Analytical and Bioanalytical Chemistry

, Volume 410, Issue 19, pp 4587–4596 | Cite as

Toward continuous amperometric gas sensing in ionic liquids: rationalization of signal drift nature and calibration methods

  • Lu Lin
  • Xiangqun Zeng
Paper in Forefront
Part of the following topical collections:
  1. Ionic Liquids as Tunable Materials in (Bio)Analytical Chemistry


Sensor signal drift is the key issue for the reliability of continuous gas sensors. In this paper, we characterized the sensing signal drift of an amperometric ionic liquid (IL)-based oxygen sensor to identify the key chemical parameters that contribute to the signal drift. The signal drifts due to the sensing reactions of the analyte oxygen at the electrode/electrolyte interface at a fixed potential and the mass transport of the reactant and product at the electrode/electrolyte interface were systematically studied. Results show that the analyte concentration variation and the platinum electrode surface activity are major factors contributing to sensing signal drift. An amperometric method with a double potential step incorporating a conditioning step was tested and demonstrated to be useful in reducing the sensing signal drift and extending the sensor operation lifetime. Also, a mathematic method was tested to calibrate the baseline drift and sensing signal sensitivity change for continuous sensing. This study provides the understanding of the chemical processes that contribute to the IL electrochemical gas (IL-EG) sensor signal stability and demonstrates some effective strategies for signal drift calibration that can increase the reliability of the continuous amperometric sensing.

Graphical Abstract


Gas sensors Ionic liquid Continuous sensing Signal drift Calibration 


Detection and quantification of gases are important in many applications, particularly in workplace safety and environmental health [1]. The concentrations of gas analytes vary in both time and space depending on the time of day as well as locations relative to the contaminants and/or in environments where hazardous gaseous substances are not a factor in daily activities (e.g., offices and on the streets). Thus, measuring the gaseous analytes in ambient conditions requires gas sensors that can measure the concentrations of the gaseous analytes in real-time or continuously throughout some defined time period. Although many hand-held portable gas sensors have been developed, none meet the challenging set of cost/utility/capability requirements that prevent the pervasive use of personal, continuous-use gas monitors in many real-world applications. We have demonstrated a room temperature ionic liquid (RTIL) electrochemical gas sensor technology that could potentially solve many prevailing problems with existing gas sensor technologies [2, 3, 4, 5, 6, 7, 8, 9]. We have shown that the accuracy, selectivity, and detection limits of IL-based electrochemical gas (IL-EG) sensors are suitable for the real-time monitoring of key gases in work safety applications (e.g., O2, CH4, H2). One of the major challenges of chemical sensors for real-time sensing applications is sensor stability and/or sensor drift [10]. As yet, it has not been possible to fabricate drift-free chemical sensors for long-term continuous sensing [11, 12, 13, 14]. Sensor signal drift is a very fundamental problem for real-time and continuous sensing in a dynamic gaseous environment. The prevailing IL-EG sensors also exhibit signal drift during long-term sensing in real-world gas mixtures [3, 15, 16].

Typically, there are two main signal drift sources for a general chemical sensor [10, 17, 18]. The first source is due to the chemical and physical interaction processes of the chemical analytes, occurring at the sensing film microstructure, such as aging (e.g., the reorganization of the sensor surface over long periods of time) and poisoning (e.g., irreversible binding due to external contamination) [10]. The second source is due to measurement system drift, produced by the external and uncontrollable alterations of the experimental operating system, including but not limited to: changes in the environment (e.g., temperature and humidity variations), measurement delivery system noise (e.g., tubes condensation, sample conditioning, etc.), and thermal and memory effects (e.g., hysteresis or remnants of previous gases) [17]. The second source of signal drift can often be eliminated with engineering approaches (e.g., packing of the sensor with a water-resistant filter) and calibration methods. In this work, we focus on understanding the nature of the first source of signal drift in the IL-EG sensor as it is directly related to the fundamental chemical sensing processes and mechanisms. Owing to the chemical and thermal stability of the IL sensing material, the aging of the sensing films can be minimized or eliminated. Thus, signal drift of the IL-EG sensor predominately results from the analyte mass transport at the IL/electrode interface and the interface sensing reactions under an applied potential. We selected an IL-EG oxygen sensor as our model system because the sensing mechanism of the oxygen sensor is based on the oxygen reduction [3]. It has been shown that the directional polarizability of ILs facilitates the formation of an ordered interface structure, which has been used in making a proton-conducting gelatinous electrolyte and nanostructure anatase TiO2 monoliths [19, 20]. This unique IL/electrode interface structure and the high viscosity of an IL could result in the accumulation of reaction by-product leading to baseline drift. The by-product can also be adsorbed at the electrode surface, which can result in the change of the sensing sensitivity. We systematically studied the nature of the signal drift focusing on analyte mass transport at IL/electrode interface and the electrode reactivity, and demonstrated the advantage of applying a double potential step chronoamperometry to minimize the signal drift using an oxygen IL-EG sensor that our lab previously reported as an example [3]. The understanding of the drift nature of the amperometric sensing signal of the IL-EG sensor and the calibration methods developed here serve as an important step toward the development of IL-EG sensors for reliable real-time and continuous monitoring of the gas analytes in real-world applications.


Chemicals and instruments

IL 1-Butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)-imide ([Bmpy][NTf2], Io-LiTec Inc. 99%) was used in all electrochemical measurements. Both pyrrolidinium-based cation and NTf2 anion of [Bmpy][NTf2] have relatively high electrochemical stability [4, 21] and were also used in our early IL-EG oxygen sensor development by Wang et al. [3]. Fifty μL of the IL was added to the electrochemical cell. The electrochemical cell structure is shown in Fig. S1 in the Electronic Supplementary Material (ESM). Based on the dimensions of the electrochemical cell, the IL thickness on the working electrode was ~637 μm. The electrochemical cell was purged with continuous nitrogen (PRAXAIR compressed nitrogen) via Tygon PVC gas tubing (ϕ = 1 mm) for a minimum of 12 h until no visible infrared peak from water was observed by FTIR transmission mode [22]. All cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS), and chronoamperometry measurements were performed using AMETEK VersaSTAT MC 4-channel electrochemical workstation.

Measurement set-up

The electrochemical sensor structure was similar to a Clark-type cell [16]. As shown in the ESM (Fig. S1), it consisted of a porous gas permeable membrane (GPM) membrane (Interstate Specialty Products, type PM71W, with a pore size of 5 μm, a thickness of 0.15 ± 0.01 mm, and 35% porosity), a polycrystalline platinum gauze working electrode (Sigma Aldrich, 100 mesh, with a diameter of 0.076 mm), a reference electrode and a counter electrode (both are polycrystalline platinum wires, Sigma Aldrich, with a diameter of 0.5 mm). All electrochemical cell components were stacked in a layered structure. Cellulose filter paper infused with the IL (Whatman, GF/A) was inserted between electrodes to avoid short circuit. The electrochemical cell body was made with a very inert Kel-F material. The electrochemical active surface area of the working electrode was 0.83 cm2, measured via 0.01 M ferro-/ferricyanide redox probe (in KCl solution, 1 M) at different scan rates and calculated according to the Randles-Sevcik equation (Eq. 1):
$$ {i}_p=268600{n}^{3/2}A{D}^{1/2}C{v}^{1/2} $$
where ip is the redox peak current (A), n is the stoichiometric number of electrons, A is the electroactive area (cm2), DO is the diffusion coefficient (0.726*10-5 cm2/s) [23], C is the ferro-ferricyanide concentration (M), and v is the scan rate (V/s).

Our experimental set-up and protocol are designed to allow for the investigation of the signal drift source from (electro)-chemical reactions at the electrolyte/electrode interface rather than environmental variables such as humidity and temperature. All measurements in this work were performed in a lab condition with a constant temperature (25 °C) and humidity (RH=30%). The analyte oxygen gas and the background gas are dry. The outlet of the gas flow system was connected to a vacuum hood to avoid the buildup of gas in the testing system. The gas flow was controlled using MKS (MKS Instruments, Inc.) type 247 4-channel readout (bundled with Mass-Flo Controller). Two mass-flow controllers were used to adjust the volume ratio to reach the target gas concentration. One of the mass flow controllers was used to control the background gas (nitrogen) flow and the second was used to control the analyte gas (oxygen) flow. Oxygen UHP cylinder (Specialty Gases of America, 99.993%) was used as the source of oxygen. The gases were purged into the electrochemical sensor directly via a GPM.

Results and discussion

The principle of amperometric sensing oxygen in the IL

An amperometric gas sensor is operated under an applied external voltage, which drives the electrode reaction in one direction. The current, due to the redox reaction of a target gas at the working electrode, is measured as the sensing signal, which can be measured at fixed or variable electrode potentials. The electroactive analyte that participates in the electrochemical reaction diffuses from the gas phase into the electrochemical cell through the porous layers (e.g., GPM) and dissolves in the electrolyte through which it proceeds to further diffuse to the working electrode surface. In this work, the target gas is oxygen and the background gas is nitrogen. The current of the redox reaction is limited by the rate of the reaction that occurs at the surface or the rate of gas diffusion toward the electrode surface. If the electrode has high reactivity and the electrode reaction is not the rate limiting process, the current is proportional to the concentration of the analyte in the surrounding gas. Wang et al. [3] demonstrated a methodology of an IL-EG oxygen sensor based on the oxygen reduction reaction. Figure S2 in the ESM presents the cycle voltammogram of oxygen redox reactions in [Bmpy][NTf2]. From the voltammogram of the oxygen redox process, a constant cathodic potential –1.2 V versus Pt was selected as the potential for the amperometric sensor method. Scheme 1 shows the processes of oxygen sensing. According to Buzzeo and colleagues [24] and Xiao et al. [25], at –1.2 V versus Pt, oxygen can be reduced to form the superoxide anion O2- (Eq. 2). The electrochemically generated O2- species has a greater stability in the aprotic IL versus that in aqueous solvents, protic organic, or other IL solvents [25]. The O2- can be oxidized back to O2 (Eq. 2) at an anodic potential.
$$ {\mathrm{O}}_2+{\mathrm{e}}^{-}\leftrightarrow {\mathrm{O}}_2{\cdot}^{-} $$
Scheme 1

An illustration of oxygen sensing involving the redox reaction of the analyte, surface adsorption and desorption of both analyte and product, mass transport of both the analyte and product, and the ion-pairing of product superoxide radical with the IL cation

We are aware that both temperature and humidity changes will result in signal drifts. The variation of temperature will affect the solubility of the gas analyte in the ILs. This can lead to signal drift as the sensing signal correlates the concentration of the target gas in the IL in our IL-EG sensor. The trace water in the IL electrolyte can be a proton source. The presence of trace water can lead to signal drift since it can change the sensing reactions. As detailed in our Experimental section, all experiments were done at lab condition with constant temperature. IL [Bmpy][NTf2] is hydrophobic with little trace water and gases (N2 and O2) used are all dry. This allows us to systematically study the effects of the interfacial reactions and processes on the analytical performance of the amperometric sensor. If the products of sensing reaction are sensor poisons (e.g., change the IL/electrode microenvironment and/or electrode poison), the sensor lifetime or response characteristics will be severely limited and will exhibit a signal drift with time. Our early work shows that superoxide generated from sensing reactions at the electrode interface affects the signal stability for the continuous amperometric sensor [16]. As the superoxide radical O2- is negatively charged, it could interact with the cations of the IL (i.e., [Bmpy]+ in this case) and form ion-pairs [5]. Superoxide radical O2- and/or the ion-pair complex [Bmpy]+[O2-] could adsorb and/or accumulate at the electrode interface, which would hinder the continuous sensing of oxygen [5]. In the following section, we systematically studied the sensing reactions and how they impact the signal stability. The mechanisms of the sensing signal drift allow for the testing of several strategies to calibrate and improve the sensing accuracy and precision using the IL amperometric oxygen sensor as a model system.

Sensing signal drift causes

Following our early IL-EG oxygen sensor method [3], the amperometric oxygen sensor was set at –1.2 V, in which oxygen was reduced to the superoxide radical that resulted in a reductive current proportional to the oxygen concentration. In a nonaqueous electrolyte, a quasi-reference electrode is commonly used as the reference electrode. If the absolute potential of the quasi-reference electrode is not stable, the sensor signal will drift since the stepped potential will not be the same. We studied the potential stability of the quasi-reference electrode by using the Fc+/Fc redox probe. As shown in ESM Fig. S3, the quasi-reference electrode is stable within our measurement time period. The uncertainty of the concentration of gas samples could also prompt the signal drift in addition to the sensing reactions. A gas sampling system using mass flow controllers was employed in our study to continuously and reproducibly provide controlled known concentrations of the gas sample. A systematic investigation of gas sampling flow rates (ESM Fig. S4) suggested the optimum flow rate was 100 sccm, which minimized IL/electrode interface perturbation while providing quicker mass transport of the gas molecules from the gas phase to the sensing electrode interface. Thus, 100 sccm was the gas sampling flow rate setup in the rest of this work.

Sensing signal drift cause: oxygen concentration change

A demonstration of our IL-EG oxygen sensor for 2-h continuous oxygen sensing (ESM Fig. S5) showed a signal drift of –0.93% at a constant 21% oxygen exposure in the measurement period of 3600–11,600 s. However, in real-world gas monitoring applications, a sensor should be able to precisely respond to the change of the target gas as well as exhibit its capability to return and be consistent with the baseline of the previous exposure [26]. Thus, one measurement (Fig. 1a) was conducted and it was observed that the sensing signal (after the removal of 21% oxygen) could not return to the same value of the baseline before the introduction of oxygen gas. This measurement started in an inert nitrogen condition and 21% oxygen was introduced in the time scale of 350–1100 s, and the oxygen was removed after the 1100th s. Following the concept of Wenzel et al. [27] who stated that the the sensing response is the difference between the limiting current by baseline current, the sensing response of 21% oxygen exposure in this case was –134.91 μA, with a relative standard deviation (RSD) of ±0.61%. The T90 value (the time needed to approach 90% of the sensing signal) was 108.4 s. Prior to oxygen introduction, the baseline current was –19.10 μA. After oxygen was removed for 900 s (i.e., time scale 1100–2000 s), the new baseline current was –59.34 μA, which was not the same as orignial baseline current (–19.10 μA). This showed that there was an obvious delay in the sensing signal’s return to the prior baseline after the removal of the analyte. Owing to the ion-pairing effect and the small diffusion coefficient of oxygen in [Bmpy][NTf2] (D0 = 1.1*10-10 m2/s) [5, 24], oxygen residue remained in the IL after the oxygen gas supply was paused, giving a faradic current-time decay controlled by diffusion. We thus ascribed this observation to the slow mass transport of both the reactant (oxygen) and the product (O2- species) when there is a concentration gradient between the IL/electrode interface and the bulk electrolyte, and therefore hypothesized that a smaller analyte concentration change would mitigate this delay and lead to less signficant signal drift.
Fig. 1

(a) Constant potential chronoamperometric recording of the sensing response toward 21% oxygen exposure in time scale of 350–1150 s (green dash rectangular mark the time period when the IL-EG responded to the ±0.08% oxygen concentration variation, see ESM Fig. S6). (b) Constant potential chronoamperometric recording of the sensing response toward the oxygen concentration change of ±1% when the sensing signal was stabilized at 5% [O2%]

Therein, in a separate measurement (Fig. 1b), a smaller oxygen concentration variation (±1%) was conducted when the sensing signal was stabilized at constant 5% oxygen concentration. Results indicated that the sensing signal was able to return to the value of 5% oxygen baseline, either by increasing [O2%] from 4% to 5% or decreasing [O2%] from 6% to 5%, which supported our hypothesis that smaller anlayte concentration changes exhibit better sensing signal stability. In this measurement, the sensor was exposed to a step-wise oxygen concentration change of ±1% and then returned back to the baseline oxygen concentration of 5%. This experiment mimics the real-world dynamic oxygen concentration change. The sensing responses of ±1% [O2%] variation from a-b, b-a’, a’-c, and c-a’ were 7.64 μA, 7.92 μA, 8.23 μA, and 8.24 μA, respectively. The RSD value was ±0.34%.

As real-world gas monitoring in leak situations may experience an increase with time, a measurement of a stepwise oxygen concentration increment (Fig. 2a) was performed to validate our hypothesis. In this measurement, oxygen was introduced at the 600th s and increased 5% per step (each step was 300 s) from 0% to 20%, followed by a decrease to 0% [O2%] at the 1800th s. The oxygen sensing signal increased with the stepwise oxygen concentration increase. However, the plot of sensing signal as the function of oxygen concentration (inset of Fig. 2a) did not yield a perfect linearity (R = 0.93). In fact, the sensing signal of 5% [O2%] change from a to e exhibited a decreasing trend. For example, the sensing signal of 5% [O2%] change from a to b was –55.2 μA, and that of 5% [O2%] change from b to c dropped to –34.7 μA. This is another observation we ascribed to the slow mass transport of the product. The accumulation of O2- species in the IL/electrode interface would inhibit the forward reaction of Eq. 2 and result in a smaller oxygen reduction signal. In this measurement, the baseline current before the introduction of oxygen was –9.01 μA, and the sensing response of 20% oxygen exposure was –132.48 μA. After the removal of oxygen, the baseline current slowly decreased to –14.9 μA at the end of the measurement (which took nearly 2200 s), suggesting the slow mass transport of the reactant.
Fig. 2

(a) Constant potential chronoamperometric recording of the sensing response toward a stepwise increase of oxygen concentration, followed by a switch to inert nitrogen condition. The oxygen concentration increment was 5% per step. (b) Constant potential chronoamperometric recording of the sensing response toward a stepwise increase of oxygen concentration, followed by a stepwise decrease of oxygen concentration, and continued for four subsequent cycles. The oxygen concentration increment and decrement were both 1.05% per step

Compared with a previous measurement presented in Fig. 1a, after the removal of oxygen for 900 s (i.e., time scale 1800~2700 s), the new baseline current was –29.05 μA. This value was closer to the corresponding original baseline current than the measurement in Fig. 1a, implying a better restoration of the IL/electrode interface double layer initial condition. In both measurements, we picked the periods when oxygen was present (i.e., time scale 350–1100 s of the measurement in Fig. 1a, and time scale 600–1800 s of the measurement in Fig. 2a), integrated the total charge values, and found that IL/electrode interface double layer could better return to its initial condition in meausurements where less amount of O2- species was produced. According to Faraday’s Law (Q = nFN, in which Q is the total charge in the unit of C, n is the number of moles of the analyte, F is the Faraday’s constant, and N is the amount of electron transferred in the redox reaction), the total charge directly reflects the amount of analyte reacted. The total charge of both measurements were 0.1178 C (measurement in Fig. 1a) and 0.1056 C (measurement in Fig. 2a). This result further validated our hypothesis that the measurement in Fig. 2a, having smaller total charge and less amount of O2- species, exhibited less baseline signal drift.

In summary, better sensing signal stability could be approached with smaller [O2%] change, which was also validated by the measurement presented in Fig. 2b. The measurement started recording after the oxygen concentration was stabilized at 15.75%. The oxygen concentration was first increased stepwise to 19.95%, then decreased stepwise back to 15.75%, completing one measurement cycle. This cycle was repeated four times continuously. The oxygen concentration change was 1.05% at each step. The plot of sensing signal as the function of oxygen concentration (ESM Fig. S7) and the summary of the corresponding sensing signal sensitivity and the linearity coefficient (ESM Table S1) indicated that the sensing signals were quite repeatable and there was nearly no sensing signal sensitivity loss over the course of the entire measurement. It is noteworthy that the sensing sensitivy value when [O2%] increased (1.778 ± 0.045 μA/%) was slightly less than that of when [O2%] decreased (1.800 ± 0.034 μA/%), due to the slow diffusion of the oxygen molecules from the bulk electrolyte to the IL/electrode interface. Nevertheless, in this measurement of small [O2%] change, the slow mass transport effect of both the reactant and product was not significantly observed.

Sensing signal drift cause: electrode reactivity

Although small [O2%] changes influence the sensing signal stability slightly, the increased trend of the sensor functioning time to maintain stable sensing signals in Fig. 1b drew the reactivity of the working electrode to our attention. In our study, platinum gauze was used as the working electrode as a platinum electrode generally exhibits excellent stability under polarized potentials that may be corrosive to other metal electrodes. Platinum metals are also excellent catalysts for many analyte reactions. Our group previously reported the formation of the Au-O2-∙ adsorbate in the oxygen reduction process on a gold surface in [Bmpy][NTf2] [25]. Similar interactions may occur on a platinum electrode as it exhibits similar physicochemical properties to gold [e.g., platinum has an electron affinity value of 205 kJ/mol, which is close to that of gold (223 kJ/mol)]. In the IL, the negatively charged superoxide radical O2- could have interactions with the platinum electrode and the cation of the IL, [Bmpy]+ (Eq. 3) when a cathodic potential is applied [28]. These interactions not only influence the ion arrangement in the inner and outer Helmholtz planes near the electrode surface but also impair the reactivity of the platinum surface, as is illustrated in Scheme 1. We hypothesized that these interface interactions are one of the causes of the sensing signal drift of the continuous oxygen sensor.

$$ \mathrm{Pt}+{\mathrm{O}}_2+{\mathrm{e}}^{\hbox{-}}\overset{\left[\mathbf{Bmpy}\right]\left[{\mathbf{NTf}}_{\mathbf{2}}\right]}{\to}\mathrm{Pt}\cdots {\left[\mathrm{Bmpy}\right]}^{+}\cdots {\mathrm{O}}_2{\cdot}^{-} $$
To validate this hypothesis, using a newly assembled IL-EG oxygen sensor, we performed two measurements of this sensor in two different device conditions. The first measurement was conducted on the newly assembled sensor with a fresh IL/electrode interface. The measurement began from an inert nitrogen condition and 5% oxygen was introduced at the 100th s (shown in Fig. 3, black curve). The maximum sensing current was 67.61 ± 0.08 μA (data obtained from the mean of current signal in the period of 1400–1800 s). This sensing signal drifted to 63.87 ± 0.18 μA at the end of the measurement (data obtained from the mean of current signal in the period of 3600–4000 s); this was a drift of –5.53%. However, a comparison of the CV between the newly assembled sensor and the used sensor (ESM Fig. S8) suggested the presence of an additional species in the used sensor (e.g., the moisture [29]). The Bode plots of the impedance measurement (ESM Fig. S8) indicated the presence of electrode surface adsorbates in the used sensor (and most likely the Pt∙∙∙[Bmpy]+∙∙∙O2- species) because the Z’ is smaller but the Z” value is larger. (i.e., smaller Z’ means lower resistance, larger Z” means higher capacitance) [30]. One of benefit of the electrochemical sensor is that the electrode can be regenerated by repetitive electrode reduction and oxidation reactions via CV method [31, 32]. Using this approach, we performed CV sweeping at the platinum electrode in the potential range of –1.5-0.5 V every 20 min in an inert nitrogen environment 10 times. After this electrochemical activation step, a reproducible amperometric current-time curve (Fig. 3, blue curve) was obtained. Analyzing the same time period, the peak sensing current was 65.86 ± 0.39 μA (time 1400–1800 and drifted to a final value of 63.98 ± 0.35 μA (time 3600–4000 s). The sensing drift after the working electrode activation dropped to –2.85%. By integrating the current-time curves from time 150–4000 s in Fig. 3, the charge values of the “new sensor” and the “used sensor after electrode activation” were –0.3588 C and –0.3531 C, respectively. After the reactivity of the working electrode was restored, the charge value indicated that 98.41% of the electrode reactivity was restored. EIS measurements in Fig. S9 (see ESM) further validate the restoration of the platinum electrode reactivity in that the Nyquist plot of the “used sensor after electrode activation” was similar to that of the “new sensor”.
Fig. 3

Comparison of constant potential (E = –1.2 V) chronoamperometry of continuous oxygen monitoring using a sensor at different device conditions: new sensor (black), and used sensor after electrode activation (blue), demonstrating the regeneration of the electrode/electrolyte interface through electrode activation process; 5% oxygen was introduced at the 100th second of all measurements. Gas flow rate was 100 sccm

Amperometric methods for signal drift minimization

Above discussions show that the drift of an IL-EG oxygen sensor predominately results from the electrode surface reactivity and the oxygen concentration change relating to the slow mass transport of both the reactant and the product, which is attributable to the presence of the O2- species. Since O2/O2- is a redox couple in an aprotic IL (see the CV in ESM Fig. S2), applying a double potential step method can be an effective way to remove the O2- species via oxidization to oxygen, which can maintain the stability of IL/electrode interface double layer condition. In this work, we programmed the double potential step amperometric measurement to alternate between 0 V and –1.2 V. The 0 V potential is sufficient to oxidize the O2- species, as the oxidation peak of O2- occurs at –0.9 V (Fig. S2). Moreover, as Jewell reported, the alternating reduction and oxidation of a species at the platinum surface could function to enhance platinum reactivity [32]. Thus, the double potential step method should be effective to not only remove the O2- species but to also maintain the electrode reactivity for the purpose of continuous sensing.

Because IL is an aggregation of cations and anions, we hypothesized that applying a –1.2 V potential will not only influence the arrangement of the ions in such a way that cations will tend to accumulate on the electrode surface, it will also impair the sensing signal stability as more redox by-products will form. However, the conditioning step can mitigate this effect and facilitate the restoration of the IL/electrode interface double layer condition [33, 34]. Hence, measurements were designed with a double potential step chronoamperometry setup and a comparison of a long versus short period of the conditioning step (Fig. 4). Results validated our hypothesis that the longer period of the conditioning step enhanced the sensing signal stability and extended the sensor operation lifetime. Each of these measurements consisted of a total of 600 repetitive double potential step cycles. Each double potential step cycle started with a conditioning step setting the potential at 0 V followed by a sensing step setting the potential at –1.2 V. Two different periods of conditioning steps were studied (i.e., 59 s, and 159 s). The measurement began in an inert nitrogen environment and the 5% oxygen/nitrogen gas mixture was introduced at the 11th cycle. The data acquisition rate was 100 data points per s. Mean values of the last 10 data points of the sensing step of each cycle were plotted as a function of the repetitive measurement cycle (Fig. 4). At a shorter conditioning step (Fig. 4a), the peak sensing signal was –41.24 μA. The sensing signal decreased to –27.29 μA after 600 sections, which was a 33.83% decrease in value. The lifetime of this sensor in terms of a 5% sensing signal loss was calculated as 0.55 h. This value was calculated from the total time between the 80th section (when the sensor exhibited maximum sensing signal) and the 113rd section (when 5% sensing signal loss was detected). However, with prolonged conditioning step (Fig. 5b), the sensor lifetime was extended and exhibited less sensing drift. The sensing signal at the 80th section was –39.98 μA, and the sensing signal decreased to –30.22 μA after 600 sections (a 24.41% decrease in value). Therein, the sensor’s lifetime in terms of 5% sensing signal loss was extended to 9.82 h (calculated from the 80th section to the 301st section). This is a tremendous improvement of the sensor’s operation lifetime.
Fig. 4

Plots of the mean value of the last 10 data points in the sensing period of all 600 repetitive measurements in the double potential step potential chronoamperometry. Period of the sensing step was set as 1 s for both measurements, but the period of the conditioning step varied: (a) 59 s and (b) 159 s

Fig. 5

The offline calibration for constant-potential oxygen amperometric sensor test. (a) The real-time sensing raw data, showing signal responses of oxygen exposure (20%) and no oxygen exposure in turn. (b) The steps of performing calibration. (c) The sensing signal after calibration. (d) The profile of amperometric test parameter settings

Signal drift calibration with mathematic tools

One significant advantage of continuous sensing is that we can utilize the characteristics of the sensor data obtained in the prior detection at earlier times to calibrate the oxygen-sensing signal via data analysis using two simple algorithms. The use of the sensing data immediately before the following sensing measurement provides the best baseline and sensitivity calibration because it calibrates the immediate change of the sensor system. Our objective is to calibrate the sensitivity decrease and the baseline drift due to the accumulative influences (i.e., electrode surface deactivation, sensing by-product formation, and aggregation). Scientists and engineers have been working on developing intelligent algorithms for on-the-fly sensor calibration [35, 36], but here we attempted to establish a simple offline mathematical method to calibrate the sensing signal of an electrochemical gas sensor, which can be incorporated on-line. Following the algorithms reported by Gutierrez-Osuna [37], we use the IL-EG oxygen sensor as a model system to attempt the offline signal calibration. In the process of oxygen sensing, oxygen molecules are reduced to form the O2- species. The presence of the O2- species will cause a change in the micro-environment of the electrode/electrolyte interface, which manifests as the sensing signal baseline fluctuates. We employ a differential signal manipulation model (Eq. 4) that is capable of satisfying drift compensation if no other interference coexists. However, the O2- species will also affect the electrode surface activity, resulting in a decrease of sensing sensitivity. In this case, the relative signal manipulation model (Eq. 5) can meet the need for sensitivity drift correction.
$$ {i}^{'}\left(\mathrm{t}\right)=i\left(\mathrm{t}\right)\hbox{--} i\left(\mathrm{t}-1\right) $$
$$ {i}^{'}\left(\mathrm{t}\right)=i\left(\mathrm{t}\right)/i\left(\mathrm{t}-1\right) $$
in which, i(0) is the sensing signal at time zero (or the background sensing signal), i(t) is the sensing signal at time t with the exposure to the target gas analyte, and i’(t) is the calibrated sensing signal.

Figure 4 shows an example of how both the differential and relative signal manipulation methods were applied to calibrate the amperometric sensing signal when the sensor was exposed to alternating 0% and 20% oxygen concentrations (see the test parameter profile in Fig. 4d). After calibration, the sensing drift was reduced from ±3.05% to ±0.19%. The detailed calibration procedure was as follows: the amperometric curve was corrected according to a baseline correction line (marked as the purple dash line in Fig. 4b) in the calibration equation of y = 5.14e-4 x -11.67. According to Wenzel et al. [27], applying a difference measurement between the steady-state signal and the conditioning signal can minimize the influence of baseline drift in analyte quantification. Thus, the baseline calibration curve was picked by aligning the mean current value of the last 50 data points in the conditioning steps in segments 1 and 2 (marked by the star symbols in Fig. 4b). After the baseline subtraction, the sensing signal was further manipulated by means of the relative signal manipulation (Eq. 5). The final outcome of this offline calibration is presented in Fig. 4c.


Viewing the importance of sensing signal drift compensation to enhance the reliability of the IL-EG sensors, we use an established oxygen sensor as a model system to investigate the fundamental chemical causes of the sensing signal drift. Analyte concentration variations and the electrode surface reactivity were studied to understand the influences toward sensor signal drift. Analyte concentration variation influences the sensing signal baseline as it modifies the IL/electrode interface double layer condition, whereas electrode surface reactivity causes the decrease of the sensing sensitivity.

Aiming to address the sensor signal drift, we reported a double potential step chronoamperometric method and demonstrated its effectiveness in mitigating the sensor signal drift while extending the sensor operation lifetime. In additional, two mathematical models (differential signal manipulation and relative signal manipulation) were employed to calibrate both the baseline drift and the sensor sensitivity change for continuous constant potential amperometric sensing of oxygen.

This is the first systematic study in identifying the chemical sensing drift issue in IL-EG sensors. Although it is impossible to eliminate the sensor signal drift, we are setting some perspective starting points to address, minimize, and calibrate the sensing drift so that both the sensing accuracy/precision and sensitivity can be improved for real-world applications, which demand the reliable continuous sensing and monitoring of important gaseous analytes with extended life-time. It is envisaged that these studies form the foundation toward improving the reliability of the IL-based electrochemical gas sensors by establishing the calibration methods to compensate for both short-term (hours) and long-term (days, months) signal drift to increase the reliability of the IL-EG sensor performance for continuously monitoring many gas analytes in real-world conditions.



X. Zeng acknowledges grant support from the National Institute of Environmental Health (R01ES022302) and the Alpha Foundation AFC518-2 for this research. The authors thank Dr. Xiaojun Liu and Dr. Jessica Koppen for helpful comments and proofreading.

Compliance with Ethical Standards

Conflict of interest

The authors declare no conflict of interest of this work.

Supplementary material

216_2018_1090_MOESM1_ESM.pdf (579 kb)
ESM 1 (PDF 578 kb)


  1. 1.
    Moeller DW, Moeller DW (2009) Environmental health. Harvard University Press, Cambridge, Massachusetts and London, EnglandGoogle Scholar
  2. 2.
    Wang Z, Guo M, Baker GA, Stetter JR, Lin L, Mason AJ, et al. Methane–oxygen electrochemical coupling in an ionic liquid: a robust sensor for simultaneous quantification. Analyst. 2014;139:5140–7.CrossRefPubMedPubMedCentralGoogle Scholar
  3. 3.
    Wang Z, Lin P, Baker GA, Stetter J, Zeng X. Ionic liquids as electrolytes for the development of a robust amperometric oxygen sensor. Anal Chem. 2011;83:7066–73.CrossRefPubMedGoogle Scholar
  4. 4.
    Mu X, Wang Z, Zeng X, Mason AJ. A robust flexible electrochemical gas sensor using room temperature ionic liquid. IEEE Sensors J. 2013;13:3976–81.CrossRefGoogle Scholar
  5. 5.
    Xiao C, Rehman A, Zeng X. Dynamics of redox processes in ionic liquids and their interplay for discriminative electrochemical sensing. Anal Chem. 2012;84:1416–24.CrossRefPubMedGoogle Scholar
  6. 6.
    Tang Y, Lin L, Kumar A, Guo M, Sevilla M, Zeng X. Hydrogen electrooxidation in ionic liquids catalyzed by the NTf2 radical. J Phys Chem C. 2017;121:5161–7.CrossRefGoogle Scholar
  7. 7.
    Tang Y, Zeng X. Electrochemical oxidation of hydrogen in Bis (trifluoromethylsulfonyl) imide ionic liquids under anaerobic and aerobic conditions. J Phys Chem C. 2016;120:23542–51.CrossRefGoogle Scholar
  8. 8.
    Tang Y, Wang Z, Chi X, Sevilla MD, Zeng X. In situ generated platinum catalyst for methanol oxidation via electrochemical oxidation of Bis (trifluoromethylsulfonyl) imide anion in ionic liquids at anaerobic condition. J Phys Chem C Nanomater Interfaces. 2015;120:1004–12.CrossRefPubMedPubMedCentralGoogle Scholar
  9. 9.
    Lin L, Rehman A, Chi X, Zeng X. 2, 4-Toluene diisocyanate detection in liquid and gas environments through electrochemical oxidation in an ionic liquid. Analyst. 2016;141:1519–29.CrossRefPubMedPubMedCentralGoogle Scholar
  10. 10.
    Vergara A, Vembu S, Ayhan T, Ryan MA, Homer ML, Huerta R. Chemical gas sensor drift compensation using classifier ensembles. Sensors Actuators B Chem. 2012;166:320–9.CrossRefGoogle Scholar
  11. 11.
    Buzzeo MC, Hardacre C, Compton RG. Use of room temperature ionic liquids in gas sensor design. Anal Chem. 2004;76:4583–8.CrossRefPubMedGoogle Scholar
  12. 12.
    Silvester DS. Recent advances in the use of ionic liquids for electrochemical sensing. Analyst. 2011;136:4871–82.CrossRefPubMedGoogle Scholar
  13. 13.
    Stetter JR, Penrose WR. Understanding chemical sensors and chemical sensor arrays (electronic noses): past, present, and future. Sensors Updat. 2002;10:189.CrossRefGoogle Scholar
  14. 14.
    Holmberg M, Winquist F, Lundström I, Davide F, DiNatale C, D’Amico A. Drift counteraction for an electronic nose. Sensors Actuators B Chem. 1996;36:528–35.CrossRefGoogle Scholar
  15. 15.
    Dossi N, Toniolo R, Pizzariello A, Carrilho E, Piccin E, Battiston S, et al. An electrochemical gas sensor based on paper supported room temperature ionic liquids. Lab Chip. 2012;12:153–8.CrossRefPubMedGoogle Scholar
  16. 16.
    Lu L, Zhao P, Mason A, Zeng X. Characterization of the IL-electrode interfacial relaxation processes under potential polarization for ionic liquid amperometric gas sensor method development. ACS Sensors. 2018.
  17. 17.
    Fonollosa J, Neftci E, Huerta R, Marco S. Evaluation of calibration transfer strategies between metal oxide gas sensor arrays. Procedia Eng. 2015;120:261–4.
  18. 18.
    Arshak K, Moore E, Lyons GM, Harris J, Clifford S. A review of gas sensors employed in electronic nose applications. Sens Rev. 2004;24:181–98.CrossRefGoogle Scholar
  19. 19.
    Li Z, Liu H, Liu Y, He P, Li J. A room temperature ionic-liquid-templated proton-conducting gelatinous electrolyte. J Phys Chem B. 2004;108:17512–8.CrossRefGoogle Scholar
  20. 20.
    Liu Y, Li J, Wang M, Li Z, Liu H, He P, et al. Preparation and properties of nanostructure anatase TiO2 monoliths using 1-butyl-3-methylimidazolium tetrafluoroborate room temperature ionic liquids as template solvents. Cryst Growth Des. 2005;5:1643–9.!divAbstract
  21. 21.
    Yu L, Garcia D, Ren R, Zeng X. Ionic liquid high temperature gas sensors. Chem Commun. 2005;0:2277–79.Google Scholar
  22. 22.
    Endres F, El Abedin SZ. Air and water stable ionic liquids in physical chemistry. Phys Chem Chem Phys. 2006;8:2101–16.CrossRefPubMedGoogle Scholar
  23. 23.
    Xiong L, Aldous L, Henstridge MC, Compton RG. Investigation of the optimal transient times for chronoamperometric analysis of diffusion coefficients and concentrations in non-aqueous solvents and ionic liquids. Anal Methods. 2012;4:371–6.CrossRefGoogle Scholar
  24. 24.
    Buzzeo MC, Klymenko OV, Wadhawan JD, Hardacre C, Seddon KR, Compton RG. Voltammetry of oxygen in the room temperature ionic liquids 1-ethyl-3-methylimidazolium bis [(trifluoromethyl) sulfonyl] imide and hexyltriethylammonium bis [(trifluoromethyl) sulfonyl] imide: one-electron reduction to form superoxide. Steady-state and tr. J Phys Chem A. 2003;107:8872–8.CrossRefGoogle Scholar
  25. 25.
    Xiao C, Rehman A, Zeng X. Evaluation of the dynamic electrochemical stability of ionic liquid--metal interfaces against reactive oxygen species using an in situ electrochemical quartz crystal microbalance. RSC Adv. 2015;5:31826–36.CrossRefGoogle Scholar
  26. 26.
    Campbell JL, Rustad LE, Porter JH, Taylor JR, Dereszynski EW, Shanley JB, et al. Quantity is nothing without quality: automated QA/QC for streaming environmental sensor data. Bioscience. 2013;63:574–85.CrossRefGoogle Scholar
  27. 27.
    Wenzel MJ, Mensah-Brown A, Josse F, Yaz EE. Online drift compensation for chemical sensors using estimation theory. IEEE Sensors J. 2011;11:225–32.CrossRefGoogle Scholar
  28. 28.
    Panchenko A, Koper MTM, Shubina TE, Mitchell SJ, Roduner E. Ab initio calculations of intermediates of oxygen reduction on low-index platinum surfaces. J Electrochem Soc. 2004;151:A2016–27.CrossRefGoogle Scholar
  29. 29.
    Pandey S. Analytical applications of room-temperature ionic liquids: a review of recent efforts. Anal Chim Acta. 2006;556:38–45.CrossRefPubMedGoogle Scholar
  30. 30.
    Bard AJ, Faulkner LR, et al. Fundamentals and applications. Electrochem Methods, 2nd ed. New York: Wiley; 2001.Google Scholar
  31. 31.
    Parsons R. The kinetics of electrode reactions and the electrode material. Surf Sci. 1964;2:418–35.CrossRefGoogle Scholar
  32. 32.
    Jewell RC. Platinum in the glass industry. Platin Met Rev. 1964;8:122–7.Google Scholar
  33. 33.
    Adelman WJ, Palti Y. The effects of external potassium and long duration voltage conditioning on the amplitude of sodium currents in the giant axon of the squid, Loligo pealei. J Gen Physiol. 1969;54:589–606.CrossRefPubMedPubMedCentralGoogle Scholar
  34. 34.
    Aspelund A, Jordal K. Gas conditioning—the interface between CO2 capture and transport. Int J Greenh Gas Control. 2007;1:343–54.CrossRefGoogle Scholar
  35. 35.
    Schwenke H, Schmitt R, Jatzkowski P, Warmann C. On-the-fly calibration of linear and rotary axes of machine tools and CMMs using a tracking interferometer. CIRP Ann Technol. 2009;58:477–80.CrossRefGoogle Scholar
  36. 36.
    Hasenfratz D, Saukh O, Thiele L (2012) On-the-fly calibration of low-cost gas sensors. In: European Conference on Wireless Sensor Networks. pp. 228–244Google Scholar
  37. 37.
    Gutierrez-Osuna R (2003) Signal processing methods for drift compensation. In: 2nd NOSE II Workshop Linköping. pp. 18–21Google Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Authors and Affiliations

  1. 1.Department of ChemistryOakland UniversityRochesterUSA

Personalised recommendations