• ## Evaluation of Poly (vinylferrocene)/Platinum Catalyst for Methanol Sensing by Box-Behnken Design

Donglin Zhang1 and Yijun Tang2*

1Ashwaubenon High School, 2391 S Ridge Road, Ashwaubenon, Wisconsin 54304, USA

2Chemistry Department, University of Wisconsin Oshkosh, Wisconsin 54901, USA

*Correspondence to: Yijun Tang, E-mail: tangy@uwosh.edu

Citation: Zhang D, Tang Y (2018) Evaluation of Poly (vinylferrocene)/Platinum Catalyst for Methanol Sensing by Box-Behnken Design. SCIOL Biomed 2018;2:53-59.

• ## Abstract

Fast detecting and monitoring the level of methanol is critical in the studies of methanol toxicity and related pathophysiology. The mechanism of such sensing is based on the oxidation of methanol in the sample. Poly (vinylferrocene) polymer was coated onto the surface of a platinum-disk electrode so that the electrode was suitable for the oxidation of methanol. The number of ferrocene centers was determined by the concentration of the polymer (X1) and the coating time (X2). Platinum nanoparticles were then incorporated to the polymer network by a redox reaction. The number of nanoparticles was determined by the reaction time (X3). The performance of such surface modification was reflected by the methanol oxidation peak current (Y) in a cyclic voltammetry test. Box-Behnken design ensured minimum combinations of X1, X2 and X3 to generate statistically reliable evaluation of the performance. Multivariable polynomial regression and analysis of variance revealed that both the thickness of polymer network and the number of platinum nanoparticles had significant effects on the catalysis performance.

## Keywords

Methanol, Box-Behnken design, Chemical sensor, Conductive polymer, Polyvinyl ferrocene, Nanoparticles, Catalysis, Electrochemistry

## Introduction

Sensing and monitoring methanol is critical in the studies of the methanol toxicity of and related pathophysiology [1-5]. Fast and reliable detection of methanol in various locations and environments is then desired in such studies [6]. There are many techniques aiming to the detection of methanol. Contrary to enzymatic detection methods, chemical sensing takes advantages of certain chemical properties of methanol [7]. This technique is generally more robust and stable; therefore, it is suitable for long-term monitoring.

The key component in chemical sensing is the sensor which often relies on catalysts to oxidize methanol. The most effective catalyst is platinum [8-10], although non-platinum catalysts have also been reported [11-13]. The addition of some other transition metals such as gold, ruthenium, etc. enhances the catalytic ability of platinum [14-18]. The supporting materials for the catalysts are also important in designing new chemical sensors. Various materials have been reported for this purpose including newly-emerging materials such as conductive polymer and nanoscale materials [19,20]. Poly (vinylferrocene) or PVF, a conductive polymer, conducts electrons through successive redox processes or electron "hopping". It has been reported that this polymer is used in the detection of some small molecule gases: Methane, ethanol, etc [21].

Many factors influence the performance of the methanol sensors: The identity of the metals, the ratio of them, the support of the catalyst, etc [9,22-24]. A new catalyst or catalyst system is demanded when the direct methanol fuel cell changes in the electrolyte or the structure of its components. In a laboratory, the research of catalyst for methanol oxidation is often carried out by means of electrooxidation [25-28]. A controlled potential at proper level is applied to an electrode modified with the catalyst layer. The amplitude of methanol oxidation current reflects the effectiveness of the catalyst. To optimize the combination of parameters is critical in developing an efficient and effective catalysts for methanol detection. Box-Behnken design is a powerful statistical tool to reduce the number of repetitive and replicate experiment to optimize the experimental conditions [29].

In this article, we present the utilization of a three-factor three-level Box-Behnken design in a mechanistic study of catalysis for the methanol electrooxidation on the surface-modified electrode. A further regression and analysis of variance (ANOVA) were adopted in the study to figure out the most critical factors in the modification of electrode surface.

## Material and Methods/Experimental Details/Methodology

### Chemicals and materials

Poly (vinylferrocene) (PVF) (Figure 1) was purchased from Polysciences (cat no. 09746); 99.9% extra dry dichloromethane (CH2Cl2) made by ACROS was purchased from Fisher Scientific (cat no. AC326850025); potassium tetrachloroplatinate (II) (K2PtCl4) by Aldrich was purchased from Sigma-Aldrich (SKU 206075); methanol (CH3OH) made by BDH was purchased from VWR (cat. No. BDH1135); and 70% perchloric acid (HClO4) by Aldrich was purchased from Sigma-Aldrich (SKU 311421). All chemicals were used as received without further purification. Platinum-disk electrode (part no. CHI102), saturated calomel electrode or SCE (part no. CHI150) and platinum wire electrode (part no. CHI115) were all purchased from CH Instruments.

## Figure 1

Figure 1: Chemical structure of poly (vinylferrocene). The figure shows two repeating units. Fe in the ferrocene group has an oxidation state of +2. View Figure 1

### Overall procedures

Figure 2 indicates the procedures of the experiment, which were modified from an earlier report [30]. Firstly, a platinum-disk electrode was coated with PVF polymer. This was fulfilled by dipping a platinum-disk electrode in an X1 mM PVF solution for X2 minutes. Secondly, platinum nanoparticles were incorporated in the PVF network by immersing the PVF-coated electrode in a 2 mM K2PtCl4 solution for X3 minutes. Thirdly, the electrooxidation of methanol was performed with a three-electrode cyclic voltammetry (CV) tests. The anodic peak current of methanol oxidation (Y) reflected the performance of the PVF/Pt catalyst. Therefore, Y was a function of X1, X2 and X3.

## Figure 2

Figure 2: Procedures of the experiment. View Figure 2

The experiment described in Figure 2 was repeated with various combinations of X1, X2 and X3. The selection of X1, X2 and X3 followed a Box-Behnken design. The further regression and analysis of variance (ANOVA) would reveal which factors (X1, X2 or X3) or combinations of factors were more critical to the PVF/Pt catalyst. The following sections include the detailed description of each step in the overall procedures.

### Modification of electrode surface

A platinum-disk electrode (2 mm diameter) was polished with 1.0, 0.3 and 0.05 micron alumina powder, 3 minutes for each. It was then rinsed with DI water and gently blotted dry. The electrode was blown with air for 3 minutes. After the electrode was dry, it was immersed in a CH2Cl2 solution containing 2, 6 or 10 mM PVF for 5, 10 or 15 minutes. The actual concentration and time were demanded by the Box-Behnken design which will be discussed later. The concentration of PVF was designated as X1 and the time in this step was designated as X2. After X2 minutes, the electrode was blown with air for 1 minute. Now PVF polymer was coated on the surface of the platinum-disk electrode.

The PVF-coated platinum-disk electrode was then immersed in a 2 mM K2PtCl2 solution for 0.5, 1 or 1.5 minutes. This time was designated as X3. Its actual value was also demanded by the Box-Behnken design. In this step, Pt nanoparticles were formed in the PVF network according to the following reaction.

$2PVF+{K}_{2}PtC{l}_{4}\to 2PV{F}^{+}C{l}^{-}+P{t}^{0}+2KCl$

The oxidation state of Fe in ferrocene center (represented by PVF in above equation) was +2. That was +3 in ferrocenium (represented by PVF+). The above reaction only took place in the vicinity of ferrocene centers. After X3 minutes, the electrode was removed from the K2PtCl4 solution. It was rinsed gently in DI water. This concluded the preparation of one working electrode. The same surface modification was applied to another 14 platinum-disk electrodes.

### Cyclic voltammetry

Cyclic voltammetry was performed with a conventional three-electrode setup on a potentiostat. A CHI650C electrochemical workstation by CH Instruments was used for the work in this paper. The working electrode was a surface-modified platinum-disk with diameter of about 2 mm. The surface was modified with PVF and Pt nanoparticles as described above. The reference electrode was a saturated calomel electrode (SCE) and the counter electrode was a platinum wire electrode. The electrolyte was a 0.05 M HClO4 aqueous solution which also contained 2M CH3OH. Potential swept up and down for 10 cycles in the range of -0.275 V and 1.2 V vs. SCE. The scan rate was 100 mV/s.

Figure 3A shows the cyclic voltammogram of 10 cycles. Peaks 1 and 3 were the reversible oxidation/reduction of PVF polymer. Peak 2 was the oxidation of methanol. The half reaction of complete methanol oxidation is shown below.

$C{H}_{3}OH+{H}_{2}O\to C{O}_{2}+6{H}^{+}+6{e}^{-}$

CO2 (or CO if partial oxidation) was one of the products. The oxidation was irreversible. In some cases, an absorption peak might be present between 0.5 - 0.6 V in the cathodic scan when potential swept from 1.2 V back to -0.275 V. The amplitude of peak 2 reflected the performance of the PVF/Pt catalyst. To be consistent in the test, only the 10th cycle of each voltammogram was used for further data analysis. Some approximation was also used in determining the amplitude of peak 2 for the same reason. Peak 1 and peak 2 were close to each other and peak 2 was wide. It was difficult to figure out a clear baseline for peak 2 alone; therefore, an arbitrary baseline was drawn from point a to point b (Figure 3B). These two points were the inflection points adjacent to peak 2. The point c was the tip of peak 2. Point d was on the baseline ab with the same potential as point c. The distance from c to d was used as the amplitude of peak 2, which was designated as Y. This approximation was statistically meaningful although it might not generate the most accurate peak amplitude. It was critical that the same approximation was applied consistently on all 15 working electrodes.

## Figure 3

Figure 3: Cyclic voltammogram of Run 1 in 2 M methanol + 0.05 M HClO4. Scan rate: 100 mV/s. A) All 10 cycles. Peak 1 was the oxidation of ferrocene. Peak 2 was the oxidation of methanol. Peak 3 was the reduction of ferrocenium; B) The 10th cycle. Line ab was the arbitrary baseline. Line cd was the amplitude of methanol oxidation. View Figure 3

### Box-Behnken Design and ANOVA

Box-Behnken design is a statistic method to determine which factors are significant to the result by minimum replica of experiment [31]. Three factors (X1, X2 and X3) were chosen in this study. Each factor varied at three levels. For X1, they were 2, 6 and 10 mM; for X2, they were 5, 10 and 15 minutes; and for X3, they were 0.5, 1 and 1.5 minutes. Considering the three-dimensional structure of the PVF network on the electrode surface, it is reasonable to assume that X1 determined the amount of ferrocene centers in each PVF layer; X2 determined the number of layers or thickness of the PVF network on the platinum-disk electrode surface and X3 determined the amount of platinum nanoparticles in the polymer network. X1 combined with X2 determined the overall amount of ferrocene centers in working electrodes. The amplitude of peak 2 (Y) changed with X1, X2 and X3. A full factorial design would have required 3 × 3 × 3 or 27 different combinations of these three factors. Box-Behnken design reduced this number to 15. The combinations that were selected in this laboratory were listed in Table 1 in the Results section. The selected combinations were tested in a random order (as shown in the Run Order column in Table 1) to minimize the possible influence of systematic errors.

## Table 1

Table 1: Experimental conditions (combinations of PVF concentration X1, PVF contact time X2, K2PtCl4 contact time X3) and the peak current (Y) of methanol oxidation. View Table 1

## Results and Discussion

The peak current (Y) of methanol oxidation obtained with cyclic voltammetry is listed in Table 1. It reflected the performance of the PVF/Pt catalyst, which was a function of PVF concentration (X1), PVF contact time (X2) and K2PtCl4 contact time (X3).

Multivariable polynomial regression was carried out based on the data in Table 1. With three independent variables, the fitting model was as follows, where all c's were coefficients.

The result of regression had a multiple regression coefficient R of 0.8475 or R2 of 0.718. This value was acceptable for a multivariable regression. Table 2 shows some information of the regression and further analysis of variance (ANOVA).

## Table 2

Table 2: Multivariable polynomial regression and ANOVA. View Table 2

P-values were the probability of null hypotheses, which stated that the factor had no significant effect on the result (i.e. performance Y in this case). A null hypothesis was rejected if its probability was too low. Therefore, a larger p-value (close to 1) indicated that the factor had no significant effects on the performance (Y). On the contrary, a small p-value (close to zero) revealed that the factor was significant to the performance. Table 2 indicates that both X2 (the thickness of PVF polymer network) and X3 (the amount of Pt nanoparticles) had quadratic effect on the PVF/Pt catalyst performance. Effect of PVF polymer could be ascribe to the fact that ferrocene centers prevented CO-like poisoning [19]; while effect of Pt nanoparticles was due to the additional oxidation centers on the electrode surface. There was no evidence of interactive factors. This suggests that PVF and Pt nanoparticles enhanced the catalyst performance independently. The ratio of PVF/Pt-nanoparticle was not a significant factor to the catalysis.

## Conclusion

Chemistry research becomes more efficient with proper Box-Behnken design. The number of replicate tests can be reduced, yet the results are still statistically reliable. A robust Box-Behnken design tolerates and offsets random errors and some systematic errors (e.g. baseline drifting), which could not be ideally provided by other techniques. In our study of PVF/Pt catalysts, many factors could affect the catalysis performance. The density of ferrocene centers in each PVF layer, the thickness of polymer network and the loading of Pt nanoparticles all play a role. The ANOVA analysis revealed that the thickness of polymer was critical to the PVF/Pt catalysis towards methanol oxidation. This was presumably due to the prevention of CO-poisoning. The amount of Pt nanoparticles was another significant factor in the catalyst-facilitated methanol oxidation, owing to the catalytic ability of the Pt.

## Acknowledgements

This paper was supported by the Faculty Development Program at the University of Wisconsin Oshkosh.