Search

In the search for new high-performance materials in materials science, especially in polynomial science, it is important to use physicochemical laws linking materials structure and physical properties, and predict the physical properties required for the design. Recently, machine learning (ML) has enabled us to extract patterns from large datasets and construct the data-driven model to predict physical properties. However, ML approach faces challenges such as interpretability and systematic errors of the data-driven model with limited data. Here, we propose a method for extracting an interpretable law from limited data, by combining a symbolic regression method and Bayesian information criterion. We focus on extracting a physicochemical law for the refractive index of polymer materials. The goal is to correct systematic errors and capture physicochemical laws more accurately. Combining explanatory variables from experiments, property calculations, and neural potential approximations, our method involves arithmetic operations on explanatory variables and selection through Bayesian information criterion. The results show that the proposed method is able to correct the results of the neural potential approximation and obtain interpretable and concise expressions for the physicochemical laws linking material structure and physical properties.

Exploring vast material spaces efficiently is challenging in materials science. Autonomous methods for material search – integrating machine learning and ab initio calculations – have emerged as powerful alternatives to traditional approaches, which are often time-consuming and limited in scope. Although these autonomous methods have been applied to various material systems, the extensive material space of B2 structured materials for half-metallicity remains largely unexplored. Herein, we introduce a simulation-based autonomous search approach to identify B2 structured alloys exhibiting high spin polarization of sp conduction electrons (Psp), sp minority spin band gap (Gsp), and Curie temperature (Tc). The proposed method explores the material space of disordered quaternary B2 magnetic alloys using the Korringa – Kohn – Rostoker coherent potential approximation and Bayesian optimization. Over a continuous search of approximately 100 days, the system identified Co1.0Mn0.7Al0.3 as a promising candidate, demonstrating high values of Psp, Gsp, and Tc. Although additional experimental and theoretical validation is necessary, this study demonstrates the potential of autonomous material search methods to expedite material discovery and enhance material property optimization.

Efficient exploration of vast material spaces is a challenging task in materials science. Autonomous material search methods utilizing machine learning and ab initio calculations have emerged as powerful alternatives to traditional material discovery through synthesis and analysis, which is time-consuming and scope-limited. Although autonomous search methods have already been applied to various material spaces, they have not explored the extensive material space of Curie temperatures. Herein, we show a simulation-based autonomous search method that suggests ternary alloys with high Curie temperatures. The material space – consisting of disordered ternary magnetic alloys – is explored through Korringa – Kohn – Rostoker coherent potential approximation and Bayesian optimization. Over a continuous 10-day search, the system proposed several alloys – CoAuIr, CoPtPd, and CoFeBi – with Curie temperatures surpassing that of pure face-centered cubic Co. Although the insights gained through these predictions require further experimental and theoretical validation, this study demonstrates that autonomous material search methods can potentially accelerate material discovery and optimize material properties.

Recently, autonomous material search methods combining machine learning and experiments/simulations have become indispensable for exploring the extremely vast material exploration space. However, conventional autonomous material search methods focus solely on target material properties and their descriptors, leaving room for improvement in search efficiency. More efficient autonomous material search can be realized by utilizing information on nontarget properties that are dormant in databases. Here, we propose a novel method for autonomous material search using transfer learning and an ensemble neural network. This method can perform an autonomous material search to optimize the target properties while transferring information on nontarget properties. To demonstrate the usefulness of this method, we applied it to search for ternary magnetic alloys with a high Curie temperature while transferring information on magnetic moment and spin polarization. Results indicate that the proposed method improves the efficiency of autonomous material search, particularly in the early stages of the search process.

We developed an autonomous and efficient system for synthesising ferromagnetic materials with large magnetocrystalline anisotropy by integrating theoretical, informatics, and experimental approaches. By combining the first-principles calculation of the magnetic anisotropy with Bayesian optimisation, we virtually screened candidate materials, comprising four elements and four-layer periods, from various magnetic multilayers. We employed the expected improvement as the acquisition function and Matern52 as the kernel function, to develop a robust machine learning model. We fabricated the top three predicted magnetic materials under laboratory conditions by monoatomic layer deposition and evaluated their magnetic anisotropy using a superconducting quantum interference device (SQUID). Ultimately, we demonstrated that [Fe/Co/Fe/Ni]13 is a novel ferromagnetic material whose magnetic anisotropy exceeds that of L10-FeNi- and L10-FeCo-type alloys. Furthermore, the origin of the perpendicular magnetic anisotropy was derived from the spin-conserving as well as the spin-flip terms. We determined that Bayesian optimisation offers promising configurability features in terms of the electronic structure that extend beyond the empirical knowledge and human intuition.

Reflection high-energy electron diffraction (RHEED) data are important for the in-situ characterization of surface conditions during physical vapor deposition. Surface superstructures obtained by adsorbing exotic atoms onto a clean silicon surface, which exhibit various physical properties, were identified using RHEED. However, this information is too abundant for quantitative analysis; therefore, scientists rely on their expertise to interpret RHEED patterns to assess surface structures and evaluate film thickness, and a large amount of information remains unused. In this study, we adapted machine learning for a RHEED pattern dataset of a clean Si(111) surface during indium deposition in molecular-beam epitaxy growth to use the entire RHEED pattern image information and investigated appropriate machine leaning analysis methods. First, we aimed to determine RHEED pattern similarities in the dataset. Then, five structural phases, 7 × 7 (clean surface), √3×√3, √31×√31, 4 × 1, and 4 × 1 (Room Temperature), were automatically detected by hierarchical clustering using Ward’s method. Next, we aimed to extract the information for each surface superstructure from the dataset. Using non-negative matrix factorization, we successfully estimated the optimal forming conditions for each surface superstructure separately more accurately than the conventional methods. Our proposed strategies could be widely applied to surface structural analysis.

Autonomous material search systems that combine ab initio calculations and Bayesian optimization are very promising for exploring huge material spaces. Setting up an appropriate material search space is necessary for efficient autonomous material search. However, performing the autonomous search within the material space set up using the prepared descriptors is not sufficient to obtain an efficient search, which can be achieved by prioritizing specific descriptors and properties. Here, a material search system that can autonomously search the huge material space while performing multi-objective optimization that considers similarities among elements and emphasizes specific descriptors is proposed. This system has been used for a material exploration of Heusler alloys. The system has successfully proposed several candidate materials with half-metallic properties. The proposed system is very versatile and can be applied to various properties and material systems.

Finding materials with large magnetization is highly desirable for technological applications. Here, a machine learning autonomous search and automated combinatorial synthesis reveal that multi-element alloys with Ir and Pt impurities have a magnetization exceeding the Slater-Pauling limit of Fe75Co25.

High-throughput experiments (HTEs) have been powerful tools to obtain many materials data. However, HTEs often require expensive equipment. Although high-throughput ab-initio calculation (HTC) has the potential to make materials big data easier to collect, HTC does not represent the actual materials data obtained by HTEs in many cases. Here we propose using a combination of simple HTEs, HTC, and machine learning to predict material properties. We demonstrate that our method enables accurate and rapid prediction of the Kerr rotation mapping of an FexCoyNi1-x-y composition spread alloy. Our method has the potential to quickly predict the properties of many materials without a difficult and expensive HTE and thereby accelerate materials development.

Machine learning: Spying enhanced materials with x-ray vision Using algorithms to automatically spot variations in massive X-ray diffraction data sets may improve design of multi-component alloys. Having three or more metals in an alloy can lead to overwhelming combinations of possible materials, each with different properties. A. Gilad Kusne from the National Institute of Standards and co-workers examined how machine learning techniques could simplify alloy discovery through ‘dissimilarity measures’ that quantify how key structural data points, such as the positions and intensities of X-ray peaks, change with sample makeup. The team fabricated a compositional spread of iron–cobalt–nickel thin film alloys, and then evaluated different software approaches to finding X-ray dissimilarities for both processing speed and accuracy. Several algorithms suitable for high-throughput generation of color-coded maps that display relations between alloy composition and phase in both two and three-dimensions were identified.