The broader fields that we work in are complex fluids and soft matter. Our mission is to address the grand challenges in modeling and simulating complex fluids and soft matter through accurate, robust, and scalable numerical methods as well as machine learning and data-driven model order reduction techniques. Our research currently spans the following areas.
Data-driven model order reduction:
Data-driven coarse-grained modeling of soft matter:
Soft matter, such as polymers and nanoparticles, can provide the structural and chemical plenitude and tunability to yield paradigm-shifting multifunctional materials for broad applications. A grand challenge in designing multifunctional materials made from soft matter is to link the emergent functionality to the underlying multiscale structures. To address this challenge, atomistic simulations are computationally prohibitive. Thus, coarse-grained (CG) modeling is particularly attractive, which projects atomistic details onto a coarser representation and is well suited to bridge multiscale structure-function relationships in materials.
- Preserve properties (dynamic, structural) under coarse-graining
“Implicit-solvent coarse-grained modeling for polymer solutions via Mori-Zwanzig formalism”, Soft Matter, 15, 7567-7582 (2019). Back Cover
“Data-driven coarse-grained modeling of polymers in solution with structural and dynamic properties conserved”, Soft Matter, 16, 8330-8344 (2020). Front Cover
- Transferable CG modeling enabled by transfer learning
“Transfer learning of memory kernels for transferable coarse-graining of polymer dynamics”, Soft Matter, 17, 5864-5877 (2021). Front Cover
- Coarse-graining of non-equilibrium systems
“Data-driven Coarse-grained Modeling of Non-equilibrium Systems”, Soft Matter, 17, 6404-6412 (2021).
Data-driven non-intrusive reduced-order modeling (ROM):
For large-scale applications and applications requiring multi-query loops (e.g., optimization and control), high-fidelity simulations that solve the full governing PDEs are too expensive. Therefore, the research in this area aims to establish accurate and efficient ROM for dynamical systems with freely moving boundaries/interfaces, e.g., in fluid-solid interactions or multiphase flows.
- Data-driven nonintrusive reduced order modeling for dynamical systems with moving boundaries using Gaussian process regression
High-fidelity simulations for fluid-solid interactions:
Fluid-solid interactions are ubiquitous in complex fluids. Computer simulations can enhance fundamental understanding and predict before experimental realization on how fluid and solids interact with each other and exhibit rich behaviors under different conditions and external forces. However, simulating fluid-solid interactions is challenging because of moving fluid–solid interfaces of arbitrary geometries, intractable cost to resolve point singularities that govern lubrication effects, and deteriorated convergence in the presence of singularities. Our research in this area aims to tackle these challenges.
- High-fidelity simulations enabled by high-order, spatially adaptive, meshless discretization methods
- Generalized moving least square (GMLS) with adaptive h-refinement and scalable multigrid preconditioner
- Consistent, spatially adaptive smoothed particle hydrodynamics (SPH); implicit and incompressible SPH
Computer Methods in Applied Mechanics and Engineering, 347, 402-424 (2019); Computer Methods in Applied Mechanics and Engineering, 324, 278-299 (2017); Journal of Computational Physics 334, 125–144 (2017)
Broader aspects of complex fluids in energy storage and thermal sciences:
By collaborating with experimentalists and domain scientists, we have also made contributions in the areas of energy storage and thermal sciences through numerical studies and machine learning-enhanced modeling. The efforts include, to give a few examples: optimizing discharge capacity of lithium-oxygen batteries by design of the air-electrode porous structure [Link]; modeling heat fluxes through boundaries for heat transfer in the presence of cracks or fractures [Link]; predicting electrokinetics in ionic liquids [Link]; and predicting device-scale gas-liquid multiphase flow and amine absorption for post-combustion CO2 capture [Link].
High-Performance Computing Tools:
- Coarse-grained simulation methods implemented in LAMMPS
- Parallel scalable in-house code for adaptive GMLS