Ab-initio calculation driven machine learning based prediction of quantum capacitance of titanium-doped graphene for efficient supercapacitor electrode design

Abstract

In this work, for the first time, the effects of Titanium (Ti) doping concentration on the quantum capacitance of Ti-doped graphene electrode is theoretically investigated by combining the first principle-based density functional theory (DFT) simulation and machine learning (ML) modelling approach for electrical double-layer (EDL) supercapacitor design. The DFT simulation reveals that Ti-doping in graphene can significantly improve the quantum capacitance , thereby enhancing the total interfacial capacitance of the electrode, wherein the electrode performance can be efficiently tuned by varying the doping concentration. Furthermore, the accuracy of quantum capacitance prediction over different local electrode potential and Ti-doping concentrations using different ML algorithms , i.e. adaptive boost (AdaBoost), artificial neural network (ANN), and extreme gradient boosting (XGBoost) are systematically assessed based on error metrics (RMSE, MAE, R 2 ). The results reveal that the ANN is the best-suited algorithm for such predictive analysis, with >99 % accuracy for individual Ti-doping concentrations. Next, the robustness of the ANN algorithm is further validated by calibrating the predicted quantum capacitances with simulated profiles for different doping concentrations within a range of 3 % to 12 %. The innovative methodology introduced in this work can efficiently guide the experimental research on high-performance EDL Supercapacitor electrode design.

Topics & Concepts