- Email: [email protected]
Humidity-insensitive and low oxygen dependence tungsten oxide gas sensors
Sensors and Actuators B 113 (2006) 265–271
Humidity-insensitive and low oxygen dependence tungsten oxide gas sensors Mana Sriyudthsak a,∗ , Sitthisuntorn Supothina b b
a Department of Electrical Engineering, Chulalongkorn University, Bangkok 10330, Thailand National Metal and Materials Technology Center, 114 Thailand Science Park, Klong Luang, Pathumthani12120, Thailand
Received 14 July 2004; received in revised form 20 February 2005; accepted 21 February 2005 Available online 1 April 2005
Abstract Gas sensing characteristics of WO3 powder and its physical properties under different heat treatment conditions have been investigated. The WO3 powder was synthesized by wet process from ammonium tungstate parapentahydrate and nitric solution. The precipitated product was then calcined at 300–800 ◦ C for 2–12 h. The physical properties of the products were characterized by using X-ray diffractometer (XRD), scanning electron microscope (SEM), and BET method. It was found that the crystallite size, particle size and surface area of the WO3 powders were in the range of 30–45 nm, 0.1–3.0 m and 1.2–3.7 m2 /g, respectively. Calcination at higher temperature and longer time led to the increase of particle size by more than 300%, and reduction in specific surface area by more than 60%. However, the crystallite size was found to increase only by ∼30% under identical heat treatment. These results inferred that such heat treatment had more profound effect on crystallite aggregation than on crystallite growth. Gas sensing measurement showed that the largest change of output voltage to both ethyl alcohol and ammonia was obtained from the sensor calcined at 600 ◦ C for 2 h, which had the highest surface area. However, the highest sensitivity which is defined as the ratio of sensor’s resistance in air to that in the sample gas, Rair /Rgas , was obtained from the sensor calcined at 600 ◦ C for 6 h due to its highest background resistance in air. Moreover, it was also found that the sensors were less sensitive to the oxygen content in the carrier gas and did not sensitive at all to water vapor. © 2005 Elsevier B.V. All rights reserved. Keywords: WO3 ; Humidity; Oxygen; Ammonia; Ethyl alcohol; Calcination
1. Introduction It is well known that electrical properties of metal oxide semiconductors were sensitive to a gas composition of surrounding atmosphere. By utilizing this phenomenon, tin oxide (SnO2 ) and zinc oxide (ZnO) semiconductors were first demonstrated as gas sensing devices in early 1960s [1–2]. Since then the semiconductor sensors based on this property have been studied and used extensively for the chemical and gas detection. It was found that most metal oxide gas sensors were sensitive to relative humidity and oxygen content in background environment, which, in many cases, can interfere the sensor performance, particularly in the equatorial zone, ∗
Corresponding author. Tel.: +66 2218 6517; fax: +66 2251 8991. E-mail address: [email protected] (M. Sriyudthsak).
0925-4005/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2005.02.057
which has high variation of humidity. Nevertheless, some oxides are so sensitive to humidity that they have been used as humidity sensor [3]. Therefore, it has been our prime interest to explore the materials, which, while show good response to targeted gases, are less sensitive to humidity. Tungsten oxide (WO3 ) was our first priority for this study since there have been many reports during the past decade on its sensing performance to many gases, such as nitrogen oxides [4–6] and sulfides [7]. The tungsten oxides have been prepared by various techniques, such as sol–gel [6], pyrolysis [7] and sputtering [8]. Like sensing behavior of other oxides, it was found that the tungsten oxides obtained from different preparation techniques resulted in different response characteristics. It was also known that the characteristics of the gas sensors depended strongly on base material as well as preparation conditions. In this paper, the response characteristics
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
266
of the tungsten oxide prepared from precipitation method to ethyl alcohol and ammonia were investigated to study the effects of crystallite size, particle size, specific surface area and initial electrical resistance on sensing performance. The effects of humidity and oxygen content in the carrier gas were also reported.
2. Experimental Fig. 1. Schematic illustration of the measuring circuit used in this study. Rs : sensor resistance; Rf : reference resistor used to adjust the baseline; Vin : input voltage and Vout : output voltage.
2.1. WO3 synthesis and characterizations The tungsten oxide (WO3 ) powders were synthesized by precipitation technique from the precursor containing ammonium tungstate parapentahydrate ((NH4 )10 W12 O41 ·5H2 O, Wako) and nitric acid (HNO3 , Carlo). A 0.1 M nitric acid was slowly added into a 3 mM ammonium tungstate solution pre-heated at 80 ◦ C. After reaction completion, the yellowish precipitate was allowed to settle, separated by filtration and then dried at 200 ◦ C for 2 h to remove the water. Finally, the dried precipitate was calcined in air at temperature ranging from 300 to 800 ◦ C for 2, 6 and 12 h. The crystal structures of the dried and calcined powders were identified using an X-ray diffractometer (XRD, JDX 3503). The crystallite size was estimated from peak broadening of the (2 0 0), (0 2 0) and (0 0 2) reflections using Scherrer approximation, which is defined as: D=
0.9λ B cos θ
(1)
˚ λ is the wavelength of where D is the crystallite size (A), ˚ B is the full width at half maximum the X-ray (1.5418 A), (radian) and θ is the Bragg angle (degree). The microstructure was observed using a scanning electron microscope (SEM, JSM-5410). The specific surface area was measured based on nitrogen adsorption by Brunauer, Emmett and Tellet (BET) method using gas chromatography. 2.2. Sensor fabrication and characterizations The calcined WO3 powders were formed into paste by mixing the powders with ethyl alcohol, de-ionized water and dispersing agent (Dispex® ). Thick-film sensors were formed by painting the paste onto the electrode coated on the glass substrate. The electrode was prepared by evaporating a 50 nm-thick titanium layer, followed by a 100 nm-thick platinum layer on the glass slide using electron beam evaporation (ULVAC, EBV-6DH). The fabricated sensors were installed in a measuring chamber, which was specially designed so that six sensors were measured simultaneously under the identical conditions. Responses of the sensors towards ethyl alcohol and ammonia were characterized at operating temperatures of 150–400 ◦ C. Nitrogen and oxygen with flow rates of 400 and 100 ml/min, respectively, were mixed together and used as carrier gas. For each measurement, a 50 l of either ethyl al-
cohol or ammonium solution (0.01–100% (v/v) in de-ionized water) was injected into the injection port where it was instantly vaporized and transported into the chamber having working volume of 452 cm3 . Conductivity change of the sensor was measured using an operational amplifier measuring circuit (Fig. 1) to prevent the loading effect, which frequently occurs in the conventional voltage divider circuit. Detailed merit of this circuitry approach has been previously reported elsewhere [9]. A bias voltage (Vin ) was fixed at 5 V. Note that Rs represents sensor’s resistance while Rf is a resistor used to adjust the baseline output voltage (Vout ). The output from the measuring circuit was sent to a PC and converted to a digital signal for monitoring and recording. Finally the effect of oxygen content in the carrier gas was also investigated by adjusting the flow rate ratio between nitrogen and oxygen gases at 0:500, 250:250 and 500:0 ml/min. In each measurement, the carrier gas was flowed into the chamber until the base line was stable before introducing the gas sample into the system. The sensor performance was compared by it sensitivity (S) which was defined as: S=
Rair Rgas
(2)
where Rair is the resistance of the sensor in air environment, and Rgas is the sensor’s resistance in gas sample. According to the illustration in Fig. 1, Rair = −
Rf Vin Vout-air
(3)
Rf Vin Vout-gas
(4)
and Rgas = −
where Vout-air and Vout-gas are the output voltage in air and the output voltage in gas, respectively. Hence, the sensitivity of the sensor could be derived directly from Eq. (5). S=
Vout-gas Vout-air
(5)
This equation shows that the calculated sensitivity can be obtained directly from the output of the measuring circuit and it is not necessary to calculate the resistance of the sensors.
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
267
Fig. 2. XRD patterns of the tungsten oxide calcined at various conditions; (a) 300 ◦ C, 2 h, (b) 600 ◦ C, 2 h, (c) 600 ◦ C, 6 h, (d) 600 ◦ C, 12 h, (e) 800 ◦ C, 2 h, (f) 800 ◦ C, 6 h and (g) 800 ◦ C 12 h. Fig. 5. Particle size of the WO3 powders calcined at 600 ◦ C and 800 ◦ C for 2, 6 and 12 h measured from the SEM images. ((䊉): 600 ◦ C; (): 800 ◦ C).
Fig. 3. Crystallite size derived from XRD peaks of the WO3 powders calcined at 600 and 800 ◦ C for 2, 6 and 12 h. (): 600 ◦ C (4 0 0); (): 600 ◦ C, (2 0 0);(): 800 ◦ C, (4 0 0); (): 800 ◦ C, (2 0 0).
Moreover, the calculated sensitivity does not depend on the resistance, Rf , in the measuring circuit. These are merits of our proposed measuring circuit.
3. Results and discussion 3.1. WO3 synthesis and characterizations Fig. 2 is XRD patterns of the precipitates calcined in air at various temperatures and times. The precipitate calcined at 300 ◦ C was identified as hydrated tungsten oxide (WO3 ·H2 O)
while the ones calcined at higher temperatures were anhydrous tungsten oxide (WO3 ). Moreover, it was observed from the relative intensities that increasing calcination temperature increased degree of crystallinity whereas calcination time in the rage of this study had no observable effect. It should be noted that the major crystalline planes of the prepared WO3 were (0 0 2), (0 2 0) and (2 0 0). Fig. 3 shows the crystallite size of the WO3 powders calcined at various conditions calculated from the broadening of the (2 0 0), (0 2 0), (2 0 0) and (4 0 0) peaks by performing curve fitting. It was found that the crystallite sizes derived from these planes are similar within experimental error. The crystallite sizes were in the range of 30–35 nm and 41–45 nm for the WO3 powders calcined at 600 and 800 ◦ C, respectively. It is obvious that calcination at 800 ◦ C caused a crystal growth by ∼30% compared to the powder calcined at 600 ◦ C, whereas prolong calcination (up to 12 h) essentially did not affect the crystal growth. Fig. 4 are SEM images of the WO3 powders calcined at 600 and 800 ◦ C for 2, 6 and 12 h. The particle sizes obtained by direct measurement from the SEM images are shown in Fig. 5. It was clearly observed that the particle size became larger with increasing calcination temperature and time, although the effect of calcination time was not significant at 600 ◦ C. For the powders calcined at 600 ◦ C, the particle size
Fig. 4. SEM images of the WO3 powders calcined at (a) 600 ◦ C, 2 h, (b) 600 ◦ C, 6 h, (c) 600 ◦ C, 12 h, (d) 800 ◦ C, 2 h, (e) 800 ◦ C, 6 h and (f) 800 ◦ C, 12 h.
268
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
ranged from 100 nm for 2 h calcination to 400 nm for 12 h calcinations (four times increase), while the particle size ranged from 300 nm to 3 m for the powders calcined at 800 ◦ C for 2 and 12 h, respectively (10 times increase). Note that under the heat treatment conditions used in this study, the particle growth were in the range of 4–10 times depending on the heating condition. For instance, at calcination time of 12 h, the particle size increased for 700% when the calcination temperature increased from 600 to 800 ◦ C. This particle growth was very high compared to the crystallite growth (∼30%) under the same heat treatment condition. By comparing the crystallite size and the particle size (i.e. Figs. 3 and 5), the particle sizes were 3–70 times larger than the crystallite size, suggesting that the particles formed by aggregation of very fine crystals. This particle growth caused reduction of specific surface area of the sensor for interacting with the sample gases. Also, since the range of the particle size was much wider than the range of crystallite size (i.e. 0.1–3 m compared to 30–45 nm), it inferred that heat treatment had more profound effect on crystallite aggregation than on crystallite growth.
Fig. 6. Specific surface area of the WO3 powders calcined at 600 and 800 ◦ C for 2, 6 and 12 h. ((): 600 ◦ C (calc.), (): 600 ◦ C (BET), (): 800 ◦ C (calc.), (): 800 ◦ C (BET)).
Fig. 6 shows the specific surface area of the WO3 powders calcined at various conditions. The specific surface area significantly decreased with increasing calcination temperature but only slightly decreased with increasing calcination time. It decreased from 3.7 m2 /g for the powder calcined at 600 ◦ C for 2 h to 1.1 m2 /g for the one calcined at 800 ◦ C for
Fig. 7. Response of the WO3 sensors to alcohol and ammonia at 400 ◦ C (a) WO3 600 ◦ C, 2 h (b) WO3 600 ◦ C, 6 h (c) WO3 600 ◦ C, 12 h (d) WO3 800 ◦ C, 2 h (e) WO3 800 ◦ C, 6 h (f) WO3 800 ◦ C, 12 h.
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
269
12 h. The specific surface area calculated from the particle size measured directly from SEM images, by assuming that the particles were spherical, is also shown. Comparing the calculated values of the powder calcined at 600 ◦ C for 2 h and at 800 ◦ C for 12 h, the particle size difference was more than 10 times, whereas the specific surface area difference was only about three times. Knowing from XRD results that the crystallite sizes of these powders were not much different, the specific surface area results implied that the particles were formed by loose aggregation of the crystallites. Then the measured surface area was attributed mainly to the crystallites, not to the particles. 3.2. Gas responses Fig. 7 shows typical responses of various WO3 sensors to ethyl alcohol and ammonia samples at operating temperature of 400 ◦ C. The sensors fabricated from the powders calcined at 600 ◦ C showed fast rise and recovery times whereas the ones fabricated from the powders calcined at 800 ◦ C essentially had no response or very low response to the sample even at high concentration. For the sensors fabricated from the powders calcined at 600 ◦ C, the rise time to reach the maximum value was about a few seconds while the 90% recovery time was ∼2–5 min, depending on the sample concentrations. It was obvious that the change of the output voltage of the sensors decreased with increasing calcination temperature and time of the WO3 powders, indicating that the output change related to some extend directly to the specific surface area of the WO3 particles. That is, the sensors having high surface area showed good responses whereas the ones had low surface area showed very low response to both ethyl alcohol and ammonia. Fig. 8 shows the sensitivities of the sensors to various concentrations of ethyl alcohol and ammonia liquid samples. It should be note that the values of concentrations indicated in the figure is the concentration of the liquid samples injected into the measuring chamber, not the final concentration of the sample in the gas phases. Considering the huge ratio of the measuring chamber volume (452 cm3 ) to the sample volume (50 l), the final concentration of the sample in the gas phase was expected to be in the range of 1–10,000 ppm. Figs. 9 and 10 show the sensitivity of the sensors to 10% (v/v) ethyl alcohol and 25% (v/v) ammonia samples at various operating temperatures. At some operating temperatures, the sensitivity of the sensor fabricated from the powder calcined at 600 ◦ C for 6 h was higher than that calcined for 2 h which had larger specific surface area (i.e. 3.7 versus 1.2 m2 /g). These results were observed in both ethyl alcohol and ammonia samples. These gas sensing results suggested that the specific surface area did not the only parameter determining the sensitivities of the sensors. As the operating temperature to obtain the highest sensitivity to ethyl alcohol (200 ◦ C) and ammonia (300 ◦ C) were different, respectively, then it could be used the same material to detect both gases by using different operating temperature.
Fig. 8. Sensitivities of the WO3 gas sensors to (a) 0–100% ethyl alcohol and (b) 0–25% ammonia at operating temperature of 200 ◦ C. ((䊉): 600 ◦ C, 2h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h).
Fig. 9. Sensitivities of the WO3 gas sensors to 10% (v/v) ethyl alcohol liquid sample at various operating temperatures. ((): 600 ◦ C, 2 h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h).
Fig. 10. Sensitivities of the WO3 gas sensors to 25% (v/v) ammonia liquid sample at various operating temperatures. ((): 600 ◦ C, 2 h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h).
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
270
Fig. 11. Resistance of the WO3 sensors in air. ((): 600 ◦ C, 2 h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h). Fig. 12. Effect of oxygen content in the carrier gas on the response of WO3 sensor to 1% (v/v) ethyl alcohol at operating temperature of 350 ◦ C.
According to the results in Fig. 7, it is important to note that the WO3 sensors prepared in this study did not sensitive to water vapor (DI as indicated in the figure). As mention earlier that the ethyl alcohol and ammonia samples were injected in an “aqueous solution” form. After being vaporized at the injection port, both alcohol or ammonia and water vapors were transported into the measuring chamber. As also shown in Fig. 8, the sensors had no response to deionized water. Therefore, the response curves shown were attributed only to interaction between the WO3 and ethyl alcohol or ammonia. This water vapor insensitive characteristic is very useful for gas detection in the present of high humidity. As shown in Fig. 7 that the highest changes in output voltage of the sensor were obtained from the sensor calcined at 600 ◦ C for 2 h, in which case the sensor had highest specific surface area. However, from the calculated sensitivities shown in Figs. 8–10, it was clearly seen that the highest sensitivity was obtained from the sensor calcined at 600 ◦ C for 6 h. The explanation to these results can be made by considering the sensor’s resistance (Rair ) or output voltage (Vout-air ) before exposed to gas as shown in Fig. 11. At the operating temperature used in this study (200 ◦ C), the sensors fabricated from the powder calcined at 600 ◦ C for 6 h somewhat had highest resistance or low Vout (see eq. (3)). Referring to the definition of the sensitivity described in eq. (1), it is therefore reasonable that this sensor had the highest sensitiv-
ity although the output change (i.e., V = Vout-gas − Vout-air ) was slightly smaller than that of the sensor calcined for 2 h. Table 1 shows an example for the calculation of the sensitivities and the response changes of the sensors operating at 200 ◦ C. However, in real application of the gas sensor, the highest change of the gas response, either the highest resistance change (R) or output change (V), is preferred than the high calculated sensitivity value (S). In this case, the sensor calcined at 600 ◦ C for 2 h might be more suitable in real application. Fig. 12 is the response of the sensor fabricated from the powder calcined at 600 ◦ C for 6 h at different carrier gas compositions to 1% (v/v) ethyl alcohol at operating temperature of 350 ◦ C. It is interesting that the rise time, recovery time and sensitivity were not obviously changed even the oxygen content in the carrier gas was changed from 0 to 100%. This result, together with the results shown in Fig. 8, implied that oxygen did not play a dominant role in oxidation–reduction with WO3 sensors during gas adsorption and desorption processes. This behavior was different from that of other oxide gas sensors [10]. It was believed that the observed response of the sensor might be due to the mechanism of the physical adsorption and desorption of the sample gas molecules on the surface of the WO3 . To proof this assumption, we are trying to confirm the interaction between the sample and the gas molecules by determining the output
Table 1 Calculation of the sensitivities and the response changes of the sensors operating at 200 ◦ C Sensor WO3 600 ◦ C, 2 h WO3 600 ◦ C, 6 h
Ethyl alcohol (% v/v)
Vout-air (V)
Vout-gas (V)
V (V)
Rair (k)
Rgas (k)
R (k)
S
1 10 100 1 10 100
0.038 0.044 0.075 0.009 0.021 0.038
0.139 0.842 2.009 0.095 0.686 1.257
0.101 0.798 1.934 0.086 0.665 1.219
132 114 67 556 238 132
36 6 2 53 7 4
96 108 64 503 231 128
3.7 19.1 26.8 10.6 32.7 33.1
Note: Vin = 5 V, Rf = 1 k, V = Vout-gas − Vout-air , R = Rair − Rgas , S = Vout-gas /Vout-air.
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
gas using mass spectrophotometer. The oxygen-independent nature of this WO3 sensor may be applied in low oxygen environment.
4. Conclusions The physical properties and gas sensing characteristics of the WO3 gas sensors prepared by precipitation method have been investigated. It was found that increasing calcination temperature caused crystal growth and aggregation resulting to larger particle size, and therefore reduction in the specific surface area. On the other hand, increasing calcination time, while causing the increase of the particle size, did not significantly affect the crystallite size. It was found that the output change of the sensor increased with increasing specific surface area of the powders. However, it was found that the sensitivity calculated from the resistance of the sensor in air to that in gas did not depend only on the specific surface area but also related to the base resistance of the sensors in air. It was also found that, unlike other oxide semiconductors, sensing behavior of the WO3 did not depend on oxygen content in the carrier gas. Moreover, they had no response to water vapor. These unique characteristics would make the sensors beneficial in environments having low oxygen content or having high variation in relative humidity. Since the highest sensitivity to ethyl alcohol and ammonia were obtained at different operating temperature (200 and 300 ◦ C, respectively), it could be used the same WO3 sensor for detecting both gases by selecting different operating temperatures.
Acknowledgements This research was supported by the National Metal and Materials Technology Center (MTEC), Thailand. The authors would like to acknowledge Professor Piyasarn Prasertthum, Dr. Tharathorn Mongkolsri and Ms.Sirinya Jaiharn for BET analysis, Mr. Kamolchai Mungkornrit for system assembly and Ms. Prathumporn Hinthao for some technical assistants.
271
References [1] T. Seiyama, A. Kato, K. Fujushi, M. Nagatani, A new detector for gaseous components using semiconductive thin films, Anal. Chem. 34 (1962) 1502f. [2] N. Taguchi, Japanese Patent 45-38200 (1970); N. Taguchi, Japanese Patent 47-38840 (1972); N. Taguchi; US patent 3644795 (1970); UK Patent 1280809 (1970). [3] B. Kulwicki, Humidity Sensors, J. Am. Ceram. Soc. 74 (4) (1991) 697–708. [4] A. Tomchenko, V. Khatko, I. Emelianov, WO3 thick-film gas sensors, Sens. Actuators B 46 (1998) 8–14. [5] D. Lee, S. Han, J. Huh, D. Lee, Nitrogen oxides-sensing characteristics of WO3-based nanocrystalline thick film gas sensor, Sens. Actuators B 60 (1999) 57–63. [6] Y. Choi, G. Sakai, K. Shimanoe, N. Miura, N. Yamazoe, Wet process-prepared thick films of WO3 for NO2 sensing, Sens. Actuators B 95 (2003) 258–265. [7] M. Ando, S. Suto, T. Suzuki, T. Tsuchida, C. Nakayama, N. Miura, N. Yamazoe, H2 S and CH3 SH using a thick film of gold-loaded tungsten oxide, Chem. Lett. (1994) 335–338. [8] H. Voit, US. Patent 4,581,204 (1986). [9] A. Teeramongkolrasmee, M. Sriyudthsak, Problems in gas sensor measuring circuit and proposal of new circuit, Sens. Mater. 11 (3) (1999) 149–162. [10] M. Sriyudthsak, L. Promsong, S. Panyakeow, Effect of carrier gas on response of oxide semiconductor gas sensor, Sens. Actuators B 13-14 (1993) 139–142.
Biographies Mana Sriyudthsak received BEng, MEng, and DEng. degrees in electrical engineering in 1984, 1986 and 1989, respectively, from Tokyo Institute of Technology, Japan. He is currently an associate professor at Chulalongkorn University, Thailand. His current research interest includes semiconductor gas sensors and biosensors. Sitthisuntorn Supothina received a BSc degree in chemistry from Khon Khaen University in 1993, MS and PhD degrees in materials science and engineering from Case Western Reserve University, USA, in 1995 and 1998, respectively. He is currently a researcher in ceramic division of the National Metal and Materials Technology Center, Thailand. His current research interest includes preparation of nanoparticles for sensor and photocatalyst applications.
Humidity-insensitive and low oxygen dependence tungsten oxide gas sensors Mana Sriyudthsak a,∗ , Sitthisuntorn Supothina b b
a Department of Electrical Engineering, Chulalongkorn University, Bangkok 10330, Thailand National Metal and Materials Technology Center, 114 Thailand Science Park, Klong Luang, Pathumthani12120, Thailand
Received 14 July 2004; received in revised form 20 February 2005; accepted 21 February 2005 Available online 1 April 2005
Abstract Gas sensing characteristics of WO3 powder and its physical properties under different heat treatment conditions have been investigated. The WO3 powder was synthesized by wet process from ammonium tungstate parapentahydrate and nitric solution. The precipitated product was then calcined at 300–800 ◦ C for 2–12 h. The physical properties of the products were characterized by using X-ray diffractometer (XRD), scanning electron microscope (SEM), and BET method. It was found that the crystallite size, particle size and surface area of the WO3 powders were in the range of 30–45 nm, 0.1–3.0 m and 1.2–3.7 m2 /g, respectively. Calcination at higher temperature and longer time led to the increase of particle size by more than 300%, and reduction in specific surface area by more than 60%. However, the crystallite size was found to increase only by ∼30% under identical heat treatment. These results inferred that such heat treatment had more profound effect on crystallite aggregation than on crystallite growth. Gas sensing measurement showed that the largest change of output voltage to both ethyl alcohol and ammonia was obtained from the sensor calcined at 600 ◦ C for 2 h, which had the highest surface area. However, the highest sensitivity which is defined as the ratio of sensor’s resistance in air to that in the sample gas, Rair /Rgas , was obtained from the sensor calcined at 600 ◦ C for 6 h due to its highest background resistance in air. Moreover, it was also found that the sensors were less sensitive to the oxygen content in the carrier gas and did not sensitive at all to water vapor. © 2005 Elsevier B.V. All rights reserved. Keywords: WO3 ; Humidity; Oxygen; Ammonia; Ethyl alcohol; Calcination
1. Introduction It is well known that electrical properties of metal oxide semiconductors were sensitive to a gas composition of surrounding atmosphere. By utilizing this phenomenon, tin oxide (SnO2 ) and zinc oxide (ZnO) semiconductors were first demonstrated as gas sensing devices in early 1960s [1–2]. Since then the semiconductor sensors based on this property have been studied and used extensively for the chemical and gas detection. It was found that most metal oxide gas sensors were sensitive to relative humidity and oxygen content in background environment, which, in many cases, can interfere the sensor performance, particularly in the equatorial zone, ∗
Corresponding author. Tel.: +66 2218 6517; fax: +66 2251 8991. E-mail address: [email protected] (M. Sriyudthsak).
0925-4005/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2005.02.057
which has high variation of humidity. Nevertheless, some oxides are so sensitive to humidity that they have been used as humidity sensor [3]. Therefore, it has been our prime interest to explore the materials, which, while show good response to targeted gases, are less sensitive to humidity. Tungsten oxide (WO3 ) was our first priority for this study since there have been many reports during the past decade on its sensing performance to many gases, such as nitrogen oxides [4–6] and sulfides [7]. The tungsten oxides have been prepared by various techniques, such as sol–gel [6], pyrolysis [7] and sputtering [8]. Like sensing behavior of other oxides, it was found that the tungsten oxides obtained from different preparation techniques resulted in different response characteristics. It was also known that the characteristics of the gas sensors depended strongly on base material as well as preparation conditions. In this paper, the response characteristics
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
266
of the tungsten oxide prepared from precipitation method to ethyl alcohol and ammonia were investigated to study the effects of crystallite size, particle size, specific surface area and initial electrical resistance on sensing performance. The effects of humidity and oxygen content in the carrier gas were also reported.
2. Experimental Fig. 1. Schematic illustration of the measuring circuit used in this study. Rs : sensor resistance; Rf : reference resistor used to adjust the baseline; Vin : input voltage and Vout : output voltage.
2.1. WO3 synthesis and characterizations The tungsten oxide (WO3 ) powders were synthesized by precipitation technique from the precursor containing ammonium tungstate parapentahydrate ((NH4 )10 W12 O41 ·5H2 O, Wako) and nitric acid (HNO3 , Carlo). A 0.1 M nitric acid was slowly added into a 3 mM ammonium tungstate solution pre-heated at 80 ◦ C. After reaction completion, the yellowish precipitate was allowed to settle, separated by filtration and then dried at 200 ◦ C for 2 h to remove the water. Finally, the dried precipitate was calcined in air at temperature ranging from 300 to 800 ◦ C for 2, 6 and 12 h. The crystal structures of the dried and calcined powders were identified using an X-ray diffractometer (XRD, JDX 3503). The crystallite size was estimated from peak broadening of the (2 0 0), (0 2 0) and (0 0 2) reflections using Scherrer approximation, which is defined as: D=
0.9λ B cos θ
(1)
˚ λ is the wavelength of where D is the crystallite size (A), ˚ B is the full width at half maximum the X-ray (1.5418 A), (radian) and θ is the Bragg angle (degree). The microstructure was observed using a scanning electron microscope (SEM, JSM-5410). The specific surface area was measured based on nitrogen adsorption by Brunauer, Emmett and Tellet (BET) method using gas chromatography. 2.2. Sensor fabrication and characterizations The calcined WO3 powders were formed into paste by mixing the powders with ethyl alcohol, de-ionized water and dispersing agent (Dispex® ). Thick-film sensors were formed by painting the paste onto the electrode coated on the glass substrate. The electrode was prepared by evaporating a 50 nm-thick titanium layer, followed by a 100 nm-thick platinum layer on the glass slide using electron beam evaporation (ULVAC, EBV-6DH). The fabricated sensors were installed in a measuring chamber, which was specially designed so that six sensors were measured simultaneously under the identical conditions. Responses of the sensors towards ethyl alcohol and ammonia were characterized at operating temperatures of 150–400 ◦ C. Nitrogen and oxygen with flow rates of 400 and 100 ml/min, respectively, were mixed together and used as carrier gas. For each measurement, a 50 l of either ethyl al-
cohol or ammonium solution (0.01–100% (v/v) in de-ionized water) was injected into the injection port where it was instantly vaporized and transported into the chamber having working volume of 452 cm3 . Conductivity change of the sensor was measured using an operational amplifier measuring circuit (Fig. 1) to prevent the loading effect, which frequently occurs in the conventional voltage divider circuit. Detailed merit of this circuitry approach has been previously reported elsewhere [9]. A bias voltage (Vin ) was fixed at 5 V. Note that Rs represents sensor’s resistance while Rf is a resistor used to adjust the baseline output voltage (Vout ). The output from the measuring circuit was sent to a PC and converted to a digital signal for monitoring and recording. Finally the effect of oxygen content in the carrier gas was also investigated by adjusting the flow rate ratio between nitrogen and oxygen gases at 0:500, 250:250 and 500:0 ml/min. In each measurement, the carrier gas was flowed into the chamber until the base line was stable before introducing the gas sample into the system. The sensor performance was compared by it sensitivity (S) which was defined as: S=
Rair Rgas
(2)
where Rair is the resistance of the sensor in air environment, and Rgas is the sensor’s resistance in gas sample. According to the illustration in Fig. 1, Rair = −
Rf Vin Vout-air
(3)
Rf Vin Vout-gas
(4)
and Rgas = −
where Vout-air and Vout-gas are the output voltage in air and the output voltage in gas, respectively. Hence, the sensitivity of the sensor could be derived directly from Eq. (5). S=
Vout-gas Vout-air
(5)
This equation shows that the calculated sensitivity can be obtained directly from the output of the measuring circuit and it is not necessary to calculate the resistance of the sensors.
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
267
Fig. 2. XRD patterns of the tungsten oxide calcined at various conditions; (a) 300 ◦ C, 2 h, (b) 600 ◦ C, 2 h, (c) 600 ◦ C, 6 h, (d) 600 ◦ C, 12 h, (e) 800 ◦ C, 2 h, (f) 800 ◦ C, 6 h and (g) 800 ◦ C 12 h. Fig. 5. Particle size of the WO3 powders calcined at 600 ◦ C and 800 ◦ C for 2, 6 and 12 h measured from the SEM images. ((䊉): 600 ◦ C; (): 800 ◦ C).
Fig. 3. Crystallite size derived from XRD peaks of the WO3 powders calcined at 600 and 800 ◦ C for 2, 6 and 12 h. (): 600 ◦ C (4 0 0); (): 600 ◦ C, (2 0 0);(): 800 ◦ C, (4 0 0); (): 800 ◦ C, (2 0 0).
Moreover, the calculated sensitivity does not depend on the resistance, Rf , in the measuring circuit. These are merits of our proposed measuring circuit.
3. Results and discussion 3.1. WO3 synthesis and characterizations Fig. 2 is XRD patterns of the precipitates calcined in air at various temperatures and times. The precipitate calcined at 300 ◦ C was identified as hydrated tungsten oxide (WO3 ·H2 O)
while the ones calcined at higher temperatures were anhydrous tungsten oxide (WO3 ). Moreover, it was observed from the relative intensities that increasing calcination temperature increased degree of crystallinity whereas calcination time in the rage of this study had no observable effect. It should be noted that the major crystalline planes of the prepared WO3 were (0 0 2), (0 2 0) and (2 0 0). Fig. 3 shows the crystallite size of the WO3 powders calcined at various conditions calculated from the broadening of the (2 0 0), (0 2 0), (2 0 0) and (4 0 0) peaks by performing curve fitting. It was found that the crystallite sizes derived from these planes are similar within experimental error. The crystallite sizes were in the range of 30–35 nm and 41–45 nm for the WO3 powders calcined at 600 and 800 ◦ C, respectively. It is obvious that calcination at 800 ◦ C caused a crystal growth by ∼30% compared to the powder calcined at 600 ◦ C, whereas prolong calcination (up to 12 h) essentially did not affect the crystal growth. Fig. 4 are SEM images of the WO3 powders calcined at 600 and 800 ◦ C for 2, 6 and 12 h. The particle sizes obtained by direct measurement from the SEM images are shown in Fig. 5. It was clearly observed that the particle size became larger with increasing calcination temperature and time, although the effect of calcination time was not significant at 600 ◦ C. For the powders calcined at 600 ◦ C, the particle size
Fig. 4. SEM images of the WO3 powders calcined at (a) 600 ◦ C, 2 h, (b) 600 ◦ C, 6 h, (c) 600 ◦ C, 12 h, (d) 800 ◦ C, 2 h, (e) 800 ◦ C, 6 h and (f) 800 ◦ C, 12 h.
268
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
ranged from 100 nm for 2 h calcination to 400 nm for 12 h calcinations (four times increase), while the particle size ranged from 300 nm to 3 m for the powders calcined at 800 ◦ C for 2 and 12 h, respectively (10 times increase). Note that under the heat treatment conditions used in this study, the particle growth were in the range of 4–10 times depending on the heating condition. For instance, at calcination time of 12 h, the particle size increased for 700% when the calcination temperature increased from 600 to 800 ◦ C. This particle growth was very high compared to the crystallite growth (∼30%) under the same heat treatment condition. By comparing the crystallite size and the particle size (i.e. Figs. 3 and 5), the particle sizes were 3–70 times larger than the crystallite size, suggesting that the particles formed by aggregation of very fine crystals. This particle growth caused reduction of specific surface area of the sensor for interacting with the sample gases. Also, since the range of the particle size was much wider than the range of crystallite size (i.e. 0.1–3 m compared to 30–45 nm), it inferred that heat treatment had more profound effect on crystallite aggregation than on crystallite growth.
Fig. 6. Specific surface area of the WO3 powders calcined at 600 and 800 ◦ C for 2, 6 and 12 h. ((): 600 ◦ C (calc.), (): 600 ◦ C (BET), (): 800 ◦ C (calc.), (): 800 ◦ C (BET)).
Fig. 6 shows the specific surface area of the WO3 powders calcined at various conditions. The specific surface area significantly decreased with increasing calcination temperature but only slightly decreased with increasing calcination time. It decreased from 3.7 m2 /g for the powder calcined at 600 ◦ C for 2 h to 1.1 m2 /g for the one calcined at 800 ◦ C for
Fig. 7. Response of the WO3 sensors to alcohol and ammonia at 400 ◦ C (a) WO3 600 ◦ C, 2 h (b) WO3 600 ◦ C, 6 h (c) WO3 600 ◦ C, 12 h (d) WO3 800 ◦ C, 2 h (e) WO3 800 ◦ C, 6 h (f) WO3 800 ◦ C, 12 h.
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
269
12 h. The specific surface area calculated from the particle size measured directly from SEM images, by assuming that the particles were spherical, is also shown. Comparing the calculated values of the powder calcined at 600 ◦ C for 2 h and at 800 ◦ C for 12 h, the particle size difference was more than 10 times, whereas the specific surface area difference was only about three times. Knowing from XRD results that the crystallite sizes of these powders were not much different, the specific surface area results implied that the particles were formed by loose aggregation of the crystallites. Then the measured surface area was attributed mainly to the crystallites, not to the particles. 3.2. Gas responses Fig. 7 shows typical responses of various WO3 sensors to ethyl alcohol and ammonia samples at operating temperature of 400 ◦ C. The sensors fabricated from the powders calcined at 600 ◦ C showed fast rise and recovery times whereas the ones fabricated from the powders calcined at 800 ◦ C essentially had no response or very low response to the sample even at high concentration. For the sensors fabricated from the powders calcined at 600 ◦ C, the rise time to reach the maximum value was about a few seconds while the 90% recovery time was ∼2–5 min, depending on the sample concentrations. It was obvious that the change of the output voltage of the sensors decreased with increasing calcination temperature and time of the WO3 powders, indicating that the output change related to some extend directly to the specific surface area of the WO3 particles. That is, the sensors having high surface area showed good responses whereas the ones had low surface area showed very low response to both ethyl alcohol and ammonia. Fig. 8 shows the sensitivities of the sensors to various concentrations of ethyl alcohol and ammonia liquid samples. It should be note that the values of concentrations indicated in the figure is the concentration of the liquid samples injected into the measuring chamber, not the final concentration of the sample in the gas phases. Considering the huge ratio of the measuring chamber volume (452 cm3 ) to the sample volume (50 l), the final concentration of the sample in the gas phase was expected to be in the range of 1–10,000 ppm. Figs. 9 and 10 show the sensitivity of the sensors to 10% (v/v) ethyl alcohol and 25% (v/v) ammonia samples at various operating temperatures. At some operating temperatures, the sensitivity of the sensor fabricated from the powder calcined at 600 ◦ C for 6 h was higher than that calcined for 2 h which had larger specific surface area (i.e. 3.7 versus 1.2 m2 /g). These results were observed in both ethyl alcohol and ammonia samples. These gas sensing results suggested that the specific surface area did not the only parameter determining the sensitivities of the sensors. As the operating temperature to obtain the highest sensitivity to ethyl alcohol (200 ◦ C) and ammonia (300 ◦ C) were different, respectively, then it could be used the same material to detect both gases by using different operating temperature.
Fig. 8. Sensitivities of the WO3 gas sensors to (a) 0–100% ethyl alcohol and (b) 0–25% ammonia at operating temperature of 200 ◦ C. ((䊉): 600 ◦ C, 2h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h).
Fig. 9. Sensitivities of the WO3 gas sensors to 10% (v/v) ethyl alcohol liquid sample at various operating temperatures. ((): 600 ◦ C, 2 h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h).
Fig. 10. Sensitivities of the WO3 gas sensors to 25% (v/v) ammonia liquid sample at various operating temperatures. ((): 600 ◦ C, 2 h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h).
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
270
Fig. 11. Resistance of the WO3 sensors in air. ((): 600 ◦ C, 2 h; (): 600 ◦ C, 6 h; (): 600 ◦ C, 12 h). Fig. 12. Effect of oxygen content in the carrier gas on the response of WO3 sensor to 1% (v/v) ethyl alcohol at operating temperature of 350 ◦ C.
According to the results in Fig. 7, it is important to note that the WO3 sensors prepared in this study did not sensitive to water vapor (DI as indicated in the figure). As mention earlier that the ethyl alcohol and ammonia samples were injected in an “aqueous solution” form. After being vaporized at the injection port, both alcohol or ammonia and water vapors were transported into the measuring chamber. As also shown in Fig. 8, the sensors had no response to deionized water. Therefore, the response curves shown were attributed only to interaction between the WO3 and ethyl alcohol or ammonia. This water vapor insensitive characteristic is very useful for gas detection in the present of high humidity. As shown in Fig. 7 that the highest changes in output voltage of the sensor were obtained from the sensor calcined at 600 ◦ C for 2 h, in which case the sensor had highest specific surface area. However, from the calculated sensitivities shown in Figs. 8–10, it was clearly seen that the highest sensitivity was obtained from the sensor calcined at 600 ◦ C for 6 h. The explanation to these results can be made by considering the sensor’s resistance (Rair ) or output voltage (Vout-air ) before exposed to gas as shown in Fig. 11. At the operating temperature used in this study (200 ◦ C), the sensors fabricated from the powder calcined at 600 ◦ C for 6 h somewhat had highest resistance or low Vout (see eq. (3)). Referring to the definition of the sensitivity described in eq. (1), it is therefore reasonable that this sensor had the highest sensitiv-
ity although the output change (i.e., V = Vout-gas − Vout-air ) was slightly smaller than that of the sensor calcined for 2 h. Table 1 shows an example for the calculation of the sensitivities and the response changes of the sensors operating at 200 ◦ C. However, in real application of the gas sensor, the highest change of the gas response, either the highest resistance change (R) or output change (V), is preferred than the high calculated sensitivity value (S). In this case, the sensor calcined at 600 ◦ C for 2 h might be more suitable in real application. Fig. 12 is the response of the sensor fabricated from the powder calcined at 600 ◦ C for 6 h at different carrier gas compositions to 1% (v/v) ethyl alcohol at operating temperature of 350 ◦ C. It is interesting that the rise time, recovery time and sensitivity were not obviously changed even the oxygen content in the carrier gas was changed from 0 to 100%. This result, together with the results shown in Fig. 8, implied that oxygen did not play a dominant role in oxidation–reduction with WO3 sensors during gas adsorption and desorption processes. This behavior was different from that of other oxide gas sensors [10]. It was believed that the observed response of the sensor might be due to the mechanism of the physical adsorption and desorption of the sample gas molecules on the surface of the WO3 . To proof this assumption, we are trying to confirm the interaction between the sample and the gas molecules by determining the output
Table 1 Calculation of the sensitivities and the response changes of the sensors operating at 200 ◦ C Sensor WO3 600 ◦ C, 2 h WO3 600 ◦ C, 6 h
Ethyl alcohol (% v/v)
Vout-air (V)
Vout-gas (V)
V (V)
Rair (k)
Rgas (k)
R (k)
S
1 10 100 1 10 100
0.038 0.044 0.075 0.009 0.021 0.038
0.139 0.842 2.009 0.095 0.686 1.257
0.101 0.798 1.934 0.086 0.665 1.219
132 114 67 556 238 132
36 6 2 53 7 4
96 108 64 503 231 128
3.7 19.1 26.8 10.6 32.7 33.1
Note: Vin = 5 V, Rf = 1 k, V = Vout-gas − Vout-air , R = Rair − Rgas , S = Vout-gas /Vout-air.
M. Sriyudthsak, S. Supothina / Sensors and Actuators B 113 (2006) 265–271
gas using mass spectrophotometer. The oxygen-independent nature of this WO3 sensor may be applied in low oxygen environment.
4. Conclusions The physical properties and gas sensing characteristics of the WO3 gas sensors prepared by precipitation method have been investigated. It was found that increasing calcination temperature caused crystal growth and aggregation resulting to larger particle size, and therefore reduction in the specific surface area. On the other hand, increasing calcination time, while causing the increase of the particle size, did not significantly affect the crystallite size. It was found that the output change of the sensor increased with increasing specific surface area of the powders. However, it was found that the sensitivity calculated from the resistance of the sensor in air to that in gas did not depend only on the specific surface area but also related to the base resistance of the sensors in air. It was also found that, unlike other oxide semiconductors, sensing behavior of the WO3 did not depend on oxygen content in the carrier gas. Moreover, they had no response to water vapor. These unique characteristics would make the sensors beneficial in environments having low oxygen content or having high variation in relative humidity. Since the highest sensitivity to ethyl alcohol and ammonia were obtained at different operating temperature (200 and 300 ◦ C, respectively), it could be used the same WO3 sensor for detecting both gases by selecting different operating temperatures.
Acknowledgements This research was supported by the National Metal and Materials Technology Center (MTEC), Thailand. The authors would like to acknowledge Professor Piyasarn Prasertthum, Dr. Tharathorn Mongkolsri and Ms.Sirinya Jaiharn for BET analysis, Mr. Kamolchai Mungkornrit for system assembly and Ms. Prathumporn Hinthao for some technical assistants.
271
References [1] T. Seiyama, A. Kato, K. Fujushi, M. Nagatani, A new detector for gaseous components using semiconductive thin films, Anal. Chem. 34 (1962) 1502f. [2] N. Taguchi, Japanese Patent 45-38200 (1970); N. Taguchi, Japanese Patent 47-38840 (1972); N. Taguchi; US patent 3644795 (1970); UK Patent 1280809 (1970). [3] B. Kulwicki, Humidity Sensors, J. Am. Ceram. Soc. 74 (4) (1991) 697–708. [4] A. Tomchenko, V. Khatko, I. Emelianov, WO3 thick-film gas sensors, Sens. Actuators B 46 (1998) 8–14. [5] D. Lee, S. Han, J. Huh, D. Lee, Nitrogen oxides-sensing characteristics of WO3-based nanocrystalline thick film gas sensor, Sens. Actuators B 60 (1999) 57–63. [6] Y. Choi, G. Sakai, K. Shimanoe, N. Miura, N. Yamazoe, Wet process-prepared thick films of WO3 for NO2 sensing, Sens. Actuators B 95 (2003) 258–265. [7] M. Ando, S. Suto, T. Suzuki, T. Tsuchida, C. Nakayama, N. Miura, N. Yamazoe, H2 S and CH3 SH using a thick film of gold-loaded tungsten oxide, Chem. Lett. (1994) 335–338. [8] H. Voit, US. Patent 4,581,204 (1986). [9] A. Teeramongkolrasmee, M. Sriyudthsak, Problems in gas sensor measuring circuit and proposal of new circuit, Sens. Mater. 11 (3) (1999) 149–162. [10] M. Sriyudthsak, L. Promsong, S. Panyakeow, Effect of carrier gas on response of oxide semiconductor gas sensor, Sens. Actuators B 13-14 (1993) 139–142.
Biographies Mana Sriyudthsak received BEng, MEng, and DEng. degrees in electrical engineering in 1984, 1986 and 1989, respectively, from Tokyo Institute of Technology, Japan. He is currently an associate professor at Chulalongkorn University, Thailand. His current research interest includes semiconductor gas sensors and biosensors. Sitthisuntorn Supothina received a BSc degree in chemistry from Khon Khaen University in 1993, MS and PhD degrees in materials science and engineering from Case Western Reserve University, USA, in 1995 and 1998, respectively. He is currently a researcher in ceramic division of the National Metal and Materials Technology Center, Thailand. His current research interest includes preparation of nanoparticles for sensor and photocatalyst applications.