Exploring the nonlinear conductive properties of polymer/graphene composites at the molecular level: a machine learning approach

Abstract

Polymer/graphene (Py/GN) composites under the influence of external electric fields often exhibit unique nonlinear conducting behaviors. However, the underlying mechanism of this field effect at the molecular level is still obscure until now. Herein, the evolution of electrical properties of Py/GN composites induced by electric fields has been explored by combining high-throughput first-principles calculations with machine learning models. The results show that the polymer valence band maximum (PVBM) and polymer conduction band minimum (PCBM) of Py/GN composites under different electric fields can be accurately predicted by the XGBoost regression algorithm. The band arrangement of polymers in Py/GN composites can be easily altered with applied electric fields, where charges accumulate around the graphene layer and depleted around the polymer layer. Moreover, the electrons at the pyrrole/GN interface may overcome the Schottky barrier height, leading to a transition from a Schottky contact to an ohmic contact under a critical field (E_b), which can be effectively predicted with an R² value of 0.854. Then, two types of novel Py/GN composites with lower or higher E_b values were screened by reverse engineering the ML model, offering valuable guidance for the application of Py/GN composites in different electric field conditions. This work can provide new insights into the nonlinear electrical response of Py/GN composites under electric fields, which has significant implications for practical applications.