Amperometric Alcohol Vapour Detection and Mass Transport Diffusion Modelling in a Platinum-Based Sensor

Abstract

An important class of analytes are volatile organic carbons (VOCs), particularly aliphatic primary alcohols. Here, we report the straightforward modification of a commercially available carbon monoxide sensor to detect a range of aliphatic primary alcohols at room temperature. The mass transport mechanisms governing the performance of the sensor were investigated using diffusion in multiple layers of the sensor to model the response to an abrupt change in analyte concentration. The sensor was shown to have a large capacitance because of the nanoparticulate nature of the platinum working electrode. It was also shown that the modified sensor had performance characteristics that were mainly determined by the condensation of the analyte during diffusion through the membrane pores. The sensor was capable of a quantitative amperometric response (sensitivity of approximately 2.2 µA/ppm), with a limit of detection (LoD) of 17 ppm methanol, 2 ppm ethanol, 3 ppm heptan-1-ol, and displayed selectivity towards different VOC functional groups (the sensor gives an amperometric response to primary alcohols within 10 s, but not to esters or carboxylic acids).

1. Introduction

The detection of VOCs (volatile organic compounds) is considered highly useful in many aspects of modern life, as they are prevalent in many human activities and can be a hazardous component of indoor air [1,2,3,4]. Thus, the monitoring and detection of these hazardous components is vitally important. The rapid and easy detection of VOCs, such as aliphatic primary alcohols (particularly in aerosol form), has long been a goal for many areas of research [5,6,7,8], even though there are plenty of analysis methods to choose from including HPLC (high-performance liquid chromatography), NMR (nuclear magnetic resonance), LC-MS (liquid chromatography coupled with mass spectroscopy), GC-MS (gas chromatography coupled mass spectroscopy), and IR (infra-red) spectroscopy [9,10,11,12]. However, whilst these methods are accurate and reliable, they are expensive, time-consuming, and require off-line analysis [13,14,15,16]. Portable breath analysers used by the police (typically ethanol sensors) are just one example of an area where rapid and accurate VOC detection is of paramount importance [17,18,19,20]. Primary alcohols such as methanol and ethanol are widely used in the industry, but they are harmful to people and the environment. Therefore, the leak detection of such chemicals is very important to ensure a safe working environment [21,22,23,24]. Thus, the development of a small, lightweight sensor for primary alcohol detection is a worthwhile area of research. There are numerous reports in the literature involving the electrochemical detection of alcohols via platinum (Pt) electrodes [25,26,27,28], including the use of electrochemical sensors which have platinum-based working electrodes [29,30,31,32]. CO-AF sensors (commercially available from Alphasense Ltd., Great Notley, UK) are mature technology three-electrode sensors with a Pt working electrode (WE). Note that CO-AF and CO-A1 are not scientific abbreviations; they are simply product names used to identify the different sensors. Therefore, it was theorised that the CO-AF sensors could be simply and easily modified to detect primary alcohols such as methanol and ethanol. This was achieved by removing the potassium permanganate from the sensor during the build process; the modified sensor is designated CO-A1 throughout this investigation. The use of a commercially successful, mature technology sensor as a basis for the modified sensor used in this investigation is considered prudent. This is because not only is the design and build of the sensor stack and its components well established and industrialised but by only changing one aspect of the sensor, any changes in performance characteristics can be attributed to that change. It also reduces the risk of introducing the possibility of systematic errors or discrepancies in the build quality of the sensor. It also means that the sensor can be readily manufactured and deployed to customers with only a minor change in the build process. This keeps production costs down and makes it more commercially viable. It was hypothesised that primary alcohols would have their hydroxyl moiety react at the platinum working electrode and give an amperometric response during chronoamperometry, mainly via surface adsorption and subsequent electron transfer. This is expected even though in certain circumstances (such as in low pH solutions), methanol can “poison” platinum electrodes via the adsorption of intermediates on the Pt electrode surface, and this leads to a subsequent reduction in the reactivity of the Pt electrocatalyst [33,34].

1.1. Electrooxidation of Alcohols

It is known that alcohols (and some other VOCs) undergo electrooxidation in sulfuric acid solution [35,36,37,38]; modified CO-AF sensors (subsequently designated CO-A1) were, therefore, used to attempt to detect a range of primary aliphatic alcohol vapours. Owing to the relatively mild conditions of the experiments (<50 °C), it was expected that the electrochemical reaction at the platinum electrode would not proceed to completion and would stop after the formation of ethanoic acid. This is because the thermal energy available is insufficient to initiate the C-C bond cleavage within the alcohol [39,40]. It should be noted that some studies of this mechanism use acidic electrolytes with weakly coordinating anions such as HClO₄ [40], but this is not a practical electrolyte for amperometric gas sensors and the choice of sulphuric acid, which may partially poison the Pt electrocatalyst, is a necessary compromise. It is suggested that the consideration of 4e⁻ oxidation in this work is partly due to an effect resulting from the electrolyte used, subsequently reducing the reactivity of the Pt working electrode. The n-values (the number of electrons) correspond to the electrooxidation of alcohols to aldehydes (n = 2) or to carboxylic acids (n = 4) at the Pt electrode surface [41,42]. Complete electrooxidation was achieved to CO₂ where n = 12 only occurred at temperatures above the operating limits of the Alphasense CO-AF sensor (maximum recommended 50 °C) [43,44]. It was considered that for primary alcohols, n = 4 is favoured between 20 and 60 °C, which leads to the formation of carboxylic acids. We used this information when estimating the effective layer thickness (L) and diffusion coefficient (D) under steady-state, mass transport-limited conditions (Equation (1)). The calculations for the primary alcohols assume that n = 4 e⁻ [45].

$$I=\mathit{nFAC}{\displaystyle \frac{D}{L}}$$

(1)

• I = steady-state current in amps (A);
• n = no. of electrons (taken to be four);
• F = Faraday’s constant taken as 96,485 C/mol;
• A = area of working electrode (cm²);
• C = concentration of analyte (mol/dm³);
• D = effective diffusion coefficient (cm²/s);
• L = apparent diffusion layer thickness (cm).

1.2. Sensor Technology Considerations

Having discussed the electrochemical aspects of the internal workings of the sensor, we now discuss the wider engineering aspects of the complete sensor. The major differences between commercially viable gas-sensing technologies and technologies that work in a small-scale laboratory setting are cost and scalability. The cost of components is rarely a major consideration when building a prototype gas sensor in a laboratory. However, the cost of building a gas sensor in the industry is often the deciding factor when choosing which sensing technology to use. The scalability of technology in an industrial setting must also be considered. Whilst making one gas sensor by hand using skilled technicians may be feasible in a laboratory, the gas sensor made in the industry must be able to be mass-produced via automated processes with operators who are not necessarily experts in their field. The major commercial gas-sensing technologies are photoionisation detectors, solid-state sensors, and amperometric gas sensors. This is mainly because they are well-established technologies and have been widely researched over the years in numerous different disciplines, from the ppm detection range of CO gas and the monitoring of the health of stored potatoes to VOC detection for environmental conservation monitoring and oil prospecting [46,47,48,49]. Some typical examples of different types of gas detectors are now discussed. This is not a comprehensive overview of all proposed gas-sensing technologies but is restricted to those which have proven commercially viable in various applications (

Table 1. Commercial sensor applications.

Sensor Technology	Application	Advantages	Disadvantages
Photoionisation	VOC gas concentration analysis [47]	High accuracy whilst being relatively cheap to run [48]	Low selectivity which can give false readings in complex samples
Solid State (Metal Oxide) Semiconductor	High-stability analyte gas analysis [46]	Thermal energy allows the reaction analysis of otherwise stable (undetectable) compounds	Only usable where a constant large power supply is available
Amperometric	Environmental analysis in remote locations	It has a high selectivity whilst being small and compact, which allows for portable usage [43]	Lower precision than other technologies and a smaller limit of detection range

).

1.3. Statement of Objectives

The objective of this work was to construct a sensor capable of detecting primary alcohol vapours and use various electrochemical methods and mathematical modelling to determine the characteristics governing the internal mechanics of the sensor.

2. Experiment Methods

All analytes (methanol, ethanol, heptan-1-ol, ethyl ethanoate, and ethanoic acid) were purchased from ThermoFisher Scientific (99% purity, analytical grade) and shipped from Germany. All experiments were conducted at 20 °C at a standard 1 atm pressure. The experiments involved passing a gas (air/alcohol analyte mixture) over the sensors and utilising electrochemical techniques such as cyclic voltammetry (CV), chronoamperometry (CA), and impedance spectroscopy (EIS) as analytical tools to probe the mass transport of the analyte through the sensors. This requires tabulated values such as the vapour pressure of simple aliphatic alcohols [50,51]. The sensors were commercially available gas sensors from Alphasense Ltd. These are designated as CO-AF sensors.

2.1. CO-AF Sensor

The CO-AF sensor (from Alphasense Ltd.) is a small, lightweight amperometric gas sensor for carbon monoxide gas (CO) that is designed to be used as part of an integrated PCB (printed circuit board) configuration. It detects CO via electrochemical reactions with its Pt nanoparticle working electrode (WE) and is typically used with the chronoamperometry setting as it can provide quantitative current responses to different CO concentrations between 0.1 and 1000 ppm. They typically come encased in plastic and with a three-electrode configuration. As standard, the construction of the CO-AF sensor involves layering silica impregnated with KMnO₄ and a semi-permeable membrane (typically Nafion, Porex, or Millipore) above the pool of 5 M concentrated sulfuric acid (acting as the electrolyte), which contains platinum (Pt) particles pressed together to form the working electrode (WE). It is designed in a vertical stack configuration to maximise the uniform accessibility of the Pt nanoparticle disc-shaped WE. A stack configuration schematic of the CO-AF sensor is shown for clarity (Figure 1a). There is, in fact, a platinum counter electrode (CE) and a platinum reference electrode (RE) within the electrolyte reservoir beneath the working electrode (WE); The working, counter, and reference electrodes are layers of Pt nanoparticles arranged in a stack and electrically isolated from each other by glass fibre. The CO-AF sensor was the basis for the idea of the sensor used in this study. The CO-AF sensor has been available for years and is already a commercially successful sensor for Alphasense Ltd.

It was theorised that a relatively minor change to the design of the CO-AF sensor would allow for the detection of an important class of VOC, namely primary alcohols. This would enable greater commercial success for the company via a simple and cheap change in the manufacturing process. The layer of KMnO₄ acts as an oxidising filter layer at the top of the stack, which is designed to remove certain reducing gases (e.g., H₂S) that can poison the electrocatalysts and interfere with the response of the device to the analyte [52,53]. Normally (within the CO-AF sensor), this filter destroys H₂S via oxidation using KMnO₄/silica powder. However, this is not suitable for an alcohol sensor because the MnO₄⁻ oxidant will react with the primary alcohol analyte before the analyte reaches the Pt working electrode. Thus, it was removed during the assembly of a batch of CO-AF sensors, subsequently designated CO-A1 sensors (Figure 1b). It must be noted that the potassium permanganate powder was not truly removed; rather, a silica support structure that was not impregnated with the potassium permanganate powder was used in the construction of the CO-A1 sensor. The use of the term “removed” is simply used to help differentiate between the standard CO-AF sensor and the modified CO-A1 sensor used in this investigation. More information as to the typical operation of the CO-AF sensor can be found online, including its operating specifications and detection limitations [54].

The use of scanning electron microscopy (SEM) can help to show the tightly packed nature and regular spacing arrangement of the metal nanoparticles on the working electrode surface of the sensor (Figure 2a). For all SEM measurements, the instrument model Jeol JSM-5610LV was used. Figure 2a has the scale shown at the bottom for reference. The darker areas indicate that the surface is not uniform and is indeed microscopically rough. This helps to explain the large calculated capacitance of the sensor. The Pt working electrode and the PTFE (polytetrafluoroethylene) semi-permeable membrane are shown to be in good contact with each other (Figure 2b), and so the interface can be considered uniformly accessible for the sake of the diffusion coefficient calculations. Even though the measured Pt thickness is shown to be non-uniform (Figure 2b), the difference of only 0.8 µm is considered small enough to be functionally ignored.

2.2. Test Rig Setup

The sensor was housed within a plastic hood and setup as shown below (Figure 3). The plastic hood simply serves to ensure the vapour is delivered directly to the sensor and the vapour concentration does not dissipate as it would in open air. The hood has an internal volume of 15.6 cm³ when empty. Commercially supplied zero-grade air (supplied by BOC Industrial Gases Ltd., Woking, UK) was used as the carrier gas to bubble through the Dreschel bottle containing the liquid analyte. The air containing the analyte vapour was then blown onto the top of the sensor from the top of the plastic hood. Direct mass flow controllers (DMFCs) using the Brooks 0260 Smart Interface software (model 5850S) were used to regulate the gas flows via a laptop (Windows 7). The DMFCs were calibrated to zero before each test. Zero-grade air is synthetic air that has been filtered so that it contains only nitrogen and oxygen and minor trace impurities (BOC product code: 270020-V; CAS number: 132259-10-0). Analyte data was taken from the literature [55], derived from experiemntal data [56,57]. All EIS data were recorded as a three-electrode system using a PalmSens 3 potentiostat and PSTrace 5.7 software, and the raw data were subsequently imported to Microsoft Excel for analysis. All potentials are vs. Pt/Pt⁺ unless otherwise stated. All scan rates and potential ranges were standardised via the PalmSens3 potentiostat. The potentiostat was calibrated by the manufacturer one month before being used in the experiments. Thus, the potentiostat was considered to still be within its calibration window. The sensor did not undergo any preconditioning prior to testing. The manufacturer (Alphasense) recommends operating voltages between −0.1 and +0.3 V, with a nominal sensor performance at 0.0 V (for the commercial CO-AF sensor). This helps to avoid any changes in the electrode surface that can occur at high cycling voltages [58,59,60,61]. For the EIS measurements, the primary alcohols were introduced to the sensor individually, with 24 h between each alcohol vapour experiment. The EIS measurements were conducted from 10 Hz to 0.1 Hz. Exposure times are for the duration of the experiment (e.g., EIS) unless otherwise shown on the graph (e.g., chronoamperometry). The chronoamperometry experiments were conducted at a 50 mL min⁻¹ flow rate and at 0.0 V dc as per the manufacturer’s recommendation. The tests of chronoamperometric response to different functional groups were also conducted at a 50 mL min⁻¹ flow rate at 0.0 V dc. All other experiments were conducted at a 50 mL min⁻¹ flow rate unless otherwise stated.

By changing the gas flow through the Dreschel bottle at a known concentration and flow rate and combining it with a known constant gas flow rate from the zero-air cylinder, the partial pressure of alcohol reaching the sensor can be calculated (Equation (2)). The partial pressure of the analyte vapour (P_t) is calculated from the standard vapour pressure of the gas at the temperature of the Dreschel bottle measured via a thermocouple (P), the volume flow rate of analyte-saturated gas (V_g), and the volume flow rate of zero air (V_a).

$$P_t=P{\displaystyle \frac{Vg}{Vg+Va}}$$

(2)

• P_t = partial pressure (kPa);
• V_g = volume flow rate of analyte (mL/min);
• P = standard vapour pressure (kPa);
• V_a = volume flow rate of zero air (mL/min).

During the experiments, the concentration of the alcohol was determined (Equation (2)). This required a knowledge of the saturated vapour pressure at the temperature of the Dreschel bottle. The vapour pressures were found using Antoine equation parameters (Equation (3)) provided by the NIST webbook website [55], which are determined by fitting the experimental data [56,57]. A list of the primary alcohols used and their corresponding Antoine parameters is shown (

Table 2. Vapour pressures and Antoine parameters derived from the literature [55].

VOC	Vapour Pressure at 20 °C (kPa)	Concentration at 50 mL min−1 Constant Flow Rate (×10−3 mol dm−3)	A	B	C
methanol	13.3 ± 0.3	5	5.2	1581	−33.5
ethanol	5.9 ± 0.1	2	5.4	1670	−40.2
heptan-1-ol	0.02 ± 0.01	0.8	4.0	1257	−133.5

). The flow rate is 50 ± 1 mL min⁻¹.

$$logP=A-{\displaystyle \frac{B}{C+T}}$$

(3)

• P = vapour pressure (kPa);
• A, B, C = constants derived from experimental data;
• T = temperature (K).

3. Results and Discussion

3.1. Impedance Spectroscopy

The sensor was exposed to the corresponding analytes before and for the whole duration of the EIS testing. The Nyquist plots ranged from 10 Hz to 0.1 Hz, which took approximately 3 min per potential (−0.1 to +0.2 V). Consequently, four datasets are shown in each graph (twelve in total). It can be seen from the EIS data (Figure 4) that the CO-A1 sensor shows a response that is dominated by a charging current. Only at low frequencies (f < 1 Hz) did the Nyquist plots for all the alcohols show a significant dependence on the applied dc voltage, which represents the influence of faradaic processes. At high frequencies, the data indicate an ohmic resistance of about 1 ohm, although it is slightly larger for the heptan-1-ol, which may indicate a change in the composition of the three-phase interface (Pt/vapour/electrolyte) in the presence of this species. This suggests low internal resistance within the part of the sensor that acts as an electrochemical cell (comprising the sulfuric acid electrolyte and the Pt working electrode). The applied dc voltage has little effect on the behaviour of the devices at high frequencies (f > 1 Hz), which indicates that potentiodynamic measurements cannot provide information on the oxidation of the analytes under these conditions.

3.2. Differential Capacitance of the Sensors

The differential capacitance (C) of the CO-A1 sensor in the presence of each alcohol at a 0.0 V steady-state dc was determined from the low-frequency data (f < 0.5 Hz) (

Table 3. Calculated capacitances for CO-A1 sensor exposed to alcohols at 0.0 V steady-state.

Alcohol	Calculated Capacitance (±0.02 F)
methanol	0.12
ethanol	0.14
heptan-1-ol	0.15

). This is based on data from the EIS graphs (Figure 4).

It is interesting to note that the calculated capacitances are all greater than 0.1 Farads. Based upon this estimate of the double-layer capacitance, even the capacitance per unit of the geometric area is on the order of 0.1/πr² = 1.3 mF cm⁻² (assuming a circular disc sensing diameter of 10 mm [54]). This capacitance value is much greater than the value for smooth Pt(111) electrodes in aqueous media (around 20 µF cm⁻² [52]). The actual surface area of the nanoparticulate working electrode is, therefore, greater than the geometric area by a factor of 1.3/0.02 = 65. Clearly, the electrolyte penetrates the electrocatalyst and wets a much larger area than the geometric area of the device. The measured differential capacitances in the presence of alcohol are, however, smaller than in the absence of alcohol (where we observed a value of 0.257 F [62]). This suggests that a significant fraction of the liquid at the nanoparticle/electrolyte interface comprises alcohol.

3.3. Chronoamperometry Response to a Concentration Step

Owing to the large differential capacitance, potentiostatic measurements provide clearer information than voltammetric techniques. The chronoamperometry data in response to a step change in the analyte concentration (Figure 5) shows that the sensor can detect all three alcohol analytes when exposed to them individually. The chronoamperometric experiments were performed at 0.0 V as per the manufacturer’s recommendations for the commercial CO-AF sensor. Note that at low analyte concentrations, the signal-to-noise ratio is quite high and so gives rise to small discrepancies in the chronoamperometric response of the sensor. The fact that they also show a decrease in the current after the analyte pulse has been turned off would normally suggest that the electrooxidation of the analytes at the working electrode is extremely fast. However, since there is a steady, exponentially curved decrease in the current, it is likely that the slow evaporation of the analyte from the electrode’s surface through the semi-permeable membrane before subsequent reactions can occur. This suggests that the membrane pores can become blocked by the alcohols and that at room temperature, a complete electrooxidation reaction is unlikely to occur at the time scales investigated here.

It should be noted that the unmodified CO-AF sensor does not give a chronoamperometric response to alcohol vapour (due to the permanganate powder reacting with the alcohols), only carbon monoxide gas. The chronoamperometry response (Figure 6) shows no discernible response from the sensor to the alcohol vapour. Thus, the key difference between the commercial CO-AF sensor and the CO-A1 sensor is that the CO-A1 sensor can give an electrochemical response to alcohol vapours.

3.4. Modelling the Chronoamperometric Response

Under the assumption of mass-transport-limited currents, the chronoamperometry data (Figure 7 can be analysed to estimate the diffusion coefficient (D) and effective diffusion layer thickness (L) of the device using a mathematical model developed in previous work on CO-AF amperometric gas sensors [62]. Such a model is reasonable in view of the use of diffusion models for VOC transport in porous materials [63,64]. The model fits the rising portion of the data in response to switching from an air to an air/alcohol mixture (Figure 5).

The model is based on a solution to the diffusion equation for a concentration step at a hypothetical single-layer device. We have previously shown that such an effective medium approach applies to the case of more realistic multi-layered devices and with certain kinds of kinetic limitations as long as the parameters L and D are interpreted as the thickness and analyte diffusion coefficient of an effective medium and a fuller explanation of the modelling is available in previous work [62]. In order to aid the reader, a brief description is given here. The background theory used in modelling software is based on the assumption that the response of the sensor is assumed to be governed by one of three steps: the diffusion of the analyte through a silica support scaffold without the standard potassium permanganate powder impregnated in it (i); the diffusion of the analyte through the semi-permeable membrane (ii); and the diffusion of the analyte through the electrolyte to the working electrode surface (iii) (Figure 8).

The analyte is assumed to be in the gas phase during stages (i) and (ii), whilst the analyte is taken to be in the liquid phase during stage (iii). It is assumed that any reaction at the electrode surface is not affected by electrode kinetics (such as electron transfer) at the potential used during chronoamperometry; this is because the electron transfer kinetics are so fast compared to the mass transport through the sensor that they are considered functionally negligible in terms of the calculations and sensor behaviour. The potential remains fixed throughout the chronoamperometry experiments. Note that this means the Butler–Volmer equation can be functionally ignored. This is explained further in previous work [62]. Partition equilibria between the different phases are ignored for the sake of simplicity. Firstly, the response of the sensor under a steady-state current is considered. When the model is assumed to be simple and to have no kinetic barriers for the analyte (or at least so small relative to the other diffusion coefficients so as to be negligible) during the changes in phase within the sensor, a graph of concentration (C) can be plotted against the distance (x) from the working electrode and takes the form of a series of straight lines as the diffusion coefficient changes due to being in different phase media (Figure 9). Where the maximum initial concentration (C*) is outside of the sensor, the diffusion coefficient (D) and layer thickness (L) are numbered sequentially according to the distance from the working electrode.

However, if there are, in fact, significant kinetic barriers to the analyte between the different phases (gas, membranes, and liquid) within the sensor, then a slightly different concentration profile forms. Whilst the fact that it is assumed that current and potential are independent of each other means that an electron transfer kinetic barrier at the electrode surface can be ruled out, interfacial kinetic barriers (k) at other interfaces that do not respond to the potential remain a theoretical possibility. In the mathematical analysis of the model, the slopes shown in Figure 9 and Figure 10 can be calculated using the gradient of a straight line (m = Δc/Δx). The flux (j) entering one region must be the same as the flux leaving the previous one if the local concentration is at a steady state (time-independent). Indeed, the flux (j) is constant from the external surface of the sensor, where the bulk concentration (C*) is applied to the electrode surface (C = 0) under steady-state conditions (Figure 10). Thus, the limiting fluxes (j) for each layer are simply added reciprocally, and the result for any number of barriers (n) can be calculated. This is logically understood as the overall flux being limited by the slowest step in the sequence.

One can then interpret the estimated parameter values in terms of the known structure of the device. For example, if L is similar to the physical length of the filter layer (approximately 16 mm [54]), then this suggests that the rate-limiting step is diffused through this layer. Equally, D values on the order of 10⁻³ cm² s⁻¹ or greater are consistent with gas phase diffusion (possibly within the pores of the layer) [65,66,67,68]. However, if they are any on the order of 10⁻⁵ cm² s⁻¹ (which are typical values of VOCs in aqueous acids) [69,70,71,72], then the diffusion of the VOCs in the liquid phase through the aqueous sulfuric acid electrolyte is likely to be the rate-limiting step.

It can be seen (

Table 4. The calculated values for the apparent layer thickness (L) and diffusion coefficient (D) of the VOC analytes at a 0.0 V steady-state dc.

VOC	L (±0.1 µm)	D (±1 × 10−12 cm2 s−1)	L2/D (±0.2 × 106 s)
methanol	0.3	3	3.3
ethanol	0.1	1	3.1
heptan-1-ol	0.2	4	0.6

) that the apparent layer thickness (L) is on the order of 0.1 µm, and the effective diffusion coefficient (D) is on the order of 10⁻¹² cm² s⁻¹. At first glance, this suggests that the mass transport of the alcohols through the sensor stack is determined (rate limited) by the diffusion of the analyte through a thin layer of the aqueous sulfuric acid electrolyte that covers the Pt nanoparticle working electrode. However, the interpretation cannot simply be that diffusion occurs in a thin aqueous phase because the calculated values of D are extremely low, much lower than the typical 10⁻⁵ cm² s⁻¹ needed for small molecule diffusion coefficients in aqueous solutions [69,70,71,72]).

Modelling Limitations

But the D values are also much smaller than the values in the literature for alcohol diffusion rates through porous membranes [73,74]. These small values might be a result of an electrode kinetic barrier because the regression model will still fit the data [64]. This is one of the limitations to ignoring partition equilibria and electrode kinetics; whilst it is suitable for simple molecules such as carbon monoxide (CO) [62], it may not hold true for more complex molecules such as primary aliphatic alcohols. However, slow kinetics would reduce the steady-state current and lead to an underestimation of the mass transfer coefficient (D/L) and, therefore, the value of D would be underestimated. This also cannot be the complete explanation because the D values are many orders of magnitude lower than expected, and a purely electrode-kinetic limitation would mean that the current is suppressed by a similar factor. Instead, it is suggested that the membrane pores blocked by the condensation of the vapour during transport through the sensor are involved. The vapour condenses in the porous membrane and so trickles down the sensor as a liquid under gravity comparatively slowly. In support of this hypothesis, it can be noted that the decline of the signal upon switching the gas stream to air (with the gas off, as shown in Figure 5) is slower than the rise in the signal with the gas on. This is not expected for a model based on an effective medium approach for diffusional behaviour, for which chronoamperometric traces with both gas-on and gas-off ought to have the same rise and decay times. Instead, we propose that the off transient is controlled by the slow evaporation of the alcohol from the device.

3.5. Response to Different Functional Groups

In order to investigate the selectivity of the sensor’s response to the hydroxyl functional group, the sensor was exposed to ethyl ethanoate and ethanoic acid vapours. These compounds were chosen due to being the simplest molecules with functional groups analogous to primary alcohols and containing the -OH hydroxyl moiety. The sensor was exposed to ethyl ethanoate and then methanol-saturated air (Figure 11a). Similarly, the sensor was exposed to pulsed ethanoic acid-saturated air and then pulsed heptan-1-ol-saturated air (Figure 11b). Each analyte had a 50 mL min⁻¹ flow rate. The sensor displays selectivity towards different functional groups in so far as it does not give a measurable amperometric response to the compounds with ester and carboxylic acid functional groups. This is most likely because the sensor is at room temperature, and so there is not enough thermal energy to overcome the steric hindrance and inductive effects within the carboxylic acid and ester functional groups. It is also possible that the voltage range used within the experiments is insufficient to overcome these effects. The fact that the sensor can give an amperometric response to the primary alcohols even after being exposed to the other functional group compounds suggests that the compounds did not react with the Pt nanoparticle electrode to form a passivation layer. It is more likely that the compounds are dissolved within the sulfuric acid electrolyte and so do not interfere with the working of the sensor. This means that the sensor is usable in environments even where the exact type of VOC functional group is not known or if there is a mixture of VOCs present in the atmospheric environment (even though it cannot differentiate between different primary alcohols).

3.6. Quantitative Analysis

The sensor can provide quantitative data via amperometric data. Thus, known concentrations of the primary alcohols as an aerosol in dry air (zero air) were passed over the sensor. The data for heptan-1-ol are shown for clarity (Figure 12). The data show good repeatability and reproducibility due to the signal current going back to zero after each analyte pulse. For the calibration curves, the amperometric current response increases with the increasing analyte concentration (Figure 13), and the sensor is capable of detecting ppm levels of primary alcohols. The sensor has a sensitivity of approximately 2.2 µA/ppm to primary alcohol-saturated air (the largest graph gradient plus signal noise). The signal noise for the sensor has a current of 1 µA. Whilst the sensor has a theoretical limit of detection (LoD) of 1.86 (±0.02) ppm (the concentration required for the current to be 3× above signal noise), the calculated LoD for each alcohol was methanol 17 (±1) ppm, ethanol 2 (±1) ppm, and heptan-1-ol 3 (±1) ppm.

4. Conclusions

It has been demonstrated that the modified sensor (CO-A1) is able to detect a selection of aliphatic primary alcohols in the air. The calculated capacitances for the sensor help to give an insight into the structure of the Pt working electrode at the microscopic scale via non-destructive means and indicate that (i) the alcohol penetrates the metal/liquid interface and (ii) the differential capacitance of the devices remains extremely large. Standard potentiodynamic methods of electrochemistry are, therefore, not suitable for probing the mechanism of sensing because the faradaic current is dominated by the charging current. However, the chronoamperometric response to a concentration step at a fixed potential provides insights. The condensation of the alcohols inside the porous membranes leads to long and very low apparent mass transport rates due to the blocking of the pores and their slow recovery when the gas stream is switched from the analyte to blank air. The CO-A1 sensor can detect ethanol vapour with a response time of about 30 to 60 s, exhibits a level of selectivity towards different functional groups on aliphatic hydrocarbons (Figure 11a,b), and can provide quantitative data on the primary alcohol analyte concentration (Figure 12 and Figure 13). The sensor is usable in leak detection and health monitoring in industrial settings, where the exact species of the analyte is not required to be known (or is already known through other means). The fact that it can differentiate between different functional groups means it can also be used as a quality control monitor during chemical synthesis. Even though the sensor is not selective and has a limited detection range, it has the advantages of being lightweight and easily portable. Therefore, it is usable as part of an early warning system or in remote locations to let the operator know further investigations are required.