Machine learning and informatics efforts in polymer science are hindered by a lack of data: state-of-the-art techniques often require thousands of samples before accurate inferences can be made, but aside from a small handful of properties, polymer data sets typically contain only a few dozen entries. Molecular simulations offer a promising route to larger data sets but must be able to accurately capture the wide range in length and time scales in polymer systems. Here we report the design and development of an automated software pipeline for bottom-up multiscale modeling of arbitrary homo- and co-polymer systems. By integrating all-atom, coarse-grained, and mesoscopic simulations, a wide array of physical properties can be accessed, from solubility parameters to rheological responses. The various levels of simulation are connected by systematic coarse-graining methods and orchestrated by a Python module which also provides tools for data analysis and management. We discuss how this system can be integrated with active learning techniques for rapid materials discovery.
Preview Presentation
Presenter(s)
Once the content has been viewed and you have attested to it, you will be able to download and print a certificate for PDH credits.
If you have already viewed this content,
please click here
to login.