AI-predicted PT-symmetric magnets

TL;DR

This paper uses AI, DFT, and symmetry analysis to identify and verify 23 candidate PT-symmetric antiferromagnetic materials with unique quantum transport and optical properties, including 3 experimentally confirmed.

Contribution

It introduces a novel AI-driven approach combining graph neural networks and symmetry constraints to discover AFM1 materials with potential quantum applications.

Findings

23 candidate AFM1 materials identified, including 3 verified experimentally.

DFT confirms AFM1 as the lowest energy magnetic state in these materials.

The method enables efficient screening of materials for symmetry-enabled quantum effects.

Abstract

Parity-time-reversal-symmetric odd-parity antiferromagnetic (AFM1) materials are of interest for their symmetry-enabled quantum transport and optical effects. These materials host odd-parity terms in their band dispersion, leading to asymmetric energy bands and enabling responses such as the magnetopiezoelectric effect, nonreciprocal conductivity, and photocurrent generation. In addition, they may support a nonlinear spin Hall effect without spin-orbit coupling, offering an efficient route to spin current generation. We identify 23 candidate AFM1 materials by combining artificial intelligence, density functional theory (DFT), and symmetry analysis. Using a graph neural network model and incorporating AFM1-specific symmetry constraints, we screen Materials Project compounds for high-probability AFM1 candidates. DFT calculations show that AFM1 has the lowest energy among the tested magnetic configurations in 23 candidate materials. These include 3 experimentally verified AFM1 materials, 10 synthesized compounds with unknown magnetic structures, and 10 that are not yet synthesized.