A nonintrusive reduced order modelling of magnetic skyrmions

We show that a nonintrusive reduced order modeling framework can predict the magnetic field dependent internal structures of magnetic skyrmions^[1] and their density in thin films. After we train the framework using results of micromagnetic simulations, in which we vary the anisotropy, asymmetric exchange energy, and the external magnetic field, we were able to predict the ground state spin configurations for given anistorpy, asymmetric exchange energy, and external magnetic field. Besides, we implement PC^[2] (principal component) manifold to interpolate multidimensional data more accurately compared to other methods. We then confirm these spin configurations by calculating micromagnetically^[3] and found XX percent accuracy and XX orders of magnitude reduction in the simulation time. Our data-driven approach accelerates identifying the spin texture by eliminating the need of re-calculation of micromagnetic simulations for every material parameter in a functioning device.