- Email: [email protected]
Hydrogen sensors on the basis of SnO2–TiO2 systems
Sensors and Actuators B 174 (2012) 527–534
Contents lists available at SciVerse ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Hydrogen sensors on the basis of SnO2 –TiO2 systems D. Shaposhnik a,∗ , R. Pavelko b , E. Llobet a , F. Gispert-Guirado a , X. Vilanova a a b
Minos-EMAS, Department of Electronic Engineering, University Rovira i Virgili, Tarragona, Spain Kyushu University, Department of Energy and Material Sciences, Kasuga-Koen 6-1, Kasuga-shi, Fukuoka 816-8580, Japan
a r t i c l e
i n f o
Article history: Received 28 December 2011 Received in revised form 28 April 2012 Accepted 7 May 2012 Available online 14 May 2012 Keywords: Semiconductor gas sensor Hydrogen Tin oxide Titanium oxide Crystallite growth
a b s t r a c t In this study we compare two types of materials for gas sensor applications: co-precipitated SnO2 and TiO2 and their mechanical mixtures. TEM, FTIR, and TXRD analyses were used to compare the synthesized materials. It was found that co-precipitation leads to the formation at low temperatures of a rutile phase only, which does not undergo any phase transformation upon heating. In contrast, separately synthesized TiO2 with anatase-type structure induces crystallite growth in SnO2 mechanically mixed with the former oxide. Sensing properties of the materials in question were analyzed in a broad range of working temperatures and H2 concentrations. Higher signals of co-precipitated materials are discussed regarding their electrical properties, thermal stability and surface hydroxyls. © 2012 Published by Elsevier B.V.
1. Introduction Fast development of hydrogen-based technologies, including promising reports on hydrogen vehicles and fuel cells, give rise to a need for inexpensive and sensitive detectors of hydrogen leakages. Importance of hydrogen sensors was also sadly proved in atomic industry: both Chernobyl and Fukushima accidents were aggravated by hydrogen explosions. Together with application in early fire alarms, hydrogen sensors seem to become one of the most abundant gas detectors in the near future [1]. Up to now the most suitable technologies for hydrogen sensors mass production include electrochemical and metal oxide (MOx) types. Even though MOx sensors are known to be highly dependent on humidity and this disadvantage (together with slightly higher power consumption) differs them greatly from the electrochemical type, the MOx sensors are still promising for the market of hydrogen detectors [1]. Being robust, compact, energy efficient (especially MEMS type) and low-cost, this type of H2 detectors should overcome the problem of high cross sensitivity under real ambient conditions, which in general, is one of the most crucial shortcomings for many hydrogen detectors [2]. One of the ways to increase selectivity in the presence of water vapors is to modify the sensing material, without technological complication of sensor design. Many R&D efforts have been focused on this problem [3–9]. SnO2 doped with TiO2 is among the prospective materials. The material has demonstrated rather low cross sensitivity towards
∗ Corresponding author. Tel.: +34 977 25 65 71; fax: +34 977 55 96 05. E-mail address: [email protected] (D. Shaposhnik). 0925-4005/$ – see front matter © 2012 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.snb.2012.05.028
water vapors [3,8,10]. It seems therefore important for practical applications to study the SnO2 –TiO2 system in more detail. Namely, it is important to understand: what is the role of the doping, is there any difference between bulk doping and simple mechanical mixture of the phases, and finally how stable are the doped materials at elevated temperatures. This article partially answers these questions. It is well known that tin and titanium dioxides (cassiterite and rutile phases) possess iso-structural crystalline modification – both crystallize in tetragonal structure P 4(2)/mnm. Similarity in the crystalline structure ensures formation of solid solutions as well as decreases electron scattering on the interphases between contacting crystallites [10,11]. But in spite of structural similarity, the oxides differ remarkably in their electronic properties. N-type conductivity of both materials is mainly determined by understoichiometric amount of oxygen atoms in the crystalline lattice. The latter gives rise to numerous donor states within the wide band gaps. For SnO2 the surface donor states are located at ca. 114 meV below conduction band, while for TiO2 – ca. 800 meV below Fermi level [12,13]. The difference in shallow levels position explains high resistance of TiO2 based materials at sensor working temperatures. It also suggests that in SnO2 –TiO2 systems, electrons generated either due to temperature or surface reaction will migrate towards TiO2 , causing electron depletion in SnO2 phase, likewise in the case of PdO and Ag2 O [12]. Another important observation for gas sensors is thermal stability of surface and bulk composition in the dispersed SnO2 –TiO2 system. At 1430 ◦ C the phases are known to form solid solutions in all ranges of SnO2 /TiO2 ratio. However, upon cooling, the solid solution undergoes spinoidal decomposition if the ratio is within
528
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
the miscibility gap observed between ca. 15 and 85 mol% of TiO2 [14,15]. In spite of being separated, the oxide phases within the gap are rich in Ti and Sn [16]. This suggests that using either mechanically mixed phases or co-precipitated ones an interfacial diffusion between the phases can occur slowly until the phases get saturated with respective metal ions. Such effect is especially important for long-term stability of the sensors, since diffusion in polycrystalline oxides, followed by recrystallization and particle growth, leads to remarkable changes in sheet resistance of the sensing material. In this article we compare two types of SnO2 –TiO2 materials. The first one was synthesized by co-precipitation method, while the second one was prepared through mechanical mixing of SnO2 and TiO2 powders. For each of the types, identical SnO2 /TiO2 ratios were prepared. Pure SnO2 and TiO2 powders were used to contrast the effect of the admixture or dopant. FTIR spectroscopy is used to qualitatively compare the amount of hydroxyl groups on the materials in question. Using in situ XRD, we examine crystallite size evolution of the synthesized materials. Sensor signals were studied in a wide range of working temperature and humidity conditions for different H2 concentration. The sensing properties of the materials as well as their thermal stability are discussed. 2. Experimental 2.1. Material synthesis Co-precipitation method, reported in [2], was used after modification of some synthesis parameters (pH, temperature of precipitation, and concentration of the precursor solution), to synthesize SnO2 bulk doped with TiO2 . Tin (IV) hydroxide acetate and titanium (IV) isopropoxide were dissolved in glacial acetic acid in the required ratios to obtain 10 and 30 wt% of TiO2 in the SnO2 –TiO2 system. Since titanium isopropoxide is not soluble in acetic acid directly, it was primarily dissolved in ethyl acetate and the obtained solution was added to acetic acid. Ethyl acetate was also added to the solution during the synthesis of blank tin dioxide. NH3 H2 O was used to cause hydrolytic precipitation of the oxides. The base was added to the cooled (ca. 7 ◦ C) solution of the precursors in order to avoid their premature hydrolysis. After the ammonia was added, the precipitation was initiated by heating the solution up to 55 ◦ C. These conditions were determined and tested first for the synthesis of tin and titania oxides individually and were found to be appropriate for the simultaneous and complete precipitation of both of them. The suspended precipitates were centrifugated, and then dried for 12 h at 100 ◦ C, 2 h at 350 ◦ C and 30 min at 440 ◦ C. No washing was used, because all by-products were expected to be removed under calcination. Second type of the materials were prepared through mechanical mixing of the synthesized pure SnO2 and TiO2 powders. The powders, after drying and annealing, as described above, were ground together in a mortar in the same reciprocal quantities as it was used for co-precipitation. No additional thermal treatment of the powders was performed. The list of the synthesized materials together with their notation in the following text is given in Table 1.
Table 1 List of the synthesized materials and their notation in the text. Synthesis
Notation in the text
Precipitation
SnO2 TiO2 cp ST-91 cp ST-73 mm ST-91 mm ST-73
Co-precipitation Mechanical mixing
SnO2 content (wt%) 100 – 90 70 90 70
TiO2 content (wt%) – 100 10 30 10 30
2.2. Materials characterization Infrared spectra of the co-precipitated and blank oxides were recorded on JASCO 680 plus FTIR spectrophotometer. The powders in the quantity ca. 3 mg were mixed with 150 mg KBr and pressed to form the discs. The spectra were taken within the range of 500–4000 cm−1 with a resolution 1 cm−1 . TEM measurements were performed on Jeol JEM 1011 at 100 kV using Holey Carbon HC-300-Cu high resolution grids. Temperature X-ray diffraction measurements (TXRD) were made using BRUKER D8 ADVANCE diffractometer (Cu K␣, 40 kV, 40 mA, vertical – goniometer) equipped with XYZ motorized stage, GADDS (General Area Diffraction System) and a MRI BTSBASIC platinum ribbon heating stage. The GADDS detector was a HI-STAR (multiwire proportional counter of 30 cm × 30 cm with a 1024 × 1024 pixel). We collected frames (2D XRD patterns) covering 20–80◦ 2 from two different detector positions at a distance of 15 cm from each sample. The exposition time was 300 s per frame and it was chi-integrated to generate the conventional 2 vs. intensity diffractogram. The first diffractogram was collected at room temperature (30 ◦ C). Then the material was heated up to 700 ◦ C (with a heating rate of 0.1667◦ /s) and the second measurement was performed. The third and followings patterns were collected consecutively at the same temperature after 600 s. In total 91 patterns were collected during 32 h of annealing at 700 ◦ C. Static air atmosphere was used throughout the all analysis. ICDD data base (release 2007) and Diffracplus Evaluation software (Bruker 2007) were used to identify the crystalline structure of the samples. Crystallite size calculations were realized with the help of original software TOPAS 4.2 [28] and local routines. The instrumental contribution to the peak width was estimated using NIST standard material LaB6 (SRM 676b). Mean crystallite size (D) was calculated from the integral breadth of the peaks within 20–78◦ of 2, according to the modified Scherrer expression [17]: ˇS =
D cos
(1)
where is the wavelength of incident radiation and is the Bragg angle. In the calculations we used Double Voigt Approach, assuming that there is no contribution to the integral breadth of the peak from the microstrain and that the Lorentz component alone determines the crystallite size. An example of pattern fitting is shown in Fig. 1. The difference between experimental and calculated patterns (black curve) and goodness of fit Rwp (found between 7.0 and 11.0) indicate that the fitting is good enough. 2.3. Sensors preparation and characterization Synthesized materials were mixed with propanediol-1,2 in a weight ratio 2:1, grinded, and deposited onto alumina microsubstrates by drop-coating. The latter were provided with gap platinum electrodes and platinum heater on the back side. Further details regarding the substrate can be found in [18]. The substrates were dried at 100 ◦ C, annealed in a furnace at 710 ◦ C for 5 min and then soldered to TO-8 package. Sensors were stabilized at 400 ◦ C in ambient air for 72 h, followed by stabilization in a teflon chamber (17 cm3 ) under flowing synthetic air during 24 h. Environics Series 4000 gas mixing system was used to prepare following gas mixtures: 1, 3, 10, 50, 200, and 500 ppm H2 . The flow rate was set to 100 ml/min throughout the all gas tests. The resistance of the sensors was controlled by electrometer/high resistance meter Keithley 6517A. Agilent E 3631A DC power supply and digital multimeter Agilent 34401A were used to heat the sensors and control the sensor temperature.
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
529
Fig. 3. FTIR spectra of some synthesized materials.
temperature, the target gas was introduced into the chamber for 55 min and then was replaced with the background gas. The pulse was repeated three times at each specified temperature. At 400 ◦ C the sensors were tested towards 1, 3, 10, 50, 200, and 500 ppm H2 in dry air. These tests were performed in the pulse mode as well. Signal values were calculated as resistance ratio: (Rair − Rgas )/Rgas , where Rair is the resistance in air, and Rgas is the one in the target gas. 3. Results and discussion
Fig. 1. Experimental diffractograms (1), their fitting with Voigt function (2) and the difference between them (3) for sample cp ST-73 at annealing time t = 0 h (a) and t = 31 h (b).
The sensors were tested first towards 20 ppm H2 in air with RH 0, 30, and 80% at operating temperatures of 300, 350, 400, 450, and 500 ◦ C. After reaching stable resistance baseline at a given
Fig. 2 shows TEM images of some synthesized materials after drying and annealing. Mean particle size for blank SnO2 was found to be close to 4 nm, while mean crystallite size for this material amounts to 2 nm. Both values are lower by ca. 1.5 times compared to SnO2 bulk doped with TiO2 . In the case of blank TiO2 the particles are notably larger, with size between 5 and 18 nm, and mean crystallite size about 6 nm. FTIR spectroscopy was used to compare amount and acidity of surface hydroxyls for the co-precipitated and blank materials (Fig. 3). The broad band centered at ca. 3420 cm−1 is assigned as stretching vibrations of bridge-bonded or/and hydrogen-bonded OH groups [19]. It is also known that the highest frequency between 3800 and 2500 cm−1 is assigned to the most basic hydroxyl groups with the lowest coordination number of oxygen. Decrease in OH frequency is therefore associated with the increase of coordination number of oxygen and possible hydrogen bonding [20].
Fig. 2. TEM images of blank tin oxide (a), cp ST-91 (b), cp ST-73 (c), and blank TiO2 (d).
530
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
The position of the band suggests that the dominant species for all materials is the one vibrating at ca. 3420 cm−1 . This position will be taken as reference and higher or lower acidity of OH groups will be judged in respect of this value hereinafter. The shapes of the bands indicate rather different contribution of hydroxyl groups with higher and lower acidity. Let us compare blank materials first. The band of TiO2 is remarkably broader in both directions than that of SnO2 . The highest broadening is observed in low frequency region, implying higher amount of more acidic OH groups. The “partial” content of less acidic hydroxyls is also higher for TiO2 . Both these facts suggest that OH groups on TiO2 surface are more heterogeneous regarding their acidity in respect with SnO2 . They are also ca. 20% more numerous judging by the band integral, normalized by sample weight, and this is regardless of the fact that the surface of SnO2 is higher than that of TiO2 after annealing at 440 ◦ C. Doping with 10 wt% TiO2 leads to an increase in the dominant species quantity, as well as slightly decreases acidic hydroxyl groups. Further doping with 30 wt% TiO2 maintain the amount of the dominant species at high level compared to the blank SnO2 but in this case increases acidic OH groups similarly to blank TiO2 . The general tendency that can be derived from Fig. 3 is that the doping with TiO2 results in higher overall amount of hydroxyl groups compared to blank SnO2 . Apart from that, OH groups become slightly more acidic in respect to the dominant species. These results are in good agreement with the ones obtained previously for SnO2 materials doped with IVB elements [8]. According to the conventional XRD analysis, co-precipitated materials as well as blank SnO2 are crystallized in the tetragonal ˚ c = 3.18710 A˚ (cassiterite). On system: P42 /mnm, a = b = 4.73820 A, the other hand, blank TiO2 possesses anatase structure: I41 /amd, ˚ which means that mechanical mixture ˚ c = 9.5139 A, a = b = 3.7852 A, of SnO2 and TiO2 consists of two phases: cassiterite (from SnO2 ) and anatase (from TiO2 ). Upon isothermal annealing at 700 ◦ C no phase transitions were observed for co-precipitated materials and blank SnO2 (see e.g., Fig. 1). However, anatase phase of blank TiO2 transforms under ˚ the same conditions to the rutile phase: P42 /mnm, a = b = 4.5933 A, c = 2.9592 A˚ (Fig. 4). As it can be seen from Fig. 4b, the mechanical mixture consists of two phases (with ca. 1:1 weight ratio) already at the earliest stage of annealing. The fact of the ART (anatase to rutile transition) for blank TiO2 as well as its temperature are in good agreement with the literature [21]. Apart from the phase transition in blank TiO2 , abrupt growth of crystallites occurs upon isothermal annealing. This phenomenon is well known and believed to be due to breaking of old and formation of new bonds in the crystal lattice (so called reconstructive transformation) [21,22]. TXRD experiment has shown that after approximately 1 h the oxide is represented mainly by rutile modification (Fig. 4b) with crystallite size of more than 200 nm. After 5 h of annealing no anatase phase was detected with XRD. High degree of crystallinity (i.e., very sharp diffraction peaks) prevented us from estimating the kinetics of the crystallite growth for TiO2 rutile phase. Accordingly, in Fig. 5 the crystallite size evolution is shown only for SnO2 phases in co-precipitated and mechanically mixed oxides. The experimental values of crystallite sizes as a function of annealing time were fitted with generalized parabolic model: D(t) = kt1/n , where D(t) is the crystallite size at time t, k is the temperature dependent constant and n is the growth exponent [23–25]. In spite of the fact that size-dependent impediment model was found to be more meaningful for nanocrystalline oxides [26–28], we used the former model because the fitting error of the latter was remarkably higher. As it can be seen either from Fig. 5 or from Table 2, the growth is remarkably higher in the case of mechanical mixtures. Most probably the anatase phase, which undergoes dramatic structural
Fig. 4. Diffractograms (a) of blank TiO2 before annealing (1), after reaching 700 ◦ C (2) and after 31 h of annealing (3); weight evolution of rutile and anatase phases in blank TiO2 (b).
Fig. 5. Crystallite size evolution of SnO2 phase during annealing at 700 ◦ C (red lines are fittings with the parabolic function). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
531
Table 2 Fitting parameters of the model (fitting error for the last decimal is given in parenthesis). Material
k
n
SnO2 cp ST-91 cp ST-73 mm ST-91 mm ST-73
7.24(4) 5.56(1) 4.35(1) 13.5(5) 12.3(8)
10.0(1) 11.3(1) 10.0(1) 20.0(1) 20.8(1)
changes upon heating, evokes recrystallization phenomena in the rutile phase through interfacial contacts. Note, that rate constant and the mean crystallite sizes are higher for the mixture with low (10 wt%) TiO2 content. On the other hand, the growth exponent is similar for both mechanical mixtures, indicating similar mechanisms of the crystallite growth. Simultaneous precipitation of the oxides seems to effectively decrease growth rate of SnO2 crystallites. The highest growth kinetics between co-precipitated materials was observed again for the system with 10 wt% TiO2 . The fact that both co-precipitated materials and blank SnO2 manifest similar growth exponent suggests that presence of Ti4+ does not affect the growth mechanism in tin dioxide. Since blank TiO2 manifested poor structural stability upon heating, this material will not be compared with the other ones. However, it is important to mention that its sensitivity towards H2 was found to be very low (ca. 1 for 20 ppm H2 , see Fig. 7a), which suggests that SnO2 plays the dominant role in the sensing phenomenon of the mixed materials. Fig. 6 summarizes the sensor tests results obtained at different temperatures. As it can be seen, both mechanical mixing and co-precipitation increase signals towards 20 ppm H2 . The highest increase is observed for the co-precipitated materials and for the highest content of TiO2 at that. The same tendency is seen for the mixed oxides: addition of 30% TiO2 enhances the signal more compared to the 10% TiO2 . The classic volcano-shaped curve of blank SnO2 has remarkably changed due the co-precipitation. The high-temperature side of the curve becomes broader and intense. However, for mixed and co-precipitated materials the signal maximum is still observed at 400 ◦ C, which is close to the blank material. This suggests that after doping, the active surface species are more numerous (signal is higher) and slightly different in their nature (high-temperature broadening).
Fig. 6. Signals towards 20 ppm H2 in dry air at five different operating temperatures. Notation: (1) cp ST-73, (2) cp ST-91, (3) mm ST-73, (4) SnO2 , and (5) mm ST-91.
Fig. 7. Typical responses towards 20 ppm H2 in dry air (a) and calibration curves for the materials in question (b).
Fig. 7a gives typical sensor responses to 20 ppm H2 in air at the temperature of signal maximum – 400 ◦ C. The sheet resistance of the co-precipitated materials gradually increases with the increase of TiO2 content. The same phenomenon is observed for the mixed SnO2 and TiO2 . However, their resistance is ca. 10 times lower compared to the co-precipitated materials, which is probably related to the percolation effect: in the case of mechanical mixtures the carriers flow predominantly through conductive SnO2 grains, while in the co-precipitated oxides the SnO2 grains are doped with TiO2 and the conductive grains of blank SnO2 are scarce. The sensors manifest rather similar response times, suggesting similar adsorption kinetics for the materials in question. The response time (t90 ) for all sensors ranges from 12 to 14 s, while recovery time (t90 ) is between 200 and 360 s for all sensors except cp ST-91 and TiO2 . The recovery time for the latter materials was found about 30 s. Blank SnO2 , mm ST-91 and cp ST-91 show very similar sensitivity with concentration exponents 0.58–0.59 (Fig. 7b). The lowest value – 0.53 – was found in the case of mm ST-73, while the highest – 0.63 – for cp ST-73. Higher signals and sensitivity of the co-precipitated materials compared to the blank SnO2 do not seem to be associated with the different surface chemistry of hydrogen interaction, since temperature of signal maximum for both materials is close to the blank oxide (however, evidently, new active surface species have appeared after doping, resulting in high signals at 450 and 500 ◦ C, see Fig. 6). Comparing surface hydroxyls, we have shown that the dominant type is similar for all materials. The difference was found only in higher amount of hydroxyls on co-precipitated materials compared to the blank SnO2 .
532
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
Fig. 8. Signals towards 20 ppm H2 in air at different humidity rate. Notation: (1) RH 0, (2) RH 30%, and (3) RH 80%.
Also sheet resistance of the co-precipitated materials is higher, meaning lower concentration of charge carriers. Since sensor signal S was defined as (Rair − Rgas )/Rgas and R = 1/qn, where q is the carrier charge, is the carrier mobility and n is the carrier concentration, we can write the following expression, assuming that charge and mobility of carriers are constant in air and in target gas: S=
ngas nair + egas egas −1= −1= nair nair nair
(2)
where, nair and ngas are the carrier concentration in contact with air and target gas, egas is the amount of generated carriers due to interaction between the target gas and the semiconductor. As nair decreases (resistance increases) and other parameters like surface area, surface chemical potential, target gas and its concentration maintain unchanged, this will increase the sensor signal. The latter, of course, is an hypothetical approximation, helping, however, to estimate possible electrical nature of the increased sensor signals in the case of co-precipitated materials. Another reason for higher signals of the co-precipitated materials can also come from surface area. These materials manifested better thermal stability compared to the blank oxide. Therefore materials are expected to be less aggregated after annealing on the substrate.
The case of mechanically mixed oxides is quite ambiguous. The TiO2 additive increases modestly the signal towards 1–50 ppm H2 . At higher H2 concentrations this advantage disappears, leading to the same signal level as for blank SnO2 . Their poor thermal stability evidently results in lower surface area compared to the blank oxide. However, the signals are very close for both types of materials. It seems that even there is evidence of interphase interaction between blank oxides and new interphase forms upon thermal treatment (leading to remarkable crystallite growth of SnO2 ), it has negligible effect on sensing properties of SnO2 . Fig. 8 shows sensors responses to 20 ppm of hydrogen in air at different humidity levels. The materials show two general types of signal changes upon increasing water concentration: their signal maxima drop and shift towards higher temperatures. This specific behavior of SnO2 -based gas sensors has been reported for many reducing (except CO and ethanol) and oxidizing gases [29–33]. Both blank SnO2 and mechanical mixtures manifest pronounced changes of the signal: it drops by ca. 1.5–2 times at 400 ◦ C and its maximum shifts by more than 50 ◦ C. Blank SnO2 and mm ST-91 manifest very similar behavior, suggesting that addition of 10% TiO2 has almost no effect on reducing humidity effect. Another mechanical mixture, mm ST-73, being more sensitive to low concentration of H2 (Fig. 7b), seems to be an intermediate case between blank and
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
co-precipitated materials. For the latter the drop at 400 ◦ C is about 10–20%, while the shift is ca. 30 ◦ C. The results show that both coprecipitated SnO2 –TiO2 materials possess much better selectivity to hydrogen in humid air, than blank oxide or its mechanical mixtures with TiO2 . The signals to hydrogen are also higher; however, their signal maxima in dry and humid air occur at temperatures 400–430 ◦ C, which is high for low-power devices. More study is needed to decrease temperature of signal maximum and in the same time maintain the achieved selectivity. 4. Conclusions Co-precipitation of SnO2 and TiO2 results in the formation of dispersed materials with high structural homogeneity already at low temperatures. These materials manifested better thermal stability compared to the blank SnO2 , and higher sensitivity towards 1–500 ppm H2 in a wide operation temperature range. Higher signals are most probably due to combination of several factors: higher surface area, higher amount of surface hydroxyls, and lower carrier concentration. Together with low sensitivity towards water vapors these findings make co-precipitated materials rather attractive for practical applications. Mechanically mixed SnO2 and TiO2 demonstrated remarkably poorer thermal stability, compared to the blank SnO2 . The phenomenon is believed to be due to the anatase–rutile phase transition in blank TiO2 , which evokes recrystallization in SnO2 . Sensing properties of this type of materials seem to be determined mainly by the SnO2 phase. Acknowledgments This work has been financially supported by the Spanish Ministry of Science and Innovation (project TEC2009-07107) and European (FEDER) Funds. References [1] L. Boon-Brett, J. Bousek, P. Moretto, Reliability of commercially available hydrogen sensors for detection of hydrogen at critical concentrations. Part II: selected sensor test results, International Journal of Hydrogen Energy 34 (2009) 562–571. [2] T. Hübert, L. Boon-Brett, G. Black, U. Banach, Hydrogen sensors – a review, Sensors and Actuators B: Chemical 157 (2011) 329–352. [3] T. Antonio, et al., Minimal cross-sensitivity to humidity during ethanol detection by SnO2 –TiO2 solid solutions, Nanotechnology 20 (2009) 315502. [4] S.H. Hahn, N. Barsan, U. Weimar, S.G. Ejakov, J.H. Visser, R.E. Soltis, CO sensing with SnO2 thick film sensors: role of oxygen and water vapour, Thin Solid Films 436 (2003) 17–24. [5] M. Hübner, C.E. Simion, A. Tomescu-St˘anoiu, S. Pokhrel, N. Bârsan, U. Weimar, Influence of humidity on CO sensing with p-type CuO thick film gas sensors, Sensors and Actuators B: Chemical 153 (2011) 347–353. [6] G. Korotcenkov, V. Brinzari, Y. Boris, M. Ivanov, J. Schwank, J. Morante, Influence of surface Pd doping on gas sensing characteristics of SnO2 thin films deposited by spray pyrolysis, Thin Solid Films 436 (2003) 119–126. [7] R.G. Pavelko, H. Daly, C. Hardacre, A.A. Vasiliev, E. Llobet, Interaction of water, hydrogen and their mixtures with SnO2 based materials: the role of surface hydroxyl groups in detection mechanisms, Physical Chemistry Chemical Physics 12 (2010) 2639–2647. [8] R.G. Pavelko, A.A. Vasiliev, E. Llobet, V.G. Sevastyanov, N.T. Kuznetsov, Selectivity problem of SnO2 based materials in the presence of water vapors, Sensors and Actuators B: Chemical 170 (2012) 51–59. [9] T. Itoh, I. Matsubara, M. Kadosaki, Y. Sakai, W. Shin, N. Izu, M. Nishibori, Effects of high-humidity aging on platinum, palladium, and gold loaded tin oxide – volatile organic compound sensors, Sensors 10 (2010) 6513–6521. [10] K. Zakrzewska, Mixed oxides as gas sensors, Thin Solid Films 391 (2001) 229–238. [11] Y. Yoshida, S. Tokashiki, K. Kubota, R. Shiratuchi, Y. Yamaguchi, M. Kono, S. Hayase, Increase in photovoltaic performances of dye-sensitized solar cells – modification of interface between TiO2 nano-porous layers and F-doped SnO2 layers, Solar Energy Materials and Solar Cells 92 (2008) 646–650. [12] M. Batzill, U. Diebold, The surface and materials science of tin oxide, Progress in Surface Science 79 (2005) 47–154. [13] U. Diebold, The surface science of titanium dioxide, Surface Science Reports 48 (2003) 53–229.
533
[14] M. Park, T.E. Mitchell, A.H. Heuer, Subsolidus equilibria in the TiO2 –SnO2 system, Journal of the American Ceramic Society 58 (1975) 43–47. [15] P.R. Bueno, E.R. Leite, L.O.S. Bulhões, E. Longo, C.O. Paiva-Santos, Sintering and mass transport features of (Sn,Ti)O2 polycrystalline ceramics, Journal of the European Ceramic Society 23 (2003) 887–896. [16] R.G. Palgrave, A. Bourlange, D.J. Payne, J.S. Foord, R.G. Egdell, Interfacial diffusion during growth of SnO2 (1 1 0) on TiO2 (1 1 0) by oxygen plasma assisted molecular beam epitaxy, Crystal Growth and Design 9 (2009) 1793–1797. [17] A.R. Stokes, A.J.C. Wilson, A method of calculating the integral breadths of Debye–Scherrer lines, Proceedings of the Cambridge Philosophical Society 38 (1942) 313–322. [18] V.V. Malyshev, A.V. Pislyakov, Investigation of gas-sensitivity of sensor structures to carbon monoxide in a wide range of temperature, concentration and humidity of gas medium, Sensors and Actuators B: Chemical 123 (2007) 71–81. [19] M.A. Henderson, The interaction of water with solid surfaces: fundamental aspects revisited, Surface Science Reports 46 (2002) 5–308. [20] A. Davydov, The nature of oxide surface centers, in: Molecular Spectroscopy of Oxide Catalyst Surfaces, John Wiley & Sons, Ltd., 2003, pp. 27–179. [21] D. Hanaor, C. Sorrell, Review of the anatase to rutile phase transformation, Journal of Materials Science 46 (2011) 855–874. [22] P.I. Gouma, P.K. Dutta, M.J. Mills, Structural stability of titania thin films, Nanostructured Materials 11 (1999) 1231–1237. [23] C. Rock, K. Okazaki, Grain growth kinetics and thermal stability in a nanocrystalline multiphase mixture prepared by low-energy ball milling, Nanostructured Materials 5 (1995) 657–671. [24] G. Li, L. Li, J. Boerio-Goates, B.F. Woodfield, High purity anatase TiO2 nanocrystals: near room-temperature synthesis, grain growth kinetics, and surface hydration chemistry, Journal of the American Chemical Society 127 (2005) 8659–8666. [25] N. Audebrand, J.P. Auffredic, D. Louer, X-ray diffraction study of the early stages of the growth of nanoscale zinc oxide crystallites obtained from thermal decomposition of four precursors. General concepts on precursor-dependent microstructural properties, Chemistry of Materials 10 (1998) 2450–2461. [26] H. Natter, M. Schmelzer, M.S. Loffler, C.E. Krill, A. Fitch, R. Hempelmann, Graingrowth kinetics of nanocrystalline iron studied in situ by synchrotron real-time X-ray diffraction, Journal of Physical Chemistry B 104 (2000) 2467–2476. [27] A. Michels, C.E. Krill, H. Ehrhardt, R. Birringer, D.T. Wu, Modelling the influence of grain-size-dependent solute drag on the kinetics of grain growth in nanocrystalline materials, Acta Materialia 47 (1999) 2143–2152. [28] R.G. Pavelko, A.A. Vasiliev, F. Gispert-Guirado, N. Barrabes, J. Llorca, E. Llobet, V.G. Sevastyanov, Crystallite growth kinetics of highly pure nanocrystalline tin dioxide: the effect of palladium doping, Materials Chemistry and Physics 121 (2010) 267–273. [29] R.G. Pavelko, Material science of SnO2 -based oxide systems: chemical ways to improve sensor performance, in: X. Qi (Ed.), Gas Sensors: Developments, Efficacy and Safety, Nova Science Publishers, Hauppauge, NY, 2011. [30] G. Martinelli, M.C. Carotta, Influence of additives on the sensing properties of screen-printed SnO2 gas sensors, Sensors and Actuators B: Chemical 16 (1993) 363–366. [31] W. Schmid, N. Bârsan, U. Weimar, Sensing of hydrocarbons with tin oxide sensors: possible reaction path as revealed by consumption measurements, Sensors and Actuators B: Chemical 89 (2003) 232–236. [32] T. Sahm, W. Rong, N. Bârsan, L. Mädler, S.K. Friedlander, U. Weimar, Formation of multilayer films for gas sensing by in situ thermophoretic deposition of nanoparticles from aerosol phase, Journal of Materials Research 22 (2007) 850–857. [33] G. Korotcenkov, I. Blinov, V. Brinzari, J.R. Stetter, Effect of air humidity on gas response of SnO2 thin film ozone sensors, Sensors and Actuators B: Chemical 122 (2007) 519–526.
Biographies Dmitry Shaposhnik graduated from Voronezh State University (Russia) in 2007. At present he is PhD student in University Rovira i Virgili (Tarragona, Spain) in the Electronic Engineering Department. His research interests concern synthesis of dispersed and nanostructured materials, their material science and applications for semiconductor gas sensors. Roman G. Pavelko graduated from People’s Friendship University of Russia (Moscow, 2003) and from University Rovira i Virgili (Tarragona, 2007), obtained his PhD in Chemistry in 2007 at Institute of General and Inorganic Chemistry (Russian Academy of Science, Moscow) and PhD in Electronic engineering in 2010 at University Rovira i Virgili. At present he is JSPS fellow at Kyushu University (Japan). His research interests are related to synthesis of dispersed materials, material science, experimental and theoretical study of surface processes related to semiconductor metal oxides. Eduard Llobet graduated in telecommunication engineering from the Universitat Politècnica de Catalunya (UPC) (Barcelona, Spain) in 1991, and received his PhD in 1997 from the same university. He is currently full professor of Electronic Technology in the Electronic Engineering Department at the Universitat Rovira i Virgili (Tarragona, Spain). His main areas of interest are in the design of semiconductor and carbon nanotube based gas sensors and in the application of intelligent systems to complex odor analysis. He is the Director of the Research Centre on Engineering of Materials and micro/nanosystems (EMaS) and Senior Member of the IEEE.
534
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
Francesc Gispert-Guirado graduated in geology from the University of Barcelona (UB) (Barcelona, Spain) in 1991, and received his PhD in Crystallography in 1998 at the same university. At present he is working as an XRD technician at the Scientific Resources Service of the University Rovira i Virgili (Tarragona, Spain). His research is focused on profile analysis of powder X-ray diffraction data with programmable software and the Rietveld method.
Xavier Vilanovagraduated in telecommunication engineering from the Universitat Politècnica de Catalunya (UPC) (Barcelona, Spain) in 1991, and received his PhD in 1998 from the same university. He is currently Full Professor in the Electronic Engineering Department at the Universitat Rovira i Virgili (Tarragona, Spain). His research activities are related to semiconductor gas sensors development and characterization, as well as, micro-preconcentration units and gas sensors microsystems design.
Contents lists available at SciVerse ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Hydrogen sensors on the basis of SnO2 –TiO2 systems D. Shaposhnik a,∗ , R. Pavelko b , E. Llobet a , F. Gispert-Guirado a , X. Vilanova a a b
Minos-EMAS, Department of Electronic Engineering, University Rovira i Virgili, Tarragona, Spain Kyushu University, Department of Energy and Material Sciences, Kasuga-Koen 6-1, Kasuga-shi, Fukuoka 816-8580, Japan
a r t i c l e
i n f o
Article history: Received 28 December 2011 Received in revised form 28 April 2012 Accepted 7 May 2012 Available online 14 May 2012 Keywords: Semiconductor gas sensor Hydrogen Tin oxide Titanium oxide Crystallite growth
a b s t r a c t In this study we compare two types of materials for gas sensor applications: co-precipitated SnO2 and TiO2 and their mechanical mixtures. TEM, FTIR, and TXRD analyses were used to compare the synthesized materials. It was found that co-precipitation leads to the formation at low temperatures of a rutile phase only, which does not undergo any phase transformation upon heating. In contrast, separately synthesized TiO2 with anatase-type structure induces crystallite growth in SnO2 mechanically mixed with the former oxide. Sensing properties of the materials in question were analyzed in a broad range of working temperatures and H2 concentrations. Higher signals of co-precipitated materials are discussed regarding their electrical properties, thermal stability and surface hydroxyls. © 2012 Published by Elsevier B.V.
1. Introduction Fast development of hydrogen-based technologies, including promising reports on hydrogen vehicles and fuel cells, give rise to a need for inexpensive and sensitive detectors of hydrogen leakages. Importance of hydrogen sensors was also sadly proved in atomic industry: both Chernobyl and Fukushima accidents were aggravated by hydrogen explosions. Together with application in early fire alarms, hydrogen sensors seem to become one of the most abundant gas detectors in the near future [1]. Up to now the most suitable technologies for hydrogen sensors mass production include electrochemical and metal oxide (MOx) types. Even though MOx sensors are known to be highly dependent on humidity and this disadvantage (together with slightly higher power consumption) differs them greatly from the electrochemical type, the MOx sensors are still promising for the market of hydrogen detectors [1]. Being robust, compact, energy efficient (especially MEMS type) and low-cost, this type of H2 detectors should overcome the problem of high cross sensitivity under real ambient conditions, which in general, is one of the most crucial shortcomings for many hydrogen detectors [2]. One of the ways to increase selectivity in the presence of water vapors is to modify the sensing material, without technological complication of sensor design. Many R&D efforts have been focused on this problem [3–9]. SnO2 doped with TiO2 is among the prospective materials. The material has demonstrated rather low cross sensitivity towards
∗ Corresponding author. Tel.: +34 977 25 65 71; fax: +34 977 55 96 05. E-mail address: [email protected] (D. Shaposhnik). 0925-4005/$ – see front matter © 2012 Published by Elsevier B.V. http://dx.doi.org/10.1016/j.snb.2012.05.028
water vapors [3,8,10]. It seems therefore important for practical applications to study the SnO2 –TiO2 system in more detail. Namely, it is important to understand: what is the role of the doping, is there any difference between bulk doping and simple mechanical mixture of the phases, and finally how stable are the doped materials at elevated temperatures. This article partially answers these questions. It is well known that tin and titanium dioxides (cassiterite and rutile phases) possess iso-structural crystalline modification – both crystallize in tetragonal structure P 4(2)/mnm. Similarity in the crystalline structure ensures formation of solid solutions as well as decreases electron scattering on the interphases between contacting crystallites [10,11]. But in spite of structural similarity, the oxides differ remarkably in their electronic properties. N-type conductivity of both materials is mainly determined by understoichiometric amount of oxygen atoms in the crystalline lattice. The latter gives rise to numerous donor states within the wide band gaps. For SnO2 the surface donor states are located at ca. 114 meV below conduction band, while for TiO2 – ca. 800 meV below Fermi level [12,13]. The difference in shallow levels position explains high resistance of TiO2 based materials at sensor working temperatures. It also suggests that in SnO2 –TiO2 systems, electrons generated either due to temperature or surface reaction will migrate towards TiO2 , causing electron depletion in SnO2 phase, likewise in the case of PdO and Ag2 O [12]. Another important observation for gas sensors is thermal stability of surface and bulk composition in the dispersed SnO2 –TiO2 system. At 1430 ◦ C the phases are known to form solid solutions in all ranges of SnO2 /TiO2 ratio. However, upon cooling, the solid solution undergoes spinoidal decomposition if the ratio is within
528
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
the miscibility gap observed between ca. 15 and 85 mol% of TiO2 [14,15]. In spite of being separated, the oxide phases within the gap are rich in Ti and Sn [16]. This suggests that using either mechanically mixed phases or co-precipitated ones an interfacial diffusion between the phases can occur slowly until the phases get saturated with respective metal ions. Such effect is especially important for long-term stability of the sensors, since diffusion in polycrystalline oxides, followed by recrystallization and particle growth, leads to remarkable changes in sheet resistance of the sensing material. In this article we compare two types of SnO2 –TiO2 materials. The first one was synthesized by co-precipitation method, while the second one was prepared through mechanical mixing of SnO2 and TiO2 powders. For each of the types, identical SnO2 /TiO2 ratios were prepared. Pure SnO2 and TiO2 powders were used to contrast the effect of the admixture or dopant. FTIR spectroscopy is used to qualitatively compare the amount of hydroxyl groups on the materials in question. Using in situ XRD, we examine crystallite size evolution of the synthesized materials. Sensor signals were studied in a wide range of working temperature and humidity conditions for different H2 concentration. The sensing properties of the materials as well as their thermal stability are discussed. 2. Experimental 2.1. Material synthesis Co-precipitation method, reported in [2], was used after modification of some synthesis parameters (pH, temperature of precipitation, and concentration of the precursor solution), to synthesize SnO2 bulk doped with TiO2 . Tin (IV) hydroxide acetate and titanium (IV) isopropoxide were dissolved in glacial acetic acid in the required ratios to obtain 10 and 30 wt% of TiO2 in the SnO2 –TiO2 system. Since titanium isopropoxide is not soluble in acetic acid directly, it was primarily dissolved in ethyl acetate and the obtained solution was added to acetic acid. Ethyl acetate was also added to the solution during the synthesis of blank tin dioxide. NH3 H2 O was used to cause hydrolytic precipitation of the oxides. The base was added to the cooled (ca. 7 ◦ C) solution of the precursors in order to avoid their premature hydrolysis. After the ammonia was added, the precipitation was initiated by heating the solution up to 55 ◦ C. These conditions were determined and tested first for the synthesis of tin and titania oxides individually and were found to be appropriate for the simultaneous and complete precipitation of both of them. The suspended precipitates were centrifugated, and then dried for 12 h at 100 ◦ C, 2 h at 350 ◦ C and 30 min at 440 ◦ C. No washing was used, because all by-products were expected to be removed under calcination. Second type of the materials were prepared through mechanical mixing of the synthesized pure SnO2 and TiO2 powders. The powders, after drying and annealing, as described above, were ground together in a mortar in the same reciprocal quantities as it was used for co-precipitation. No additional thermal treatment of the powders was performed. The list of the synthesized materials together with their notation in the following text is given in Table 1.
Table 1 List of the synthesized materials and their notation in the text. Synthesis
Notation in the text
Precipitation
SnO2 TiO2 cp ST-91 cp ST-73 mm ST-91 mm ST-73
Co-precipitation Mechanical mixing
SnO2 content (wt%) 100 – 90 70 90 70
TiO2 content (wt%) – 100 10 30 10 30
2.2. Materials characterization Infrared spectra of the co-precipitated and blank oxides were recorded on JASCO 680 plus FTIR spectrophotometer. The powders in the quantity ca. 3 mg were mixed with 150 mg KBr and pressed to form the discs. The spectra were taken within the range of 500–4000 cm−1 with a resolution 1 cm−1 . TEM measurements were performed on Jeol JEM 1011 at 100 kV using Holey Carbon HC-300-Cu high resolution grids. Temperature X-ray diffraction measurements (TXRD) were made using BRUKER D8 ADVANCE diffractometer (Cu K␣, 40 kV, 40 mA, vertical – goniometer) equipped with XYZ motorized stage, GADDS (General Area Diffraction System) and a MRI BTSBASIC platinum ribbon heating stage. The GADDS detector was a HI-STAR (multiwire proportional counter of 30 cm × 30 cm with a 1024 × 1024 pixel). We collected frames (2D XRD patterns) covering 20–80◦ 2 from two different detector positions at a distance of 15 cm from each sample. The exposition time was 300 s per frame and it was chi-integrated to generate the conventional 2 vs. intensity diffractogram. The first diffractogram was collected at room temperature (30 ◦ C). Then the material was heated up to 700 ◦ C (with a heating rate of 0.1667◦ /s) and the second measurement was performed. The third and followings patterns were collected consecutively at the same temperature after 600 s. In total 91 patterns were collected during 32 h of annealing at 700 ◦ C. Static air atmosphere was used throughout the all analysis. ICDD data base (release 2007) and Diffracplus Evaluation software (Bruker 2007) were used to identify the crystalline structure of the samples. Crystallite size calculations were realized with the help of original software TOPAS 4.2 [28] and local routines. The instrumental contribution to the peak width was estimated using NIST standard material LaB6 (SRM 676b). Mean crystallite size (D) was calculated from the integral breadth of the peaks within 20–78◦ of 2, according to the modified Scherrer expression [17]: ˇS =
D cos
(1)
where is the wavelength of incident radiation and is the Bragg angle. In the calculations we used Double Voigt Approach, assuming that there is no contribution to the integral breadth of the peak from the microstrain and that the Lorentz component alone determines the crystallite size. An example of pattern fitting is shown in Fig. 1. The difference between experimental and calculated patterns (black curve) and goodness of fit Rwp (found between 7.0 and 11.0) indicate that the fitting is good enough. 2.3. Sensors preparation and characterization Synthesized materials were mixed with propanediol-1,2 in a weight ratio 2:1, grinded, and deposited onto alumina microsubstrates by drop-coating. The latter were provided with gap platinum electrodes and platinum heater on the back side. Further details regarding the substrate can be found in [18]. The substrates were dried at 100 ◦ C, annealed in a furnace at 710 ◦ C for 5 min and then soldered to TO-8 package. Sensors were stabilized at 400 ◦ C in ambient air for 72 h, followed by stabilization in a teflon chamber (17 cm3 ) under flowing synthetic air during 24 h. Environics Series 4000 gas mixing system was used to prepare following gas mixtures: 1, 3, 10, 50, 200, and 500 ppm H2 . The flow rate was set to 100 ml/min throughout the all gas tests. The resistance of the sensors was controlled by electrometer/high resistance meter Keithley 6517A. Agilent E 3631A DC power supply and digital multimeter Agilent 34401A were used to heat the sensors and control the sensor temperature.
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
529
Fig. 3. FTIR spectra of some synthesized materials.
temperature, the target gas was introduced into the chamber for 55 min and then was replaced with the background gas. The pulse was repeated three times at each specified temperature. At 400 ◦ C the sensors were tested towards 1, 3, 10, 50, 200, and 500 ppm H2 in dry air. These tests were performed in the pulse mode as well. Signal values were calculated as resistance ratio: (Rair − Rgas )/Rgas , where Rair is the resistance in air, and Rgas is the one in the target gas. 3. Results and discussion
Fig. 1. Experimental diffractograms (1), their fitting with Voigt function (2) and the difference between them (3) for sample cp ST-73 at annealing time t = 0 h (a) and t = 31 h (b).
The sensors were tested first towards 20 ppm H2 in air with RH 0, 30, and 80% at operating temperatures of 300, 350, 400, 450, and 500 ◦ C. After reaching stable resistance baseline at a given
Fig. 2 shows TEM images of some synthesized materials after drying and annealing. Mean particle size for blank SnO2 was found to be close to 4 nm, while mean crystallite size for this material amounts to 2 nm. Both values are lower by ca. 1.5 times compared to SnO2 bulk doped with TiO2 . In the case of blank TiO2 the particles are notably larger, with size between 5 and 18 nm, and mean crystallite size about 6 nm. FTIR spectroscopy was used to compare amount and acidity of surface hydroxyls for the co-precipitated and blank materials (Fig. 3). The broad band centered at ca. 3420 cm−1 is assigned as stretching vibrations of bridge-bonded or/and hydrogen-bonded OH groups [19]. It is also known that the highest frequency between 3800 and 2500 cm−1 is assigned to the most basic hydroxyl groups with the lowest coordination number of oxygen. Decrease in OH frequency is therefore associated with the increase of coordination number of oxygen and possible hydrogen bonding [20].
Fig. 2. TEM images of blank tin oxide (a), cp ST-91 (b), cp ST-73 (c), and blank TiO2 (d).
530
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
The position of the band suggests that the dominant species for all materials is the one vibrating at ca. 3420 cm−1 . This position will be taken as reference and higher or lower acidity of OH groups will be judged in respect of this value hereinafter. The shapes of the bands indicate rather different contribution of hydroxyl groups with higher and lower acidity. Let us compare blank materials first. The band of TiO2 is remarkably broader in both directions than that of SnO2 . The highest broadening is observed in low frequency region, implying higher amount of more acidic OH groups. The “partial” content of less acidic hydroxyls is also higher for TiO2 . Both these facts suggest that OH groups on TiO2 surface are more heterogeneous regarding their acidity in respect with SnO2 . They are also ca. 20% more numerous judging by the band integral, normalized by sample weight, and this is regardless of the fact that the surface of SnO2 is higher than that of TiO2 after annealing at 440 ◦ C. Doping with 10 wt% TiO2 leads to an increase in the dominant species quantity, as well as slightly decreases acidic hydroxyl groups. Further doping with 30 wt% TiO2 maintain the amount of the dominant species at high level compared to the blank SnO2 but in this case increases acidic OH groups similarly to blank TiO2 . The general tendency that can be derived from Fig. 3 is that the doping with TiO2 results in higher overall amount of hydroxyl groups compared to blank SnO2 . Apart from that, OH groups become slightly more acidic in respect to the dominant species. These results are in good agreement with the ones obtained previously for SnO2 materials doped with IVB elements [8]. According to the conventional XRD analysis, co-precipitated materials as well as blank SnO2 are crystallized in the tetragonal ˚ c = 3.18710 A˚ (cassiterite). On system: P42 /mnm, a = b = 4.73820 A, the other hand, blank TiO2 possesses anatase structure: I41 /amd, ˚ which means that mechanical mixture ˚ c = 9.5139 A, a = b = 3.7852 A, of SnO2 and TiO2 consists of two phases: cassiterite (from SnO2 ) and anatase (from TiO2 ). Upon isothermal annealing at 700 ◦ C no phase transitions were observed for co-precipitated materials and blank SnO2 (see e.g., Fig. 1). However, anatase phase of blank TiO2 transforms under ˚ the same conditions to the rutile phase: P42 /mnm, a = b = 4.5933 A, c = 2.9592 A˚ (Fig. 4). As it can be seen from Fig. 4b, the mechanical mixture consists of two phases (with ca. 1:1 weight ratio) already at the earliest stage of annealing. The fact of the ART (anatase to rutile transition) for blank TiO2 as well as its temperature are in good agreement with the literature [21]. Apart from the phase transition in blank TiO2 , abrupt growth of crystallites occurs upon isothermal annealing. This phenomenon is well known and believed to be due to breaking of old and formation of new bonds in the crystal lattice (so called reconstructive transformation) [21,22]. TXRD experiment has shown that after approximately 1 h the oxide is represented mainly by rutile modification (Fig. 4b) with crystallite size of more than 200 nm. After 5 h of annealing no anatase phase was detected with XRD. High degree of crystallinity (i.e., very sharp diffraction peaks) prevented us from estimating the kinetics of the crystallite growth for TiO2 rutile phase. Accordingly, in Fig. 5 the crystallite size evolution is shown only for SnO2 phases in co-precipitated and mechanically mixed oxides. The experimental values of crystallite sizes as a function of annealing time were fitted with generalized parabolic model: D(t) = kt1/n , where D(t) is the crystallite size at time t, k is the temperature dependent constant and n is the growth exponent [23–25]. In spite of the fact that size-dependent impediment model was found to be more meaningful for nanocrystalline oxides [26–28], we used the former model because the fitting error of the latter was remarkably higher. As it can be seen either from Fig. 5 or from Table 2, the growth is remarkably higher in the case of mechanical mixtures. Most probably the anatase phase, which undergoes dramatic structural
Fig. 4. Diffractograms (a) of blank TiO2 before annealing (1), after reaching 700 ◦ C (2) and after 31 h of annealing (3); weight evolution of rutile and anatase phases in blank TiO2 (b).
Fig. 5. Crystallite size evolution of SnO2 phase during annealing at 700 ◦ C (red lines are fittings with the parabolic function). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
531
Table 2 Fitting parameters of the model (fitting error for the last decimal is given in parenthesis). Material
k
n
SnO2 cp ST-91 cp ST-73 mm ST-91 mm ST-73
7.24(4) 5.56(1) 4.35(1) 13.5(5) 12.3(8)
10.0(1) 11.3(1) 10.0(1) 20.0(1) 20.8(1)
changes upon heating, evokes recrystallization phenomena in the rutile phase through interfacial contacts. Note, that rate constant and the mean crystallite sizes are higher for the mixture with low (10 wt%) TiO2 content. On the other hand, the growth exponent is similar for both mechanical mixtures, indicating similar mechanisms of the crystallite growth. Simultaneous precipitation of the oxides seems to effectively decrease growth rate of SnO2 crystallites. The highest growth kinetics between co-precipitated materials was observed again for the system with 10 wt% TiO2 . The fact that both co-precipitated materials and blank SnO2 manifest similar growth exponent suggests that presence of Ti4+ does not affect the growth mechanism in tin dioxide. Since blank TiO2 manifested poor structural stability upon heating, this material will not be compared with the other ones. However, it is important to mention that its sensitivity towards H2 was found to be very low (ca. 1 for 20 ppm H2 , see Fig. 7a), which suggests that SnO2 plays the dominant role in the sensing phenomenon of the mixed materials. Fig. 6 summarizes the sensor tests results obtained at different temperatures. As it can be seen, both mechanical mixing and co-precipitation increase signals towards 20 ppm H2 . The highest increase is observed for the co-precipitated materials and for the highest content of TiO2 at that. The same tendency is seen for the mixed oxides: addition of 30% TiO2 enhances the signal more compared to the 10% TiO2 . The classic volcano-shaped curve of blank SnO2 has remarkably changed due the co-precipitation. The high-temperature side of the curve becomes broader and intense. However, for mixed and co-precipitated materials the signal maximum is still observed at 400 ◦ C, which is close to the blank material. This suggests that after doping, the active surface species are more numerous (signal is higher) and slightly different in their nature (high-temperature broadening).
Fig. 6. Signals towards 20 ppm H2 in dry air at five different operating temperatures. Notation: (1) cp ST-73, (2) cp ST-91, (3) mm ST-73, (4) SnO2 , and (5) mm ST-91.
Fig. 7. Typical responses towards 20 ppm H2 in dry air (a) and calibration curves for the materials in question (b).
Fig. 7a gives typical sensor responses to 20 ppm H2 in air at the temperature of signal maximum – 400 ◦ C. The sheet resistance of the co-precipitated materials gradually increases with the increase of TiO2 content. The same phenomenon is observed for the mixed SnO2 and TiO2 . However, their resistance is ca. 10 times lower compared to the co-precipitated materials, which is probably related to the percolation effect: in the case of mechanical mixtures the carriers flow predominantly through conductive SnO2 grains, while in the co-precipitated oxides the SnO2 grains are doped with TiO2 and the conductive grains of blank SnO2 are scarce. The sensors manifest rather similar response times, suggesting similar adsorption kinetics for the materials in question. The response time (t90 ) for all sensors ranges from 12 to 14 s, while recovery time (t90 ) is between 200 and 360 s for all sensors except cp ST-91 and TiO2 . The recovery time for the latter materials was found about 30 s. Blank SnO2 , mm ST-91 and cp ST-91 show very similar sensitivity with concentration exponents 0.58–0.59 (Fig. 7b). The lowest value – 0.53 – was found in the case of mm ST-73, while the highest – 0.63 – for cp ST-73. Higher signals and sensitivity of the co-precipitated materials compared to the blank SnO2 do not seem to be associated with the different surface chemistry of hydrogen interaction, since temperature of signal maximum for both materials is close to the blank oxide (however, evidently, new active surface species have appeared after doping, resulting in high signals at 450 and 500 ◦ C, see Fig. 6). Comparing surface hydroxyls, we have shown that the dominant type is similar for all materials. The difference was found only in higher amount of hydroxyls on co-precipitated materials compared to the blank SnO2 .
532
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
Fig. 8. Signals towards 20 ppm H2 in air at different humidity rate. Notation: (1) RH 0, (2) RH 30%, and (3) RH 80%.
Also sheet resistance of the co-precipitated materials is higher, meaning lower concentration of charge carriers. Since sensor signal S was defined as (Rair − Rgas )/Rgas and R = 1/qn, where q is the carrier charge, is the carrier mobility and n is the carrier concentration, we can write the following expression, assuming that charge and mobility of carriers are constant in air and in target gas: S=
ngas nair + egas egas −1= −1= nair nair nair
(2)
where, nair and ngas are the carrier concentration in contact with air and target gas, egas is the amount of generated carriers due to interaction between the target gas and the semiconductor. As nair decreases (resistance increases) and other parameters like surface area, surface chemical potential, target gas and its concentration maintain unchanged, this will increase the sensor signal. The latter, of course, is an hypothetical approximation, helping, however, to estimate possible electrical nature of the increased sensor signals in the case of co-precipitated materials. Another reason for higher signals of the co-precipitated materials can also come from surface area. These materials manifested better thermal stability compared to the blank oxide. Therefore materials are expected to be less aggregated after annealing on the substrate.
The case of mechanically mixed oxides is quite ambiguous. The TiO2 additive increases modestly the signal towards 1–50 ppm H2 . At higher H2 concentrations this advantage disappears, leading to the same signal level as for blank SnO2 . Their poor thermal stability evidently results in lower surface area compared to the blank oxide. However, the signals are very close for both types of materials. It seems that even there is evidence of interphase interaction between blank oxides and new interphase forms upon thermal treatment (leading to remarkable crystallite growth of SnO2 ), it has negligible effect on sensing properties of SnO2 . Fig. 8 shows sensors responses to 20 ppm of hydrogen in air at different humidity levels. The materials show two general types of signal changes upon increasing water concentration: their signal maxima drop and shift towards higher temperatures. This specific behavior of SnO2 -based gas sensors has been reported for many reducing (except CO and ethanol) and oxidizing gases [29–33]. Both blank SnO2 and mechanical mixtures manifest pronounced changes of the signal: it drops by ca. 1.5–2 times at 400 ◦ C and its maximum shifts by more than 50 ◦ C. Blank SnO2 and mm ST-91 manifest very similar behavior, suggesting that addition of 10% TiO2 has almost no effect on reducing humidity effect. Another mechanical mixture, mm ST-73, being more sensitive to low concentration of H2 (Fig. 7b), seems to be an intermediate case between blank and
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
co-precipitated materials. For the latter the drop at 400 ◦ C is about 10–20%, while the shift is ca. 30 ◦ C. The results show that both coprecipitated SnO2 –TiO2 materials possess much better selectivity to hydrogen in humid air, than blank oxide or its mechanical mixtures with TiO2 . The signals to hydrogen are also higher; however, their signal maxima in dry and humid air occur at temperatures 400–430 ◦ C, which is high for low-power devices. More study is needed to decrease temperature of signal maximum and in the same time maintain the achieved selectivity. 4. Conclusions Co-precipitation of SnO2 and TiO2 results in the formation of dispersed materials with high structural homogeneity already at low temperatures. These materials manifested better thermal stability compared to the blank SnO2 , and higher sensitivity towards 1–500 ppm H2 in a wide operation temperature range. Higher signals are most probably due to combination of several factors: higher surface area, higher amount of surface hydroxyls, and lower carrier concentration. Together with low sensitivity towards water vapors these findings make co-precipitated materials rather attractive for practical applications. Mechanically mixed SnO2 and TiO2 demonstrated remarkably poorer thermal stability, compared to the blank SnO2 . The phenomenon is believed to be due to the anatase–rutile phase transition in blank TiO2 , which evokes recrystallization in SnO2 . Sensing properties of this type of materials seem to be determined mainly by the SnO2 phase. Acknowledgments This work has been financially supported by the Spanish Ministry of Science and Innovation (project TEC2009-07107) and European (FEDER) Funds. References [1] L. Boon-Brett, J. Bousek, P. Moretto, Reliability of commercially available hydrogen sensors for detection of hydrogen at critical concentrations. Part II: selected sensor test results, International Journal of Hydrogen Energy 34 (2009) 562–571. [2] T. Hübert, L. Boon-Brett, G. Black, U. Banach, Hydrogen sensors – a review, Sensors and Actuators B: Chemical 157 (2011) 329–352. [3] T. Antonio, et al., Minimal cross-sensitivity to humidity during ethanol detection by SnO2 –TiO2 solid solutions, Nanotechnology 20 (2009) 315502. [4] S.H. Hahn, N. Barsan, U. Weimar, S.G. Ejakov, J.H. Visser, R.E. Soltis, CO sensing with SnO2 thick film sensors: role of oxygen and water vapour, Thin Solid Films 436 (2003) 17–24. [5] M. Hübner, C.E. Simion, A. Tomescu-St˘anoiu, S. Pokhrel, N. Bârsan, U. Weimar, Influence of humidity on CO sensing with p-type CuO thick film gas sensors, Sensors and Actuators B: Chemical 153 (2011) 347–353. [6] G. Korotcenkov, V. Brinzari, Y. Boris, M. Ivanov, J. Schwank, J. Morante, Influence of surface Pd doping on gas sensing characteristics of SnO2 thin films deposited by spray pyrolysis, Thin Solid Films 436 (2003) 119–126. [7] R.G. Pavelko, H. Daly, C. Hardacre, A.A. Vasiliev, E. Llobet, Interaction of water, hydrogen and their mixtures with SnO2 based materials: the role of surface hydroxyl groups in detection mechanisms, Physical Chemistry Chemical Physics 12 (2010) 2639–2647. [8] R.G. Pavelko, A.A. Vasiliev, E. Llobet, V.G. Sevastyanov, N.T. Kuznetsov, Selectivity problem of SnO2 based materials in the presence of water vapors, Sensors and Actuators B: Chemical 170 (2012) 51–59. [9] T. Itoh, I. Matsubara, M. Kadosaki, Y. Sakai, W. Shin, N. Izu, M. Nishibori, Effects of high-humidity aging on platinum, palladium, and gold loaded tin oxide – volatile organic compound sensors, Sensors 10 (2010) 6513–6521. [10] K. Zakrzewska, Mixed oxides as gas sensors, Thin Solid Films 391 (2001) 229–238. [11] Y. Yoshida, S. Tokashiki, K. Kubota, R. Shiratuchi, Y. Yamaguchi, M. Kono, S. Hayase, Increase in photovoltaic performances of dye-sensitized solar cells – modification of interface between TiO2 nano-porous layers and F-doped SnO2 layers, Solar Energy Materials and Solar Cells 92 (2008) 646–650. [12] M. Batzill, U. Diebold, The surface and materials science of tin oxide, Progress in Surface Science 79 (2005) 47–154. [13] U. Diebold, The surface science of titanium dioxide, Surface Science Reports 48 (2003) 53–229.
533
[14] M. Park, T.E. Mitchell, A.H. Heuer, Subsolidus equilibria in the TiO2 –SnO2 system, Journal of the American Ceramic Society 58 (1975) 43–47. [15] P.R. Bueno, E.R. Leite, L.O.S. Bulhões, E. Longo, C.O. Paiva-Santos, Sintering and mass transport features of (Sn,Ti)O2 polycrystalline ceramics, Journal of the European Ceramic Society 23 (2003) 887–896. [16] R.G. Palgrave, A. Bourlange, D.J. Payne, J.S. Foord, R.G. Egdell, Interfacial diffusion during growth of SnO2 (1 1 0) on TiO2 (1 1 0) by oxygen plasma assisted molecular beam epitaxy, Crystal Growth and Design 9 (2009) 1793–1797. [17] A.R. Stokes, A.J.C. Wilson, A method of calculating the integral breadths of Debye–Scherrer lines, Proceedings of the Cambridge Philosophical Society 38 (1942) 313–322. [18] V.V. Malyshev, A.V. Pislyakov, Investigation of gas-sensitivity of sensor structures to carbon monoxide in a wide range of temperature, concentration and humidity of gas medium, Sensors and Actuators B: Chemical 123 (2007) 71–81. [19] M.A. Henderson, The interaction of water with solid surfaces: fundamental aspects revisited, Surface Science Reports 46 (2002) 5–308. [20] A. Davydov, The nature of oxide surface centers, in: Molecular Spectroscopy of Oxide Catalyst Surfaces, John Wiley & Sons, Ltd., 2003, pp. 27–179. [21] D. Hanaor, C. Sorrell, Review of the anatase to rutile phase transformation, Journal of Materials Science 46 (2011) 855–874. [22] P.I. Gouma, P.K. Dutta, M.J. Mills, Structural stability of titania thin films, Nanostructured Materials 11 (1999) 1231–1237. [23] C. Rock, K. Okazaki, Grain growth kinetics and thermal stability in a nanocrystalline multiphase mixture prepared by low-energy ball milling, Nanostructured Materials 5 (1995) 657–671. [24] G. Li, L. Li, J. Boerio-Goates, B.F. Woodfield, High purity anatase TiO2 nanocrystals: near room-temperature synthesis, grain growth kinetics, and surface hydration chemistry, Journal of the American Chemical Society 127 (2005) 8659–8666. [25] N. Audebrand, J.P. Auffredic, D. Louer, X-ray diffraction study of the early stages of the growth of nanoscale zinc oxide crystallites obtained from thermal decomposition of four precursors. General concepts on precursor-dependent microstructural properties, Chemistry of Materials 10 (1998) 2450–2461. [26] H. Natter, M. Schmelzer, M.S. Loffler, C.E. Krill, A. Fitch, R. Hempelmann, Graingrowth kinetics of nanocrystalline iron studied in situ by synchrotron real-time X-ray diffraction, Journal of Physical Chemistry B 104 (2000) 2467–2476. [27] A. Michels, C.E. Krill, H. Ehrhardt, R. Birringer, D.T. Wu, Modelling the influence of grain-size-dependent solute drag on the kinetics of grain growth in nanocrystalline materials, Acta Materialia 47 (1999) 2143–2152. [28] R.G. Pavelko, A.A. Vasiliev, F. Gispert-Guirado, N. Barrabes, J. Llorca, E. Llobet, V.G. Sevastyanov, Crystallite growth kinetics of highly pure nanocrystalline tin dioxide: the effect of palladium doping, Materials Chemistry and Physics 121 (2010) 267–273. [29] R.G. Pavelko, Material science of SnO2 -based oxide systems: chemical ways to improve sensor performance, in: X. Qi (Ed.), Gas Sensors: Developments, Efficacy and Safety, Nova Science Publishers, Hauppauge, NY, 2011. [30] G. Martinelli, M.C. Carotta, Influence of additives on the sensing properties of screen-printed SnO2 gas sensors, Sensors and Actuators B: Chemical 16 (1993) 363–366. [31] W. Schmid, N. Bârsan, U. Weimar, Sensing of hydrocarbons with tin oxide sensors: possible reaction path as revealed by consumption measurements, Sensors and Actuators B: Chemical 89 (2003) 232–236. [32] T. Sahm, W. Rong, N. Bârsan, L. Mädler, S.K. Friedlander, U. Weimar, Formation of multilayer films for gas sensing by in situ thermophoretic deposition of nanoparticles from aerosol phase, Journal of Materials Research 22 (2007) 850–857. [33] G. Korotcenkov, I. Blinov, V. Brinzari, J.R. Stetter, Effect of air humidity on gas response of SnO2 thin film ozone sensors, Sensors and Actuators B: Chemical 122 (2007) 519–526.
Biographies Dmitry Shaposhnik graduated from Voronezh State University (Russia) in 2007. At present he is PhD student in University Rovira i Virgili (Tarragona, Spain) in the Electronic Engineering Department. His research interests concern synthesis of dispersed and nanostructured materials, their material science and applications for semiconductor gas sensors. Roman G. Pavelko graduated from People’s Friendship University of Russia (Moscow, 2003) and from University Rovira i Virgili (Tarragona, 2007), obtained his PhD in Chemistry in 2007 at Institute of General and Inorganic Chemistry (Russian Academy of Science, Moscow) and PhD in Electronic engineering in 2010 at University Rovira i Virgili. At present he is JSPS fellow at Kyushu University (Japan). His research interests are related to synthesis of dispersed materials, material science, experimental and theoretical study of surface processes related to semiconductor metal oxides. Eduard Llobet graduated in telecommunication engineering from the Universitat Politècnica de Catalunya (UPC) (Barcelona, Spain) in 1991, and received his PhD in 1997 from the same university. He is currently full professor of Electronic Technology in the Electronic Engineering Department at the Universitat Rovira i Virgili (Tarragona, Spain). His main areas of interest are in the design of semiconductor and carbon nanotube based gas sensors and in the application of intelligent systems to complex odor analysis. He is the Director of the Research Centre on Engineering of Materials and micro/nanosystems (EMaS) and Senior Member of the IEEE.
534
D. Shaposhnik et al. / Sensors and Actuators B 174 (2012) 527–534
Francesc Gispert-Guirado graduated in geology from the University of Barcelona (UB) (Barcelona, Spain) in 1991, and received his PhD in Crystallography in 1998 at the same university. At present he is working as an XRD technician at the Scientific Resources Service of the University Rovira i Virgili (Tarragona, Spain). His research is focused on profile analysis of powder X-ray diffraction data with programmable software and the Rietveld method.
Xavier Vilanovagraduated in telecommunication engineering from the Universitat Politècnica de Catalunya (UPC) (Barcelona, Spain) in 1991, and received his PhD in 1998 from the same university. He is currently Full Professor in the Electronic Engineering Department at the Universitat Rovira i Virgili (Tarragona, Spain). His research activities are related to semiconductor gas sensors development and characterization, as well as, micro-preconcentration units and gas sensors microsystems design.