arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.05683cond-mat.str-elcond-mat.mes-hallphysics.comp-ph
收藏

构建天空子系统详细相图的机器学习技术

Machine learning techniques to construct detailed phase diagrams for skyrmion systems

F. A. Gómez Albarracín, H. D. Rosales

AI总结:

该研究使用卷积神经网络结合蒙特卡罗模拟,对包含铁磁交换和DM相互作用的自旋模型进行相分类,成功构建了详细的磁场-温度相图,并准确识别了复杂的中间拓扑相。

摘要:

近年来,将机器学习(ML)技术应用于凝聚态物理中各种问题的兴趣日益浓厚。在这方面,对简单和复杂物质相的表征尤为重要。在此,我们使用ML方法构建了一个结合铁磁交换和Dzyaloshinskii-Moriya(DM)相互作用的著名自旋模型的完整相图,其中出现了拓扑相。在低温下,系统通过磁场从螺旋相转变为天空子晶体。然而,热涨落引发了两种中间相,即双子和天空子气,它们不像螺旋或天空子晶体那样容易确定。我们借助大规模蒙特卡罗模拟获取低温自旋构型,并训练了一个卷积神经网络(CNN),仅使用特定DM耦合值处的快照来对不同相进行分类,重点关注中间的复杂拓扑纹理。然后,我们将CNN应用于更高温度的构型和其他DM值,以构建详细的磁场-温度相图,取得了出色的结果。我们讨论了包含无序顺磁相以确定相边界的重要性,最后将我们的方法与其他ML算法进行了比较。

英文摘要:

Recently, there has been an increased interest in the application of machine learning (ML) techniques to a variety of problems in condensed matter physics. In this regard, of particular significance is the characterization of simple and complex phases of matter. Here, we use a ML approach to construct the full phase diagram of a well known spin model combining ferromagnetic exchange and Dzyaloshinskii-Moriya (DM) interactions where topological phases emerge. At low temperatures, the system is tuned from a spiral phase to a skyrmion crystal by a magnetic field. However, thermal fluctuations induce two types of intermediate phases, bimerons and skyrmion gas, which are not as easily determined as spirals or skyrmion crystals. We resort to large scale Monte Carlo simulations to obtain low temperature spin configurations, and train a convolutional neural network (CNN), taking only snapshots at specific values of the DM couplings, to classify between the different phases, focusing on the intermediate and intricate topological textures. We then apply the CNN to higher temperature configurations and to other DM values, to construct a detailed magnetic field-temperature phase diagram, achieving outstanding results. We discuss the importance of including the disordered paramagnetic phases in order to get the phase boundaries, and finally, we compare our approach with other ML algorithms.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑