Net
Prediction of Quantum Yields of Monolayer WS2 by Machine Learning
페이지 정보
| PUBLICATION | Advanced Optical Materials (IF 8.0, Top 17%), 2023 |
| AUTHORS | Jolene W. P. Khor, Trang Thu Tran, Anir S. Sharbirin, Sammy X. B. Yap, Hyunseung Choo, Jeongyong Kim |
ABSTRACT
Abstract
Monolayer transition metal dichalcogenides (1L-TMDs) exhibits distinct light emissions in the visible range, making them suitable for 2D optoelectronic applications. Photoluminescence quantum yield (PLQY) is a key factor for practical applications of 1L-TMDs. However, the methods for PLQY measurements of 1L-TMDs suffer from limitations due to the small sample size and typically low PLQY, which require a complex measurement setup. In this study, machine learning (ML) models are developed to predict the PLQY of monolayer tungsten disulfide (1L-WS2) using data extracted from 1208 PL spectra and corresponding measurement conditions as the ML training and testing data set. The ML model shows a high accuracy with R2 value of 0.744 and a mean absolute percentage error of 44% in the prediction of widely ranged PLQYs of 1L-WS2 from 0.07% to 38%. This data-driven prediction not only enables the convenient PLQY estimation of 1L-TMDs, but also helps in identifying key parameters influencing PLQYs.
Monolayer transition metal dichalcogenides (1L-TMDs) exhibits distinct light emissions in the visible range, making them suitable for 2D optoelectronic applications. Photoluminescence quantum yield (PLQY) is a key factor for practical applications of 1L-TMDs. However, the methods for PLQY measurements of 1L-TMDs suffer from limitations due to the small sample size and typically low PLQY, which require a complex measurement setup. In this study, machine learning (ML) models are developed to predict the PLQY of monolayer tungsten disulfide (1L-WS2) using data extracted from 1208 PL spectra and corresponding measurement conditions as the ML training and testing data set. The ML model shows a high accuracy with R2 value of 0.744 and a mean absolute percentage error of 44% in the prediction of widely ranged PLQYs of 1L-WS2 from 0.07% to 38%. This data-driven prediction not only enables the convenient PLQY estimation of 1L-TMDs, but also helps in identifying key parameters influencing PLQYs.