Results and discussion
4-1 Nitrogen-doped Titanium Dioxide
(I) XRD
N-doped TiO2 powder prepared by annealing the TiO2 powder at various nitrogen pressures is evidenced by XRD θ-2θ scans shown in Fig. 16. The XRD pattern for pure TiO2 without the treatment is also included for comparison. All samples are single phase (rutile) with no evidence of secondary phase within the XRD measurements. Therefore, the possibility of the contribution of ferromagnetism from the secondary phase can be excluded. In addition, a color change in the TiO2 powders from white to black was noticed after treatment. It is called F-center which is a type of crystallographic defect in which an anion vacancy in a crystal is filled by one or more electrons, depending on the charge of the missing ion in the crystal. Electron in such a vacancy tends to absorb light in the visible spectrum such that the originally transparent material becomes colored [69, 70].
30 40 50 60
(II) SQUID
Fig. 17(a) shows the magnetization (M) versus field (H) curves taken at 300K for pure TiO2 (rutile), V0-TiO2, and N-TiO2 powders. From the partial hysteresis loop shown in Fig. 17(b), it is evident that pure TiO2 is completely paramagnetic, while V0-TiO2 is predominantly paramagnetic with minute sign of ferromagnetism. On the contrary, the N-TiO2 exhibits marked ferromagnetic feature at room temperature. The magnetic parameters for V0-TiO2 are estimated to have 6×10-6 emu/g for remnant magnetization and 35.7 Oe for coercive force, respectively. The slight ferromagnetism is believed to arise from the contribution of oxygen vacancies introduced by heating the TiO2 powders at high temperature in vacuum. Coey et al. [71] proposed that the ferromagnetism is originating from direct exchange interaction between the molecular orbital consisting of the valence electrons of the three titanium ions and the oxygen vacancy surrounded by them. The molecular orbital results in ferromagnetism when the concentration of oxygen vacancies, namely, the concentration of molecular orbital, is large enough. However, if the concentration of oxygen vacancies is low, the isolated molecular orbital will result in paramagnetism. This model is similar to that described by the BMP model. In the BMP scenario, oxygen vacancies act both as electron donors and electron traps which can bind the electron, and then combining with the three titanium ions in the vicinity to form a bound magnetic polaron. In Coey’s case, it is named as molecular orbital. If neighboring BMPs do not interact strongly with each other, a paramagnetic, insulating phase results. However, for certain BMP–BMP distances and combinations of electron–electron and electron–local moment exchange constants, the BMPs may couple in a ferromagnetic fashion.
On the other hand, when nitrogen is introduced by high temperature annealing, it is clear from Fig. 17(b), an obvious M-H hysteresis loop representing the existence of
ferromagnetism is seen. From the hysteresis loop, we estimate a saturation magnetization (Ms) of 9×10-4 emu/g, a remnant magnetization of 7×10-5 emu/g, and a coercive force of 74.1 Oe for sample N-TiO2, respectively. Those values are apparently orders of magnitude larger than those obtained for V0-TiO2, indicating a much stronger ferromagnetic characteristic. Based on the density functional theory (DFT) calculation by Di Valentin et al. [72], it has been shown that a large decrease in the formation energy of oxygen vacancies can be gained owing to the presence of nitrogen atoms in the lattice. It also suggests that oxygen vacancies are most probably induced by N-doping of TiO2. We will confirm this assertion by XPS in the next section. In any case, all of these arguments imply that TiO2 powders will lose more oxygen or have more oxygen vacancies when doped with nitrogen and, more importantly, it also induces strong ferromagnetism. However, it is not clear whether the stronger ferromagnetism is correlated solely to the larger concentration of oxygen vacancies or something else.
Fig. 17 (a). Magnetization versus magnetic field (M-H) at 300K for V0-TiO2 and N-TiO2. (b) The partial hysteresis for all samples.
Alternatively, Shen et al. [53] proposed p-p coupling interaction for ferromagnetism in N-doped ZnO. It was believed that substitution of nitrogen atoms
N 2p localized spin via the p-p interaction, similar to p-d hybridation in TM-doped DMS. This interaction follows essentially from quantum-mechanical level repulsion, which “pushes” the minority states upward, crossing the Fermi-level. A strong p-p coupling interaction between the impurity state and valence-band state is allowed near the Fermi-level. In other words, nitrogen-doping induces interaction between the impurity state and valence-band state so that the minority states cross the Fermi-level.
It effectively has similar results as the Stoner splitting both induces extra states near the Fermi-level. Consequently, if it satisfies the Stoner criterion it would promote spontaneous ferromagnetism. As will be described later, the valence band XPS indeed indicates that there are extra states near the Fermi-level in N-doped TiO2. Briefly, we have found that nitrogen-doping not only creates more oxygen vacancies but also introduces p-p interaction between the N 2p and O 2p. Both factors are probably working together to result in the significant room-temperature ferromagnetism observed in N-TiO2.
(III) XPS
As mentioned above, the XPS spectra may provide crucial information for clarifying the effect of nitrogen dopant and oxygen vacancies on the very different M-H results observed in TiO2 treated under various conditions. In Fig. 18, all the peaks reported were charge corrected using C 1s peak position at 284.5eV as the reference point. Fig. 18(a) shows the high-resolution O 1s XPS spectra of all samples and Fig. 18(b. I, II) is the fitting results for the pure TiO2 and the N-doped TiO2
samples, respectively. From Fig. 18(b. II), the peak can be divided into three symmetric peaks. The origin of the peak at the lowest binding energy (peak 1) is not clear. For the medium binding energy (peak 2), it is ascribed to the O 1s core peak of O2- bound to Ti4+. Perhaps, the most prominent peak is peak 3 with the highest
binding energy, which is referred to as the high binding energy component (HBEC) and the peak 2 is referred to as the low binding energy component (LBEC) [73]. It has been previously reported that the HBEC component develops with the increasing loss of oxygen [74]. Furthermore, the relative area under the curve (area of HBEC peak/area of LBEC peak) is determined to be 0.369 for pure TiO2, 0.387 for V0-TiO2, and 0.680 for N-TiO2, respectively. The relatively large contribution of the HBEC peak for the case of annealing in nitrogen gas strongly suggests the presence of nitrogen may have introduced more oxygen vacancies in our case (Fig.19). This may be also relevant to the obvious M-H curve hysteresis loop observed for the N-TiO2
sample. At this stage, it appears that oxygen vacancies may have played an important role in the origin of ferromagnetism of doped TiO2.
Fig. 18 (a) XPS spectra of O 1s core level for pure TiO2, V0-TiO2, and N-TiO2 samples.
(b) Fitting results of O 1s XPS spectra for (I) pure TiO2 and (II) N-TiO2
535 534 533 532 531 530 529 528 527
In order to further clarify the role played by nitrogen, we have also investigated the XPS of Ti ions. From Fig. 20(a), the high-resolution Ti 2p XPS spectra evidently show that, pure TiO2 only has Ti4+ signal with characteristic 2p3/2 and 2p1/2 spin doublet at 459.3 eV and 465.2 eV, respectively, corresponding to a peak separation of 5.9 eV. Ti4+ is non-magnetic in stoichometric TiO2, since it does not have any unpaired 3d electron. However, unpaired 3d electron in Ti3+ or Ti2+ can lead to some magnetic moment. It is known that oxygen vacancies may change the charge balance and therefore there is a possibility to generate Ti3+ or Ti2+. In Fig. 20(b) we can observe clearly the difference between N-doped TiO2 and pure TiO2. Therefore, we can fit the extra spectra appeared in N-TiO2 and try to understand these. From Fig.
20(b), in addition to the above mention peaks, there are extra peaks labeled as peak 3 and peak 4. These two peaks might correspond to 2p3/2 and 2p1/2 of Ti3+ or Ti2+. The slightly lower binding energy for peak 3 suggests that the removal of oxygen may lead to a higher electron cloud density than in the case of Ti4+. However, at present, we are unable to identify the exact valence of the Ti ions, albeit that a reduction is evident. In any case, the features for both the core level peaks of O 1s and Ti 2p strongly suggest the existence of oxygen vacancies.
Fig. 20 (a) XPS spectra of Ti 2p core level for pure TiO2, V0-TiO2, and N-TiO2samples.
(b) Fitting result of Ti 2p XPS spectra for N-TiO2samples.
Next, we turn to discuss the effect on N-doping on the density of states (DOS) near the Fermi-level. The XPS spectra of valence band for all samples are shown in Fig. 21.
Curves are plotted by setting the Fermi-level to be equal to zero. For one thing, we can observe slight state appeared near the Fermi-level in sample V0-TiO2. On the other hands, it is clear that there exists marked difference in the DOS near the Fermi-level between N-doped TiO2 and the rest of the samples. Pure TiO2 has primarily a filled O 2p derived valence band separated from an empty Ti 3d, 4s and 4p
derived conduction band by a bulk band-gap of 3.2 eV [75]. The valence band spectra show the emission from O 2p band with its upper edge lying 3 eV away from the Fermi-level. It has been pointed out that the feature close to 0 eV in the reduced TiO2
or anion-doped TiO2 can be attributed to the occupied defect states (or impurity state) corresponding to the partial population of Ti 3d band [76]. Thus, in our case, the doping of nitrogen into TiO2 must have somehow induced significant amount of oxygen vacancies, as reflected in the significant growth of the occupied-defect states related feature near the Fermi-level. This is also consistent with the first-principle calculation by Rumaiz [55], wherein it was shown that N doping in TiO2 leads to the formation of oxygen vacancies. It appears that the observed occupied states of Ti-3d near the Fermi-level and the modification of electronic structure are intimately related not only the doped N-impurities but also the oxygen vacancies.
Furthermore, Fig. 21 also shows that the O 2p valence band is shifting to lower binding energy with nitrogen doping. It is suggestive that the incorporation of less tightly bound N 2p level with O 2p level further cause slight band-gap shrinkage.
Alternatively, it may also relate to the p-p interaction between the impurity p states (N 2p) and the host p states (O 2p) at the top of the valence band. Though this interaction, the Stoner splitting might be induced to promote the ferromagnetism given by the Stoner criterion. The simultaneous presence of the nitrogen dopants and oxygen vacancies, in addition to result in the formation of new states near the Fermi-level, may push the system all the way to surpass the Stoner criterion. If we denote the density of states at the Fermi-level as g(εf) and U as the Coulomb energy, the condition for spontaneous ferromagnetism given by the Stoner criterion is U g(εf)≧1.
Thus, it should be interesting to see if our system is in the conditions of leading to Stoner splitting of the band and then induce ferromagnetism.
This exchange splitting of the conduction band can be observed in Fig. 22, where the binding energy shift is attributed to a chemical potential shift. Fig. 22 shows Ti 2p and O 1s core level spectra for pure TiO2, V0-TiO2 and N-TiO2. We can observe that the core peaks of both O 1s and Ti 2p are shifted to lower binding energy. The fact that the N-TiO2 sample has the most binding energy shift among all samples is most likely due to the Stoner splitting [77]. This also explains the marked enhancement of ferromagnetism displayed in the M-H hysteresis for the N-TiO2 sample.
Fig. 21 Valence band XPS spectra for pure TiO2, V0-TiO2, and N-TiO2 samples.
Fig. 22 O 1s and Ti 2p core level peaks for all samples showing the chemical potential shift
In short, in this section, we have shown that nitrogen doping in TiO2 generates oxygen vacancies and new states near the Fermi-level simultaneously, such that
exchange splitting of the band and lead to ferromagnetism.
(IV) EPR
The next question to be asked is how the N-doping cope with the oxygen vacancies to give rise to the observed phenomena? To probe the exact electronic configuration and to gain more insight about the mechanism of the ferromagnetism in the N-TiO2
sample, EPR measurements were carried out. It is an effective tool to investigate the origin and nature of observed ferromagnetism in a material, in particular, to extract information about the oxidation state of the dopant cation involved in the spin coupling.
For the electron structure of N doped TiO2, Di Valentin et al. [78] have proposed two models in their DFT calculations. In substitutional model, nitrogen atom is assumed to replace oxygen in TiO2, so that nitrogen atom accepts the extra charge from oxygen atom and is in the state of negative oxidation. On the other hand, in the interstitial model, the nitrogen atom is assumed to bind with one lattice oxygen to form a NO species. These species, then, interact with the Ti atoms through their π bounding state in TiO2. In either case, however, the unpaired electrons are almost localized on the surrounding of nitrogen atoms.
On the contrary, it was shown experimentally [78,79] that most of the unpaired electrons do not distribute in the vicinity of nitrogen atom based on the observations of the EPR signals which gave g values of 1.99-1.93 attributable to the Ti3+ species.
Fig. 23 reveals the presence of active species trapped on Ti3+ site (1.94) in V0-TiO2
sample, which appears to be consistent with the previous results. Nevertheless, on the right hand side of the Ti3+ signal for N-TiO2 sample, the position apparently shifts by 0.02 to higher magnetic field. The line width (ΔHpp) of V0-TiO2 and N-TiO2 is 70G and 170 G, respectively. There are not any EPR signals in pure TiO2. It conforms that
the removal of oxygen will lead to a higher electron cloud density than that of intrinsic Ti4+, so that the valence of Ti is effectively decreased. However, by comparing with the M-H loop results, the sample V0-TiO2 does not appear to possess obvious ferromagnetism even though the Ti3+ species are already existent. Hence, we speculate that Ti3+ alone may not be the primary factor for ferromagnetism.
Thus, there must be something else involved in giving rise to the significant ferromagnetism seen in N-TiO2 sample. It is noted that the line width of N-TiO2
extends to lower magnetic field and higher magnetic field for the left-hand side and the right-hand side EPR peak, respectively, as compare to that of the V0-TiO2 sample (Fig. 23). We believe that this is due to the supposition of other overlapping signals, presumably originated from the nitrogen impurities. Since the EPR signals for N-centers are having g-values between 2.003-2.005 [78, 79]. This broad signal is attributed to ferromagnetic resonance (FMR) arising from exchange interaction between BMPs. The fact that both V0-TiO2 and N-TiO2 exhibit the existence of the Ti3+ signal but only N-TiO2 showed ferromagnetism, further implies that the combination of nitrogen and the associated increasing in oxygen vacancies is the key to induce ferromagnetism.
(V) Summary
In summary, standard DMSs usually dope transition elements which possess unpaired d or f electron in materials. However, the standard theories for explaining DMSs are found inapplicable to the newly discovered 2p-light elements doped materials. There are several important differences between the 2p and 3d orbital which determine the different magnetic properties exhibited in DMSs doped with 2p LE (anion) and 3d TM (cation). First, the anion 2p bands of the element are usually full in ionic states, leaving no room for unpaired spins compared to 3d bands for transition metal. Second, p states are inefficient in the spin-orbital interaction compared to d states. Third, valence electrons in p states are more delocalized than those in d- or f- states. It means that they have much larger spatial extension which can facilitate long-range exchange coupling interactions. Therefore, 2p light-element doped DMSs can have weak ferromagnetism even with low doping concentration.
Consequently, 2p light elements are appropriate candidate for dopant in DMSs that are able to induce room temperature ferromagnetism. In our experiment, nitrogen doping promotes the formation of excessive oxygen vacancies, as indicated by the marked enhancement of HBEC component in the XPS results. From the valence band XPS spectra (Fig. 22), it is evident that the O 2p valence band moves to lower binding energy with nitrogen doping due to the engagement of p-p interaction. In other words, doping nitrogen into TiO2 not only induces the formation of oxygen vacancies but also activates the p-p interaction between the impurity state (N 2p) and valence band state (O 2p). This interaction pushes the minority states upward, crossing the Fermi-level and triggers the Stoner splitting to extend the tail of the valence-band maximum to lower binding energy and produces extra states below the maximum as compared to undoped TiO2. It not only satisfies the Stoner criterion but also creates
unpaired electrons when charge transfer happens so as to promote spontaneous ferromagnetism. The unpaired electrons originated from the interaction between impurities and the lattice oxygen in TiO2 manifests itself by the broad signal associated with the ferromagnetic resonance in EPR spectra. It acts as spin-polarized carrier sources.
Within the context of current scenario, the combined effects of the p-p coupling interaction and charge transfer near each impurity will tend to align spins and form moment carried by the impurity ion. The moment carrying impurities then couples strongly with carrier spin generated by oxygen vacancies to form long range BMPs.
The primary reason is due to the spatially-extended p-states inherent to the impurities are able to facilitate long-range magnetic coupling. If the concentration of BMPs is sufficiently high, it is able to effectively mediate indirect ferromagnetic coupling among impurities by carriers. Finally, not only spin of impurities but also carrier spins are aligned to result in ferromagnetism (Fig. 24). This model gives a reasonable explanation to the experimentally observe ferromagnetism in TiO2 doped with small amount of nitrogen.
Fig. 24 Schematic showing ferromagnetic coupling between impurities. (a) The magnetic moment of impurities polarizes carriers and aligns the spins of the carriers in the same direction as BMP. (b) If the carrier concentration is sufficiently high, it is able to effectively mediate indirect ferromagnetic coupling among nearly all BMP.
Cation is not shown in the diagram.