arXiv subjects

Paolo Pegolo

Publications and source records attributed to Paolo Pegolo.

18 recordsLinked to original sources

arXiv2026

High-quality, high-information datasets for universal atomistic machine learning

The quality, consistency, and information content of training data is often what determines the practical value of machine-learning models for atomistic simulations. Yet, many widely used electronic-structure databases are assembled having materials screening as primary goal rather than robust force-field learning, are limited in their scope to a specific class of chemical compounds, and/or employ inconsistent DFT functionals and settings. Here we introduce MAD-1.6, a highly curated dataset designed explicitly for training broadly applicable atomistic models across the periodic table at high levels of theory. MAD-1.6 extends the MAD dataset with targeted enrichment strategies that improve the coverage of chemical space to 102 elements while keeping the total number of configurations compact. All structures are computed with a single, standardized all-electron DFT workflow using the r$^2$SCAN meta-GGA functional and consistent convergence settings, ensuring uniformity across chemically heterogeneous systems. The dataset encompasses molecules, clusters, bulk crystals, surfaces, and low-dimensional structures, and its quality and consistency are further enhanced by outlier removal using uncertainty quantification. We demonstrate the high accuracy that can be achieved with the proposed dataset by training PET-MAD-1.6, a generally applicable r$^2$SCAN interatomic potential that covers 102 elements in the periodic table and achieves exceptional levels of benchmark accuracy and stability in challenging simulation protocols.

cond-mat.mtrl-sci↗

arXiv2026

Pushing the limits of unconstrained machine-learned interatomic potentials

Machine-learned interatomic potentials (MLIPs) are increasingly used to replace computationally demanding electronic-structure calculations to model matter at the atomic scale. The most commonly used model architectures are constrained to fulfill a number of physical laws exactly, from geometric symmetries to energy conservation. Evidence is mounting that relaxing some of these constraints can be beneficial to the efficiency and (somewhat surprisingly) accuracy of MLIPs, even though care should be taken to avoid qualitative failures associated with the breaking of physical symmetries. Given the recent trend of scaling up models to larger numbers of parameters and training samples, a very important question is how unconstrained MLIPs behave in this limit. Here we investigate this issue, showing that -- when trained on large datasets -- unconstrained models can be superior in accuracy and speed when compared to physically constrained models. We assess these models both in terms of benchmark accuracy and in terms of usability in practical scenarios, focusing on static simulation workflows such as geometry optimization and lattice dynamics. We conclude that accurate unconstrained models can be applied with confidence, especially since simple inference-time modifications can be used to recover observables that are consistent with the relevant physical symmetries.

physics.chem-ph↗

arXiv2026

How unconstrained machine-learning models learn physical symmetries

The requirement of generating predictions that exactly fulfill the fundamental symmetry of the corresponding physical quantities has profoundly shaped the development of machine-learning models for physical simulations. In many cases, models are built using constrained mathematical forms that ensure that symmetries are enforced exactly. However, unconstrained models that do not obey rotational symmetries are often found to have competitive performance, and to be able to \emph{learn} to a high level of accuracy an approximate equivariant behavior with a simple data augmentation strategy. In this paper, we introduce rigorous metrics to measure the symmetry content of the learned representations in such models, and assess the accuracy by which the outputs fulfill the equivariant condition. We apply these metrics to two unconstrained, transformer-based models operating on decorated point clouds (a graph neural network for atomistic simulations and a PointNet-style architecture for particle physics) to investigate how symmetry information is processed across architectural layers and is learned during training. Based on these insights, we establish a rigorous framework for diagnosing spectral failure modes in ML models. Enabled by this analysis, we demonstrate that one can achieve superior stability and accuracy by strategically injecting the minimum required inductive biases, preserving the high expressivity and scalability of unconstrained architectures while guaranteeing physical fidelity.

cs.LG↗

arXiv2026

Journal Research Data Policies in Materials Science

Open and reproducible research in materials science relies on the availability of data, code, and common metadata standards. Journal research data policies (RDPs) remain a primary mechanism by which publication norms are defined and enforced. We survey RDPs for 171 materials science journals spanning 17 publishers, using an expanded coding framework that captures both data-and-code sharing behavior as well as refereeing standards. We find clear signs of progress in comparison to earlier research on RDPs: nearly all journals provide an RDP, and most mention data availability statements. However, enforceable requirements remain uncommon, public deposition of underlying data is rarely mandatory, and FAIR publication is typically encouraged rather than required. Expectations for research software are substantially less developed than those for data, with limited attention to versioning and persistent identifiers, dependency disclosure, reproducible execution environments, or software quality practices. Aggregating the findings on policy features into an open research data score reveals pronounced heterogeneity across journals. Neither impact factor nor access model reliably predicts policy strength. Double-coding further shows that more complex policies and stricter policies can be more challenging to interpret consistently, and we highlight challenges in consistent RDP encoding across studies. Lastly, we conclude with recommended best practice directions for the future.

cs.DL↗

arXiv2025

Metatensor and metatomic: foundational libraries for interoperable atomistic machine learning

Incorporation of machine learning (ML) techniques into atomic-scale modeling has proven to be an extremely effective strategy to improve the accuracy and reduce the computational cost of simulations. It also entails conceptual and practical challenges, as it involves combining very different mathematical foundations, as well as software ecosystems that are very well developed in their own right, but do not share many commonalities. To address these issues and facilitate the adoption of ML in atomistic simulations, we introduce two dedicated software libraries. The first one, metatensor, provides multi-platform and multi-language storage and manipulation of arrays with many potentially sparse indices, designed from the ground up for atomistic ML applications. By combining the actual values with metadata that describes their nature and that facilitates the handling of geometric information and gradients with respect to the atomic positions, metatensor provides a common framework to enable data sharing between ML software -- typically written in Python -- and established atomistic modeling tools -- typically written in Fortran, C or C++. The second library, metatomic, provides an interface to store an atomistic ML model and metadata about this model in a portable way, facilitating the implementation, training and distribution of models, and their use across different simulation packages. We showcase a growing ecosystem of tools, including low-level libraries, training utilities, and interfaces with existing software packages that demonstrate the effectiveness of metatensor and metatomic in bridging the gap between traditional simulation software and modern ML frameworks.

physics.chem-ph↗

arXiv2025

Revealing Fast Ionic Conduction in Solid Electrolytes through Machine Learning Accelerated Raman Calculations

Fast ionic conduction is a defining property of solid electrolytes for all-solid-state batteries. Previous studies have suggested that liquid-like cation motion associated with fast ionic transport can disrupt crystalline symmetry, thereby lifting Raman selection rules. Here, we exploit the resulting low-frequency, diffusive Raman scattering as a spectral signature of fast ionic conduction and develop a machine learning-accelerated computational pipeline to identify promising solid electrolytes based on this feature. By overcoming the steep computational barriers to calculating Raman spectra of strongly disordered materials at finite temperatures, we achieve near-ab initio accuracy and demonstrate the predictive power of our approach for sodium-ion conductors, revealing clear Raman signatures of liquid-like ion conduction. This work highlights how machine learning can bridge atomistic simulations and experimental observables, enabling data-efficient discovery of fast-ion conductors.

cond-mat.mtrl-sci↗

arXiv2025

Representing spherical tensors with scalar-based machine-learning models

Rotational symmetry plays a central role in physics, providing an elegant framework to describe how the properties of 3D objects -- from atoms to the macroscopic scale -- transform under the action of rigid rotations. Equivariant models of 3D point clouds are able to approximate structure-property relations in a way that is fully consistent with the structure of the rotation group, by combining intermediate representations that are themselves spherical tensors. The symmetry constraints however make this approach computationally demanding and cumbersome to implement, which motivates increasingly popular unconstrained architectures that learn approximate symmetries as part of the training process. In this work, we explore a third route to tackle this learning problem, where equivariant functions are expressed as the product of a scalar function of the point cloud coordinates and a small basis of tensors with the appropriate symmetry. We also propose approximations of the general expressions that, while lacking universal approximation properties, are fast, simple to implement, and accurate in practical settings.

physics.chem-ph↗

arXiv2025

PET-MAD, a lightweight universal interatomic potential for advanced materials modeling

Machine-learning interatomic potentials (MLIPs) have greatly extended the reach of atomic-scale simulations, offering the accuracy of first-principles calculations at a fraction of the cost. Leveraging large quantum mechanical databases and expressive architectures, recent ''universal'' models deliver qualitative accuracy across the periodic table but are often biased toward low-energy configurations. We introduce PET-MAD, a generally applicable MLIP trained on a dataset combining stable inorganic and organic solids, systematically modified to enhance atomic diversity. Using a moderate but highly-consistent level of electronic-structure theory, we assess PET-MAD's accuracy on established benchmarks and advanced simulations of six materials. Despite the small training set and lightweight architecture, PET-MAD is competitive with state-of-the-art MLIPs for inorganic solids, while also being reliable for molecules, organic materials, and surfaces. It is stable and fast, enabling the near-quantitative study of thermal and quantum mechanical fluctuations, functional properties, and phase transitions out of the box. It can be efficiently fine-tuned to deliver full quantum mechanical accuracy with a minimal number of targeted calculations.

cond-mat.mtrl-sci↗

arXiv2025

Exploring the design space of machine-learning models for quantum chemistry with a fully differentiable framework

Traditional atomistic machine learning (ML) models serve as surrogates for quantum mechanical (QM) properties, predicting quantities such as dipole moments and polarizabilities, directly from compositions and geometries of atomic configurations. With the emergence of ML approaches to predict the "ingredients" of a QM calculation, such as the ground state charge density or the effective single-particle Hamiltonian, it has become possible to obtain multiple properties through analytical physics-based operations on these intermediate ML predictions. We present a framework to seamlessly integrate the prediction of an effective electronic Hamiltonian, for both molecular and condensed-phase systems, with PySCFAD, a differentiable QM workflow that facilitates its indirect training against functions of the Hamiltonian, such as electronic energy levels, dipole moments, polarizability, etc. We then use this framework to explore various possible choices within the design space of hybrid ML/QM models, examining the influence of incorporating multiple targets on model performance and learning a reduced-basis ML Hamiltonian that can reproduce targets computed from a much larger basis. Our benchmarks evaluate the accuracy and transferability of these hybrid models, compare them against predictions of atomic properties from their surrogate models, and provide indications to guide the design of the interface between the ML and QM components of the model.

physics.chem-ph↗

arXiv2024

Effects of colored disorder on the heat conductivity of SiGe alloys from first principles

Semiconducting alloys, in particular SiGe, have been employed for several decades as high-temperature thermoelectric materials. Devising strategies to reduce their thermal conductivity may provide a substantial improvement in their thermoelectric performance also at lower temperatures. We have carried out an ab initio investigation of the thermal conductivity of SiGe alloys with random and spatially correlated mass disorder employing the Quasi-Harmonic Green-Kubo (QHGK) theory with force constants computed by density functional theory. Leveraging QHGK and the hydrodynamic extrapolation to achieve size convergence, we obtained a detailed understanding of lattice heat conduction in SiGe and demonstrated that colored disorder suppresses thermal transport across the acoustic vibrational spectrum, leading to up to a 4-fold enhancement in the intrinsic thermoelectric figure of merit.

cond-mat.mtrl-sci↗

arXiv2024

Seebeck coefficient of ionic conductors from Bayesian regression analysis

We propose a novel approach to evaluating the ionic Seebeck coefficient in electrolytes from relatively short equilibrium molecular dynamics simulations, based on the Green-Kubo theory of linear response and Bayesian regression analysis. By exploiting the probability distribution of the off-diagonal elements of a Wishart matrix, we develop a consistent and unbiased estimator for the Seebeck coefficient whose statistical uncertainty can be arbitrarily reduced in the long-time limit. To validate the effectiveness of our method, we benchmark it against extensive equilibrium molecular dynamics simulations conducted on molten $\mathrm{CsF}$ using empirical force fields. We then employ this procedure to calculate the Seebeck coefficient of molten $\mathrm{NaCl}$, $\mathrm{KCl}$ and $\mathrm{LiCl}$ using neural-network force fields trained on ab initio data over a range of pressure-temperature conditions.

cond-mat.mtrl-sci↗

arXiv2024

Thermal transport of glasses via machine learning driven simulations

Accessing the thermal transport properties of glasses is a major issue for the design of production strategies of glass industry, as well as for the plethora of applications and devices where glasses are employed. From the computational standpoint, the chemical and morphological complexity of glasses calls for atomistic simulations where the interatomic potentials are able to capture the variety of local environments, composition, and (dis)order that typically characterize glassy phases. Machine-learning potentials (MLPs) are emerging as a valid alternative to computationally expensive ab initio simulations, inevitably run on very small samples which cannot account for disorder at different scales, as well as to empirical force fields, fast but often reliable only in a narrow portion of the thermodynamic and composition phase diagrams. In this article, we make the point on the use of MLPs to compute the thermal conductivity of glasses, through a review of recent theoretical and computational tools and a series of numerical applications on vitreous silica and vitreous silicon, both pure and intercalated with lithium.

cond-mat.dis-nn↗

arXiv2023

Self-interaction and transport of solvated electrons in molten salts

The dynamics of (few) electrons dissolved in an ionic fluid--as when a small amount of metal is added to a solution while upholding its electronic insulation--manifests interesting properties that can be ascribed to nontrivial topological features of particle transport (e.g., Thouless' pumps). In the adiabatic regime, the charge distribution and the dynamics of these dissolved electrons are uniquely determined by the nuclear configuration. Yet, their localization into effective potential wells and their diffusivity are dictated by how the self-interaction is modeled. In this article, we investigate the role of self-interaction in the description of localization and transport properties of dissolved electrons in non-stoichiometric molten salts. Although the account for the exact (Fock) exchange strongly localizes the dissolved electrons, decreasing their tunneling probability and diffusivity, we show that the dynamics of the ions and of the dissolved electrons are largely uncorrelated, irrespective of the degree to which the electron self-interaction is treated, and in accordance with topological arguments.

physics.chem-ph↗

arXiv2023

Hydrodynamic finite-size scaling of the thermal conductivity in glasses

In the past few years, the theory of thermal transport in amorphous solids has been substantially extended beyond the Allen-Feldman model. The resulting formulation, based on the Green-Kubo linear response or the Wigner-transport equation, bridges this model for glasses with the traditional Boltzmann kinetic approach for crystals. The computational effort required by these methods usually scales as the cube of the number of atoms, thus severely limiting the size range of computationally affordable glass models. Leveraging hydrodynamic arguments, we show how this issue can be overcome through a simple formula to extrapolate a reliable estimate of the bulk thermal conductivity of glasses from finite models of moderate size. We showcase our findings for realistic models of paradigmatic glassy materials.

cond-mat.mtrl-sci↗

arXiv2023

Unearthing the foundational role of anharmonicity in heat transport in glasses

The time-honored Allen-Feldman theory of heat transport in glasses is generally assumed to predict a finite value for the thermal conductivity, even if it neglects the anharmonic broadening of vibrational normal modes. We demonstrate that the harmonic approximation predicts that the bulk lattice thermal conductivity of harmonic solids inevitably diverges at any temperature, irrespective of configurational disorder, and that its ability to represent the heat-transport properties observed experimentally in most glasses is implicitly due to finite-size effects. Our theoretical analysis is thoroughly benchmarked against careful numerical simulations. Our findings thus reveal that a proper account of anharmonic effects is indispensable to predict a finite value for the bulk thermal conductivity in any solid material, be it crystalline or glassy.

cond-mat.dis-nn↗

arXiv2022

Topology, oxidation states, and charge transport in ionic conductors

Recent theoretical advances, based on a combination of concepts from Thouless' topological theory of adiabatic charge transport and a newly introduced gauge-invariance principle for transport coefficients, have permitted to connect (and reconcile) Faraday's picture of ionic transport -- whereby each atom carries a well-defined integer charge -- with a rigorous quantum description of the electronic charge-density distribution, which hardly suggests its partition into well defined atomic contributions. In this paper we review these progress and in particular that, by relaxing some general topological conditions, charge may be transported in ionic conductors without any net ionic displacements. After reporting numerical experiments which corroborate these findings, we introduce a new connection between our topological picture and the well-known Marcus-Hush theory of electron transfer, which we are able to connect with the topology of adiabatic paths drawn by atomic trajectories. As a significant byproduct, the results reviewed here permit to classify different regimes of ionic transport according to the topological properties of the electronic structure of the conducting material. We finally report on a few recent applications to energy materials and planetary sciences.

cond-mat.mtrl-sci↗

arXiv2021

Temperature- and vacancy-concentration-dependence of heat transport in Li$_3$ClO from multi-method numerical simulations

Despite governing heat management in any realistic device, the microscopic mechanisms of heat transport in all-solid-state electrolytes are poorly known: existing calculations, all based on simplistic semi-empirical models, are unreliable for superionic conductors and largely overestimate their thermal conductivity. In this work, we deploy a combination of state-of-the-art methods to calculate the thermal conductivity of a prototypical Li-ion conductor, the Li$_3$ClO antiperovskite. By leveraging ab initio, machine-learning, and force-field descriptions of inter-atomic forces, we are able to reveal the massive role of anharmonic interactions and diffusive defects on the thermal conductivity and its temperature dependence, and to eventually embed their effects into a simple rationale which is likely applicable to a wide class of ionic conductors.

cond-mat.mtrl-sci↗

arXiv2020

Oxidation states, Thouless' pumps, and non-trivial ionic transport in non-stoichiometric electrolytes

Thouless' quantization of adiabatic particle transport permits to associate an integer topological charge with each atom of an electronically gapped material. If these charges are additive and independent of atomic positions, they provide a rigorous definition of atomic oxidation states and atoms can be identified as integer-charge carriers in ionic conductors. Whenever these conditions are met, charge transport is necessarily convective, i.e. it cannot occur without substantial ionic flow, a transport regime that we dub trivial. We show that the topological requirements that allow these conditions to be broken are the same that would determine a Thouless' pump mechanism if the system were subject to a suitably defined time-periodic Hamiltonian. The occurrence of these requirements determines a non-trivial transport regime whereby charge can flow without any ionic convection, even in electronic insulators. These results are first demonstrated with a couple of simple molecular models that display a quantum pump mechanism upon introduction of a fictitious time dependence of the atomic positions along a closed loop in configuration space. We finally examine the impact of our findings on the transport properties of non-stoichiometric alkali-halide melts, where the same topological conditions that would induce a quantum pump mechanism along certain closed loops in configuration space also determine a non-trivial transport regime such that most of the total charge current results to be uncorrelated from the ionic ones.

cond-mat.mtrl-sci↗