October 3 Simulation Work

***Being posted on October 6th after having to leave for lunch on the 3rd***

I've always been quite grateful for the breadth that my project encompasses (while the other half of the time I'm panicking about how any of the objectives will be completed within the given time frame).

As an experimentalist, I work on two very different behaving materials, polycrystalline nickel and nanocrystalline nickel. While on the characterization side the primary source of my data comes from EBSD, but the CMU Physics group has also provided an HEDM dataset which provides an immense amount of information to be found.

However, to play around with simulation work really adds another side to my research work. While we design experiments with controls and variables to investigate what we like, ultimately natural laws and physics still dictate the behavior of atoms and molecules. What we discover, helps us understand the things that were not clear before. But when one works with simulations, they dictate the laws, they make the rules, the make things that cannot happen. We, the scientist, get the opportunity to truly manipulate the experiment.

Of course, we could design anything, but in the end, we have to design simulated microstructures that behave similar to real microstructures in order to contribute to the field of science.

The model I primarily work with for microstructure evolution is the Potts model, which labels each individual grain as a spin "number", and grows or shrinks based on the number of neighbors. As a result, the model is curvature driven and aims to reduce the overall energy by eliminating the number of grains with time, thereby eliminating grain boundaries. It should be important to point out that grain boundaries are indiscrete in this model, that is, they are not defined, but exist at the points where two different spins interact. This brings about an interesting point that we assign factors such as grain boundary mobilities, energy, and so forth, based on the original spin site and the neighboring spin sites that we pick at the simulation step. As we add more factors and complexity to our model, the probability to change a spin, that results in grain growth, becomes the aggregate of several energy arguments. Because all terms considered end up in the same summation, this leads to interesting competing effects or even domination of one term over all the others.

The simplicity of the Potts model is ideal as it keeps calculations to a minimum, maximizing the simulation speed. Potts model has been thoroughly applied to grain growth and recrystallization, and in turn, I eventually hope to develop a feasible Potts model exhibiting twinning.