Introduction

Photoelectrochemical solar cells are the most ambitious targets in the utilization of solar energy field (Grätzel 2001, 2003, 2004; O’Regan and Grätzel 1991; Chen et. al. 2006). Among the new materials being developed for solar cells, photocatalytic and other applications, titanium dioxide (TiO2) remains one of the most promising materials because of its high efficiency, low cost, chemical inertness, eco-friendly nature and photostability (Fujishima and Honda 1972; Neumann et al. 2005; Kisch et al. 1998). However, the widespread use of TiO2 is hindered by its low utilization of solar energy in the visible region (about 3–5%) because of the wide band gap (3.0 eV for rutile and 3.2 eV for anatase crystalline phase). Therefore, many efforts were concentrated on narrowing the optical band gap of titania. In dye-sensitized solar cells, this is being implemented successfully by anchoring low band gap ruthenium dye on to the surface of TiO2. But the ruthenium dyes are expensive and unstable in aqueous solutions (Peter et al. 2003).

Narrowing of the band gap has been reported in polycrystalline powders and films of TiO2 doped with anion impurities such as nitrogen (Asahi et al. 2001; Morikawa et al. 2001), carbon (De Vos et al. 2002; Khan et al. 2002; Sakthivel and Kisch 2003; Irie et al. 2003; Bard and co-workers 2006; Milani and co-workers 2005; Kisch and co-workers 2001) and sulfur (Ohno et al. 2003, 2004). In particular, carbon-doped TiO2 was indicated as one of the best in terms of band gap narrowing. There are mainly several methods for preparing the carbon-doped TiO2, such as direct oxidation of the Ti metal in the flame of a burner, oxidizing commercial TiC powders, the wet process synthesis, etc. Khan et al. (2002) reported incorporation of carbon during n-TiO2 (rutile) synthesis by flame oxidation of Ti metal sheet lowered the band gap of n-TiO2 to 2.32 eV (535 nm) and thereby exhibited water-splitting efficiency under xenon lamp illumination, but there is still disagreement about its solar-to-hydrogen conversion efficiency of 8.35 % (Fujishima 2003; Grätzel and co-workers 2003; Lackner 2003). Sakthivel and Kisch (2003) reported the wet process synthesis of carbon-doped TiO2 (anatase) by hydrolysis of TiCl4 with tetrabutylammonium hydroxide, and the samples (band gap of 3.02 eV) containing 2.98% carbon possessed the maximum band-gap narrowing of 0.14 eV. Irie et al. (2003) reported that the carbon-doped anatase TiO2 powders (the percentage of carbon was 0.32 %) could be obtained by oxidizing commercial TiC powders. Recently, Bard and co-workers (2006) reported that novel carbon-doped TiO2 nanotube arrays (band gap of 2.22 eV, the percentage of carbon was 8–42%) with high aspect ratios for efficient solar water splitting were prepared by controlling the anodizing conditions and heating with CO at 600 °C. Milani and co-workers (2005) obtained carbon-doped nanostructured TiO2 films (rutile, the percentage of carbon was 17%) by cluster beam deposition with a band gap of 2.06 eV, but this method required high temperature.

The wet process synthesis of carbon-doped TiO2 has received significant attention owing to its low cost, ease to scale up and eco-friendly nature (Kisch and co-workers 2004; Cong et al. 2006; Reddy et al. 2005; Khan and co-workers 2006a, b). Tetrabutylammonium hydroxide (C16H36NOH) (Kisch and co-workers 2004; Cong et al. 2006; Reddy et al. 2005) and glucose (C6H12O6) (Khan and co-workers 2006a, b) were used as the carbon sources in the wet process synthesis. In either case, the percentage of carbon in carbon-doped TiO2 was no more than 11.2% (Khan and co-workers 2006a, b). Although these carbon-doped TiO2 samples synthesized in the presence of either of the two carbon sources show the shift of optical response from UV to near infrared region, the tetrabutylammonium hydroxide (C16H36NOH) and glucose (C6H12O6) have large size, which are difficult to diffuse. Compared to glucose (C6H12O6) and tetrabutylammonium ions (C16H36N+), oxalic acid (C2H2O4) will be easier diffusion due to its smaller size. However, there is little report on preparing carbon-doped TiO2 film in the presence of oxalic acid (C2H2O4) and using as photoanode for photoelectrochemical solar cells.

Recently, our group found that nanostructured TiO2 (ns-TiO2) films prepared using titanium ethanolate (Ti(OEt)4) ethanol solution (containing carbon) as precursors by sol–gel process showed excellent photovoltaic activity to absorb visible light. The precursors of Ti(OEt)4 ethanol solution were electrosynthesized firstly in the presence of oxalic acid and tetrabutylammonium bromide (Bu4N·Br) in our laboratory (Zhou et al., 2002). The gel of ns-TiO2 films coated on the fluorine-doped SnO2 (FTO) conducting glass was calcined at 450 °C for 30 min in air. Subsequently, the successful incorporation of the carbon into the ns-TiO2 host was confirmed by X-ray photoelectron spectroscopy (XPS), X-ray diffraction (XRD), scanning electron microscope (SEM), and UV-Visible spectra. The synthesized carbon-doped nanostructured TiO2 (CD ns-TiO2) films (anatase) absorbs well into the visible light to near infrared region up to 800 nm and have narrower band gap energy of 1.98 (628 nm), 1.64 (758 nm) and 1.26 eV (958 nm). The doping concentration of carbon in the CD ns-TiO2 was 41% (as determined from the XPS peaks). We investigated the photoelectrochemical behavior of the CD ns-TiO2 film photoanode for solar cells, it was found that the CD ns-TiO2 film photoanode exhibited photocurrent densities (l.34 mA/cm2) and an overall conversion efficiency (η) of 4.42 % under visible light illumination at 11 mW/cm2.

Experimental

All reagents were analytical pure, 1,2-propanediol carbonate, Bu4N·Br, oxalic acid and Iodine were purchased from ShangHai Chemical Company, and lithium iodide was obtained from Alfa. The fluorine-doped SnO2 (FTO) conducting glass was placed in deionized water and sonicated for 30 min, then washed thoroughly with deionized water and ethanol, respectively, then dried at room temperature. The precursors of Ti(OEt)4 and TiO2 sols were prepared as described in our previously works (Chu et al. 2005), the TiO2 sols were coated on treated conducting glass, after calcining at 450 °C in air for 30 min and cooling to room temperature, a transparent CD ns-TiO2 film electrode was obtained. The undopant ns-TiO2 film was prepared in the same way.

XRD spectra were collected on a Philips X’ Pert-pro X-ray diffractometer (Cu Kα). UV-visible absorption was collected with a U-4100 spectrophotometer (Hitachi, Kyoto Japan), UV-visible absorption of films on FTO were measured against the FTO substrate as a blank. SEM was collected using a FE-SEM (JSM-6700F, JEOL). XPS spectra were obtained by a VG ESCALAB MK II.

The current-voltage (I–V) characteristics were measured using LK 2005 microcomputer electroanalysis system. Appropriate cut-off filters were used. A 150 W xenon lamp was used as the light source, the incident light intensity calibrated by using a photometer. The photocurrent density versus measured potential for FTO/CD ns-TiO2 photoanode characteristics were carried out using electrochemical workstation with a three-electrode system. The CD ns-TiO2 film electrode, platinum foil, and an aqueous SCE served as the working, counter, and reference electrodes, respectively. The electrolyte solution consisted of 0.3 M LiI and 0.03 M I2 in 1,2-propanediol carbonate. The current-voltage characteristics of the cell were measured two-electrode system under 11 mW/cm2. The CD ns-TiO2 film electrode served as photoanode, the cathode was obtained by electroplated FTO conducting glass in the solution of H2PtCl6. All experiments were carried out at room temperature.

Results and discussion

Figure 1 gives XRD pattern of CD ns-TiO2 powders. XRD results demonstrated that the CD ns-TiO2 nanoparticles were of homogeneous anatase structure with low crystallinity, and the average grain sizes calculated from Scherrer equation (crystallite size or grain size, d = 0.9λ/β 1/2 cosθ, where λ is the characteristic X-ray wavelength applied (0.154056 nm), β 1/2 is the half width of the peak at the 2θ value) were found to be about 12 nm.

Fig. 1
figure 1

XRD pattern of CD ns-TiO2 powders

Full size image

Figure 2 shows SEM image of CD ns-TiO2 film. The surface morphology of films was a multiple-porous network structure. The average size of nanoparticle was 40 nm. The pore diameter was about 80 nm.

Fig. 2
figure 2

SEM image of the surface morphology of CD ns-TiO2 film

Full size image

For further analysis of the chemical structure of the CD ns-TiO2 films, we have examined three areas of the XPS spectra (Fig. 3): the C 1s region near 280 eV, Ti 2p the region near 460 eV and the O 1s region.

Fig. 3
figure 3

XPS spectra for the (a) C 1s and (b) Ti 2p and (c) O 1s of the CD ns-TiO2 films

Full size image

XPS measurements of the C 1s binding energy indicated the complete absence of carbon in the sample, peaks at binding energies at 284.6, 286.3 and 288.5 eV were found. The two features of 288.5, 286.3 eV at higher binding energies were attributed to the C–O binding, the carbon peak at 284.6 eV was assigned to the C remained in the precursor organic compound. No appreciable signal related to carbide (281.5 eV) was observed, showing that the amount of Ti-bonded carbon was very low.

When scanning the Ti 2p XPS regions, three peak structures at binding energies at 464.4, 458.6, and 457.3 eV. The first value arises from the 2p1/2 peaks. Compared to the classic paper of Saha and Tompkins (1992), we suggest that the Ti 2p3/2 XPS peak at 458.6 eV is consistent with the formation of a crystalline TiO2 sample. The carbon incorporation shifts the XPS spectrum to a lower binding energy (a peak is observed at 457.3 eV) because of surface strain and lattice distortion induced by the incorporation of the carbon.

The O 1s core levels of CD ns-TiO2 films revealed four peak structures at binding energies at 533.2, 531.7, 529.8, and 528.4 eV in the CD ns-TiO2 films. The peaks appeared after the air calcinations at the binding energy of 533.2 and 531.7 eV were assigned to O–H bonds, the peak at 529.8 eV ascribes to Ti–O bond, the peak at 528.4 eV further testified that incorporation shifted the XPS spectra to a lower binding energy, although it can not be seen in the XRD pattern.

We found that the doping concentration of carbon in the CD ns-TiO2 lattice was estimated to be 41% as determined from the XPS peaks. This suggested that further increase of the doped amount content in the CD ns-TiO2 films were realized by using oxalic acid and tetrabutylammonium bromide as carbon sources by sol–gel process. This may be attributed to easier diffusion of oxalic acid (C2H2O4) due to its smaller size compared to glucose (C6H12O6) and tetrabutylammonium ions (C16H36N+). And the photoresponse in visible region can be enhanced by increasing the doped carbon content (Bard and co-workers 2006; Milani and co-workers 2005; Khan and co-workers 2006a, b).

The UV-visible absorption measurements are used to evaluate the nanostructured TiO2 (ns-TiO2) and CD ns-TiO2 films. Figure 4 shows the optical absorbance spectra of ns-TiO2 and CD ns-TiO2 films. Compared with ns-TiO2, it reveals that the new powerful absorption at 400–800 nm is related to the high carbon content. The CD ns-TiO2 films show two optical absorption thresholds at 628 and 758 nm in the visible range, which corresponds to a band gap energy of 1.98 (628 nm) and 1.64 eV (758 nm). Particularly, there is a wide absorption peak in the CD ns-TiO2 films at 605 nm and extends to near infrared region. It is clear that an optical absorption threshold lies in at about 980 nm in the infrared region, which corresponds to a band gap energy of 1.26 eV. Nie and Sohlberg (2004) showed that carbon-doping gives rise to two band gaps by theoretical calculation, which was confirmed by experiments (Khan et al. 2002; Khan and co-workers 2006a, b). The finding of three band gaps can be attributed to the higher carbon content in the CD ns-TiO2 films. It is evident that the amazing photoresponse observed from the CD ns-TiO2 films was attributed to the band gap lowering and the formation of the intragap band (Kamisaka et al. 2005; Gole and Stout, 2004).

Fig. 4
figure 4

UV-Visible spectra of FTO/ ns-TiO2 and FTO/ CD ns-TiO2 films

Full size image

Figure 5 shows a comparison of the photocurrent density versus applied potential curves of the FTO/CD ns-TiO2 photoanode in the dark and under visible illumination. The observed dark current densities were found to be negligible. The increased photocurrent generation at anodic potentials shows that the CD ns-TiO2 film possesses n-type semiconducting behavior, which delivers a maximum photocurrent of l.34 mA/cm2 under anodic bias 0.3 V (versus SCE). The CD ns-TiO2 film structure also shows a steeper increase in the photocurrent with applied potential. With nanoporous network structure, e−h+ pairs induced by photon absorption split more readily, and also produced a high photoanodic response probably because of an increased light penetration depth and better scattering within a regular pore structure.

Fig. 5
figure 5

Variation of photocurrent density versus measured potential for FTO/CD ns-TiO2 photoanode in the dark and under visible illumination (cell active area: 0.56 cm2)

Full size image

Figure 6 presents a photocurrent density-voltage curve of CD ns-TiO2 solar cell. The photovoltaic characteristics are analyzed using equations 1 and 2,

Fig. 6
figure 6

Current density-voltage curve of CD ns-TiO2 solar cell under 11 mW/cm2 visible illumination (cell active area: 0.56 cm2)

Full size image
$$ FF = \frac{{W_{\max } }} {{V_{{\hbox{oc}}} \times I_{{\hbox{sc}}} }} = \frac{{V_{\max } \times I_{\max } }} {{V_{{\hbox{oc}}} \times I_{{\hbox{sc}}} }} $$
(1)
$$ \eta \left( \% \right) = \frac{{V_{{\hbox{oc}}} \times I_{{\hbox{sc}}} \times FF}} {I} \times 100 $$
(2)

Where the fill factor (FF) is the maximum power delivered to an external load, calculated from the short-circuit current density (I sc) and the open-circuit voltage (V oc). The overall conversion efficiency (η) is normalized by the intensity of the incident light (I).

The open circuit voltage (V oc), short-circuit photocurrent (I sc) and the fill factor (FF) of the cell were 0.50 V, 0.256 mA/cm2 and 0.47, respectively. The η obtained from the cell system was 4.42 %. The results indicated that incorporation of carbon into nanoporous space of TiO2 was very successful.

Conclusions

In conclusion, we have demonstrated that CD ns-TiO2 films with three optical band gap of 1.98, 1.64, and 1.26 eV and 41% carbon content can be obtained with oxalic acid and Bu4N·Br as the carbon sources by the sol–gel process. The evidence for the presence of carbon states within CD ns-TiO2 film is demonstrated by UV-Visible spectra and XPS. The CD ns-TiO2 films have revealed a superior harvesting light capability and improvement photovoltaic activity under visible light. Carbon doped into substitutional sites of TiO2 has also proven to be indispensable for band gap narrowing and photovoltaic effect. The CD ns-TiO2 film was firstly used as photoanode for solar cells, exhibiting photocurrent densities (l.34 mA/cm2) and yielding an overall conversion efficiency (η) of 4.42 %.

This investigation shows that our method of narrowing the band gap has low cost, simple processing, excellent reproducibility, easy large-scale production and extensive application in the field of solar energy conversion and photocatalysis. However the influence of calcination temperature for nanostructure and doping concentration of carbon are not disscussed, we will study these aspects for next steps.