EUROPEAN CONGRESS AND EXHIBITION ON ADVANCED MATERIALS AND PROCESSES - EUROMAT 2021

Computational studies of the phase microstructure can assist greatly in the design of Cu-Sn material for joining and energy applications. The phase field method is a very popular method for mechanistic studying of the spatial-temporal evolution of phases in the material’s microstructures. In Cu-Sn system consisting of LIQUID (Sn-rich), IMC (Cu6Sn5) and FCC (Cu-rich) phases at 523.15 K, the knowledge of interfacial energies at the interphase and intergrain boundaries is essential during the development of the multi-phase field model. While the interface energy for the solid/solid (FCC/IMC and IMC/IMC) interfaces has been clearly outlined as 0.3 J/m2 in the literature of previous works, there are several discrepancies regarding the reporting of liquid/solid (LIQUID/IMC and LIQUID/FCC) interfaces. For a computational study having pre-existing IMC grains at the interface of LIQUID and FCC , the LIQUID/FCC interfacial energy (σ_(L/F)) plays a lesser role in guiding the phase morphologies once the IMC attains a layered structure, and so it is assumed to be of a value equal to that of LIQUID/IMC interface energy (σ_(L/I)). This simplifies the system to possessing the uncertainty in only one variable, namely, σ_(L/I). For the present study, based upon the varying interface energy (VIE) formalism, the microstructural image datasets are created by performing simulations with σ_(L/I) designated at different values in the range 0.05-0.25 J/m2. These image datasets are assigned different labels corresponding to values of σ_(L/I) and are then used to train a convolutional neural network. Then the label in the prediction model that best matches with the experimental microstructural image is considered as the optimum LIQUID/IMC interface energy for the Cu-Sn system. Through this study it can be illustrated how microstructural simulation can serve as the source of datasets in the data-driven study of materials’ interfacial energies.