Background
SOT-MRAM is based on the reversible switching of the magnetization of one of two ferromagnetic layers in the magnetic tunnel junction (MTJ) due to the magnetic field created by an in-place current pulse. This causes the magnetization of the switchable ferromagnetic layer to precess toward an in-plane direction. After the current passes, the magnetization relaxes randomly to either the original or opposite state. The switching is non-deterministic, occurring only 50% of the time. To induce deterministic switching a static magnetic field, the so-called “zero-field,” needs to exist in the region of the MTJ. The footprint needed for the zero-field undermines the potential density gains of SOT-MRAM.
SOT-MRAM has the potential for much higher density and much lower power consumption than existing memories which makes it superior for use in nearly all chips. The primary challenge is that reproducible bipolar switching of the SOT-MRAM memory element, the magnetic tunnel junction (MTJ), requires a static magnetic field. The university’s deterministic MTJ does not require a static magnetic field, enabling an SOT-MRAM able to achieve both performance and density goals.
Technology Overview
By introducing geometric anisotropy into the MTJ, the precession of the magnetization of the switchable ferromagnetic layer 1) depends upon the direction of current flow in the read line, and 2) doesn’t end in an in-plane direction. This allows the magnetization of the switchable ferromagnetic layer to relax deterministically to the opposite state after a correctly oriented current pulse has passed.
Benefits
Because a zero-field is not required to achieve deterministic bipolar switching, the elliptical MTJ enables an SOT-MRAM device which can achieve the intended performance of SOT-MRAM while retaining its advantageous density enhancement over existing memory solutions. Importantly, this method uses existing MTJ and CMOS technologies allowing for scalable implementation.
Compared to patents on related techniques which are currently being pursued by SOT-MRAM startups, our patents have a much earlier priority date.
Applications
SOT-MRAM