Mitchell Falgoust / Chemistry & Biochemistry / Faculty Mentor: Peter Kroll
We have developed a machine learning interatomic potential using moment tensor potentials (MLIP-MTP) to describe systems comprising Si-C-O-H. We use DFT as the ideal model for our MLIP to fit. Using an interface with LAMMPS, we Its purpose is to describe low temperature vibrations and higher temperature reactions involving SiCOH polymeric precursors and SiCO ceramics. In our initial results we show vibrational density of states plots (vDOS) for models which are remarkably similar to DFT. This includes polymeric models as well as ceramic models with included excess carbon. We also show large scale reaction simulations, include a system of nearly one million atoms with local DFT-like behavior, which see the conversion from SiCOH polymers to SiCO ceramics.
Leave a Reply