Searcharxiv⌕ Search

arXiv subjects

Ralf Drautz

Publications and source records attributed to Ralf Drautz.

At least 19 recordsLinked to original sources

Efficient Machine Learning Interatomic Potentials

Machine learning interatomic potentials (MLIPs) reach an accuracy close to quantum mechanical reference calculations at a small fraction of the cost. Efficiency is the purpose of MLIPs, yet strategies for efficient MLIPs have not been reviewed systematically. This review takes efficiency as its guideline. To compare MLIPs on a common footing, we represent them in a complete basis expansion, the Atomic Cluster Expansion (ACE) and its graph generalization (GRACE). The expansion is canonical, built from products of single-particle functions. Classical interatomic potentials, such as the embedded-atom method, and many modern MLIPs, including equivariant transformers, can be represented within this expansion. Much of the apparent diversity of MLIP architectures therefore reflects how non-linearity, layer coupling and optimization are handled rather than which interactions the models can represent. We show how the cost of an MLIP becomes linear in system size and, at fixed rank of the tensor decomposition, in model complexity, and we discuss the role of machine learning, which provides non-linear parameterizations and optimization. Because efficiency is meaningful only at a given accuracy, we also compare accuracy and cost. On the Matbench Discovery leaderboard, models with three to five layers approach the accuracy of deeper models, and one basis, truncated along the Pareto front of error and speed, spans more than two orders of magnitude in cost. Ultimately, training data determine the quality of any MLIP.

cond-mat.mtrl-sci↗

Computing binary alloy phase diagrams with explicit configurational and vibrational entropy

Phase stability in multicomponent solid solutions depends on configurational entropy beyond the ideal mixing limit, but capturing it together with vibrational entropy within the same atomistic framework remains challenging. Here, we extend non-equilibrium thermodynamic integration to composition-dependent transformations through an alchemical interpolation of the interactions, combined with Monte Carlo identity exchange moves and molecular dynamics that sample the vibrational and non-ideal configurational entropy along the integration path. We apply the framework to the Au-Cu binary alloy using Atomic Cluster Expansion (ACE) potentials trained on density functional theory data using the LDA, PBE, and r2SCAN functionals, and construct composition-temperature phase diagrams directly from atomistic free energies. We find that explicit configurational sampling lowers the AuCu order-disorder transition temperature predicted by an ACE potential trained on LDA data by about 126 K, closer to the experimental value of 683 K, and substantially widens the stability range of the solid solution. At the same time, the much larger sensitivity to the exchange-correlation functional shows that this level of agreement should not be interpreted as general predictive accuracy. No single functional is uniformly accurate across the thermodynamic observables considered here, showing that functional sensitivity must be assessed alongside configurational effects when constructing atomistic phase diagrams.

cond-mat.mtrl-sci↗

A thermodynamic correction layer for machine learning interatomic potentials

Machine learning interatomic potentials enable atomistic simulations with near \textit{ab initio} accuracy over large length and time scales. Foundation models extend this capability across the periodic table but inherit the thermodynamic description of the electronic structure methods used to generate their training data. Small errors in relative phase stability can therefore produce large errors in transformation temperatures. We introduce an additive thermodynamic correction layer that shifts phase free energies towards an assessed reference without retraining or modifying the base potential. A first order thermodynamic construction converts the required phase free energy shifts directly into configuration energy targets, which are fitted using a linear atomic cluster expansion and refined iteratively from measured free energies. We apply the method to melting and solid to solid transformations across elements with different crystal structures using a foundation GRACE model. After the correction, all transformation temperatures lie within the prescribed tolerance of the assessed values, including Ca, Ti, and Zr, where a solid to solid transition and melting are corrected together. Static energy curves, bulk moduli, and liquid structure remain close to those of the base model. Vacancy formation energies, thermal expansion, and liquid density depend on the corrected phase balance and change where the correction is large. The method requires neither retraining, differentiable dynamics, nor access to the parameters of the base potential. It provides a modular route for combining transferable atomistic models with assessed phase thermodynamics.

cond-mat.mtrl-sci↗

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↗

A latent-space extrapolation grade built into graph atomic cluster expansion foundation potentials

Foundation machine-learning interatomic potentials cover broad configurational and chemical spaces, but their reliability can vary across the atomic environments encountered during a simulation. Here we introduce the calibrated Mahalanobis (CALM) extrapolation grade $γ$, a piecewise differentiable per-atom quantity integrated into GRACE foundation models and evaluated alongside energies and forces in a single model pass. We define $γ$ from nearest-cluster Mahalanobis distances in latent feature space, setting $γ=1$ from the training-distance distribution separately for each element and cluster. Controlled tests show that a normalized random projection of the invariant many-body basis detects structural and chemical extrapolation. On different foundation datasets, OMat24 and SMAX, $γ$ correlates with atomic force errors and separates structures with different error distributions. The CALM grade adds percent-level computational cost, and its spatial gradient guides uncertainty-biased data collection toward configurations with larger absolute force errors.

cond-mat.mtrl-sci↗

Reactive calcium carbonate precipitation from an atomic cluster expansion potential and enhanced sampling

Calcium carbonate formation from aqueous solution is central to biomineralization and to carbon sequestration through mineral carbonation. At near-neutral pH, the process is highly reactive, with proton transfer mediating the interconversion between carbonate species. Most atomistic simulations to date either treat carbonate speciation as fixed or consider proton transfer only in small clusters. Here, we combine an ab initio trained atomic cluster expansion (ACE) machine-learning potential for molecular dynamics with enhanced sampling to enable reactive simulations of the early stages of calcium carbonate precipitation at previously inaccessible length and time scales. We study proton transfer and carbonate speciation in ion pairs and triplets, as well as in the collective aggregation of many ions. Our simulations with few ions show that ion association provides a favorable pathway for proton transfer, facilitating interconversion between carbonate, bicarbonate, and carbonic acid. In many-ion systems, proton transfer occurs spontaneously alongside aggregation, and we observe significant changes in the coordination environments as species evolve during the simulations. These results show that ion aggregation and chemical reactivity can be strongly coupled during the early stages of nucleation from solution.

physics.chem-ph↗

A hidden low-temperature transformation pathway in compositionally complex materials

Most compositionally complex materials (CCMs, frequently referred to as high entropy alloys) are metastable and their attractive properties often belong to kinetically trapped states. However, pathways towards lower-free-energy phase states governing long-term stability, can remain hidden because diffusion-controlled atomic redistribution is too slow to be revealed at experimentally accessible timescales. This blind spot is acute in CCM design: enormous compositional spaces are screened for performance, yet the low-temperature kinetics and the associated transformation pathways determining whether that performance persists are rarely considered in material selection. Here we use defect-rich nanoscale volumes coupled with atom-probe tomography to access and reconstruct the hidden phase-evolution pathway in a metastable Ag24Au20Pd50Pt6 electrocatalyst, without relying on elevated temperatures to accelerate the transformation. By varying microstructural starting state, annealing temperature and time, we reveal precipitation of a Pt-rich phase within the fcc matrix, its coarsening and re-homogenization. The Pt-rich phase recurs after homogenization with delayed kinetic accessibility, while prolonged annealing extends the pathway to 300°C. Atomistic simulations independently predict the same Pt-rich phase selection. The transformation is accompanied by a 3.7-fold loss of catalytic activity for hydrogen evolution. These results establish hidden phase-evolution pathways as a materials-design variable: resolving them can guide the selection of metastable CCMs not only for their as-synthesized properties, but also for the phase states and associated functionalities they may access over time.

cond-mat.mtrl-sci↗

Aromatic Molecule Solvation in Liquid Water with Coupled Cluster Accuracy: The Balance of Pi-Interactions and Hydrophobicity

Aromatic organic solutes in water exhibit a delicate balance between hydrophobic solvation and directional O-H$\cdots π$ hydrogen bonds, yet widely used force fields and state-of-the-art density functional approaches struggle to provide a consistent picture of these pivotal interactions. We introduce a data-efficient upfitting strategy to train a machine learning interatomic potential (MLIP) based on the graph atomic cluster expansion for aqueous aromatic molecules with CCSD(T) accuracy for condensed phase simulations, using only finite molecular clusters. We apply our method to aqueous toluene (C$_6$H$_5$CH$_3$). The resulting CCSD(T)-quality MLIP reproduces coupled cluster energies and forces in bulk and reveals that commonly employed methods do not capture the crucial balance between hydrophilic and hydrophobic solvation, distorting the interactions of aromatic molecules with their environment. Representative biomolecular force fields substantially understructure the hydrophobic solvation shell and misorient interfacial water, while overestimating $π$-contacts, yielding an inconsistent solvation balance. Even hybrid DFT and MP2 overestimate barriers to breaking of water-$π$ hydrogen bonds. Our workflow provides a practical, general route to CCSD(T)-quality condensed-phase simulations of aqueous solutions, and thus constructed interaction potentials now open the door to consistent, highly accurate benchmark studies of $π$-contacts and hydrophobic effects in biomolecular contexts such as solvation of proteins and DNA in aqueous environments.

physics.chem-ph↗

A general-purpose atomic cluster expansion interatomic potential for niobium

Niobium, a body-centered cubic transition metal, poses a challenge for interatomic potentials, which struggle to capture its properties, such as phonons, high-pressure behavior, energy barriers to dislocation glide, and others. To tackle this challenge, we constructed a general-purpose atomic cluster expansion (ACE) potential for niobium. We trained our ACE on thousands of density functional theory (DFT) structures spanning a diversity of local environments. We validated it across a range of properties and compared it with existing empirical and machine learning (ML) potentials, including a novel universal ML potential. The resulting ACE balances accuracy, efficiency, and robustness, enabling large-scale exploration of niobium with near-DFT precision. Finally, our ACE held its own in a stringent test: a near-million-atom molecular dynamics simulation of fracture

cond-mat.mtrl-sci↗

Fast contracted Clebsch--Gordan tensor products for equivariant graph neural networks

We present an $\mathcal{O}(L^3)$ algorithm for evaluating contracted Clebsch--Gordan tensor products in $\mathrm{O}(3)$-equivariant machine learning potentials at fixed Canonical Polyadic (CP) rank. Mapping the angular integral to a structured Gauss--Legendre and Fourier tensor-product grid decouples the radial channel contractions from the angular transforms. The antisymmetric parity-odd Clebsch--Gordan channels, unreachable by the symmetric pointwise product on a scalar $S^2$ grid, are recovered through the surface-curl pairing $\hat r \cdot [\nabla_{S^2} A \times \nabla_{S^2} B]$, the spherical Poisson bracket, which supplies the $L=1$ angular momentum on the grid while preserving rotational equivariance. The construction extends to parity-aware equivariant message passing in atomic-cluster-expansion-style architectures and is verified by direct numerical quadrature. The full uncontracted Clebsch--Gordan tensor product remains subject to the $\mathcal{O}(L^4)$ output-size lower bound. A benchmark shows wall-clock scaling empirically as $L^2$ across the practical $l_{\max}$ range. For the on-site contraction this is pre-asymptotic, giving way to $L^3$ at large $l_{\max}$. For message passing it is structural and the runtime is memory-bandwidth bound on $L^2$-sized grid tensors.

physics.comp-ph↗

Hydride formation and phase separation in palladium nanoparticles from a transferable atomic cluster expansion potential

The palladium-hydrogen system is a prototype for hydrogen-metal interactions and underpins technologies such as hydrogen storage, catalysis and purification. Yet its nanoscale behaviour -- where surface and interface energetics, elastic coherency strain and size-dependent thermodynamics govern phase separation -- has eluded accurate atomistic simulation. Empirical potentials misrepresent the energetics of interstitial hydrogen, while existing machine-learning models are restricted to bulk phases at low-hydrogen environments. Here we introduce an atomic cluster expansion (ACE) for Pd-H that reproduces formation energies, phonon spectra, elastic constants, hydrogen migration barriers and surface adsorption with near-DFT accuracy, benchmarked directly against neutron-scattering, high-pressure and lattice-expansion experiments. Its near-linear scaling and CPU efficiency make molecular dynamics of PdH$_x$ nanoparticles exceeding 28,000 atoms ($\sim$12 nm in diameter) tractable over nanosecond timescales. These simulations resolve, at the atomic scale, the kinetic separation of $α$- and $β$-PdH$_x$ into a core-shell architecture, reproduce the experimentally observed size dependence of the lattice parameter, and uncover a pronounced hydrogen-induced lowering of the nanoparticle melting temperature. The potential brings experimentally relevant scales of metal-hydride dynamics within quantitative reach.

cond-mat.mtrl-sci↗

Data-efficient machine-learning of complex Fe-Mo intermetallics using domain knowledge of chemistry and crystallography

Atomistic simulations of multi-component systems require accurate descriptions of interatomic interactions to resolve details in the energy of competing phases. A particularly challenging case are topologically close-packed (TCP) phases with close energetic competition of numerous different site occupations even in binary systems like Fe-Mo. In this work, machine learning (ML) models are presented that overcome this challenge by using features with domain knowledge of chemistry and crystallography. The resulting data-efficient ML models need only a small set of training data of simple TCP phases $A$15, $σ$, $χ$, $μ$, $C$14, $C$15, $C$36 with 2-5 WS to reach robust and accurate predictions for the complex TCP phases $R$, $M$, $P$, $δ$ with 11-14 WS. Several ML models with kernel-ridge regression, multi-layer perceptrons, and random forests, are trained on less than 300 DFT calculations for the simple TCP phases in Fe-Mo. The performance of these ML models is shown to improve systematically with increased utilization of domain knowledge. The convex hulls of the $R$, $M$, $P$ and $δ$ phase in the Fe-Mo system are predicted with uncertainties of 20-25 meV/atom and show very good agreement with DFT verification. Complementary X-ray diffraction experiments and Rietveld analysis are carried out for an Fe-Mo R-phase sample. The measured WS occupancy is in excellent agreement with the predictions of our ML model using the Bragg Williams approximation at the same temperature.

cond-mat.mtrl-sci↗

AI-Driven Expansion and Application of the Alexandria Database

We present a novel multi-stage workflow for computational materials discovery that achieves a 99% success rate in identifying compounds within 100 meV/atom of thermodynamic stability, with a threefold improvement over previous approaches. By combining the Matra-Genoa generative model, Orb-v2 universal machine learning interatomic potential, and ALIGNN graph neural network for energy prediction, we generated 119 million candidate structures and added 1.3 million DFT-validated compounds to the ALEXANDRIA database, including 74 thousand new stable materials. The expanded ALEXANDRIA database now contains 5.8 million structures with 175 thousand compounds on the convex hull. Predicted structural disorder rates (37-43%) match experimental databases, unlike other recent AI-generated datasets. Analysis reveals fundamental patterns in space group distributions, coordination environments, and phase stability networks, including sub-linear scaling of convex hull connectivity. We release the complete dataset, including sAlex25 with 14 million out-of-equilibrium structures containing forces and stresses for training universal force fields. We demonstrate that fine-tuning a GRACE model on this data improves benchmark accuracy. All data, models, and workflows are freely available under Creative Commons licenses.

cond-mat.mtrl-sci↗

Hydrogen uptake and hydride formation in Al$_x$CoCrFeNi high-entropy alloys: First-principles, universal-potential, and experimental study

Hydrogen uptake in complex multicomponent alloys, including high-entropy alloys (HEAs), governs both hydrogen storage capacity and resistance to hydrogen-induced degradation. We combine high-pressure experiments, density-functional theory (DFT), and a GRACE universal interatomic potential to investigate hydrogen absorption in Al$_{0.3}$CoCrFeNi and Al$_3$CoCrFeNi HEAs. In H$_2$ as a pressure-transmitting medium, the FCC Al$_{0.3}$CoCrFeNi alloy forms hydrides at ambient temperature above 3 GPa, whereas the Al-rich B2 Al$_3$CoCrFeNi alloy shows no evidence of hydride formation even upon heating at pressures up to 50 GPa. Experiments and calculations show that aluminum suppresses hydrogen uptake by increasing solution energies and destabilizing interstitial sites. The universal potential, employed in the calculations and pretrained on large DFT databases, closely reproduces DFT energetics and demonstrates transferability from the dilute limit to the hydride-forming regime. Simulations further disentangle the roles of local ordering, volume changes, composition, and crystal structure. Overall, our results indicate that hydrogen solubility in Al-containing HEAs is governed primarily by composition, with Al-driven B2 ordering as a strong secondary effect.

cond-mat.mtrl-sci↗

Exploring the extremes: atomic basis for multi-elemental materials science under complex thermodynamic conditions

Modern materials science has historically been founded on combining restricted subsets of the periodic table, favoring high-purity, few-element systems. However, the demands of an emerging circular economy, together with the need to understand materials behavior under planetary and industrial extremes, increasingly require mastering Mendeleev materials - chemically and structurally complex systems that span large portions of the periodic table. In these regimes, current universal machine-learning interatomic potentials often fail, largely due to systematic gaps in traditional training datasets that heavily emphasize low-energy, near-equilibrium structures. We address this limitation by introducing a chemistry-agnostic, information-entropy-maximization protocol for data generation. By decoupling structural sampling from thermodynamic bias, our approach provides a robust physical prior for atomic interactions across the entire periodic table, including regimes far from equilibrium and under extreme conditions. Training a Graph Atomic Cluster Expansion (GRACE) model on the resulting statistically maximized entropy (SMAX) dataset yields markedly improved robustness across a range of stringent benchmarks. These include large-strain phase transformations in tin, defect evolution in tungsten-based alloys, and catalytic reaction barrier prediction. More broadly, our approach establishes a scalable and principled methodology for navigating the vast chemical and configurational space relevant to future materials design. It enables a paradigm of discovery by simulation in which unbiased sampling protocols autonomously resolve emergent structures in multi-elemental mixtures-such as systems containing the nine most abundant elements in the Earth's crust-without reliance on a priori chemical assumptions.

cond-mat.mtrl-sci↗

Water Phase Diagram from a General-Purpose Atomic Cluster Expansion Potential

Water's phase diagram remains one of the most intricate and challenging benchmarks in molecular modeling. In this study, we compute the phase diagram of water using an Atomic Cluster Expansion (ACE) potential trained on density-functional theory (DFT) calculations based on the revPBE-D3 exchange and correlation functional. We compute solid-liquid chemical potential differences and melting points using biased coexistence simulations with the On-the-Fly Probability Enhanced Sampling (OPES) method. Starting from these points, we trace coexistence lines using Gibbs-Duhem integration. This combination of methods allows us to consistently map pressure-temperature phase boundaries and reconstruct the full phase diagram between approximately 100-500 K and 0-4 GPa. The stability regions of the main ice polymorphs (Ih, II, V, VI, and VII) are reproduced in close agreement with experiments. As in earlier studies based on DFT, ice III is metastable and there are systematic shifts of coexistence lines with respect to experimental results. Our results demonstrate the capability of our general-purpose ACE potential to capture the complex phase behavior of water across wide thermodynamic conditions.

cond-mat.mtrl-sci↗

Graph atomic cluster expansion for foundational machine learning interatomic potentials

Foundational machine learning interatomic potentials that can accurately and efficiently model a vast range of materials are critical for accelerating atomistic discovery. We introduce universal potentials based on the graph atomic cluster expansion (GRACE) framework, trained on several of the largest available materials datasets. Through comprehensive benchmarks, we demonstrate that the GRACE models establish a new Pareto front for accuracy versus efficiency among foundational interatomic potentials. We further showcase their exceptional versatility by adapting them to specialized tasks and simpler architectures via fine-tuning and knowledge distillation, achieving high accuracy while preventing catastrophic forgetting. This work establishes GRACE as a robust and adaptable foundation for the next generation of atomistic modeling, enabling high-fidelity simulations across the periodic table.

cond-mat.mtrl-sci↗

Conservative adaptive-precision interatomic potentials

Adaptive precision molecular dynamics simulations have developed along energy- and force-coupling approaches, which allow for a continuous transition between different particle descriptions or interaction potentials. Most approaches consider different (fixed) spatial regions, which control the transition between the descriptions and consequently avoid a consistent momentum-conserving Hamiltonian description. We present here a new approach to fully integrate the coupling into a Hamiltonian, therefore allowing for a conservative description, which, by design, guarantees both energy and momentum conservation. By coupling a fast EAM potential to a highly accurate ACE potential, we verify numerically the conservation properties and show that one can achieve - dependent on both the potential and the atomistic system - a speedup of one or two orders of magnitude compared to a pure ACE simulation.

physics.comp-ph↗