Continuous Flow Chemistry\nand Bayesian Optimization\nfor Polymer-Functionalized Carbon Nanotube-Based Chemiresistive Methane\nSensors

We report the preparation of poly(ionic) polymer-wrapped\nsingle-walled\ncarbon nanotube dispersions\nfor chemiresistive methane (CH<sub>4</sub>) sensors with improved\nhumidity tolerance. Single-walled CNTs (SWCNTs) were noncovalently\nfunctionalized by poly(4-vinylpyridine) (P4VP) with varied amounts\nof a poly(ethylene glycol) (PEG) moiety bearing a Br and terminal\nazide group (Br-R<sub>1</sub>). The quaternization of P4VP with Br-R<sub>1</sub> was performed using continuous flow chemistry and Bayesian\noptimization-guided reaction selection. Polymers (PyBrR<sub>1</sub>) with different degrees of functionalization were used to disperse\nSWCNTs and subsequently incorporated into sensors containing a platinum\ncomplex as an aerobic oxidative catalyst with a polyoxometalate (POM)\nredox mediator to facilitate room-temperature CH<sub>4</sub> sensing.\nAs the degree of quaternization in the PyBrR<sub>1</sub>-CNT composites\nincreased, improvements in response magnitude were observed, with\nnominally 10% quaternized PyBrR<sub>1</sub> giving the largest response.\nIncorporation of PEG improved sensor stability at relative humidities\nbetween 57–90% versus sensors fabricated from CNT dispersions\nwith unfunctionalized P4VP. Devices fabricated with these dispersions\noutperformed those prepared in situ under dry conditions, and exhibited\ngreater stability at elevated humidities. The influence of Keggin-type\nPOM character was also evaluated to identify alternative POMs for\nenhanced sensor performance at high humidity. In an effort to identify\nareas for further improvement in algorithm performance for polymer\nfunctionalization, a kinetically informed machine learning model was\nexplored as a route to predict reactivity of pyridine units and alkyl\nbromides under flow conditions.