J. MATER. CHEM., 1994,4(8), 1259-1262 Improvement of Copper Oxide-Tin Oxide Sensor for Dilute Hydrogen Sulfide Tomoki Maekawa, Jun Tamaki, Norio Miura and Noboru Yamazoe Department of Materials Science and Technology, Graduate School of Engineering Sciences, Kyushu University, Kasuga-shi, Fukuoka 816, Japan The influence of the loading and dispersion of CuO on the H,S-sensing properties of CuO-SnO, sensors has been investigated. The response rates to H,S depend on various factors such as CuO loading, specific surface area (or size) of SnO, grains, method of loading the SnO, grains with CuO and the operating temperature. For the CuO-SnO, sensor prepared by a chemical fixation method, the response rates to dilute H,S (1 -1 0 ppm) increased with decreasing CuO loading, while the sensitivity to H2S decreased monotonically with decreasing amount of CuO dispersed per unit surface area of SnO,.By optimizing these factors, it was possible to obtain a CuO-SnO, sensor which responded to H,S above 1 ppm at 160 "C with sufficient sensitivity and fairly good response kinetics. The sensing of odours has become increasingly important for the control of living environments and food processing. Hydrogen sulfide (H2Sj is a typical bad-smelling component, in addition to its rather strong toxicity. High performance sensors for H2S are demanded for the purposes of auto- ventilation, diagnosis in dentistry and safety. It has been attempted to develop semiconductor-type sensors for H2S. The sensitivity and/or selectivity of Sn0,-based sensors to H,S are reportedly improved by adopting unusual operation modes such as quick cooling' and thermal cycling,2 or by modifying the sensors with hydrophobic silica,, ZrO,,'' basic oxides5 or Ag,6 but further improvements seem to be necessary.Recently an Au-doped WO, film sputtered on LiNb0, was reported to respond to sub-ppm levels of H2S.7 An Sn0,-based sensor impregnated with a small amount (typically 5 wt.%) of CuO is extremely sensitive and selective to 50 ppm H,S in air.' However, the rate of response is highly dependent on the concentration of H2S and becomes extremely sluggish in dilute H,S conditions (<20 ppm). A clue to improving this problem seems to exist in the fact that the response rate also depends strongly on the amount of CuO as well as the choice of the copper salt used as the starting reagent for the impregnation of CUO.~ This suggests that the dispersion of CuO particles on the surface of SnO, grains strongly affects the H,S-sensing properties of the sensor.From this viewpoint, the influence of the specific surface area (or grain size) of SnO, and the CuO loading were investigated in more detail in this paper. A chemical fixation method was partly adopted to attain a fine dispersion of CuO. As a result, it was possible to extend the lower detection limit of the CuO-SnO, sensor down to ca. 1ppm H,S. Experimental Preparation of Sensing Material The powder of SnO, was prepared from SnC1, by a wet process' and was calcined at 400°C for 1h (Sn0,-A), 600°C for 5 h (Sn0,-B) or 900°C for 4 h (Sn0,-C).The specific surface area of SnO, was 62 (A), 16 (B), and 6.4 (C)m2 g-'. Each sample was loaded with CuO by one of the following two methods. (1) Impregnation method: SnO, powder was impregnated with a solution of basic copper carbonate, followed by calcination in air at 400 "C for 1h (Sn02-A) or at 700 "C for 4 h (Sn0,-B and C). The CuO loading was fixed at 5 wt.%. (2) Fixation method:" SnO, powder was suspended in an aqueous solution of CuC1, (0.05mol dmF3j and CH3CO2NH4 (1mol dm-3) at room temperature for 24 h under agitation, to obtain copper chloro complexes that are chemically fixed onto the SnO, surface as shown in Fig. 1. The suspension was then filtered off, washed with deionized water and calcined in air at 400°C for 1h.The CuO loading was determined using an X-ray fluorescence analyser. When the fixation reaction was carried out at pH 6, the CuO loading on SnO, (A), (Bj and (C) was 3.3, 1.3, and 0.42 wt.%, respectively, which is equivalent to 0.53,0.81 and 0.66 mg CuO rn-, of SnO, surface in surface loading density. The CuO loading on SnO, could be reduced by decreasing the pH of the CuC1,-containing solution with the addition of HC1. With Sn02-C, for example, CuO loadings were 0.15 and 0.091 wt.%, or 0.23 and 0.14 mg CuO mP2 of SnO, surface, at pH 4 and 2, respectively. Measurement of Sensing Properties Sensors of a sintered block type were fabricated as described elsewhere.8 The electrical resistance of each sensor was meas- ured in a flow (200cm3 rnin-') of dry air as well as H2S ( 1-48 ppm) in dry air at given temperatures (typically 200 "C).Prior to the acquisition of electrical resistance data, each sensor was exposed to H,S (48 ppm) in an air stream at 200°C for 1 h. The gas sensitivity (S)was defined as the ratio (R,/R,) of the electrical resistance in air (R,) to that in the sample gas (RJ. H Hoi oi. I I + [CuCh(OAc),f--Sn-0-Sn-I I 2 NH4'I CI, ,CI 0 I I -Sn-0-Sn-I1 I Fig. 1 Chemical fixation of the Cu complex on the surface cif SnO, J. MATER. CHEM., 1994, VOL. 4 Results and Discussion Sensing Properties of CuO-impregnated SnOz Sensors Fig. 2 shows the response transients of three CuO(5 wt.%)-SnO, sensors to H,S at 200°C.The response on turning on H2S was in many cases sluggish in contrast to the extraordinarily fast recovery on turning off H2S, which is one of the unique characteristics of the CuO-SnO, sensor. Up to 48 or 40ppm H,S in air, the response rates were very dependent on the SnO, powder used, resulting in the order of A >I3 >C; The sensor using Sn02-A had a 90% response time as short as 1 min, but that using Sn0,-C could not reach steady state within 30 min. The sensors using Sn0,-B and C have an induction period of a few minutes in the response transients [Fig. 2(d) and (e)]. The sensitivity to H2S at the steady state is seen to be unusually high, i.e. 16200 (48 ppm H,S) for A and MOO0 (40 ppm H2S) for B, although the value for C could not be determined.This is another unique character- istic of the CuO-SnO, sensor. The order in the response rate coincided with that of specific surface area. With the CuO loading fixed at 5 wt.%, an increase in specific surface area of SnO, means an increase in the dispersion of CuO. In this connection, the amount of CuO per unit surface area of SnO, (surface loading density, Y) appear to be important, which are 0.85, 3.2 and 8.2 mg CuO of SnO, surface, for the sensors using Sn0,-A, B and C, respectively. These results indicate that the response rate to H2S tends to increase with decreasing although the sensitivity to H2S can behave in the opposite manner. The response rates were also highly dependent on the H,S concentration, decreasing drastically with H2S concen-tration.As is seen from Fig. 2, for example, the sensor using Sn02-A could not respond to 15 or 9 ppm at the rates that are acceptable practically. It is suggested that a lower Y will be more advantageous for detecting dilute H,S. Sensing Properties of Chemically Fixed CuO-SnO, Sensors In order to attain a finer CuO dispersion, CuO was dispersed on the SnO, grains by means of the chemical fixation method. In this case, Cu complexes were first supported on the SnO, surface up to monolayer coverage or less and then converted to CuO particles during the subsequent calcination. In prin- ciple, this method should be superior to the impregnation method for attaining high CuO dispersion, especially at small CuO loadings. Fig.3 shows the response transients of the resulting sensors to 9 ppm H2S at 200 "C. The two sensors, CuO( 1.3 wt.%)-Sn0,-B and CuO(O.42 wt.%)-Sn0,-C, could respond sharply to 9 ppm H2S, with 90% response times of ca. 2 min, in marked contrast to Fig. 2. On the other hand, CuO( 3.3 wt.%-Sn0,-A could not attain a steady state within 30 min, despite the fact that it should have a surface loading density of CuO (Y)comparable to the other sensors. This H2S off (a) 48 ppm S= 16200 + tv HZS on H2Son Fig. 2 Response transients of CuO(5 wt.%)-SnO, sensors to H2S in air at 200 "C (impregnation method). (a)-(c) SnO, (A) (surface area 62 m2 g-'), H2S 48 ppm (4, 15 ppm (b),9 ppm (c); (4 SnO, (B)(16 m2 g-'), H2S 40 ppm; (e)SnO, (C)(6.4 m2 g-'), H,S 40 ppm (b)S=47100 Fig.3 Response transients of CuO-SnO, sensors to 9 ppm H,S in air at 200°C (fixation method): (a) CuO(3.3 wt.%)-SnO, (A), (b)CuO(1.3 wt.%)-SnO, (B), (c) CuO(O.42 wt.%)-SnO, (C) suggests that factors other than such as porosity of the sensor, are also important for the response rates. As estimated from the specific surface area of SnO,, the mean grain size of Sn0,-A is 4.8 nm while those of Sn0,-B and C are 11 and 27 nm, respectively. Thus the micropores as well as macro- pores of the sensors can be significantly different, depending on the crystalline state of the SnO,. The diffusion of H2S molecules into and/or inside the pores is thus affected by the crystalline state of Sn02, probably causing the slower response of the CuO-Sn0,-A sensor.Even with the CuO-Sn0,-B and C sensors. the response became increasingly' sluggish as H,S was diluted further. Lowering the operating temperature from 200 to 160°C increased the response rates slightly. Fig. 4 shows the response transients of the CuO(O.42 wt.%)-Sn0,-C sensor to varying concentrations of H2S at 160°C. The detection of H,S at 6ppm and above could be achieved in a few minutes, while that of 2ppm H,S took ca. 20min, indicating that further improvement is still necessary for detecting such dilute concen- trations of H,S. It was possible to increase the rate of response to dilute H2S by decreasing the CuO loading further. For example, Fig.5 shows the response transient of the CuO(O.091 wt.%)-Sn02-C sensor on turning on and off 1.2 ppm H2S at 160"C. On exposure of the sensor to H,S, the output voltage reached 90% of the full change in 4 min. The gas sensitivity of the sensor was reduced significantly because of the smaller CuO loading, but it was still as high as 130. Fig. 6 shows the stationary electrical resistances (R,) of three sensors having different CuO loadings under exposure to H,S as a function of H,S concentration. The logarithm of R, is linearly correlated with the logarithm of H2S concentration for each sensor, assuring that each is applicable for the detection of H,S in the respective concentration range. So far attention was mainly focused on the response rate. H2S Off H~S9 ppm S=49000 6 ppm S = 37000 al 0, ,c > c 2a c 0 4 H2Son Fig.4 Response transients of CuO(O.42 wt.%)-SnO, (C) sensor to H2S at 160 "C (fixation method) J. MATER. CHEM., 1994, VOL. 4 8.7 x lo3 12 5 min--Qc 0 1.1 x 106R Fig. 5 Response transient of CuO(O.091 wt.%)-SnO, (C) sensor to 1.2 ppm H,S at 160°C (fixation method) 106f 1 23 5 10 H2Sconc. (ppm) Fig. 6 Electrical resistances of CuO-Sn02 (C) sensors at 160 "C as a function of H,S concentration (fixation method). CuO loading (wt.%): 0,0.091; A, 0.15; I?, 0.42 It was found that the sensitivity to H,S (S),another important characteristic, correlated well with the surface loading density of CuO (Y).Fig.7 shows the sensitivity to 9 ppm H,S at 200 "C as a function of Y. S increases sharply with Y,obeying the following equation: log S=2.34 log Y +4.68 The correlation is seen to hold for both Sn0,-B and C, indicating that S is independent of the grain size of SnO,. In this way, an increase in Y brings about a sharp increase in S, while it is not favourable for the response kinetics to dilute 'O2I10' 1000 0.1 0.2 0.5 1 Y/mg m-2 Fig. 7 Sensitivity of CuO-SnO, sensors to 9 ppm H2S at 200°C as a function of the amount of CuO per unit SnO, surface area (Y) (fixation method) H2S as mentioned previously. Thus the optimum Y value as a compromise of these two factors depends on the range of H,S concentration to be detected. Influenceof CuO Dispersion on H2SSensitivity The H2S-sensing mechanism of CuO-SnO, sensors involves the reaction between CuO and H2S,'y9 as schematically shown in Fig.8. In air, SnO, grains with CuO dispersed on their surface are strongly depleted of electrons due to p-n junctions at CuO(p)/SnO,(n) interfaces. On exposure to H,S. the CuO is converted into CuS according to the reaction CuO +H,S +CuS +H,O resulting in the destruction of the p-n junctions and the reduction of the electron depletion layer. When H,S is turned off, CuS is quickly oxidized to CuO by cus +3/20, +cue +So2 thus restoring the p-n junctions. The whole electrical resist- ance (R)of the sensor is assumed to be mainly determined by the height of the potential barrier at the grain bchundaries.The conversion between CuO and CuS is thus accompanied by a drastic change of resistance between R, and Rg,giving rise to the extraordinarily high sensitivity of the (hO-pro- moted sensor to H,S. Based on such a sensing scheme, the dispersion of CuO particles on the SnO, surfacc will be critically important for the H2S sensitivity; CuO should be dispersed as finely as possible to increase the density of p-n junctions on the SnO, surface. Generally speaking, the chemi- cal fixation method can give finer CuO dispersion than the impregnation method; this is why the former method is advantageous for obtaining high H,S sensitivity at reduced CuO loadings. When the former method was adopted, the sensitivity to a fixed concentration of H2S was shown to correlate well with the amount of CuO per unit surface area of SnO, (Fig.7). This can be understood as reflecting the increase in the density of p-n junctions. H2SConcentratio2-dependent Resistance The electrical resistance of the sensor (R,)correlated well with the H,S concentration (Fig. 6). It is not very obvious how R, varies with H2S concentration on the basis of the above sensing mechanism. A possible interpretation is that the degree Fig. 8 Schematic drawings for CuO dispersed SnO, under :xposure to air (u) and H,S-air (b) J. MATER. CHEM.. 1994, VOL. 4 Fig. 9 Part of the microstructure of a sintered block type sensor using CuO-SnO, under exposure to H,S in air. CuO-dispersed SnO, grains coagulate to form secondary particles.CuO in the outer region of each secondary particle is sulfurized while CuO in the inner region is not of sulfurization of CuO particles on all SnO, grains changes with the H2S concentration. However, this fatally ignores the microstructure of the H,S-sensing entity. In the actual sensor, SnO, grains are coagulated to form large clusters (secondary particles) which are in contact to each other, as depicted in Fig. 9. H,S molecules should gain access to Sn02 grains by diffusion through macropores and micropores which are developed among and inside the secondary particles, respect- ively. Since H2S is very reactive with CuO, its concentration should decrease when H,S diffuses from the outer to the inner part of the microstructure.Such a concentration gradient of H,S would be especially steep inside the micropores. In other words, the sulfurization of CuO proceeds favourably on the SnO, grains that are located at the outer region of the secondary particles, the sulfurized region in the steady state becoming thicker with increasing H,S concentration. Such a heterogeneous progress of sulfurization depending on the H2S concentration is a more plausible cause for the H2S concentration-dependent resistance of the sensor. However, further investigations are necessary to explain the observed correlations quantitatively. Rates of Response to H2S The rate of response of the CuO-SnO, sensor to H,S was shown to deteriorate sharply when the H,S concentration dropped below a certain level which depended on the CuO loadings as well as on the specific surface area (or mean size) of the SnO, grains.To illustrate these kinetic properties more quantitatively, the times for 70% of the full response were evaluated from the response transients to H2S depicted in terms of the logarithm of electrical resistances for the three CuO-Sn0,-C sensors (Fig. 10). For the same CuO loading, the 70% response time is seen to increase almost linearly with a decreasing logarithm of H2S concentration, and the slope tends to become steeper as the CuO loading increases. This figure explains well why reduced CuO loading is necessary for detecting dilute H2S. What then determines the rate of response to H,S? Based on the scheme depicted in Fig.8 and 9, the rate of response involves not only the rate of sulfur- ization of CuO (or H,S consumption) but also the rate of 1 2 5 10 H2Sconc. (ppm) Fig. 10 70% response time of CuO-SnO, (C) sensors as a function of H,S concentration at 160"C. (fixation method 1. CuO loading (Wt.Yo): 0,0.091; A,0.14; 0,0.42 diffusion of H2S inside the sensor microstructure. The rate of response appears to reflect how quickly the H,S concentration profile inside the microstructure reaches a steady state. It seems that this process becomes quicker as the rate of diffusion of H,S increases relative to the rate of consumption of H2S. The former rate increases with H2S Concentration while the latter decreases with CuO loading, both changes resulting in the reduction of the 70% response times as observed in Fig.10. These considerations suggest the importance of sensor structure other than the sintered block type investigated in the present study. For example, thick or thin film type sensors may be more advantageous with regard to the detection of dilute H2S, because the diffusion path for H2S can be made far shorter. Conclusions The H2S-sensing properties of CuO-SnO, sensors are influ- enced strongly by the CuO loading amongst other factors. The response rates to dilute H2S increase with decreasing CuO loading, while the sensitivity to H2S decreases sharply with the amount of CuO per unit surface area of SnO,. Such characteristics can be understood well from the H2S-sensing scheme which involves the sulfurization of CuO with H2S and the diffusion of H2S inside the macro- and micro-pores of the sensors. References 1 V. Lantto and P. Romppainen, J. Electrochem. SOC., 1988, 135, 2550. 2 V. Lantto, P. Romppainen, T. S. Rantala and S. Leppavurori, Sensors Actuators B, 1991,4,451. 3 S. Kanefusa, M. Nitta and M. Haradome, J. Electrochem. Soc., 1985,132,1770. 4 S. Kanefusa, M. Nitta and M. Haradome, 7ech. Digest, 7th Chemical Sensor Symp., Saitama, Japan, 1988, p. 145. 5 T. Nakahara, Proc. Symp. Chem. Sensors, Honolulu, 1987, p. 55. 6 V. Lantto and J. Mizsei, Sensors Actuators B, 1991,5, 21. 7 P. J. Smith, J. F. Vetelino, R. S. Falconer and E. L. Wittman, Sensors Actuators B, 1993,13-14,264. 8 T. Maekawa, J. Tamaki, N. Miura and N. Yamazoe, Chem. Lett., 1991,575. 9 J. Tamaki, T. Maekawa, N. Miura and N. Yamazoe, Sensors Actuators B, 1992,9, 197. 10 T. Kaji, H. Ohno, T. Nakahara, N. Yamazoe and T. Seiyama, Nippon Kagaku Kaishi, 1980, 1088. Paper 4/02042A; Received 6th April, 1994