An artificial intelligence model estimation for functionalized graphene quantum dot-based diode characteristics

Abstract

The use of artificial intelligence (AI) approaches, one of the most significant technological advances of the 21st century, to determine the diode parameters that can be obtained from Schottky diode characterization allows data to be collected, processed, analyzed, and results obtained faster than ever before, with high accuracy. It also supports the development of a selection and modeling tool for future studies and, most importantly, facilitates modeling implementation with faster and fewer experimental results. In this context, this study presents a machine learning (ML) model to predict diode parameters from current-voltage (I-V) measurements of polyethyleneimine-functionalized graphene quantum dots (GQDs)-based Schottky hybrid diode. This study used K-Nearest Neighbor, Random Forest, Multilayered Perceptron, and Support Vector Machine algorithms. In research, the lowest model error of each model was compared, and the performance of the models obtained was evaluated. In addition, out of 30 diodes on the fabricated structure, the diode with the best rectification ratio(RR) was identified within a few seconds using machine learning, verified to be the same as the diode selected by the researchers, and the optimal estimation of RR and ideality factor(n) was made from the diode parameters using the thermionic emission (TE)method. The experimental results were compared with machine learning models. Among these algorithms, the RF algorithm performed best with a root mean square error (RMSE)value of 4.1 E-05 and an R-squared value of(R2) 0.999998. The success of RF depends on the characteristics of the dataset used, its size, and data distribution. RF reduces overfitting by taking the average of multiple decision trees and is less sensitive to noise and outliers in the data. The results obtained will allow the development of a selection and modeling tool for future studies and, most importantly, facilitate modeling faster and with fewer experimental results, enabling the integration of AI into science.