The application of machine learning to polymer physics has traditionally struggled with two major challenges: a lack of large, curated datasets and the need to understand the physics behind the machine learning prediction. Here we aim to simultaneously tackle these challenges through the incorporation of domain knowledge often in the form of theory, thus, providing improved predictions for smaller datasets while improving explainability. This effort is under the Polymer Analytics project.
To spur advances in machine learning for polymer science, we also collaborate with external partners to develop the polymeric databases that such algorithms require. In collaboration with MIT, University of Chicago, Citrine Informatics, and Dow we are developing a Community Resource for Innovation in Polymer Technology (CRIPT). This project is led by Prof. Brad Olsen and funded by the NSF Convergence Accelerator. Using FAIR data principles, CRIPT helps users input, visualize and share polymeric data. In collaboration with Prof. Ian Foster, Prof. Juan de Pablo and colleagues at University of Chicago we are developing the Polymer Property Predictor and Database. Specifically, we are exploring methods for extracting polymer properties from the literature using machine learning and are also working on building a polymer rheology database based on experiments from the Center for Hierarchical Materials Design (CHiMaD). These efforts are under the Polymer Analytics project.
The need to address the growing waste stream of plastics has become a global challenge. To enable next generation recycling, we apply simulation techniques and machine learning to improve characterization of a common polymer, linear low density polyethylene and to identify methods for upcycling mixed waste streams through compatibilization. We also are investigating how to improve near infrared measurements of polyolefins through correlation with other slower, measurement techniques. All of these efforts are in the Polymer Analytics project and in collaboration with the Macromolecular Architectures project.
We help develop ZENO, a Monte Carlo based code that computes several quantities including the intrinsic viscosity and hydrodynamic radius. It has been previously shown to generate results within experimental uncertainty. More information including links to the source code and web app version can be found at https://zeno.nist.gov/
Aqueous solutions of oppositely charged polymers, under certain conditions, can macrophase separate yielding a polymer rich phase known as a complex coacervate and a polymer poor phase often with vanishing polymer concentrations. We explore this phase separation in systems of copolymers containing charged blocks and neutral, hydrophilic blocks, as well as homopolymers using a combination of simulation and theory. This effort is a part of the Fundamentals of Macromolecular Self Assembly project.
A selected list of publications is below. For a complete list see Google Scholar.
"Incorporating polymer theory in machine learning for improved prediction and understanding," Polymer Physics Gordon Research Conference, South Hadley, MA, 2022
"Leveraging polymer theory for improved machine learning," Materials Research Society Spring Meeting, Virtual, 2022
"Enabling machine learning for polymer science," University of Delaware NRT Community Hour, Virtual, 2022
"Overcoming data scarcity in polymer science," CHiMaD/NIST ODI Seminar Series, Virtual, 2022
"Improving prediction and understanding with theory aware machine learning," American Physical Society March Meeting, Chicago, IL, 2022
“Enabling polymer informatics through databases,” Center for Integrated Nanotechnologies Annual Meeting Symposia, Santa Fe, NM, 2019
Contact me for further details. Current programs include: