Machine Learning Enhanced Computational Reverse Engineering Analysis for Scattering Experiments (CREASE) to Determine Structures in Amphiphilic Polymer Solutions.

In this article, we present a machine learning enhancement for our recently developed "Computational Reverse Engineering Analysis for Scattering Experiments" (CREASE) method to accelerate analysis of results from small angle scattering (SAS) experiments on polymer materials. We demonstrate this novel artificial neural network (NN) enhanced CREASE approach for analyzing scattering results from amphiphilic polymer solutions that can be easily extended and applied for scattering experiments on other polymer and soft matter systems. We had originally developed CREASE to analyze SAS results [i.e., intensity profiles, I(q) vs q] of amphiphilic polymer solutions exhibiting unconventional assembled structures and/or novel polymer chemistries for which traditional fits using off-the-shelf analytical models would be too approximate/inapplicable. In this paper, we demonstrate that the NN-enhancement to the genetic algorithm (GA) step in the CREASE approach improves the speed and, in some cases, the accuracy of the GA step in determining the dimensions of the complex assembled structures for a given experimental scattering profile.

Self-assembly of amphiphilic polymers of varying architectures near attractive surfaces.

We use coarse-grained molecular dynamics simulations to investigate the assembly of A-B amphiphilic polymers near/on surfaces as a function of polymer architecture and surface attraction to the solvophobic B-block in the polymer. We study four polymer architectures: linear, bottlebrush with a backbone that is longer than each of the side chains, bottlebrush where the solvophobic backbone is similar in length to each of the side chains, and 'star-like' architectures where the backbone is significantly shorter than the side chain lengths. For each architecture and surface-B attraction, we quantify the assembled aggregate structure (i.e., aggregation number, domain shapes and sizes) and the chain conformations (i.e., components of the chain radius of gyration) on and away from the surface. For all the architectures and surface-B attraction strengths, the assembled structure away from the surface is similar to the assembly observed in bulk systems without surfaces. Near/on the surface, the assembled B-blocks form domains whose shapes and sizes are dependent on the surface-B attraction strength and the ability of the B-block in the polymer architecture to change conformations and pack on the surface. Domain sizes formed from linear and 'star-like' polymer architectures show the highest sensitivity to surface-B-block attraction strength, transitioning from hemispherical to disordered domains with increasing attraction strength. In contrast, bottlebrushes with long backbones and short side chains transition from hemispherical to striped to continuous domains with increasing surface-B attraction strength. Bottlebrushes with similar solvophobic backbone and side chain lengths form hemispherical domains that do not change significantly with the surface-B-block attraction strength. These computational results can guide experimentalists in their choices of surface chemistry and polymer architecture to achieve desired assembled domain shapes and sizes on the surface.

Computational Reverse-Engineering Analysis for Scattering Experiments on Amphiphilic Block Polymer Solutions.

In this paper, we present a computational reverse-engineering analysis for scattering experiments (CREASE) based on genetic algorithms and molecular simulation to analyze the structure within self-assembled amphiphilic polymer solutions. For a given input comprised of scattering intensity profiles and information about the amphiphilic polymers in solution, CREASE outputs the structure of the self-assembled micelles (e.g., core and corona diameters, aggregation number) as well as the conformations of the amphiphilic polymer chains in the micelle (e.g., blocks' radii of gyration, chain radii of gyration, monomer concentration profiles). First, we demonstrate CREASE's ability to reverse-engineer self-assembled nanostructures for scattering profiles obtained from molecular simulations (or in silico experiments) of generic coarse-grained bead-spring polymer chains in an implicit solvent. We then present CREASE's outputs for scattering profiles obtained from small-angle neutron scattering (SANS) experiments of poly(d-glucose carbonate) block copolymers in solution that exhibit assembly into spherical nanoparticles. The success of this method is demonstrated by its ability to replicate, quantitatively, the results from in silico experiments and by the agreement in micelle core and corona sizes obtained from microscopy of the in vitro solutions. The primary strength of CREASE is its ability to analyze scattering profiles without an off-the-shelf scattering model and the ability to provide chain and monomer level structural information that is otherwise difficult to obtain from scattering and microscopy alone.

Molecular dynamics simulation study of linear, bottlebrush, and star-like amphiphilic block polymer assembly in solution.

In this study we investigate the effect of varying branched polymer architectures on the assembly of amphiphilic block polymers in solution using coarse-grained molecular dynamics simulations. We quantify assembly structure (e.g., aggregation number, assembly morphology, and micelle core size) and thermodynamics (e.g., unimer to micelle transition conditions) as a function of increasing solvophobicity of the solvophobic block in the copolymer for three broad categories of polymer architectures: linear, 'bottlebrush' (with many short side chains on a long backbone), and 'star-like' (with few long side chains on a short backbone). Keeping the total number of coarse-grained beads in each polymer (or polymer molecular weight) constant, as we go from either linear or 'star-like' to 'bottlebrush' polymer architectures, the micelle aggregation number and micelle core size decrease, and the solvophobicity required for assembly (i.e., transition solvophobicity) increases. This trend is linked to the topological/steric hinderance for making solvophobic bead contacts between neighboring polymers for the 'bottlebrush' polymer architecture compared to the linear or 'star-like' architectures. We are able to identify some universal trends in assembly by plotting the assembly structure and thermodynamics data as a function of branching parameter defined as the ratio of the branched chain to the linear chain radius of gyration in the unimer state, and the relative lengths of the backbone versus side chain. The results in this paper guide how one could manipulate the amphiphilic block polymer assembly structure and thermodynamics by choosing appropriate polymer architecture, block sequence, and composition.

Experiments and Simulations of Complex Sugar-Based Coil-Brush Block Polymer Nanoassemblies in Aqueous Solution.

In this work, we investigated the fundamental molecular parameters that guide the supramolecular assembly of glucose-based amphiphilic coil-brush block polymers in aqueous solution and elucidated architecture-morphology relationships through experimental and simulation tools. Well-defined coil-brush polymers were synthesized through ring-opening polymerizations (ROP) of glucose carbonates to afford norbornenyl-functionalized poly(glucose carbonate) (NB-PGC) macromonomers, followed by sequential ring-opening metathesis polymerizations (ROMP) of norbornene N-hydroxysuccinimidyl (NHS) esters and the NB-PGC macromonomers. Variation of the macromonomer length and grafting through ROMP conditions allowed for a series of coil-brush polymers to be synthesized with differences in the brush and coil dimensions, independently, where the side chain graft length and brush backbone were used to tune the brush, and the coil block length was used to vary the coil. Hydrolysis of the NHS moieties gave the amphiphilic coil-brush polymers, where the hydrophilic-hydrophobic ratios were dependent on the brush and coil relative dimensions. Experimental assembly in solution was studied and found to yield a variety of structurally dependent nanostructures. Simulations were conducted on the solution assembly of coil-brush polymers, where the polymers were represented by a coarse-grained model and the solvent was represented implicitly. There is qualitative agreement in the phase diagrams obtained from simulations and experiments, in terms of the morphologies of the assembled nanoscopic structures achieved as a function of coil-brush design parameters ( e.g., brush and coil lengths, composition). The simulations further showed the chain conformations adopted by the coil-brush polymers and the packing within these assembled nanoscopic structures. This work enables the predictive design of nanostructures from this glucose-based coil-brush polymer platform while providing a fundamental understanding of interactions within solution assembly of complex polymer building blocks.