“Theory, simulations, and machine learning for design and structural characterization of macromolecular materials ”
The Jayaraman research group members’ expertise lies in development of coarse-grained models and computational approaches involving liquid state theory, molecular simulation, and machine learning for designing and characterizing macromolecular materials.
In the first part of my talk, I will talk about recent work involving Polymer Reference Interaction Site Model (PRISM) theory and coarse-grained molecular dynamics simulations to predict polymer blend morphology (i.e., macrophase separated, disordered with concentration fluctuations, microphase separated) as a function of placement and fraction of associating groups along polymer chains at varying strengths of association. The features in structure factors [S(k) vs. k] for varying polymer design and association strengths are used to identify the morphologies within the phase diagram. For the disordered morphologies that exhibit concentration fluctuations, we calculate how the length scales of concentration fluctuations change with the associating group placement for similar fraction of association groups. Using this combination of PRISM theory and molecular simulations we are able to explore a large polymer design space with reduced computational intensity and more reliable structure factors than would be possible with an approach involving only molecular simulations.
In the second part of my talk, I will introduce a method ‘Computational Reverse Engineering of Scattering Experiments (CREASE)’ that we have developed for analysis of small angle scattering profiles and interpretation of the assembled structure in macromolecular solutions. CREASE is comprised of two steps: the first step involves a genetic algorithm (GA) to determine the shape and dimensions of the domains in the assembled structure and the second step uses molecular simulations to reconstruct chain conformations and monomer level arrangements within the assembled structure. We validate the GA step within CREASE by taking input scattering intensity profiles from a variety of assembled shapes with known shapes and dimensions, and by producing outputs that match those known shapes and target dimensions. Besides applying CREASE to scattering data where the shape is known, I will demonstrate how CREASE would be applied to experimental scattering profiles and realistic situations where microscopy may hint at the potential shapes without the user knowing the dimensions with certainty, to test hypotheses on the shape and calculate the resulting dimensions for that shape. CREASE’s power lies in its ability to interpret structural detail at a range of length scales for macromolecular solutions without relying on fitting with off-the-shelf analytical models that may be too approximate for novel polymers and/or unconventional assembled structures.
