Ionic liquids (ILs) have been shown to have a significant effect on the redox behavior of organic compounds. These effects can be the shifting of oxidation and/or reduction potentials, stabilizing the redox reaction products, inducing certain additional chemical reactions, or even changes in the electrochemical reaction mechanism (e.g. the two one-electron reduction waves of dinitrobenzene in acetonitrile is reportedly collapsed to a single two-electron wave in ILs such as [C₄mim][BF₄]).³⁷ Such anomalies due to the presence of IL electrolytes are usually ascribed to the strong ion-pairing in ILs. In a series of computational studies, Fry et al.³⁸ have examined and reported the interactions between solvent, electrolytic ions, and electroactive species, indicating that the redox potentials can be lowered when ion-pairing with the electrolyte was included. We hypothesize that similar shifts in the oxidation mechanisms of the analytes (i.e. 2,4-TDI) can have significant implications on the development of a more efficient isocyanate electrochemical sensor. Therefore, cyclic voltammetry (CV) was first utilized to characterize an isocyanate redox activity in two non-aqueous solvents (i.e. acetonitrile and IL [C₄mpy][NTf₂]). Figure 1A presents the cyclic voltammogram of 2,4-TDI in both solvents, whereas nitrogen was used as the background. In the case of acetonitrile, the oxidation of 2,4-TDI occurred at 1.96 V. On the other hand, there was no redox peak observed in the pure IL indicating that the electrochemical system comprising IL is stable in this potential region. In the presenece of 2,4-TDI in the IL, the electrochemical oxidation peak of 2,4-TDI was observed at a less positive anodic potential (1.38 V) than in acetonitrile (1.96 V). Moreover, though the oxidation peak currents have similar values in both electrolytes, multiple shoulder peaks and a broad small reduction peak were observed before the major oxidation peak in the acetonitrile system, suggesting that the 2,4-TDI oxidation process in acetonitrile involves multiple steps. Such complicated redox mechanisms usually cause difficulty in quantitative analysis due to poor reproducibility and are not desirable for electroanalytical method development. In contrast, in the [C₄mpy][NTf₂] system, it showed one single and clear oxidation peak, which can be used for analytical quantification. Thus, utilization of [C₄mpy][NTf₂] was validated to be suitable for electrochemical detection of 2,4-TDI, as has been done by our group for various different analyte targets.^{28, 30, 31, 33, 39} In [C₄mpy][NTf₂], the oxidation peak current of 2,4-TDI was approximately seven times larger than that of the reduction peak, as shown in Figure 1A. Therefore, it is also clear that the oxidation process is capable of providing better sensitivity for the detection of 2,4-TDI (Figure 1B) and can be used for further quantifications. It was also found that the reduction process would interfere with the current signals of the oxidation process as well. As shown in Figure 1A, when the reduction process occurred first, the 2,4-TDI concentration would decrease in the region near the electrode surface leading to a decreased TDI oxidation signal as comapared to the narrower potential scan in Figure 1B (i.e. the oxidation peak current in Figure 1B was 381 μA, higher than the value of 297 μA in Figure 1A). Thus, in real-time sensing, the potential region wherein only oxidation of 2,4-TDI occurs is selected. The abovementioned experiments supported our rationale to study the optimal 2,4-TDI sensing conditions as well as demonstrating the benefit of employing IL in these sensor developments.

To further investigate the electrochemical process of 2,4-TDI and to mimic real-world sensing conditions, similar cyclic voltammetric experiments were conducted using air as the background gas, as shown in Figure 2. Oxygen is ubiquitous in the ambient environment and plays significant roles in any sensing method development. Especially because its presence can also modify electrochemical processes in ILs, even though the reductively generated superoxide anion radical is more stable than in other solvents. Xiao et al. have reported the effect of the superoxide radical on carbon dioxide electrochemistry.³³ Herein, it is shown that for a narrow potential scan wherein oxygen reduction processes are not involved, a clear and large 2,4-TDI oxidation peak was observed (red curve). The same was true for the experiment with a wider potential window, but with initiating the scan in the anodic direction, thereby performing the 2,4-TDI oxidation before its reduction (green curve). However, when the scan direction was first in the cathodic potential, involving the generation of superoxide shown by a reduction peak at ~−1.4 V, the peak for the oxidation of 2,4-TDI almost disappeared (see the blue curve). From this data, two important conclusions can be drawn. First, 2,4-TDI can react with the reduction products of the most important environmental interferent, oxygen,⁴⁰ thereby leaving no TDI at the interface to be detected. This property can be used for removing toxic TDI. But, it supports the choice of using only the oxidation processes for quantification, as shown by the results of these experiments, especially if the electrochemical 2,4-TDI sensor has to be used in real ambient environments. Second, this experiment also shows the selectivity of the sensing system for 2,4-TDI under the chosen experimental conditions. The peak at 1.38 V is clearly due to the presence of 2,4-TDI, which remains there in the presence of oxygen and all other air gases, although it disappears in the presence of superoxide due to their reaction. Thus, if the fomation of superoxide radical is avoided by choosing a specific potential window, 2,4-TDI can be selectively detected in real ambient environments.

For further elaboration of the electrochemical reaction mechanisms for the abovementioned processes, we performed the CV at various scan rates. Figure 3 shows the cyclic voltammogram of 0.1% v/v 2,4-TDI in [C₄mpy][NTf₂] at scan rates of 10 mV/s – 500 mV/s. An oxidation peak potential shift (from 1.77 V to 1.24 V) was observed when the scan rate was changed from 500 mV/s to 10 mV/s. For a fast, electrochemically reversible system, the peak potential does not shift at different scan rates. In the case of 2,4-TDI, the electrode reaction is an irreversible process even at a slow scan rate (e.g. 10 mV/s). Thus, the peak potential shift is suggested to be the results of multiple coupled reactions, e.g. chemical reactions coupled with the electrochemical reactions or the electron transfer reaction reactions rather than the uncompensated IR drop, based on the following results: (1) the peak current has a linear relationship with the scan rate, supporting a surface controlled electron transfer process; (2) the increase of impedance vs. frequency in EIS measurement at various time intervals (Figure 5) supporting the surface reactions discussed below.

Furthermore, it is important to note in Figure 3 that a shoulder peak started to emerge at about 0.8 V. This is more evident at faster scan rates. Kuroiwa et al.⁴¹ have reported a similar shoulder peak for the oxidation reaction of chlorophyll a in IL/acetonitrile mixture, whereas this peak can be de-convoluted from the voltammogram and the resulting peak current can be estimated by using Gaussian curve fitting methods. The peak current ratio of P_I/P_II decreases as the scan rate increases (from 0.251 at 10 mV/s to 0.0254 at 500 mV/s) where P_I is the shoulder peak and P_II is the main oxidation peak, which suggests that the first oxidation process (P_I) has relatively slower kinetics and can be related to the aggregation of the analyte molecules as was the case with chlorophyll a. Furthermore, the sufficiently large oxidation current for this peak strongly supports the idea that analyte molecules form a low-order aggregate, such as a dimer, because high-order aggregates would not yield sizeable oxidation currents due to slow diffusion. Such aggregation processes have been reported to lower the oxidation potential of the analyte in addition to the similar effects caused by IL ion-pairing. To further investigate this peak P_I and to validate the possibility of a dimerization process, a series of CV measurements with different time intervals were designed. In this experiment, 2,4-TDI was equilibrated in the IL electrolytes at different times and then the measurements were performed. The 2,4-TDI oxidation peak current decreased when the CV measurements were performed at longer time intervals (Figure 4A). No peak shift was observed as was the case during the variation of scan rates. The peak current values followed a first-order exponential decrease (Figure 4B) with a correlation coefficient equal to 0.994. This exponential decay of the 2,4-TDI oxidation peak current can also be utilized for a selectivity study of the future sensor device. In Figure 4A, the shoulder oxidation peak (P_I, located at 0.8 V) was also observed. This peak was more clear at longer time intervals (e.g. 1000 min) and fairly insignificant when the measurement was carried out instantaneously (0 min). The reduced current signals for P_II and increasing current signals for P_I with increasing time of measurement could only be attributed to the adsorption of 2,4-TDI at the electrode surface by the formation of dimers, trimers and even polymers, through isocyanate self-addition reactions (eq.1 & eq.2, where R is defined as the toluene aromatic ring), which lowered the concentration of electrochemically active species near the electrode surface. (eq. 1)(eq. 2)

We performed electrochemical impedance analysis to understand further about the interface reactions at different time scales. EIS results can provide valuable information about the dynamic surface adsorption processes of 2,4-TDI in the [C₄mpy][NTf₂] electrolyte. EIS has been used to study many surface reactions such as dye and surfactant dimerization processes.⁴² Thus, EIS experiments were carried out at several time intervals in the presence of 2,4-TDI at an open circuit potential. As shown in Figure 5A and Figure S3, the Nyquist plots change at various measurement times. As shown in Figure S3, we simulated the EIS curves by carefully selecting the equivalent circuit according to the literature. Among all the equivalent circuit models we employed to fit the experimental data, the one that obtained the best fit is presented in the inset of Figure S3. These simulated results are not good fits, especially at the zero hour. We believe that the deviation from a perfect fit is an indication for the dynamic chemical reactions (i.e. the TDI self-polymerizations). Figure 5B, the Bode curves of the EIS experiment present the correlation of log |Z| versus log frequency at different time intervals. The Z’ versus frequency and Z” versus frequency plots (Figure S3 C-D) show that the real part of the impedance increases with time at low frequencies much more than the imaginary part of the impedance. It is clear that the surface reactions cause the impedance to increase at longer time intervals, especially at low frequencies. Such trends can also be observed in the Nyquist plots too (Figure 5A) and are similar to the ones reported earlier for different surfactants and dyes in dimerization/aggregation reactions.

Results from the EIS experiments confirm the presence of dynamic interface processes of 2,4-TDI and are consistent with the results presented in Figure 4. Toluene was used as a negative control to further understand the 2,4-TDI oxidation processes in the IL at the glassy carbon electrode. The selection of toluene as a reference sample was due to its similar chemical structure as 2,4-TDI except the two isocyanate groups (–N=C=O). No oxidation peak was observed in the toluene sample with the same conditions as for 2,4-TDI. Thus, the oxidation peak and the consequent surface modification/dimerization are attributed to the isocyanate groups.

UV-Vis and FT-IR spectroscopy experiments were employed to determine the 2,4-TDI oxidation products for additional insights on the reaction mechanism. There are two possible molecular sites for 2,4-TDI oxidation: 1) the C-N single bond cleavage at the 2 and/or 4 position of the benzene ring or 2) the N=C double bond cleavage in the isocyanate functional group. First, UV-Vis measurements (Figure 6A) were performed to characterize samples of different 2,4-TDI concentrations in [C₄mpy][NTf₂] as well as the sample of pure [C₄mpy][NTf₂] as the background. As the concentration of 2,4-TDI increased, the absorbance peak at 290-300 nm⁴³ also increased with a red shift, indicating an enhanced delocalized conjugation effect. This could be attributed to isocyanate self-addition forming isocyanate polymers. We also characterized the bulk IL near the glassy carbon electrode surface after an electrochemical oxidation of 2,4-TDI was conducted. A significant decrease of the absorbance and a blue shift were observed from this bulk IL sample. The absorbance peak changes indicated that there was a reduced conjugation of the 2,4-TDI molecular structure, which could be attributed to the loss of the isocyanate functional group due to electrochemical oxidation of 2,4-TDI, causing a decrease of the delocalized conjugation effect.

FT-IR measurements (Figure 6B) were carried out to obtain additional information in terms of identifying the bond breaking of the isocyanate functional group. An absorbance peak at 2279 cm⁻¹ was observed from [C₄mpy][NTf₂] in the presence of 0.5% v/v 2,4-TDI. This peak was not found in either pure [C₄mpy][NTf₂] or the [C₄mpy][NTf₂] sample that was taken from the bulk solution near the glassy carbon electrode surface after an electrochemical oxidation of 2,4-TDI was performed. According to Thomson et al.,⁴⁴ the absorbance peak at 2279 cm⁻¹ could be assigned to the asymmetric vibration of N=C=O functional group. This illustrated that the NCO functional group disappeared. Thus, it is plausible that the N=C double bond was cleaved during 2,4-TDI oxidation. Thus, the plausible mechanism is an CECE mechanism depicted as follows (R is defined as the toluene aromatic ring, we use the dimer as a case study), in which the chemical process is the rate limiting step: (eq. 4)(eq. 5)(eq. 6)

2,4-TDI monomer undergoes a self-addition reaction in the initial state (eq. 3). When an external anodic potential is applied, eq. 4 occurs yielding a stable radical product due to the delocalized electron conjugation effect. It is followed by a surface adsorption process (eq. 5, C stands for glassy carbon) at anodic conditions. This step is rate-limiting because of the reactivity of the carbon surface. Then, a second oxidation occurs involving the C-N bond breaking forming adsorbate residues (eq. 6). This proposed mechanism complies with CV data in Figure 4A that the shoulder peak becomes more significant at longer testing time intervals.

2,4-TDI was selected as the representative molecule of the isocyanates class, because it has a high vapor pressure of 0.05 mm Hg and is the most common airborne isocyanate. As described earlier, toluene can be used as a negative control for 2,4-TDI oxidation process as it has the same structure with the absence of isocyanate groups. However, it can also be tested as an interfering molecule for 2,4-TDI detection. When toluene was subjected to the electrochemical oxidation process, no oxidation peaks were observed, as shown in Figure S1, even for a concentration of 4%. In the same experiment, TDI at a much lower concentration showed reproducible signals. In the other experiment, aniline, which is quite well known as an oxidizable compound, was tested in comparison with 2,4-TDI. At the same concentration and under the same conditions, aniline showed an oxidation current even larger than the target (Figure S2), though the peak potential of aniline is quite different. Thus, different electrochemical techniques, such as chronoamperometry, can be used to selectively detect the target isocyanates by choosing a certain measurement potential.

However, the most important observation in this regard came from the time lapse experiment carried out with aniline and 2,4-TDI, the data is shown in Figure 7. As indicated, no significant change in the oxidation current signal occurs in the IL that contains aniline but the same shoulder peak is present in IL containing TDI, thereby decreasing the oxidation peak current with every time step. This indirectly proves the self-addition reaction of TDI, and alongside, this mechanism can serve to provide an extremely selective response mechanism for TDI detection with signal diminishing as a function of time. In addition, Yakabe et al.⁴⁵ reported that aniline and 2,4-TDI exhibit different side reactions in the presence of trace water (i.e. 2,4-TDI produces CO₂ when reacted with water), which can be utilized to enhance the selectivity of the detection.

Finally, the sensitivity experiments were performed to evaluate LODs and T₉₀ for the developed sensor mechanism. To perform liquid phase 2,4-TDI detection, 2,4-TDI was directly added into the electrochemical cell with [C₄mpy][NTf₂] electrolyte in an air environment. The concentration of 2,4-TDI was varied from 0% v/v to 0.5% v/v, with 0.1% v/v increments. The cyclic voltammograms and the corresponding calibration curves are presented in Figure 8. The value of the correlation coefficient (R²) was calculated to be 0.997. The detection sensitivity could be obtained from the slope of the linear fitting curve, which was 252.6 μA/μL. By analyzing the mean background signal (S_bm) and the corresponding standard deviation (s_bm) value, the minimum distinguishable signal (S) can be obtained via eq. 7. Thus, the limit of detection (LOD) can then be converted from S in the calibration equation shown in Figure 8B. LOD in this liquid phase 2,4-TDI sensor is 0.0651 μL in 500 μL, which equals to 130.2 ppm. (eq. 7)S=Sbm+3∗Sbm

For gas sensing, one of the most important aspects is the sampling of analyte gases. The use of hydrophobic, non-volatile IL as an electrolyte enables a simple gas sampling system that can be open to the ambient environment. This not only allows pre-concentration to further increase the sensitivity of the detection, but can also significantly simplify future sensor system integration and provide long sensor lifetimes. Real-time current response recording (Figure S4) in gas phase sensing was performed via chronoamperometry at different 2,4-TDI concentrations. Figure 9A presents the chronoamperometry data with time offsets by showing the sensor response when 2,4-TDI is added and by removing the non-Faradic current decay period. Becasue the 2,4-TDI sample is in liquid form, a sampling system consisting of two gas paths was designed and depicted in Scheme S2. By adjusting the flowrates of the two gas paths, an accurate amount of gas phase 2,4-TDI can be calculated (i.e. the gas phase 2,4-TDI is 32.89 ppm when the flowrate of both gas paths is 100 sccm). The external potential was set at 1.4 V, based on the cyclic voltammogram in Figure 8. Each measurement was carried out in samples that were separately prepared. Air was used as the background gas in order to fully explore the real life potential of the sensor. The sensing current reached a signal plateau at about the 600^th second. It is interesting to note that the sensing current curve exhibits two stages. This observation is consistent with the CV data presented in Figure 3 and Figure 4A, indicating that there are multiple reactions involved.

The plot of sensitivity values at different sensing time scales (Figure 9B) also demonstates that there are two separate electrochemical steps, of which the transition point is the 200^th second. Detailed calibration curves at various sensing time scales are shown in Figure S5. By analyzing the sensor response at different time scales, properties such as sensitivity and LOD values vary (Table S2). For example, when picking the signal plateau at the 600^th second, the sensitivity is 1.942 μA/ppm with a value of LOD at 0.7693 ppm. However, if picking one of the times before the plateau (e.g. 100^th second), the sensitivity decreases to 0.2083 μA/ppm with a relative higher value of LOD (12.46 ppm). Moreover, T90 values of this sensor to function for 100 and 600 seconds are calculated to be 90.80 seconds and 432.3 seconds, respectively. Thus, by sacrificing the sensor response time, higher sensitivity and lower LOD values can be approached, which gives this sensor a potential to tune different functioning times for different sensing needs.