Carnival on Theoretical/Computational Sciences: Molecular Dynamics

Thanks to GMP:


As I've talked about before, my specialty within materials science in molecular dynamics. I wanted a change of pace from the research I had done as an undergraduate, which was in metal fatigue and creep. I had spent a lot of time (some creep tests literally last years) on studying *what* happened, and I wanted to start answering *why*. Fracture mechanics is a surprisingly young field, with the earliest theoretical models dating back to WWI, but really only achieving broad interest in the 1950s. Many of the models are empirically derived, especially in composite systems.


In the interests of remaining pseudonymous, I can't talk too much about the specific systems I study without very quickly revealing my research group.  Basically, I study interfaces between polymer and nonpolymer systems. These are particularly interesting, because there's no good way to study them experimentally without fundamentally changing the structure, and therefore the properties, of the interface. 


Molecular dynamics (MD) is an atomistic method, which can generally tackle problems in nanometer length scales and nanosecond time scales. If you're designing an engine, it may not seem particularly applicable, but the information we gather at these scales can be fed into continuum level models. In turn, MD often turns to quantum mechanical techniques like density functional theory to improve our models.


I love how broadly applicable molecular dynamics is. You can study everything from mechanics of everyday composites to the structure of materials in the earth's core. I also like bridging the gap between fundamental science and practical engineering: by studying the atomistic mechanics at an interface, you can get a better idea of what controls those interacts, and predict novel materials for adhesion without some of the expenses associated with traditional experiments. 


I don't always love the coding side of my job: my formal training in programming is limited, and most of the intro classes at GiantU teach Java and Python instead of C++ and Fortran. Fortunately, there are some fantastic open source codes for MD, such as LAMMPS. Because GiantU has it's own high performance computing facility, I'm lucky enough to have some very awesome system administrators to help compile code that's very well parallelized, and so that's not a major part of my job, like it might be at a smaller school.  


Everything I do is fundamentally informed by experimental results, even when exact experiments don't exist. If my bulk materials don't have a realistic density, or realistic elastic properties, I can't confidently draw conclusions from my simulations. This does seem to mean that I end up reading twice as many papers as my experimental counterparts, who tend not to explore the computational literature they way I end up exploring the experimental literature. However, I really do like what I'm doing, or else I would definitely be doing something else by now...