Title: Methodological Advances in Multiscale Simulation of Surfactant Self-Assembly

 

Surfactant formulations can form micelles, liquid-crystalline mesophases, and coexisting mesostructured phases whose stability depends sensitively on molecular chemistry, composition, temperature, and additives. Predicting this phase behavior remains challenging because the interactions originating at the atomistic length scale determine the structures and thermodynamics that emerge over mesoscopic length scales. Molecular simulations have the potential to aid in surfactant formulation design as a predictive tool for self assembly and phase behavior. With particle-based simulations, all-atom models are the most chemically detailed and predictive but are generally too computationally expensive for routine phase behavior prediction, whereas coarse-grained models often rely on experimental data, limiting the scope of their applicability in the chemical design space. This dissertation develops a multiscale route for bridging chemical information from all-atom simulations with coarse-grained statistical field theories that are efficient for determining surfactant self-assembly behavior and thermodynamics. 

The approach begins by using relative entropy minimization to parameterize coarse-grained models from all-atom reference simulations. These particle models are then transformed into equivalent field-theoretic representations and sampled using self-consistent field theory. Applied to aqueous alkyl quaternary ammonium surfactants, this multiscale framework predicts the experimentally observed micellar and liquid-crystalline phases with the correct ordering with composition and self-assembly trends associated with temperature, surfactant tail length, and salt concentration without any experimental input. The same model also provides access to equilibrium quantities such as domain spacing, micelle aggregation number, and critical micelle concentration under a unified free energy-based framework. 

A key methodological development is introduced to enable coarse-grained model parameters learned from strongly phase separated reference states to remain accurate under other thermodynamic conditions. This method applies external potentials in the all-atom simulations to promote composition fluctuations and sample local environments that would otherwise be rare. A particle fluctuation number is derived to measure the thermodynamic information contained in these reference ensembles and to guide the selection of external potential strengths. This strategy improves the coarse-grained model predictions of the transfer free energy for water-alkane systems and substantially improves the predicted critical micelle concentration for sodium dodecylsulfate when applied in the multiscale framework. 

These developments are then combined to study novel, multi-component trisiloxane surfactant formulations containing water, salt, and silicone oil. Common tangent and Gibbs ensemble calculations are used to determine complex types of binary and ternary component phase coexistence and component partitioning under the effects of temperature and silicone oil molecular weight. This extension of the multiscale framework enables detailed construction of ternary phase diagrams consisting of precise boundaries between competing types of phase coexistence for the trisiloxane system. Taken together, this work connects all-atom simulation, transferable coarse-grained model development, and efficient field-theoretic free energy calculations into a practical framework for de novo prediction of surfactant phase behavior from molecular level information.