Visible-Light Photocatalysis in Nitrogen-Doped Titanium Oxides

See allHide authors and affiliations

Science  13 Jul 2001:
Vol. 293, Issue 5528, pp. 269-271
DOI: 10.1126/science.1061051


To use solar irradiation or interior lighting efficiently, we sought a photocatalyst with high reactivity under visible light. Films and powders of TiO2-xNxhave revealed an improvement over titanium dioxide (TiO2) under visible light (wavelength < 500 nanometers) in optical absorption and photocatalytic activity such as photodegradations of methylene blue and gaseous acetaldehyde and hydrophilicity of the film surface. Nitrogen doped into substitutional sites of TiO2 has proven to be indispensable for band-gap narrowing and photocatalytic activity, as assessed by first-principles calculations and x-ray photoemission spectroscopy.

Since photoinduced decomposition of water on TiO2 electrodes was discovered (1), semiconductor-based photocatalysis has attracted extensive interest. One particular focus has been on applications in which organic molecules are photodegraded, such as water and air purifications (2–4). Most of the investigations have focused on TiO2(5–7), which shows relatively high reactivity and chemical stability under ultraviolet (UV) light [wavelength (λ) < 387 nm], whose energy exceeds the band gap of 3.2 eV in the anatase crystalline phase.

The development of photocatalysts that can yield high reactivity under visible light (λ > 380 nm) should allow the main part of the solar spectrum, and even poor illumination of interior lighting, to be used. One approach has been to dope transition metals into TiO2 (8–10), and another has been to form reduced TiOx photocatalysts (11,12). However, doped materials suffer from a thermal instability (9), an increase of carrier-recombination centers, or the requirement of an expensive ion-implantation facility (10). Reducing TiO2introduces localized oxygen vacancy states located at 0.75 to 1.18 eV below the conduction band minimum (CBM) of TiO2(12), so that the energy levels of the optically excited electrons will be lower than the redox potential of the hydrogen evolution (H2/H2O) located just below the CBM of TiO2 (13) and that the electron mobility in the bulk region will be small because of the localization.

We have considered whether visible-light activity could be introduced in TiO2 by doping, and we set the following requirements: (i) doping should produce states in the band gap of TiO2 that absorb visible light; (ii) the CBM, including subsequent impurity states, should be as high as that of TiO2 or higher than the H2/H2O level to ensure its photoreduction activity; and (iii) the states in the gap should overlap sufficiently with the band states of TiO2 to transfer photoexcited carriers to reactive sites at the catalyst surface within their lifetime. Conditions ii and iii require that we use anionic species for the doping rather than cationic metals, which often give quite localized d states deep in the band gap of TiO2 and result in recombination centers of carriers. We have calculated densities of states (DOSs) of the substitutional doping of C, N, F, P, or S for O in the anatase TiO2 crystal, by the full-potential linearized augmented plane wave (FLAPW) formalism (14, 15) in the framework of the local density approximation (LDA) (16) (Fig. 1). The substitutional doping of N was the most effective because its p states contribute to the band-gap narrowing by mixing with O 2p states. Although doping with S shows a similar band-gap narrowing, it would be difficult to incorporate it into the TiO2 crystal because of its large ionic radius, as evidenced by a much larger formation energy required for the substitution of S than that required for the substitution of N (17). The states introduced by C and P are too deep in the gap to satisfy condition iii. The calculated imaginary parts of the dielectric functions of TiO2-xNx indeed show a shift of the absorption edge to a lower energy by the N doping (Fig. 2A) (18). Dominant transitions at the absorption edge have been identified with those from N 2p π to Ti dxy , instead of from O 2p π as in TiO2 (19). Jansen and Letschert demonstrated tunable colorings in modified perovskite oxinitrides, which suggested a similar band-gap control by N, although they did not address the detailed electronic states in these pigments (20).

Figure 1

(A) Total DOSs of doped TiO2 and (B) the projected DOSs into the doped anion sites, calculated by FLAPW. The dopants F, N, C, S, and P were located at a substitutional site for an O atom in the anatase TiO2 crystal (the eight TiO2 units per cell). The results for N doping at an interstitial site (Ni-doped) and that at both substitutional and interstitial sites (Ni+s-doped) are also shown. The energy is measured from the top of the valence bands of TiO2, and the DOSs for doped TiO2 are shifted so that the peaks of the O 2s states (at the farthest site from the dopant) are aligned with each other. Arb. unit, arbitrary units.

Figure 2

Optical properties of TiO2-xNx (thick lines) compared with TiO2 (thin lines). (A) Calculated imaginary parts of the dielectric functions (ɛ2), which are averaged over three (x, y, andz) polarization vectors. (B) Experimental optical absorption spectra of TiO2-xNx and TiO2 films.

Photocatalysts for the present purpose, however, require more elaborate controls for doping besides the optical absorption; i.e., conditions ii and iii. To this end, we have made a theoretical comparison among three systems: substitutional N doping, interstitial N doping, and both types of doping in the anatase TiO2. In optimizing the positions of N in the eight TiO2 units per cell, we observed molecularly bonding states—NO and N2—for the last two cases; the obtained bond lengths of N–O and N–N were 1.20 and 1.16 Å, respectively, which are compared with those of the NO molecule (1.15 Å) and the N2 molecule (1.10 Å). Such molecularly existing dopants give rise to the bonding states below the O 2p valence bands and antibonding states deep in the band gap (Fig. 1B, Ni and Ni+s). However, these states are well screened and hardly interact with the band states of TiO2, and thus are unlikely to be effective for photocatalysis because of condition iii. The importance of substitutional site N doping is emphasized in the experimental results discussed below.

We prepared TiO2-xNx films by sputtering the TiO2 target in an N2(40%)/Ar gas mixture. After being annealed at 550°C in N2 gas for 4 hours, the films were crystalline, with features assignable to a mixed structure of the anatase and rutile crystalline phases, as determined by x-ray diffraction (XRD). The TiO2-xNx films were yellowish and transparent. TiO2 films were prepared in a similar fashion by sputtering the TiO2 target in an O2(20%)/Ar gas mixture and subsequently annealing it at 550°C in O2 gas for 4 hours. XRD showed that the homogeneous anatase crystalline phase was formed. Optical absorption spectra (Fig. 2B) show that the TiO2-xNx films noticeably absorb the light at less than 500 nm, whereas the TiO2films do not, which is in good agreement with the theoretical results.

Photocatalytic activity was evaluated by measuring decomposition rates of methylene blue as a function of the cutoff wavelength of the optical high-path filters under fluorescent light (Fig. 3A). Substantial photocatalytic activity under 500 nm has been observed in TiO2-xNx films, and the cutoff wavelength for photocatalytic activity corresponds well with the optical absorption spectra (21, 22). Both TiO2-xNx and TiO2 films reveal similar activity under UV light [represented by the results under black light (BL) illumination].

Figure 3

Photocatalytic properties of TiO2-xNx samples (solid circles) compared with TiO2 samples (open squares). (A) Decomposition rates [measuring the change in absorption of the reference light (Δabs)] of methylene blue as a function of the cutoff wavelength of the optical high-path filters under fluorescent light with the integrated photon flux of 2.45 × 10−9 einstein (E) s−1 cm−2between 350 and 520 nm, compared with the results under BL illumination with the integrated photon flux of 3.51 × 10−9 E s-1 cm-2 in the UV range. (B) CO2 evolution as a function of irradiation time (light on at zero) during the photodegradation of acetaldehyde gas [with an initial concentration of 485 parts per million (ppm)] under UV irradiation (BL with a peak at 351 nm and the light power of 5.4 mW cm−2) and visible irradiation [fluorescent light cut by the optical high-path filter (SC42, Fuji Photo Film), with a peak intensity at 436 nm and a light power of 0.9 mW cm−2]. (C) Contact angles of water as a function of time under interior lighting (with light powers of 28.5 and 159.4 μW cm−2 in the UV and visible ranges, respectively). All these light powers were measured by the UV radiometer (UVR-2, TOPCON, Tokyo, Japan) with detectors (UD-36 for the UV range and UD-40 for the visible range).

We also evaluated the photodecomposition of gaseous acetaldehyde of TiO2-xNx powder samples that were prepared by treating anatase TiO2 powder (ST01, Ishihara Sangyo Kaisha, Osaka, Japan) in the NH3(67%)/Ar atmosphere at 600°C for 3 hours. The Brunaer-Emmett-Teller surface areas of the TiO2-xNx and TiO2 powders were measured as 67 and 270 m2/g, respectively. Figure 3B shows CO2 concentrations, evolved as a result of the photodecomposition of acetaldehyde, as a function of irradiation time. The photocatalytic activity of the TiO2-xNx sample is superior to that of the TiO2 sample in the visible range of irradiation, whereas both samples yield similar UV activity. We used the photodegradation process of acetaldehyde proposed in (23) to estimate quantum yields for the CO2evolution, based on the number of incident photons being 0.42% (TiO2-xNx) and 0.14% (TiO2) at 436 nm, and 3.0% (TiO2-xNx) and 2.6% (TiO2) at 351 nm.

Our final evaluation of the photocatalytic activity was to measure contact angles of water on the sample films under interior lighting. The so-called photoinduced hydrophilic surface is known as an important application of TiO2 (24,25). In this experiment, we used the TiO2-xNx and TiO2 films, on which SiO2 with a nominal thickness of 5 nm was deposited to hold adsorbed water (25). The results show an excellent hydrophilic surface of the SiO2/TiO2-xNxfilm, which maintained a contact angle of 6° even after 30 days, in contrast to the SiO2/TiO2 film, whose contact angle constantly increased with time (Fig. 3C). As with usual TiO2 specimens, our TiO2-xNx samples resisted the attack of acid and alkaline solvents such as H2SO4, HCl, H2O2, and NaOH at the ambient temperature, and their photocatalytic performance was stable during successive use under 100-W mercury lamp irradiation for more than 3 months.

To investigate N states in TiO2-xNx, we measured N 1s core levels with x-ray photoemission spectroscopy (XPS). Three peak structures at the binding energies of 402, 400, and 396 eV were observed for the TiO2-xNx films (Fig. 4A). We observed similar XPS spectra for the TiO2-xNx powder samples. The TiO2 films also included a small amount of N; however, the peak at 396 eV was not observed. Saha et al. (26) investigated the N 1s XPS spectra during the oxidation process of TiN and assigned the peaks as atomic β-N (396 eV) and molecularly chemisorbed γ-N2 (400 and 402 eV). To elucidate relations to photocatalytic activity systematically, we prepared TiO2-xNx samples by annealing the TiO2 powder in the NH3(67%)/Ar atmosphere at 550° to 600°C. The powder samples were used here, because the reaction temperature and the N concentration in the samples can be easily controlled by changing temperature and time of the NH3 treatment rather than by changing the sputtering conditions.

Figure 4

(A) N 1s XPS spectra of the (upper lines) TiO2-xNx and (lower lines) TiO2 films. (B) Decomposition rates (measuring the change in absorption of the reference light after 10 hours) of methylene blue in the aqueous solution under visible light (the same light source as the visible irradiation in Fig. 3B) as a function of the ratio of the decomposed area in the XPS spectra with the peak at 396 eV to the total area of N 1s. The total N concentrations for the powder samples were evaluated to be 1.0 atomic %, a; 1.1 atomic %, b; 1.4 atomic %, c; 1.1 atomic %, d; and 1.0 atomic %, e.

Figure 4B shows the decomposition rates of methylene blue under visible light (λ > 400 nm) as a function of the ratio of the decomposed area in the XPS peak at 396 eV to the total area of N 1s. Each powder sample included the total N concentration of about 1 atomic %. An increase of photocatalytic activity with an increase of the component of N with the XPS peak at 396 eV is clearly observed. The decrease in photocatalytic activity observed in the powder sample (e) may be attributed to a change in the crystal structure of the sample caused by the high doping of N in the crystal, although we did not find a noticeable change in XRD. An optimum concentration of N with the XPS peak at 396 eV, which may depend on the preparation process, was found at ∼0.25 atomic % (read from Fig. 4B and the total concentrations of N in the samples) or TiO1.9925N0.0075. All these results, combined with the theoretical analyses described before (along with Fig. 1), consistently show that the active sites of N for photocatalysis under visible light are the substitutional ones that can be identified with the atomic β-N states peaking at 396 eV in the XPS spectra.

The active wavelength of TiO2-xNx, of less than 500 nm, promises a wide range of applications, as it covers the main peak of the solar irradiation energy beyond Earth's atmosphere (around 460 nm) and an excellent light source peaking at 390 to 420 nm, provided by recently developed light-emitting indium gallium nitride diodes (27).

  • * To whom correspondence should be addressed. E-mail: rasahi{at}


View Abstract

Navigate This Article