[16a-70A_101-1~5] AI x Robotics x Human: Opening New Frontiers in Autonomous Materials Research

9:45 AM - 10:25 AM JST(12:45 AM - 1:25 AM UTC)

[16a-70A_101-3] Fermi Surface Analysis Using Principal Component Analysis and Anomaly Detection

Daichi Ishikawa¹, 〇Kentaro Fuku^1,2, Yoshio Miura^3,4, Yasuhiko Igarashi⁵, Yuma Iwasaki³, Yuya Sakuraba³, Koichiro Yaji^3,6, Alexandre Lira Foggiatto¹, Takahiro Yamazaki¹, Naoka Nagamura^1,3,6, Masato Kotsugi¹ (1.Tokyo Univ. of Science, 2.Nagoya Univ., 3.NIMS, 4.KIT, 5.Tsukuba Univ., 6.Tohoku Univ.)

Keywords:

machine learning,electronic structure

The Fermi surface (FS) is central to understanding electronic properties, however, its analysis remains expertise-intensive and increasingly burdensome in high-throughput studies. We present a compact, data-driven framework that combines principal component analysis (PCA) with anomaly detection to extract correlations between FS morphology and physical properties. We apply the method to the Heusler alloy Co₂MnGa_xGe_1-x using FS images constructed from first-principles calculations. In the PCA space, we observe discontinuous changes in FS features that strongly correlate with extrema and inflection points of the spin polarization changes. Anomalies around x = 0.94–0.95 are attributed to the emergence of a nodal line; difference analysis visualize the nodal-line location. Robustness tests confirm stable detection performance. Finally, applying to experimental graphene data demonstrates Dirac-point detection, highlighting the method’s applicability to real ARPES datasets.