arXiv subjects

Silvia Bonfanti

Publications and source records attributed to Silvia Bonfanti.

At least 19 recordsLinked to original sources

arXiv2026

Dyna-Mat: End-to-end benchmarking of foundation machine learning interatomic potentials in finite-temperature ensembles

Foundation machine learning interatomic potentials (MLIPs) are increasingly being used as drop-in replacements for first-principles calculations, enabling simulations of materials at length and time scales that were previously inaccessible. However, due to lack of ground truth data, their accuracy on structural and dynamical observables in finite thermodynamic ensembles is yet to be established. Here, we introduce Dyna-Mat-v1.0, a benchmark dataset of condensed-phase first-principles molecular dynamics trajectories designed to test foundation MLIPs at realistic finite-temperature conditions. Using this dataset, we evaluate 15 foundation MLIPs across four model tiers by comparing both single-point energy and force errors on first-principles configurations and observables generated from MLIP-driven trajectories. We find that "on average" models with lower single-point force errors also yield lower errors for structural and dynamical observables. However, there are individual systems for which low force errors lead to qualitative failures in the predicted structure. Pressure remains poorly described across most models, pointing to limitations in the density functional theory stress labels available in current large-scale training datasets. Finally, we construct an accuracy-cost Pareto frontier to identify the best trade-offs for molecular dynamics with foundation MLIPs, finding that the latest generation of cross-trained models is close to Pareto-optimal according to the accuracy metrics considered here. Overall, Dyna-Mat-v1.0 shows that end-to-end finite-temperature validation is essential for quantifying the predictive behaviour of foundation MLIPs, and provides a simple, scalable route for assessing them beyond static and harmonic benchmarks relevant to materials design.

cond-mat.mtrl-sci↗

arXiv2026

Transmutation-accelerated sampling method for multi-component ZrCu(Al) metallic glasses

We investigate multi-component metallic glass systems using a hybrid Molecular Dynamics (MD) and Variance-Constrained Semi-Grand Canonical approach. This method enables us to generate samples with properties consistent with experimental observations, at deeply supercooled states that are typically inaccessible with conventional MD simulations. Using a realistic interatomic potential, we investigate the dynamics, kinetic stability, and rheology of a ZrCu(Al) metallic glass, together with the widely studied ZrCu system, in the low-temperature glassy regime. Our results demonstrate how the hybrid method enhances relaxation and provides a generic framework for modeling realistic complex metallic glasses.

cond-mat.dis-nn↗

arXiv2026

Data-driven body-centered cubic phase prediction in cobalt free high-entropy alloys

High-entropy alloys (HEAs) are known for superb combination of performance attributes, making them ideal for advanced applications, e.g., nuclear engineering. The concept of cobalt-free HEAs aims to mitigate concerns about cobalt's radioactivity, however, predicting their phase formation remains challenging due to their complex compositions. In this work, we integrate six semiempirical parameters, i.e., mixing entropy (ΔSmix), mixing enthalpy (ΔHmix), atomic size difference (δ), valence electron concentration (VEC), d-orbital energy level (Md), and the Ω parameter, along with machine learning (ML) to predict the body-centered cubic phase stability in Co free HEAs. To address the limitations of experimental data, generative adversarial networks were used to augment the dataset, thus improving the accuracy of the Gaussian process classification model used for phase prediction. After dimensionality reduction to five principal components, the model achieved an accuracy of 84%, with ΔHmix and δ identified as the key descriptors influencing phase formation. This approach highlights the synergy of ML and data augmentation in accelerating the design of HEAs for advanced applications.

cond-mat.mtrl-sci↗

arXiv2026

Formalizing and validating properties in Asmeta with Large Language Models (Extended Abstract)

Writing temporal logic properties is often a challenging task for users of model-based development frameworks, particularly when translating informal requirements into formal specifications. In this paper, we explore the idea of integrating Large Language Models (LLMs) into the Asmeta framework to assist users during the definition, formalization, explanation, and validation of temporal properties. We present a workflow in which an LLM-based agent supports these activities by leveraging the Asmeta specification and the feedback produced by the model checker. This work serves as a proof of concept that illustrates the feasibility and potential benefits of such an integration through representative examples.

cs.SE↗

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

Recent Advances in Metallic Glasses

This paper reviews recent advances in the field of metallic glasses, focusing on the development of novel experimental techniques and in silico models. We discuss progress in experimental characterization, additive manufacturing, multiscale modeling approaches, and the growing role of machine learning in understanding and designing these complex materials. On the experimental side, we highlight measurements of thermophysical properties of supercooled liquids via fast chip calorimetry and enhancements in mechanical properties through rejuvenation treatments. This work underscores the crucial role of short-range order and medium-range order in controlling metallic glass mechanical properties. Recent progress in structural probes allows in situ observations of deformation mechanisms, positioning the field well to further advance our understanding of mechanical properties. Additive manufacturing of metallic glasses is discussed as one encouraging new manufacturing route for metallic glasses. We examine laser powder-bed fusion process physics and the central trade-off between amorphicity and densification, including heat affected zone devitrification and defects formation, together with emerging mitigation strategies and applications. On the theoretical and simulation side, we review advances in nanoscale, mesoscale, and continuum modeling of metallic glasses that have led to promising approaches by which multiscale schemes can incorporate data sourced from atomic-scale simulations. These efforts have helped to elucidate the connection between the glass structure and mechanical and rheological responses. We also cover the development of machine learning interatomic potentials for metallic glasses, along with machine learning driven prediction of glass forming ability and inverse design methods. Finally, challenges and directions for future research are presented and discussed.

cond-mat.mtrl-sci↗

arXiv2025

Fracture toughness and auxeticity in disordered metamaterials

Auxetic metamaterials are commonly thought to exhibit favorable mechanical properties, notably high energy absorption. Here we investigate disordered metamaterials obtained from random beam networks by optimizing simultaneously auxeticity and the energy absorbed before fracture. By giving different weights to these optimization targets, we demonstrate that the optimal configurations are connected along a Pareto front where high auxeticity implies comparatively low energy absorption and vice versa. We study the mechanical properties of the resulting metamaterials and characterize the different deformation modes obtained for distinct optimization targets. The simulation and optimization results are validated by comparison with the deformation behavior of additively manufactured samples. Our work provides an illustration of the potentials and limitations of multi-objective optimization in the design of disordered mechanical metamaterials

cond-mat.dis-nn↗

arXiv2025

Efficient training of machine learning potentials for metallic glasses: CuZrAl validation

Interatomic potentials are key to uncovering microscopic structure-property relationships, essential for multiscale simulations and high-throughput experiments. For metallic glasses, their disordered atomic structure makes the development of potentials particularly challenging, resulting in the scarcity of chemistry-specific parametrizations for this important class of materials. We address this gap by introducing an efficient methodology to design machine learning interatomic potentials (MLIPs), benchmarked on the CuZrAl system. Using a Lennard-Jones surrogate model, swap-Monte Carlo sampling, and single-point Density Functional Theory (DFT) corrections, we capture amorphous structures spanning 14 decades of supercooling. These representative configurations, competing with the experimental time scale, enable robust model training across diverse states, while minimizing the need for extensive DFT datasets. The resulting MLIP matches the experimental data and predictions of the classical embedded atom method (EAM) for structural, dynamical, energetic, and mechanical properties. This approach offers a scalable path to develop accurate MLIPs for complex metallic glasses, including emerging multi-component and high-entropy systems.

cond-mat.mtrl-sci↗

arXiv2025

Growth and prediction of plastic strain in metallic glasses

Predicting the failure and plasticity of solids remains a longstanding challenge, with broad implications for materials design and functional reliability. Disordered solids like metallic glasses can fail either abruptly or gradually without clear precursors, and the mechanical response depends strongly on composition, thermal history and deformation protocol -- impeding generalizable modeling. While deep learning methods offer predictive power, they often rely on numerous input parameters, hindering interpretability, methodology advancement and practical deployment. Here, we propose a macroscopic, physically grounded approach that uses plastic strain accumulation in the elastic regime to robustly predict deformation and yield. This method reduces complexity and improves interpretability, offering a practical alternative for disordered materials. For the Cu-Zr-(Al) metallic glasses prepared with varied annealing, we identify two limiting regimes of plastic strain growth: power-law in poorly annealed and exponential in well-annealed samples. A physics-informed framework with Bayesian inference extracts growth parameters from stress-strain data within $\sim$5\% strain, enabling early prediction of bulk response and yield point, well before the failure. The predictive performance improves with annealing, and bulk plasticity correlates with the microscopic plastic activity from scattered to growth near yielding. This work presents a physically interpretable and experimentally relevant framework for predicting plasticity and failure in metallic glasses from early mechanical response, offering both theoretical insights and practical tools for material characterization and design.

cond-mat.mtrl-sci↗

arXiv2025

Exploring the impact of Ti/Al on L12 nanoprecipitation and deformation behavior in CoNiFeAlTi multi-principal element alloys through atomistic simulations

Recent studies on CoNi-based multi-principal element alloys (MPEAs) have demonstrated high strength and ductility, attributed to the formation of stable L12 nanoscale precipitates. However, the fundamental mechanisms behind such impressive properties in these complex alloys are not well understood. In this work, we investigate the effects of Ti and Al concentrations on the formation of L12 precipitates in (CoNiFe)84(Al8Ti8), (CoNiFe)86(Al7Ti7), (CoNiFe)88(Al6Ti6), and (CoNiFe)94(Al4Ti2) MPEAs using hybrid molecular dynamics/Monte Carlo (MD/MC) simulations and a MEAM interatomic potential for the CoNiFeTiAl system. Additionally, we study the effect of L12 precipitation on the mechanical properties and stacking fault energy (SFE) of these MPEAs using MD. Our hybrid MD/MC simulations show that the (CoNiFe)86(Al7Ti7) alloy exhibits the highest amount of L12 nanoprecipitates. We find that L12 precipitation increases the SFE, with higher Al and Ti contents leading to greater increases. Tensile simulations reveal that L12 precipitates enhance yield strength, with alloys exhibiting higher precipitation showing increased flow stress. We also investigate dislocation-nanoprecipitate interactions with different precipitate sizes in the (CoNiFe)86(Al7Ti7) alloy. Larger nanoprecipitate sizes result in stronger dislocation pinning. Dislocations predominantly shear through 4-8 nm precipitates instead of looping around them (Orowan mechanism), enhancing strength while maintaining good ductility. Although the lattice mismatch between the L12 nanoprecipitate and the matrix is low (0.139%), the significant difference in SFE between the L12 nanoprecipitate and the matrix results in stronger dislocation pinning. This understanding can guide the design of MPEAs with tailored properties by controlling nanoscale precipitation.

cond-mat.mtrl-sci↗

arXiv2025

Uncertainty in the era of machine learning for atomistic modeling

The widespread adoption of machine learning surrogate models has significantly improved the scale and complexity of systems and processes that can be explored accurately and efficiently using atomistic modeling. However, the inherently data-driven nature of machine learning models introduces uncertainties that must be quantified, understood, and effectively managed to ensure reliable predictions and conclusions. Building upon these premises, in this Perspective, we first overview state-of-the-art uncertainty estimation methods, from Bayesian frameworks to ensembling techniques, and discuss their application in atomistic modeling. We then examine the interplay between model accuracy, uncertainty, training dataset composition, data acquisition strategies, model transferability, and robustness. In doing so, we synthesize insights from the existing literature and highlight areas of ongoing debate.

physics.chem-ph↗

arXiv2024

Bayesian exploration of the composition space of CuZrAl metallic glasses for mechanical properties

Designing metallic glasses in silico is a major challenge in materials science given their disordered atomic structure and the vast compositional space to explore. Here, we tackle this challenge by finding optimal compositions for target mechanical properties. We apply Bayesian exploration for the CuZrAl composition, a paradigmatic metallic glass known for its good glass forming ability. We exploit an automated loop with an online database, a Bayesian optimization algorithm, and molecular dynamics simulations. From the ubiquitous 50/50 CuZr starting point, we map the composition landscape changing the ratio of elements and adding aluminium to characterize the yield stress and the shear modulus. This approach demonstrates with relatively modest effort that the system has an optimal composition window for the yield stress around aluminium concentration $c_{\rm Al} = 15$ \% and zirconium concentration $c_{\rm Zr} = 30$ %. We also explore several cooling rates ("process parameters") and find that the best mechanical properties for a composition result from being most affected by the cooling procedure. Our Bayesian approach paves the novel way for the design of metallic glasses with "small data", with an eye toward both future in silico design and experimental applications exploiting this toolbox.

cond-mat.mtrl-sci↗

arXiv2024

Avalanches in CuZrAl metallic glasses

Metallic glasses have mechanical properties, which exhibit avalanches in the disguise of stress drops. We study these phenomena in a classical metallic glass system CuZrAl by athermal quasistatic shear and varying the element concentrations and for pure CuZr 50/50 case the cooling rate. The resulting mechanical properties are close to the behaviour found experimentally. At small strains, the pristine systems are akin to other glassy systems with a so-called gap distribution with a small positive exponent. Critical avalanching behaviour is found only approaching the yield point. The post-yield stress drops are universal, and the gap distribution becomes flat.

cond-mat.stat-mech↗

arXiv2023

Quasi-localized modes in crystalline high entropy alloys

High Entropy Alloys (HEAs) are designed by mixing multiple metallic species in nearly the same amount to obtain crystalline or amorphous materials with exceptional mechanical properties. Here we use molecular dynamics simulations to investigate the role of positional and compositional disorder in determining the low-frequency vibrational properties of CrMnFeCoNi HEAs. Our results show that the expected dependence of the density of states on the frequency as $D(ω)\simω^4$ is recovered for amorphous HEAs and is also observed for partially crystallized alloys with deviations that depend on the degree of crystallization. We find that the quasi-localized vibrations are still visible in crystalline HEAs, albeit suppressed compared to the corresponding amorphous alloys. Our work offers a unified perspective to describe HEA mechanical properties in terms of their vibrational density of states.

cond-mat.dis-nn↗

arXiv2022

Universal Density of Low Frequency States in Silica Glass at Finite Temperatures

The theoretical understanding of the low-frequency modes in amorphous solids at finite temperature is still incomplete. The study of the relevant modes is obscured by the dressing of inter-particle forces by collision-induced momentum transfer that is unavoidable at finite temperatures. Recently, it was proposed that low frequency modes of vibrations around the {\em thermally averaged} configurations deserve special attention. In simple model glasses with bare binary interactions, these included quasi-localized modes whose density of states appears to be universal, depending on the frequencies as $D(ω) \sim ω^4$, in agreement with the similar law that is obtained with bare forces at zero temperature. In this work, we report investigations of a model of silica glass at finite temperature; here the bare forces include binary and ternary interactions. Nevertheless we can establish the validity of the universal law of the density of quasi-localized modes also in this richer and more realistic model glass.

cond-mat.dis-nn↗

arXiv2022

Predicting the failure of two-dimensional silica glasses

Being able to predict the failure of materials based on structural information is a fundamental issue with enormous practical and industrial relevance for the monitoring of devices and components. Thanks to recent advances in deep learning, accurate failure predictions are becoming possible even for strongly disordered solids, but the sheer number of parameters used in the process renders a physical interpretation of the results impossible. Here we address this issue and use machine learning methods to predict the failure of simulated two dimensional silica glasses from their initial undeformed structure. We then exploit Gradient-weighted Class Activation Mapping (Grad-CAM) to build attention maps associated with the predictions, and we demonstrate that these maps are amenable to physical interpretation in terms of topological defects and local potential energies. We show that our predictions can be transferred to samples with different shape or size than those used in training, as well as to experimental images. Our strategy illustrates how artificial neural networks trained with numerical simulation results can provide interpretable predictions of the behavior of experimentally measured structures.

cond-mat.mtrl-sci↗

arXiv2021

Developing a Prototype of a Mechanical Ventilator Controller from Requirements to Code with ASMETA

Rigorous development processes aim to be effective in developing critical systems, especially if failures can have catastrophic consequences for humans and the environment. Such processes generally rely on formal methods, which can guarantee, thanks to their mathematical foundation, model preciseness, and properties assurance. However, they are rarely adopted in practice. In this paper, we report our experience in using the Abstract State Machine formal method and the ASMETA framework in developing a prototype of the control software of the MVM (Mechanical Ventilator Milano), a mechanical lung ventilator that has been designed, successfully certified, and deployed during the COVID-19 pandemic. Due to time constraints and lack of skills, no formal method was applied for the MVM project. However, we here want to assess the feasibility of developing (part of) the ventilator by using a formal method-based approach. Our development process starts from a high-level formal specification of the system to describe the MVM main operation modes. Then, through a sequence of refined models, all the other requirements are captured, up to a level in which a C++ implementation of a prototype of the MVM controller is automatically generated from the model, and tested. Along the process, at each refinement level, different model validation and verification activities are performed, and each refined model is proved to be a correct refinement of the previous level. By means of the MVM case study, we evaluate the effectiveness and usability of our formal approach.

cs.FL↗

arXiv2020

Universal low-frequency vibrational modes in silica glasses

It was recently shown that different simple models of glass formers with binary interactions define a universality class in terms of the density of states of their quasi-localized low-frequency modes. Explicitly, once the hybridization with standard Debye (extended) modes is avoided, a number of such models exhibit a universal density of state, depending on the mode frequencies as $D(ω) \sim ω^4$. It is unknown however how wide is this universality class, and whether it also pertains to more realistic models of glass formers. To address this issue we present analysis of the quasi-localized modes in silica, a network glass which has both binary and ternary interactions. We conclude that in 3-dimensions silica exhibits the very same frequency dependence at low frequencies, suggesting that this universal form is a generic consequence of amorphous glassiness.

cond-mat.soft↗