In order to meet the rising demand for greater computational power, semiconductor manufacturing processes continue to become more and more complicated [2]. To fabricate these devices, a silicon substrate undergoes upwards of 500 process steps to create a single product [1]. These steps typically consist of a rotation between bulk deposition of a material, photolithography to form the desired structure, and etch to remove the undesired material. Recently, a new fabrication method known as “Area-Selective Atomic Layer Deposition” (ASALD) has received attention due to its ability to combine the three aforementioned process steps into a singular process step. This comes by taking advantage of the patterns in the underlying layer. Specifically, the substrate can be separated into a Non-Growth Area with a very low deposition rate and a Growth Area with a high deposition rate [3]. Thus, deposition will selectively occur on the Growth Area, removing the need for a photolithography and etch step. One ASALD reaction that has great potential for reducing the number of process steps is the deposition of Al₂O₃ on SiOH as the Growth Area with SiH as the Non-Growth Area. However, due to the heavy reliance on the selectivity between the Non-Growth Area and the Growth Area, reaction selection is critical, as a low selectivity will result in undesired deposition in the Non-Growth Area.
Mesoscopic and atomistic modeling offer a way to analyze and predict this selectivity behavior. In particular, mesoscopic simulations using kinetic Monte Carlo (kMC) methods can capture how selectivity evolves over time and across deposition cycles. This approach allows for time-resolved modeling that can reveal when the selectivity begins to diverge from experimentally observed values—especially in scenarios where the ideal high selectivity cannot be maintained across multiple batches. Rather than seeking an optimal operating condition, this work aims to identify the batch number at which selectivity loss begins to occur. To support this, atomistic modeling is used to first break down the ASALD process into its elementary reactions and extract relevant reaction coefficients such as activation energies and sticking coefficients. These coefficients are generated using numerical methods such as electronic structure optimization and the nudged elastic band technique [5]. With these constants, reaction rates in both the Growth Area and Non-Growth Area can be calculated and fed into the mesoscopic kMC model [4]. The combined approach allows us to study how selectivity degrades under realistic processing conditions, enabling better understanding of process limitations and informing future experimental or industrial implementations.
- Richard, C., 2023. Understanding Semiconductors: A Technical Guide for Non-Technical People. Apress, Berkeley, CA.
- Jurczak, M., Collaert, N., Veloso, A., Hoffmann, T., Biesemans, S., 2009. Review of FINFET technology. In: 2009 IEEE International SOI Conference, 1–4, Foster City, CA, USA.
- Yarbrough, J., et al., 2023. Area-Selective Atomic Layer Deposition of Al2O3 with a Methanesulfonic Acid Inhibitor. Bent Chemistry of Materials 35 (15), 5963-5974
- Wang, H., M. Tom, F. Ou, G. Orkoulas and P. D. Christofides, ''Multiscale Computational Fluid Dynamics Modeling of an Area-Selective Atomic Layer Deposition Process Using a Discrete Feed Method,'' Dig. Chem. Eng., 10, 100140, 2024.
- Yun, S., F. Ou, H. Wang, M. Tom, G. Orkoulas and P. D. Christofides, "Atomistic-Mesoscopic Modeling of Area-Selective Thermal Atomic Layer Deposition," Chem. Eng. Res. & Des., 188, 271-286, 2022.
