In 1972, Fujishima et al. [1] discovered photoelectron chemical splitting of water on TiO2 electrodes, which led to extensive investigations of TiO2 as a photocatalyst. TiO2 is one of the most promising photocatalysts for the photocatalytic degradation of organic pollutants and photocatalytic dissociation of water. However, the wide bandgap of anatase TiO2 (3.2 eV) limits its photocatalytic applications to the ultraviolet (UV) light range. Meanwhile, the relatively high rate of electron-hole recombination results in a low quantum yield and poor efficiency in promoting photocatalytic reactions. These fundamental problems prevent the use of anatase TiO2 in practical applications. Therefore, one of the endeavours to improve the performance of TiO2 is to increase its optical efficiency by shifting the onset of its response from the UV to the visible region [2, 3, 4, 5]. One strategy is to dope TiO2 with an anion such as B, C, N, F, P, or S to decrease its band gap.
Compared to other anion dopants, N has been proven to be one of the most efficient for visible light-responsive TiO2 photocatalysts [3, 5], and this system has been studied extensively by experiments. Various methods have been developed to prepare N-doped TiO2 photocatalysts (powders and films), such as sputtering [3, 6, 7, 8], ion implantation [9, 10, 11, 12], controlled hydrolysis or sol-gel [13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24], and chemical treatment of TiO2 [3, 9, 12, 25, 26, 27, 28, 29, 30]. Although these N-doped TiO2 photocatalysts are visible light-active, there are debates on their structures, particularly on the location of the active N atoms and the origin of the bandgap narrowing in N-doped TiO2. Asahi et al. [3] indicated that N atoms substituted for O atoms in N-doped TiO2 and the decrease of the bandgap originated from the mixing of the O 2p and substitutional N 2p states. Diwald et al. [9] concluded that N atoms can also be located at interstitial sites in N-doped TiO2, and they further confirmed that this nitrogen state was responsible for the observed shift of the photochemical threshold of rutile TiO2(110) down to 2.4 eV. Irie et al. [12] proposed that the isolated narrow N 2p band formed by substitutional N in N-doped TiO2 above the O 2p valence band was responsible for the visible light response. Serpone [31] argued that the visible light activation of N-doped TiO2 was due to defects associated with oxygen vacancies that gave rise to color centers. Recently, Livraghi et al. [19] investigated N-doped TiO2 by a combined experimental and theoretical approach and suggested that Nb· centers (single atom nitrogen in the bulk of TiO2) played an essential role in the absorption of visible light, in the promotion of electrons to the conduction band, and in photoinduced electron transfer to reducible adsorbates. So far, the most common technique to detect doped N atoms in TiO2 is by X-ray photoelectron spectroscopy (XPS). N 1s XPS peaks with different binding energies have been reported for N-doped TiO2 prepared by different methods, but their assignments are still not conclusive. Compared with N in TiN, the N 1s XPS feature with its binding energy at 397.0 eV is generally assigned to N2- anions that are substituted for O in the TiO2 lattice. However, Chen et al. [32] attributed a N 1s peak with its binding energy at 401.3 eV to substitutional N in O-Ti-N. Diwald et al. [9] prepared N-doped rutile TiO2(110) by heating in NH3 and observed two N 1s peaks with binding energy at 396.7 and 399.6 eV, which were assigned to substitutional and interstitial N atoms, respectively. Rodriguez et al. [33] reported the N 1s binding energy of atomic N adsorbed on rutile TiO2(110) to be 399.0 eV. Therefore, a clear understanding of the structures of N-doped TiO2 photocatalysts is of great importance. Recently, other nonmetal-doped TiO2 with dopants like B, C, or F have also received growing interest from both experiments and theoretical calculations [19, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46].
In theoretical studies, Di Valentin et al. [19, 36, 41, 42, 43, 44, 45, 46] have published a number of papers on nonmetal-doping of TiO2. In their work, they performed a lot of electronic state analysis on the localized states formed in the band gap of TiO2. In addition, they also performed core level shift (CLS) studies for different doping positions. Their CLS value of different N locations in N-doped anatase TiO2 is 1.6 eV with the initial state approximation. They also showed in their calculations that there existd a difference of 2.6 eV between interstitial B and substitutional B with initial state relaxation effects considered [36]. However, their CLS results may be more convincing if they had considered final state relaxation in their calculations.
The locations of anions and thermodynamic stability are two aspects of most concern to experimentalists. Therefore, in this work, we comprehensively studied the CLS and thermodynamic stability for different doping anions (B, C, N, F, P, and S), different doping types (substitutional and interstitial), and different TiO2 phases (rutile and anatase). In the energetics calculations, especially with N-doping, we have considered the difference between the PBE functional and hybrid HSE06 functional.
In the first part of the work, we determined the optimized structures of the anion-doped TiO2. Then in the second part, we employed standard and hybrid density function theory (DFT) calculations for an energetics study of anion-doped TiO2. We study the thermodynamics in an effort to find the fundamental reason for the different N-doped TiO2 types identified by the different preparation methods. Although there are less experimental studies than on N-doped TiO2, some conclusions for the systems of B-, C-, F-, P-, or S-doped TiO2 can be given. In the last part, DFT calculations with final state relaxation effects included were employed to provide comprehensive and precise insight into the CLS results of anion-doped TiO2. In this part, we studied different doping locations, anions, and TiO2 phases. We hope these CLS calculations help distinguish the close peaks in the XPS spectra.
Spin-polarized total energy calculations were performed based on the all-electron projected augmented wave (PAW) method and DFT within the generalized gradient approximation (GGA-PBE) using a HSE06 hybrid functional as implemented in the Vienna Ab Initio Simulation 5.2 Package (VASP) [47, 48, 49, 50, 51, 52, 53, 54]. A cutoff of 400 eV was used for the plane wave expansion. Throughout the present work, optimized equilibrium lattice constants (anatase TiO2: a = 3.82 Å, c = 9.62 Å; rutile TiO2: a = 4.66 Å, c = 2.97 Å), which agreed well with previous calculations [45] and experiments (anatase TiO2: a = 3.78 Å, c = 9.52 Å; rutile TiO2: a = 4.59 Å, c = 2.96 Å) [5], were used unless otherwise stated. To study anion doping in the TiO2 bulk, one anion atom was included in the anatase (3×3×1) supercell with dimensions of 11.46 Å × 11.46 Å × 9.62 Å and rutile (2×2×3) supercell with dimension of 9.32 Å × 9.32 Å × 8.91 Å, which corresponded to the doping concentration of 1.4% and 2.1%, respectively. For the Brillouin zone integration, we employed a Monkhorst-Pack (2×2×4) and (4×4×4) Γ-centered k-point grid for the anatase (3×3×1) and rutile (2×2×3) supercell. All the ions in the super cell were relaxed until the residual force on each ion was less than 0.01 eV/Å.
CLS were calculated as the energy difference between the core level binding energy of the atom of interest (Eicl) and a reference atom (Erefcl), i.e., CLS = Eicl - Erefcl. The core level binding energy (Ecl) can be calculated in both the initial state and final state approximations. In the initial state approximation, Ecl is the negative eigenvalue (-ɛc) of the orbital from which the core electron has been excited, where Ecl = -ɛc, and ɛc refers to the Fermi level for the solid and vacuum for the gas. In the final state approximation, Ecl was calculated by the total energy difference between the two configurations, which were an excited configuration in which an electron was removed from a particular core state, Etot(nc - 1) and the ground state configuration, Etot(nc). Thus, Ecl = Etot(nc - 1) - Etot(nc), in which the core-hole was assumed to remain localized in the excited atom.
In this work, for the final state calculation, a core electron was removed from the core by generating the corresponding core excited ionic PAW potential in the calculation, and the remaining core states were relaxed in the self-consistent calculation.
In this work both anion substitution and interstitial doping were considered. As all the O atoms are in the same position, there is only one possibility for the anion substitution location. However, for the interstitial-doped anion, there are three possible equivalent positions in anatase TiO2 and two possible positions in rutile TiO2, as shown in Fig. 1. The optimized coordinate of the interstitial N atoms in each structure and the total energy of the optimized structure are summarized in Fig. 2. For interstitial N-doped anatase, the A3 type exhibited the lowest total energy, in agreement with the previous result [45]. For interstitial N-doped rutile, the R2 type is more stable. Therefore, the a2 type and the R2 type were chosen as the models for the further theoretical investigations of interstitial N-doped anatase and rutile, respectively. Similar work was done on the other anion-doped TiO2. The results are listed in Table 1. It can be seen that the location depended on the size of the atomic radius. Larger anions prefer the larger interspace. Compared with bulk TiO2, the atoms surrounding the doped anion were slightly relaxed in the anion-doped TiO2.
To compare the reaction enthalpy of the reactions producing interstitial and substitutional anion-doped TiO2, we considered the reaction, which is that of one anion with TiO2 to produce anion-doped TiO2. The source of the anion atom would offset in the enthalpy comparison of two types of anion-doped TiO2, thus for interstitial doping, there is only one process in which anions get into the favorable interstitial space of the TiO2 lattice. However, for substitutional doping, there are three processes in which O atoms first are removed from the TiO2 lattice leading to the formation of oxygen vacancies, and anions fill these vacancies at the same time. As the sources of the anion and final form of the O atom are not unique, the doping reaction can be described as the following for anatase TiO2 on the basis of the structural model:
Interstitial anion-doped reaction
Substituting anion-doped reaction
For rutile TiO2, the doping reaction equations are the following:
DH was calculated by the equation DH = ΣE(products) − ΣE(reactants), in which E represents the total energy.
Reaction (2) is not the same as reaction (1), which has stable products, so the final form of the O atom should be included in reaction (2). In other words, there should be an additional enthalpy DH5 added to DH2 in the case of the comparison with DH1, where DH5 is the enthalpy of the process where the O atom changed into the final form, which is determined by the particular reaction route.
To identify the error from using different methods, both the PBE functional and HSE06 hybrid functional were used for N-doped TiO2, as listed in Table 2. It can be seen that the energetics did not exhibit a strong dependence on the functional used in this study. Because the HSE06 functional is more time consuming than the PBE, the other anion-doped TiO2 energetics calculations were performed with the standard PBE functional. The results are shown in Fig. 3 and Fig. 4. As the anions do not appear in reaction (3), the values of DH3 do not change for the same system, which are 7.11 and 7.20 eV for anatase and ruile TiO2, respectively. The substitutional anion-doped reaction was strongly endothermic for both anatase and rutile TiO2, but the interstitial anion-doped reaction can be exothermic or slightly endothermic. There existed a large energy barrier for substitutional N-doped TiO2 for both anatase and rutile TiO2, and the value of DH3 was a quite positive number compared with DH1. The larger energy value can explain why substitutional N-doped TiO2 usually requires severe reaction conditions, such as ion implantation and sputtering, or nitridizing TiO2 in N2 or NH3 at high temperatures, whereas interstitial N-doped TiO2 can be prepared via mild wet chemistry reactions. However, this comparison cannot provide insights into which product is formed at a given severe reaction condition, and the process of how the O atom changed into its final form therefore should be taken into account. In the following part, we used O2 as the final form of the O atom to address this issue. The total energy of O2 was also included in our work, with the energy of a O2 E(O2) = -9.857 eV and a single O atom E(O) = -1.90 eV. Thus another reaction is O = 0.5O2 + DH5, with the value of DH5 = -3.02 eV. DH5 should be added to DH2, and then compared with DH1. Fig. 3 shows that for B- or P-doped TiO2, interstitial doping was still more favorable than substitutional doping, which means that under severe conditions and with the final form of O, interstitial B- or P-doped TiO2 is still more easily prepared than substitutional B- or P-doped TiO2. N-, C-, or S-doped TiO2 tended to have a similar enthalpy change, while F-doped TiO2 showed the reversed trend. As the value of DH5 depended on the final form of the O atoms, different preparation solutions can lead to different mixtures of interstitial and substitutional doped products. In experimental preparations when severe conditions were provided for N doping of TiO2, there were always two N 1s peaks in the XPS spectra [3, 9, 12] whereas when the conditions used were mild, there was only one peak [23, 24]. From the energetics study, we can identify that with a mild environment, only interstitial N-doped TiO2 is produced, and both types of N-doped TiO2 exist when severe conditions are used.
For the core level of the anion doped into TiO2, the two different types of doping: substitutional and interstitial doping showed two close peaks in the XPS spectra. These two peaks cannot be distinguished just by the experimental data, and the CLS by a DFT study can be used to identify these close peaks. We have found that the most stable location is the interstitial site. Then, we performed a CLS study of the two different doping types with standard DFT for B-, C-, N-, F-, P-, and S-doped TiO2. The results are shown in Fig. 5 and Table 3.
From Fig. 5, we can see that for B-, C-, N-, F-, P-, and S-doped TiO2, there exists a clear difference between substitutional and interstitial doping. Most interstitial anion dopants type had a higher core level binding energy than that of the substitutional dopant, except for F-doped TiO2. For N-doped TiO2, it was found that the N 1s core level binding energy of the interstitial N species was higher than that of the substitutional N species for both N-doped anatase TiO2 and rutile TiO2. The CLSs were 2.52 and 2.17 eV for N-doped anatase and rutile TiO2, respectively. Experimentally, N 1s XPS peaks with different binding energies have been observed by XPS for N-doped TiO2 prepared by different methods. Asahi et al. [3] assigned the N 1s peak with a binding energy at 396 eV to substitutional N in N-doped anatase TiO2 prepared by sputtering a TiO2 target in a N2 atomsphere. Chen et al. [32]attributed the N 1s peak with a binding energy at 401.3 eV to substitutional N in N-doped anatase TiO2 prepared by treating TiO2 with an excess of triethylamine. Diwald et al. [9] prepared N-doped rutile TiO2(110) by heating in NH3 and observed two N 1s peaks with binding energies at 396.7 and 399.6 eV, which were assigned to substitutional and interstitial N atoms, respectively. Fang et al. [23] assigned the N 1s peak with a binding energy at 399.6 eV to interstitial N in N-doped anatase TiO2 prepared by a wet chemistry method. Comparing the N in TiN with a N 1s binding energy at 397 eV [57], it has been generally accepted that the N 1s peak with a binding energy at 396-397 eV is from substitutional N in N-doped TiO2. However, the evidence for the assignment of the N 1s peak with a binding energy at 399.6 eV to interstitial N in N-doped TiO2 is lacking. Our theoretical calculations fully support the assignments: substitutional N in N-doped TiO2 exhibits N 1s binding energy between 396 and 397 eV and interstitial N in N-doped TiO2 exhibits the N 1s binding energy at 399.6 eV. Our results also suggested the assignment of the N 1s peak with a binding energy at 401.3 eV to substitutional N in N-doped anatase TiO2 already made by Chen et al. [32]. Meanwhile, on basis of the calculated enthalpy for the N-doping reaction, it is also reasonable to conclude that treating TiO2 with an excess of triethylamine that was employed by Chen et al. [32] should not be able to form the substitutional N species.
We summarized the CLS results of different anion-doped TiO2 in Fig. 5 and Table 3. Our calculations showed that the ECLS value of anion-doped TiO2 was positive except for F-doped TiO2. This means that in the XPS spectra of anion-doped TiO2, the interstitial doping peak is located higher than that of substitutional doping. The CLS results (Fig. 5) showed that the F atom gave a reverse trend compared with other anions. This can be understood through the Pauli electronegativity of these atoms [56] (Table 1). The electronegativity value of O is larger than those of B, C, N, P, and S, while smaller than that of F. After an O atom is substituted by an anion, the Ti atoms are closer than with the O atoms, while for interstitial doping, the O atoms are closer than the anion. The difference in the structure between substitutional and interstitial doping is that more nearby O atoms surround the anion in interstitial doping. According to the sequence of the electronegativity [56], there is more electron transfer from O to F in interstitial doping than substitutional doping; while for B-, C-, N-, P-, or S-doping of TiO2, there is the reverse electron transfer direction, and substitutional doping gets more electrons than interstitial doping. As we know, when more electrons are moved away, the binding energy would be larger. Thus, the B, C, N, P, and S atoms in interstitial doping sites have a higher core level energy than those in substitutional doping sites; and F-doped TiO2 has a reverse trend. When this conclusion is applied to the XPS spectra, for B-, C-, N-, P-, and S-doped TiO2, the peak from interstitial doping is higher than that from substitutional doping; and for F-doped TiO2, that from substitutional doping is higher than interstitial doping.
We have studied the anion CLS for both the rutile and anatase, as shown in Fig. 5 and Table 3. The results revealed that both anatase and rutile showed the same sequence of the substitutional and interstitial dopant peaks. However, for the same dopant, the core level binding energy of the anion in anatase is a little higher than that in rutile. The probable reason may be that the density of rutile is larger than that of anatase. As the distance between two neighboring atoms is less, the interaction between Ti and the anion is stronger, thus, the anion can accept more electrons in rutile than anatase. As discussed above, when more electrons are lost, the core level binding energy is higher. Therefore, the binding energy of the anion in anatase is higher than in rutile.
We have performed comprehensive and improved DFT calculations of anion-doped anatase and rutile TiO2. The N 1s core level binding energy calculations suggested that interstitial N species exhibit a higher N 1s binding energy than substitutional N species in N-doped TiO2. The calculated core level shift between interstitial and substitutional N species agreed well with experimental data, which strongly supports the assignments of the experimental XPS results. Our calculations provide a fundamental understanding of N-doped TiO2 photocatalysts. We also performed CLS calculations for B-, C-, F-, P-, and S-doped TiO2 and show that in F-doped TiO2, the interstitial dopant peak is lower than that of the substitutional dopant. However, for B-, C-, N-, P-, and S-doped TiO2, the peak of the interstitial dopant is higher than that of the substitutional dopant. The enthalpy calculations demonstrated that the substitutional anion-doped reaction is strongly exothermic for both anatase and rutile TiO2 due to the intermediate production of oxygen vacancies, whereas the interstitial anion-doped reaction is modestly endothermic. Our calculations suggested that for B, C, N, F, P, and S anion-doped TiO2, severe experimental conditions are required for substitutional doping, while for interstitial doping, wet chemistry methods would be enough.
2015, Vol. 36 Issue (2): 181-187
PDF (21287 KB)
2015, Vol. 36 Issue (2): 181-187
PDF (21287 KB)