arXiv subjects

Luca Leoni

Publications and source records attributed to Luca Leoni.

5 recordsLinked to original sources

arXiv2026

Machine-learned dynamics of surface polarons at reduced oxide surfaces

Reducible oxides exhibit a rich interplay of electronic, structural, and chemical properties that underpins applications in catalysis, photovoltaics, batteries, and energy storage. This interplay is strongly shaped by excess electrons, often introduced by oxygen vacancies, that localize as small polarons and influence charge transport and surface chemistry. At surfaces, these polarons play a central role in charge localization, mobility, and reactivity, yet their finite-temperature dynamics remain difficult to access from first principles due to the long time scales needed to adequately sample polaron's hopping. To overcome this limitation, we extend machine-learning-assisted polaron dynamics to redox-active oxide surfaces, using oxygen-deficient rutile TiO$_2$(110) as a paradigmatic case. By accessing several nanoseconds of dynamics over a range of temperatures, we show that small-polaron mobility at the reduced rutile TiO$_2$(110) surface is suppressed by several orders of magnitude relative to the corresponding bulk material, providing a microscopic interpretation of the lower electron mobilities observed in porous rutile TiO$_2$ compared with single-crystal samples. This suppressed mobility arises from the loss of favourable hopping pathways: surface polaron motion is largely confined to planar inter-row trajectories within the second topmost layers, with only rare interlayer hopping events. These results establish a transferable machine-learning strategy for investigating polaron dynamics in reducible oxides.

cond-mat.mtrl-sci↗

arXiv2026

Machine Learning the order-disorder Jahn-Teller transition in LaMnO$_3$

We investigate the Jahn-Teller structural phase transition in LaMnO$_3$ at $T_{JT} \simeq 750$ K using molecular dynamics simulations based on machine-learning force fields trained on ab initio data. Analysis of the site-site correlation function of the distortions reveals that the transition is driven by the ordering of the $Q_2$ Jahn-Teller distortion of the MnO$_6$ octahedra, which acts as the order parameter and establishes the order-disorder nature of the transition. Dynamical local distortions are found to persist above $T_{JT}$. Our results reproduce the experimental temperature dependence of both structural and phonon properties and highlight the presence of anharmonic effects at finite temperature. More broadly, the combined use of machine-learning molecular dynamics and velocity autocorrelation function analysis provides a robust framework for uncovering the microscopic mechanisms of structural phase transitions in correlated materials. In particular, this approach enables a clear distinction between order-disorder transitions and alternative mechanisms, such as displacive behavior, through the temperature evolution of vibrational properties.

cond-mat.stat-mech↗

arXiv2024

Machine Learning Small Polaron Dynamics

Polarons are crucial for charge transport in semiconductors, significantly impacting material properties and device performance. The dynamics of small polarons can be investigated using first-principles molecular dynamics (FPMD). However, the limited timescale of these simulations presents a challenge for adequately sampling infrequent polaron hopping events. Here, we introduce a message-passing neural network combined with FPMD within the Born-Oppenheimer approximation, that learns the polaronic potential energy surface by encoding the polaronic state, allowing for simulations of polaron hopping dynamics at the nanosecond scale. By leveraging the statistical significance of the long timescale, our framework can accurately estimate polaron (anisotropic) mobilities and activation barriers in prototypical polaronic oxides across different scenarios (hole polarons in rocksalt MgO and electron polarons in pristine and F-doped rutile TiO$_2$) within experimentally measured ranges.

cond-mat.mtrl-sci↗

arXiv2024

Hydrogen Diffusion in Magnesium Using Machine Learning Potentials: a comparative study

Understanding and accurately predicting hydrogen diffusion in materials is challenging due to the complex interactions between hydrogen defects and the crystal lattice. These interactions span large length and time scales, making them difficult to address with standard ab initio techniques. This work addresses this challenge by employing accelerated machine learning (ML) molecular dynamics simulations through active learning. We conduct a comparative study of different ML-based interatomic potential schemes, including VASP, MACE, and CHGNet, utilizing various training strategies such as on-the-fly learning, pre-trained universal models, and fine-tuning.By considering different temperatures and concentration regimes, we obtain hydrogen diffusion coefficients and activation energy values which align remarkably well with experimental results, underlining the efficacy and accuracy of ML-assisted methodologies in the context of diffusive dynamics. Particularly, our procedure significantly reduces the computational effort associated with traditional transition state calculations or ad-hoc designed interatomic potentials. The results highlight the limitations of pre-trained universal solutions for defective materials and how they can be improved by fine-tuning. Specifically, fine-tuning the models on a database produced during on-the-fly training of VASP ML force-field allows the retrieving of DFT-level accuracy at a fraction of the computational cost.

cond-mat.mtrl-sci↗

arXiv2024

Global sampling of Feynman's diagrams through Normalizing Flow

Normalizing Flows (NF) are powerful generative models with increasing applications in augmenting Monte Carlo algorithms due to their high flexibility and expressiveness. In this work we explore the integration of NF in Diagrammatic Monte Carlo (DMC), presenting an architecture designed to sample the intricate multidimensional space of Feynman's diagrams through dimensionality reduction. By decoupling the sampling of diagram order and interaction times, the flow focuses on one interaction at a time. This enables constructing a general diagram by employing the same unsupervised model iteratively, dressing a zero-order diagram with interactions determined by the previously sampled order. The resulting NF-augmented DMC is tested on the widely used single-site Holstein polaron model in the entire electron-phonon coupling regime. The obtained data show that the model accurately reproduces the diagram distribution reducing sample correlation and statistical error for observables such as the polaron binding energy and the interacting Green's function.

cond-mat.str-el↗