arXiv subjects

Atsuto Seko

Publications and source records attributed to Atsuto Seko.

At least 37 records · Page 2Linked to original sources

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↗

arXiv2017

Compositional descriptor-based recommender system accelerating the materials discovery

Structures and properties of many inorganic compounds have been collected historically. However, it only covers a very small portion of possible inorganic crystals, which implies the presence of numerous currently unknown compounds. A powerful machine-learning strategy is mandatory to discover new inorganic compounds from all chemical combinations. Herein we propose a descriptor-based recommender-system approach to estimate the relevance of chemical compositions where stable crystals can be formed [i.e., chemically relevant compositions (CRCs)]. As well as data-driven compositional similarity used in the literature, the use of compositional descriptors as a prior knowledge can accelerate the discovery of new compounds. We validate our recommender systems in two ways. Firstly, one database is used to construct a model, while another is used for the validation. Secondly, we estimate the phase stability for compounds at expected CRCs using density functional theory calculations.

cond-mat.mtrl-sci↗

arXiv2017

Linearized machine-learning interatomic potentials for non-magnetic elemental metals: Limitation of pairwise descriptors and trend of predictive power

Machine-learning interatomic potential (MLIP) has been of growing interest as a useful method to describe the energetics of systems of interest. In the present study, we examine the accuracy of linearized pairwise MLIPs and angular-dependent MLIPs for 31 elemental metals. Using all of the optimal MLIPs for 31 elemental metals, we show the robustness of the linearized frameworks, the general trend of the predictive power of MLIPs and the limitation of pairwise MLIPs. As a result, we obtain accurate MLIPs for all 31 elements using the same linearized framework. This indicates that the use of numerous descriptors is the most important practical feature for constructing MLIPs with high accuracy. An accurate MLIP can be constructed using only pairwise descriptors for most non-transition metals, whereas it is very important to consider angular-dependent descriptors when expressing interatomic interactions of transition metals.

cond-mat.mtrl-sci↗

arXiv2017

Matrix- and tensor-based recommender systems for the discovery of currently unknown inorganic compounds

Chemically relevant compositions (CRCs) and atomic arrangements of inorganic compounds have been collected as inorganic crystal structure databases. Machine-learning is a unique approach to search for currently unknown CRCs from vast candidates. Herein we propose matrix- and tensor-based recommender system approaches to predict currently unknown CRCs from database entries of CRCs. Firstly, the performance of the recommender system approaches to discover currently unknown CRCs is examined. A Tucker decomposition recommender system shows the best discovery rate of CRCs as the majority of the top 100 recommended ternary and quaternary compositions correspond to CRCs. Secondly, systematic density functional theory (DFT) calculations are performed to investigate the phase stability of the recommended currently unknown CRCs. The phase stability of the 27 compositions reveals that 23 currently unknown compounds are stable. These results indicate that the recommender system has great potential to accelerate the discovery of new compounds.

cond-mat.mtrl-sci↗

arXiv2017

Descriptors for Machine Learning of Materials Data

Descriptors, which are representations of compounds, play an essential role in machine learning of materials data. Although many representations of elements and structures of compounds are known, these representations are difficult to use as descriptors in their unchanged forms. This chapter shows how compounds in a dataset can be represented as descriptors and applied to machine-learning models for materials datasets.

cond-mat.mtrl-sci↗

arXiv2017

Conceptual and practical bases for the high accuracy of machine learning interatomic potential

Machine learning interatomic potentials (MLIPs) based on a large dataset obtained by density functional theory (DFT) calculation have been developed recently. This study gives both conceptual and practical bases for the high accuracy of MLIPs, although MLIPs have been considered to be simply an accurate black-box description of atomic energy. We also construct the most accurate MLIP of the elemental Ti ever reported using a linearized MLIP framework and many angular-dependent descriptors, which also corresponds to a generalization of the modified embedded atom method (MEAM) potential.

cond-mat.mtrl-sci↗

arXiv2017

Temperature-dependent phonon spectra of magnetic random solid solutions

A first-principles-based method for computing phonons of magnetic random solid solutions including thermal magnetic fluctuations is developed. The method takes fluctuations of force constants (FCs) due to magnetic excitations as well as due to chemical disorder into account. The developed approach correctly predicts the experimentally observed unusual phonon hardening of a transverse acoustic mode in Fe-Pd an Fe-Pt Invar alloys with increasing temperature. This peculiar behavior, which cannot be explained within a conventional, harmonic picture, turns out to be a consequence of thermal magnetic fluctuations.

cond-mat.mtrl-sci↗

arXiv2016

Representation of compounds for machine-learning prediction of physical properties

The representations of a compound, called "descriptors" or "features", play an essential role in constructing a machine-learning model of its physical properties. In this study, we adopt a procedure for generating a systematic set of descriptors from simple elemental and structural representations. First it is applied to a large dataset composed of the cohesive energy for about 18000 compounds computed by density functional theory (DFT) calculation. As a result, we obtain a kernel ridge prediction model with a prediction error of 0.041 eV/atom, which is close to the "chemical accuracy" of 1 kcal/mol (0.043 eV/atom). The procedure is also applied to two smaller datasets, i.e., a dataset of the lattice thermal conductivity (LTC) for 110 compounds computed by DFT calculation and a dataset of the experimental melting temperature for 248 compounds. We examine the performance of the descriptor sets on the efficiency of Bayesian optimization in addition to the accuracy of the kernel ridge regression models. They exhibit good predictive performances.

cond-mat.mtrl-sci↗

arXiv2016

Mode-decomposition based on crystallographic symmetry in the band-unfolding method

The band-unfolding method is widely used to calculate the effective band structures of a disordered system from its supercell model. The unfolded band structures show the crystallographic symmetry of the underlying structure, where the difference of chemical components and the local atomic relaxation are ignored. It has been, however, still difficult to decompose the unfolded band structures into the modes based on the crystallographic symmetry of the underlying structure, and therefore detailed analyses of the unfolded band structures have been restricted. In this study, we develop a procedure to decompose the unfolded band structures according to the small representations (SRs) of the little groups. For this purpose, we derive the projection operators for SRs based on the group representation theory. We also introduce another type of projection operators for chemical elements, which enables us to decompose the unfolded band structures into the contributions of different combinations of chemical elements. Using the current method, we investigate the phonon band structure of disordered face-centered cubic Cu0.75Au0.25, which has large variations of atomic masses and force constants among the atomic sites due to the chemical disorder. We find several peculiar behaviors such as discontinuous and split branches in the unfolded phonon band structure. They are found to occur because different combinations of the chemical elements contribute to different regions of frequency.

cond-mat.mtrl-sci↗

arXiv2015

A machine learning-based selective sampling procedure for identifying the low energy region in a potential energy surface: a case study on proton conduction in oxides

In this paper, we propose a selective sampling procedure to preferentially evaluate a potential energy surface (PES) in a part of the configuration space governing a physical property of interest. The proposed sampling procedure is based on a machine learning method called the Gaussian process (GP), which is used to construct a statistical model of the PES for identifying the region of interest in the configuration space. We demonstrate the efficacy of the proposed procedure for atomic diffusion and ionic conduction, specifically the proton conduction in a well-studied proton-conducting oxide, barium zirconate BaZrO3. The results of the demonstration study indicate that our procedure can efficiently identify the low-energy region characterizing the proton conduction in the host crystal lattice, and that the descriptors used for the statistical PES model have a great influence on the performance.

cond-mat.mtrl-sci↗

arXiv2015

Prediction model of band-gap for AX binary compounds by combination of density functional theory calculations and machine learning techniques

Machine learning techniques are applied to make prediction models of the G0W0 band-gaps for 156 AX binary compounds using Kohn-Sham band-gaps and other fundamental information of constituent elements and crystal structure as predictors. Ordinary least square regression (OLSR), least absolute shrinkage and selection operator (LASSO) and non-linear support vector regression (SVR) methods are applied with several levels of predictor sets. When the Kohn-Sham band-gap by GGA (PBE) or modified Becke-Johnson (mBJ) is used as a single predictor, OLSR model predicts the G0W0 band-gap of a randomly selected test data with the root mean square error (RMSE) of 0.54 eV. When Kohn-Sham band gap by PBE and mBJ methods are used together with a set of various forms of predictors representing constituent elements and crystal structures, RMSE decreases significantly. The best model by SVR yields the RMSE of 0.18 eV. A large set of band-gaps estimated in this way should be useful as predictors for materials exploration.

cond-mat.mtrl-sci↗

arXiv2015

Discovery of low thermal conductivity compounds with first-principles anharmonic lattice dynamics calculations and Bayesian optimization

Compounds of low lattice thermal conductivity (LTC) are essential for seeking thermoelectric materials with high conversion efficiency. Some strategies have been used to decrease LTC. However, such trials have yielded successes only within a limited exploration space. Here we report the virtual screening of a library containing 54,779 compounds. Our strategy is to search the library through Bayesian optimization using for the initial data the LTC obtained from first-principles anharmonic lattice dynamics calculations for a set of 101 compounds. We discovered 221 materials with very low LTC. Two of them have even an electronic band gap < 1 eV, what makes them exceptional candidates for thermoelectric applications. In addition to those newly discovered thermoelectric materials, the present strategy is believed to be powerful for many other applications in which chemistry of materials are required to be optimized.

cond-mat.mtrl-sci↗

arXiv2015

First-principles interatomic potentials for ten elemental metals via compressed sensing

Interatomic potentials have been widely used in atomistic simulations such as molecular dynamics. Recently, frameworks to construct accurate interatomic potentials that combine a systematic set of density functional theory (DFT) calculations with machine learning techniques have been proposed. One of these methods is to use compressed sensing to derive a sparse representation for the interatomic potential. This facilitates the control of the accuracy of interatomic potentials. In this study, we demonstrate the applicability of compressed sensing to deriving the interatomic potential of ten elemental metals, namely, Ag, Al, Au, Ca, Cu, Ga, In, K, Li and Zn. For each elemental metal, the interatomic potential is obtained from DFT calculations using elastic net regression. The interatomic potentials are found to have prediction errors of less than 3.5 meV/atom, 0.03 eV/Å and 0.15 GPa for the energy, force and the stress tensor, respectively, which enable the accurate prediction of physical properties such as lattice constants and the phonon dispersion relationship.

cond-mat.mtrl-sci↗

arXiv2014

Phonon softening in paramagnetic bcc Fe and relationship with pressure-induced phase transition

Structural stability of paramagnetic (PM) body-centered cubic (bcc) Fe under pressure is investigated based on first-principles phonon calculations. Spin configurations of the PM phase are approximated using a binary special quasi-random structure (SQS) with a supercell approach. The behavior of phonon modes can be associated with pressure-induced phase transitions to the face-centered cubic (fcc) and hexagonal close-packed (hcp) structures as follows: For the PM phase, it is found that the low-frequency transverse mode at the N point (N$_4^-$ mode), which corresponds to a bcc-hcp phase transition pathway, exhibits strong softening under isotropic volume compression. The frequency of this mode becomes zero by $2\%$ volume decrease within the harmonic approximation. This result is not consistent with the experimental fact that phase transition from the PM bcc to hcp phases does not occur under volume compression. The seeming contradiction can be explained only when anharmonic behavior of the N$_4^-$ mode is taken into consideration; a potential energy curve along the N$_4^-$ mode becomes closer to a double-well shape for the PM phase under the volume compression. On the other hand, softening of the longitudinal mode at the 2/3[111] point under the volume compression is also found for the PM phase, which indicates the pressure-induced bcc-fcc phase transition along this mode. Such behaviors are not seen in ferromagnetic (FM) bcc Fe, implying that the magnetic structure plays essential roles on the phase transition mechanism.

cond-mat.mtrl-sci↗

arXiv2014

Special quasirandom structure in heterovalent ionic systems

The use of a special quasirandom structure (SQS) is a rational and efficient way to approximate random alloys. A wide variety of physical properties of metallic and semiconductor random alloys have been successfully estimated by a combination of an SQS and density functional theory (DFT) calculation. Here, we investigate the application of an SQS to the ionic multicomponent systems with configurations of heterovalent ions, including point-charge lattices, MgAl$_2$O$_4$ and ZnSnP$_2$. It is found that the physical properties do not converge with the supercell size of the SQS. This is ascribed to the fact that the correlation functions of long-range clusters larger than the period of the supercell are not optimized in the SQS. However, we demonstrate that the physical properties of the perfectly disordered structure can be estimated by linear extrapolation using the inverse of the supercell size.

cond-mat.mtrl-sci↗

arXiv2014

A sparse representation for potential energy surface

We propose a simple scheme to estimate potential energy surface (PES) with which the accuracy can be easily controlled and improved up to the level of the density functional theory (DFT) calculations. It is based on a model selection within the framework of linear regression using the least absolute shrinkage and selection operator (LASSO) technique. Basis functions are selected from a systematic large set of candidate functions. The sparsity of PES significantly reduces the computational demands for evaluation of the energy and the force in molecular dynamics simulations without losing the accuracy. The usefulness of the scheme is well demonstrated for describing elemental metals of Na and Mg.

cond-mat.mtrl-sci↗

arXiv2014

Efficient determination of alloy ground-state structures

We propose an efficient approach to accurately finding the ground-state structures in alloys based on the cluster expansion method. In this approach, a small number of candidate ground-state structures are obtained without any information of the energy. To generate the candidates, we employ the convex hull constructed from the correlation functions of all possible structures by using an efficient algorithm. This approach is applicable to not only simple lattices but also complex lattices. Firstly, we evaluate the convex hulls for binary alloys with four types of simple lattice. Then we discuss the structures on the vertices. To examine the accuracy of this approach, we perform a set of density functional theory calculations and the cluster expansion for Ag-Au alloy and compare the formation energies of the vertex structures with those of all possible structures. As applications, the ground-state structures of the intermetallic compounds CuAu, CuAg, CuPd, AuAg, AuPd, AgPd, MoTa, MoW and TaW are similarly evaluated. Finally, the energy distribution is obtained for different cation arrangements in MgAl$_2$O$_4$ spinel, for which long-range interactions are essential for the accurate description of its energetics.

cond-mat.mtrl-sci↗

arXiv2014

Cluster expansion of multicomponent ionic systems with controlled accuracy: Importance of long-range interactions in heterovalent ionic systems

We have been examining factors determining the accuracy of cluster expansion (CE), which is used in combination with many density functional theory (DFT) calculations. With the exception of multicomponent metallic or isovalent ionic systems, the contributions of long-range effective cluster interactions (ECIs) to configurational energetics are not negligible, which is ascribed to long-range electrostatic interactions. The truncation of ECIs in such systems leads to systematic errors. A typical problem with such errors can be seen in Monte Carlo (MC) simulations since simulation supercells composed of a larger number of atoms than those of the input DFT structures are used. The prediction errors for long-period structures beyond the cell size of the input DFT structures in addition to those for short-period structures within the cell size of the input DFT structures need to be carefully examined to control the accuracy of CE. In the present study, we quantitatively discuss the contribution of the truncation of long-range ECIs to the accuracy of CE. Two types of system, namely, a point-charge spinel lattice and a real MgAl2O4 spinel crystal, are examined.

cond-mat.mtrl-sci↗