1 Introduction

There has been increasing focus on the development of renewable and clean energies based on solar power. Nanostructured semiconducting materials have been developed as potential solutions for the utilisation of solar energy for hydrogen production and the photodegradation of organic pollutants [1,2,3]. Although photocatalytic TiO2 has been investigated extensively, its wide band gap and high electron–hole recombination rate have limited its applicability [4]. In order to improve the photocatalytic performance, doping has been investigated extensively with the aim of improving the photoresponse under solar light [5,6,7,8].

One of the most common methods to improve these characteristics in semiconducting oxides is through the use of dopants. One of the most important of these parameters is the optical indirect band gap, which can be lowered by convergence of the conduction and valence band positions as well as the introduction of shallow midgap energy levels, which derive from the introduction of lattice defects [9], the presence of biaxial tension in the a–b plane of TiO2 imposed by dopants [10], increasing the crystallinity [11], and/or activating intervalence charge transfer (IVCT) [11]. Another important parameter that can be manipulated is extension of the electron–hole pair recombination time, which can be enhanced through the presence of deep midgap energy levels, which again derive from lattice defects [12]. Two other important parameters are charge carrier diffusion distance and density of surface-active sites, the effects of which can be enhanced through reductions in grain size [13, 14]. Finally, charge carrier mobility can be improved by reduction in scattering by lattice defects, grain boundaries, and surfaces [15].

Doping of TiO2 using transition metal elements has attracted considerable interest as it has been shown to have the capacity to enhance the photocatalytic efficiency [16,17,18,19,20,21,22,23,24]. Of the various transition metal ions, Fe and Mo are of considerable interest owing to multiple valence states (Fe3+, Fe2+, Mo6+, Mo5+, Mo4+) and the resultant potential for the imposition of different midgap defect energies and IVCT. Table 1 summarizes prior work on Fe or Mo single doped TiO2.

Table 1 Survey of experimental work on Fe-doped and Mo-doped TiO2 photocatalysts
Full size table

However, there is only a limited amount of work that has been done on TiO2 codoped with two transition metals [37,38,39], which enhances the capacity to increase charge transfer through intervalence charge transfer [40, 41]. Wang et al. [37] demonstrated that 0.10 mol% Fe/0.40 mol% Co codoped TiO2 nanocrystals revealed the highest photoactivity under visible light, which was attributed to the promotion of charge carrier separation and interfacial charge transfer. Lin et al. [38] demonstrated that undoped TiO2 thin films revealed the best photocatalytic performance relative to Fe/Mn codoped thin films, which was attributed to the increased density of recombination centres and the enhancement of lattice distortion. Chen et al. [39] observed that 0.05 mol% Co/0.05 mol% V codoped TiO2 thin film exhibited the best photocatalytic performance, which was attributed to modification of the band gap and associated semiconducting effects.

The present work reports the preparation of Fe/Mo codoped TiO2 thin films by spin coating on polished fused silica glass substrates, followed by annealing at 450 °C for 2 h. The thin films were characterised by glancing angle X-ray diffraction (GAXRD), laser Raman microspectoscopy (Raman), atomic force microscopy (AFM), UV–Vis spectrophotometry (UV–Vis), and X-ray photoelectron spectrometry (XPS). The photocatalytic efficiency was determined comparatively in terms of methylene blue (MB) degradation under UV light for 24 h.

2 Experimental procedure

2.1 Sample fabrication

The fabrication process for TiO2 thin films using spin coating has been described in detail elsewhere [42,43,44]. Polished fused silica substrates were selected because they risk contamination by only a single cation dopant and Si doping of TiO2 has been reported to have a neutral or positive effect on the photocatalytic performance [45]. However, the most important study is that of Kabir et al. [46], who determined that Si contamination of TiO2 thin films deposited and annealed on fused silica substrates derived solely from grain boundary diffusion. Consequently, the only effect was blockage of active sites rather than alteration of the defect state.

The fabrication procedures are detailed as follows:

  • Precursor solutions: Titanium tetraisopropoxide (TTIP, Reagent Grade, 97 wt%, Sigma-Aldrich) was dissolved in isopropanol (Reagent Plus, ≥ 99 wt%, Sigma-Aldrich) at 0.1 M titanium concentration.

  • Doping and codoping: The Fe3+ or Mo5+ dopant level was varied in the range 0.00–0.10 mol% (metal basis) by adding FeCl3 (Reagent Grade, 99 wt%, Sigma-Aldrich) and/or MoCl5 (Reagent Grade, 95 wt%, Sigma-Aldrich) to the solution.

  • Mixing: The precursor solution was mixed by manual stirring for 10 min in a Pyrex beaker without heating.

  • Repeat coating process: Spin coating (Laurell Technologies WS-65052) was done by depositing ~ 0.2 mL of precursor solution onto a polished fused silica substrate spinning at 2000 rpm in nitrogen over a period of ~ 10 s. The films were dried by spinning for an additional 15 s and the overall process was repeated six more times (~ 1.4 mL), ultimately yielding films of thickness 300 ± 10 nm.

  • Annealing: Annealing in air was done in a muffle furnace at 450 °C for 2 h; the heating rates were 0.5 °C/min from room temperature to 200 °C and 1 °C/min from 200 to 450 °C, followed by natural cooling.

2.2 Characterisation

The resultant films were characterised using the following techniques:

  • Glancing angle X-ray diffraction (GAXRD; 45 kV, 40 mA, PANalytical Empyrean Thin-Film XRD).

  • Laser Raman microspectroscopy (Raman; green argon ion laser (514 nm, 25 mW, 50X, spot size 1.5 mm, Renishaw inVia Raman Microscope).

  • Atomic force microscopy (AFM; tapping mode, scan size 1 µm × 1 µm, Bruker Dimension Icon Scanning Probe Microscope).

  • Ultraviolet–visible spectrophotometry (UV–Vis; dual-beam, 300–800 nm, PerkinElmer Lambda 35 UV–Visible Spectrophotometer).

  • X-ray photoelectron spectroscopy (XPS; 20 °C, 10−7 Pa vacuum, 13 kV, 12 mA, spot size 500 mm, 2–5 nm beam penetration, Thermo Scientific ESCALAB 250Xi X-ray Photoelectron Spectrometer Microprobe).

2.3 Photocatalytic performance

The photocatalytic performances of the TiO2 thin films were assessed in terms of photo-bleaching of methylene blue (MB) dye solutions, which has been described in detail elsewhere [42,43,44]. This testing was done by immersing each TiO2 thin film in MB solution and then irradiating under UV light for 24 h. The MB solutions were prepared using methylene blue (M9140, dye content ≥ 82 wt%, Sigma-Aldrich) dissolved in deionized water at 10−5 M concentration. The solutions were magnetically stirred in a Pyrex beaker for 1 h without heating. The samples were placed in MB solution in a dark container for saturation for ~ 12 h prior to testing. The samples then were placed in separate small beakers filled with ~ 10 mL of MB solutions and exposed to UV radiation (3UV-38, 8 W, UVP) for 24 h. The vertical lamp-liquid and liquid-sample distances were ~ 6 cm and ~ 4 cm, respectively. After irradiation, the tested MB solutions were analysed by UV–Vis spectrophotometry in order to determine the extent of degradation.

3 Results and discussion

Figure 1 show the GAXRD patterns and Raman spectra, respectively, of the annealed undoped and doped thin films with varying doping concentrations. All of the thin films consist of anatase as the only crystalline phase. These data suggest that Fe or Mo doping causes only a slight reduction in the crystallinity (i.e., lattice stability) and that codoping has little or no effect.

Fig. 1
figure 1

GAXRD patterns (a Fe-doped, b Mo-doped, c Fe/Mo codoped) and Raman laser microspectra (d Fe-doped, e Mo-doped, f Fe/Mo codoped) of thin films annealed with varying doping levels at 450 °C for 2 h

Full size image

Images of the topographies of the TiO2 thin films are shown in Figs. 2, 3, and 4 and the associated data for the grain sizes and surface roughnesses are given in Table 2. These data show that there are few differences between the samples, so the data are not sufficiently distinctive to allow more than the conclusion that doping appears to increase the grain size marginally. An increase in grain growth upon doping is not unexpected as it is well known that defects enhance this.

Fig. 2
figure 2

AFM images of Fe-doped TiO2 thin films annealed with varying doping levels at 450 °C for 2 h

Full size image
Fig. 3
figure 3

AFM images of Mo-doped TiO2 thin films annealed with varying doping levels at 450 °C for 2 h

Full size image
Fig. 4
figure 4

AFM images of Fe/Mo codoped TiO2 thin films annealed with varying doping levels at 450 °C for 2 h

Full size image
Table 2 Variation in grain sizes and surface roughnesses of annealed doped TiO2 thin films based on AFM data
Full size table

Figure 5 shows the UV–Vis transmission spectra of the annealed doped TiO2 thin films at varying doping concentrations. The films are flat and highly transparent and the absorption edges show no particular trends as a function of doping but there are significant blue shifts at the lower wavelengths, which are not advantageous.

Fig. 5
figure 5

UV–Vis transmission spectra of thin films annealed with varying doping levels at 450 °C for 2 h

Full size image

Table 3 gives the optical indirect band gaps (Eg) of the doped TiO2 thin films, which were calculated by the Tauc method [47]. These data suggest that doping causes a slight increase in the band gap, which is consistent with the blue shift of the absorption edges.

Table 3 Optical indirect band gaps of annealed doped TiO2 thin films
Full size table

Surface chemical analysis by XPS has emerged as a standard tool in materials characterisation because it quantitatively identifies valence states [11, 35, 36]. Consequently, the peak shifts indicating valence increases (greater binding energies) and decreases (lower binding energies) provide information on charge transfer effects. In particular, electron transfer and complementary valence shifts between codopants, between dopant and matrix, and between codopants and matrix can be used to support the conclusion of IVCT [11, 39] or multivalence charge transfer (MVCT) [41]. In addition, these data can indicate the presence of direct redox effects, such as 2Ti4+ → 2Ti3+ + \({\text{V}}_{\text{O}}^{ \cdot \cdot }\), where the latter is a charge compensating oxygen vacancy.

Figure 6 shows the XPS spectra for the annealed doped TiO2 thin films at the different doping concentrations. In Fig. 6a, the two main Ti2p peaks for TiO2 at 458.30 eV (Ti2p3/2) and 464.05 eV (Ti2p1/2) confirm the presence of Ti4+ [48, 49]. It can be seen that Fe-doping caused the binding energies to decrease (indicating Ti4+ valence decrease), Mo-doping caused the binding energy to increase (indicating Ti4+ valence increase), and the codoping did not change the Ti valence.

Fig. 6
figure 6

a Ti2p, b, c Fe2p, and d, e Mo3d XPS spectra for 1.00 mol% doped thin films annealed with varying doping levels at 450 °C for 2 h

Full size image

In Fig. 6b, c, the two main Fe2p peaks at ~ 710.0 eV (Fe2p3/2) and ~ 723.5 eV (Fe2p1/2) suggest the presence of Fe3+/2+ [50] for both Fe-doping and codoping. Although the Fe3+ and Fe2+ peaks cannot be differentiated due to overlap, there appears to be shifts to lower binding energies relative to the Fe3+ precursor, suggesting decreases in valence for both doping types. In contrast, Fe3+ is thermodynamically more likely to be present in the films after annealing at 450 °C in air [38], so a shift to higher binding energy would be expected.

In Fig. 6d, e, the two main Mo3d peaks represent Mo2O5 at 232.18 eV (Mo3d5/2) and 235.28 eV (Mo3d3/2) [32, 51], and Mo2O4 at 230.56 eV and 233.78 eV [36] for both Mo-doping and codoping. It is clear that some of the Mo5+ has reduced to Mo4+ but the presence of the most thermodynamically stable Mo6+ valence is not certain as the shoulder on the Mo5+ is not unambiguous. As Mo5+ is not thermodynamically stable after annealing at 450 °C in air [36], its conversion to Mo4+ and possibly Mo6+ would be expected.

The preceding suggests that the processes of dissolution and partial equilibration may be influenced by both thermodynamics and IVCT. Figure 7 shows the Gibbs standard free energies, which are available, for the stepwise oxidation reactions for the dopants and matrix as a function of temperature. These data indicate that Fe2+ → Fe3+ oxidation is favoured over Ti3+ → Ti4+ oxidation, so the thermodynamics of the Fe–O system will dominate those of the Ti–O system. In contrast, for Fe-doping by precursor Fe3+, the XPS data indicate valence decreases (i.e., reduction) for both Ti and Fe. Since the Gibbs free energies of reaction at 450 °C for Fe2O3 → 2FeO + 1.5O2 (+184.8 kJ/mol) and TiO2 → 0.5Ti2O3 + 0.25O2 (+155.6 kJ/mol) are positive, neither is favoured thermodynamically. This contradiction between the XPS data and thermodynamics can be explained by IVCT, as indicated in Reaction 1 (where Ti3+ is the minority species). Following IVCT, which alters the balance of the valences in solution, the final valences are equilibrated during annealing, which effectively results in reduced valence states for the dominant species in both matrix (Ti4+) and dopant (Fe3+) relative to the precursors, thereby matching the XPS data. As these are contrary to the free energy calculations for oxidation at 450 °C, then IVCT drives the equilibria both in solution and during annealing.

Fig. 7
figure 7

Gibbs standard free energies for stepwise oxidation reactions of dopants and matrix as a function of temperature

Full size image
$${\text{Ti}}^{3 + } + {\text{ Fe}}^{3 + } \,\mathop \leftrightarrow \limits^{\text{Solution}} \,{\text{Ti}}^{4 + } + {\text{ Fe}}^{2 + } \,\mathop \leftrightarrow \limits^{{{\text{O}}_{2} }} \,{\text{Ti}}^{{(4{-}{\text{x}}) + }} + {\text{Fe}}^{(2 + {\text{x}}) + }$$
(1)

For Mo-doping by precursor Mo5+, assuming that Mo6+ is not formed, the XPS data indicate a valence increase (i.e., oxidation) for Ti and a valence decrease (i.e., reduction) for Mo. Since the thermodynamic stability diagram for Mo–O [36] shows that neither Mo5+ nor Mo3+ are stable and Fig. 1 does not provide useful data for this system, then the driving force for redox in solution could be either thermodynamics (as Ti3+ → Ti4+ oxidation is favoured) and/or IVCT (the mechanism alternative to thermodynamics). These are given in Reaction 2, which matches the XPS data. It is concluded that the equilibria during annealing are driven by IVCT because the majority Ti4+ → Ti5+ oxidation is not thermodynamically favoured as is the case for Mo5+ → Mo4+ reduction.

$${\text{Ti}}^{3 + } + {\text{Mo}}^{5 + } \,\mathop \leftrightarrow \limits^{\text{Solution}} \,{\text{Ti}}^{4 + } + {\text{Mo}}^{4 + } \,\mathop \leftrightarrow \limits^{{ {\text{O}}_{2} }} \,{\text{Ti}}^{(4 + {\text{x}}) + } + {\text{Mo}}^{{(4{-}{\text{x}}) + }}$$
(2)

However, although the XPS data are ambiguous, it is possible that Mo6+ is formed, as suggested by a thermodynamic modelling study that indicated MoO3 stability at temperatures up to ~ 850 °C. Accordingly, Reaction 3 also matches the XPS data, where the Mo5+ peaks shift to lower binding energies. This supports the conclusion that Mo6+ is not formed because, while Reaction 3 gives the only feasible valence conversions, the majority Ti4+ → Ti5+ oxidation is not thermodynamically favoured and IVCT does not generate Mo6+ as a final reaction product.

$${\text{Ti}}^{4 + } + {\text{Mo}}^{5 + } \,\mathop \leftrightarrow \limits^{\text{Solution}} \,{\text{Ti}}^{3 + } + {\text{Mo}}^{6 + } \,\mathop \leftrightarrow \limits^{{{\text{O}}_{2} }} \,{\text{Ti}}^{(4 + {\text{x}}) + } + {\text{Mo}}^{{(5{-}{\text{x}}) + }}$$
(3)

For codoping, the XPS data indicate effectively no valence change for Ti but valence decreases (i.e., reduction) for both Fe and Mo, again unfavourable thermodynamically. This situation cannot be assessed unambiguously because any shifts in the deconvoluted XPS peaks for Mo4+ cannot be determined and there are no thermodynamic data for Mo2O5 or Mo2O3, which are relevant to the reactions Mo2O5 → 2MoO2 + 0.5O2 or MoO2 → 0.5Mo2O3 + 0.25O2. Since Mo5+ and Mo3+ do not appear to be thermodynamically stable [36], the first reaction would be favoured but the second wouldn’t. Nonetheless, Reaction 4 harmonises the XPS data by illustrating that, relative to the precursors, IVCT between matrix and dopant drives the reaction in solution (since Fe reduction at room temperature is not favoured thermodynamically) but IVCT between dopants drives the reaction during annealing (since neither dopant valence is favoured thermodynamically):

$$2{\text{Ti}}^{3 + } + {\text{Fe}}^{3 + } + {\text{Mo}}^{5 + } \,\mathop \leftrightarrow \limits^{\text{Solution}} \,2{\text{Ti}}^{4 + } + {\text{Fe}}^{2 + } + {\text{Mo}}^{4 + } \,\mathop \leftrightarrow \limits^{{{\text{O}}_{2} }} \,2{\text{Ti}}^{4 + } + {\text{Fe}}^{(2 + {\text{x}}) + } + {\text{Mo}}^{{(4{-}{\text{x}}) + }}$$
(4)

An alternative approach is consideration of the effect of electronegativity [52]. Table 4 summarises the different scales of electronegativities of Ti matrix and Fe and Mo dopants. These data tend to show that the electronegativities of dopants are the greater value than that of the Ti matrix, indicating that the electron cloud of the O in the Mo–O–Ti and Fe–O–Ti bonding configuration would move toward that of dopants, causing Fe and Mo valences to increase and that of Ti to decrease. However, it is unlikely that electronegativity plays a significant role since the Ti valence decreases only with Fe doping of TiO2.

Table 4 Summary of electronegativities of Ti, Fe, and Mo in different scales
Full size table

For size considerations, the crystal radii of Ti, Fe, and Mo in sixfold (substitutional) and fivefold (interstitial) coordinations [60] are given in Table 5. According to Hume-Rothery’s rules for substitutional solid solubility [36], it is very likely that this type of solid solubility occurred owing to the similar crystal radii of Ti4+, Fe4+, Fe3+ (high spin), Fe2+ (low spin) for Fe-doping, and Mo6+, Mo5+, Mo4+ and Mo3+ for Mo-doping. However, it also is possible that interstitial solid solubility occurred since each of the two interstices adjacent to the central Ti in the elongated TiO6 octahedron of anatase has a radius of 0.0782 nm [11]; Fe4+, Fe3+ (low spin), and Mo6+ can fit in this position. According to the crystal field stabilisation energies, Fe3+ and Fe2+ are likely to be high-spin complexes [36]. For the latter case, Fe2+ is unlikely to be soluble, which suggests that the Fe(2+x)+ of Reactions 1 and 3 is more appropriately represented at Fe(3−x)+ in terms of the size of x, which would be small.

Table 5 Relevant Shannon crystal radii in sixfold and fivefold coordinations [60]
Full size table

The potential dopant effects on defect equilibria, using Kröger–Vink notation [61], that can arise from doping with Fe or Mo are summarised in Table 6. The defects in the form of oxygen vacancies, metal substitutionals, metal interstitials, and metal vacancies would provide midgap states, which could have the capacity to improve the photocatalytic performance. With electron charge compensation, the electron and hole charge carriers are subject to traps in the form of the defects, which could have the capacity to increase the charge separation and carrier lifetime.

Table 6 Potential dopant effects on defect equilibria
Full size table

Figure 8 shows the photocatalytic performances of the undoped and doped TiO2 thin films evaluated by photobleaching of MB solution for 24 h. The performances of the Fe-doped TiO2 thin films generally were slightly inferior to that of the undoped sample while the performance of the Mo-doped TiO2 thin films generally were superior to that of the undoped sample. For the codoped TiO2 thin films, the performances were more consistent in revealing the degradation in performance in proportion to the codoping concentration.

Fig. 8
figure 8

Degradation of MB solutions photocatalysed for 24 h by thin films annealed with varying doping levels at 450 °C for 2 h

Full size image

Considering the three sets of data globally, the performance data do not correlate with the degrees of crystallinity (Fig. 1), absorption edge, indirect band gap (Fig. 5 and Table 3), or microstructure (Figs. 2, 3, 4). In more conventional thought, which would consider oxygen vacancy formation, only Fe doping would create oxygen vacancies in TiO2−x, which would be expected to enhance the photocatalytic performance. However, Fig. 4 shows that this was not the case. Considering the formation of midgap states, both Fe and Mo doping have the capacity to reduce the band gap, but Table 3 shows that, again, this was not the case. Since all of these typically key factors do not appear to be dominant, then this suggests the importance of the valences as revealed by the XPS data:

  • Fe doping: The presence of Ti(4−x)+ and Fe(2+x)+ reflects Ti4+ → Ti3+ and Fe3+ → Fe2+ reductions (IVCT leaves the oxygen vacancy concentration unaffected), where Ti(4−x)+ and Fe(2+x)+ (effectively Fe(3−y)+, where y = 1 − x) would act as an electron traps, resulting in diminished performance since TiO2 is an n-type semiconductor and the principal charge carrier is electrons [62].

  • Mo doping: The presence of Ti(4+x)+, Mo(5−x)+, and Mo(4−x)+ reflects Ti4+ → Ti5+ oxidation and Mo5+ → Mo4+ reduction, where the majority Ti(4+x)+ would act as a hole trap and minority Mo(5−x)+ and Mo(4−x)+ would act as electron traps. These valences would serve as effective hole traps while decreasing the majority electron charge carriers to a much lower extent. This would result in enhanced performance owing to increased charge separation.

  • Codoping: The Ti valence is unchanged at Ti4+ but the dopants are converted from Fe3+ to Fe(2+x)+ (effectively Fe(3−y)+) and from Mo5+ to Mo(5−x)+ and Mo(4−x)+. Hence, Ti4+ would play no role in charge carrier trapping but all of the dopants would act as electron traps, so their combination would increase both the deleterious effect of codoping and the ability to detect it.

4 Conclusions

Fe/Mo single doped and codoped TiO2 thin films were spin coated on polished fused silica substrates and annealed in air at 450 °C for 2 h. All of the thin films consisted of anatase; doping appeared to have only a slight negative effect on the crystallinity. There were few differences in grain sizes, with only a slight increase for the doped thin films. The thin films were flat and highly transparent, where doping caused a slight blue shift and associated increase in Eg. The XPS data were distinctive in showing that Fe-doping caused Ti4+ reduction, Mo-doping caused Ti4+ oxidation, and codoping did not alter the Ti valence. The XPS data also showed that the precursor valences of Fe3+ and Mo5+ were reduced upon annealing. Of the thermodynamically stable valences Mo4+ and Mo6+, the former was a final reaction product but it is unlikely that the latter was formed.

Analysis of the potential roles of thermodynamics and IVCT shows conclusively that the latter drives the equilibria both in solution and during annealing for all three types of doping. Photocatalytic performance testing indicated that Fe-doping was slightly deleterious, Mo doping had a consistent positive effect, and codoping exhibited a clear negative trend as a function of doping concentration. The data do not support the conclusion that crystallinity, microstructure, band gap, or midgap states played a dominant role in the photocatalytic performance. These results are interpreted in terms of the effect of IVCT on the matrix valence, dopant valences, and charge carrier trapping. Fe doping exhibited reduced performance because both matrix Ti(4−x)+ and dopant Fe(3−y)+ acted as electron traps. Mo doping exhibited enhanced performance because the matrix Ti(4+x)+ acted as a hole trap and the dopant Mo(5−x)+ and Mo(4−x)+ acted as electron traps, thereby promoting charge separation. The role of IVCT was shown most distinctly through the negative trend that codoping had on the photocatalytic performance. Here, Ti4+ played no role but Fe(3−y)+, Mo(5−x)+, and Mo(4−x)+ all acted as traps for the majority charge carrier electrons. This explains both why the trend occurred and why it was clearly detectable.