Data-driven and Simulation-assisted Synthesis of Hydrocarbon Polymer Electrolyte Membranes

Abstract

Polymer Electrolyte Membrane Fuel Cells (PEMFCs) are versatile energy devices that provide useable electrical energy for a wide range of stationary and automo tive applications. Polymer Electrolyte Membranes (PEMs) are solid ionomeric poly mer membranes that conduct proton (H+) from anode to cathode. These proton conducting electrolyte membranes are the most important component of the fuel cell as the rate of proton transport taking place in the water channels governs the over all performance of the PEMFC. Currently, the commercially available perfluorinated membranes have limitations at higher temperatures and low humidity operations as usually faced in automobile applications. Sulfonated Polyimides (SPIs) are a class of versatile hydrocarbon ionomeric polymers that are being explored as a polymer electrolyte material for fuel cells due to their superior thermal and mechanical stabil ity. Experimentation to discover alternative PEMs is extensive time-consuming and resource-intensive. There often occurs a loss of effort as the proton conductivity of the developed PEM is not able to be at par with the perfluorinated PEMs. Also, there is a lack of investigations into the correlation of the nano-scale mor phology of the PEM with the behavior of proton transport in SPI PEMs. Previously, researchers have gained some understanding of structure-property interplay through multi-scale computational models and extensive experimental synthesis and testing of SPI-based PEMs. However, combined efforts supported by the simulation-informed synthesis of hydrocarbon-based PEM can provide an understanding of the hydrocar bonstructure-property relationship that is still elusive. Moreover, data-driven polymer discovery is a promising methodofselectingpolymersfortargetapplications. Looking at the potential of data-driven polymer discovery of novel polymers as PEM, this the sis combines the effort put into ML-based identification of potential novel SPI PEMs as an alternative to Nafion and their validation using MD simulation and extensive experimentation. The thesis work was divided into three different objectives. In the first objective of the present work, a data set was prepared to comprise the physicochemical properties and proton conductivity data of SPI-based PEMs col lected and organized from the reported literature. The data set also included the chemical structures of the repeat units of the SPIs in computer-parsable SMILES for mat. Semi-empirically calculated properties and Quantitative structure-property re lationships(QSPR) properties were also included in the data set. Decision trees were trained to obtain certain rules for designing novel PEMs whose high proton conduc tivity could be ascertained with a high accuracy rate even before synthesizing them. Thus, following the rules, one SPI PEM, namely 1,4,5,8-naphthalene tetracarboxylic di anhydride/ 4,4’-diamino stilbene-2,2-disulfonic acid/4,4’-Diaminodiphenyl methane (NTDA/DSDSA/MDP) was designed for computational modeling using Molecular Dynamics(MD). In the second objective, all-atom molecular dynamics simulations were used to model the nano-phase segregation, the morphology of the ionic domains, and the dy namics of proton transport in a novel hydrocarbon-based PEM identified through the workdoneinthefirstobjective. Thediffusioncoefficientsofhydroniumionsandwater molecules, radial distribution function (RDF) plots between sulfur atoms of sulfonate groups and solvent phase (hydronium ions and water molecules), as well as polymer solvent volume fractions and fractional free volumes, have been calculated at increas ing levels of hydration (λ = 1, 5, 10 and 15) to understand the proton transport in the novel SPI PEM ionomers In the final objective, the NTDA/DSDSA/MDP SPI membrane, identified through data-driven and computational techniques mentioned in the first and second objec tives, was synthesized and proton conductivity was determined. Proton conductivity was found to be in the range of 0.1588- 0.28636 S·cm−1 which is exceptionally well for a PEM while those obtained in the MD simulations were 0.03- 0.18 S·cm−1. Thus, a goodagreementwasobservedbetweentheprotonconductivityvaluespredictedusing MDsimulations and the values for stable stand-alone SPI PEMs.