Evaluation of Spin Lifetime in Thin-Body FETs: A High Performance Computing Approach

Abstract

Silicon, the prominent material of microelectronics, is perfectly suited for spin-driven applications because of the weak spin-orbit interaction resulting in long spin lifetime. However, additional spin relaxation on rough interfaces and acoustic phonons may strongly decrease the spin lifetime in modern silicon-on-insulator and trigate transistors. Because of the need to perform numerical calculation and appropriate averaging of the strongly scattering momenta depending spin relaxation rates, an evaluation of the spin lifetime in thin silicon films becomes prohibitively computationally expensive. We use a highly parallelized approach to calculate the spin lifetime in silicon films. Our scheme is based on a hybrid parallelization approach, using the message passing interface MPI and OpenMP. The algorithm precalculates wave functions and energies, and temporarily stores the results in a file-based cache to reduce memory consumption. Using the precalculated data for the spin relaxation rate calculations drastically reduces the demand on computational time. We show that our approach offers an excellent parallel speedup, and we demonstrate that the spin lifetime in strained silicon films is enhanced by several orders of magnitude.