Following the historical development of informatics, data-intensive science, machine learning and materials informatics, we introduce important applications to surface science. As an indication of first application of multivariate analysis to surface science, application of target factor analysis to depth profiling data by Auger electron spectroscopy is introduced including the detailed analysis principle. We believe the way the future measurement system should be is a “smart measurement system” linked with cyber space and physical space. In such a data driven measurement system, novel informatics that can be applied to big measurement data are required. From such a viewpoint, the fusion of surface analysis and informatics, that is, “surface measurement informatics” is expected as a new paradigm of surface science.

Interface plays crucial roles for mechanical and functional properties. Despite its importance, the structure-property relationships of the interface have not been fully understood because a determination of the interface structure accompanies huge computations which is caused by the degrees of geometrical freedom of the interface. The acceleration of the interface structure searching is important for the comprehensive understanding of the structure-property relationships of the interface. Here, we applied machine learning techniques for the interface structure searching. Two methods, virtual screening and kriging, are mainly introduced. Our methods achieved considerably high efficiency and can be applied to any materials.

The prediction model of the result of computed fluid dynamics simulation in SiC solution growth was constructed on neural network using machine learning. Utilizing the prediction model, we can optimize quickly crystal growth conditions. In addition, the real-time visualization system was also made using the prediction model.

In this review, we present the current situation of designs of biomaterials using techniques of informatics. In particular, we discuss the prediction the responses of proteins and cells toward materials and data-driven design of new biomaterials. We introduce our recent work, in which we analyzed the correlation among chemical structures of molecules constituting self-assembled monolayers (SAMs), amounts of adsorbed protein, a density of adhered platelets by machine learning using an artificial neural network model. The main conclusion is that the quality of the database is a critical factor in determining the accuracy of the prediction and material design. We also discuss technical issues to develop databases efficiently and systematically to expand the possibility of data-driven strategies to design biomaterials.

We present an adaptive design of experiment (DoE) by machine learning for X-ray spectroscopy to improve its efficiency. One of the machine learning techniques, Gaussian process regression predicts a spectrum from the experimental data and determines the optimal energy points to measure. Adaptive DoE successfully reduces total energy points to measure as compared to an X-ray magnetic circular dichroism spectroscopy experiment by a conventional DoE. This method has potential applicability to various measurements and reduces the time and cost of experiments.

We executed the topological data analysis of labyrinth magnetic domain structure to visualize pinning site during the magnetization reversal process. We utilized persistent homology to extract the topological feature of the magnetic domain structure, and principal component analysis was used to construct the correlation between persistence diagram and magnetic hysteresis loop. As a result, we could automatically visualize the pinning site as topological defect on the original magnetic domain structure.