Method for rapid screening of diatomic doping bcn adsorbed h structure by machine learning potential

By employing machine learning potential models and periodic loop correction techniques, we can automatically screen for H-adsorbed structures in diatomic-doped BCN, solving the problems of high-throughput computation time and structural stability determination, and achieving rapid and accurate structure screening and resource optimization.

CN122177315APending Publication Date: 2026-06-09GUIZHOU NORMAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUIZHOU NORMAL UNIVERSITY
Filing Date
2026-04-01
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing technologies face challenges in screening biatomic BCN-doped H-adsorbed structures, including long high-throughput computation time, high waste computing power, and a lack of automated quantitative standards for eliminating unstable and distorted structures. These issues lead to large human errors and inconsistent screening results.

Method used

A machine learning potential model is used for geometric optimization iteration, combined with periodic wrap-around correction operations, to construct a multi-dimensional quantitative structural judgment index. An automatic screening is achieved through a veto mechanism to eliminate coordinate errors that cross the unit cell boundary and ensure structural stability.

Benefits of technology

It significantly shortens the relaxation time of individual structures, improves the utilization of computing resources, reduces human error, achieves consistency in structural stability screening, and has good versatility and scalability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122177315A_ABST
    Figure CN122177315A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of computational materials science, and discloses a method for rapidly screening a diatomic doped BCN adsorbed H structure by using a machine learning potential, which comprises the following steps: generating three-dimensional coordinate data of a candidate structure by high-throughput automatic modeling; obtaining a relaxed structure by iterative geometric optimization by using a machine learning potential model; solving a final real displacement module length value of an atom by performing coordinate conversion and period winding correction; constructing a local neighbor atom set, extracting a displacement module length, and generating three quantitative structure judgment indexes; comparing the indexes with a judgment threshold value, executing a one-vote veto mechanism, and outputting a structure abnormal state marker; and classifying and archiving the structure according to the marker and outputting a statistical report. The application replaces traditional first-principle calculation with a machine learning potential, combines a quantitative displacement index, realizes automatic pre-screening of the stability of a large-scale candidate structure, and improves material calculation efficiency.
Need to check novelty before this filing date? Find Prior Art