Stuart Group Seminar

This talk explores the predictive computational modeling of a mesoscale phenomenon called the self-assembly of diblock copolymer (Di-BCP) thin films. Di-BCPs are polymers consisting of two blocks of distinct monomers. Upon thermal or solvent annealing, the two blocks in a Di-BCP melt spontaneously segregate due to their incompatibility. Periodically-ordered structures emerge as the melt achieves equilibrium. This phenomenon has attracted attention for its potential applications in the cost-effective fabrication of nano-scale devices aided by computer simulations. This talk presents studies that aim to systematically enhance the quality and reliability of these simulations via bridging theory and data.

The first part of this talk is dedicated to formulating the self-consistent field theory (SCFT) of Di-BCP self-assembly from the first principles of statistical mechanics. SCFT is a continuum model posed as a PDE-constrained optimization problem. The connections between SCFT and fundamental concepts in probability theory are highlighted. The second part of this talk is dedicated to the Bayesian inference of model parameters from experimental X-ray scattering or microscopy image data when considering uncertainties in data acquisition and material behavior. We propose to use likelihood-free inference approaches based on the pseudo-marginal method and measure transport together with the construction of summary statistics for image data. Combining the two parts, we briefly discuss the opportunities in deploying predictive models of Di-BCP self-assembly in optimal design and control problems for defect elimination in nanofabrication.