Enabling Controlled Material Synthesis and Processing via Predictive Modeling and Simulation

Abstract

Computational models that can predict materials’ evolution under synthesis and processing conditions, along with complementary experimental characterizations, are powerful tools that provide insights into the process designs for materials with tailored properties. This dissertation presents a collection of computational models and methods that enable predictions of material structures under different conditions and sample-temperature control during experiments. The first part of this dissertation describes computational models for simulating phase transformations, microstructure evolution, and mechanical behavior. A phase-field model that captures the evolution of ionic concentrations and phase fractions during solid-state metathesis (SSM) reactions was first presented. This model was employed to investigate the effect of mobilities of ions on the reaction dynamics. We identified the expressions for effective mobility for each type of ions and showed that the type of ions with a larger effective mobility dominates the diffusion and that the rate of the overall process is set by an overall characteristic mobility of the reaction. This phase-field model was then utilized to predict the phase evolutions during the FeS2 synthesis reaction. The simulation predicted nonplanar phase evolution, which was recently observed in the experiment via transmission electron microscopy. The model was also applied to study the effect of particle packing on the reaction rate, along with a lattice model. Simulations suggested that reactions could occur more rapidly when the sample is densified since each particle within the sample is more likely to have a larger number of reactive neighbors and a particle with more reactive neighbors tends to react faster. Two phase-field models that describe the evolution of microstructure with dislocations were then discussed. A simple model assuming a uniform intra-granular dislocation density was employed to study the macroscale translation of grains during non-isothermal annealing. Grain translation was observed from the simulations, which was the net effect of a grain with a medium dislocation density consuming an adjacent grain with a higher dislocation density and simultaneously being consumed by another adjacent grain with a lower dislocation density. An extended model that considers intra-granular dislocation densities variation was utilized to study the effect of cyclic heat treatment on the microstructure evolution. We showed that this extended model yielded results that closely resemble the experimental data and that non-self-similar evolution leading to multi-modal grain size distribution was observed in the simulations after dislocations were injected into the microstructure three times. Finally, the relaxation of flat and buckled triangular monolayers of atoms was achieved using a phase-field-crystal model. The second part of this dissertation presents heat transfer models designed to assist in sample-temperature control and a machine learning algorithm for parameter optimization. First, a heat transfer model that describes the temperature distribution within a sample in an optical floating zone (OFZ) experiment was discussed. The effectiveness of the machine learning algorithm was demonstrated by applying it to determine uncertain parameters in the heat transfer model for the OFZ experiment. The parameterized OFZ model accurately reproduced both steady-state thermal profiles and time-dependent temperatures. Additionally, a coupled thermal and Joule heating model was implemented to predict the thermal profiles of a sample heated by a gradient heater designed to produce a temperature gradient within the sample. The parameterized model was used to study the effect of the heater geometry on the resulting thermal profiles within the sample.