Deep characterization of the electrical features of Ag/P3HT/SiNWs Schottky diodes by machine learning models based on experimental study

Abstract

In the present paper, silicon nanowires (SiNWs) surface covered by organic semiconducting polymer (P3HT) were used as the base material for the formation of the Ag/P3HT/SiNWs Schottky diode. Based on Current-Voltage (I-V) experimental study, different machine learning models are used to predict the current output of the Ag/P3HT/SiNWs diode taking into account changes in the applied voltage and immersion time of SiNWs in P3HT polymer solution. We used mathematical, decision tree, and Artificial Neural Network (ANN) models. The ANN model is the best model describing the changes of I-V characteristics with different immersion durations. The ANN, which is a perception layer with 48 neurons in its hidden layer, was trained using real current readings collected at different immersion times in P3HT solution ranging from 30 min to 210 min and voltages from -10 V to +10 V. ANN model achieves a mean squared error of 4.24 10−7, a mean absolute error of 3.82 10−4, a root mean square error of 6.7 10−4, and an R-squared value of 0.999997, indicating its exceptional predictive accuracy. The input layer of the model includes 244 data points representing time and voltage values, while the output layer provides the corresponding values of 844 data points. Using only two hidden layers, this model accurately predicts the current output for the Ag/P3HT/SiNWs Schottky diode. Comparing it with other models, the ANN method offers very accurate and fast predictions.