A direct comparison of amperometric gas sensors with gas-diffusion and ion-exchange membrane based electrodes

Abstract

The effect of the nature of the working electrode used in amperometric gas sensors on the performance criteria of sensitivity, detection limit, gas flow rate and humidity dependence was evaluated. The arrangement based on metallized ion-exchange membranes (Nafion) was compared with gas-diffusion electrodes based on porous poly(tetrafluoroethylene) (PTFE) with metallic electrodes deposited on the rear side. Two representative analyte gases were chosen: SO₂, which has fast reaction kinetics, and NO, which has slow reaction kinetics. It was found that both types of electrodes showed a similar performance. A dependence on the flow rate of the sample gas was found in both cases. The sensitivities were higher for the ion-exchange membrane-backed electrodes; however, the 3σ detection limits were all in the lower ppb range and for NO were significantly lower on the Nafion membrane than on the PTFE membrane. The Nafion electrode was found to show a dependence on the relative humidity of the gas stream, but not the PTFE-based electrode.

---

Introduction

Amperometric gas sensors for the direct determination of electroactive species in the gas phase have evolved from the Clark electrode for dissolved oxygen. These sensors are usually based on a gas-permeable membrane such as porous poly(tetrafluoroethylene) (PTFE) bearing a precious metal electrode on the rear side which is in contact with an internal electrolyte solution. The analyte gas diffuses through the not-wetted pores in a non-selective manner in this case. Devices based on such gas-diffusion electrodes (GDEs) have been commercially available for some time for the monitoring of a number of toxic species such as NO_x, SO₂, H₂S, NH₃, HCN and CO, mainly for applications in industrial hygiene. The lower limits of detection are usually taken to be at about 1 ppm. These levels are not adequate for many applications such as environmental monitoring. General reviews on such amperometric gas sensors are available, e.g., refs. 1 and 2.

A fundamentally different arrangement is based on metallized ion-exchange membranes [also often termed solid polymer electrolyte (SPE) membranes in this context] with electrodes on the front-side, i.e., in direct contact with the gas phase. This means that the analyte gas does not have to diffuse through the membrane in this case and the mass transport rate to the electrode can be expected to be high. Such sensors have been studied by several researchers working in the field for approximately the last 10 years^3–24 and this topic has been reviewed by Bontempelli et al.²⁵ and Opekar and Stulik.²⁶ These devices were found to perform similarly in terms of selectivity to the diffusion membrane-based type, which is not surprising considering that the fundamental electrode reactions are identical. However, very high sensitivities and the achievement of unprecedentedly low detection limits in the low ppb range have been reported for direct contact gas electrodes based on solid polymer electrolytes.^5,17–19,23 This performance makes them attractive for applications which cannot be covered by the commercially available sensors based on gas-diffusion electrodes. However, it is not clear if the high sensitivity of the electrodes based on metallized ion-exchange membranes is mainly a feature of the different electrode arrangement or is caused by other experimental parameters. A comparison of the performance reported in the literature for the two arrangements has been impossible because the operating conditions used by different workers have always varied considerably. A direct one-to-one comparison of the two types of sensors has to our knowledge not been carried out.

The two gases SO₂ and NO were chosen for this study, representing species which possess fast and slow electrode reaction kinetics, respectively. Chiou and Chou²⁷ described an SO₂ sensor for the ppm range using a gas-diffusion working electrode based on dispersed Au. They found a sensitivity of 1 μA ppm⁻¹ and a fast response time. The oxidation current was constant at potentials more positive than 0.6 V vs. Ag/AgCl, indicating fast reaction kinetics. Schiavon et al.⁵ and Hodgson et al.¹⁹ described the detection of SO₂ with sensors based on ion-exchange membranes as electrode supporting material (SPE). Different membrane materials were used to support gold electrodes and acidic and basic electrolyte solutions were investigated. High sensitivities of 1–12 μA ppm⁻¹ and low detection limits down to 1 ppb were reported, in addition to very fast electrode kinetics.^5,19 Sedlak and Blurton^28,29 described the sensing of NO employing a gas-diffusion type of membrane. Do and Wu⁹ and Jacquinot et al.²³ reported sensors for NO utilizing ion-exchange membranes. Sensitivities between 1 and 5 μA ppm⁻¹ and a detection limit in the lower ppb range were reported. The absence of an effect of a diffusion barrier, which was deliberately introduced in front of the electrode for test purposes, and the strong influence of the real surface area of the electrode²³ both point to a strong effect of the reaction kinetics on the measured current.

Experimental

The design of the electrochemical cell has been described previously.^15,21 Gas-diffusion type electrodes were obtained from Transducer Research (Chicago, IL, USA) and are based on a porous PTFE membrane (Zitex) of 0.26 mm thickness with a powder-coated gold layer deposited on one side. The SPE-based electrodes were produced by chemical deposition of gold on Nafion 117 (ElectroChem, Woburn, MA, USA).³⁰ In this procedure, tetrachloroauric acid in solution, which is in contact with one side of the membrane, is reduced by sodium borohydride, which diffuses through the membrane from the opposite side. Porous electrodes are obtained the real surface area of which can be varied by controlling the reaction time and the concentration of the NaBH₄.^15,31 The geometric surface of the working electrodes was always 0.79 cm². The real surface areas of the electrodes were determined from the charges due to the reduction of the oxide monolayer in the cathodic sweep of the cyclic voltammogram in 1 M H₂SO₄ with a value of 420 C cm⁻².³¹ The electrode areas were generally between 130 and 180 cm² on Nafion and between 12 and 24 cm² on PTFE membranes. The reference electrode used was a mercury/mercurous sulfate electrode (MSE) (Ref-601, Radiometer Analytical, Lyon, France) with a potential of 640 mV vs. SHE. A gold wire was employed as counter electrode (Advent Research Materials, UK) and 1 M sulfuric acid (Baker, Deventer, The Netherlands) was used as internal electrolyte solution. All measurements were carried out with a potentiostat (Model 263A, EG&G Instruments, Princeton, NJ, USA). Different concentrations of SO₂ and NO in nitrogen were prepared by mixing of certified gas standards (Carbagas, Basel, Switzerland) of 10 ppm with pure nitrogen at the correct ratio while adjusting the individual flows to yield the desired total flow rate. Mass flow controllers (MFCs) (Type 1179A and 1159B, with maximum flow rates of 10, 100 and 200 cm³ min⁻¹; MKS Instruments, Munich, Germany) were employed for dilution and adjustment of gas flow rates. Humidification was achieved by passing a nitrogen stream through a Drechsel bottle containing water. The Au–Nafion electrode was allowed to equilibrate for 30 min after a change of humidity. Measurements involving stagnant gases were carried out by placing the open cell in a closed container filled with the standard gas mixture. All experiments were done at room temperature. Detection limits were determined from five consecutive exposures of the sensor to low concentrations in alternation with pure nitrogen. The values given correspond to a signal of three times the standard deviation.

Results and discussion

The dependence of the response of the cell for SO₂ to changes in flow rate and applied working potential is illustrated in Fig. 1(A) and (B) for the metallized ion-exchange membrane and the porous gas diffusion membranes, respectively. For both electrodes, the reaction of SO₂ sets in at 0 mV, a maximum is obtained at 100 mV and the currents remain fairly constant up to the highest potential investigated for each flow rate. Note that beyond a potential of 600 mV a loss of sensitivity is encountered commensurate with the onset of the gold oxide formation. For the higher applied potentials the sensitivity is dependent only on the gas flow rate (i.e., a plateau region is obtained). This indicates fast reaction kinetics and that the current must be limited by the rate of transport of the analyte gas to the electrode (mass transport current limitation). Note the different scales on the current axis for the two types of electrodes. From the linear plot of resulting current vs. gas flow rate at the highest possible potential (with least expected kinetic current limitation) of Fig. 2, it is evident that a different effect of flow rate is observed for the two membranes. The relative increase at high flow rates is limited for the PTFE-based electrode in comparison with the behaviour of the ion-exchange membrane-backed electrode. This may be an indication that the porous PTFE membrane (in front of the electrode) indeed acts as a diffusion barrier. In the case of the Nafion-backed electrode a depletion layer may be present in the gas phase; however, considering the small cell volume in front of the electrode, an active transport (enforced flow) of gas into the cell is mandatory to maintain the measured current. Considering the mass transport of analyte into the cell, the highest Faradaic efficiency is obtained at a flow rate of 50 cm³ min⁻¹ for the Nafion-based electrode (32%). It also has to be considered that the effective surface area of the PTFE-based electrode is much smaller than that of the Nafion-backed electrode (14 vs. 176 cm²). This aspect is discussed further below.

Fig. 1 3D plots of current for 1 ppm SO₂ in N₂vs. applied potential and gas flow rate at Au–Nafion (A) and PTFE–Au (B) electrodes. Real surface areas are 176 cm² for Nafion and 14 cm² for PTFE; flow rates 5, 10, 20, 30, 50, 100 and 200 cm³ min⁻¹; potential steps from –50 to +650 mV vs. MSE at 50 mV intervals. Note the non-linear scale on the flow rate axis.

Fig. 2 Plots of current vs. gas flow rate for SO₂ measured at Au–Nafion (open circles) and Au–PTFE (closed circles) electrodes; +600 mV vs. MSE.

In contrast to the case with SO₂, the response of the sensor to NO is much more strongly dependent on the applied potential. As evidenced by the data in Fig. 3(A) and (B), a current plateau is not reached. This indicates that the current for the oxidation of NO is largely limited by the reaction kinetics on the electrode surface. However, on the Nafion electrode a pronounced effect of the gas flow rate was still observed, a sign of current limitation by mass transport. This indicates, overall, mixed current control on this electrode. The difference in behaviour of the two electrodes again is likely to be partly due to the fact that the PTFE membrane acts as a small diffusion barrier.

Fig. 3 3D plots of current for 1 ppm NO in N₂vs. applied potential and gas flow rate at Au–Nafion (A) and PTFE–Au (B) electrodes. Real surface areas are 176 cm² for Nafion and 14 cm² for PTFE; flow rates 5, 10, 20, 30, 50, 100 and 200 cm³ min⁻¹; potential steps from +350 to +700 mV vs. MSE at 50 mV intervals. Note the non-linear scale on the flow rate axis.

Another factor to be considered is the effective surface area, or roughness, of the porous electrodes. It is possible to control the effective area in the manufacture of the Nafion electrodes by adjusting the concentrations of the gold salt and the reducing agent as well as the deposition time. As shown in Fig. 4, for the kinetically hindered species NO a strong dependence on the effective surface area is obtained on the Nafion membrane. The relationship is practically linear, indicating almost complete kinetic current control. In contrast, for SO₂ a limited effect of the surface area on the sensitivity was found for the smallest areas only. This is in agreement with the expected fast reaction kinetics. When comparing the performance of the two types of membranes the effect of the surface area must be taken into consideration. However, the fact that for SO₂, which does not show kinetic current limitation, higher sensitivities are obtained even for electrodes with areas comparably small compared with the PTFE electrodes corroborates the finding from above that the porous PTFE membrane acts as a diffusion barrier. On the PTFE-based electrode, with a relatively small active surface area, saturation is obtained even at low flow rates and an increase cannot lead to improved sensitivity [Fig. 3(B)].

Fig. 4 Influence of real surface area of Au–Nafion electrodes on the currents for SO₂ (open circles) and NO (open squares); potential +600 mV vs. MSE; flow rate 100 cm³ min⁻¹.

The response of both types of membranes to low concentrations of the two gases was tested for a gas flow rate of 100 cm³ min⁻¹ and an applied potential 600 mV vs. MSE. The results for SO₂ are illustrated in Fig. 5. Clearly, high sensitivity is obtained for both arrangements and the response is linear. The response times are also comparable for the two cases. For SO₂ the sensitivities were determined as 5.3 and 0.93 μA ppm⁻¹ for the ion-exchange and gas-diffusion membranes and the 3σ detection limits were determined as 0.7 and 0.5 ppb, respectively. For NO the sensitivities were measured as 5.6 and 0.18 μA ppm⁻¹ and the detection limits as 1.8 and 10 ppb again for the ion-exchange and gas-diffusion membranes, respectively. Good detection limits can therefore generally be obtained with both types of membranes for the working conditions employed. The higher sensitivities obtained with the ion-exchange membrane-based sensor do not necessarily lead to significantly lower detection limits. Only in the case of the kinetically hindered NO is a distinct difference observed.

Fig. 5 Current vs. time profile at an Au–Nafion electrode (a) (176 cm²) and a PTFE–Au electrode (b) for 0–200 ppb SO₂ in N₂, in 20 ppb steps.

Commercially available amperometric gas sensors are usually employed without a forced gas flow as in our case. The sensors are exposed directly to the ambient atmosphere, and often contain a deliberate diffusion barrier in front of the membrane. The purpose of this mechanical restriction is to achieve independence of the signal from draughts in the air. To simulate these conditions, the gas inlet/outlet of our sensor was opened to a diameter of about 0.7 cm and this sensor was placed in a container of about 500 cm³ volume. The salient response data achieved with this arrangement were sensitivities of 0.16 and 0.15 μA ppm⁻¹ for SO₂ and 0.28 and 0.19 μA ppm⁻¹ for NO on the Nafion- and PTFE-based electrodes, respectively. In this situation the currents are thus all low compared with the situation with active gas transport and therefore limited by mass transport. Correspondingly, the detection limits were all relatively high and determined as 60 and 19 ppb for SO₂ and 25 and 15 ppb for NO on the Nafion- and PTFE-based membranes, respectively.

In Fig. 6, the effect of the relative humidity of the gas stream on the current sensitivity for both SO₂ and NO is shown. The absence of such an influence is desirable as the humidity of the sample gas cannot be assumed to be constant in an application. For the Nafion electrode a pronounced dependence on the moisture content is found, whereas the effect on the PTFE electrode is almost negligible. As the latter electrode is effectively placed in the internal electrolyte solution of the cell, the freedom from interference is not surprising. The effect on the Nafion electrode may be rationalized by the fact that water is directly involved in the oxidation of SO₂ according to the following equation:¹⁹

SO₂ + 2H₂O → SO₄²⁻ + 4H⁺ + 2e⁻

It appears that the rate of reaction in this case may be limited by the flux of water. Even though the Nafion membrane is hydrated, there is probably a local depletion on the surface in a dry gas stream as the diffusion of water through the membrane must be slow relative to the mass transport in the gas phase. However, it is also possible that the relative humidity, by affecting the degree of swelling on the surface, also influences the condition of the porous electrode or the rate of transport of reaction products away from the electrode and thus exerts an indirect effect.

Fig. 6 Effect of relative humidity on the sensitivity for SO₂ (open and closed circles) and NO (open and closed squares) on an Au–Nafion electrode (open circles and squares) and a PTFE–Au electrode (closed circles and squares). Gas flow rate 100 cm³ min⁻¹.

In contrast, in the case of NO the sensitivity was found to decrease with increasing humidity. The highest sensitivities are measured with a dry gas stream. The oxidation reaction also consumes water according to the equation:²

NO + 2H₂O → NO₃⁻ + 4H⁺ + 3e⁻

Obviously, the explanation offered for the behaviour of SO₂ cannot be applied to NO. A more subtle, not understood, phenomenon must take place. Possibly water leads to a blockage of active electrode sites.

Conclusion

The similarities and differences of amperometric gas sensors based on metallized ion-exchange membranes and more conventional gas-diffusion membranes were highlighted. It was found that detection limits in the low ppb range can obtained for both types of membranes if active gas transport is utilized. This implies that such gas sensors based on either membrane type could be more widely employed for applications not currently considered if they were combined with a pump to create a controlled flow of gas. If the sensors are employed with passive diffusion, the detection limits are higher by one to two orders of magnitude for both membranes and independent of the kinetic behaviour of the analyte gas. A disadvantage of the metallized ion-exchange membranes is their humidity dependence, which is not present with the gas-diffusion type electrodes.

A metallized membrane with a very high active surface area may used to advantage for gases with relatively slow reaction kinetics when very low detection limits are desired. Of the different gases studied by our group, it was found that fast reaction kinetics comparable to those for SO₂ were exhibited only by acetaldehyde.²¹ Ethanol,²¹ formaldehyde,²⁴ ethylene,¹⁵ acetylene,¹⁴ ethylene oxide²⁰ and NO₂²³ all showed a behaviour similar to that of NO.

References

1. Z. Cao, W. J. Buttner and J. R. Stetter, Electroanalysis, 1992, 4, 253.
2. S. C. Chang, J. R. Stetter and C. S. Cha, Talanta, 1993, 40, 461.
3. G. Schiavon, G. Zotti and G. Bontempelli, Anal. Chim. Acta, 1989, 211, 27.
4. G. Schiavon, G. Zotti, G. Bontempelli, G. Farnia and G. Sandona, Anal. Chem., 1990, 62, 293.
5. G. Schiavon, G. Zotti, R. Toniolo and G. Bontempelli, Analyst, 1991, 116, 797.
6. G. Schiavon, G. Zotti, R. Toniolo and G. Bontempelli, Anal. Chem., 1995, 67, 318.
7. G. Schiavon, N. Comisso, R. Toniolo and G. Bontempelli, Electroanalysis, 1996, 8, 544.
8. J. S. Do and R. Y. Shieh, Sens. Actuators, B, 1996, 37, 19.
9. J. S. Do and K. J. Wu, Sens. Actuators B, 2000, 67, 209.
10. F. Opekar, J. Langmaier and Z. Samec, J. Electroanal. Chem., 1994, 379, 301.
11. H. Yan and C. C. Liu, Sens. Actuators, B, 1994, 17, 165.
12. L. R. Jordan and P. C. Hauser, Anal. Chem., 1997, 69, 2669.
13. L. R. Jordan, P. C. Hauser and G. A. Dawson, Electroanalysis, 1997, 9, 1159.
14. L. R. Jordan and P. C. Hauser, Anal. Chem., 1997, 69, 2669.
15. L. R. Jordan, P. C. Hauser and G. A. Dawson, Anal. Chem., 1997, 69, 558.
16. L. R. Jordan, P. C. Hauser and G. A. Dawson, Analyst, 1997, 122, 811.
17. A. W. E. Hodgson, P. Jacquinot, L. R. Jordon and P. C. Hauser, Electroanalysis, 1999, 11, 782.
18. A. W. E. Hodgson, P. Jacquinot, L. R. Jordan and P. C. Hauser, Anal. Chim. Acta, 1999, 393, 43.
19. A. W. E. Hodgson, P. Jacquinot and P. C. Hauser, Anal. Chem., 1999, 71, 2831.
20. A. W. E. Hodgson, P. Jacquinot and P. C. Hauser, Anal. Chem., 2000, 72, 2206.
21. P. Jacquinot, A. W. E. Hodgson, B. Müller, B. Wehrli and P. C. Hauser, Analyst, 1999, 124, 871.
22. P. Jacquinot, B. Müller, B. Wehrli and P. C. Hauser, Anal. Chim. Acta, 2001, 432, 1.
23. P. Jacquinot, A. W. E. Hodgson and P. C. Hauser, Anal. Chim. Acta, 2001, 443, 53.
24. R. Knake, P. Jacquinot and P. C. Hauser, Electroanalysis, 2001, 13, 631.
25. G. Bontempelli, N. Comisso, R. Toniolo and G. Schiavon, Electroanalysis, 1997, 9, 433.
26. F. Opekar and K. Stulik, Anal. Chim. Acta, 1999, 385, 151.
27. C. Y. Chiou and T. C. Chou, Electroanalysis, 1996, 8, 1179.
28. J. M. Sedlak and K. F. Blurton, J. Electrochem. Soc., 1976, 123, 1476.
29. J. M. Sedlak and K. F. Blurton, Talanta, 1976, 23, 811.
30. R. C. Cook, R. C. MacOuffard and A. F. Sammelli, J. Electrochem. Soc., 1981, 132, 187.
31. H. Kita and H. Nakajima, Electrochim. Acta, 1986, 31, 193.

---