Difference in structural and chemical properties of sol–gel spin coated Al doped TiO2, Y doped TiO2 and Gd doped TiO2 based on trivalent dopants

In this research, pure titanium dioxide (TiO2) and doped TiO2 thin film layers were prepared using the spin coating method of titanium(iv) butoxide on a glass substrate from the sol–gel method and annealed at 500 °C. The effects on the structural and chemical properties of these thin films were then investigated. The metal doped TiO2 thin film which exists as trivalent electrons consists of aluminium (Al), yttrium (Y) and gadolinium (Gd). The anatase phase of the thin films was observed and it was found that the crystal size became smaller when the concentration of thin film increased. The grain size was found to be 0.487 to 13.925 nm. The types of surface morphologies of the thin films were nanoporous, with a little agglomeration and smaller nanoparticles corresponding to Al doped TiO2, Y doped TiO2 and Gd doped TiO2, respectively. The trivalent doping concentration of the thin films increased with a rising of thickness of the thin film. This can contribute to the defects that give advantages to the thin film when the mobility of the hole carriers is high and the electrons of Ti can move easily. Thus, Ti3+ existed as a defect state in the metal doped TiO2 thin film based on lattice distortion with a faster growth thin film that encouraged the formation of a higher level of oxygen vacancy defects.


Introduction
Titanium dioxide (TiO 2 ) is nanomaterial with a vast range of applications for energy production, 1 electronics, 2 solar devices 3 and gas sensors at the nanometre length scale. TiO 2 thin lms can be obtained using the sol-gel technique, which is a reliable and low-cost chemical route. 4 The spin coating method is widely used for the deposition of materials. 5 TiO 2 is an n-type semiconductor with low conductivity. 6 TiO 2 materials need to be synthesised with a high surface to bulk ratio with a smaller particle size to attain an efficient contact in the target gas and photocatalytic or photochemical reactions arise on the surface of TiO 2 . 7 Therefore, there has been much effort has been directed to doping TiO 2 with metal atoms, where the electrons move from one atom to another atom which causes species with an overall electric charge to be formed. 8 Such species are called ions in which the ionic state is 3+, from the metal atoms. Species with overall positive charges are called cations. Therefore, individual Ti atoms losing electrons give monatomic ions, because of the Ti 4+ reduction to Ti 3+ . 9 When Ti atoms lose electrons this causes a change of the characteristic number of electrons in which the holes' carrier mobility of the metal is high and the electrons can move easily. Thus, the ionic state of 3+ charges of metal atoms can also modify TiO 2 , which shows excellent reproducibility. 10 The Ti 3+ reactions are one of the effects of point defects. Bharti et al. stated that the point defects in terms of charged oxygen vacancies and interstitial titanium ions with three or four charges affect the conductivity of the thin lms. 11 Xiong et al. reported that Ti 3+ plays a role as a defect which increases the conductivity and stability. 9 Stability here means the orientation phase of the TiO 2 doped with metal. Yang et al. explained that the anatase phase of TiO 2 thin lm also can also contribute to the stability by controlling its parameter properties. 12 There is also an indication that anatase is the most promising phase because of its higher surface reactivity to gases 13 and photocatalysts. 14 Trivalent dopants need to be used to improve the conductivity, decrease grain growth, increase the crystallinity of the peak and increase the surface area. Xu et al. showed that the defects which existed in TiO 2 from a trivalent dopant also contribute excellent results. 15 Trivalent dopants which can be used for doping are aluminium (Al), 10 gallium (Ga), 16 yttrium (Y), 17 niobium (Nb) 18 and gadolinium (Gd). 19 Apart from this, the optimisation of the doping concentration decreases the resistivity of the TiO 2 and increases the charge carrier concentration. 20 The aim of this research is to introduce Ti 3+ as a defect state of the metal doped TiO 2 thin lm based on the lattice distortion in which the oxygen vacancies have been created. Thus, it is rst necessary to study the effects of Al, Y and Gd doping on the structural and chemical properties of the TiO 2 thin lm. This study discovered that the differences have been attributed to various causes such as doping elements, phase relations, crystal size, chemical elements, structural defects, surface roughness and grain size of the thin lm.

Experimental
Firstly, for the substrate preparation, a glass sheet with a size of 2.5 cm Â 2.5 cm was cut, and this was used as a substrate. The glass was cleaned with acetone in an ultrasonic bath for 5 min at 50 C. This cleaning process was needed to remove the organic contamination on the glass substrate surface. Aer that, the substrate was rinsed in deionized water. Then, the substrate was purged to a dry state with nitrogen gas (N 2 ). The N 2 was used to reduce the oxidation, which is established using an inert atmosphere with a very low dew point. The TiO 2 sol was prepared using a method reported in previous work. The materials used were titanium(IV) butoxide [Ti(OC 4 H 9 ) 4 ; Sigma-Aldrich, 97%] as a precursor, ethanol (C 2 H 5 OH) as a solvent, deionised water as a source for adding oxygen (O), acids [glacial acetic acid (CH 3 CO 2 H) and hydrochloric acid (HCl)] and Triton X-100 (Sigma-Aldrich) as a stabilizer to avoid precipitation in the solution. Titanium(IV) butoxide mixed with ethanol, acid catalysts, Triton X-100 and aluminium nitrate nonahydrate [Al(NO 3 ) 3 $9H 2 O; Sigma-Aldrich, $98%] were stirred for 3 h (ageing process) using sol-gel synthesis under ambient conditions to obtain the gel at room temperature. Dopant precursors such as aluminium nitrate nonahydrate, yttrium(III) nitrate hexahydrate (N 3 O 9 Y$6H 2 O; Sigma-Aldrich, 99.8%) and gadolinium(III) acetate hydrate (C 6 H 9 GdO 6 $xH 2 O; Sigma-Aldrich, 99.9%) were added last. All the steps for the undoped sample and doping samples were executed sequentially. A metal ion doped TiO 2 solution can be used for the deposition of thin lms using the spin-coating method. The acquired sol was spincoated on the glass substrate at a speed of 3000 rpm over 30 s to deposit ve layers of uniform lms. The TiO 2 solution was dropped up to 10 times onto the substrates. Aer spin-coating, the layers formed were preheated at 100 C for 5 min. All the layers were annealed at 500 C for 1 h to achieve better crystallization. The annealing process was a heat treatment process which altered the microstructure of the material to change its mechanical or electrical properties. Aer the annealing process, the thin lms needed to be cooled in ambient air to room temperature. Finally, the TiO 2 thin lms were structurally and chemically, characterised. 21 The structural properties were characterized using a PAnalytical Smartpowder X-ray diffractometer (XRD). XRD is a non-destructive technique which can be used to detect the crystalline phases of unknown material phases by comparing them with records of the crystal structures of materials in the Inorganic Crystal Structure Database (ICSD).
The surface morphologies and cross-sections of the thin lms were observed using eld emission scanning electron microscopy (FESEM). The surface morphology images were magnied by 200 K and 100 K. The magnication of the thickness images was 100 K. The surface topologies and roughness were characterized using a Park Systems XE-100 atomic force microscope (AFM), at room temperature. In this research, a non-contact mode was used in which the cantilever oscillates almost on the surface of the sample and senses the van der Waals attractive force that causes a frequency move in the resonant frequency of a stiff cantilever. Furthermore, it was also realised that it was necessary to use the information on grain size and roughness of the TiO 2 thin lm. X-ray Photoelectron Spectroscopy (XPS) surface chemical analysis techniques were used to determine a stoichiometry change of the oxide for the uppermost top layer of the chemical materials' surface. Fig. 1 shows the XRD data of the TiO 2 thin lms, as-deposited undoped thin lm and Al doped TiO 2 thin lm at different doping concentrations. The anatase TiO 2 peaks (I4 1 /amd) became smaller with an increase of the Al doping concentration because of the disorder caused by the size of ionic radii of Al 3+ and Ti 4+ . Doping of Al into TiO 2 can lead to a crystallite sized anatase TiO 2 . Thus, this indicates that the crystallite sizes were smaller when compared to the undoped thin lms.

Results and discussion
Al doped TiO 2 has a smaller average size than the pure TiO 2 . This phenomenon can be described as the quantum size effect, because the accumulation of Al causes a signicant decrease in the crystal size as shown Table 1. The formula of crystal size (D) is simplied as: Strain (3) is dened as the fractional change in length and the dislocation density (d) is dened as the length of dislocation lines per unit volume of the crystal, and it is measured from the following relationship using the simple approach of Freund and Suresh 22 (see eqn (2)).
The strain and dislocation density increases with the increase in doping concentration and was attributed to the combination of the atoms in TiO 2 in the substitution. The stress (s) of the prepared thin lms is calculated using: 22 where, c 11 ¼ 208.8 GPa, c 33 ¼ 213.8 GPa, c 12 ¼ 119.7 GPa and c 13 ¼ 104.2 GPa. 23 Doping with a larger amount of dopant with 6 wt% of Al caused no characteristic peaks to be observed and the peaks related to Al were not detected, suggesting that Al did not form a signicant second phase. This means that the XRD did not detect the dopant phase either, because of the low concentration of Al doping. The samples became amorphous when a higher Al doping concentration was applied. This also applied to the optimisation of Al substitution on the Ti sites.
When an atom with a larger atomic radius replaced a small atom in the lattice, it was observed that the peak was shied towards the lower 2q value. The crystallinity of the lm inuences the transport properties of the photo-generated electrons and holes to the lm surface as well as the bandgap energy. 24 A lattice is dened as the periodic arrangement of atoms in the crystal. The smaller the ions or the higher the charge on the ions, the stronger is the attraction between the ions. This creates a strong bond. Lattice energy is the measurement of the strength of the ionic bond. The stronger the ionic bond, the more exothermic is the lattice energy. The lattice energy relies on the size of the ions and the charge on the ions 25 Therefore, during the formation of ionic bonds, the ionization energy required should be low and at the same time the electron affinity should be high, meaning that the reaction is more exothermic. The anatase peak of pure TiO 2 thin lm was measured at 25.3735 which corresponded to ICSD le no. 98-002-4276. The anatase peak with 1 wt% Al doping concentration was measured at 25.3574 which corresponded to ICSD le no. 98-020-0392 and the anatase nanocrystalline peak with 3 wt% Al doping concentration was measured at 25.2528 which corresponded to ICSD le no. 98-015-4601. Table 1 shows the lattice constants and crystallite size of the synthesised samples. The position of 2q was shied to the le and the intensity decreased when the Al doping concentration increased because of the increase in the arrangement of the atoms. In addition, the full width at half maximum (FWHM) became wider and the crystal size decreased because of the increase of the dislocation of the Al doping concentration. When 1 wt% Al doping concentration was applied, the strain of the thin lm decreased, so the stress caused a compressive thin lm because of the negative sign. However, when 3 wt% Al doping concentration was applied, the strain of the thin lm increased, so the stress of the thin lm caused a tensile thin lm because of the positive sign. Pure TiO 2 has a lattice constant of a ¼ 3.7760 and c ¼ 9.4860. The lattice of a and c were changed to a ¼ 3.7960 and c ¼ 9.4440 when Al doping concentration 3 wt% was applied. Doping Al atoms increased the lattice constant "a" of the tetragonal TiO 2 structure which may show that Al has entered into the TiO 2 lattice. However, when a 3 wt% Al doping concentration was applied, the crystallite size was 18.85 nm. A comparison of the (101) peaks of as-synthesized undoped and Al doped TiO 2 showed that the FWHM increased based on doping concentrations. Furthermore, the trend was valid for all peaks of the samples as well. Fig. 2 reveals the XRD patterns of the TiO 2 nanoparticles, which show the apparent characteristic peaks of the anatase phase. The anatase nanocrystalline peak of 1 wt% Y doping concentration was measured at 25.2599 which corresponded to ICSD le no. 98-015-4601. The anatase peak of 4 wt% Y doping concentration was measured at 25.2065 which corresponded to ICSD le no. 98-000-9854 and the anatase nanocrystalline peak of 5 wt% Y doping concentration was measured at 25.2887 which corresponded to ICSD le no. 98-015-4602. No other phases were detected and at a larger than 5 wt% Y doping concentration indicated that they were amorphous in nature. Likewise, the intensity of the diffraction peaks decreased with increasing Y doping concentration in the TiO 2 lattice. This can be ascribed to Y impurities dislocating  the crystal structure of the anatase TiO 2 . This could be because some Ti 4+ was replaced by Y 3+ and by some Y 3+ entering into the TiO 2 lattice, or the yttrium(III) oxide (Y 2 O 3 ) content was an extremely small size and could not be detected. XRD patterns of pure and Y doped TiO 2 nanoparticles show that the characteristic peaks appear with wider FWHM and the intensity of the diffraction peaks decreased with an increase in the Y doping concentration. The ionic radius of Y 3+ was 0.088 nm, which was much larger than that of Ti 4+ at 0.074 nm, so as Y 3+ entered into the lattice of TiO 2 , there was a lower diffraction intensity, and a wider diffraction peak and the crystal lattice changed, thus resulting in lower crystallinity. These results show that the appropriate Y doping concentration can maintain the crystallinity of TiO 2 nanoparticles. From the XRD analysis results, it was found that there were differences in the lattice constants between Y doped and pure TiO 2 thin lm, which also inferred that Y 3+ entered into the lattice of TiO 2 and substitutes for the Ti 4+ ion which would eventually be benecial for separating the charge carriers, extending their lifetime and effectively hindering the recombination of the electronhole pairs. 17 The position of 2q was shied to the le and the intensity decreased when the Y doping concentration increased because of the increase in the arrangement of the atoms. Furthermore, the FWHM became wider and the crystal size decreased because of the increase in the dislocation of the Al doping concentration. When a 1 wt% Y doping concentration was applied, the strain of the thin lm increased, so the stress caused it to become a tensile thin lm because of the positive sign. However, when a 4 wt% Y doping concentration was applied, the strain of the thin lm increased, so the stress caused it to become a compressive thin lm because of the negative sign.
The lattice constants of a and c were altered to a ¼ 3.7970 and c ¼ 9.5790 when 4 wt% Y doping concentration was applied as shown in Table 2. Doping Y atoms increased the lattice constant "a" of the tetragonal TiO 2 structure, which may show that Y has entered into the TiO 2 lattice. This shows that an increase in doping concentration weakens the lm's crystallinity, which may be because of the stresses caused by the difference in ion size between titanium and the dopant and the segregation of dopants in the grain boundaries for high doping concentrations. The calculated crystallite sizes of Y doped TiO 2 thin lms with 1 wt%, 4 wt% and 5 wt% were 17.2 nm to 45.23 nm. A comparison of the (101) peaks between the undoped and Y doped TiO 2 shows that the FWHM increased based on the doping concentrations. 26 Fig. 3 shows XRD patterns of undoped and Gd doped TiO 2 thin lm using 2 wt%, 4 wt% and 5 wt% of Gd. The TiO 2 indicated a dominant anatase phase for Gd-undoped sample. Then, no secondary phase was detected for Gd doped and undoped TiO 2 thin lms and more than 4 wt% Gd doping concentration indicated that they were amorphous in nature. The Gd dopant did not cause any shi in peak position of TiO 2 which may be because the amount of Gd doping was too small. Gd doped TiO 2 has smaller average size particles than the pure TiO 2 as shown in Table 1. The diffraction peak of anatase indicated an offset to the lower angle region aer doping Gd ions into the anatase TiO 2 lattice and this demonstrated the incorporation of Gd 3+ ions into TiO 2 lattice by substituting them for the Ti 4+ ion. The Gd 3+ (with an ionic radius of 0.0938 nm) doping into Ti sites was challenging because of the large difference in the ionic radii. The results of the XRD studies le a lot of uncertainty about the content of crystalline and amorphous phases in the deposited thin lms, and only veried that the modication of the chemical composition of deposited Gd doped TiO 2 thin lm really affects the growth mechanism, which has a direct impact on the microstructure. 27 The anatase nanocrystalline peak of 2 wt% Gd doping concentration was measured at 25.2608 which corresponded to ICSD le no. 98-015-4601. The anatase peak of 4 wt% Gd doping concentration was measured at 25.3202 which corresponded to ICSD le no. 98-008-2084 and the anatase nanocrystalline peak of 5 wt% Gd doping concentration was measured at 25.2887 which corresponded to ICSD le no. 98-015-4601. Table 3 shows the lattice constants and crystallite size of the synthesized   Table 4. This shows the thickness, and veried the increase when the concentration of the Al doping increased. Al doped TiO 2 thin lms revealed the size of nanoparticles to be between 14 nm and 49 nm. The nanoparticle sizes decreased when the Al doping concentration process decreased, resulting in the nanoporous porosity. In order to determine the roughness of the surface as a function of the Al concentration, it was necessary to perform AFM. Fig. 4 depicts the two-dimensional AFM images of the deposited lms using 1 wt%, 3 wt% and 6 wt% concentrations of Al doping. Table 4 shows that the grain size increased and the surface roughness of the Al-TiO 2 thin lm decreased. The lower roughness value showed that there was good homogeneity of the Al-TiO 2 nanoparticles on the surface. The change in the grain shape was determined by analysing the XRD data. Based on the XRD data it can be seen that there is considerable variation in the maximum   intensity peak, indicating the preferred growth direction of TiO 2 for each phase formed with a decreasing Al doping concentration. The decomposition reaction kinetics at the substrate surface dened the grain shape and size. Larger grain sizes provided higher surface contact between the Al doped TiO 2 thin lms and the electrode thus improving electron migration. 29 Fig. 6 and 7 show the surface morphology and the crosssectional view of Y doped TiO 2 thin lms for all thin lm samples. For samples deposited using a Y doping concentration, the thicknesses are shown in Table 5. This shows the increasing thickness that occur when the concentration of the Y doping increased. Morphological properties of the undoped and Y doped TiO 2 thin lms were examined using FESEM, and the images of the grainy looking structure became less grainy with Y doped in the TiO 2 at 1 wt%, 4 wt% and 5 wt% concentrations. Nevertheless, the granules became smaller at 4 wt% doping concentration and the morphology shied to a more uniform lm structure. The AFM images of pure and Y doped TiO 2 were prepared by using the spin coating method with yttrium(III) nitrate [Y(NO 3 ) 3 ] as a doping source as shown in Fig. 7. The Y doping had little inuence on the morphology of the TiO 2 thin lm when the Y doping concentration was increased. It can be seen that the crystalline size of the nanoparticles was about 20 nm to 49 nm. The crystalline size of pure TiO 2 nanoparticles was about 45.23 nm with a little agglomeration. With the increase in Y doping content, the crystalline size increased, and the agglomeration gradually increased when the Y doping content was at 1 wt%. The crystalline size of the nanoparticles increased to about 49 nm, and the agglomeration became more serious. This was mainly because of the sol-gel process, where the precursor of titanate reacted with NO 3 À and Ti(NO 3 ) 2 was obtained when the doped source Y(NO 3 ) 3 was added. Therefore, NO 3 À can also act as a chelating agent as well as a doping agent, both of which can slow down the condensation reaction rate, thus controlling the nucleation and growth process. Instead, the Y elements can strongly prevent the conversion of the TiO 2 crystal phase. The AFM roughness analysis also shows the values of the roughness parameters. 30 Fig. 8 and 9 show the surface morphology and crosssectional view of Gd doped TiO 2 thin lms. The thicknesses are shown in Table 6. This shows the increasing thickness when the concentration of Gd doping was increased. Fig. 9 shows that the surface morphology of the Gd doping samples were nearly nanoparticles and had high porosity as the pure TiO 2 and Gd doped nanoparticles became very small when an increase in the doping process was observed. Fig. 9 also reveals that the two-   dimensional size was increased and as well there was an increase in roughness as shown in Table 6. The surface roughness determined the number of active surface sites. In addition, the lower roughness value indicated that the TiO 2 particles had a very smooth surface. The AFM image shows that at 5 wt% Gd doping the thin lm had the lowest roughness. The average grain sizes range around 40.760 nm to 63.000 nm. The changes in the grain shape were explained by examining the XRD data. 28 When the AFM images of the deposited lms were examined, it was found that as the Gd doping increased, then the grain size also increased and the roughness decreased. The thickness increased with increasing Gd doping concentration. It can be clearly observed from the results in Table 6 that the thickness is directly proportional to the Gd doping concentration.
The Raman spectrum of the pure TiO 2 at a spectral range of 100-800 cm À1 showed that four of the peaks were prominent and the corresponding modes at 144.913 cm À1 (E g ), 398.918 cm À1 (B 1g ), 517.139 cm À1 (a combination of A 1g and B 1g ) and 636.698 cm À1 (E g ), were assigned to the anatase phase as shown in Fig. 10. The spectra of the doped TiO 2 were slightly shied as a result of the crystal structure modication from the doping. Raman spectroscopy was carried out to conrm the phase identication of the Al doped TiO 2 thin lms as well as to investigate the defects in the materials as shown in Fig. 9(a). The E g mode of the pure TiO 2 at 144.913 cm À1 arose from external vibrations of the anatase structure and indicated the formation of a long-range order, thus verifying the XRD data. Fig. 10 shows the room temperature Raman spectra of the Al doped TiO 2 , its E g mode when 1 wt% of Al is used remains at 144.913 cm À1 but when a concentration of 3 wt% of Al is applied then a high intensity of Raman peak at 145.723 cm À1 was observed.
The sharpest and strongest peak at about 145.723 cm À1 was assigned to the high frequency branch of the E g mode of Al doped TiO 2 , which was the strongest mode in the anatase phase. The strong E g mode indicated good crystallinity. Thus, from the Raman studies it was determined that the anatase phase was not altered by the presence of a trivalent aluminium dopant. It was noted that doping was considered to be the main factor that would cause the lattice distortion of the crystals, and this is usually due to the differences of the ionic radii of different elements. In this study, the crystallite quality of the Al   doped TiO 2 thin lm with 3 wt% of Al was better than the crystallite quality of the other sample of TiO 2 with a concentration of 1 wt% of Al doping. 31 The crystallite quality of the doped TiO 2 with 6% of Al was worse than with the other concentrations because no peak for the anatase phase of TiO 2 was detected. Furthermore, FWHM showed pure TiO 2 > Al doped TiO 2 at the strongest peak, meaning that the FWHM decreased when the doping concentration increased. Fig. 11 shows the Raman spectra of TiO 2 with 1 wt%, 4 wt% and 5 wt% of Y concentration at a spectral range with high intensity Raman peaks of anatase TiO 2 and no Raman peak for the rutile phase was noted. Four normal modes were observed at 144.913 cm À1 (E g ), 395.864 cm À1 (B 1g ), 517.139 cm À1 (a combination of A 1g and B 1g ) and 636.698 cm À1 (E g ). The sharpest and strongest peak at about 144.913 cm À1 was assigned to the high frequency branch of the E g modes of TiO 2 , which was the strongest mode in the anatase phase. The strongest E g mode indicated good crystallinity as the Y doping concentration reached a concentration of 5 wt% of Y. Thus from the Raman studies it could be determined that the anatase phase was not changed by the presence of the trivalent yttrium dopant. It can also be observed that doping was considered to be the main factor that would cause the lattice distortion of the crystals, and this veried the XRD data, for it was generally different from the ionic radii of different elements. 32 In this study, the crystallite quality of the Y doped TiO 2 nanoparticles with 5 wt% of Y were better than the crystallite quality of the other sample of TiO 2 with a doping concentration of 1 wt% and 4 wt% of Y. However, the value of the FWHM showed that pure TiO 2 y Y doped TiO 2 at the strongest peak of 5 wt% doping, meaning that the FWHM rose slightly when the doping concentration was increased. 17 Fig . 12 shows the Raman spectra of the TiO 2 with a Gd doping concentration of 2 wt%, 4 wt% and 5 wt% at a spectral range of 100-800 cm À1 with the highest intensity of Raman peaks of anatase TiO 2 when a 4 wt% Gd doping concentration was applied. Four normal modes were observed at 144.913 cm À1 (E g ), 395.953 cm À1 (B 1g ), 514.128 cm À1 (a combination of A 1g and B 1g ) and 636.698 cm À1 (E g ). The sharpest and strongest peak at about 144.913 cm À1 was assigned to the high frequency branch of the E g mode of TiO 2 . This indicated good crystallinity. Thus, from the Raman studies it was determined that the anatase phase was not altered by the presence of a trivalent gadolinium dopant. The change in the position and shape of the Raman active E g line of anatase TiO 2 was also detected so that the doping concentration was thought to be the main aspect that could produce the lattice distortion of the crystals, which was normally different from the ionic radii of different elements. In this research, the crystallite quality of the Gd doped TiO 2 with a doping concentration of 5 wt% of Gd was better than the crystallite quality of the other samples of TiO 2 with 2 wt% and 4 wt% Gd doping concentrations. However, the value of FWHM showed Gd doped TiO 2 < pure TiO 2 at the strongest peak of 5 wt%, meaning that FWHM decreased slightly when the doping concentration increased. 33 XPS was carried out to check the surface chemical composition, purity and the oxidation valence states of TiO 2 , and Al doped TiO 2 . In order to understand the mechanism resulting in the changes in the band gap of Al doped TiO 2 lms, the lms were investigated using XPS. The XPS being a surface sensitive technique provided information about the change in the chemical state of the constituent species of lm. Here, the variation in the chemical state of the O and Ti elements was determined in detail to correlate it to the observed variations in the band gap of the lms. 34 Fig. 13(a) shows a high resolution XPS spectrum of pure TiO 2 lm. In this spectrum, the doublet Ti2p 3/2 (binding energy 458.20 eV) and Ti2p 1/2 (binding energy 463.92 eV) occurs from the spin-orbit splitting. These peaks were consistent with Ti 4+ in the TiO 2 lattice. Next, the O1s spectrum of pure TiO 2 thin lm is shown in Fig. 13(b), which was tted the three peaks. The peaks at binding energies of 529.65 eV, 531.85 eV and 534.50 eV were attributed to lattice oxygen, titanium(III) oxide (Ti 2 O 3 ) and non-lattice oxygen, respectively. Fig. 14(a) shows the peaks in the Al doped samples now located at binding energies of 458.60 eV (Ti2p 3/2 ) and 464.32 eV (Ti2p 1/2 ). The sample was 3 wt% Al doped TiO 2 thin lms. The alteration in the position of these peaks shows the effect of an Al increase on the electronic state of the Ti element. It is most likely that some of the Ti ions are replaced with Al ions in the lattices. The rise in the area of the Ti 4+ peak shows that either TiO 2 was created in a large amount or some of the oxide structure mingled with the Al (having oxidation state Ti 4+ ) which was created aer the doping. Similarly, for the doped sample, the O1s spectrum of Al doped TiO 2 thin lm tted with three peaks is shown in Fig. 14(b). In this spectrum, three peaks  Paper at binding energies of 529.65 eV, 530.13 eV and 531.52 eV were detected which were assigned to lattice oxygen, lattice oxygen (Ti 2 O 3 ) and lattice oxygen (Ti 2 O 3 + OH), respectively. This shows that in the doping process the TiO 2 is created together with some mixed oxide.
The alteration in stoichiometry was expected because of the change in area of the relative peaks. The Ti 3+ state existed in the Al doped TiO 2 in which the oxygen vacancy defects were created. Al doping results in a small alteration in the binding energy, revealing that the Al ions were better dispersed in the replacement sites of the TiO 2 lattice and more mingled oxide structure, which could be Al-O-Ti. Fig. 14(c) shows the XPS spectrum at binding energies of 74.07 eV and 76.15 eV corresponding to the Al2p and Al2s of Al doped TiO 2 lm, respectively. The appearance of these peaks supports the existence of Al in the Al 3+ ionic state. 35 Fig. 15(a) shows the peaks in the Y doped samples that were detected at binding energies of 458.49 eV (Ti2p 3/2 ) and 464.19 eV (Ti2p 1/2 ). These peaks were assigned to 4 wt% Y doped TiO 2 . The shi in the position of these peaks shows the effect of a Y increase on the electronic state of the Ti element. Possibly some of the Ti ions were replaced with Y ions in the lattices. For the Y doped sample, the O1s spectrum of the Y doped TiO 2 thin lm tted with three peaks is shown in Fig. 15(b). In this spectrum, two peaks at binding energies of 529.88 eV and 531.15 eV were observed which were assigned to lattice oxygen (TiO 2 ) and Ti 2 O 3 , respectively. The Ti 3+ state existed in the Y doped TiO 2 in which oxygen vacancy defects were created. This showed that in the doping process TiO 2 was created together with some mingled oxide. The alteration in stoichiometry was expected because of the change in area of the relative peaks. Nevertheless, the peak at a binding energy of 530.03 eV appeared to correspond to the lattice oxygen. The Y doping concentration results in a small alteration in the binding energy, showing that Y ions were better dispersed in the replacement sites of the TiO 2 lattice and created a more mingled oxide structure, which could be Y-O-Ti. Fig. 15(c) shows the XPS spectrum at binding energies of 157.68 eV and 159.74 eV which corresponded to Y3d 5/2 and Y3d 3/2 , respectively, of the Y doped TiO 2 thin lm. The creation of these peaks preserved the existence of Y in the Y 3+ ionic state. These peaks also show that the usual Y3d spectra with spin-doublets (d 3/2 and d 5/2 ) were separated by two peaks, both of 2.06 eV. 36 Fig. 16(a) shows the peaks in the Gd doped samples that were found at binding energies of 458.77 eV (Ti2p 3/2 ) and 464.49 eV (Ti2p 1/2 ). The alteration in the position of these peaks shows the effect of Gd accumulation on the electronic state of the Ti element. Possibly some of the Ti ions were replaced with Gd ions in the lattices. The area of the Ti 4+ peak shows that either TiO 2 was created in a large amount or some mingled oxide structure with Gd (having the oxidation state of Ti 4+ ) was created aer doping. Meanwhile, the reducing area of the Ti 4+ indicated a reduction of TiO 2 in the sample, and probably the creation of the Ti-O-Gd structure in the TiO 2 lattice through the replacement of transition metal ions. The detected alteration in the peaks also showed contact between the Ti and Gd atoms and an overlapping of their 4d orbital. Fig. 16(b) shows the Gd doped sample, in which the O1s spectrum of the Gd doped TiO 2 thin lm was tted with two peaks that were detected at binding energies of 530.21 eV and 531.43 eV which were assigned to lattice oxygen (Ti 2 O 3 ) and lattice oxygen (Ti 2 O 3 ), respectively. The Ti 3+ state existed in the Y doped TiO 2 in which the oxygen  This journal is © The Royal Society of Chemistry 2018 vacancy defects were created. This showed that defects in the doping process of TiO 2 were created together with some mingled oxide. The alteration in stoichiometry was expected because of the alteration in the area of the relative peaks. Nevertheless, the TiO 2 peak at the binding energy 529.91 eV appeared to correspond to the lattice energy. Fig. 16(c) revealed that the Gd ions were better dispersed in the replacement sites of the TiO 2 lattice and created a more mingled oxide structure, which could be Gd-O-Ti. This also showed that the XPS spectrum at binding energies of 141.73 eV and 143.79 eV corresponded to Gd 4 d 5/2 and gadolinium(III) oxide (Gd 2 O 3 ) of the Gd doped TiO 2 lm, respectively. The creation of these peaks encouraged the existence of Gd in the Gd 3+ ionic state. A shake up satellite peak at 148 eV also encouraged Gd to be exhibited in the Gd 3+ state as an oxide. These shake-up satellite peaks were related to the Gd3d-O2p hybridisation. Thus, the XPS analysis indicated that Gd ions were doped into the TiO 2 matrix in the form of Gd-O-Ti. 37

Conclusions
Pure titanium dioxide (TiO 2 ) and doped TiO 2 thin lms were prepared by spin coating titanium(IV) butoxide on a glass substrate from a sol-gel, followed by annealing at 500 C. The anatase phase was observed to be annealed at 500 C. The doped TiO 2 thin lm was produced by doping with metal atoms for which the ionic state was a 3+ ion, which were Al, Gd, and Y. The ionic state, 3+, of the metal atom doping based on lattice distortion could contribute oxygen vacancy defects that provided advantages to the thin lm. The conductivity increase because of a faster growth of the thin lm encouraged the formation of a higher level of oxygen vacancy defects. These thin lms can be applied in gas sensor applications.

Conflicts of interest
There are no conicts of interest to declare.