molecular shape through point cloud priors and supports conditional generation. We also release an interactive web tool that allows users to generate structures with SiMGen online (https://zndraw.icp.uni-stuttgart.de).

physics.chem-ph↗
arXiv2024

Equivariant Matrix Function Neural Networks

Graph Neural Networks (GNNs), especially message-passing neural networks (MPNNs), have emerged as powerful architectures for learning on graphs in diverse applications. However, MPNNs face challenges when modeling non-local interactions in graphs such as large conjugated molecules, and social networks due to oversmoothing and oversquashing. Although Spectral GNNs and traditional neural networks such as recurrent neural networks and transformers mitigate these challenges, they often lack generalizability, or fail to capture detailed structural relationships or symmetries in the data. To address these concerns, we introduce Matrix Function Neural Networks (MFNs), a novel architecture that parameterizes non-local interactions through analytic matrix equivariant functions. Employing resolvent expansions offers a straightforward implementation and the potential for linear scaling with system size. The MFN architecture achieves stateof-the-art performance in standard graph benchmarks, such as the ZINC and TU datasets, and is able to capture intricate non-local interactions in quantum systems, paving the way to new state-of-the-art force fields.

stat.ML↗
arXiv2023

Efficiency, Accuracy, and Transferability of Machine Learning Potentials: Application to Dislocations and Cracks in Iron

Machine learning interatomic potentials (ML-IAPs) enable quantum-accurate, classical molecular dynamics simulations of large systems, beyond reach of density functional theory (DFT). Yet, their efficiency and ability to predict systems larger than DFT supercells are not fully explored, posing a question regarding transferability to large-scale simulations with defects (e.g. dislocations, cracks). Here, we apply a three-step validation approach to body-centered-cubic iron. First, accuracy and efficiency are assessed by optimizing ML-IAPs based on four state-of-the-art ML packages. The Pareto front of computational speed versus testing root-mean-square-error (RMSE) is computed. Second, benchmark properties relevant to plasticity and fracture are evaluated. Their average relative error Q with respect to DFT is found to correlate with RMSE. Third, transferability of ML-IAPs to dislocations and cracks is investigated by using per-atom model uncertainty quantification. The core structures and Peierls barriers of screw, M111 and three edge dislocations are compared with DFT. Traction-separation curve and critical stress intensity factor (K_Ic) are also predicted. Cleavage on the pre-existing crack plane is found to be the zero-temperature atomistic fracture mechanism of pure body-centered-cubic iron under mode-I loading, independent of ML package and training database. Quantitative predictions of dislocation glide paths and KIc can be sensitive to database, ML package, cutoff radius, and are limited by DFT accuracy. Our results highlight the importance of validating ML-IAPs by using indicators beyond RMSE. Moreover, significant computational speed-ups can be achieved by using the most efficient ML-IAP package, yet the assessment of the accuracy and transferability should be performed with care.

cond-mat.mtrl-sci↗