Fast and realistic 3D feature profile simulation platform for plasma etching process

Abstract

We present a topographic simulation platform that simultaneously considers three-dimensional surface movement, neutral and ion transport, and surface reactions in plasma high-aspect-ratio (HAR) oxide etching. The hash map data structure is considered for an effective 3D level set algorithm with parallelized computations to calculate surface moving speed. Neutral and ion transport within nanoscale semiconductor geometry is parallelized with a graphics processing unit (GPU) so that the speedup ratio as compared to a single central processing unit (CPU) is approximately 200. The surface reaction based on a two-layer model was incorporated into a 3D feature profile simulation platform with CPU parallelization. Finally, our simulation platform demonstrates that adaptive surface meshing can drastically decrease the computational load with a parallelized numerical platform.

Review and future perspective of feature scale profile modeling for high-performance semiconductor devices

“…As an advanced shock-tracking method, a level-set method with more accuracy and stability of the etch front was proposed and used to predict the feature scale profile by many researchers. [29][30][31][32][33][34][35][36][37][38][39] In this method, a level-set function Φ, in which the distance from lattice points in the calculation region to the interface was an indicator, was used, and a zero value of Φ was recognized as the etch front by solving the Hamilton-Jacobi equation. 29 Makabe et al used a grid-weighted and level-set method to simulate the variations in the 2D feature scale profiles of the SiO 2 trenches etched by CF 4 (5%)/Ar plasma.…”

“…As an advanced shock-tracking method, a level-set method with more accuracy and stability of the etch front was proposed and used to predict the feature scale profile by many researchers 29 – 39 In this method, a level-set function Φ, in which the distance from lattice points in the calculation region to the interface was an indicator, was used, and a zero value of Φ was recognized as the etch front by solving the Hamilton–Jacobi equation 29 …”

“…Adalsteinsson and Sethian 36 , 37 proposed feature scale profile models using 2D and 3D level-set methods. Recently, introducing a GPU parallelized algorithm, Yook et al 38 . improved the K-SPEED code, 39 which can simulate the 3D feature scale profiles of SiO2 HARC etching with a Monte Carlo gas transportation process, thereby shortening the calculation time.…”

Science-based, data-driven developments in plasma processing for material synthesis and device-integration technologies

“…The feature-scale kinetic MC etch model determines the asymmetries in the postetched feature profiles. 137) The topographic simulation platform simultaneously considers 3D surface movement, neutral and ion transport, and surface reactions in plasma HAR oxide etching. 137) The stochastic charging of the inside surfaces of features resulted in tilting of the HAR features in random directions.…”

“…137) The topographic simulation platform simultaneously considers 3D surface movement, neutral and ion transport, and surface reactions in plasma HAR oxide etching. 137) The stochastic charging of the inside surfaces of features resulted in tilting of the HAR features in random directions. 138) The ion energy and angular distribution functions (IED and IAD) of ions and fast neutrals were estimated computationally by a hybrid model of fluid and particles and they were validated across a range of operating conditions of experiments.…”