In the present study, MgCo 1.

In this study, we used a benchmark dataset to evaluate the impact of scaling with the extrapolation domain on the prediction performance of machine learning algorithms. We pseudo-divided the data into the interpolation domain (training data) and the extrapolation domain (test data) using a combination of UMAP (Uniform Manifold Approximation and Projection) and material domain knowledge. In anticipation of bridging interpolation and extrapolation domains in nonlinear machine learning algorithms, we evaluated how the scaling considering the extrapolation domain affects prediction performance in the extrapolation domain. For this evaluation, we used three nonlinear algorithms widely used in the MI (Materials Informatics) domain: XGB (XGBoost) regression, GP (Gaussian Process) regression, and SVR (Support Vector Regression). In this study, by defining the pseudo extrapolation domain, we established the approach for evaluating the prediction accuracy of machine learning models in the extrapolation domain, which is considered difficult to evaluate quantitatively. We also demonstrated that this method, which uses scaling that considers the extrapolation domain, is an effective method for improving prediction accuracy in the extrapolation domain while maintaining prediction accuracy in the interpolation domain.

1. Introduction The Mg secondary battery of which a high volume energy densities is expected as the next generation secondary battery. However the Mg diffusion is difficult, and it don’t achieve to practical use. The previous study reported the cathode material MgCo2-xMnxO4 with spinel type structure (x= 0.5)1) and Mg (Mg0.33V1.67-yNiy)O4 (y= 0.1) 2) so far at our laboratory. The former cathode can exhibit a high initial discharge capacity. However, the capacity retention ratio is low. On the other hand, the latter cathode can exhibit high capacity retention ratio. This research was performed to determine synthetic method of solid solution cathode αMgCo1.5Mn0.5O4-(1-α) Mg(Mg0.33V1.57Ni0.1)O4 (α = 0.1, 0.3, 0.5, 0.7, 0.9) which is new cathode material and to measure battery characteristics for the purpose of utilizing an advantage of these materials. The electrode was characterized by the Rietveld analysis to reveal charge-discharge mechanism. 2. Experimental Three synthetic routes were examined to obtain the solid solution materials: (1) solid phase method (2) mechanochemical method and (3) reverse coprecipitation method. The synthesis method (1), mixed each endmembers by an agate mortar, and calcined. The synthesis method (2), mixed each endmembers and obtained a target materials by only the mechanical milling. The synthesis method (3), mixed each metal-nitrate aqueous solutions as it was α = 0.1, 0.3, 0.5, 0.7, 0.9, calcined to get a target materials. The obtained materials were identified by XRD,ICP-AES, and performed the charge-discharge cycle tests using three electrode cell. Synchrotron X-ray diffraction measurement (BL19B2, SPring-8) was performed and the average structure was analyzed by a Rietveld analysis (RIETAN - FP). Furthermore, XAFS (BL14B2, SPring-8) revealed pristine and charging/discharging electrode redox mechanism. 3. Result and discussion The sample synthesized by method (1) wasn’t assigned to spinel structure, while the samples synthesized by method (2) and (3) were assigned to spinel structure with the space group of Fd-3m from the powder X-ray diffraction. The sample of synthesis method (1) and (2), approximately controlled on a metal composition by ICP - AES. However, V composition of synthesis method (3) controlled smaller than preparation composition. Good fitting was obtained about a powder sample of a synthesis method (2) and (3) as a result of the Rietveld analysis, however, it revealed Mg occupied 8a/16d site for each sample. A charge-discharge cycle test was carried out at 90 ° C using a three-electrode cell. the sample of a synthesis method (2) performed 15 cycles by the discharge capacity of about 80 mAh/g, and the sample of method (3) (α =0.5) performed 24 cycles by the discharge capacity of about 100 mAh/g while the sample of method (3) (α =0,3) achieved 60 cycles at the discharge capacity of about 160 mAh/g and it had high capacity ratio (Fig.1). Acknowledgement This work was supported by JST ALCA-SPRING Grant Number JPMJAC1301, References 1) M. Ichiyama, N. Ishida, N. Kitamura, Y. Idemoto, the 86th Electrochemical Society of Japan Meet summary and 3O20 (2019). 2) N. Kawakami, N. Ishida, N. Kitamura, Y. Idemoto, the 59th Battery forum summary P.375(2018). Figure 1