Materials Curation^®, which has been proposed by the author as a method to search materials by combining scientific principles in multi-diciplinary way, is reported. This method has been developed to overcome a problem that experimental or calculation data on materials in a issue are far below enough for big-data type analysis with machine learning. By utilizing scientific principles interdisciplinary, material search is made possible with data in a very small size.

View full abstract

The charge compensation mechanism of a LiNi_⅓Co_⅓Mn_⅓O₂ (NCM333) positive electrode, used in a commercial lithium ion battery, was investigated by operando soft X-ray absorption fine structure (SX-XAFS) spectroscopy. The SX-XAFS spectral analyses recorded with the partial fluorescence yield revealed followings. The pristine NCM333 consisted of Ni²⁺, Co³⁺, Mn⁴⁺ and O^2-. During charge, Ni²⁺ was oxidized to Ni⁴⁺, and Mn⁴⁺ did not change at all. O^2- was oxidized by creating O 2p holes during charge, which caused an appearance of a new pre-edge peak in the O K-edge XAFS spectra. Co L₃-edge peak shifted to higher energy, indicating that Co³⁺ is oxidized during charge. Hence nickel ions play the central role of the charge compensation in the NCM333 positive electrode, and Mn ions are not contribute to the charge compensation. O ions and Co ions are not as large as Ni ions, but they are also contribute to the charge compensation.

View full abstract

Time-of-Flight secondary ion mass spectrometry (TOF-SIMS) and scanning electron microscope (SEM) images were fused and then evaluated by means of principal component analysis. As a result, TOF-SIMS spatial resolution could be improved by adding SEM image information to TOF-SIMS data without drastic change of TOF-SIMS spectrum information. Sparse modeling and machine learning were applied to TOF-SIMS data to interpret complex TOF-SIMS spectra. Least Absolute Shrinkage and Selection Operator (LASSO) provided a simplified TOF-SIMS spectrum with less noise. Machine learning using Random Forest and k-Nearest Neigbour appropriately predicted unknown test samples by learning TOF-SIMS data similar the test samples.

View full abstract

This article [J. Surf. Anal. 21, 2-9 (2014)] was originally published with three errors.

View full abstract