(340bj) Investigation into Ion Transport Properties of Nanoparticle-Based Single-Ion Conducting Electrolytes Using Multiscale Simulations and Machine Learning

Funded by an NSF Materials Research Science and Engineering Center (MRSEC), my research aims at: (i) employing multiscale molecular simulations to investigate ion transport mechanisms in nanoparticle-based electrolyte materials to understand the experimentally observed conductivity characteristics for this relatively novel class of electrolyte materials, and (ii) employ machine learning techniques to guide design considerations (for our experimental collaborators) in the efforts to realize nanoparticle-based electrolytes with enhanced ionic conductivity and stimuli-responsive ion transport properties.

Electrolytes with attractive ionic conductivity, high lithium transference (single-ion conduction) and sufficient mechanical strength are desired for application in lithium-ion battery technologies. Nanoparticle-based electrolytes, in which nanoparticles are embedded in ion-conducting solid polymers or liquids, have recently emerged as a relatively novel yet promising class of electrolyte material for achieving high mechanical strength and single-ion conduction in next-generation lithium ion batteries. Furthermore, if there exists a strong dependence of ionic conductivity on the configuration of nanoparticles, such electrolytes with relatively low nanoparticle volume fraction (~0.1) could potentially serve as excellent candidates for stimuli-responsive battery electrolytes through reconfiguration of nanoparticle microstructure on applying an external stimuli.

Although nanoparticle-based electrolytes offer better safety characteristics than the conventional liquid electrolytes, the major bottleneck for the commercialization of these electrolytes seems to be the relatively modest ionic conductivities reported by handful of experimental studies. Further, no mechanistic understanding of transport mechanisms was suggested to interpret the experimental observations and identify key design parameters influencing ionic conductivity.

To understand and improve upon the moderate ionic conductivities observed in nanoparticle-based electrolytes, we employed multiscale simulations (sequential combination of coarse-grained Molecular Dynamics in LAMMPS software and mesoscale kinetic Monte Carlo simulations) to simulate the experimentally reported systems in the literature and identify the mechanisms underlying the experimentally observed moderate conductivities as well as identify the key design parameters influencing conductivity. Our results in this regard suggest that for the experimental systems reported thus far, the dominant pathway for cation transport is along the surface of nanoparticles, in the vicinity of nanoparticle-tethered anions. We discuss the impacts of nanoparticle concentration, cation and anion choice as well as solvent polarity with this mechanistic understanding of ion transport. On tuning the key design parameters and analyzing the previously unexplored design space, our results suggest that functionalizing nanoparticles with high dielectric constant polymers and dispersing them in high dielectric constant solvents promotes cation transport through the faster solvent medium and results in elevated ionic conductivity.

These results inspired our experimental collaborators to synthesize of single-ion conducting copolymer electrolytes with a wide range of key design parameters (identified from our previous results). We eventually intend to functionalize the nanoparticles with the optimal polymer chemistry and composition for effective solvation of cations. In this context, we are currently pursuing atomistic-level molecular dynamics simulations of the polymer electrolytes to complement understanding of their experimentally observed conductivity trends.

Since the dominant transport of cations was elucidated to be along the surface of the nanoparticles, we expected significant modulation of ionic conductivity on tuning the spatial arrangement of the nanoparticles, and thereby the connectivity of surface transport pathways. Explicitly, at a fixed nanoparticle loading, certain nanoparticle microstructures are expected to exhibit better conductivities than others. In this context, we also demonstrate that some morphologies can exhibit simultaneously enhanced ion transport and mechanical properties, thus illustrating the potential of nanoparticle-based electrolytes to decouple ion transport and mechanical strength generally observed in polymer electrolytes.

We also resorted to machine learning techniques for accelerated screening of the vast morphological search space. To overcome the computational cost issues accompanying the conventional multiscale simulation approach to probe the ionic conductivity, we developed a 3-D Convolutional Neural Network (CNN) model to quantitatively link the ionic conductivity to the spatial arrangement of the nanoparticles. Such an approach allows for faster yet fairly accurate prediction of the ionic conductivities. The model is trained using the “true” conductivity values computed for diverse nanoparticle configurations (at a fixed nanoparticle loading of 0.1 volume fraction) using multiscale simulations. By integrating the trained CNN models with a nanoparticle topology optimization algorithm, we demonstrated accelerated search of morphological space to identify nanoparticle networks exhibiting a wide range of ionic conductivities. Finally, by using data-driven approaches, we explore how simple descriptors of nanoparticle morphology correlate with cation diffusivity, thus providing a physical rationale for the observed optimal microstructures. This could in turn provide design guidelines to decide on potential design parameters for experimental realization of stimuli-responsive nanoparticle-based electrolytes.

Overall, my research aims to clarify the mechanistic underpinnings of ion transport in relatively novel nanoparticle-based electrolytes for lithium-ion battery applications, and guide design considerations for realizing improved performance. Our results provide quantitative support to the hypothesis that such electrolyte systems possess the potential to achieve high ionic conductivity along with high mechanical strength and desirable safety characteristics. Furthermore, we demonstrate the suitability of nanoparticle-based electrolytes to achieve stimuli-responsive ionic conductivity.

Currently, we are working on an invited review article titled “Machine learning for property design in materials” for publication in Annual Review of Chemical and Biomolecular Engineering journal with me as the first author. In the near future, using computer simulations we intend to provide design considerations to our experimental collaborators for self-assembly of polymer-grafted nanoparticles in the presence of thermoresponsive depletants to realize stimuli-responsive electrolytes.

In addition to above research projects, I also intend to highlight in the Industry Candidate poster my other computational skills (not highlighted above) and my summer research internship experience at Kraton Corporation. Explicitly, I am well versed with Gaussian software for DFT calculations and programming languages/tools such as Python (Numpy, Tensorflow, Pandas, Keras), Fortran, Bash, C++, MATLAB, Latex, OpenMP. During my time at Kraton as a research intern, I was responsible for compounding polymer blends (using internal mixer) and characterizing the blends using Electrochemical Impedance Spectroscopy, Differential Scanning Calorimetry, Dynamic Mechanical Analysis, Melt Flow Indexer, etc.