Continuous Flow Chemistry and Bayesian Optimization for Polymer-Functionalized Carbon Nanotube-Based Chemiresistive Methane Sensors

Abstract

We report the preparation of poly(ionic) polymer-wrapped single-walled carbon nanotube dispersions for chemiresistive methane (CH₄) sensors with improved humidity tolerance. Single-walled CNTs (SWCNTs) were noncovalently functionalized by poly(4-vinylpyridine) (P4VP) with varied amounts of a poly(ethylene glycol) (PEG) moiety bearing a Br and terminal azide group (Br-R₁). The quaternization of P4VP with Br-R₁ was performed using continuous flow chemistry and Bayesian optimization-guided reaction selection. Polymers (PyBrR₁) with different degrees of functionalization were used to disperse SWCNTs and subsequently incorporated into sensors containing a platinum complex as an aerobic oxidative catalyst with a polyoxometalate (POM) redox mediator to facilitate room-temperature CH₄ sensing. As the degree of quaternization in the PyBrR₁-CNT composites increased, improvements in response magnitude were observed, with nominally 10% quaternized PyBrR₁ giving the largest response. Incorporation of PEG improved sensor stability at relative humidities between 57-90% versus sensors fabricated from CNT dispersions with unfunctionalized P4VP. Devices fabricated with these dispersions outperformed those prepared in situ under dry conditions, and exhibited greater stability at elevated humidities. The influence of Keggin-type POM character was also evaluated to identify alternative POMs for enhanced sensor performance at high humidity. In an effort to identify areas for further improvement in algorithm performance for polymer functionalization, a kinetically informed machine learning model was explored as a route to predict reactivity of pyridine units and alkyl bromides under flow conditions.

Keywords: Bayesian optimization; chemiresistor; flow chemistry; methane; polymer wrapped carbon nanotube; sensors.

Figure 1

Summary of polymer synthesis and dispersions prepared in this study.

Figure 2

P4VP Functionalization Optimization Campaign Metrics. (A) Schematic of a typical GPR model and the EDBO+ campaign workflow. (B) Average quaternization per campaign round. (C) Maximum % quaternization observed (convergence) toward the 20% target during the optimization campaign. Three experiments performed per round.

Scheme 1. Schematic of (A) “Direct-from-Flow” Preparation for Polymer-Wrapped CNT Dispersions and (B) Deposition of [1-DMSO][MeOSO₃] Followed by the Choice of Polyoxometalate

Figure 3

XPS spectra of polymers and polymer-CNT composites. (A) XPS survey spectra of PyBrR₁-CNT samples, (B) high-resolution N 1s spectra of 5-PyBrR₁ polymer, (C) 10-PyBrR₁ polymer, and (D) 20-PyBrR₁ polymer.

Figure 4

Sensing response to 0.5% CH₄ for (A) devices fabricated using CNT dispersions containing P4VP with different degrees of Br-R₁ functionalization and (B) at increasingly higher levels of RH. All devices were functionalized with the Pt complex ([1-DMSO][OSO₃Me]) and H₅PV₂Mo₁₀O₄₀. Each data point represents an averaged response from three exposure-recovery cycles. Error bars represent one standard deviation from the mean.

Figure 5

Impact of removing components relevant to the CH₄ oxidation system on sensing response at different humidity levels, using spray coated 5-PyBrR₁-CNT. Each data point represents an average over three CH₄ exposure-recovery cycles and the error bar represents one standard deviation away from the mean.

Figure 6

Influence of choice of POM on CH₄ sensing response for 5-PyBrR₁-CNT sensors. (A) Integrated changes in conductivity during 0.5% CH₄ exposure at different humidity levels. No measurable response at 90% RH was observed for PtW₁₂. (B) Devices fabricated using the different POMs while excluding Pt. Each data point represents an averaged integral value over three exposure-recovery cycles and error bars represent one standard deviation away from the mean.

Figure 7

Dimensionality reduction study on generated data set. (Top) Chemical reaction schemes for a previously published small molecule (butylpyridinium bromide) and the polymers in this work. (Left) Illustration of the dimensionality reduced modeling configuration for 2D and 4D parameter spaces relative to a typical black box model. (A) The natural log of the reaction rate constant as a function of the reciprocal of the temperature for the generated polymer data set with corresponding Gaussian process and linear regressions on the correlation. (B) The MSE for the polymer yield prediction for a 10-sample testing set as a function of training set size for the black box and kinetics-informed linear and Gaussian process regressions. Data points indicate the median of 500 replicates with randomized model initialization and subsampling, and the shaded regions represent the inner quartile range. (C) The natural log of the reaction rate constant as a function of the reciprocal of the temperature for the small molecule reactions data set generated in prior work. (D) The MSE for the yield prediction as a function of training size for the small molecule reaction.