Chemistry-Informed Machine Learning for Polymer Electrolyte Discovery

Abstract

Solid polymer electrolytes (SPEs) have the potential to improve lithium-ion batteries by enhancing safety and enabling higher energy densities. However, SPEs suffer from significantly lower ionic conductivity than liquid and solid ceramic electrolytes, limiting their adoption in functional batteries. To facilitate more rapid discovery of high ionic conductivity SPEs, we developed a chemistry-informed machine learning model that accurately predicts ionic conductivity of SPEs. The model was trained on SPE ionic conductivity data from hundreds of experimental publications. Our chemistry-informed model encodes the Arrhenius equation, which describes temperature activated processes, into the readout layer of a state-of-the-art message passing neural network and has significantly improved accuracy over models that do not encode temperature dependence. Chemically informed readout layers are compatible with deep learning for other property prediction tasks and are especially useful where limited training data are available. Using the trained model, ionic conductivity values were predicted for several thousand candidate SPE formulations, allowing us to identify promising candidate SPEs. We also generated predictions for several different anions in poly(ethylene oxide) and poly(trimethylene carbonate), demonstrating the utility of our model in identifying descriptors for SPE ionic conductivity.

Machine Learning-Enabled Environmentally Adaptable Skin-Electronic Sensor for Human Gesture Recognition

“…According to the allowable limits of the equipment and thermogravimetric analysis (TGA) in Figure S10a, the ionic conductivity is investigated at a temperature range of 20 to 80 °C to ensure no decomposition reaction. The change in ionic conductivity is fitted by the Arrhenius equation: σ = A exp true[ prefix− E a italicRT true] there are the constants related to carrier concentration ( A ), activation energy ( E a ), and the ideal gas constant ( R ) on overall conductivity (σ) at a given temperature ( T ) . In Figure b, the ion gel with 10 wt % ionic liquid exhibits higher conductivity sensitivities to temperature changes compared with 20 and 30 wt % because of the high activation energy of the ion gel at low ionic liquid.…”

“…there are the constants related to carrier concentration (A), activation energy (E a ), and the ideal gas constant (R) on overall conductivity (σ) at a given temperature (T). 48 In Figure 5b, the ion gel with 10 wt % ionic liquid exhibits higher conductivity sensitivities to temperature changes compared with 20 and 30 wt % because of the high activation energy of the ion gel at low ionic liquid. According to Eq 4, the activation energies of EMIM-TFSI 10, 20, and 30 wt % are about 17.7, 6.02, and 5.66 kJ/mol, respectively.…”

Computational and data-driven modelling of solid polymer electrolytes

“…Bradford et al built a chemistry-informed ML model that could predict SPE ionic conductivity based on the electrolyte and composition. 145 They gathered data set of SPE ionic conductivity values from 217 experimental publications. They adopted a message passing NN, which is a special type of GNN, to learn optimal representations of the molecular components.…”

Prediction and Interpretability of Glass Transition Temperature of Homopolymers by Data-Augmented Graph Convolutional Neural Networks

“…As an alternative to the experimental approaches, data-driven machine learning (ML) methods have been extensively used to predict distinct properties of chemical compounds, ranging from heterogeneous catalysts to pharmaceutical molecules. − In comparison with the high computational complexity and the cost of molecular modeling, the ML methods have the capability to achieve a similar accuracy but less computational cost in the establishment of structure–property relationship. Recently, with the rapid advancements of polymer informatics, the ML methods are applied to acquire the complicated structure–property relationship of polymers. − As a typical example, integration of descriptor- or string-based representations (like Simplified Molecular-Input Line-Entry System, abbreviated as SMILES) with the ML methods (especially deep neural networks), is widely used for establishing the relationship between the monomers and T g of polymers. − In particular, by virtue of the one-hot encoding method, Miccio and Schhwartz encoded the SMILES-based structures of polymers into binary images, and then used the convolutional neural network to forecast the T g values of polymer molecules . The SMILES representation is further extended to text-based BigSMILES representation, , which is more suitable to describe nonperiodic and stochastic characteristics of polymer molecules.…”