Butane detection: W-doped TiO₂ nanoparticles for a butane gas sensor with high sensitivity and fast response/recovery

Abstract

This article describes a new option for butane detection: W-doped TiO₂ nanoparticles with high sensitivity and fast response/recovery toward butane, which were obtained from a simple, non-aqueous sol–gel route. The structure, morphology, surface chemical state and specific surface area were analyzed by X-ray powder diffraction (XRD), transmission electron microscopy (TEM), X-ray photoelectron spectrum (XPS) and N₂-sorption isotherm, respectively. The obtained products are anatase-type TiO₂ with a small grain size (7.5 ± 1.4 nm) and a high specific surface area (181.15 m² g⁻¹). Tungsten element presents in the +6 oxidation state. The resistance–temperature measurements indicate that tungsten dopant leads to the decrease in resistance. The as-prepared pure and W-doped TiO₂ nanoparticles were used to fabricate gas sensor devices. Gas response toward 3000 ppm butane is increased from 6 to 17.8 through the doping of 5% tungsten. Meanwhile, the response and recovery time toward 3000 ppm butane are as fast as 2 and 12 s, respectively. Moreover, the sensor also possesses low detection limit, good linear dependence, good repeatability and long-term stability, indicating the potential of using W-doped TiO₂ nanoparticles for butane gas detection. In addition, a possible mechanism for the enhanced sensitivity of W-doped TiO₂ nanoparticles toward butane is also offered.

---

1. Introduction

With the rapid rise of living standards, more and more chemical reagents, which could be poisonous, cancerogenic, and flammable, enter human life and take an irreplaceable role. Such is butane, a flammable and colorless transparent gas with a distinct odor. Owing to a high critical liquefaction temperature (151.9 °C), butane gas is easily liquefied by pressure even at room temperature; meanwhile, the liquefied butane is also easily gasified by decompression, which is beneficial for its storage, transportation and use. Hence, butane is widely used as portable and convenient fuel in various situations, such as in lighters, outdoor cooking, residential heating and industry. For such a widely used fuel, safety during its storage, transportation and use cannot be ignored. An air mixture that contains 1.5–8.5% butane by volume may cause an explosion or flash fire at a temperature higher than its flash point (−60 °C). Excessive heat exposure and, consequently, excessive pressure in the enclosed vessel containing liquid butane also will lead to violent explosion. Due to its high density and the high storage density caused by its physical nature and liquid storage mode, respectively, the affected area would be quite large once the leakage of butane happens. On the other hand, its combustion products contain some suffocating gases such as CO and CO₂, which further increases the risk in rescue operations. Thus, to safeguard life and property, the fabrication of high-performance gas sensors to monitor butane is needed. For this task, the key is obtaining desirable butane gas-sensing materials.¹

The earlier gas-sensing effects were discovered by T. J. Gray around 1948 on Cu₂O.² Since then, gas-sensing materials have gained great focus on a global scale and, consequently, rapid development, especially for metal oxide semiconductors (MOS).³ The gas-sensing properties of tens of thousands of MOSs with different morphologies and components have been investigated. A series of promising gas-sensing materials have been reported due to their abundant micro-structural defects, which are beneficial to forming absorbed oxygen ions, such as SnO₂,⁴ ZnO,⁵ Fe₂O₃,⁶ WO₃,⁷ ZnSnO₃,⁸ CdIn₂O₄,⁹ LaFeO₃,¹⁰ and ZnCo₂O₄.¹¹ Because butane, which only consists of C–H and C–C bonds, possesses much more stable molecule structure than other common target gases such as ethanol, isopropanol, acetone and so on, the sensitivity toward butane cannot match that toward other gases, and only a few well-known gas-sensing MOSs, such as SnO₂,¹² ZnO,¹³ Fe₂O₃,¹⁴ LaFeO₃,¹⁵ and ZnSnO₃,¹⁶ have been investigated for fabricating butane-specific gas sensors. Unfortunately, their performance, particularly in the sensitivity and response/recovery time, is still not ideal and cannot meet the need of practical butane detection. For example, Min et al. reported that the sensitivity of Pt–SnO₂ thin film toward 5000 ppm butane only reaches around 4.5.¹⁷ Pati et al. reported that the response of ZnO thin film toward 1660 ppm butane cannot reach 3.¹³ Chakraborty et al. reported that the response of γ-Fe₂O₃ toward 500 ppm butane is estimated at around 6.7.¹⁴ It is also reported that the sensor response of hollow ZnSnO₃ microspheres is 5.79 toward 500 ppm butane.¹⁶ In this case, attempting to search for some missed potential butane-sensing MOSs, like TiO₂, is deserved.

TiO₂, with its wide energy band gap, is one of the most versatile semiconductor materials and plays an important role in various technological domains, such as UV absorption,¹⁸ photovoltaic cells,¹⁹ photocatalysis²⁰ and Li-ion battery,²¹ because of its good chemical stability, non-toxicity, abundance and low cost.²² Its versatility seems to be equally promising for gas-sensing application. However, the high resistivity of TiO₂ limits studies on its sensing performance as a gas sensor, especially for the resistance-type gas sensor, let alone fabricating a butane-specific gas sensor. In our previous work, it was found that doping tungsten into TiO₂ can effectively decrease the resistivity of TiO₂,²³ which breaks the barrier that the high resistivity imposes on the investigation of TiO₂ sensing properties. As far as we know, studies on butane-sensing properties of W-doped TiO₂ are seldom reported. At the same time, it is well known that the sensing reactions mainly occur on the surface of sensing materials. Hence, it is believed that increasing the specific surface area of the sensitive materials can provide more active sites and further improve the sensitivity.²⁴ Among the different morphologies, nanoparticles ought to have the highest specific surface area within same scale, and the specific surface area will be larger with smaller grain size.

In this article, W-doped TiO₂ nanoparticles with high specific surface area (181.15 m² g⁻¹) and size of about 7.5 nm were obtained by a simple, non-aqueous solvothermal route. The as-prepared samples were used to fabricate a resistance-type gas sensor. The influence of tungsten dopant on the resistance of TiO₂ was investigated by testing the resistance–temperature curves of the as-fabricated sensors. The gas-sensing properties of the sensors were also tested, showing good sensing properties toward butane, particularly with the high sensitivity and fast response/recovery time. In order to further understand the related mechanisms, we also investigated the structure, morphology, chemical state as well as specific surface area of as-prepared W-doped TiO₂ nanoparticles.

2. Experimental section

All the chemical reagents used in the experiments were obtained from commercial sources, were of analytical purity and were used without further purification.

2.1 Preparation of W-doped TiO₂ nanoparticles

As the most common reagents of tungsten and titanium, WCl₆ and TiCl₄ were chosen as the precursors of W-doped TiO₂ nanoparticles. As we know, both WCl₆ and TiCl₄ are sensitive to water, which will lead to uncontrollable hydrolysis. To avoid the influence of water, nonaqueous sol–gel method was used to prepare W-doped TiO₂ nanoparticles.²⁵ In a typical synthetic experiment, the appropriate amount of WCl₆ was dissolved into 40 mL 1-propanol. Five minutes later, 0.255 mL TiCl₄ was transferred to the WCl₆ solution to form a slight yellow solution. The amount of WCl₆ was calculated according to [W]/[Ti] = 0%, 2.5%, 5.0% and 7.5%, respectively. After magnetically stirring for another 5 minutes, 36 mL of the mixture was transferred into a Teflon-lined stainless steel autoclave with a capacity of 40 mL and reacted under solvothermal conditions at 180 °C for 12 h. Then the autoclave was cooled down to room temperature in standard atmosphere. The resulting products were centrifuged and washed three times with ethanol, and the final W-doped TiO₂ nanoparticles were obtained after drying at 80 °C.

2.2 Characterization of as-prepared W-doped TiO₂ nanoparticles

X-ray diffraction (XRD, Rigaku D/MAX-3B powder diffractometer) with a copper target and K_α radiation (λ = 1.54056 Å) was used for the phase identification, where the diffracted X-ray intensities were recorded as a function of 2θ. The sample was scanned from 10° to 80° (2θ) in steps of 0.02°. The transmission electron micrographs (TEM) were obtained with a Zeiss EM 912 Ω instrument at an acceleration voltage of 120 kV, while high-resolution transmission electron microscopy (HRTEM) characterization was done using a Philips CM200-FEG microscope (200 kV, C_s = 1.35 mm). The samples for TEM were prepared by dispersing the final powders in ethanol, and this dispersion was then dropped onto carbon–copper grids. The nitrogen adsorption isotherm was measured at 77.3 K with a Micromeritics ASAP 2010 automated sorption analyzer. Prior to the measurement, the sample was degassed at 300 °C for 3 h under a vacuum. X-ray photoelectron spectroscopy (XPS) was measured at room temperature in ESCALAB 250. During XPS analysis, Al Kα X-ray beam was adopted as the excitation source, and power was set to 250 W. Vacuum pressure of the instrument chamber was 1 × 10⁻⁷ Pa as read on the panel. Measured spectra were decomposed into Gaussian components by a least-squares fitting method. Bonding energy was calibrated with reference to C 1s peak (285.0 eV).

2.3 Fabrication and measurement of the gas sensor

The indirect heating sensor was fabricated according to the literature.²⁶ The as-prepared W-doped TiO₂ nanoparticles were used as a sensing material fabricated on an alumina tube with Au electrodes and platinum wires. A Ni–Cr alloy wire crossing the tube was used as a resistor. The schematic structure of the gas sensor is shown in Fig. S1(a).† In order to improve the stability and repeatability, the fabricated sensors were aged at 320 °C for 48 h in air. The resistance–temperature curves as well as gas sensing properties of as-fabricated sensors were measured using a WS-30A system (Weisheng Instruments Co. Zhengzhou, China); see photograph in Fig. S1(b).† When the resistances of all the sensors were stable, the desired amounts of target gases were injected into the chamber (18 L in volume) by a micro-syringe. The injected gas was mixed with air immediately by two installed fans in the chamber. The values of sensor resistance and dynamic responses were recorded by the analysis system automatically, and the basic testing principle is shown in Fig. S1(c).† The gas response β was defined as the ratio of the electrical resistance in air (R₀) to that in gas (R_g). In addition, the response time was defined as the time required for the gas response to reach 90% of the final equilibrium value after a test gas was injected, and the recovery time was the time needed for gas response to decrease by 90% after the gas sensor was exposed in air again. All of the values of electrical and gas sensing properties were tested under a 30% relative humidity.

3. Results and discussion

3.1 Structure, morphology and chemical state of W-doped TiO₂ nanoparticles

Fig. 1 shows the X-ray diffraction patterns of as-prepared samples. The well-developed diffraction lines demonstrate the high crystallinity of the products. Compared with pattern a, the patterns of all the samples match well with the Bragg positions of anatase TiO₂ (JCPDS no. 78-2486), space group I4₁/amd (141), and no other peaks belonging to tungsten-related byproducts, such as WO₃ and WO₂, are found. These indicate that tungsten element has been successfully doped into the anatase-type TiO₂ and therefore the high purity of the products. Meanwhile, obvious peak widening can be observed over the whole of the patterns, which further leads to the overlapping of neighboring peaks, for instance, the overlapping of (103), (004) and (112). On the other hand, the broadened peaks generally point to the small grain size of the samples. In order to further verify the small grain size as well as the structural features of as-prepared W-doped TiO₂ nanoparticles, refinement of the diffraction pattern of 5% W–TiO₂ was carried out with the Rietveld method.²⁷ The experimental data, together with the calculated pattern obtained from Rietveld refinement and difference profiles, are shown in Fig. 2. The fact that no obvious peak was found in the difference curve indicates good agreement between the experimental and calculated curves. According to the refinement, the average grain size was calculated to be 7.75 nm, which is small enough to explain the widening of the peaks observed in Fig. 1. The other structural parameters are listed in Table S1.† From the table, changed cell parameters compared with standard value of TiO₂ and significant microstrain can be observed, which are thought to be the results of the heterosubstitution of Ti by W.

Fig. 1 XRD patterns of as-prepared pure and W-doped TiO₂ nanoparticles.

Fig. 2 Typical Rietveld output plot of as-prepared 5% W-doped TiO₂ nanoparticles. The experimental data, calculated pattern and the difference curves are shown in red, black and blue, respectively. The short vertical bars in green represent the positions of the Bragg reflections.

The morphology of as-prepared 5% W–TiO₂ was examined using transmission electron microscopy (TEM) as shown in Fig. 3. From Fig. 3(a), one can find ultra-small particles with few agglomerations uniformly distributed on the field-of-view. Fig. S2† shows the TEM image of pure TiO₂ nanoparticles, and a morphology similar to the doped sample can be observed. Fig. 3(b) shows the high-resolution TEM (HRTEM) images of the particles. The clear and well-developed lattice fringes demonstrate the high crystallinity of the samples. To verify the structural features of the samples, selected-area electron-diffraction (SAED) was carried out and shown as the inset of Fig. 3(b). Four clear characteristic diffraction rings corresponding to the reflections (101), (004), (200) and (105) of anatase can be observed, indicating the polycrystal and random orientation natures of the sample, which is in good agreement with the results of XRD. Fig. 3(c) exhibits a HRTEM imagine of one isolated W-doped TiO₂ nanoparticle. The interplanar spacing is measured to be 0.356 nm, which is close to the (101) lattice plane of anatase-type TiO₂, and the size of such a well-defined particle is estimated to be 7.1 nm. In order to have deeper understanding of the grain size of the powder, the size distribution curve of the nanoparticles was obtained by measuring around 500 particles from HRTEM images, shown in Fig. 3(d). The obtained result, 7.5 ± 1.4 nm, matches well with that of Rietveld refinement, which further verifies the as-prepared samples surely possess an ultra-small grain size. The specific surface areas for both pure and W-doped TiO₂ nanoparticles were also measured according to the Brunauer–Emmett–Teller (BET) method. The test value for TiO₂ is 173.08 m² g⁻¹; meanwhile, the value of 5% W–TiO₂ is 181.15 m² g⁻¹. Such high specific surface area is believed to be the result of the ultra-small grain size and few agglomerations of the samples. Similar values for pure and doped samples mean that the tungsten dopant has little influence on the specific surface area, which also can be attributed to the similar morphology.

Fig. 3 (a) TEM image and (b) HRTEM image of 5% W-doped TiO₂ nanoparticles, (c) HRTEM image of an isolated 5% W-doped TiO₂ nanoparticle, and (d) the size distribution of the 5% W-doped TiO₂ nanoparticles obtained by measuring the size of five hundreds nanoparticles from HRTEM images. Inset is the selected area electron diffraction of 5% W-doped TiO₂ nanoparticles.

XPS was also carried out to study the chemical state of the as-prepared samples. Fig. 4 is the XPS spectra of 5% W–TiO₂, and Fig. 4(a) shows the survey. One can find that only O, Ti and W-related peaks exist apart from C, indicating the high purity of the sample. In Fig. 4(b), Ti 2p spectra revealing two peaks of Ti 2p_1/2 and Ti 2p_3/2 at 464.7 eV and 458.9 eV demonstrate that Ti ions in the samples present in a single form of Ti⁴⁺.²² W 4f spectra shown in Fig. 4(c) reveal two peaks of W 4f_5/2 and W 4f_7/2 at 37.6 eV and 35.8 eV, with a splitting of 1.8 eV, indicating that the tungsten element in the W-doped TiO₂ exists in the form of W⁶⁺.²⁸ On the other hand, the presence of tungsten in XPS and the unobservable tungsten-related phase in the XRD pattern further illustrate that tungsten is surely doped into the TiO₂ lattice. The high-resolution XPS spectra of O 1s are shown in Fig. 4(d). Obviously, the oxygen in the W-doped TiO₂ consists of two different components: O_latt and O_x⁻. O_latt centered at 530.2 eV represents the O atoms in the crystal lattice, and this kind of oxygen atom is thought to be pretty stable and has no contribution to the gas response, while the O_x⁻ at 531.4 eV are absorbed oxygen ions, which are believed to take a very important role in the gas-sensing property.²⁹ As we know, the area ratio of these two peaks can reflect their relative amounts. By calculation, the concentrations of O_latt and O_x⁻ are estimated to be 73.43% and 26.57% in the reference of O 1s, respectively. At the same time, XPS of pure TiO₂ was also carried out, and the high-resolution XPS spectrum of O 1s is shown in Fig. S3.† It can be found that the O 1s spectrum of TiO₂ is very similar to that of W-doped TiO₂. The concentrations of O_latt and O_x⁻ are calculated to be 71.41% and 28.58%, respectively. Apparently, the relative amounts of O_x⁻ in pure and doped TiO₂ are similar; O_x⁻ and its relative amount will be further discussed in the sensing mechanism section.

Fig. 4 XPS spectra of 5% W-doped TiO₂ nanoparticles: (a) XPS survey spectra and the high-resolution spectra of (b) Ti 2p, (c) W 4f and (d) O 1s.

3.2 Effect of tungsten dopant on the resistance of TiO₂ nanoparticles

The effect of tungsten dopant on the resistance of TiO₂ was investigated by testing the resistance–temperature curves of as-fabricated sensors. Because the sensors' resistance is too high to be tested at low temperature, the resistance–temperature curves of the sensor were measured from 340 °C to 485 °C and are shown in Fig. 5. It can be found that tungsten dopant leads to a decrease in resistance of the sensors. With a lower doping level (2.5%), the sensor's resistance is slightly lower than that of the sensor made from pure TiO₂. But for higher doping levels (5% and 7.5%), the decreased resistance at the same operating temperature is distinct, which verifies our previous work that tungsten dopant can decrease the resistance of TiO₂ and allow the devices with proper resistance to be used as gas sensors. This phenomenon can be explained by the following equation:²³

Fig. 5 Relation between the resistance and the operating temperature for the as-fabricated gas sensors.

According to the conclusion drawn in the XPS section, tungsten and titanium present 6+ and 4+ oxidation states in the obtained W-doped TiO₂ nanoparticles. It is assumed that donor impurities of tungsten in the +6 oxidation state (W⁶⁺) replace the titanium atom and maintain the same oxidation state of +4 as the titanium atom, which leads to the increase of electrical conductivity. Hence, W-doped TiO₂ shows a lower resistance with same operating temperature. Meanwhile, it is obvious that all the sensors present typical negative coefficients of resistance to temperature during the temperature range. And, a good linear dependence can be observed. All the curves were fitted in the form of following:

lg R = mT + k (2)

where R is the resistance of the sensor, m is the coefficient of resistance to temperature, T is the operating temperature, and k is a constant. The fitted equations for all the sensors are shown in Table S2.† From the table, the lowest coefficient belonging to 5.0% W–TiO₂ indicates that 5% is probably the most effective doping amount of tungsten.

3.3 Gas-sensing property of W-doped TiO₂ nanoparticles

Generally, there is a temperature region in which the sensor shows the highest sensitivity when other conditions are kept the same. Hence, the relation between the response and operating temperature for both pure and doped TiO₂ toward 3000 ppm butane was investigated before the evaluation of other fundamental gas-sensing properties. The obtained results are summarized in Fig. 6. Owing to the high resistance of pure TiO₂, the corresponding curve only shows the sensitivity from 400 °C. All the curves possess the same tendency, showing a rise in sensitivity in the first stage, followed by a decrease. Due to the high stability of the butane molecule, the optimal operation temperature appears at a high temperature region, from 420 °C to 440 °C.¹⁵ Meanwhile, the conclusion can be made that tungsten dopant leads to an unexpected increase of sensitivity. TiO₂ doped with 5% tungsten shows the highest gas response, which increases from 6 in the sensor fabricated from pure TiO₂ to 17.8 at an operating temperature of 420 °C. In the subsequent tests on the gas-sensing properties of W-doped TiO₂ nanoparticles, only the sensor fabricated from TiO₂ doped with 5% tungsten was investigated, and 420 °C was chosen as the optimal operating temperature because these parameters led to the highest sensitivity.

Fig. 6 Gas response of the sensors based on pure and W-doped TiO₂ nanoparticles toward 3000 ppm butane at different operating temperatures.

Fig. 7(a) exhibits the dynamic response–recovery curve of 5% W-doped TiO₂ toward different butane concentrations from 1 to 5000 ppm. Obviously, the sensor possesses a good response–recovery property. It remains almost constant, with only a small vibration when it reaches the dynamic balance in both air and butane; meanwhile, a fast increase and decrease in response and recovery situations also can be found. To have a quantitative and accurate understanding on the response–recovery property, the dynamic response–recovery curve toward 3000 ppm butane is magnified and shown in Fig. 7(b). The response and recovery times, calculated according to the definition in the Experimental section, were calculated to be 2 and 12 seconds, respectively. On the other hand, the sensor also possesses a high sensitivity toward butane. To be more specific, the response toward 1, 10, 50, 100, 500, 1000, 2000, 3000, 4000 and 5000 ppm butane are 1.46, 3.20, 3.60, 4.55, 8.47, 11.09, 14.93, 17.8, 20.16 and 22.18, respectively. Moreover, the sensor shows good response/recovery in all the concentrations, even toward 1 ppm, which is a very low detection concentration for butane and seldom reported. As we have announced in the introduction, the low explosive limit for butane is 1.5%, which corresponds to 15 000 ppm. Hence, the sensitivity of the sensor obtained from 5% W–TiO₂ well satisfies the detection needs for butane. In practical application, a good dependence of gas response on concentration benefits the quantitative measurement of the gas concentration. Fig. 7(c) shows the calibration curve, and the experimental data from 100 to 5000 ppm were fitted as:

lg β = −0.1621 + 0.4059 lg C (3)

where β is the gas response and C is the concentration of butane. The correlative coefficient R² is 0.9992, indicating the sensor has a good linear dependence in the region from 100 to 5000 ppm. According to this fitted curve, the response toward 1 ppm butane is calculated to be 1.45, which matches well with the experimental result (1.46), which means eqn (3) not only is suitable for the region from 100 to 5000 ppm, but also is suitable for a much lower concentration. From this point of view, the low detection limit of the sensor can be reliably estimated on the basis of the fitted equation to be 399 ppb.

Fig. 7 The butane gas-sensing properties of the sensor fabricated from 5% W-doped TiO₂ nanoparticles at an operating temperature of 420 °C. (a) The dynamic response from 1 to 5000 ppm, (b) the response/recovery time toward 3000 ppm butane gas, and (c) variation of gas response to different butane concentrations from 100 to 5000 ppm.

Furthermore, the gas response of common flammable gases, including methane and carbon monoxide, were tested and are summarized in Fig. 8. It is obvious that the sensitivity toward butane is the highest among the others for the sensor fabricated from 5% W–TiO₂. For 5000 ppm tested gas, the response toward butane is 18.4 and 4.9 times that toward methane and carbon monoxide, which means the sensor possesses selectivity to some degree.

Fig. 8 Variation in gas response of the sensor based on 5% W-doped TiO₂ nanoparticles to different tested gases at an operating temperature of 420 °C.

In order to verify the reliability of the obtained butane-sensing data, the repeatability of the sensor was investigated by testing 3000 ppm butane five times under same conditions, and the test dynamic response/recovery curve is shown in Fig. 9. The average value of the gas response in these five tests is calculated to be 17.68, with a small relative deviation of 0.62%, and the response/recovery time does not show any distinct difference. Both results indicate that the butane sensor based on as-prepared 5% W–TiO₂ nanoparticles has pretty good repeatability, and the gas-sensing data used previously are reliable. Meanwhile, it is known to us that the long-term stability of the gas sensor is also an important parameter for evaluating reliability and service life. So, the long-term stability of the sensor was investigated, and Fig. 10 shows the evolution of the gas response toward 3000 ppm butane over two weeks. During the two weeks, the response value fluctuated around its average value (17.55), and the relative deviation is calculated to be 4.6%, illustrating that the sensor possesses good long-term stability.

Fig. 9 Dynamic response–recovery cycles toward 3000 ppm butane of the sensor based on 5% W-doped TiO₂ nanoparticles at an operating temperature of 420 °C.

Fig. 10 The gas response of the sensor based on 5% W-doped TiO₂ nanoparticles toward 3000 ppm butane, tested once a day at an operating temperature of 420 °C.

Table 1 presents a brief summary of the sensing performances of various reported butane-specific gas sensors. Obviously, the sensor presented in this work shows the highest sensitivity toward the same concentration. Moreover, most of the reported sensors show a bad response/recovery property, which seldom reported accurate values. For example, the recovery time of γ-Fe₂O₃ nanocrystalline toward 250 ppm butane is around 10 min.¹⁴ Su et al. reported the recovery time for Pd–SnO₂ thick film toward 500 ppm butane is more than 400 s.¹² The response time for LaFeO₃ nanocrystalline toward 1000 ppm butane is more than 100 s.¹⁵ The slow response and recovery time naturally cannot meet the need of real-time dynamic measurement. But the as-prepared W-doped TiO₂ only needs 2 and 12 s to response and recovery with 3000 ppm butane, which is a very high concentration. Furthermore, the fast response/recovery time might result from the high operation temperature, which also is higher than the reported values. Apart from these, most of the reported sensors only can detect high-concentration butane (hundreds ppm). The sensor presented in this work shows good response/recovery property toward 1 ppm butane, which further illustrates the high sensitivity of the sensor. From these results, one can find that the sensor fabricated from W-doped TiO₂ nanoparticles shows pretty good integrated sensing property toward butane.

Table 1 Comparison of the sensing performances of various gas sensors toward butane

Materials	Concentration (ppm)	Operating temperature (°C)	Response	References
ZnO thin film	1660	380	3	13
γ-Fe2O3 nanocrystalline	500	280	6.67	14
LaFeO3 nanocrystalline	500	250	6.25	15
ZnSnO3 hollow microspheres	500	380	5.79	16
Pt–SnO2 films	5000	400	4.5	17
Fe-doped SnO2 powder	1000	325	≈5.88	30
Pd–CdO nanorods	1000	275	1.54	31
	500		8.47
W-doped TiO2 nanoparticles	1000	420	11.09	This work
	5000		22.18

---

3.4 Mechanisms of enhanced sensitivity of W-doped TiO₂ nanoparticles toward butane

According to the results of the electrical and butane-sensing tests, the tungsten dopant in TiO₂ not only decreases the sensor's resistance as we have designed, but also leads to an unexpected remarkable increase in sensitivity toward butane. In general, the enhanced sensitivity can be attributed to the different morphology, specific surface area and the relative content of O_x⁻ with respect to the oxygen element.²⁶ However, it has been found that these values for pure and W-doped TiO₂ are similar. Generally, it is thought that the relative amount of absorbed oxygen ions can be detected easily by XPS, which has been carried out and proves that the relative amounts of absorbed oxygen ions are similar in pure and W-doped TiO₂ nanoparticles. However, the XPS was carried out by placing the sample in a chamber with high vacuum at room temperature, while the butane-sensing tests were conducted at 420 °C in ambient air. In our previous work, we concluded that the relative amount of absorbed oxygen ions changes with different temperatures, especially for ambient air.³² The free electrons excited by high temperatures are easily captured by the oxygen absorbed on the surface of sensing materials, which leads to the increase of the amount of absorbed oxygen ions. This process can be described by the following equations:⁹

O_2gas ↔ O_2ads (4)

O_2ads + e⁻ ↔ O_2ads⁻ (5)

O_2ads⁻ + e⁻ ↔ 2O_ads⁻ (6)

O_ads⁻ + e⁻ ↔ O_ads²⁻ (7)

Naturally, the formation of absorbed oxygen ions in this process depends on the high temperature and abundant ambient oxygen. So, the XPS results obtained in vacuum at room temperature were not suitable in explaining the enhanced sensitivity observed in ambient air at 420 °C. Taking the different components between the samples into consideration, we propose a possible mechanism for the enhanced sensitivity from the perspective of absorbed oxygen ions. For pure TiO₂, the excitation of free electrons only relies on the intrinsic thermal excitation when temperature increases. But for W-doped TiO₂, the ionization of donor impurity (tungsten atoms) also leads to the release of free electrons (as eqn (1)) apart from the intrinsic thermal excitation. According to the reaction kinetics of eqn (4)–(6), a higher concentration of electrons naturally leads to a higher extent of reaction and, consequently, more absorbed oxygen ions. On the basis of the butane gas-sensing mechanisms shown in Fig. S4† of TiO₂, it is obvious that the increased relative amount of absorbed oxygen ions results in enhanced sensitivity. However, such an assumption still needs further investigation to verify its validity.

4. Conclusion

In summary, W-doped anatase TiO₂ nanoparticles with high crystallinity were successfully synthesized by a simple, non-aqueous sol–gel route. Owing to the ultra-small grain size and few agglomerations, as-prepared W-doped TiO₂ nanoparticles possess an ultra-high specific surface area (181.15 m² g⁻¹). It is found that doping with tungsten decreases the resistance of TiO₂, and the decreased resistance is attributed to the increased free electrons caused by the donor impurity ionization from tungsten atoms. It is also observed that doping tungsten can significantly enhance the sensitivity of TiO₂ toward butane, and the best doping concentration is 5%. The sensor fabricated from 5% W-doped TiO₂ also possesses excellent response/recovery, good linear dependence, repeatability as well as long-term stability and certain selectivity, illustrating the potential of using W-doped TiO₂ nanoparticles for butane detection. Meanwhile, the significant enhancement of the sensitivity toward butane is attributed to the increased absorbed oxygen ions caused by the doping of tungsten.

Acknowledgements

This work was supported by National Natural Science Foundation of China (Grant No. 51262029) and the Key Project of the Department of Education of Yunnan Province (ZD2013006).

Notes and references

1. T. Samerjai, N. Tamaekong, K. Wetchakun, V. Kruefu, C. Liewhiram, C. Siriwong, A. Wistoraat and S. Phanichphat, Sens. Actuators, B, 2012, 171–172, 43.
2. T. J. Gray, Nature, 1948, 162, 260.
3. M. Shahabuddin, A. Sharama, J. Kumar, M. Tomar, A. Umar and V. Gupta, Sens. Actuators, B, 2014, 194, 410.
4. J. W. Yoon, Y. J. Hong, Y. C. Kang and J. H. Lee, RSC Adv., 2014, 4, 16067.
5. M. R. Alenezi, T. H. Alzanki, A. M. Almeshal, A. S. Alshammari, M. J. Beliatis, S. J. Henley and S. R. P. Silva, RSC Adv., 2014, 4, 49521.
6. Y. Li, Y. Gao, D. Jia, Y. Wang and J. Xie, Sens. Actuators, B, 2014, 198, 360.
7. S. Y. Liu, F. Zhang, H. Li, T. Chen and Y. D. Wang, Sens. Actuators, B, 2012, 161, 259.
8. P. Song, Q. Wang and Z. Yang, Sens. Actuators, B, 2011, 156, 983.
9. M. H. Cao, Y. D. Wang, T. Chen, M. Antonietti and M. Niederberger, Chem. Mater., 2008, 20, 5781.
10. T. Chen, Z. L. Zhou and Y. D. Wang, Sens. Actuators, B, 2009, 143, 124.
11. S. Vijayanand, P. A. Joy, H. S. Patil and P. Patil, Sens. Actuators, B, 2011, 152, 121.
12. P. C. Su and I. C. Chen, Sens. Actuators, B, 2004, 99, 304.
13. S. Pati, A. Maity, P. Banerji and S. B. Majumder, Sens. Actuators, B, 2013, 183, 172.
14. I. Ray, S. Chakraborty, A. Chowdhury, S. Majumdar, A. Prakash, R. Pyare and A. Sen, Sens. Actuators, B, 2008, 130, 882.
15. K. K. Bhargav, A. Maity, S. Ram and S. B. Majumder, Sens. Actuators, B, 2014, 195, 303.
16. H. Fan, Y. Zheng, X. Xu, N. Lv and T. Zhang, Sens. Actuators, B, 2011, 153, 170.
17. B. K. Min and S. D. Choi, Sens. Actuators, B, 2005, 108, 125.
18. D. Koziej, F. Fischer, N. Kränzlin, W. R. Caseri and M. Niederberger, ACS Appl. Mater. Interfaces, 2009, 1, 1097.
19. M. Stefik, F. J. Jeiligtag, M. Niederberger and M. Grätzel, ACS Nano, 2013, 7, 8981.
20. J. Hong, C. Yang, Z. Wang, S. Jiao and H. Zhu, Appl. Catal., B, 2013, 129, 333.
21. Y. D. Wang, B. M. Smarsly and I. Djerdj, Chem. Mater., 2010, 22, 6624.
22. A. L. Linsebigler, G. Lu and J. T. Yates, Chem. Rev., 1995, 95, 735.
23. Y. D. Wang, T. Chen and Q. Y. Mu, J. Mater. Chem., 2011, 21, 6006.
24. J. Xu, Q. Pan, Y. Shun and Z. Tian, Sens. Actuators, B, 2000, 66, 277.
25. (a) C. Wang, Z. Deng, G. Zhang, S. Fan and Y. Li, Powder Technol., 2002, 125, 39; (b) M. Niederberger and G. Ganweitner, Chem.–Eur. J., 2006, 12, 7282; (c) N. Pinna and M. Niederberger, Angew. Chem., Int. Ed., 2008, 47, 5292.
26. Y. D. Wang, I. Djerdj, M. Antoniett and B. Smarsly, Small, 2008, 10, 1656.
27. L. B. Mccusker, R. B. V. Dreele, D. E. Cox, D. Louër and P. Scardi, J. Appl. Crystallogr., 1999, 32, 36.
28. S. S. Lin, Ceram. Int., 2014, 40, 217.
29. X. Y. Cai, D. Hu, S. J. Deng, B. Q. Han, Y. Wang, J. M. Wu and Y. D. Wang, Sens. Actuators, B, 2014, 198, 402.
30. S. Chakraborty, A. Sen and H. S. Maiti, Sens. Actuators, B, 2006, 115, 610.
31. R. R. Salunkhe, D. S. Dhawale, U. M. Patil and C. D. Lohande, Sens. Actuators, B, 2009, 136, 39.
32. X. Liu, K. M. Pan, W. B. Li, D. Hu, S. Y. Liu and Y. D. Wang, Ceram. Int., 2014, 40, 9931.

---

Footnote

† Electronic supplementary information (ESI) available: The gas sensor, the testing principle diagram, formation process, the typical Rietveld output plot, FE-SEM images, XPS spectra of as-prepared pure TiO₂ nanoparticles and the butane-sensing mechanism. See DOI: 10.1039/c5ra20886f

---