Design of thermal conductivity of mercapto group-activated graphene/epoxy nanocomposites using the molecular dynamics simulation and Gaussian process regression-based Bayesian optimization

An excessively high coverage rate (CR) can decrease the effective thermal conductivity (TC) of nanocomposites owing to a decrease in the intrinsic TC of the nanofillers. In this study, we propose a design framework to predict the optimal CR for achieving the highest TC in mercapto-group-activated graphene (SH@GNP)/epoxy nanocomposites. This framework integrates molecular dynamics (MD) simulations, effective medium theory, and Gaussian process-regression-based Bayesian optimization (GPR-BO). The interfacial phonon vibrational coupling (i.e., the overlap factor), positively correlated with the interfacial thermal conductivity (ITC) between the nanofiller and the matrix, was employed to efficiently determine the optimal CR for the highest ITC and consequently accelerate the design framework. The obtained optimal CR for the highest ITC was used for the initial sampling points of the GPR-BO to determine the optimal CR for the highest effective TC of the nanocomposites because the optimal CR for the effective TC was within the initial sampling points owing to the competitive relationship between the TC of SH@GNP and ITC. The optimization results of the design framework indicated that the proposed framework effectively reduced the computational time required for repetitive MD modeling and simulations.