Molecular nitrogen in N-doped TiO2 nanoribbons

Chimie des Interactions Plasma Surface, C Parc, B-7000 Mons, Belgium. E-mail: carla.b Jožef Stefan Institute, Jamova cesta 39, SI-1 Jožef Stefan International Postgraduate Sch Slovenia Helmholtz-Zentrum Berlin für Materialien Matter and Functional Materials, Albert-Ein Faculty of Mathematics and Physics, Unive 1000 Ljubljana, Slovenia Graduate School of Advanced Integration Sc Japan † Electronic supplementary informa 10.1039/c4ra14410d Cite this: RSC Adv., 2015, 5, 23350


Introduction
Nitrogen doping titanium dioxide (TiO 2 ), with N at substitutional sites has been reported to be indispensable for enhancing the use of TiO 2 as a visible-light photocatalytic material. 1 The main effect of the N-doping is the narrowing of the energy band gap of TiO 2 due to the mixing of N 2p and O 2p states. 2 Additionally, an isolated narrow band responsible for the visible light photoactivity is formed above the valence band. 2 In substitutional N-doping, the N atom bonds to three Ti atoms replacing a lattice O atom in TiO 2 . Otherwise, N atoms can also occupy interstitial sites where N is bound to one or more O atoms thus creating paramagnetic centres. The electronic properties of N-doped titania, the corresponding bonding conguration, and the oxidation states of nitrogen in the lattice are yet to be unambiguously determined. [3][4][5] Recent reports based on density functional theory (DFT) calculations and are consistent with both substitutional and interstitial nitrogen doping, [6][7][8][9] where interstitial doping is slightly lower in energy especially for surface states with low concentration of oxygen vacancies. 3 For further improving the photocatalytic performance of titania, it is thus critical to identify the local bonding congurations around N atoms at the atomic level.
Nitrogen doping into the TiO 2 lattice is rather challenging as N atoms must be accommodated geometrically and electronically. 2,[10][11][12] To address the fundamental issues discussed above we explored a chemical route and report here a comprehensive investigation of N-doped TiO 2 nanoribbons. As a precursor material for N-doped TiO 2 nanoribbons we used hydrogen titanate nanoribbons (HTiNRs) because upon heating HTiNRs easily transform rst to the monoclinic TiO 2 (ref. 13) b-phase (TiO 2 -B) and then to anatase (tetragonal phase) while retaining the morphology of the parent nanostructure intact. 14 In the used chemical route, the precursor HTiNRs were calcined at different temperatures for different times in an NH 3 (g)/Ar(g) ow (Table 1). When calcined at 400 C the HTiNRs successfully converted to TiO 2 -B, whereas calcination at 580 C and 650 C yielded in a mixture of two TiO 2 phases. Morphology probing transmission electron microscopy (TEM) and structure sensitive X-ray diffraction (XRD) techniques were employed to extract critical local and global bonding information. The nitrogen chemical environment was inspected with X-ray photoelectron spectroscopy (XPS), while its bond geometry was probed by transmission X-ray microscopy combined with near-edge X-ray absorption ne structure spectroscopy (NEXAFS-TXM). The nitrogen chemical environment (i.e., substitutional or interstitial) was found to depend on the calcination temperature. In N-doped TiO 2 nanoribbons calcined at 580 C and 650 C, besides the expected covalently bonded nitrogen, surprisingly, also N 2 -like species were detected through N 1s XPS and N K-edge NEXAFS. The present results provide a coherent description of N-sites in TiO 2 nanostructures and may thus facilitate further improvements in their photocatalytic performance.

Experimental
Synthesis of N-doped TiO 2 nanoribbons N-doped TiO 2 nanoribbons were synthesized from hydrogen titanate nanoribbons (HTiNRs) by calcination in a dynamic NH 3 (g)/Ar(g) atmosphere. The reaction was carried out in a horizontal furnace, which was purged with a mixture of NH 3 (g)/Ar(g). The NH 3 (g)/Ar(g) ow ratio was set to 30 mL min À1 /10 mL min À1 , respectively. In a typical experiment 225 mg of HTiNRs, prepared from sodium trititanate nanoribbons 15 by an ion-exchange process, 14 is weighted on a quartz boat, placed into the hot zone of the furnace and heated at a ramp rate of 7 C min À1 to a selected temperature. The samples were calcined at the selected temperature for four or six hours. The experimental conditions for the transformation of HTiNRs to TiO 2 nanoribbons, nitrogen content and phase composition are summarized in Table 1.

Characterization techniques
The morphology of the N-doped TiO 2 nanoribbon samples was investigated with a eld emission scanning (FE-SEM, Jeol, 7600F) and a transmission (TEM, Jeol 2100) electron microscopes.
The phase analysis was performed on the cut surface by X-ray powder diffraction (XRD) using a diffractometer with Cu K a radiation (l ¼ 1.5406Å) and a Sol-X energy-dispersive detector (Endeavor D4, Bruker AXS, Karlsruhe, Germany).
The XPS measurements were performed in a VERSAPROBE PHI 5000 from Physical Electronics, equipped with a Monochromatic Al Ka X-ray source. The energy resolution was 0.7 eV. For the compensation of built up charge on the sample surface during the measurements a dual beam charge neutralization composed of an electron gun ($1 eV) and an Argon ion gun (#10 eV) was used. The samples for XPS measurements were prepared by pressing the sample into a pellet. A conductive double face tape was used to attach the pellet to a sample holder.
The near edge X-ray adsorption ne structure (NEXAFS) spectra were recorded with the TXM (transmission X-ray microscope) end-station installed at the undulator beamline U41-XM at the electron storage ring BESSY II, Helmholtz-Zentrum Berlin (HZB). 16 The spectra were recorded at room temperature in a transmission mode by taking a sequence of images over a range of photon energies covering the investigated absorption edges with E/DE $ 4500. The exit slit of the monochromator was set to 5 mm, which corresponds to a calculated spectral resolution of E/DE ¼ 20 000. The data analysis was accomplished by using aXis2000. 17 The NEXAFS spectra were normalized by using the signal intensity in the proximity of the sample to correct for variations of the photon ux with photon energy (hn) and acquisition time. For comparison and calibration molecular nitrogen N 2 (g) was analyzed using Nitrogen 5.0 with 99.999% purity (Air Liquide).
For the TEM and NEXAFS-TXM analyses the samples were sonically dispersed in ethanol and a drop of the solution was deposited onto a lacey carbon lm supported by a copper grid. Note that due to electron-beam damage effects on the N-doped TiO 2 nanostructures, different grids were used for the TEM and for the NEXAFS-TXM analysis.
The EPR measurements were conducted using a commercial Bruker E580 spectrometer equipped with an Oxford Cryogenics He ow cryostat that has an optical window. The spectra were measured by recording the intensity of the echo signal as a function of magnetic eld. A Hahn-echo pulse sequence (p/2-s-p-s-echo) together with appropriate phase cycling has been used. The length of p/2 pulse was set to 20 ns and the delay between the two pulses was s ¼ 200 ns. The sample light-illumination was performed in situ at 35 K with a standard halogen lamp. A sample for EPR measurements was sealed under dynamic vacuum in a standard Suprasil quartz EPR tube.

Results and discussion
The typical morphology of the nanostructures prepared by calcination in the NH 3 (g)/Ar(g) ow, presented in Fig. 1, show that transformation from hydrogen titanate to N-doped TiO 2 does not affect the nanoribbon morphology. The nanoribbons reach a length of several micrometres with diameters in the range of 30 to 250 nm and are thus comparable to the starting hydrogen titanate nanoribbon (HTiNRs) sample. 18 Table 1 Experimental conditions for the transformation of hydrogen titanate nanoribbons to N-doped TiO 2 nanoribons in an NH 3 (g)/Ar(g) flow. T ( C) temperature of transformation, t (h) duration time of the transformation, N nitrogen relative atomic concentration (as determined by XPS) and phase composition Sample The porous structure is characteristic of TiO 2 -B nanoribbons and with further transformation to anatase disappears as we reported in ref. 14.
The crystal structure of samples treated at different calcination temperatures was examined by powder XRD (Fig. 2). The XRD pattern of the sample calcined at 400 C reveals only the presence of the TiO 2 -B phase (JCPDS no. 35-0088), whereas the samples treated at 580 C and 650 C come as a mixture of TiO 2 -B and anatase phases. 13,19 The amount of anatase in the samples increases with increasing calcination temperature up to 650 C. At this temperature TiO 2 already starts to transform to titanium oxynitride as indicated with the appearance of a new diffraction peak at 2q $ 43 . 20 The XPS analysis of the samples conrmed the presence of nitrogen in all calcined samples ( Table 1). The relative content of nitrogen increases with the calcination temperature in agreement with data published in the literature. [21][22][23] The samples with different nitrogen contents have colours that range from white over light yellow to light green. The higher the nitrogen content, the more intense is the colour of the sample. The XPS Ti 2p spectra show a Ti 2p 3/2 binding energy of 458.7 eV with a splitting of 5.7 eV between the doublets, characteristic of titanium in the oxidation state 4+ (Fig. 3A, S2 and S3 †). 3,24 The absence of peaks corresponding to titanium in lower oxidation states in the XPS spectra of N400-6h and N580-6h suggests that there is no signicant net electron transfer to titanium atoms from oxygen vacancy formation. 3 Conversely, in the spectrum recorded on the sample calcined at 650 C (N650-4h) we observed a low intensity shoulder at low binding energies characteristic of the photoelectrons generated in titanium species having lower oxidation states (Ti +3 at 457.3 eV and Ti +2 455.5 eV) 25 (Fig. S3 †). Fig. 3B shows the typical O 1s XPS spectrum recorded on the studied samples ( Fig. S2B and S3B †). It exhibits a major peak centred at 530.0 eV, which is characteristic of transition metal oxides. The broad peak centred at 531.2 eV was attributed to hydroxyl groups and carbon compounds contamination at the TiO 2 surface. 2,26 Most likely the carbon contamination  originates from the preparation of HTiNRs 14 where ethanol was used for the nal rinsing of HTiNRs aer the ion exchange.
The N 1s XPS peaks recorded on N580-6h and N650-4h, show four components whereas in the N1s peak of N400-6h only three components are observed (Fig. 4). The component with binding energy at 396.1 eV is assigned to O-Ti-N linkage in the lattice, i.e. lattice oxygen atoms are replaced by nitrogen atoms (substitutional N-doping). The relative intensity of this component and the total amount of nitrogen incorporated increased with increasing calcination temperature in accordance with the results reported by Zhang et al. on the N-doping of on titania nanoparticles prepared by a sol-gel technique. 21 The component at 396.7 eV was assigned to carbon contamination 26 as observed in the O 1s XPS spectra, while the component at 400.1 eV to nitrogen species bonded to oxygen sites (N-O-Ti linkage); i.e. an interstitial N-doping mode formed by N-atoms bonded to one or more lattice oxygen atoms. 2 The contribution of the component associated to photoelectrons emitted from N atoms in O-Ti-N linkages to the N 1s XPS spectrum increases at the expense of the other three components as the calcination temperature increases. DFT calculations showed that adsorbed and interstitial nitrogen atoms have a tendency to bond to oxygen vacancy sites and accept electrons, 6 explaining the high intensity of the component associated to N atoms in the O-Ti-N linkage. The highest binding energy peak at 403.7 eV has been ascribed to molecular N 2 . 24,27 This binding energy is signicantly lower than for gas phase N 2 molecules (409.9 eV), indicating that photocreated core holes maybe screened by neighbouring atoms as was observed in physisorbed N 2 molecules on various surfaces. 28 In order to investigate the presence of N 2 molecules in the N-doped TiO 2 nanoribbon samples (N580-6h and N650-4h), we next studied the Ti L-edge and the N K-edge using the NEXAFS-TXM, which allows recording NEXAFS spectra from isolated nanoribbons or selected regions of the sample. 16 The features of these edges are sensitive to the local symmetry and ligand coordination, providing detailed information on crystal structure, oxidation states of solid samples and vibrational ne-structure of gas molecules.
The TXM-NEXAFS spectra were recorded on a region of interest of an isolated nanoribbon (Fig. S4 †). The Ti L-edge spectrum of the N-doped TiO 2 nanoribbons contains two sets of doublets (Fig. S5, † column A), Ti L 3 at lower energy and Ti L 2 at higher energy. The two doublets originate primarily from electronic transitions from the Ti 2p 3/2 and Ti 2p 1/2 levels to Ti 3d orbitals, which split into t 2g and e g levels in the  approximately octahedral symmetry such as in the anatase TiO 2 phase. 29 The features of these Ti L-edge indicate that the coordination environment of the titanium atoms in the N-doped TiO 2 samples corresponds to Ti(IV) in a tetragonal structure as suggested by the XPS measurements. 16 The N K-edge spectra recorded on N580-6h and N650-4h show a very sharp peak at 401 eV that is not present in the spectrum recorded on N400-6h (Fig. 5). To examine this more carefully, we recorded detailed spectra in the 399-403 eV photon energy range with step energy of 0.05 eV and compared with the spectrum recorded on a standard sample of molecular nitrogen (Fig. 6). In both cases we observed ve well-resolved peaks that are characteristic of free N 2 molecules. 24,30,31 The peak at 400.8 eV of the gas phase N 2 molecules was assigned to a transition from the N 1s to p* orbital and the ne feature is a hallmark of N 2 molecule due to its vibrational structure. 32 Therefore, the observed ne structure in the N K-edge undoubtedly proves the presence of N 2 molecules in the N-doped TiO 2 nanoribbons. This is further supported by the observation that the energy position of the 5 components of the peak, their area ratio, and energy separation are similar to those of gas phase N 2 molecule. This means that N-N bond order of N 2 in the N-doped TiO 2 nanoribbons is comparable to that of gas phase N 2 molecule and the interaction of N 2 with the N-doped TiO 2 NRs can be considered as very weak. 30 Therefore, the N 1s XPS and the N K-edge NEXAFS spectra support the idea that nitrogen atoms trapped in N-doped TiO 2 nanoribbons may combine to form N 2 molecules evolving into the gas phase upon heating at elevated temperatures. This behaviour agrees well with reported DFT calculations which show that nitrogen embedded in TiO 2 is unstable energetically and spontaneously forms trapped or gaseous N 2 . 3 Therefore, the XPS and the NEXAFS analyses suggest that for the doping parameters used, the incorporated nitrogen atoms remain in the oxide structure as substitutional or interstitial sites and do not segregate into the ordered titanium oxynitride phase for calcination temperature in the range of 400 to 580 C; segregation was observed at 650 C.
The formation of paramagnetic NO x defects and the possible changes in the electronic structure aer nitrogen doping prompted us to carry out electron paramagnetic resonance (EPR) measurements. EPR has proved to be a very powerful tool for the investigation of local properties of various paramagnetic centres in TiO x nanostructures. 18,[33][34][35][36][37] Moreover, it can also distinguish between different NO x centers through the values of electronic gfactors and hyperne coupling tensors A. 18,38,39 The X-band EPR spectrum of N580-6h sample measured at 35 K with the echodetected eld-sweep technique is shown in Fig. 7A. Clearly, at least three different signals can be recognised. The broad component at high elds can be simulated with the uniaxial gfactor anisotropy yielding g t ¼ 1.907 and g k ¼ 1.939. These eigenvalues match those expected for the Ti 3+ surface sites, (Ti 3+ ) S . 40 The large homogenous broadening (DB 1/2 ¼ 4.7 mT) used in the simulation for this component is consistent with the strains and local inhomogeneities associated with such surface states and the t could be improved if a distribution of g-factor values is considered. Much narrower homogenous broadening (DB 1/2 ¼ 0.9 mT) is characteristic for the second component with typical g t ¼ 1.974 and g || ¼ 1.986, also obtained from the powder lineshape t. These values allow us to assign this signal to the bulk Ti 3+ sites, (Ti 3+ ) B . 41 Finally, the tting of the low-eld signal at g ¼ 2.0054 with two characteristic powder hyperne satellite lines anking the sharp peak yield the hyperne coupling values of A t ¼ 18 MHz and A || ¼ 94 MHz. Both, g-factor and hyperne coupling values are characteristic values for NO paramagnetic centres 18 and provide independent support for the formation of N-O-Ti linkage. The EPR data fully corroborates  the results of the XPS study (Fig. 4) and thus unambiguously prove the incorporation of N into the TiO 2 lattice.
Additionally, since the nitrogen doping is also known to enhance visible light photoactivity in TiO 2 , we measured EPR spectra of illuminated samples. Interestingly, different components of the EPR spectrum respond differently to light illumination. The spectrum measured at 35 K aer illumination (Fig. 7B) still contains the same NO, (Ti 3+ ) B , and (Ti 3+ ) S signals as their respective g-factors and hyperne coupling constants coincide with those before the illumination (Fig. 7A). However, while the signal assigned to surface Ti 3+ sites remains almost the same, the signals of NO and bulk Ti 3+ centres double in intensity aer 10 min of illumination. This nding thus implies that photoexcited charges preferentially trap in the bulk of the nanoribbons where they form stable paramagnetic centres.

Conclusions
We successfully prepared N-doped TiO 2 nanoribbons using hydrogen titanate nanoribbons as a precursor by a thermal treatment in a dynamic ammonia atmosphere. Two types of nitrogen species with bonding energies of 396.1 eV and 400.1 eV were observed by X-ray photoelectron spectroscopy. The two nitrogen species can be assigned to substitutional nitrogen doping and nitrogen bonded to oxygen sites (interstitial), respectively. The amount of the substitutional nitrogen increases with the increasing calcination temperature. The NEXAFS-TXM studies show that local bonding environments of oxygen and titanium atoms in N-doped TiO 2 nanoribbons are similar to the undoped TiO 2 . Depending on the calcination temperature N 2 molecules can be formed and remain weakly trapped in the N-doped TiO 2 structure. The changes observed in the EPR spectrum upon light illumination indicate that photoexcited charges preferentially trap in the bulk of the nanoribbons where they form stable paramagnetic centres. Fig. 7 X-band electron paramagnetic resonance spectra of N580-6h measured at 35 K with an echo-detected field-sweep technique (grey shaded area) before (A) and after (B) illumination with light. The spectra fit to the sum of three different signals (solid red line) ascribed to paramagnetic NO and to bulk (Ti 3+ ) B and surface (Ti 3+ ) S sites, respectively.