Parallelized Level-Set Velocity Extension Algorithm for Nanopatterning Applications

Abstract

We present a parallelized algorithm for accelerating the velocity extension calculations in a level-set method, which is essential for surface velocity based topography simulations, such as etching or deposition simulations for nanopatterning applications. The proposed algorithm improves the prevailing fast marching method by optimizing the heap data structure and efficiently reordering the calculations. We implemented the algorithm into Silvaco’s Victory Process simulator, which is utilized for evaluating our algorithm with a three-dimensional simulation of an ion beam etching process used for spin-transfer torque magnetoresistive random access memory devices. Our results show a significant serial speed-up by a factor of at least 1.4 and a total speed-up by a factor of up to 8 using 8 threads for the velocity extension.