Polymer-based thermal interface materials (TIMs) are indispensable for reducing the thermal contact resistance of high-power electronic devices. Owing to the low thermal conductivity of polymers, adding multiscale dispersed particles with high thermal conductivity is a common approach to enhance the effective thermal conductivity. However, optimizing multiscale particle matching, including particle size distribution and volume fraction, for improving the effective thermal conductivity has not been achieved. In this study, three kinds of filler-loaded samples were prepared, and the effective thermal conductivity and average particle size of the samples were tested. The finite element model (FEM) and the random thermal network model (RTNM) were applied to predict the effective thermal conductivity of TIMs. Compared with the FEM, the RTNM achieves higher accuracy with an error less than 5% and higher computational efficiency in predicting the effective thermal conductivity of TIMs. Combining the abovementioned advantages, we designed a set of procedures for an RTNM driven by the genetic algorithm (GA). The procedure can find multiscale particle-matching ways to achieve the maximum effective thermal conductivity under a given filler load. The results show that the samples with 40 vol %, 50 vol %, and 60 vol % filler loading have similar particle size distribution and volume fractions when the effective thermal conductivity reaches the highest. It should be emphasized that the optimized effective thermal conductivity can be improved obviously with the increase in the volume fraction of the filler loading. The high efficiency and accuracy of the procedure show great potential for the future design of high-efficiency TIMs.
Keywords: effective thermal conductivity; genetic algorithm; optimization; random thermal network model; thermal interface materials.