Flame-Made Nb-Doped TiO2 Ethanol and Acetone Sensors

Undoped TiO2 and TiO2 nanoparticles doped with 1–5 at.% Nb were successfully produced in a single step by flame spray pyrolysis (FSP). The phase and crystallite size were analyzed by XRD. The BET surface area (SSABET) of the nanoparticles was measured by nitrogen adsorption. The trend of SSABET on the doping samples increased and the BET equivalent particle diameter (dBET) (rutile) increased with the higher Nb-doping concentrations while dBET (anatase) remained the same. The morphology and accurate size of the primary particles were further investigated by high-resolution transmission electron microscopy (HRTEM). The crystallite sizes of undoped and Nb-doped TiO2 spherical were in the range of 10–20 nm. The sensing films were prepared by spin coating technique. The mixing sample was spin-coated onto the Al2O3 substrates interdigitated with Au electrodes. The gas sensing of acetone (25–400 ppm) was studied at operating temperatures ranging from 300–400 °C in dry air, while the gas sensing of ethanol (50–1,000 ppm) was studied at operating temperatures ranging from 250–400 °C in dry air.


Introduction
TiO 2 is used extensively as a gas sensing material due to its change in electrical conductivity under analyte gas exposure. Sensing capability has been improved with the addition of foreign atoms such as Cr [1], Mo and W [2], Pt and Nb [3], Fe [4], and La and Cu [5]. Nb doping of TiO 2 has been used for O 2 , CO, NO 2 , and ethanol sensing. Nb doping modifies the microstructure of TiO 2 , controls grain growth mechanisms, introduces electronic defects at the surface or in the bulk of grains and so modifies TiO 2 conductivity and gas sensing. TiO 2 presents three crystalline structures: brookite, anatase, and rutile.
The Nb 2 O 5 -TiO 2 system has been prepared by several methods such as the solid state reaction of Nb 2 O 5 and TiO 2 [6][7][8], sol-gel [9,10], RF-sputtering of thin films [11][12][13], laser induced pyrolysis [14], pulsed laser deposition [15] and thick film using powder screen printing [16]. The Nb 2 O 5 -TiO 2 system can be used in applications such as varistors [17], catalysts [18], photocatalysts [19,20], and electrodes applicable to photoelectonic devices such as p-n type solar cells [21] and hybrid solar cells [22]. For gas sensing applications, it has been reported that the Nb 2 O 5 -TiO 2 system shows higher sensitivity and shorter response time as an oxygen gas sensor than undoped TiO 2 [11]. The Nb 2 O 5 -TiO 2 system can be used for sensing other gases as well, such as CO [23], CO, and ethanol [9,16,24], CO and NO 2 [16,24], and ethanol [3]. Table 1 lists literature examples of the use of Nb-TiO 2 for gas sensing applications, showing the authors, method of preparation, % Nb, sensing gas, range of detection, type of titania, size, and some remarks.
Comini et al. [3] reported that Nb-and Pt-doped TiO 2 thin films could be used for ethanol and methanol sensors. The thin films were prepared using the sol-gel process by the spin coating technique on Al 2 O 3 substrate. The sensors were tested under exposure of ethanol and methanol gases at 300 C with the concentration ranging from 500-1,250 ppm, making them feasible for development of breath analyzers (detection limit is 200 ppm). The thicknesses of the film were ranging from 60-100 nm. It was noticed that 1 at.% Nb and 0.5 at.% Pt/TiO 2 showed the best sensing performance. The TiO 2 sensors developed were sensitive at up to 500 ppm of ethanol. The response and recovery dynamics to ethanol were particularly promising for applications in food analysis, electronic noses, and breath analyzers.
In comparison to the same materials, Teleki et al. [16] reported the preparation of a flame-made TiO 2 spherical particles film of about 30 μm thickness by drop-coating of a heptanol suspension of these powders, and sensing tests at 500 C with ethanol at concentrations ranging from 10-75 ppm. The sensor showed the highest sensor signal at 75 ppm (S = 30) ethanol concentration. Secondly, Teleki et al. [23] reported in 2008 the effect on ethanol and CO gas sensing of flame-made Nb-and Cu-doped TiO 2 thick film (5 m) sensors fabricated by doctor-blading. All sensors were tested with gas concentrations ranging from 25-300 ppm during forward and backward cycles at 400 C. Niobium stabilized the anatase phase and retarded grain growth up to 600 C. The sensitivity increased by addition of either Cu and Nb to titania, and the best improvement was found for the 4 at.% Nb/TiO 2 sensor. Decreasing the Nb concentrations from 10 to 4 at.% increased the response. The response of 10 at.% Nb/TiO 2 sensor was high, but the baseline was not stable. The response time decreased with increased ethanol concentration, from 180 to 15 s for 25 and 300 ppm, respectively. The recovery time was very slow, within the 5-10 min range. This, however, decreased with increasing ethanol concentration. The anatase phase seems to be crucial as the Nb/TiO 2 sensor showed the highest response while Cu doping had no influence on the response relative to undoped TiO 2 .
Flame spray pyrolysis (FSP) is a very promising technique for synthesis of high purity nanosized materials with controlled size and high surface area in one step. FSP has been demonstrated to produce high surface area of tin dioxide nanoparticles for gas sensing [24]. The aim of this research was to apply this technique to synthesize niobium-doped TiO 2 nanoparticles. Characterization of the nanoparticles and their acetone and ethanol sensing properties were performed.

Flame Synthesis of Nanopowders
Titanium isopropoxide (Aldrich, 97%) and niobium (IV) 2-ethlyhexanoate (Strem Chemicals) were used as titanium and niobium precursors, respectively. Both precursors were dissolved in xylene (Fluka, >98.5%) and acetonitrile (Fluka, >99.5%) in equal volume with the total metal atom concentration maintained at 0.5 mol/L. The niobium concentration was varied between 1 and 5 at.%. The precursor was fed into a flame spray pyrolysis reactor [24] by a syringe pump (Inotech) with a rate of 5 mL/min and was dispersed into droplets by 5 L/min of oxygen (Pan Gas, purity > 99%) using a gas assisted nozzle. The pressure drop at the nozzle tip was kept at 1.5 bar. The water-cooled system of the reactor avoided any evaporation of the precursor within the liquid feed lines or overheating of the nozzle. The spray flame was maintained by a concentric supporting flamelet ring of premixed methane/oxygen (CH 4 1.5 L/min, O 2 3.2 L/min). In order to assure the presence of enough oxidant for complete conversion of the reactants, an additional outer oxygen flow (5 L/min) was supplied. The powder was collected with the aid of a vacuum pump (Vaccubrand) on a glass fiber filter (GF/D Whatman, 25.7 cm in diameter). During the experiment, the filter was placed in a water-cooled holder, 40 cm above the nozzle, keeping the off-gas temperature below 200 C. Scheme 1 shows the formation of Nb-doped TiO 2 by flame spray pyrolysis. Scheme 1. The formation of Nb-doped TiO 2 by flame spray pyrolysis.

Powder Characterization
X-ray diffraction (XRD) patterns were recorded with a Bruker AXS D8 Advance (40 kV, 40 mA) operating with Cu K  . The relative amounts of anatase and rutile and their respective crystallite sizes were calculated from the XRD data using the Rietveld method. BET powder-specific surface area (SSA), was measured by nitrogen adsorption at 77 K (Micromeritics Tristar) after degassing the sample for 1 h at 150 °C in nitrogen. The equivalent average primary particle diameter d BET was calculated by d BET = 6/(SSA ρ P ). Here, ρ P is the average density of TiO 2 calculated from weight percent and density of anatase and rutile where d anatase and d rutile are 3.97 g/cm 3 and 4.17 g/cm 3 respectively. Morphologies of all the flame-made powders were investigated by Transmission Electron Microscopy (TEM, Hitachi H600, operated at 100 kV).

Paste and Sensor Preparations
An appropriate quantity of homogeneous mixed solution (0.28 mL) was prepared by stirring and heating at 80 °C for 12 h ethyl cellulose (Fluka, 30-70 mPa· s) as the temporary binder and terpineol (Aldrich, 90%) as a solvent. The liquid mixture was combined with 60 mg of samples 1-5 at.% Nb/TiO 2 nanopowders and mixed for 30 min to form a paste prior to spin-coating. The resulting paste was firstly spin-coated (700 rpm) 1 time for 10 s, and then subsequently at 3,000 ppm, 2 times for 30 s on the Al 2 O 3 substrates interdigitated with Au electrodes (0.5 × 0.5 cm) to deposit sensing films. The resulting substrates were annealed in an oven at 150 °C for 1 h with an annealing rate of 1 °C/min and at 400 °C for 1 h with an annealing rate of 1 °C/min for binder removal prior to the sensing test.

Sensor Measurement
The sensor characteristics of the sensing films were determined with acetone (25-400 ppm) and ethanol (50-1,000 ppm). The flow through technique was used to test the gas-sensing properties of sensing films. A constant flux of synthetic air of 2 L/min as gas carrier was flowed to mix with the desired concentration of pollutants dispersed in synthetic air. All measurements were conducted in a temperature-stabilized sealed chamber at 20 C under controlled humidity. The gas flow rates were precisely manipulated using a computer controlled multi-channel mass flow controller. The external NiCr heater was heated by a regulated DC power supply to different operating temperatures. The operating temperature was varied from 250 C to 400 C. The resistances of various sensors were continuously monitored with a computer-controlled system by voltage-amperometric technique with 5 V DC bias and current measurement through a picoammeter. The sensor was exposed to the gas mixed sample for ~5 min for each gas concentration testing and then the air flux was restored for 15 min. The response (S) is defined in the following as the resistance ratio R a /R g [25], where R a is the resistance in dry air, and R g is the resistance in the test gas. The response time (T res ) is defined as the time required until 90% of the response signal is reached. The recovery time (T rec ) denotes the time needed until 90% of the original baseline signal is recovered [25]. After the sensors fabricated using samples undoped TiO 2 , 1 at.% Nb/TiO 2 , 3 at.% Nb/TiO 2 , and 5 at.% Nb/TiO 2 had been tested with varied operating temperatures, they were designated as S0, S1, S3, and S5, respectively.

Nanopowder Properties
The XRD technique was used to study the relative amounts of anatase and rutile phases and the evolution of crystallite sizes as a function of composition of the flame-made undoped TiO 2 and Nb-doped TiO 2 . Figure 1 shows the XRD patterns of nano-sized undoped TiO 2 and 1-5 at.% Nb/TiO 2 samples. The nanopowders were highly crystalline, and all peaks can be confirmed to be the anatase phase (JCPDS file no. . No amorphous phase and the characteristic peaks attributed to Nb or NbO 2 were found in the XRD patterns. It can be assumed that the amount of Nb doping particles was very low, which resulted in non-appearance of the Nb peaks. The relative amounts of anatase and rutile and their respective crystallite sizes were calculated from the XRD data using the fundamental parameter approach Rietveld method [26]. The average crystal sizes (d XRD ave.) were calculated based on the half-maximum widths in Scherrer equation [27] using the TOPAS-3 software, which compared with the average BET-equivalent particle diameter (d BET ) as shown in Table 2. The d XRD (rutile) was slightly increased with increasing Nb-doped concentrations. On the other hand, the d XRD (anatase) decreased with increasing Nb-doping concentrations. The rutile weight fraction percentage and d BET remained almost the same with increasing Nb concentrations. It can be concluded from Table 2 that (1) d XRD anatase were smaller than d XRD rutile (2) d BET and d XRD anatase were not affected by the amount of dopant but d XRD rutile were.   Figure 2 shows HR-TEM bright-field images of (a,b) undoped TiO 2 and (c,d) 5 at.% Nb/TiO 2 nanoparticles with different magnifications. The corresponding diffraction patterns were shown in the insets. Both samples were highly crystalline as seen from the intense electron diffraction patterns (Figure 2(a,c): insets), which were in good agreement with the XRD data. Figure 2(a-d) show the TEM bright-field images of the FSP-made (5/5) nanoparticles, which were aggregated of primary particles. Nb-doped TiO 2 nanopowder formed Ti 1−x Nb x O 2 with a fully solid solution due to diffusion of Nb atom into the TiO 2 nanoparticles because Nb 4+ has a similar ionic radius (0.64 Å) to Ti 4+ (0.605 Å). Teleki et al. [23] reported that niobium was partly incorporated in the titania lattice promoting anatase formation.  Figures 2(a,b) show the morphologies of flame-made (5/5) undoped TiO 2 and 5 at.% Nb/TiO 2 nanoparticles containing spherical nanoparticles with average diameters of 13 and 11 nm, respectively. The primary particle diameters observed by TEM were consistent with both the d BET and the d XRD . Particularly, the lattice fringes of 5 at.% Nb/TiO 2 nanoparticles were also clearly visible in a HRTEM image at higher magnification (Figures 2(b,d)). TEM bright-field images can reveal internal structure and a more accurate measurement of particle size and morphology.

Gas Sensing Properties
The gas sensitivity is usually dependent on the sensor operating temperature and the dopant. Figure 3 shows the response as a function of sensor operating temperature for undoped and doped with different Nb concentrations (1, 3, and 5 at.% Nb) for ethanol (Figure 3(a)) and acetone (Figure 3(b)) vapors in dry air atmosphere. The measurement of the resistance vs. temperature (R/T) profile of undoped TiO 2 sensor revealed a strong temperature dependence of their resistance and quite low response. It was evident from these results that the response of the undoped sensor (S0) was very poor compared to the doped sensors in different concentration concentrations. With 3 at.% Nb doped TiO 2 sensor (S3), the sensor showed maximum response to ethanol vapor at an operating temperature of 350 °C for sensor S3 and 400 °C for the other sensors, and for acetone vapor maximum sensitivity at an operating temperature of 400 °C for all sensors. The best sensitivities can be seen at the highest concentration of gases (to 1,000 ppm; S eth = 41.4 (350 °C), S eth = 31.7 (400 °C), to 400 ppm, S acet = 13.0) and response time was extremely fast about 1 s (400 °C) and 9 s (350 °C) for ethanol vapor and of about 33 s for acetone vapor. On the other hand, the response decreased at higher Nb concentration (5 at.% Nb/TiO 2 ; S5). Further increase of the dopant concentration decreased the ethanol and acetone sensitivity and deteriorated the response time. The optimum concentration of Nb doping on TiO 2 sensor was found to be 3 at.% Nb. Possibly a segregated Nb phase was formed on the surface at a higher Nb content (5 at.% Nb/TiO 2 ). For low Nb content (3 at.%), Nb nanoparticles are very small compared to TiO 2 nanoparticles and they can be well dispersed on TiO 2 nanoparticles. Thus, Nb nanoparticles are very effective catalyst. In contrast, larger Nb nanoparticles, which are formed at higher Nb contents, cannot be well dispersed and cause possible separation among TiO 2 nanoparticles. Therefore, catalytic action of Nb becomes considerably less effective. This is the reason why the gas sensitivity decreases significantly at the higher Nb content of 5 at.%.
Traversa et al. [9] reported a solubility of up to 5 at.% Nb in anatase TiO 2 . They attributed this to small segregated crystalline domains of niobia (Nb 2 O 5 ), which were not visible in the TEM bright-field images ( Figure 2). Our results agreed well with those of Comini et al. [28] showing that Nb doping improved the response to ethanol with respect to undoped TiO 2 . Teleki et al. [23] reported that 4 at.% Nb/TiO 2 showed higher response towards 25-300 ppm ethanol than 10 at.% Nb/TiO 2 at 400 °C. Figures 4(a,b) show the change in resistance of sensor S0 (Undoped TiO 2 ), S1(1 at.% Nb/TiO 2 ), S3 (3 at.% Nb/TiO 2 ), and S5 (5 at.% Nb/TiO 2 ) under exposure to reducing gas ethanol (Figure 4(a)) and acetone (Figure 4(b)) at the concentration ranging from 50-1,000 ppm and 25-400 ppm, respectively with the same operating temperature of 400 °C during the backward cycle. The original baseline (dry air) of ethanol sensing was stable during the sensing test. The resistance drastically decreased during the gas exposure with increasing VOC analyte gas concentration, typical for anatase TiO 2 as an n-type semiconductor. Nb-doped TiO 2 could exhibit a stronger n-type character and a higher electronic conductivity than undoped TiO 2 as the electron concentration in the titania lattice might increase and the Fermi level might be shifted closer to the conduction band level. The stabilized original baselines of sensors led to sensor response accuracy in terms of sensitivity and response time detection. The base-resistance of the sensor S3 (3 at.% Nb/TiO 2 ) was the lowest compared to the other sensors. This is because the sensor S3 had the appropriate amount of concentration and also could perform the sensing properties on the surface and by bulk interaction. The gas sensing behavior of semiconducting oxide sensors could be attributed to both regions of surface and bulk interactions, depending on the small grain size and the appropriate thickness of sensing films. The effects of a high conductivity of the sensor are described clearly from the interactions between VOCs gases and the surface-absorbed oxygen species such as peroxide ion (     (25-400 ppm) in dry air at 400 °C. Figures 5(a,b) show the plots of response (S) and response times (T res ) versus the ethanol and acetone vapor concentrations ranging from 50-1,000 ppm and 25-400 ppm plot for the sensors S0, S1, S3, and S5 during the backward cycle at the operating temperature of 400 °C. Because the particle size of TiO 2 was in the nanometer range and Nb is known as an excellent catalyst for VOC gases, we paid close attention to the gas sensing activity of this material. From the data, the response of all Nb doping concentrations (S1, S3, and S5) appeared to be higher than that of an undoped TiO 2 sensor (S0). The response of both gases (filled symbols, left axis) increased linearly and the response time (open symbols, right axis) decreased drastically with increasing ethanol and acetone concentrations. Moreover, it was found that the 3 at.% Nb concentration (S3) sensor showed the best sensing performance in terms of response (S = 31.7) and response time. The response time of 3 at.% Nb/TiO 2 sensor (S3) for 1,000 ppm at 400 °C was very fast-within 1 s (open triangles, right axis)-which was better than that of undoped TiO 2 (6 s) (open circles, right axis) and other doping concentrations (1 at.% Nb/TiO 2 = 3 s (open rectangles, right axis), and 5 at.% Nb/TiO 2 = 2 s (open diamonds, right axis)). The fast response time suggests a surface controlled sensing mechanism, where a steady-state adsorption of ethanol and desorption of CO 2 [25] on the sensing films was rapidly reached. This is the common interaction between the reducing gas ethanol and surface-adsorbed oxygen species of sensing layer including ethanol with those of its oxidation products (CO 2 and H 2 O) versus times in dry air. This is because CO 2 was the majority product oxidized with oxygen on the surface of semiconductor materials. This indicates a partial combustion of ethanol to CO 2 and H 2 O, as well as a release of ethoxides formed during the adsorption of ethanol on the sensing surface. Also, Nb possible increases the number of surface-adsorbed oxygen species, thus promoting reaction sizes for CO oxidations. Teleki et al. [23] reported the response of 30 with response time of 375 seconds towards 75 ppm ethanol at 500 °C for undoped TiO 2 sensing film by the drop-coating technique. With the acetone response, the sensing performances were lower than for ethanol vapor in terms of the sensor response, sensitivity, and response time. It was noticed that the Nb concentration (3 at.%; S3) sensor showed the best sensing performance at an operating temperature of 400 °C for the highest acetone concentration to 400 ppm in terms of response (S = 13.0) and response time. The response time of acetone sensors were with a few minutes. The best response time of 3 at.% Nb/TiO 2 sensor (S3) for 400 ppm at 33 s which was better than undoped TiO 2 (147 s) (open circles, right axis) and the other doping concentrations (1 at.% Nb/TiO 2 = 93 s (open rectangles, right axis), and 5 at.% Nb/TiO 2 = 126 s (open diamonds, right axis)). The response time of 3 at.% Nb/TiO 2 sensor (S3) for 400 ppm at 400 °C was slightly sluggish compared to 300 ppm of an ethanol (2 s) ( Figure 5(a), open triangles, right axis) sample. Doping the TiO 2 with 3 at.% Nb resulted in a much steeper calibration curve and the highest sensor signal compared to undoped TiO 2 (see Figures 5(a,b)). The higher sensor signal and especially the higher response (i.e., the steeper response curve) increased sensor performance.

Conclusions
FSP was successfully used for the preparation of undoped TiO 2 and 1-5 at.% Nb/TiO 2 nanopowders for application to acetone and ethanol gas sensing. The trend of SSA BET on the doping samples increased and the d XRD (rutile) increased with the higher Nb-doping concentrations while d BET remained almost the same. Nb could form a solid solution in the crystal structure of TiO 2 due to the fact the size of Nb 5+ (0.64 Å) is similar to that of Ti +4 (0.605 Å), thus the size of particles in the doping samples were not affected by Nb atoms as shown from the HRTEM. The crystallite sizes of undoped and Nb-doped TiO 2 spherical were in the 10-20 nm range. The gas sensing of acetone (25-400 ppm) was studied at operating temperatures ranging from 300-400 °C in dry air while the gas sensing of ethanol (50-1,000 ppm) was studied at operating temperatures ranging from 250-400 °C in dry air. The 3 at.% Nb-dispersed on TiO 2 sensing film showed a response of 31.7 and a very fast response time of 1 second towards 400 ppm ethanol, as compared to an undoped TiO 2 sensing film. The 3 at.% Nb-dispersed on TiO 2 sensing film also showed a response of 13 and a response time of 33 seconds towards 400 ppm acetone. The response times in our study were faster than the previously reported values [16,23]. The highest responses for acetone and ethanol occurred at 400 and 350 °C , respectively.