Largely tunable compensation temperature in a rare-earth ferrimagnetic metal and deterministic spin-orbit torque switching for artificial neural network application

Abstract

Ferrimagnets are important for next-generation high-density ultrafast spintronic device applications. Magnetization compensation temperature (T_M) is a fundamentally critical magnetic parameter for ferrimagnets besides their Curie temperature. Around T_M, the spin-orbit switching efficiencies are extraordinarily high. Therefore, the accurate manipulation of T_M from the material fabrication process is essential for the electrical steering of ferrimagnetic spins. In this work, CoTb thin films, with the 3d and 4f magnetic sublattices antiferromagnetically coupled to each other, are deposited at different temperatures. The magnetotransport and magnetic properties of these films are systematically investigated. It was found that the T_M of this rare-earth ferrimagnet largely depends on the growth temperature and it can be tuned by over 100 K. Accordingly, the spins of an optimized ferrimagnetic CoTb thin film with its T_M close to room temperature can be efficiently switched by the current-pulse-induced spin-orbit torque. Moreover, an artificial neural network utilizing the spin-orbit torque device was constructed, demonstrating an image recognition accuracy of approximately 92.5%, which is comparable to that of conventional software solutions. Thus, this work demonstrates the large tunability of T_M of a rare earth ferrimagnet by chemical ordering and the great potential of such a ferrimagnet for electrically operated spintronic devices.