Application of Neural Networks for the Identification of the Skyrmionic Phases on a 2D Semiregular Ferromagnetic Lattice

Application of Neural Networks for the Identification of the Skyrmionic Phases on a 2D Semiregular Ferromagnetic Lattice

• Autor
• Juan Rafael Gomez Quispe
• Co-autores
• Tony Wenceslao Milla Robles , Pedro Alves Da Silva Autreto , Chachi Rojas Ayala
• Resumo

• Magnetic skyrmions, intricate topological formations within magnetic systems, have become a focal point in materials research due to their notable stability and nanoscale mobility. The potential applications of these magnetic patterns in data storage and information processing technologies have fueled a concerted effort to develop efficient methods for their characterization [1]. Simultaneously, the advent of neural networks, inspired by the intricacies of the human brain, marks a transformative shift in artificial intelligence. Their remarkable ability to learn patterns and execute intricate tasks positions them as indispensable tools across a spectrum of academic disciplines. Particularly within the domain of materials physics, neural networks have proven effective in modelling and predicting complex phenomena [2,3]. The Skyrmionic number stands as a pivotal metric delineating the presence and distribution of skyrmions within a material. A positive Skyrmionic number represents skyrmions with a specific orientation, while a negative number indicates the presence of skyrmions with the opposite orientation [2,4]. The calculation of the Skyrmionic number can be challenging and its difficulty depends on several factors, including the complexity of the magnetic system, the presence of defects or irregularities, and the precision required in the results. The objective of this study is to employ neural networks for the identification of skyrmionic phases in low temperature (t = 0.001) on a semi-regular 2D magnetic system comprising tetragonal and octagonal rings. To accomplish this, our methodology integrates Monte Carlo Metropolis simulations within an optimal space sampling dynamic to minimize the energy of initial magnetic configurations [5]. Subsequently, neural networks are harnessed to identify the magnetic phases and construct a comprehensive D/H phase diagram.

  References:

  [1] K. Wang, V. Bheemarasetty, J. Duan, S. Zhou, and G. Xiao, “Fundamental physics and applications of skyrmions: A review,” J. Magn. Magn. Mater., vol. 563, no. August, p. 169905, 2022, doi: 10.1016/j.jmmm.2022.169905.

  [2] I. A. Iakovlev, O. M. Sotnikov, and V. V. Mazurenko, “Supervised learning magnetic skyrmion phases,” 2018, doi: 10.1103/PhysRevB.98.174411.

  [3] D. Kapitan et al., Application of machine learning in solid state physics, vol. 74. Elsevier, 2023.

  [4] J. C. Criado, S. Schenk, M. Spannowsky, P. D. Hatton, and L. A. Turnbull, “Simulating anti-skyrmions on a lattice,” Sci. Rep., vol. 12, no. 1, pp. 1–11, 2022, doi: 10.1038/s41598-022-22043-0.

   

  [5] J. D. Alzate-Cardona, D. Sabogal-Suárez, R. F. L. Evans, and E. Restrepo-Parra, “Optimal phase space sampling for Monte Carlo simulations of Heisenberg spin systems,” J. Phys. Condens. Matter, vol. 31, no. 9, 2019, doi: 10.1088/1361-648X/aaf852.

   

• Palavras-chave
• Neural Network, Monte Carlo Metropolis, Skyrmions
• Modalidade
• Pôster
• Área Temática
• Nanociência e Nanotecnologia