Preparation , Characterization , and Evaluation of Humidity-Dependent Electrical Properties of Undoped and Niobium Oxide-Doped TiO 2 : WO 3 Mixed Powders

CEMUC, Electrical and Computers Engineering Department, Faculty of Sciences and Technology, University of Coimbra, Polo 2, Pinhal de Marrocos, 3030-290 Coimbra, Portugal Physics Department, Technological Institute of Aeronautics (ITA), Pça. Marechal Eduardo Gomes 50, 12228-900 São José dos Campos, SP, Brazil Industrial Fostering and Coordination Institute (IFI), Pça. Marechal Eduardo Gomes 50, Vila das Acácias, São José dos Campos, SP, Brazil Institute of Materials Science, Federal University of Sao Francisco Valley, 48920-310 Juazeiro, BA, Brazil


Introduction
e development of new ceramic metal oxide materials provides a promising platform for diverse applications such as optoelectronics, microelectronics, dye-sensitized solar cells, and tunneling devices [1][2][3].
In particular, the production of moisture sensors of metal oxide materials requires improved selectivity and stability for water sorption.ese processes are intrinsically dependent on microstructure and prevailing transport mechanisms of resulting materials.
e doping process induced by metal oxides introduces atomic defects which a ect the overall conduction mechanisms of blends, characterizing the relative concentration of a dopant as a tuning parameter in the optimization of electrical response in terms of RH variation [4][5][6][7].
e typical process of structural modi cation is provided by mechanical mixing of the powders, molding, and sintering of the pelletized samples [8,9].e introduction of a doping agent in a mixed metal oxide ceramic has been considered an interesting strategy for improvement in the dependence of impedance value with relative humidity (RH) variation [10][11][12][13].e incorporation of dopants a ects the structure and morphology of the ceramic, providing additional path-structural water layer interaction.
In this work, the authors explored the incorporation of niobium pentoxide (Nb 2 O 5 )-an n-type transition metal oxide (E g of 3.2-4 eV) applied as a dopant for the TiO 2 : WO 3 pair.e microstructure was evaluated from SEM images, X-ray di raction patterns, Raman spectra, BET analysis, and porosimetry.
ese techniques were explored in order to evaluate the in uence of each component on the overall electrical response of the blends under controlled variation of RH.

Experimental
2.1.Materials and Methods.TiO 2 , WO 3 , and Nb 2 O 5 (Fluka) were used as received.Grain size was determined using an Autosizer II C (Malvern Instruments).e X-ray di raction patterns were obtained by means of a Philips X'Pert, PW 3040/00 using Cu-K α radiation (K α 1.5418 Å) (20 °< 2θ < 75 °) with 0.04 °of step and 0.5 s per point.Raman spectra of the samples were obtained in a HORIBA Jobin-Yvon Raman spectrometer in which the excitation wavelength was adjusted to 532 nm with a power of 25 mW.SEM images were obtained in a Philips scanning electron microscope (model XL 30 TMP), operated at 30 kV. e pore area was determined using a Micromeritics PoreSizer 9320 mercury porosimeter (as a standard procedure, chamber containing the samples was degasi ed, and then, mercury intrusion pressured analysis was performed in the pressure range from 0.5 up to 30000 psi).Brunauer-Emmett-Teller (BET) surface area experiments were provided by a Micrometrics ASAP using nitrogen gas.
e impedance of the samples was measured in the range between 100 Hz and 10 MHz, with an AC voltage of 0.5 V-no bias, in a Hewlett-Packard (model HP4294A) impedance/ gain-phase analyzer.
e experimental setup for impedance measurements consists in a 6.5-liter chamber in which the temperature is controlled with a 1 °C precision in response to the RH variation with steps of 10% in an overall range from 10 to 100%.All the experiments were performed at 20 °C.e RH values were obtained by mixing water-saturated air and dry synthetic air, in which the respective amount of each part was regulated by mass ow controllers.
e impedance spectra for each RH value were reached at a continuous ow rate of 5 l/h, after at least 90 minutes of stabilization.
e measuring electrical contacts were made of gold on opposite sides of the top surface of the samples.

Preparation of the Samples.
e pristine TiO 2 : WO 3 powder was prepared by mechanical mixing of TiO 2 and WO 3 in the ratio of 48.92 : 51.08 wt.%, respectively.e dopant (Nb 2 O 5 ) was incorporated by direct mixing with speci c amounts (2, 4, and 6 wt.%), and named as TiW-Nb2, TiW-Nb4, and TiW-Nb6 samples.All the samples were prepared in the form of pellets for the microstructural and electrical characterization.e mixtures (pristine and doped powders) were initially pelletized in the samples of 8 mm × 6 mm × 1 mm under 8 MPa of uniaxial pressure and then isostatically pressed at 200 MPa.Afterwards, the thermal treatment (under air) was conducted at 700 °C for 2 h with heating and cooling rates of 20 °C/min in accordance with a previous used procedure [8,9].

Results and Discussion
3.1.Morphology of the Mixed Pellets.SEM images of TiW-Nb2, TiW-Nb4, and TiW-Nb6 are shown in Figure 1.As can be seen, the resulting material is characterized by a porous surface with a distribution of di erent aggregation sizes of particles disposed in overlaid layers, providing free sites for molecule percolation along the structure.Statistical analysis of the images indicates a distribution of smaller grains with size around 225.4 ± 88.2 nm.A slight increase in the size of the smaller aggregates is observed for TiW-Nb4 sample, reaching 284.5 ± 131.6 nm. e aggregation level is favored by the 2 Advances in Materials Science and Engineering progressive incorporation of a doping agent (Figure 1(c)) (TiW-Nb6 reveals aggregates with a diameter of about 386.7 ± 252.3 nm).By comparing the images, it is possible to identify a slight dependence of the aggregation level on the dopant concentration.ese data can be con rmed by the di erential mercury intrusion curves (shown in Figure 2).As shown, a broad peak in the range of 0.07-0.5 μm with maximum in 0.275 μm is observed for all the samples.A weak contribution is observed in the range of 3-30 μm, shown in details in the inset, characterizing the distribution of aggregates with di erent sizes, as observed in the previous SEM images.Using the total volume of intruded mercury, the corresponding values for the total porosity were calculated: percentages of 34.4,37.1, and 37.0 were determined for the samples TiW-Nb2, TiW-Nb4, and TiW-Nb6, respectively, indicating that doping load introduces minimal di erences on the pore structure of the resulting material, which remain distinct due to the previously de ned ratio of the semiconductors used in the preparation, preserving the structural humidity properties of the template.
e BET surface area of samples is summarized in Table 1.
e results (Table 1) con rm that progressive incorporation of the dopant until 4 wt.%introduces negligible in uence on the surface area of samples.e higher level of the dopant in the sample TiW-Nb6 a ects the response (32% in terms of undoped ones) a ecting the water adsorption if compared with samples prepared under low doping level condition.

Microstructural Characterization.
e X-ray di raction patterns of a pristine sensor (mixed TiO 2 : WO 3 ) sintered at 700 °C for 120 minutes in air are compared with those of a nonsintered sample in the inset of Figure 3(a).It is noteworthy that the anatase phase of the nonsintered sample is converted into rutile phase as a consequence of annealing.Both TiO 2 polymorphs are characterized by tetragonal conguration consisting in TiO 6 octahedra that share four edges with anatase and two with rutile.e rutile is identi ed by the ICDD card no.21-1276 according to the following crystal system: tetragonal space group: P4 2 /mmm − D 14 4h with the unit cell parameters a b 4.5933 Å and c 2.9592 Å. e spectrum of tungsten trioxide, shown in the inset (B) of Figure 3(a), reveals the decrease in the intensity of the initial monoclinic phase due to the sintering process-by introduction of impurities and defects.It is also noticeable the appearance of the anorthic phase identi ed by ICDD card no.083-0951 and of the monoclinic identi ed by ICDD card no.071-0131.ese di raction patterns con rm the existence of a polycrystalline material that results from a mixture between the oxides.In Figures 3(b)-3(d), the XRD patterns of the samples TiW-Nb2, TiW-Nb4, and TiW-Nb6 are shown.
e possible incorporation of Nb 5+ ions into the crystalline structure of TiO 2 [14][15][16] can be assumed, once the corresponding ionic radius of Ti 4+ and Nb 5+ can be identi ed as a source for the absence of niobium peaks in the corresponding curves.e insets of Figures 3(b)-3(d) con rm the absence of the rutile phase and the incorporation of Nb atoms into the TiO 2 structure.XRD peaks of anatase TiO 2 were identi ed at 2θ 25.3 °, 36.9 °, 37.8 °, 38.5 °, 48.1 °, 55.1 °, 62.7 °, and 68.6 °. e latter phase also has tetragonal organization with the space group D 19  4h I4 1 and with the unit cell parameters a b 3.7842 Å and c 9.5146 Å, all identi ed by the ICDD card no.21-1272.erefore, in order to state more clearly the substitution of Ti atoms by Nb atoms, the unit cell parameters of samples under study were calculated from the anatase peaks located at 2θ 25.28 °and 2θ 48.05 °. e calculated cell parameters are a b 3.7818 Å and c 9.6261 Å for the sample TiW-Nb2, corresponding to an increase along c-direction and a decrease along a-b directions, when compared with the standard values.Such distortions reinforce the niobium incorporation: they were also observed for the sensors doped with 4 wt.%(a b 3.7822 Å and c 9.5971 Å) and 6 wt.% (a b 3.7838 Å and c 9.5410 Å).Moreover, the anatase phase showed a slight shift towards smaller 2θ angles as doping content increased, as veri ed in the changes in the peak positions 2θ 48.08 °(2 wt.%) and 2θ 48.05 °(6 wt.%).
e average crystallite size of WO 3 and TiO 2 of doped and nondoped sensors was calculated Advances in Materials Science and Engineering by the Scherrer equation (D k.λ/β.cosθ,where D denotes the crystallite size, β corresponds to the full width at half maximum, θ is the Bragg peak, in radians, λ is the wavelength of incident radiation, and k is the Scherrer constant).Table 1 summarizes the values obtained for the higher intensity re ections (101) and (002) of the TiO 2 and WO 3 , respectively.Comparing both WO 3 and TiO 2 crystallite sizes of doping sensors with the value of nondoped sensor, it is observable that niobium incorporation promoted a slight decrease of crystallite dimensions for both oxides.ese results indicate that niobium has been incorporated into the crystal lattice of titanium and provides considerable changes in the electrical response to humidity of the pellets.
e Raman spectrum of nondoped TiO 2 : WO 3 and TiW-Nb2 samples is compared in Figure 4(a).
e incorporation of niobium originates a less intense peak in the spectrum, corresponding to the position at 635 cm −1 , and two other even less intense peaks at 513 and 390 cm −1 , attributed to the anatase phase, con rming that niobium hinders the phase change of titanium dioxide.e peak at 440 cm −1 in the spectrum of doped pellet is assigned to tungsten trioxide (W 5+ O), also visible in the Raman spectra of TiW-Nb4 and TiW-Nb6 (Figure 4(b)).e asymmetric shoulder formed is more evident for the higher doped sensor and might be associated with the proximity of the monoclinic WO 3 (mode G) vibrational mode at 605 cm −1 [17,18].

Electrical Impedance Characterization of Samples.
Electrical impedance spectroscopy has been progressively reported in the literature as a promising tool applied in the identi cation of phase transitions in materials [19][20][21][22] and transport mechanisms in structures of solid state (ceramics) and soft matter.
e frequency-dependent excitation provides information about transport and polarization (transport, ionic di usion, and charge separation).e graphical representation (Nyquist diagrams) o ers a direct visualization of the di erent

4
Advances in Materials Science and Engineering mechanisms (relaxation process-from a characteristic semicircle-and di usion-from a straight line at lowfrequency limit) [23].
In terms of surface conductivity in ceramics due to the progressive water adsorption [24], three di erent mechanisms are present [25].At low RH range, the monomolecular adsorption process takes place as a response to surface modi cation due to water molecule incorporation, reaching a rst complete layer, the chemisorbed one, in the range 20 to 40% RH.Above 40% RH, the proton conductivity of water dominates and di usion e ects tend to be more e ective, improving the surface conductivity.
e continuous formation of water layers, physisorbed ones, favors the ionic transport participation in the overall conduction process.e porous structure tends to be lled by water molecules, allowing proton transport between adjacent water molecules.With the increasing water adsorption rate, the surface conductivity tends to assume a constant value.Nyquist diagrams in Figures 5(a) and 5(b) reveal that a minimal variation is observed for all the samples if the considered RH concentration is above 70%, con rming the previous analysis; below 70% RH, the relaxation dominates over di usion and two overlapping semicircles are present.At increasing RH (in the range of 40%-60%), di usion contributions are added to the surface contribution due to the tunneling process along water layers.Above 70% RH, di usion tends to be the dominant mechanism-as a result of mutual contribution of surface water layer and bulk water-lled pores. is pronounced behavior is similarly observed for the sample TiW-Nb4 (Figure 5(b)).Increasing dopant concentration a ects the dependence of the electrical spectrum with RH variation.As expected, the minimal variation in the Nyquist curves' characteristic diameter reveals that saturation in the transport is reached.
ese results are con rmed in Figure 6 (impedance measured at f 1.3 kHz and 4 kHz, at 20 °C for di erent doping levels).As shown, the incorporation of niobium in the composite reduces the range of variation of resulting  devices as a function of RH due to the doping level established by the additive.

Electrical Circuit Modelling for Sample's
Response.Equivalent circuits have been considered as an interesting source of parameters [25][26][27] that have been associated with di erent mechanisms in overall electrical response.e circuit represented in Figure 7 has been explored by the authors in recent works [8,9], and the components are described below: C geo is used to represent the geometrical capacitance, while the bulky granular response is assigned to R 1 //C 1 .
e grain boundary contribution is assigned to the parallel circuit R 2 //C 2 , while the component R 3 //C 3 characterizes the surface contribution from electron tunnelling along the water layers disposed above the semiconductor surface.e charge di usion is represented by a constant phase element, CPE [25,28].For both referred di usion mechanisms, the interfacial character of their impedance makes it partly capacitive as well as resistive: CPE po has to do with the contribution of the pores, due to di usion phenomena taking place inside the water-lled pores, and CPE el is related to the electrode-water layer interface di usion phenomena that take place at that interface.
In Tables 2-4, the best-t parameters for the proposed electrical equivalent circuit are summarized.A el and n el represent the two parameters of the impedance CPE el , while A po and n po represent the two parameters of the impedance CPE po .
Due to the strong variation observed in two of these parameters, R 1 and R 2 were chosen as relevant parameters for RH dependence of impedance data.
Figure 8 shows the dependence of R 1 as a function of RH for di erent doping levels.A general tendency of reduction of the corresponding bulk resistance is observed with increasing RH concentration, as a consequence of progressive di usion of water molecules into the bulk of the devices.In addition to this, it is possible to identify a niobiumdependent variation of R 1 with RH.
e minimal variation is observed for the sample TiW-Nb6, while the maximum one is reached for the sample TiW-Nb4.
e water di usion in the corresponding structure introduces a competition for transport mechanisms with intrinsic electrical properties.e higher doping level samples are minimally a ected by water incorporation, since e cient channels for current transport are established under doping; the water impregnation in the bulk of the samples represents a minimal perturbation in the corresponding transport process.In the opposite direction, for the sample TiW-Nb4, this process is extremely dependent on water incorporation.e dependence of grain boundary resistance (R 2 )shown in Figure 9-con rms the observed behavior.Maximum variation in R 2 is observed for the sample TiW-Nb4, while negligible variation is observed for the sample TiW-Nb6.6 Advances in Materials Science and Engineering RH (%) ) ) ) RH (%) ) ) ) RH (%) Advances in Materials Science and Engineering e sensitivity of all three doped samples taken at 1.3 and 4 kHz is plotted in Figure 10.
e sensitivity was calculated using the ratio between the conductivity of the sensor exposed to a certain moisture concentration and the conductivity of the sensor under a dry air atmosphere.As can be seen, the sample that exhibits the best sensitivity is the one doped with 4% of niobium, in accordance with the discussed morphology of the resulting material and with the impedance changes with moisture previously discussed.ese results con rm that structural modi cation provided by niobium returns the best sensibility to RH at 4% of niobium.At this condition, the competition established between the doping level, induced by additive, and electrical response of structure to water layer incorporation is maximized, characterizing an optimal condition for TiW-Nb-based RH samples.8

Conclusion
Advances in Materials Science and Engineering is induced structural modi cation requires speci c and low concentration of the dopant (sample TiW-Nb4) in order to optimize the RH sensitivity of the resulting composite.Above this critical value, the high conductivity of the obtained devices a ects the sensibility degree of the structure, due to the progressive aggregation of grains and higher surface conductivity at low RH condition.

Figure 7 :
Figure 7: Proposed equivalent circuit for the sensors.

Figure 6 :
Figure 6: Impedance change with RH, at 1.3 kHz (a) and 4 kHz (b), at 20 °C for all three sensors (it was not possible to calculate values of the impedance modulus for 10% RH in the TiW-Nb2 sensor at 4 kHz).

e
doping level established by Nb 2 O 5 in TiO 2 : WO 3 composites preserves the anatase phase and provides modi cations at the atomic level of the resulting structure (niobium modi es the crystal lattice of titanium) as detected by XRD data.

Figure 9 :
Figure 9: Dependence of R 2 -tting parameter of equivalent circuit with RH variation at di erent doping levels.

Table 1 :
BET surface area and crystal dimensions calculated from Scherrer's equation.Log di erential intrusion as a function of pore diameter of the TiO 2 : WO 3 -doped samples with 2, 4, and 6% of Nb 2 O 5 (inset: macropore region zoom).

Table 2 :
Fit parameters for presented Nyquist plots of the sensor TiW-Nb2 at 20 °C.

Table 3 :
Fit parameters for presented Nyquist plots of the sensor TiW-Nb4 at 20 °C.

Table 4 :
Fit parameters for presented Nyquist plots of the sensor TiW-Nb6 at 20 °C.