Hydride Content Control of Perovskite Oxyhydride BaTiO_3–<i>x</i>H_<i>x</i> Supported by Image-Based Machine Learning

Hydride content in perovskite oxyhydrides represents a crucial parameter that governs properties of the materials. However, the synthesis of the oxyhidrides with precise control over the hydride content x remains a challenging task because of the time-consuming and/or destructive analytical methods typically used for the determination of x. Here, we report an image-based machine learning (ML) system for prediction of the hydride contents x in a perovskite oxyhydride BaTiO3–xHx using a picture of powder material as a quick evaluation method of hydride content x. The ML system, which employs the ExtraTrees algorithm, enabled the prediction of x with a mean absolute error of 0.013 and a low running cost. Although a similar ML system constructed by convolutional neural networks (CNN) that we used in the previous study demonstrated comparable accuracy in the prediction, the system with ExtraTrees was more computationally efficient, thus was applied as a nondestructive and quick analytical method for hydride content x. Using the ML system, detailed profiles of thermal hydrogen release of BaTiO3–xHx were quantitatively analyzed to demonstrate the feasibility of fine-tuning the hydride content within 0 ≤ x ≤ 0.4 by adjusting the processing temperature. In the case of powder samples comprising mixtures of the oxyhydride with different hydride contents, the ML-based prediction provided an almost averaged value for x. The present results demonstrate an application of image-based ML for the fine-tuning of oxyhidride materials.