Data-driven assessment of chemical vapor deposition grown MoS2 monolayer thin films

Abstract

Growth of high quality two-dimensional transition metal dichalcogenide monolayers with the desired microstructure and morphology is critical for enabling key technological solutions. This is a non-trivial problem because the processing space is vast and lack of a priori guidelines impedes rapid progress. A machine learning approach is discussed that leverages the data present in published growth experiments to predict growth performance in regions of unexplored parameter space. Starting from the literature data on MoS2 thin films grown using chemical vapor deposition (CVD), a database is manually constructed. Unsupervised and supervised machine learning methods are used to learn from the compiled data by extracting trends that underlie the formation of MoS2 monolayers. Design rules are uncovered that establish the phase boundaries classifying monolayers from other possible outcomes, which offers future guidance of CVD experiments.

Chemistry-Informed Machine Learning Enables Discovery of DNA-Stabilized Silver Nanoclusters with Near-Infrared Fluorescence

“…(Random forest is an ensemble learning method consisting of many distinct decision trees, where the collective predictions of the decision trees are used to determine the model’s output.) This approach has provided insights into nanomaterial synthesis conditions 54 and methane uptake by metal–organic frameworks. 53 For each pair of color classes, at most 16 of the 144 staple features scored higher than the most important shadow feature: i.e., sufficiently higher than random.…”

Machine-Learning-Enabled Prediction of Adiabatic Temperature Change in Lead-Free BaTiO3-Based Electrocaloric Ceramics

“…Random forest is an ensemble-based learning method, where many unbiased or decorrelated tree-based models are generated from the training data and the final outcome is decided based on aggregating the results from each tree. , The process of constructing many models is enabled by the bootstrap sampling method. It has two hyperparameters: (1) number of trees in the forest (ntree) and (2) number of variables to be chosen per node split (mtry) .…”

Chemical vapor deposition of 2D materials: A review of modeling, simulation, and machine learning studies

“…Copyright © 2020 Elsevier Ltd. All rights reserved (B) Feature importance plot depicting the relative significance of the growth parameters in the order Mo precursor temperature > growth pressure > growth time > sulfur precursor temperature > highest growth temperature. Adapted from Costine et al, 2020 . Copyright © 2020, the Authors.…”

“…Recently, Costine et al. developed an ML-based approach to predict the growth of MoS 2 using previously unexplored process parameters, using data available from the literature on the CVD synthesis of MoS 2 ( Costine et al, 2020 ). An unsupervised metric multidimensional scaling (MDS) analysis was performed using the growth variables, namely, Mo and S precursor temperature, maximum temperature of growth, growth time, and pressure.…”

“…No tangible “favorable regions for growth” were found using the MDS analysis and hence, a supervised ML strategy was applied. The importance of various features was extracted from a random forest ML model ( Figure 9 B), indicating the Mo precursor temperature to have the most significant effect on the growth process ( Costine et al, 2020 ). Lee et al.…”