(365g) Mesoscale Modeling of Complex Fluids in Chemical and Biological Applications

Research Interests: molecular simulation, colloid, phase transition, electrokinetics, complex fluids, high-performance computing, machine learning

Complex fluids are mixtures of two coexisting phases, and they are ubiquitous in natural and engineered materials. Examples include biological fluids such as blood and cytoplasm, industrial coatings such as paint, fluids in food industry such as milk and salad dressing, and cosmetics such as shaving foam. Complex fluids are useful precisely because they exhibit both liquid- and solid-like behaviors, due to their internal microstructure formed by the presence of a second, non-continuum phase dispersed in a suspending continuum phase. However, such multi-phase systems pose a challenge in modeling them and understanding their material properties and transport: the vast separation of time- and length-scale between the coexisting phases. Computational methods to study complex fluids have been split into continuum methods and atomistic methods. On one hand, continuum methods perform computation efficiently, but can miss important mesoscale effects such as microstructural rearrangements and electric double layers. On the other hand, atomistic methods such as molecular dynamics resolve molecular details, but currently it is computationally infeasible to scale up to industrial applications. To overcome these limitations on both types of methods, my research interest focuses on developing mesoscale simulation methods for complex fluids to facilitate better chemical and biological process design. Following this theme, I have research experiences developing mesoscale computational techniques for the following chemical and biological applications.

(1) Modeling colloidal glass transition using Brownian dynamics and characterizing its physical aging behavior. Upon rapid cooling, many molecular fluids undergo glass transition where liquid-like structure is frozen into an amorphous solid. Colloidal suspensions exhibit similar behavior upon a fast increase in the concentration of colloidal particles. In collaboration with three experimentalists from Texas Tech University who use polymethylmethacrylate (PMMA) to trigger colloidal glass transition, I developed a similar “concentration quench” protocol in silico using Brownian dynamics. We investigated the aging behavior of colloidal glasses by conducting a time-concentration superposition (in analogous to a time-temperature superposition in molecular systems). Moreover, my simulations show microstructural rearrangements through particle tracking, Voronoi analysis and scattering function, suggesting a non-diverging relaxation time in colloidal glass transition and that osmotic pressure is a driving force for physical aging behavior.

(2) Modeling electrokinetic flows using an efficient hybrid method between fluctuating hydrodynamics and immersed boundary method. Electrokinetic flows are important in many applications such as microfluidic pumping and desalination. Electrokinetic effect is inherently related to the development of the electric double layers due to fluctuations of ions in an electrolyte solution, where molecular-scale physics that cannot be captured by a continuum model are still important. While molecular dynamics can also model the double layer, it is computationally expensive and is challenging to scale the problem size to real applications. To overcome these challenges, I developed simulations of induced-charge electroosmosis (ICEO) over a metallic plate to study non-linear electrokinetic effects, using the recently developed Discrete-Ion Stochastic Continuum Overdamped Solvent (DISCOS) algorithm at Lawrence Berkeley National Laboratory, with collaboration from a diverse team of physicists, mathematicians and software engineers. This algorithm leverages AMReX framework (supported by the Department of Energy (DOE)’s Exascale Computing Project (ECP)), where solvent is modeled under the grid-based, fluctuating hydrodynamics framework, and the ions are treated by the particle-based, immersed boundary method. I show that the steric effects play a crucial role at strong electric fields where electric double layer is thin, which leads to several intriguing phenomena such as non-local distribution of ions and overcharging of co-ions to the surface charge.

(3) Computational modeling of the E. coli ribosome L12 complex and its impacts on mRNA translation rate. Protein synthesis is essential to life and as such, requires extensive cell machinery, which is done via decoding (translating) messenger ribonucleic acid (mRNA) within the ribosome. Previous studies on translation have focused primarily on two disparate limits: structural biology that resolves atomistic details of tRNA and ribosome but is limited to translation on a single ribosome, and system biology that resolves kinetic rates of the well-known intra-ribosome translation pathway but abstracts away spatial resolution. A recent work showed that colloidal-scale physical transport in crowded cytoplasm plays a crucial role in limiting translation rate, in addition to chemical kinetics. My current research focuses on improving the detailed representation of the E. coli cytoplasm’s molecules, to predict absolute protein synthesis rates. I start with one of the constituents of the E. coli cytoplasm, ribosome, which is the key bio-factory to translation process. I develop a coarse-grained framework to model the ribosomal L10 N-terminus domain (NTD), ribosomal L12 stalk, hinges, and ribosomal L12 C-terminus domain (CTD) as a set of flexible arms endowed with the shape and charge determined from the Protein Data Bank (PDB) and experimental reports of L12 mobility. The long, tether-like L12 arms permit a larger search volume for translation molecules to bind to ribosomes, which can impact translation rate. Using this model, I probe the effects of cell growth-rate-mediated cellular stoichiometry, diffusive fluctuations of the L12 subunit, and combinatoric sampling efficiency on translation rate.

With my experiences and expertise in mesoscale modeling of various complex fluids, for the next step I am interested in leveraging machine learning to develop multi-level coarse-graining simulation techniques based on existing experimental data (such as PDB) and/or molecular structure predictions (such as AlphaFold) to scale up complex fluid modeling. The outcome of this work can potentially advance modern material engineering and therapeutic design.