Spin-transfer Torque based random access memory (STTRAM) has the potential to be a universal memory -- it is faster than flash and non-volatile compared to SRAM and DRAM. Furthermore, it is readily integrable with CMOS technology, and has excellent scaling properties. The main idea is to encode memory in the up and down orientations of the magnetism in a ferromagnetic layer that is free to rotate. Instead of using a magnetic field to write the information onto the magnetism, the spins are rotated by driving a small current perpendicular to the layer, using another fixed ferromagnet to spin polarize the current. The component of the incoming spin perpendicular to the free layer magnetism is completely absorbed, so that angular momentum conservation leads to a torque on the magnet that rotates the spin. The main project we have is to study the fundamental physics of the spin torque and rotation, their dependence on the materials involved (the magnetic layers and the intervening oxide) and their device and circuit-level ramifications.

A self-consistent bandstructure-spin torque-micromagnetic simulator

We have developed a self-consistent simulator that couples bandstructures of the ferromagnets with a quantum calculation of the currents, spin currents, torques and tunnel magnetoresistances, and finally with the micromagnetic simulation of the spin precession in the contacts. The bandstructures are fully atomistic, especially critical for the study of magnetic alloys and interfacial strains. The torque is fully quantum, including both in and out-of-plan components. The precession dynamics includes temperature and multiple domain effects. The results are allowing us to explore a wide range of parallel and perpendicular materials, geometries and magnetic structures.

Study of Symmetry filtering materials

Fe on MgO has a dual property that allows for extremely high TMRs. One is the presence of Delta_1 up spin bands in Fe that align with those in MgO. Furthermore, there is symmetry-filtering due to a non-exponential decay of the wavefunctions into MgO due to the co-operative effect of conduction and valence bands in the center of the MgO band-gap. The calculated TMR and torque are shown below. While symmetry filtering enhances the TMR, the elimination of minority carriers makes it hard to switch from parallel to antiparallel, as shown by the strong R-V asymmetry below. Symmetry filtering materials will thus need a dual MTJ structure for efficient operation.

PUBLICATIONS

"Self-consistent Quantum Transport and Magnetization Dynamics in STT-RAM", K. Munira and A. W. Ghosh, Intermag (2010).