Sensors and Actuators B 101 (2004) 107–111 Wet process-based fabrication of WO 3 thin film for NO 2 detection Yong-Gyu Choi, Go Sakai, Kengo Shimanoe, Noboru Yamazoe ∗ Department of Materials Science, Faculty of Engineering Sciences, Kyushu University, Kasuga-shi, Fukuoka 816-8580, Japan Received 12 November 2003; received in revised form 6 February 2004; accepted 7 February 2004 Available online 8 April 2004 Abstract It was tried to fabricate a WO 3 thin film device through a wet process starting from an aqueous sol of WO 3 ·2H 2 O. When mixed with polyethylene glycol (PEG, molecular weight 6000), the sol was made compatible to spin coating on an alumina substrate and the coating could be converted into a WO 3 thin film by calcination at 300 ◦ C for 2 h. Starting with a typical coating dispersion containing WO 3 ·2H 2 O by 5 mass% on WO 3 basis and 2 mass% PEG, the WO 3 film obtained was 450 nm in mean thickness. The film was a slim pack of square plates, each of which was a stack of thin lamellar crystals of WO 3 . The device was sensitive enough to detect 50 ppb NO 2 in air at 200 or 250 ◦ C, although the response and recovery transients were rather sluggish. Unexpectedly, the transients were found to be sharpened drastically in humid atmosphere, while sensor response (sensitivity) to NO 2 was hardly degraded with humidity. © 2004 Elsevier B.V. All rights reserved. Keywords: Tungsten oxide; Gas sensor; NO 2 ; Thin film; Sol 1. Introduction For semiconductor gas sensors, WO 3 is an important base oxide which exhibits high sensitivity to non-hydrocarbon gases, like NO 2 and NH 3 [1–3]. So far various methods have been adopted for preparing WO 3 as a sensor material, in- cluding pyrolysis of (NH 4 ) 10 W 12 O 41 ·5H 2 O [1–3], sputter- ing or evaporation from a source of WO 3 [4–9], and sol–gel processes starting from W-alkoxide [10]. However, it is dif- ficult to control the microstructure of WO 3 -based devices, i.e. the shape, size and stacking of WO 3 particles included, by these methods. Recently, we reported that an aqueous colloidal dispersion (sol) of WO 3 ·2H 2 O could be prepared though wet processes starting from an ion-exchange reac- tion of Na 2 WO 4 [11]. The sol could be converted into a gel by centrifuge, making it possible to fabricate a thick film by a screen-printing method. The resulting film of WO 3 was found to exhibit fairly excellent NO 2 sensing properties if the sol had been subjected to ultrasonic or centrifugal treatments under proper conditions, which affected the mor- phology of WO 3 crystals [12,13]. Apart from thick films, it is also of interest to fabricate thin films of WO 3 from the WO 3 ·2H 2 O sol. Unfortunately, this has been postponed be- cause it was difficult to obtain a uniform thin film from the sol by a spin-coating method. However, it has been found ∗ Corresponding author. Tel.: +81-92-583-7537; fax: +81-92-583-7538. E-mail address: yamazoe@mm.kyushu-u.ac.jp (N. Yamazoe). that, when mixed with an organic binder (polyethylene gly- col, PEG), the sol gives a thin film on an alumina substrate. This paper aims at reporting morphology and NO 2 sensing properties of WO 3 thin films thus obtained. 2. Experimental A colloidal dispersion (sol) of WO 3 ·2H 2 O was prepared in the same way as reported elsewhere [11]. An aqueous so- lution of Na 2 WO 4 (0.15 M) was let to pass through a column packed with protonated cation exchange resin (Diaion SK 1B). The effluent was kept standing for 3 days before it de- posited a gel. The gel, collected by decantation, was washed with deionized water, and recollected by centrifuge and de- cantation. The washed gel was dispersed in deionized water again to form a WO 3 ·2H 2 O sol. The sol contained colloidal particles of crystalline WO 3 ·2H 2 O of about 30 nm in size. The content of WO 3 ·2H 2 O in the sol was set to be 5mass% on the WO 3 basis unless otherwise noted. PEG with aver- age molecular weight of 6000 was added to the above sol by 2mass% unless otherwise noted to obtain a spin-coating dispersion. The dispersion was spin coated on an alumina substrate attached with a pair of interdigited gold electrodes (300 ␮m in separation between electrodes) under the rotation speed of 1500rpm. The obtained thin film of WO 3 ·2H 2 O was calcined at 300 ◦ C for 2 h for conversion of WO 3 ·2H 2 O to WO 3 as well as sintering. Gas sensing experiments were carried out in a conventional flow apparatus equipped with a 0925-4005/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2004.02.031 108 Y G. Choi et al. /Sensors and Actuators B 101 (2004) 107–111 heating facility at a gas flow rate of 100cm 3 /min. The con- centration of NO 2 was varied between 10 and 1000 ppb by diluting a parent NO 2 gas (1 ppm in dry air) with dry air. When necessary, part of the air was humidified by babbling through water phase. 3. Results and discussion 3.1. Morphology of WO 3 thin films Thin films of WO 3 could be derived from the WO 3 ·2H 2 O sol added with PEG (2 wt.%) by one time spin coating and calcination at 300 ◦ C for 2 h. Fig. 1 shows SEM images of the thin films obtained. The films were an irregular packing of square plates of WO 3 , 0.5–1␮m in width and 0.2–0.5 ␮m in thickness. Each square plate was a stack of thin plates (lamellae) of less than 100 nm in thickness. This morphology was unchanged regardless of the WO 3 ·2H 2 O contents in the spin-coating dispersion, i.e. 1mass% Fig. 1a and 5 mass% Fig. 1b. Actually the outlook of WO 3 crystals, featured by lamellar structure, was essentially the same as what we ob- served for the WO 3 thick films prepared by a screen-printing method [12,13]. The lamellar structure has its origin in the crystal habit of WO 3 ·2H 2 O, which tends to grow into thin Fig. 1. SEM images (top and cross-sectional views) of WO 3 thin films derived from WO 3 ·2H 2 O sols (PEG 2mass%). WO 3 ·2H 2 O content: (a) 1mass%, (b) and (c) 5 mass%. plates in the sol. During the drying process, the thin plates are stacked together to form square plates having lamel- lar structure. The lamellar structure is preserved even after WO 3 ·2H 2 O is dehydrated finally into WO 3 by calcination (dehydration temperature about 200 ◦ C). We have shown that topotaxy holds well between the crystals of WO 3 ·2H 2 O and that of WO 3 : The basal plane (010) of WO 3 ·2H 2 Ois converted into the basal plane (002) of WO 3 [12]. As shown by the cross-sectional view Fig. 1c, the thin films were a slim layer of discrete square plates of WO 3 ly- ing on the substrate. In the direction normal to the substrate, a few plates were stacked in some spots, while only a single plate existed in other spots. With such a gross population, the films were hardly uniform in thickness. Nevertheless, the thickness of the layer was measured visually at arbitrarily selected five spots to obtain their mean value for each film. Fig. 2 shows the mean thickness thus obtained as a func- tion of WO 3 ·2H 2 O content (1–5 mass%) of the spin-coating dispersion, together with the maximum and minimum in thickness measured. Although the thickness fluctuated sig- nificantly depending on spots within a single film, the mean thickness tended to increase monotonically with increasing WO 3 content. The film derived from 5mass% WO 3 ·2H 2 O dispersion, 450nm in the mean thickness, was supplied to the subsequent gas sensing experiments. Y G. Choi et al. /Sensors and Actuators B 101 (2004) 107–111 109 Fig. 2. Mean thickness of WO 3 thin films derived from WO 3 ·2H 2 O sols as a function of WO 3 ·2H 2 O content (mass% on the WO 3 basis). Vertical bars show maxima and minima in thickness. Remarks are given here to the role and fate of PEG added to the spin-coating dispersion. As stated earlier, it was unable to deposit a layer of WO 3 ·2H 2 O crystals on the substrate from a neat WO 3 ·2H 2 O dispersion by spin coating. Obvi- ously, the deposition of WO 3 ·2H 2 O crystals in the present case was assisted greatly by PEG. It is known that PEG in- creases the viscosity of the dispersion. In fact, the viscosity of the spin-coating dispersions containing 0 and 2 mass% PEG was 1.0 and 1.4cP, respectively. An increase in viscos- ity would make the dispersion more adhesive to the substrate, giving the WO 3 ·2H 2 O crystals more chances to deposit in the spin-coating process. It is thus considered at present that the primary role of PEG is to increase the viscosity of the dis- persion. Part of PEG would be left behind in the spin-coated film. As revealed by thermogravimetric analysis, however, PEG molecules began to decompose at about 200 ◦ C, and no residues remained after calcination at 300 ◦ C for 2 h. Fig. 3. Response and recovery transients of electrical resistance to switching-on and -off 100 and 500 ppb NO 2 in dry air at three selected temperatures. (Film derived from a dispersion of 5 mass% WO 3 ·2H 2 O and 2 mass% PEG.) 3.2. NO 2 sensing properties in dry atmosphere The WO 3 thin film derived from the dispersion of 5 mass% WO 3 ·2H 2 O was examined for NO 2 sensing prop- erties at 200, 250 and 300 ◦ C. Response and recovery transients in electrical resistance to switching-on and -off 100 and 500 ppb NO 2 are shown in Fig. 3. Response and recovery were both rather sluggish, taking about 10 min or more for 90% of full response or of full recovery at all temperatures. This was in contrast to the case of thick films (about 6 ␮m in thickness) reported previously [12,13], where the 90% response or recovery times were less than 2 min for the films calcined at 300 ◦ C. The electrical re- sistance of the thin film device in air (R a ) was fairly high, exceeding 10 7  at 200 ◦ C. Accordingly, the resistance under exposure to NO 2 in air (R g ) easily went beyond the limit of reliable measurement range (∼10 9 )even for small concentrations (e.g. 500ppb) of NO 2 . The upper limiting concentration of NO 2 practically acceptable was rather low, for example, being smaller than a few hundred ppb at 200 ◦ C. Within this limitation, the present film was quite sensitive to NO 2 . As seen from Fig. 4, where sensor response as expressed by normalized resistance (R g /R a )is shown as a function of NO 2 concentration at three different temperatures. At 200 ◦ C, sensor response was as large as 7–50 ppb NO 2 . With a rise in temperature, sensor response reduced rather sharply, showing 50 and 10–500 ppb NO 2 at 250 and 300 ◦ C, respectively. Nevertheless, sensor response to 50 ppb NO 2 at 250 ◦ C was still 3, assuring that the device was sensitive enough even at this temperature to meet the detection of environmental NO 2 (environmental standard: 40–60 ppb in Japan). For comparison, sensor response of the screen-printed thick film device was also shown in the same figure. It is seen that the present device is more sensitive to NO 2 than the thick film device at 200 ◦ C, although such 110 Y G. Choi et al. /Sensors and Actuators B 101 (2004) 107–111 Fig. 4. Normalized resistance (R g /R a ) as a function of NO 2 concentration for a thin film device (full lines and filled marks) and a thick film device (broken lines and open marks) at three selected temperatures (dry atmosphere). superiority is reduced and almost lost at 250 and 300 ◦ C, respectively. 3.3. Preparation under different conditions The thin film device prepared above was found to have a problem that the response and recovery transients were too sluggish. Thus, thin film devices were prepared under dif- ferent conditions. First, the content of PEG in the dispersion was increased from 2mass% to 7, 15 or 20 mass%, while keeping the WO 3 ·2H 2 O content the same (5mass%). The viscosity of the spin-coating dispersion was increased from 1.4 to 3.3, 8.0 or 11.3 cP, correspondingly, so that the film thickness would be expected to increase in this order. Fig. 5a shows the response and recovery transients to 50 ppb NO 2 for the resulting thin film devices, calcined at 300 ◦ C. The rate of response tended to be quicker with 15 and 20 mass% PEG (transient (3) and (4)) than with 2 and 7 mass% PEG ((1) and (2)), whereas the rate of recovery was almost un- changed regardless of the PEG content. As also seen from the figure, the magnitudes of sensor response were about the same, being large enough to safely detect 50ppb NO 2 for all the devices. These results indicate the necessity of investigating the influences of high PEG content on the re- sulting thin films in more detail in the future. Second, the calcination temperature was raised from 300 to 400 ◦ C. Re- sulting transients are shown in Fig. 5b. The rates of both response and recovery were made more sluggish as com- pared with the case of Fig. 5a. The deterioration of response and recovery rates with increasing calcination temperature has been observed for the thick film devices, and the phe- nomenon has been interpreted as reflecting an increase of micro-pores in the WO 3 lamellae included [13]. In the case of the thick films, however, fairly quick transients, taking less than 2 min for 90% response or recovery, have been obtained even after calcination at 400 ◦ C. The sluggish transients of the present thin films, shown in Fig. 5a and b, suggest that the micropore structure may differ significantly from that Fig. 5. Response and recovery transients to 50 ppb NO 2 in dry air for WO 3 thin film devices, derived from WO 3 ·2H 2 O sols different in PEG content and calcined at 300 ◦ C (a) and 400 ◦ C (b). PEG content in mass%: (1) 2, (2) 7, (3) 15 and (4) 20. of the thick films, although this is also to be confirmed experimentally. 3.4. Response to NO 2 in humid atmosphere Finally, NO 2 sensing properties in humid atmosphere were tested briefly by using a fresh thin film device, which was derived from the dispersion added with 2 mass% PEG and calcined at 300 ◦ C. Fig. 6 shows response and recovery transients to 1000 and 500 ppb NO 2 in air of 10% relative hu- midity (RH) at 250 and 300 ◦ C. Unexpectedly, the response and recovery transients were found to be improved drasti- cally under the humid atmosphere. The 90% response (re- covery) times were about 1min (about 0.5min) at both tem- peratures. It is remarkable that the recovery is quicker than the response, because the reverse is the case usually. Almost the same conclusion was drawn when relative humidity was increased to 40 and 70% (data not shown here). The mech- anism by which the humidity improves the transients is not known well at present. It may be possible that the adsorption of NO 2 in the micro-pores of WO 3 lamellae, which is consid- ered to be responsible for the sluggish transients, is prevented Y G. Choi et al. /Sensors and Actuators B 101 (2004) 107–111 111 Fig. 6. Response and recovery transients to 1000 and 500ppb NO 2 in humid air (10% relative humidity) at 250 and 300 ◦ C. (Device derived from a dispersion of 5 mass% WO 3 ·2H 2 O and 2 mass% PEG and calcined at 300 ◦ C.) from occurring due to preferential condensation of water va- por. Any way, it turned out that the problem of the present de- vice, i.e. being sluggish in response and recovery, could thus be eliminated in humid atmospheres. In addition, although the resistance of the device in air (R a ) was lowered by some extent in the humid conditions, sensor response (R g /R a ) hardly degraded with humidity, as seen from the comparison between Figs. 3 and 6. This suggests that the NO 2 sensing may be carried out more favorably in humid atmospheres than in dry atmosphere provided that the humidity level is known separately. More detailed investigations are desired for the NO 2 sensing properties in humid atmospheres. 4. Conclusions An aqueous sol of WO 3 ·2H 2 O, added with PEG, could be applied successfully for fabricating a thin film of WO 3 on an alumina substrate by spin coating and calcination. The thin film device obtained was sensitive enough to detect dilute NO 2 in environments, though the response and recov- ery transients were rather sluggish. Remarkably, the tran- sients were made very sharp in humid atmospheres, while sensor response (sensitivity) to NO 2 hardly degraded with humidity. Acknowledgements This study was partially supported by a grant-in-aid for Scientific Research from The Ministry of Education, Cul- ture, Sport, Science and Technology of Japan. References [1] M. Akiyama, J. Tamaki, N. Miura, N. Yamazoe, Tungsten oxide-based semiconductor sensor highly sensitive to NO and NO 2 , Chem. Lett. (1991) 1611. [2] T. Maekawa, J. Tamaki, N. Miura, N. Yamazoe, Gold-loaded tungsten oxide sensor for detection of ammonia in air, Chem. Lett. (1992) 639. [3] M. Ando, S. Suto, T. Suzuki, T. Tsuchida, C. Nakayama, N. Miura, N. Yamazoe, H 2 S and CH 3 SH sensor using a thick film of gold-loaded tungsten oxide, Chem. Lett. (1994) 335. [4] G. Sberveglieri, L. Depero, S. Groppelli, P. Nelli, WO 3 sputtered thin film for NO x monitoring, Sens. Actuators B 26/27 (1995) 89. [5] P. Nelli, L.E. Depero, M. Ferroni, S. Groppelli, V. Guidi, F. Ron- coni, L. Sangaletti, G. Sberveglieri, Sub-ppm NO 2 sensors based on nanosized thin films of titanium–tungsten oxides, Sens. Actuators B 31 (1996) 89. [6] D.J. Smith, J.F. Vetelino, R.S. Falconer, E.L. Wittman, Stability, sen- sitivity and selectivity of tungsten trioxide films for sensing appli- cations, Sens. Actuators B 13/14 (1993) 264. [7] M. Penza, M.A. Tagliente, L. Mirenghi, C. Gerardi, C. Martucci, G. Cassano, Tungsten trioxide (WO 3 ) sputtered thin films for a NO x gas sensor, Sens. Actuators B 50 (1995) 9. [8] B.T. Marquis, J.F. Vetelino, A semiconducting metal oxide sensor array for the detection of NO x and NH 3 , Sens. Actuators B 77 (2001) 100. [9] J.L. Solis, S. Saukko, L. Kish, C.G. Granqvist, V. Lantto, Semi- conductor gas sensors based on nanostructured tungsten oxide, Thin Solid Films 391 (2001) 255. [10] K. Galatsis, Y.X. Li, W. Wlodarski, E. Comini, G. Sberveglieri, C. Cantalini, S. Santucci, M. Passacantando, Comparison of single and binary oxide MoO 3 , Sens. Actuators B 83 (2002) 276. [11] Y G. Choi, G. Sakai, K. Shimanoe, N. Miura, N. Yamazoe, Prepa- ration of aqueous sols of tungsten oxide dihydrate from sodium tungstate by an ion exchange method, Sens. Actuators B 87 (2002) 63. [12] Y G. Choi, G. Sakai, K. Shimanoe, Y. Teraoka, N. Miura, N. Ya- mazoe, Preparation of size and habit-controlled nano crystallites of tungsten oxide, Sens. Actuators B 93 (2003) 486. [13] Y G. Choi, G. Sakai, K. Shimanoe, Y. Teraoka, N. Miura, N. Ya- mazoe, Wet process-prepared thick films of WO 3 for NO 2 sensing, Sens. Actuators B 95 (2003) 258. Biographies Yong-Gyu Choi received his BE degree in Materials Science and Engi- neering in 1996 and ME degree in 1998 from Kyungsung University in Korea. He received PhD in Engineering in 2003 from Kyushu Univer- sity. His current research interest is development of a NO x sensor by spin-coating method with WO 3 sol provided by ion exchange method. Go Sakai has been a research associate at Kyushu University since 1996. He received his BE degree in Applied Chemistry in 1991, ME degree in 1993 and PhD in Engineering in 1996 from Kyushu University. His current research work is focused on development of chemical sensors as well as functional inorganic materials. Kengo Shimanoe has been an Associate Professor at Kyushu University since 1999. He received his BE degree in Applied Chemistry in 1983 and ME degree in 1985 from Kagoshima University and Kyushu University, respectively. He joined the advanced materials and technology labora- tory in Nippon Steel Corp. and studied the electronic characterization on semiconductor surface and the electrochemical reaction on materials. He received PhD in Engineering in 1993 from Kyushu University. His cur- rent research interests include the development of chemical sensors and the analysis of solid surface. Noboru Yamazoe has been a Professor at Kyushu University since 1981. He received his BE degree in Applied Chemistry in 1963 and PhD in Engineering in 1969 from Kyushu University. His current research inter- ests include the development and application of the functional inorganic materials. . 107–111 Wet process-based fabrication of WO 3 thin film for NO 2 detection Yong-Gyu Choi, Go Sakai, Kengo Shimanoe, Noboru Yamazoe ∗ Department of Materials. mean thickness. The film was a slim pack of square plates, each of which was a stack of thin lamellar crystals of WO 3 . The device was sensitive enough