CAREER: Multiscale Simulation and Machine Learning for Smart Polymer Design

NONTECHNICAL SUMMARY

Polymers are very long chain-like molecules that consist of repeating subunits. Depending on their underlying chemistry and architecture, certain polymers are known to be “stimuli-responsive” in the sense that they can drastically alter their characteristics based on environmental conditions. Consequently, such polymers could be used to create “smart,” adaptive materials that alter function in response to triggers like temperature, acidity, and stress.

However, while it is innately known that the nature and extent of stimuli-response depends on chemical structure, predicting whether any given polymer is suitable for a given application is elusive. This CAREER award focuses on developing predictive tools to facilitate the understanding and design of stimuli-responsive polymers. In particular, the research team aims to improve upon the accuracy of current molecular modeling strategies by implementing “environment-aware” simulation algorithms.

Furthermore, the research team will harness the power of machine learning to guide structure formation of polymers based on stimuli-response. The knowledge and methods generated via these activities will set the stage for future campaigns in polymer design across diverse applications, such as smart sensing, diagnostics, drug-delivery, coatings, clothing, and purification.

The major goals and methods of the research, which derive from the power of modern computation, will also supplement numerous education and training activities. Research activities will be coupled to “Princeton’s Laboratory Learning Program” for high school students with targeted outreach efforts to catalyze interest in using computation for engineering, across youth and by underrepresented groups.

Moreover, the principal investigator will enhance and/or develop two engineering electives predicated on machine learning and materials design, emphasizing domain-relevant examples and applications to accelerate understanding and utilization. In the same vein, the team will also develop “handbook”-style guides that illuminate common pitfalls and best practices for molecular modeling that are ubiquitously encountered but seldom formally taught.

All developed educational products will be made publicly accessible to extend the reach and utility of these materials. In the long-term, these efforts will snowball into more expansive projects that more firmly integrate molecular modeling and machine learning into traditional engineering curricula and prepare graduates to meet the ever-increasing demands of the technical workforce.

TECHNICAL SUMMARY

Research for this CAREER award will substantially advance capabilities to design “smart” functional materials based on stimuli-responsive polymers. Stimuli-responsive polymers are macromolecules that adapt their functionality in response to exposure to certain triggers/stimuli and can thus be exploited for numerous applications, such as sensing, robotics, drug-delivery, and separations.

The prospect of tailoring the chemistry and architecture of a stimuli-responsive polymer to elicit a specific, desired functional response is highly enticing; however, there are no existing robust, predictive frameworks to inform their design in a high-dimensional parameter space. The research team will resolve key technical bottlenecks that currently inhibit computationally guided design of stimuli-responsive polymers.

In particular, major projects include (i) multiscale modeling of thermo-sensitive polymers with expressive coarse-grained potential energy functions, (ii) modeling polymer dynamics in inhomogeneous environments within implicit-solvent frameworks, and (iii) leveraging machine learning to control emergent structural properties of polymeric materials. In aggregate, these activities will provide a foundation for modern computational techniques to be exploited during design of novel smart nanomaterials.

Educational and training activities as part of this CAREER award will address an urgent need to cultivate trainees for the next-generation workforce. Skills in molecular modeling and machine learning are increasingly relevant in both academic and industrial settings, yet these are rarely integrated directly into curricula for physical scientists and engineers.

Consequently, trainees often learn such skills outside of traditional pedagogical environments and in contexts that are divorced from their intended domain of application; this delays integration into and innovation by the workforce. The principal investigator will lead activities that bridge technical gaps, including (i) development of two engineering electives related to machine learning and materials design, (ii) creation of educational aids on practical considerations for molecular modeling, and (iii) expanded participation in training programs for high school students that highlight the visibility and utility of computation in engineering.

These activities will provide near-term enhancements in important technical training for young professionals and more firmly ingrain modeling/data science into future engineering curricula.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.