(2ep) Multiscale Modeling and Engineering of Low-Dimensional Material Interfaces

Design and optimization of novel materials have been crucial to counter the world’s challenge of renewable energy. This requires accurate computational prediction of the structure-property relationship and building efficient models to understand the multiscale physical phenomena at the material interface. My research career has been following the path of combining the multiscale understanding of interfacial phenomena to guide experimental design.

Starting in experimental polymer chemistry, my initial research projects involved combining rational molecular design, polymer physics, and mass transport theories to fabricate reactive polymer photonic crystals as ultrasensitive sensors. Based on these experiences, I further pursued my PhD under the supervision of Prof. Chih-Jen Shih (ETH Zürich, Switzerland) with a focus on the systematic understanding and engineering of multiscale phenomena on atomically-thin two-dimensional (2D) materials. These studies provide insight into how the intrinsic electronic and dielectric properties of 2D materials influence their behavior in a 3D world. My first few projects studied the penetration of 2D materials to the electrostatic field, i.e. how “transparent” are they to the field effect. My multiscale models based on the concept of quantum capacity successfully captured the operational mechanism of graphene-based vertical transistors and explained the later-observed asymmetric field screening in 2D heterostructures. The quantum capacitor model was extended to demonstrate the multiscale effect on real-world 2D materials applications, including i) theoretically reveal- ing the doping-induced wetting characteristics of the water-2D material interface, ii) demonstrating the impact of 2D quantum capacitance on the selective ionic transport through gated graphene nanopores (experimental collaboration), and iii) fabricating ultra-sensitive pressure sensor based on field penetration and hydrophobicity at graphene-organic-liquid semiconductor interfaces (own experimental work). Beyond electrostatics, I also worked on the frequency-dependent dielectric behavior of 2D materials. Combining a simplified perturbation model and high-throughput DFT calculations, I identified the universal scaling relation between the electronic polarizability tensor of 2D materials and their electronic bandgap and geometric parameters, providing a general rule for searching 2D materials with desired dielectric properties at almost no computational cost. By coupling the electronic polarizability scaling with the Lifshitz dispersion theory, I systematically studied how the 2D layers screen the van der Waals (vdW) forces by up to 5x10⁵ material combinations. The model further predicted the existence of repulsive vdW interactions at the 2D-bulk solid interface, which was further validated by direct force measurement and interfacial epitaxy on graphene surfaces.

My current postdoctoral research (supervisor: Prof. Zachary Ulissi, Carnegie Mellon University) focuses on using machine learning (ML) models to accelerate the high-throughput discovery of novel heterogeneous catalyst materials. In particular, I’m working to build high-accuracy ML models incorporating dispersion and long-range interactions for predicting catalyst surface energetics. My works show that the transfer learning technique serves as a general approach to building neural network models based on limited quantum chemistry data while maintaining similar prediction precision compared with the state-of-art models. When combined with active learning algorithms, the transferred model can be further used to accelerate data collection done with computationally expensive quantum chemistry methods. These research projects further extend my theoretical background and are extremely helpful in accomplishing my future research goals.

My research plan for the faculty career will be focused on the machine-learning-assisted multiscale simulation of functional energy materials and heterogeneous catalysts. As an extension to my previous research projects, a few of the short-term research directions include:

1. Building efficient machine-learning force fields for simulating interfacial molecular packing at different length scales and designing high-performance organic electronics

2. Use generative machine learning models for inverse design of high-entropy nanomaterials for optoelectronic applications that break the scaling relation

3. Developing computational pipelines for adaptive parameter passing within multiscale models to facilitate the design of electrodes for battery applications and CO₂reduction reactions.

My research group will focus on not only developing these theoretical methods but to make them transferrable to the computational community and collaborating experimental groups. Close collaboration with experimental groups will be necessary for iteratively revising the models and looking for potential new physics.