arXiv subjects

Atsuto Seko

Publications and source records attributed to Atsuto Seko.

At least 19 recordsLinked to original sources

arXiv2026

Global structure searches under varying temperatures and pressures using polynomial machine learning potentials: A case study on silicon

Polynomial machine learning potentials (MLPs) based on polynomial rotational invariants have been systematically developed for various systems and applied to efficiently predict crystal structures. In this study, we propose a robust methodology founded on polynomial MLPs to comprehensively enumerate crystal structures under high-pressure conditions and to evaluate their phase stability at finite temperatures. The proposed approach involves constructing polynomial MLPs with high predictive accuracy across a broad range of pressures, conducting reliable global structure searches, and performing exhaustive self-consistent phonon calculations. We demonstrate the effectiveness of this approach by examining elemental silicon at pressures up to 100 GPa and temperatures up to 1000 K, revealing stable phases across these conditions. The framework established in this study offers a powerful strategy for predicting crystal structures and phase stability under high-pressure and finite-temperature conditions.

cond-mat.mtrl-sci↗

arXiv2025

AI4X Roadmap: Artificial Intelligence for the advancement of scientific pursuit and its future directions

Artificial intelligence and machine learning are reshaping how we approach scientific discovery, not by replacing established methods but by extending what researchers can probe, predict, and design. In this roadmap we provide a forward-looking view of AI-enabled science across biology, chemistry, climate science, mathematics, materials science, physics, self-driving laboratories and unconventional computing. Several shared themes emerge: the need for diverse and trustworthy data, transferable electronic-structure and interatomic models, AI systems integrated into end-to-end scientific workflows that connect simulations to experiments and generative systems grounded in synthesisability rather than purely idealised phases. Across domains, we highlight how large foundation models, active learning and self-driving laboratories can close loops between prediction and validation while maintaining reproducibility and physical interpretability. Taken together, these perspectives outline where AI-enabled science stands today, identify bottlenecks in data, methods and infrastructure, and chart concrete directions for building AI systems that are not only more powerful but also more transparent and capable of accelerating discovery in complex real-world environments.

physics.soc-ph↗

arXiv2025

Systematic global structure search of bismuth-based binary systems under pressure using machine learning potentials

Machine learning potentials (MLPs) have significantly advanced global crystal structure prediction by enabling efficient and accurate property evaluations. In this study, global structure searches are performed for 11 bismuth-based binary systems, including Na-Bi, Ca-Bi, and Eu-Bi, under pressures ranging from 0 to 20 GPa, employing polynomial MLPs developed specifically for these systems. The searches reveal numerous compounds not previously reported in the literature and identify all experimentally known compounds that are representable within the explored configurational space. These results highlight the robustness and reliability of the current MLP-based structure search. The study provides valuable insights into the discovery and design of novel bismuth-based materials under both ambient and high-pressure conditions.

cond-mat.mtrl-sci↗

arXiv2024

Projector-based efficient estimation of force constants

Estimating force constants for crystal structures is crucial for calculating various phonon-related properties. However, this task becomes particularly challenging when dealing with a large number of atoms or when third- and higher-order force constants are required. In this study, we propose an efficient approach that involves constructing a complete orthonormal basis set for the force constants. We formulate this approach using projection matrices and their eigenvectors to meet the requirements for the force constants. This basis set enables precise inference of the force constants from displacement-force datasets. Our efficient algorithms for basis-set construction and force constant estimation are implemented in the symfc code. Furthermore, several applications demonstrated in this study indicate that the current approach facilitates the efficient and accurate determination of force constants.

cond-mat.mtrl-sci↗

arXiv2024

Globally-stable and metastable crystal structure enumeration using polynomial machine learning potentials in elemental As, Bi, Ga, In, La, P, Sb, Sn, and Te

Machine learning potentials (MLPs) have become indispensable for conducting accurate large-scale atomistic simulations and for the efficient prediction of crystal structures. Polynomial MLPs, defined by polynomial rotational invariants, have been systematically developed for a wide range of elemental, alloy, and ionic systems. This study introduces a highly efficient and robust methodology for enumerating globally stable and metastable structures utilizing polynomial MLPs. To develop MLPs that are sufficiently robust for structure enumeration, an iterative process involving random structure searches and subsequent updates to the MLPs is employed. This methodology has been systematically applied to elemental systems such as As, Bi, Ga, In, La, P, Sb, Sn, and Te, where numerous local minima present significant competition with the global minimum in terms of energy. The proposed approach facilitates robust global structure searches and structure enumerations by markedly accelerating the search process and expanding the search space.

cond-mat.mtrl-sci↗

arXiv2024

Polynomial machine learning potential and its application to global structure search in the ternary Cu-Ag-Au alloy

Machine learning potentials (MLPs) have become indispensable for performing accurate large-scale atomistic simulations and predicting crystal structures. This study introduces the development of a polynomial MLP specifically for the ternary Cu-Ag-Au system. The MLP is formulated as a polynomial of polynomial invariants that remain unchanged under any rotation. The polynomial MLP facilitates not only comprehensive global structure searches within the Cu-Ag-Au alloy system but also reliable predictions of a wide variety of properties across the entire composition range. The developed MLP supports highly accurate and efficient atomistic simulations, thereby significantly advancing the understanding of the Cu-Ag-Au system. Furthermore, the methodology demonstrated in this study can be easily applied to other ternary alloy systems.

cond-mat.mtrl-sci↗

arXiv2024

Predictive power of polynomial machine learning potentials for liquid states in 22 elemental systems

The polynomial machine learning potentials (MLPs) described by polynomial rotational invariants have been systematically developed for various systems and used in diverse applications in crystalline states. In this study, we systematically investigate the predictive power of the polynomial MLPs for liquid structural properties in 22 elemental systems with diverse chemical bonding properties, including those showing anomalous melting behavior, such as Si, Ge, and Bi. We compare liquid structural properties obtained from molecular dynamics simulations using the density functional theory (DFT) calculation, the polynomial MLPs, and other interatomic potentials in the literature. The current results demonstrate that the polynomial MLPs consistently exhibit high predictive power for liquid structural properties with the same accuracy as that of typical DFT calculations.

cond-mat.mtrl-sci↗

arXiv2024

On-the-fly training of polynomial machine learning potentials in computing lattice thermal conductivity

The application of first-principles calculations for predicting lattice thermal conductivity (LTC) in crystalline materials, in conjunction with the linearized phonon Boltzmann equation, has gained increasing popularity. In this calculation, the determination of force constants through first-principles calculations is critical for accurate LTC predictions. For material exploration, performing first-principles LTC calculations in a high-throughput manner is now expected, although it requires significant computational resources. To reduce computational demands, we integrated polynomial machine learning potentials on-the-fly during the first-principles LTC calculations. This paper presents a systematic approach to first-principles LTC calculations. We designed and optimized an efficient workflow that integrates multiple modular software packages. We applied this approach to calculate LTCs for 103 compounds of the wurtzite, zincblende, and rocksalt types to evaluate the performance of the polynomial machine learning potentials in LTC calculations. We demonstrate a significant reduction in the computational resources required for the LTC predictions.

cond-mat.mtrl-sci↗

arXiv2022

Systematic development of polynomial machine learning potentials for metallic and alloy systems

Machine learning potentials (MLPs) developed from extensive datasets constructed from density functional theory (DFT) calculations have become increasingly appealing for many researchers. This paper presents a framework of polynomial-based MLPs, called polynomial MLPs. The systematic development of accurate and computationally efficient polynomial MLPs for many elemental and binary alloy systems and their predictive powers for various properties are also demonstrated. Consequently, many polynomial MLPs are available in a repository website. The repository will help many scientists perform accurate and efficient large-scale atomistic simulations and crystal structure searches.

cond-mat.mtrl-sci↗

arXiv2021

Finding well-optimized special quasirandom structures with decision diagram

The advanced data structure of the zero-suppressed binary decision diagram (ZDD) enables us to efficiently enumerate nonequivalent substitutional structures. Not only can the ZDD store a vast number of structures in a compressed manner, but also can a set of structures satisfying given constraints be extracted from the ZDD efficiently. Here, we present a ZDD-based efficient algorithm for exhaustively searching for special quasirandom structures (SQSs) that mimic the perfectly random substitutional structure. We demonstrate that the current approach can extract only a tiny number of SQSs from a ZDD composed of many substitutional structures (>$10^{12}$). As a result, we find SQSs that are optimized better than those proposed in the literature. A series of SQSs should be helpful for estimating the properties of substitutional solid solutions. Furthermore, the present ZDD-based algorithm should be useful for applying the ZDD to the other structure enumeration problems.

physics.comp-ph↗

arXiv2021

Structure and lattice thermal conductivity of grain boundaries in silicon by using machine learning potential and molecular dynamics

In silicon, lattice thermal conductivity plays an important role in a wide range of applications such as thermoelectric and microelectronic devices. Grain boundaries (GBs) in polycrystalline silicon can significantly reduce lattice thermal conductivity, but the impact of GB atomic structures on it remains to be elucidated. This study demonstrates accurate predictions of the GB structures, GB energies, and GB phonon properties in silicon using machine learning potentials (MLPs). The results indicate that the MLPs enable robust GB structure searches owing to the fact that the MLPs were developed from a training dataset covering a wide variety of structures. We also investigate lattice thermal conduction at four GB atomic structures using large-scale perturbed molecular dynamics and phonon wave-packet simulations. The comparison of these results indicates that the GB structure dependence of thermal conductivity stems from anharmonic vibrations at GBs rather than from the phonon transmission behavior at GBs. The advantages of the MLPs compared with a typical empirical potential of silicon are also thoroughly investigated.

physics.comp-ph↗

arXiv2020

Machine learning potentials for multicomponent systems: The Ti-Al binary system

Machine learning potentials (MLPs) are becoming powerful tools for performing accurate atomistic simulations and crystal structure optimizations. An approach to developing MLPs employs a systematic set of polynomial invariants including high-order ones to represent the neighboring atomic density. In this study, a formulation of the polynomial invariants is extended to the case of multicomponent systems. The extended formulation is more complex than the formulation for elemental systems. This study also shows its application to Ti-Al binary system. As a result, an MLP with the lowest error and MLPs with high computational cost performance are selected from the many MLPs developed systematically. The predictive powers of the developed MLPs for many properties, such as the formation energy, elastic constants, thermodynamic properties, and mechanical properties, are examined. The MLPs exhibit high predictive power for the properties in a wide variety of ordered structures. The present scheme should be systematically applicable to other multicomponent systems.

physics.comp-ph↗

arXiv2020

Application of machine learning potentials to predict grain boundary properties in fcc elemental metals

Accurate interatomic potentials are in high demand for large-scale atomistic simulations of materials that are prohibitively expensive by density functional theory (DFT) calculation. In this study, we apply machine learning potentials in a recently constructed repository to the prediction of the grain boundary energy in face-centered-cubic elemental metals, i.e., Ag, Al, Au, Cu, Pd, and Pt. The systematic application of machine learning potentials shows that they enable us to predict grain boundary structures and their energies accurately. The grain boundary energies predicted by the MLPs are in agreement with those calculated by DFT, although no grain boundary structures were included in training datasets of the present MLPs.

physics.comp-ph↗

arXiv2020

Machine Learning Potential Repository

This paper introduces a machine learning potential repository that includes Pareto optimal machine learning potentials. It also shows the systematic development of accurate and fast machine learning potentials for a wide range of elemental systems. As a result, many Pareto optimal machine learning potentials are available in the repository from a website. Therefore, the repository will help many scientists to perform accurate and fast atomistic simulations.

physics.comp-ph↗

arXiv2020

Derivative structure enumeration using binary decision diagram

A derivative structure is a nonequivalent substitutional atomic configuration derived from a given primitive cell. The enumeration of derivative structures plays an essential role in searching for the ground states in multicomponent systems. However, it is computationally hard to enumerate derivative structures if the number of derivative structures of a target system becomes huge. In the present study, we introduce a novel compact data structure of the zero-suppressed binary decision diagram (ZDD) to enumerate derivative structures much more efficiently. The present study shows its simple applications to the enumeration of structures derived from the face-centered cubic and hexagonal close-packed lattices in binary, ternary, and quaternary systems. The present ZDD-based procedure should significantly contribute not only to various computational approaches based on the derivative structures but also to a wide range of combinatorial issues in physics and materials science.

physics.comp-ph↗

arXiv2020

Prediction of perovskite-related structures in ACuO$_{3-x}$ (A $=$ Ca, Sr, Ba, Sc, Y, La) using density functional theory and Bayesian optimization

Oxygen vacancy ordering in perovskite-type transition-metal oxides plays an important role in the emergence of exotic electronic properties, as typified by superconducting cuprates. In this study, we predict the stability of oxygen-deficient perovskite structures in ACuO$_{3-x}$ (A $=$ Ca, Sr, Ba, Sc, Y, La) by density functional theory calculation. We introduce a combination of the cluster expansion method, Gaussian process, and Bayesian optimization to find stable oxygen-deficient structures among a considerable number of candidates. Our calculations not only reproduce the reported structures but suggest the presence of several unknown oxygen-deficient perovskite structures, some of which are stabilized at high pressures. This work demonstrates the great applicability of the present computational procedure for the elucidation of the structural stability of strongly correlated oxides with a large tolerance to oxygen deficiency.

physics.comp-ph↗

arXiv2019

Group-theoretical high-order rotational invariants for structural representations: Application to linearized machine learning interatomic potential

Many rotational invariants for crystal structure representations have been used to describe the structure-property relationship by machine learning. The machine learning interatomic potential (MLIP) is one of the applications of rotational invariants, which provides the relationship between the energy and the crystal structure. Since the MLIP requires the highest accuracy among machine learning estimations of the structure-property relationship, the enumeration of rotational invariants is useful for constructing MLIPs with the desired accuracy. In this study, we introduce high-order linearly independent rotational invariants up to the sixth order based on spherical harmonics and apply them to linearized MLIPs for elemental aluminum. A set of rotational invariants is derived by the general process of reducing the Kronecker products of irreducible representations (Irreps) for the SO(3) group using a group-theoretical projector method. A high predictive power for a wide range of structures is accomplished by using high-order invariants with low-order invariants equivalent to pair and angular structural features.

physics.comp-ph↗

arXiv2018

Exploring a potential energy surface by machine learning for characterizing atomic transport

We propose a machine-learning method for evaluating the potential barrier governing atomic transport based on the preferential selection of dominant points for the atomic transport. The proposed method generates numerous random samples of the entire potential energy surface (PES) from a probabilistic Gaussian process model of the PES, which enables defining the likelihood of the dominant points. The robustness and efficiency of the method are demonstrated on a dozen model cases for proton diffusion in oxides, in comparison with a conventional nudge elastic band method.

physics.chem-ph↗