Spinelectronics is a very rapidly growing area of R&D that merges magnetism and electronics (Nobel Prize 2007). Since the discovery of GMR (Giant magnetoresistance) in 1988, several breakthroughs have further boosted this field (spin-valves 1990, tunnel magnetoresistance (TMR) 1995, spin transfer 1996, very large TMR in MgO 2004, ). Spinelectronics has found applications in read heads of hard disk drives (1998) and more recently in non-volatile standalone memories (MRAM=Magnetic Random Access Memory). MRAMs integrate CMOS components with magnetic tunnel junctions (MTJ). The later are key basic elements in spintronics. They consist of two magnetic layers separated by a thin oxide barrier (most often MgO, ~1nm thick). When a bias voltage is applied across this trilayer, the transparency of the barrier to electrons strongly depends on the relative orientation of magnetization in the two magnetic electrodes. This results in a change of resistance of the junction which can reach 600% at room temperature. Thanks to the non-volatility brought by the magnetic character of the electrodes, these elements are ideally suited for memory applications. Freescale launched the first MRAM product in 2006 (4Mbit chip). In Freescale’s technology, the switching of the magnetization in the MRAM cell is achieved by local pulses of magnetic fields. More recently, alternative write schemes have been proposed which offer good down size scalability at least down to the 22nm technological node. They are based on spin-transfer writing. It was predicted in 1996 and experimentally observed for the first time in 2000 that the injection of a spin polarized current of sufficient density in a magnetic nanostructure can be used to switch the magnetization of this nanostructure. Therefore, it became possible to switch the magnetic configuration in an MTJ from parallel (low resistance state) to antiparallel (high resistance state) and vice-versa by sending bipolar pulses of current directly through the tunnel junction. The ultimate scalability of MRAM is actually expected to be achieved by a combination of spin transfer and thermally assisted switching (TA-ST-MRAM).
Beside standalone MRAM, this hybrid CMOS/MTJ technology can also find important applications in logic by allowing to intimately mix memory and logic functions (non-volatile logic). In particular, this may help reducing the power consumption of electronic systems.