(2gd) Atomic and Molecular Design of Materials for Sustainable Energy Storage Solutions

Tackling modern engineering challenges means bridging length scales. In energy storage, understanding electrochemical performance of polycrystalline cathode materials involves unprecedented integration of first-principles computations and machine learning methods for large-scale simulations. Similarly, in liquids, modeling explicit interactions are essential for computational electrochemistry applications investigating ionization potentials, solvation shells, and ion transport. Here again, first-principles and machine learning approaches can be used together to enable large-scale molecular simulations.

My research bridges the length scale gap in solid state and liquid systems to enable sustainable energy storage solutions. I start from accurate first-principles approaches, obtained by reliable benchmarking, to investigate solids, liquids, and gases. In disordered systems with high chemical and configurational complexity, I also develop and apply statistical and machine learning models which efficiently sample phase space to evaluate thermodynamic and kinetic processes.

I completed my PhD at UC Berkeley with Professor Gerbrand Ceder modeling earth-abundant battery materials using density functional theory (DFT) and cluster expansion (CE) lattice models. We selected DFT methods after rationalizing the origin of reliable crystal structure prediction [1] and used them to understand phase stability in layered sodium-ion batteries (SIB) at high state-of-charge [2] and short-range order in ultrahigh power and energy dense lithium-ion battery (LIB) materials [3]. In the SIB work, we proposed why a specific fraction of earth-abundant iron in layered SIB cathodes increases phase stability at high state of charge, a key challenge in engineering SIB with long cycle life. In the LIB work, we developed new tools to construct the most complex CE lattice model to our knowledge [4] in order to explain the role of short range order in promoting high rate performance [3].

As an Environmental postdoctoral Fellow at Harvard University working with Professor Boris Kozinsky, I use machine learning methods to understand designer solvents, namely deep eutectic solvents (DESs), which are related to ionic liquids in that their physicochemical properties are tunable for a variety of applications [5]. We are primarily interested in type-III DESs since they are inexpensive, non-toxic, and reusable solvents for the recycling of battery materials. We are the first to benchmark DFT and computational chemistry approaches for these solvents and use them with machine-learned interatomic potentials in large-scale atomistic simulations to evaluate their utility as green solvents. Importantly, we work closely with experiments for validation.

In the future, my group will tackle urgent energy and sustainability challenges using computational and machine learning approaches. We will do deep dives to assess the processes in polycrystalline SIB and LIB undergoing chemical reactions, including topotactic delithiation, surface reactions, oxygen loss and densification, bringing the accuracy of reliable first-principles approaches into modeling explicit reaction processes far out of equilibrium. The group will also develop robust understandings of molecular solvents for metals recovery to enable a circular economy in energy storage. The goal of our interest in liquids is specifically for energy applications where immediate future projects are battery recycling and discovery of functional solvents. As a computational group, we will make it a priority to search for experimental collaborations spanning chemical engineering, materials science, and chemistry/electrochemistry groups. In summary, we will focus initially on three topics:

1. Electrochemical processes in polycrystalline cathode materials

2. Critical metals recovery from black mass using solvents

3. Modeling, simulation, and machine learning to accelerate the discovery of functional solvents

Electrochemical processes in polycrystalline cathode materials

Rechargeable ion batteries are desirable solutions for energy storage due to their design flexibility and high energy conversion efficiency. Materials based on the cubic close-packed anion lattice have higher energy density than less compact polyanion frameworks (e.g. LiFePO₄), but typically incorporate critical materials, namely cobalt or nickel due to the strong octahedral preference of Co³⁺ or Ni²⁺ which preserves the layered structure. For this reason, earth-abundant electrodes with the closed-packed cubic lattice are competitive alternatives to layered electrodes because they maintain high energy density but do not require Co or Ni [6]. While their pristine states are generally modelled as a single phase [2,3], during electro-chemical cycling, heterogeneous (de)intercalation processes become important. Our first research area focuses on the modeling of heterogenous nano-domains, where the domains themselves are metastable phases observed in first-principles calculations or CE Monte Carlo simulations. We will investigate the effect of heterogenous (de)intercalation of nano-domains on the 1) voltage profile and hysteresis due to different (de)lithiation pathways, 2) inter-domain Li mobility and effect on rate performance, moving beyond the single-phase assumption, and 3) influences of oxygen-redox on intra-domain phase stability and inter-domain chemo-mechanical effects that ultimately limit cycling stability. To investigate all these questions, we will rely on CE models for intra-domain modeling and machine learned molecular dynamics (MLMD) simulations for inter-domain interactions. We aim to identify thermodynamic and kinetic factors controlling electrochemical performance of polycrystalline cathodes.

Critical metals recovery from black mass using solvents

Recycling of critical materials in spent LIBs is an emerging challenge. In 2020, more than 500,000 tons of LIBs had reached end of life. This will grow to 11 million tons by 2030, more than 100 times the tonnage of spent batteries recycled in 2019. Unfortunately, recycling processes, namely pyrometallurgy and hydrometallurgy, have prohibitive challenges limiting their scale-up, major ones being handling changing chemistries (e.g. LiCoO₂, LiNi_1/3Mn_1/3Co_1/3O₂, LiFePO₄, LiMn₂O₄) and serious safety and environmental hazards. Direct recycling, a third route, requires laborious sorting and separation, single-cathode input steams, and reasonably-intact crystal structures. Furthermore, it has limited return-to-market viability. We will investigate a fourth route, ionometallurgy, which uses environmentally-benign organic solvents to electrodeposit high-quality films, owing to their effective solvation of transition metal compounds and wide electrochemical stability windows. We will use first-principles calculations, MLMD, and experiments to understand ionometallurgical processes. The future group will ideally have standard electrochemical analysis capabilities, which are used for baseline experimental validation. These experiments can be done by any interested student (e.g. upper-level high school, undergraduate, computational graduate students) after some training. However, for advanced characterization (e.g. NMR, ICP-OES, GC-MS), we will actively find collaborations. In summary, the aim of this computationally-guided experimental approach is to optimize for metal recovery for any chemistry found in battery waste.

Modeling, simulation, and machine learning to accelerate the discovery of functional solvents

Tunable organic solvents with binary or ternary molecular components, such as ionic liquids or deep eutectic solvents (DESs) have a vast design space. In principle, there are billions of known DES combinations and they have been used in a wide range of applications, e.g. oxaline for metals recovery, reline for methane capture, ethaline for redox flow batteries. Despite these many applications, the neat structure and descriptors that control technologically-important physicochemical properties such as viscosity, ionization potential, and thermal stability, are not characterized well from first principles. In this last approach, we will use computational chemistry, active learning, and MLMD simulations to study factors controlling physiochemical properties such as electrochemical stability, thermal stability, viscosity, and more. These approaches can be generalized to study a range of functional solvents for applications related to climate and sustainability.

Teaching Interests

Education of the next generation of students is more important than ever. We need to educate not only future research and development engineers, but also policy leaders, stakeholders, business people, and managers. Regardless of where engineering students choose to specialize, it is important they have the foundations in thermodynamics and materials science so they learn how society and innovation is limited by our materials and their availability. Having been an undergraduate and graduate-level instructor for thermodynamics at Carnegie Mellon and UC Berkeley, I would gladly teach this course so that students can gain a rigorous and intuitive understanding of this broad theoretical framework.

I also have a great interest in teaching a computational modeling course, where students learn about empirical potentials, density functional theory and computational chemistry, molecular dynamics, lattice models, and machine-learned interatomic potentials. Given the increased amount of compute and code accessibility in the last decade, along with the growing need for atomistic and molecular modeling, this course could be useful for chemists, chemical engineers, materials engineers, and mechanical engineers. I imagine extensive room for growth as certain areas, such as machine learned interatomic potentials, increase in scope and robustness for a variety of applications.

[1] J. H. Yang, D. A. Kitchaev, and G. Ceder, Phys. Rev. B, 100, 3, (2019).

[2] J. C. *. Kim, D.-H. *. Kwon, J. H. Yang*, H. Kim, S.-H. Bo, L. Wu, H. Kim, D.-H. Seo, T. Shi, J. Wang, Y. Zhu, and G. Ceder, Adv. Energy. Mater., 10, 31, (2020).

[3] J. H. Yang and G. Ceder, Adv. Energy. Mater., 13, 4, (2023).

[4] J. H. Yang, T. Chen, L. Barroso-Luque, Z. Jadidi, and G. Ceder, npj Comput. Mater., 8, 1, (2022).

[5] J. H. Yang, V. Gharakhanyan, T. Gadhiya, and A. Holiday, U.S. Patent App. 17/967,711, filed Oct. 17, 2022

[6] J. H. Yang, H. Kim, and G. Ceder, Molecules, 26, 11, (2021).