Machine learning algorithms for predicting the photoionization cross section of CdS/ZnSe core/shell spherical quantum dots surrounded by dielectric matrices

In this study, we explore the prediction of the photoionization cross section (PCS) of CdS/ZnSe core/shell spherical quantum dots (CSQD) surrounded by different dielectric matrices. The quantum dot systems, embedded in polyvinyl alcohol (PVA), polyvinyl chloride (PVC), and silicon dioxide (SiO₂) matrices, were modeled under varying core-shell dimensions and dielectric environments. Our findings show that the resonant peak of the PCS experience a redshift with improvement in their amplitude in the case of the PVA matrix, while in the case of the PVC and SiO_2, the magnitude of the PCS is reduced and their resonant peak suffers a blueshift. Three different machine learning algorithms were used to estimate the photoionization cross-section, namely Artificial Neural Networks (ANN), Decision Trees (DT), and Random Forest Regressors (RFR). Among these, Random Forest Regression proved to be the most successful algorithm, particularly for the SiO₂ matrix, achieving exceptional performance with the coefficient of determination $R^2$ = 0.999 Mean Squared Error $\left(MSE\right)={10}^{-4}$ and the Root Mean Squared Error $\left(RMSE\right)=0.0077$. While DT exhibited lower MSE, MAE, and RMSE than ANN in the SiO₂ matrix, ANN showed potential in capturing more complex nonlinear relationships. These results demonstrate the superior predictive capabilities of RFR and highlight the applicability of machine learning in modeling quantum dot systems. This work offers valuable insights into the optimization of optoelectronic device design through accurate and efficient computational methods.