Discovery of Polymer Electrolytes with Bayesian Optimization and High-Throughput Molecular Dynamics simulations

Polymer electrolytes are critical for safe, high-energy-density solid-state batteries, yet discovering candidates that balance high ionic conductivity with high transference numbers remains a significant challenge. In this work, we develop a high-throughput screening platform that utilizes molecular dynamics simulations to navigate a chemical space of 1.7 million hypothetical polymer electrolyte candidates. Data from previous literature is used to warm-start batch Bayesian optimization for iteratively selecting new polymer electrolytes to evaluate. We iteratively identified, evaluated and analyzed 767 homopolymers as potential candidates. Our results reveal several candidates with transport properties exceeding the benchmark polyethylene oxide (PEO)/LiTFSI system. Crucially, our optimization campaigns for ionic conductivity and Li-diffusivity demonstrate that branched architectures and ketone functional groups significantly enhance ion-hopping mechanisms within the polymer matrix. We provide an in-depth mechanistic comparison of Li vs. Na cation transport and offer our open-source framework to accelerate the discovery of liquid, gel, and multi-cation electrolyte systems.

💡 Research Summary

This paper presents an integrated high‑throughput workflow that couples large‑scale molecular dynamics (MD) simulations with Bayesian optimization (BO) to discover polymer electrolytes for solid‑state batteries. Starting from 1.66 million synthetically accessible homopolymers drawn from the Open Macromolecule Genome and SMiPoly databases, each candidate is encoded as a SMILES string and transformed into a 768‑dimensional MolFormer embedding. Principal component analysis reduces these embeddings to 50 dimensions, which serve as inputs to a Gaussian Process (GP) surrogate model. The GP is warm‑started with 106 experimental ionic conductivity values from the literature, providing a strong prior for early iterations.

Batch BO proceeds by clustering the chemical space into 20 groups via k‑means++ and selecting, from each cluster, the polymer with the highest Expected Improvement (EI) acquisition value. Twenty polymers per batch are then simulated with the HiTPoly pipeline: OPLS‑AA force‑field parameters (generated by LigParGen), a charge scaling factor of 0.75, and a temperature offset of +50 K (393 K) to align MD glass‑transition behavior with experiments. Each MD run lasts 100 ns, and ion transport properties are extracted using the cluster Nernst‑Einstein (cNE) method, yielding ionic conductivity, transference number, fraction of free charge carriers, and self‑diffusivities of Li⁺ and the constitutional repeat unit (CRU).

Four optimization objectives are explored across separate campaigns: (1) maximize ionic conductivity, (2) maximize the product of free‑carrier fraction and transference number, (3) maximize ionic conductivity multiplied by transference number, and (4) maximize the product of Li⁺ diffusivity/CRU diffusivity ratio and ionic conductivity. In total, 30 batches (20 polymers each) are evaluated, amounting to 767 MD simulations. Warm‑started BO outperforms a cold‑start GP and random search, achieving higher conductivities within far fewer iterations. Visualization of the convex hull in the reduced feature space shows progressive expansion, indicating balanced exploration and exploitation.

Key chemical insights emerge: branched polymer backbones and ketone functional groups markedly enhance ion‑hopping pathways, leading to conductivities that surpass the benchmark PEO/LiTFSI system. Ten top candidates with superior performance and diverse backbone chemistries are identified. Comparative analysis of Li⁺ versus Na⁺ transport reveals that while Na⁺ exhibits a higher fraction of free carriers, Li⁺ maintains a more favorable combination of conductivity and transference number for battery applications.

All data, the MolFormer embeddings, the GP surrogate, and the HiTPoly simulation scripts are released as open‑source resources, enabling rapid extension to liquid, gel, and multi‑cation electrolyte designs. The study demonstrates that coupling machine‑learning‑driven active learning with accurate MD simulations can efficiently navigate massive polymer chemical spaces and accelerate the discovery of high‑performance solid‑state electrolyte materials.