$Ab$ $Initio$ Study of Magnetic Tunnel Junctions Based on Half-Metallic and Spin-Gapless Semiconducting Heusler Compounds: Reconfigurable Diode and Inverse Tunnel-Magnetoresistance Effect
Magnetic tunnel junctions (MTJs) have attracted strong research interest within the last decades due to their potential use as nonvolatile memory such as MRAM as well as for magnetic logic applications. Half-metallic magnets (HMMs) have been suggested as ideal electrode materials for MTJs to achieve an extremely large tunnel-magnetoresistance (TMR) effect. Despite their high TMR ratios, MTJs based on HMMs do not exhibit current rectification, i.e., a diode effect, which was achieved in a magnetic tunnel junction concept based on HMMs and type-II spin-gapless semiconductors (SGSs). The proposed concept has recently been experimentally demonstrated using Heusler compounds. In the present work, we investigate from first-principles MTJs based on type-II SGS and HMM quaternary Heusler compounds FeVTaAl, FeVTiSi, MnVTiAl, and CoVTiSb. Our $ab$ $initio$ quantum transport calculations based on a nonequilibrium Green's function method have demonstrated that the MTJs under consideration exhibit current rectification with relatively high on:off ratios. We show that, in contrast to conventional semiconductor diodes, the rectification bias voltage window (or breakdown voltage) of the MTJs is limited by the spin gap of the HMM and SGS Heusler compounds. A unique feature of the present MTJs is that the diode effect can be configured dynamically, i.e., depending on the relative orientation of the magnetization of the electrodes, the MTJ allows the electrical current to pass either in one or the other direction, which leads to an inverse TMR effect. The combination of nonvolatility, reconfigurable diode functionality, tunable rectification voltage window, and high Curie temperature of the electrode materials makes the proposed MTJs very promising for room-temperature spintronic applications and opens ways to magnetic memory and logic concepts as well as logic-in-memory computing.