Exchange-Biased Magnetic Multilayers for Nanomagnet Logic Applications

Nanomagnet logic (NML) is a spin-based logic paradigm and one of the potential alternatives to CMOS based logic as we approach the end of Moore's law. In order to become a viable next generation device, NML needs an electrical means to sense the states of output magnets. Furthermore, the error rate and the power dissipation have to be sufficiently low to make NML a reliable and low-power system. To this end, we pursued experimental study on exchange-biased systems to solve these critical problems in NML. First, we investigated the switching behavior of CoFeB/MgO/CoFeB/Ru/CoFe/PtMn based magnetic tunnel junctions (MTJ). We observed a presence of parasitic bias field in the free layer of the nanoscale MTJ which can impede proper operation of the MTJ as an NML output.

As an alternative to the MTJ we developed a CoFe/Cu/CoFe/IrMn based giant magnetoresistance (GMR) stack that is more suitable as an NML output. Current-in-plane (CIP) magnetoresistance (MR) ratio and exchange bias field of the blanket GMR stack were improved up to 4.5% and 120 mT, respectively. The optimized GMR stack was then integrated with the NML platform.

The clock-field requirement and power dissipation in NML can be lowered by improving the field coupling strength through inter-magnet space reduction. In this regard, a double e-beam exposure technique with 3δ overlay accuracy of 4 nm was developed to fabricate NML datalines with sub-10-nm spacing. Reduction of spacing from 30 nm to 10 nm lowered the power dissipation by 77%.

Dependency of error rate on the coupling strength was studied by varying the inter-magnet spacing. A 46% improvement in error rate was achieved by reducing the spacing from 30 nm to 10 nm. Types and causes of error in ultra-dense NML dataline were also investigated.

Finally, we proposed a novel approach for synchronous and directional operation of in-plane NML circuits that leverages the unidirectional anisotropy properties of exchange-biased systems and asymmetric field distribution of shape-engineered magnets. This approach promises to eliminate errors in NML by avoiding the stochastic switching of nanomagnets and ensuring simultaneous arrival of input bits at the logic gates.