Multiscale transport modeling of reactive sputtering for fabrication of neuromorphic hardware

Conventional computer technology is facing scaling limitations related to hardware architecture and power consumption. To overcome these limitations, memristive devices are currently under investigation in neuromorphic hardware that draws inspiration from biological principles. However, the reliable fabrication of memristive devices remains challenging, as the variation of electrical device properties is often severe. In particular, characteristics of devices have shown a strong correlation with plasma parameters used for deposition of thin films. In this work, a multiscale model for reactive sputter deposition of metal oxides is used to complement the experiments and assess the quantities that mainly influence the resulting properties of the deposited films. The model relates on a hybrid scheme for the plasma, test-particle simulations for the sputtered particle transport, and a kinetic Monte Carlo model for thin film growth. As a results of the study, the main plasma parameters, such as electron temperature and distribution of energetic ions, and their relation with electrical properties of the memristive devices produced on one wafer are presented.