Computational approach for plasma process optimization combined with deep learning model

Abstract

As semiconductor device structures become more complex and sophisticated, the formation of finer and deeper patterns is required. To achieve a higher yield for mass production as the number of process steps increases and process variables become more diverse, process optimization requires extensive engineering effort to meet the target process requirements, such as uniformity. In this study, we propose an efficient process design framework that can efficiently search for optimal process conditions by combining deep learning (DL) with plasma simulations. To establish the DL model, a dataset was created using a two-dimensional (2D) hybrid plasma equipment model (HPEM) code for an argon inductively coupled plasma (ICP) system under a given process window. The DL model was implemented and trained using the dataset to learn the functional relationship between the process conditions and their consequential plasma states, which was characterized by 2D field data. The performance of the DL model was confirmed by comparison of the output with the ground truth, validating its high consistency. Moreover, the DL results provide a reasonable interpretation of the fundamental features of plasmas and show a good correlation with the experimental observations in terms of the measured etch rate characteristics. Using the designed DL, an extensive exploration of process variables was conducted to find the optimal processing condition using the multi-objective particle swarm optimization (MOPSO) algorithm for the given objective functions of high etch rate and its uniform distribution. The obtained optimal candidates were evaluated and compared to other process conditions experimentally, demonstrating a fairly enhanced etch rate and uniformity at the same time. The proposed computational framework substantially reduced trial-and-error repetitions in tailoring process conditions from a practical perspective. Moreover, it will serve as an effective tool to narrow the processing window, particularly in the early stages of development for advanced equipment and processes.

Review: Machine learning for advancing low-temperature plasma modeling and simulation

“…44 Instead of a grid-based approach for training and prediction (where the whole spatial domain is regressed on at once), MLPs were trained on phase-average plasma quantities sampled at varying positions within the discharge. The dynamics of an inductively coupled plasma with additional radio frequency substrate bias has been implemented based on hybrid plasma equipment model 45 simulations by Ko et al 46 2D axisymmetric simulations of an Ar etch plasma were performed with a fixed geometry (rectangular mesh) and varying process conditions (pressure, source power, bias power at 1 and 13.56 MHz). The devised data set was further used to train a multi-encoder/decoder ANN, which encodes the process variables and geometry into a latent representation to finally decode the corresponding phase-averaged discharge quantities.…”

Data-driven prediction of the output composition of an atmospheric pressure plasma jet

“…This means that the output concentrations correspond to the plasma plume and are consistent with the plasma optical emission spectrum [29,30]. Also, ML methods, especially those using neural networks, can control and automatically optimize the plasma generator performance [31][32][33]. In the last decade a new technique namely 'physics-informed data-driven modeling' was developed especially the one using a physics-informed neural network (PINN) [34].…”