Design of Polymer Blend Electrolytes through a Machine Learning Approach

Abstract

We apply a machine learning (ML) technique to the multiobjective design of polymer blend electrolytes. In particular, we are interested in maximizing electrolyte performance measured by a combination of ionic transport (measured by ionic conductivity) and electrolyte mechanical properties (measured by viscosity) in a coarse-grained molecular dynamics framework. Recognizing the expense of evaluating each of these properties, we identify that the anionic mean-squared displacement and polymer relaxation time can serve as their proxies. By employing the ML approach known as Bayesian optimization, we identify a trade-off between ion transport and electrolyte mechanical properties as a function of varied design parameters, which include host molecular weight and polarity. Our results suggest that blend electrolytes whose hosts have unequal molecular weights, such as gel polymer electrolytes, rarely maximize electrolyte performance. Overall, our results suggest the potential of a framework to design highperformance electrolytes using a combination of molecular simulation and ML.

Miscible Polyether/Poly(ether–acetal) Electrolyte Blends

“…Ternary blends with polymer components of different polarities have been discussed as a means for improving ion transport in a recent theoretical paper by using a coarse-grained bead-spring model. 10 In this work, we demonstrate that PEO is miscible with P(2EO-MO) in the neat, salt-free, state. We have also identified a range of salt concentrations over which PEO/P(2EO-MO)/ LiTFSI blends remain miscible.…”

“…I(q) = aP (q) + b (10) where P(q) is a form factor given by the Debye function (see eq 14), a is a constant scaling factor, and b is a constant assumed to be equal to I inc (q). 14,15,25 Figure 4 shows the coherent SANS profiles, I coh (q), of the miscible blends at 90 °C.…”

Bayesian optimization for goal-oriented multi-objective inverse material design

“…In this paper, this inverse design approach is called “few-solution-inverse-design”. Most few-solution-inverse-design solutions use scalarization functions (SFs), which convert multiple objective properties to a single score, which, in turn, can be optimized using any single-objective optimization methods ( Cummins and Bell, 2016 ; Wang et al., 2020a ; Wheatle et al., 2020 ; Yamawaki et al., 2018 ). Among them, weight summation of multiple objective properties is most popular one.…”

“…In the weight-summation SF, predefined weights for weight summation determine the balance of objective properties in the Pareto optimal solution. Using this weight-summation SF, several studies have successfully performed inverse material designs using machine-learning-based optimization ( Cummins and Bell, 2016 ; Wang et al., 2020a ; Wheatle et al., 2020 ). However, because, the predefined weight is not directly reflected in the optimized solution, it is difficult to balance objective properties using weight-summation SF exactly as required.…”

Discovery of Innovative Polymers for Next-Generation Gas-Separation Membranes using Interpretable Machine Learning

“…Machine learning (ML) is a promising data-centric approach for prediction of gas permeabilities by learning a functional model based on polymer chemistry 40,41 . ML methods using chemical inputs have been successfully applied to accurately predicting many polymer properties including glass transition temperature [42][43][44] , thermal conductivity 45 , dielectric constants 46 , organic photovoltaic properties 47,48 , and transport properties [49][50][51] . The primary challenge for learning a generalizable ML model is training on robust and diverse data, which requires compiling multiple databases with the most recent literature values and imputing missing values 40 .…”