Research on Flat-Band Feature Recognition Model Based on Energy Band Images

Junlong Li

doi:10.47297/taposatWSP2633-456922.20260705

PDF
Export
Share
Collection
Album

Research on Flat-Band Feature Recognition Model Based on Energy Band Images

Volume 7, Issue 5

Views: 0 Downloads: 0

Research on Flat-Band Feature Recognition Model Based on Energy Band Images
Research on Flat-Band Feature Recognition Model Based on Energy Band Images
College of Science, China Jiliang University, Hangzhou, Zhejiang, 310018, China
Vol. 7, Issue 5, Pages: 115-119(2026)
Online First：15 July 2026，

Published：30 May 2026
DOI：10.47297/taposatWSP2633-456922.20260705
Accepted：

Scan QR Code

Li Junlong.Research on Flat-Band Feature Recognition Model Based on Energy Band Images[J].Theory and Practice of Science and Technology,2026,07(05):115-119.
DOI：

Li Junlong.Research on Flat-Band Feature Recognition Model Based on Energy Band Images[J].Theory and Practice of Science and Technology,2026,07(05):115-119. DOI： 10.47297/taposatWSP2633-456922.20260705.

摘要

Abstract

Topological materials

including topological insulators

superconductors

and semimetals

have attracted extensive attention in condensed matter physics and materials science. In particular

flat-band topological semimetals are closely related to strongly correlated electron behaviors and unconventional superconductivity due to their extremely low bandwidth and high density of states. To improve the recognition of local low-dispersion bands

this work proposes a convolutional neural network framework

MEA-CNN

which integrates attention mechanisms and a mixture of experts mechanism for flat band recognition in band structure images. The model enhances key flat-band regions through channel and spatial attention

while the MoE structure adaptively models different flat band patterns. Experimental results show that MEA-CNN achieves an accuracy of 0.928 and an AUC of 0.9842

outperforming conventional convolutional neural networks with good interpretability.

关键词

Keywords

references

He Y, De Breuck P P, Weng H, et al. Machine learning on multiple topological materials datasets[J]. npj Computational Materials, 2025, 11(1): 181.

Andrejevic N, Andrejevic J, Bernevig B A, et al. Machine-learning spectral indicators of topology[J]. Advanced Materials, 2022, 34(49): 2204113.

Neves P M, Wakefield J P, Fang S, et al. Crystal net catalog of model flat band materials[J]. npj Computational Materials, 2024, 10(1): 39.

Regnault N, Xu Y, Li M R, et al. Catalogue of flat-band stoichiometric materials[J]. Nature, 2022, 603(7903): 824-828.

Bhattacharya A, Timokhin I, Chatterjee R, et al. Deep learning approach to genome of two-dimensional materials with flat electronic bands[J]. npj Computational Materials, 2023, 9(1).

Wang X, Wei Y, Bhattacharya A, et al. Structure-informed learning of flat band 2d materials[EB/OL]. (2025-06-11)[2026-05-25]. https://arxiv.org/abs/2506.07518.

Woo S, Park J, Lee J Y, et al. Cbam: convolutional block attention module[EB/OL]. (2018-07-17)[2026-03-17]. https://arxiv.org/abs/1807.06521v2.

Jacobs R A, Jordan M I, Nowlan S J, et al. Adaptive mixtures of local experts [J]. Neural Computation, 1991, 3 (1): 79-87.

Sinha P, Joshi A, Dey R, et al. Machine-learning-assisted materials discovery from electronic band structure [J]. Journal of Chemical Information and Modeling, 2024, 64 (22): 8404-8413.

Jain A, Ong S P, Hautier G, et al. Commentary: the materials project: a materials genome approach to accelerating materials innovation [J]. APL Materials, 2013, 1 (1).

Lin Z, Choi J H, Zhang Q, et al. Flatbands and emergent ferromagnetic ordering in fe3sn2 kagome lattices [J]. Physical Review Letters, 2018, 121 (9): 096401.

Huang Hao. Study on novel magnetic and electrical transport properties in CoSn, FeGe, and MnPt [D]. Hefei: China University of Science and Technology, 2024.

Lin Z, Wang C, Wang P, et al. Dirac fermions in antiferromagnetic fesn kagome lattices with combined space inversion and time-reversal symmetry [J]. Physical Review B, 2020, 102 (15): 155103.

Arora R, Basu A, Mianjy P, et al. Understanding deep neural networks with rectified linear units [EB/OL]. (2018-11-14)[2026-05-25]. http://arxiv.org/abs/1611.01491.

Hecht-Nielsen R. Theory of the backpropagation neural network[C]//International 1989 Joint Conference on Neural Networks. Piscataway: IEEE Press, 1989: 593-605.