1. Introduction

The photoactive properties of titanium dioxide (TiO₂) have been utilised for two main fields of application: (i) environmental remediation including water/air purification, and selfcleaning surfaces; (ii) solar energy conversion through photoelectrolytic water splitting to yield hydrogen and oxygen [1–4]. Nanoparticle TiO₂ electrodes have also found application in dye sensitised photovoltaic cells [5, 6]. The reported efficiency of TiO₂ for solar applications in (i) and (ii) above is low due to its wide band-gap (3.0–3.2 eV) requiring ultraviolet irradiation (UV). In attempts to increase the solar efficiency by utilising visible light, TiO₂ has been doped with other elements with the aim of producing a red shift in the absorbance spectrum. Transition metal ions are commonly employed dopants due to their strong absorption of visible light and their similar ionic radius to the parent titanium ion. However, studies into the effectiveness of transition metal ions as dopants to improve solar efficiency have reported conflicting results. For example, chromium doping of TiO₂ has been reported to yield both an increase in activity [7–10] and a decrease in activity [11–16]. The majority of dopant studies do report a red shift in the absorption spectra of the doped material, however, this does not necessarily correlate with an increase in photoactivity [13, 17–20]. Unfortunately, there are few publications which report both the photocurrent response and absorption spectra for doped TiO₂. Where they exist, the results are conflicting, with some researchers reporting a good correlation between the photoactive spectrum and the absorption spectrum [8, 11, 17] while others have not found a correlation [21].

The direct comparison of published results for the efficiency of doped TiO₂ photocatalysts is not straightforward as reported materials have been produced using different methods and have been tested under different conditions. For example, if the photocatalytic degradation of a model pollutant is used as the measure of efficacy then changes in surface adsorption, due to changes in surface PZC resulting from doping, may result in an increase in photocatalytic degradation rate of certain organics but not others [15, 22].

Only a few papers discriminate between the effects of the dopant ion energy levels and the dopant ion-induced oxygen defect energy level within band structure. An oxygen defect level is introduced when the dopant has a lower valence than titanium thus leaving oxygen bonding unsatisfied and a new energy level is produced.

The effect of this induced defect level is difficult to study within a single dopant regime. Analysis of the dopant induced oxygen level within the band structure of platinised TiO₂ was studied by Karakitsou and Verykios using dopants of different valences to titanium [23]. Using hydrogen evolution as the measure of photoactivity, they reported that dopants with a valence greater than 4 gave higher activity, and dopants with valence less than 4 had a lower activity, as compared to the undoped TiO₂. The decrease in the photoactivity with dopants of valence less than 4 was attributed to an increased rate of charge carrier recombination.

The effect of oxygen defect levels in the TiO₂ band structure have previously been reported in literature [24, 25]. Siripala and Tomkievicz were the first to examine the effect of these defects on charge transfer at the TiO₂ electrolyte interface [26]. They reported that oxygen defect levels were responsible for observed photocurrent using sub-band gap illumination. The sub-band gap photocurrent was only observed between applied potentials of −0.6 V and +0.6 V (SCE). Other workers have reported the generation of oxygen defect states via plasma modification of TiO₂ which yielded a visible photocurrent response [27].

In this work, we report the effect of doping with transition metal ions on changes in UV/Vis absorption spectra, crystal phase, current-potential response, and spectral photocurrent response.

2. Experimental

A sol gel procedure, adapted from the work of Murakami et al. [28] using ammonium acetate as a catalyst, was used to prepare TiO₂ films. A solution of 0.0973 g ammonium acetate, in 0.2 cm³ water/15 cm³ butan-1-ol, was added drop-wise to solution of 0.377 g titanium (IV) butoxide in 10 cm³ butan-1-ol. For doped samples, the relevant dopant acetate salt was added along with the ammonium acetate to give 0.2, 1.0, or 5.0 atom% (with respect to titanium). Dopants investigated were V, Cr, Mn, Fe, Co, Ni, and Cu. The resulting transparent sol gel was stirred for a further 1 hour before dip coating onto previously cleaned 2 × 2 cm² titanium foil coupons (1.27 × 10⁻² cm thick, Aldrich). Films were deposited in five coat cycles, drying between coats with an infrared lamp, before annealing at 500^°C in air for 1 hour with a temperature ramp rate 1^°C min⁻¹. Electrical contact was made to the coupons by attaching copper wire with conductive silver epoxy (Circuit Works, Chemtronics). The contact and coupons were painted with a negative photoresist (KPR resist, Casio Chemicals) and UV cured, leaving a defined area of TiO₂ exposed.

Electrochemical characterisation of samples was performed under potentiostatic control (Autolab PGSTAT30) in a one-compartment cell with quartz window, platinum basket counter electrode, and a saturated calomel reference electrode (SCE). The supporting electrolyte was 0.1 M sodium perchlorate, and irradiation was provided from a 1 kW Xe arc lamp (ss1000 Spears Robinson) either; unfiltered, with an AM1 filter (to simulate solar conditions) or with a sodium nitrite filter (λ > 420 nm) for high-intensity visible light. For monochromatic irradiation, a high-intensity monochromator (GM252, set for 10 nm band pass) was positioned between the source and the photoreactor. Light intensity was measured using a calibrated spectral radiometer (Jobin Yvonne Gemini spectral radiometer). Linear sweep voltammetry (LSV) was carried scanning from –1.0 V to +1.5 V with a scan rate of 10 mV s⁻¹. The current-time response was measured at fixed potentials using chopped irradiation (Uniblitz vmm-t1, Vincent Associates). The cell temperature was maintained at 25^°C ± 2^°C. The presence or absence of dissolved oxygen was controlled by sparging with air or oxygen free nitrogen (OFN), respectively. All potentials are reported versus SCE.

Elemental analysis was carried out by Auger electron spectroscopy (AES) (Kratos ) in a vacuum of 10⁻⁶ torr with electrons accelerated from a field-emission source at 4 kV accelerating voltage to give a beam current of 2 μA. Depth profiling was achieved using an argon etch.

UV-Vis transmission spectra of films deposited on quartz were measured using a spectrophotometer (Lambda 11, Perkin Elmer). Optical band-gap values were calculated by extrapolation of the absorption band edge. Crystal phase was analysed by Raman spectroscopy (LabRam 300, ISA instruments) by comparing to commercial anatase (99.5%, Aldrich) and rutile (created from predominantly rutile sample (Aldrich) by heating to 800^°C). The source was an argon laser  nm. Optimum analysis was performed with an acquisition time of 5 seconds, averaged over 10 accumulations, with the neutral density filter set to 50%.

3. Results and Discussion

AES measurements indicated that not all of the corresponding dopant was incorporated at the hydrolysis stage of sol-gel procedure. Determination of dopant concentrations at initial precursor levels below 1.0 atom% was not possible with the AES system used in this study. Analysis of doped samples containing 1.0 atom% and 5.0 atom%, as precursor, confirmed the presence of the dopants (Table 1), however, they were at much lower levels (ca. 40%) than that of the initial precursor dopant level.

Table 1: Elemental analysis by AES for Cu and Cr doped TiO₂ films.

For simplicity, we will continue to refer to percentage doping of samples as the initial concentration of dopant added to the sol precursor. Depth profiling of the thin films showed that the dopant species were preferentially located on the surface with lower concentrations present in the bulk. Additionally, a high proportion of carbon was measured which originated from the alkoxide precursor used in the sol-gel process. Carbon was present in all samples including the undoped sample.

Figure 1 shows the Raman for a TiO₂ film and a TiO₂:V film, compared with anatase and rutile powders. Analysis of the powder samples showed all relevant Raman vibration modes of anatase and rutile correlating with previous literature on Raman analysis of TiO₂ [29]. Comparison of the sol-gel derived thin films showed all samples to contain the vibrational mode at 144 cm⁻¹ common to both anatase and rutile. However, the rutile signal at 144 cm⁻¹ is much weaker compared to the anatase signal. Since the sol-gel film shows a large peak at 144 cm⁻¹ but no other rutile peaks, then we can assume it was predominantly anatase. The TiO₂:V sample showed peaks present at 447 cm⁻¹ () and 612 cm⁻¹ () representative of rutile and 519 cm⁻¹ () representative of an anatase crystal structure. The spectra of the undoped sample and all other doped samples were almost identical showing a predominantly anatase crystal phase, with the 399 cm⁻¹ and 519 cm⁻¹ peaks, along with the 639 cm⁻¹ peak associated with anatase.

Figure 1: Raman spectra of (a) undoped TiO₂ film, (b) vanadium doped TiO₂, (c) commercial anatase powder, (d) commercial rutile powder (following high-temperature anneal).

UV-Vis absorption spectroscopy of the doped films, deposited on quartz, showed a minor red shift in the absorbance spectra as compared to the undoped film, however, this was mainly confined to the region from 250 nm to 350 nm (examples in Figure 2).

Figure 2: UV/Vis absorption spectra of doped TiO₂ films (a) V 0.2 atom%, (b) Cr 0.2 atom%, (c) Mn 0.2 atom%, (d) undoped TiO₂.

With all dopants investigated, an increase in the dopant concentration resulted in an increase in the absorbance between 290 nm and 350 nm with a small increase in absorbance at wavelengths greater than 350 nm. For example, the change in absorbance with concentration of dopant using TiO₂:Mn is given in Figure 3. Changes in the absorption band edge, obtained by extrapolation of the UV/Vis absorption data, were in the range of 0.01–0.25 eV. The error in these calculations was not acceptable and therefore we do not report the band-gap values.

Figure 3: UV/Vis absorbance versus dopant concentration for TiO₂:Mn (a) 5 atom%, (b) 1 atom%, (c) 0.2 atom%, (d) undoped.

The photoelectrochemical characterisation of the films using linear sweep voltammetry (LSV) showed a typical -type semiconductor response with negligible anodic current in the dark and a significant increase in the anodic current under AM1 solar simulated irradiation (Figure 4).

Figure 4: Linear sweep voltammograms under AM1 solar simulated irradiation for doped TiO₂ films (0.5 atom% dopant) (a) undoped, (b) Fe, (c) Co, (d) Ni, (e) dark response for undoped film.

The photocurrent was dependent on the applied potential, and the undoped TiO₂ sample showed a larger photocurrent response than any of the doped samples. Furthermore, a decrease in photocurrent with increasing dopant concentration was observed with all doped samples. Air sparging of the electrolyte prior to LSV resulted in a positive shift in the onset potential for anodic current (Figure 5 and Table 2).

Table 2: Anodic photocurrent onset potential for different preparations in the presence of oxygen (the undoped sample gave an onset potential of –0.8 V).

Figure 5: Effect of electrolyte air sparging on the I-E response for TiO₂:Cr (0.2 atom%) under unfiltered irradiation (a) light, following air sparging, (b) light, following nitrogen sparging, (c) dark, following air sparging, (d) dark, following nitrogen sparging (SR = 10 mV s⁻¹).

The photocurrent response spectra of these films were obtained using chopped monochromatic illumination (10 nm band pass). The spectral current-time response for the undoped TiO₂ electrode is given in Figure 6. The incident photon conversion efficiency (IPCE) was calculated by

Figure 6: Current-time response under chopped irradiation for undoped sample as function of irradiation wavelength from 250 to 430 nm (left to right) in 10 nm intervals. Measured in a one-compartment cell with applied potential of +1.0 V and 0.1 M sodium perchlorate as a backing electrolyte with air sparging.

where J is the photocurrent density, is the incident light intensity, and F is Faraday’s constant.

The %IPCE versus wavelength is given in Figure 7. The monochromatic light intensity was a limiting factor in these experiments yielding very low photocurrents. A small photocurrent response at wavelengths above 400 nm was observed for a number of samples (see Figure 6), however, a change in current less than three times the deviation of the background current (i.e., 30 nA) was not taken as significant. The doped samples exhibited no red shift in the %IPCE compared to the undoped TiO₂.

Figure 7: Incident photon to current efficiency for samples doped to 0.5 atom%. Dopant: (a) none, (b) Fe, (c) Cu, (d) V, (e) Co, (f) Cr, (g) Ni, (h) Mn (measured in a one-compartment cell at an applied potential of +1.0 V and 0.1 M sodium perchlorate electrolyte with air sparging).

To overcome the limitations of the low light intensity of the monochromatic source, the photocurrent response was measured under visible excitation using a NaNO₂ filter (λ ≥410 nm) with the Xe source. All samples showed a photocurrent response under sub-band gap irradiation which was dependent on applied potential with the photocurrent maxima observed between −0.5 and −0.1 V (see Figure 8). The peak potentials and onset potentials for anodic current are given in Table 3. AES analysis confirmed the presence of carbon impurities at high levels and this is common with sol-gel routes employing alkoxide precursors. It has been previously reported that carbon doped TiO₂ gives a visible response [30]. All samples contained carbon impurities and there was a small visible response observed for the undoped TiO₂ sample. This visible photocurrent for the undoped sample, which may be due to carbon impurities, is much less than that observed for the metal ion dopants (see Figure 8 and Table 3). Recently, Asahi et al. [31] reported the carbon defect level to be too low to give an optical transition in the visible. Changes in the optical or photoactive spectra of TiO₂ due carbon doping have been correlated to an increase in the concentration of Ti³⁺ stabilised by carbon impurities [32, 33].

Table 3: Photoelectrochemical data for undoped and 0.5 atom% doped TiO₂ films under visible irradiation.

Figure 8: Visible ( nm) photocurrent response (a) TiO₂:Co 0.5 atom%, (b) undoped TiO₂, and (c) dark current for undoped TiO₂.

The presence of metal ion dopants will result in oxygen vacancy states or single-bonded oxygen (possibly hydroxyl) due to the lower coordination of the dopant with respect to the parent Ti⁴⁺. Nakamura et al. [27] reported that H₂ plasma treatment of TiO₂ resulted in the formation of oxygen vacancy states located around 2.02–2.45 eV above the valence band corresponding to a wavelength between 506 and 614 nm. Electrons may be promoted from the valence band to the vacancy states by visible light excitation. They reported that plasma modification of TiO₂ allowed visible-light-driven oxidation of NO.

Siripala and Tomkievicz [26] investigated interactions between photoinduced and dark charge transfer at single crystal TiO₂-aqueous electrolyte interfaces. They reported a sub-band gap irradiation resulted in electron injection from the valence band to states in the gap which lie around 0.6 eV below the bottom of the conduction band. Carrier excitation through such surface states should be much less potential dependent than carrier excitation through the bulk. Indeed they report that sub-band gap photocurrent is observed only between −0.5 and +0.4 V (SCE) at pH 6.5 (reported flat band potential was ca. −0.8 V). They proposed that the surface state was an intermediate of the dark reaction forming TiO₂–H, the energy for which lies within the band-gap.

The photocurrent under visible light excitation observed at negative potentials is consistent with previously published results which attribute this visible response to donor/acceptor energy levels below the conduction band caused by surface defects [24–27]. Photocurrent under sub-band gap illumination is only observed in the potential region where donor/acceptor energy levels are filled, that is, more negative than −0.1 V in this case. The presence of defect states is usually reported to have a negative effect on photoactivity because they act as recombination centres for charge carriers. However, under appropriate applied potential, these defect levels can act as electron donor states yielding a photocurrent response under sub-band gap irradiation.

To investigate the potential dependence of these defect levels, the current-time response of TiO₂ electrodes under chopped illumination was examined. Salvador [34] investigated the kinetics of recombination involving defects, surface groups, and absorbed species, by photocurrent transient measurements. Using single crystal TiO₂ electrodes, they reported that the initial photocurrent produced in the instant of illumination () decayed with time to a steady state () due to recombination on defects and surface states. By plotting the ratio of / versus potential, the critical band bending was determined.

Figures 9 and 10 show examples of transients produced in this work for the undoped and doped samples, respectively. For the undoped samples, the initial photocurrent spike is absent at potentials more positive than +0.6 V whilst for the doped samples, the initial spike was always present. Additionally, photocurrent showed -type to -type conversion at negative potentials close the . Matsumoto et al. [21] reported increasing -type character at negative potentials following doping with transition metals which resulted in a lowering of the anodic photocurrent in comparison to the undoped sample, with higher ratios correlating to higher defect densities.

Figure 9: Changes in photocurrent transients as a function applied potential. Example of undoped sample under polychromatic illumination (applied bias from 1.0 V to −1.0 V in 0.2 V steps left to right).

Figure 10: Changes in photocurrent transients as a function applied potential. Example of 0.2% chromium doped sample under polychromatic illumination (applied bias from +1.0 V to −1.0 V in 0.2 V steps left to right).

The films used in this study were made using a sol-gel route which gave oxide films consisting of nanoparticles with a primary particle size around 38 nm [35]. For nanoparticles in this size range, band bending is not significant. However, initial and steady state photocurrent measurements can still be used to examine recombination as a function of potential by subtracting from . Figure 11 shows the potential dependence of recombination in these electrodes. The peaks in recombination ( − ) with respect to potential are tabulated in Table 4.

Table 4: Potential dependence of recombination.

Figure 11: − as a function of potential from 0.2 atom% chromium doped sample.

The magnitude of − is observed to follow the same potential dependence as the sub-band gap photocurrent response. Under band-gap irradiation, the defect states act as recombination centres lowering the photocurrent response at negative potentials; however, under sub-band gap irradiation, the defect states act as electron donors showing a visible photocurrent response. This is an agreement with other reported results [36].

There is a small visible photocurrent response is observed at potentials more positive than −0.6 V due to defect states. Nevertheless, it is clear that the primary band-gap photocurrent response is decreased by the addition of metal ion dopants, which act as charge-carrier recombination centres, and that the sub-band gap photocurrent is only a fraction of the band-gap photocurrent and, therefore, may not be relevant for practical applications including the photoelectrolytic splitting of water under solar irradiation.

4. Conclusions

Thin films of transition metal doped (0.2, 1.0, and 5.0 atom%) TiO₂ were prepared on titanium foil using a sol-gel route catalyzed by ammonium acetate. Raman spectroscopy confirmed anatase crystal phase in all samples except in the case of the TiO₂:V sample where a mixture of anatase and rutile was obtained.

The films doped with transition metal ions showed lower photocurrent response than undoped samples under simulated solar irradiation. The photocurrent response spectra of the doped films were similar to the undoped sample with no major red shift. Sub-band gap irradiation of the samples gave a small photocurrent at negative potentials which is indicative of electron promotion from a filled defect level, the occupation of which is potential dependant. Examination of defect level potential dependence by analysis of photocurrent transients (from −0.8 to +1 V) showed a good correlation with the potential dependence in visible illumination studies.

The primary band-gap photocurrent response is decreased by the addition of metal ion dopants, which act as charge-carrier recombination centres, and the sub-band gap photocurrent is only a very small fraction of the band-gap photocurrent.

Acknowledgment

Department of Employment and Learning Northern Ireland for funding JWJ Hamilton.