About this Abstract |
Meeting |
MS&T24: Materials Science & Technology
|
Symposium
|
Machine Learning and Simulations
|
Presentation Title |
New Machine–Learning Interatomic Potentials (MLIPs) for Si-C-O-H Compounds Enabling Atomistic Simulations of Complex Chemical Transformations |
Author(s) |
Mitchell Falgoust, Shariq Haseen, Peter Kroll |
On-Site Speaker (Planned) |
Mitchell Falgoust |
Abstract Scope |
We present new Machine–Learning Interatomic Potentials (MLIPs) for Si-C-O-H compounds, enabling simulations of structure, properties, and thermal conversion from polymer to ceramic. The MLIPs are constructed from Moment-Tensor-Potentials (MTPs) and trained to a library of configurations with added active and hybrid learning strategies.
The MLIPs reproduce vibrational properties of polymers and SiCOH structures obtained from aiMD simulations, thus providing a tool to identify chemical units and distinct structural characteristics through their vibrational properties. Simulations of the polymer-to-ceramic transformation deliver details of chemical reaction mechanisms during the pyrolysis of polysiloxanes, including methane abstraction and Kumada-like rearrangements that transform the siloxane backbone. Simulations of millions of atoms for several nanoseconds show the development of mixed tetrahedra in SiCO ceramics and align with experimental observations. Moreover, they showcase precipitation of graphitic nanosheets from carbon-rich SiCO precursors. Overall, the new MLIPs deliver simulations for Si-C-O-H compounds with “DFT-like” quality at low and high temperatures. |