Tuning oxygen vacancy photoluminescence in monoclinic Y2WO6 by selectively occupying yttrium sites using lanthanum

The effect of isovalent lanthanum (La) doping on the monoclinic Y2WO6 photoluminescence was studied. Introducing the non-activated La3+ into Y2WO6 brings new excitation bands from violet to visible regions and strong near-infrared emission, while the bands position and intensity depend on the doping concentration. It is interesting to find that doping La3+ into Y2WO6 promotes the oxygen vacancy formation according to the first-principle calculation, Raman spectrum, and synchrotron radiation analysis. Through the Rietveld refinement and X-ray photoelectron spectroscopy results, La3+ is found to mainly occupy the Y2 (2f) site in low-concentration doped samples. With increasing doping concentration, the La3+ occupation number at the Y3 (4g) site increases faster than those at the Y1 (2e) and Y2 (2f) sites. When La3+ occupies different Y sites, the localized energy states caused by the oxygen vacancy pair change their position in the forbidden band, inducing the variation of the excitation and emission bands. This research proposes a feasible method to tune the oxygen vacancy emission, eliminating the challenge of precisely controlling the calcination atmosphere.

The effect of isovalent lanthanum (La) doping on the monoclinic Y 2 WO 6 photoluminescence was studied. Introducing the non-activated La 31 into Y 2 WO 6 brings new excitation bands from violet to visible regions and strong near-infrared emission, while the bands position and intensity depend on the doping concentration. It is interesting to find that doping La 31 into Y 2 WO 6 promotes the oxygen vacancy formation according to the first-principle calculation, Raman spectrum, and synchrotron radiation analysis. Through the Rietveld refinement and X-ray photoelectron spectroscopy results, La 31 is found to mainly occupy the Y2 (2f) site in low-concentration doped samples. With increasing doping concentration, the La 31 occupation number at the Y3 (4g) site increases faster than those at the Y1 (2e) and Y2 (2f) sites. When La 31 occupies different Y sites, the localized energy states caused by the oxygen vacancy pair change their position in the forbidden band, inducing the variation of the excitation and emission bands. This research proposes a feasible method to tune the oxygen vacancy emission, eliminating the challenge of precisely controlling the calcination atmosphere.
T ungstates are a kind of self-activated luminescence materials. That can be divided into several categories, normal metal tungstates (MWO 4 ), rare earth tungstates (Re 2 WO 6 ) and poly-tungstates 1-3 . Since Kroger concluded that the lattice group (WO 4 22 /WO 6 62 ) itself was responsible for the luminescence origin 4 , what influenced tungstates luminescence properties was explored extensively such as morphology, size and dimension 5,6 . In addition to intrinsic emission of anion-cation groups, there was also emission from defect states, inevitably incurred because of the abundant synthesis methods and flexible annealing temperatures and atmospheres [7][8][9] . For example, the photoluminescence intensity of amorphous BaWO 4 was higher than that of crystalline BaWO 4 because of different annealing temperatures 10 . Therefore, the tungstate hosts luminescent properties were of great interest 11,12 .
Though various methods have been employed to improve luminescent properties, process parameters especially annealing atmospheres, in particular oxygen partial pressure, were not controlled precisely 13 . An easy and convenient approach is to dope impurities in matrixes to enhance emission or obtain multi-color emission [14][15][16] . The impurity can be any elements for the non-isovalent doping, such as trivalent rare earth and monovalent alkaline metal ions 17,18 . This method has been extensively investigated in luminescent compounds, photocatalysts, and magnetic materials. For example, La 31 -doped ZnO has high photocatalytic activity 19 , and LaCoMnO 6 presents the coexistence of ferromagnetic and antiferromagnetic properties with increased Ca 21 substitution amounts 20 . For tungstates, this approach mainly aims at enhancing luminescent intensity, changing optical activity, or broadening emission wavelength range [21][22][23] . For instance, the emission wavelength of CaW (Mo)O 4 nanoparticles was tuned from blue-green or yellow to white by increasing the Dy 31 concentration 24 . In addition, there are considerable reports on the La 31 doping effect on the luminescence properties of PbWO 4 25,26 .
According to the first principles study, different electronic compensation effects lead to different defect states in the band gap for high and low doping conditions, which explains the 420 nm band origin and new red absorption band 27 .
Similarly, the isovalent doping can also change the luminescence properties of matrixes such as salt compounds and oxides 28,29  However, the photoluminescence mechanism of the matrix such as lanthanum, lutetium and yttrium tungstates (La 2 WO 6 , Lu 2 WO 6 , Nd 2 WO 6 and Y 2 WO 6 ) 31-37 after importing isovalent and nonactivated ions is still open to be exploited. The luminescent properties of Bi 2 WO 6 with and without La 31 doping were compared at low temperature 4 K 38 , having found that La 31 doping increased the stokes shift of the matrix luminescence.
Various activators and sensitizers have been doped to improve the emission of monoclinic Y 2 WO 6 [33][34][35][36][37]39 . The main purposes of these investigations are how to obtain white-light emission or promote energy conversion efficiency. The oxygen vacancy and local crystal structural regulation of monoclinic Y 2 WO 6 by non-activated ions have not been reported until now. Recently, we 40 found that the atmosphere and calcination temperature induced the changes of oxygen vacancy concentration and tungsten coordination number in monoclinic Y 2 WO 6 , and thus affected the appearance of longwave excitation and near-infrared emission bands. By calcining Y 2 WO 6 in the air at 1200uC, the 340 nm excitation band, caused by low-concentration oxygen-vacancy, was substantially enhanced in comparison with those calcined at high temperature or in argon. Calcining in argon resulted in strong infrared emission because of the increased oxygen vacancy concentration 40 . But the oxygen partial pressure, which depends on the airtightness of the furnace, cannot be controlled purposefully when the sample is calcined in the air. Therefore, it is ideal to tune the oxygen vacancy by a simple and feasible method on the basis of the doping approach advantages. Moreover, the mechanism behind the impurity effect on the luminescence of the matrix needs more intensive investigation to make up for the deficiency of the previous theory.
In this paper, a series of Y 2 WO 6 :xLa 31 (x 5 0 and 0.01-0.05) powders are synthesized in air condition at 1250uC through the simple solid phase reaction. These samples show strong visible emission. It is surprising to find that the 340 nm excitation intensity of the powders with dopant concentration not more than 3 at% is stronger than that of the pristine Y 2 WO 6 . The La 31 doping can also produce many new excitation bands in the ultraviolet and visible regions. These new excitation bands are ascribed to the different oxygen vacancy pair behavior induced by occupation variation of La 31 in the three Y sites. When La 31 enters into the Y2 (2f) site at low concentration, the oxygen vacancy pair energy band locates just above the valence band (VB), intensifying the 340 nm excitation band. At high doping concentration, the occupation number of La 31 in the Y3 (4g) site becomes high, bringing new localized energy states and excitation bands and weakening the 340 nm excitation intensity. The change of oxygen vacancy energy states generates different luminescence phenomenon.

Results and Discussion
Crystal structure. The crystal structure of pure Y 2 WO 6 is monoclinic phase with space group 13-P 12 /C 1 -C 2h 4 reported by Efremov 41 , whose inorganic crystal structure database (ICSD) number is 20955. In order to check the phase purity of as-prepared samples, X-ray diffraction (XRD) measurement results are plotted in Figure 1. All the XRD patterns agree well with the patterns of powder diffraction file (PDF) card 73-0118 and no peaks from other phases such as La 2 WO 6 are observed. Due to the effective ion radius difference of Y 31 (0.96 Å and 1.019 Å for VII and VIII coordination) and La 31 (1.10 Å and 1.160 Å for VII and VIII coordination) 42 , the diffraction peaks of the La 31 -doped samples slightly deviate when La 31 substitutes for Y 31 .
The structure refinement of Y 2 WO 6 :xLa 31 (x 5 0.03, 0.05) powders was performed through the general structure analysis system (GSAS) software package 43 . The calculated patterns are consistent with the experimental XRD patterns (Supporting Information, Figure S1(a) and (b)). The atomic positions, occupation numbers, crystal structure and refinement parameters are listed in Table 1. The volumes of the samples gradually become larger with increasing La 31 content, while the atomic coordination and lattice parameters change a little. From the occupation numbers, one can see that La 31 enters into three Y sites simultaneously at any concentrations, occupying mainly the Y2 (2f) and Y3 (4g) sites and minorly the Y1 (2e) site. In 3.0 at% and 5.0 at% La 31 -doped samples, the occupation  There is a W (4g) site possessing C 1 symmetry, which is surrounded by six O atoms to form distorted octahedral coordination. The bond lengths of the six W-O are not identical each other. Three kinds of Y sites were coordinated with eight (2e, 2f) and seven (4g) oxygen atoms constructing polyhedron coordination. The Y1 (2e) and Y2 (2f) sites have C 2 symmetry, while the Y3 (4g) site takes on C 1 symmetry 44 31 also possesses a stable structure as evidenced by the invariability of the XRD patterns ( Figure 1) [46][47][48][49] .
Photoluminescence properties. Figure 2 shows the photoluminescence (PL) emission and excitation spectra of all Y 2 WO 6 :xLa 31 (x 5 0 and 0.01-0.05) samples. Under 340 nm excitation, the PL emission spectra have broad band shapes covering from 365 to 650 nm with maximal value around 470 nm. This band results from charge transition emission between the local oxygen 2p states (just above the VB) and the conduction band (CB) 40 . The PL intensities in pure and low-concentration La 31 -doped samples (x # 0.03) are stronger than those of high-concentration La 31 -doped powders. From Figure 2(a), the asymmetric shapes of emission spectra do not depend on dopant concentration and remain singlepeak frameworks. In addition, the total, low-energy side, high-energy side half widths and peak values show slight dependence on La 31 concentration (Supporting Information, Table S1) 50 . Thus, the emission band is composed of at least two overlapping bands 50 . Moreover, the powders also show the near-infrared emission in the scope of 1000-1700 nm as depicted in Figure 3 . When the concentration continues to rise to 4 at% and 5 at%, the infrared luminescence is also observed in two regions (1300-1400 nm and 1450-1700 nm), besides much stronger emission from 1000 to 1150 nm. In order to exhibit the fine structure clearly, we have enlarged the emission spectra of 4 at% and 5 at% La 31 doped samples from 1300 nm to 1700 nm (supporting information, Figure S2). Tuning the La 31 content changes the infrared luminescence, suggesting the presence of ample local states in the band gap 51 .
To obtain a better understanding of photoluminescence phenomenon, Figure 2(b) displays the excitation spectra of all samples by monitoring 520 nm emission. All samples show three excitation bands containing two short wavelength bands (peaking at 280 and 310 nm) and one long wavelength band (peaking at 340 nm). When La 31 doping concentration increases from 1 at% to 3 at%, the 340 nm band gradually intensifies compared with the pristine sample. When the doping concentration exceeds 3 at%, the intensity of this band gradually weakens. A similar excitation band also appeared in other tungstates such as CaWO 4 52 and ZnWO 4 :Bi 31 , Eu 31 phosphors 53 . Their origins were ascribed to oxygen vacancy and 1 S 0 R 3 P 1 transitions of Bi 31 . For air-annealed Y 2 WO 6 samples 34,40 , this band was originated from low-concentration oxygen vacancy. Therefore, the dramatic variation of the 340 nm excitation band intensity undoubtedly originates from the oxygen vacancy defect and La 31 doping effect. The tunable defect emission intensity is obtained by changing the La 31 content. When the detector wavelength extends to the nearinfrared ranges, many new excitation peaks, such as 380, 491, and 523 nm, appear in 1-3 at% La 31 -doped powders. For samples doped with higher content of La 31 , a series of peaks at 380, 482, 522, 533, 577, and 591 nm are observed. These new excitation bands were ascribed to the oxygen vacancy pair in Y 2 WO 6 40 . In the pristine sample, only the 340 nm excitation band is observed. Hence the behavior of oxygen vacancy changes a lot due to the incorporation of La 31 in Y 2 WO 6 .
Local crystal environments. In order to explore these new excitation bands origins and intensity variation of the 340 nm excitation band, we carried out Raman measurement to determine the local vibration structure of all the samples. According to group theory 40,54 , Y 2 WO 6 crystals have 3N 5 3 3 36 5 108 distinct Raman and Infrared    vibration modes. As we know, Raman spectra of tungstates can be identified with two types of groups as external and internal vibration modes 55 . The external vibration modes concerning lattice phonons correspond to the motion of polyhedral YO 8/7 clusters. The internal vibration modes originates from the vibration of distorted octahedral WO 6 clusters 56,57 , which are written as C internal vibration~A1g z E g zF 2g z2F 1u z2F 2u . The Raman active vibration modes of all Y 2 WO 6 :xLa 31 powders ranging from 100 to 1000 cm 21 are shown in the Supporting Information ( Figure S3). Table 2 lists all the Raman peaks positions and the coordination numbers according to similar chemical formulas M 2 W(Mo)O 6 (M 5 La, Nd, Sm, and Bi) 57 . In the undoped sample, the strongest peak at 834 cm 21 is assigned to the W 5 O symmetrical stretching vibration A 1g mode and has a half width of 20 cm 21 , which indicates that the coordination number (CN) between tungsten and oxygen is 6 40 40 , it is confirmed that the local crystal structure is sensitively affected by the calcination environment which depends on the individual furnace used during the samples preparation.
The Raman spectra of La 31 -doped samples differ from that of the pure sample by exhibiting some new Raman peaks. In Y 2 WO 6 : xLa 31 (x 5 0.01-0.03), the peaks at 798, 773, 381, and 207 cm 21 suggest that some tungsten atoms have tetrahedral coordination because the peaks positions are similar to those of La 2 MoO 6 , Bi 2 MoO 6 and Nd 2 MoO 6 57 . The peak value 156 cm 21 is analogous to that of Sm 2 WO 6 , which indicates that CN of several tungsten atoms reduces to 5. Therefore, there exist tetrahedral and pentacoordinated tungsten atoms in 1-3 at% La 31 -doped samples. For highconcentration La 31 -doped samples, the peak shapes and numbers are almost similar to those of 1-3 at% La 31 -doped samples. In addition, the sudden enhanced peak at 934 cm 21 may be a combined tone with the sum of 773 and 156 cm 21 . Therefore, the tetra-and penta-fold tungsten atom numbers become larger after La ions are introduced into Y 2 WO 6 . To study the crystal environments further, the extended X-ray absorption fine structure (EXAFS) measurements of W-L III absorption in Y 2 WO 6 :xLa 31 (x 5 0 and 0.01-0.05) samples are applied to determine the local structure around W atoms. Through Fourier transformation of the fine structure signals, Figure S4 shows the radical structure functions of W atoms. A strong peak in Figure  S4 corresponds to the nearest neighbor O atoms of W ion. Furthermore, the fitting results are given in Table S2 by further Fourier fitting of this peak.
The average coordination number (CN) of tungsten and oxygen can be calculated as CN 5 (attenuation factor) 3 6/A, where A is the value of W-O in standard sample (generally 0.7-0.9) and 6 is the theoretical coordination number. From the table S2, the average number of tungsten is reduced with the increase of La 31 content. Hence, it is confirmed further that oxygen vacancy concentration becomes high gradually. Though the attenuation factor of 2 at% La 31 -doped sample is slightly larger than those of 1 at% and 3 at% Ladoped samples, the three values are assumed approximately equal. The tungsten CN is gradually reduced with increasing La 31 concentration, which hints that oxygen vacancies exist in all samples. Since the external environments are the same in the experiment, the oxygen vacancy variation is undoubtly ascribed to the doping effect of La 31 .
In order to shed more lights on the La 31 information in Y 2 WO 6 , the XPS spectra of the La 3d core level for the Y 2 WO 6 :xLa 31 with x 5 0.03 and 0.05 are shown in Figure 4. The La 3d XPS in the two samples can be fitted as a superposition of four Gaussian components. The peaks at 834 and 851 eV can be attributed to two spin orbits of La 3d 5/2 and La 3d 3/2 , respectively 58 . The other two peaks at 837 and 854eV are La 3d satellite peaks. Hence, the double peak structure of each spin-orbit split agrees with the reported literatures 59 . As we know, the split spin orbit reflects states with configurations [3d 9 ] hole 4f 0 L and [3d 9 ] hole 4f 1 [L] hole , where L indicates the oxygen ligand. Generally, the f 0 dominates the low binding energy signals, and the high binding energy is referred to the f 1 peaks. Because the doping concentration has no obvious effect on the La 3d f 1 /f 0 intensity ratio, the f 1 -f 0 energy separation and peak shift are considered.
The f 1 -f 0 separation values are 3.628 eV and 3.404 eV in these two samples, which differ from those in ABO 3 perovskites, La 2 CuO 4 (3.1 eV), LaCoO 3 (4.3 eV) and La 1.85 Ba 0.15 CuO 4 (5.3 eV) 59 . Thus,  Photoluminescence mechanism. To obtain the deep understanding of luminescence origin, the first principle method is often applied to derive electronic structures of luminescent materials 62 . The appearance of four-and five-coordination tungsten atom numbers from the pure to different-concentration La 31 -doped samples indicates that the oxygen vacancy concentration increases gradually with incorporation of La 31 into the samples. Hence, we first establish a perfect 1 3 2 3 1 Y 2 WO 6 supercell, and then one or two oxygen atoms next to tungsten are removed to constitute single and twin oxygen vacancies together with replacing the nearest Y atom with a La atom (these models are labeled as La Yk 1 V O(i) and La Yk 1 V O(ij) with k 5 1, 2, and 3 and i, j 5 1, 2, 3, 4, 5, and 6, i ? j). Under oxygen-rich atmosphere, the defect formation energies (E formation ) of various models when oxygen vacancy locates at different sites are plotted in Figure 5. As illustrated in Figure 5(a), for the models containing one oxygen vacancy, the variation rules of E formation are the same for the undoped and La 31 -doped models expect for the model with V O (6) . Though the average E formation for V O(i) is smaller than those of La Y1 1 V O(i) and La Y3 1 V O(i) , La 31 also can occupy Y1 and Y3 sites in the process of high-temperature calcination. Therefore, the probability of La entering into the Y2 site exceeds that of entering into Y1 and Y3 sites. The calculation results accord well with those of XRD refinement. For the La Yk 1 V O(ij) models, the minimal E formation is located at different sites for the four configurations (V O(ij) and La Yk 1 V O(ij) ). The average values of the four cases are 4.1819, 4.3838, 3.6547, and 4.4581 eV. Similarly, the E formation average value of La Y2 1 V O(ij) is smaller than that of the V O(ij) , and the difference of average E formation between V O(ij) and La Yk 1 V O(ij) (k 5 1 and 3) samples is very small. Hence, the La 31 doping induces oxygen vacancy increase, which is consistent with the analysis of Raman spectra and synchrotron radiation.
For self-activated luminescent tungstates, the CB and VB are mainly composed of W 5d and O 2p states. Thus, the tungstate luminescence origin is intrinsic luminescence 62 . Moreover, luminescence caused by intrinsic defects such as oxygen vacancies or interstitial atoms also exists in tungstates 63,64 . For our samples, there exists amply oxygen vacancy luminescence information. As previously reported, the origin of the 340 nm excitation band is ascribed to low-concentration oxygen vacancy namely some five coordination tungsten atoms 40 . In samples containing La 31 , the oxygen vacancy concentration increases in comparison with the pristine Y 2 WO 6 , resulting in some four and five-fold tungsten atoms.
Based on these results, we calculate the electronic structure of the samples containing both La 31 and an oxygen vacancy pair, to exploit the role of La doping on the behavior of the oxygen vacancy pair. As found previously 40 , single oxygen vacancy accounts for the excitation bands shorter than 400 nm (340 nm), while oxygen vacancy pair causes a series of excitation bands from 380 to 600 nm. We only need to find out why in the low-concentration La 31 -doped sample which contains an oxygen vacancy pair like the high-concentration La 31 -doped samples the 340 nm excitation band dominates. La 31 prefers to enter into the Y2 site at low concentration and then enters into the Y3 site at high concentration. Therefore, we studied the electronic structures of samples containing an oxygen vacancy pair, where La 31 occupies three Y sites, namely La Y1 , La Y2 , and La Y3 , respectively. When the E formation of these structures are closest to the average value of E formation in La Yk 1 V O(ij) , we calculated their electronic properties.
On the basis of the Y 2 WO 6 crystal structure 65 , three-type Y sites are surrounded by different kinds of oxygen atoms. The nearest  (24) , and La Y3 1 V O (14) configurations are calculated, because their E formation are closest to the average E formation . Figure 6 displays the total density of states (DOS) and partial DOS of the constituted atoms. The CB and VB are mainly composed of W 5d and O 2p states with small contributions of Y 4d. The contributions of all the La electron states for VB and CB are almost zero. The W 5p, Y 4s, W 5s, La 6s, La 4d, Y 4p, O 2s, and La 5p are located below the VB from 245 eV to 210 eV. The electronic structure properties of Y 2 WO 6 are similar to those of some tungstates and molybdates such as scheelite CaWO 4 and wolframite ZnMoO 4 66,67 . Therefore, the luminescence origin of Y 2 WO 6 is mainly ascribed to the charge transfer transition between W and O 68 . Moreover, the local state positions and numbers for the La Y2 1 V O(24) model are similar to those of Y 2 WO 6 with low-concentration oxygen vacancy 40 . For La Y1 1 V O (36) and La Y3 1 V O (14) models, these states resemble those of Y 2 WO 6 with high-concentration oxygen vacancy 40 . Thus, these differences of local states induce different excitation and emission phenomena. In low-concentration La 31 -doped samples, La mainly enters into the Y2 sites resulting in ample oxygen vacancy pair.   When these samples are radiated under ultraviolet (UV) light larger than the gap value (E gap ) 3.75 eV, the electrons jump from the VB to the CB then relax and finally emit photons. Thus, the 280 nm and 310 nm excitation bands can be produced. When the UV light energy is smaller than E gap , the electrons in (3) states (local O 2p states) jump to the CB, generating the 340 nm excitation peak, while electrons jumping from (3) to (4) (local W 5d states) produce the 378 nm weak excitation band. Therefore, the 340 nm peak intensifies gradually with increasing La Y2 numbers (not more than Y2 sites numbers). When the doping concentration increases, the content variation of La Y3 is larger than those of La Y2 . The electrons in (5) also jump to the CB corresponding to 340 nm excitation. However, the transition from (5) to (6) becomes a direct transition leading to strong 592 nm excitation, which weakens the electron transition for the contribution of 340 nm. Therefore, the 340 nm intensity becomes weak (does not disappear) in high-concentration La 31doped samples.
From a phenomenological viewpoint, the occupation number of Y2 site becomes higher but does not reach saturation, and La 31 tends to occupy the Y3 sites at high doping concentrations. Thus, the 340 nm band intensity strengthens at low La content and then weakens with further increasing doping concentration. In addition, a few La 31 ions enter into the Y1 site as shown in table 1, and the transitions between the VB and the local states (1) can produce some excitation peaks around 500 nm. When La 31 occupies different Y sites, along with the oxygen pair, the local crystal structure shows different changes such as bond length and electronic density. The W-O bond length was measured as shown in Table S3. Since the six W-O bond lengths are unequal, the average values are computed. The average bond length variations in models La Y1 1 V O (36) , La Y2 1 V O (24) , and La Y3 1 V O (14) are 0.048, 0.026, and 0.045 Å , respectively. One can see that the bond lengths change relative to the pristine system in model La Y2 1 V O (24) is the smallest, and thus smallest distortion promotes the 340 nm excitation intensity. For models La Y1 1 V O (36) and La Y2 1 V O (24) , bigger distortions lead to other forbidden transitions becoming allowed transitions, thus weakening the 340 nm intensity. Different bond length variations result in different local state distributions in the band gap. Table 3 lists the energy level positions of VB maximum (VBM), local states, and CB minimum (CBM). From Figure 6, Figure S5 and Table 3, one can see that the 340 nm excitation band weakens or disappears when La 31 enters into the Y3 or Y1 sites. Therefore, the first principle calculation explains the gradual death of the 340 nm excitation band and the emergence of many new peaks. Finally, the schematic oxygen vacancy forming process is plotted ( Figure S6).
Conclusion. In the present paper, we study the concentration effect of isovalent La 31 doping on the photoluminescence of monoclinic Y 2 WO 6 . The incorporating of La 31 into the matrix favors the formation of single and coupled oxygen vacancies. At low doping concentration, La 31 prefers to occupy the Y2 (2f) site, while at high concentration, it mainly occupies the Y3 (4g) site. When La occupies the Y2 (2f) site, the local states caused by the oxygen vacancy pair locate just below the CBM and connect with the VBM. When La occupies the Y3 (4g) site, new local energy bands appear. As a result, besides the emission of the W-O group in the visible region, there appear new emission and excitation bands because of the mid-gap excitation using light longer than 320 nm. At low La 31 doping concentration, the 340 nm excitation band is substantially intensified, resulting in visible emission. A series of excitation bands in the visible region also appear, causing strong nearinfrared emission. The abundant change of the excitation and emission spectra with the La doping is ascribed to the single and coupled oxygen vacancy change and the selective occupation of La to different Y sites. In our previous articles 34 , the Y 1.98 WO 6 :0.02Sm 31 phosphors can emit white light under 340 nm excitation. The luminescence is mainly originated from the tungstate group and Sm 31 emission. In this paper, La doping in monoclinic Y 2 WO 6 greatly improves luminescence intensity under 340 nm excitation. Therefore, the strong white-light emission can be anticipated through co-doping the non-activated La 31 and luminous Re 31 (Sm 31 , Eu 31 ) in self-activated Y 2 WO 6 host under the near-violet irradiation.

Experiment and calculation details
Samples preparation and characterization. Y 2 WO 6 :xLa 31 (x 5 0 and 0.01 , 0.05) powders were prepared through solid-state reaction. The detailed experiment steps and characterization methods were described previously 34,40 . For La 31 -doped samples, the raw materials were added different amounts of La 2 O 3 . The chemical state of the La element was examined by x-ray photoelectric spectra (XPS) using a Thermoelectron ESCALAB 250 spectrometer equipped with monochromatic Al X-ray source (1486.6 eV).
Calculation procedures. To determine formation energy and the electronic structure, we use VASP software to simulate the doping effect in periodic supercell structures 69 . In the following, the La Y1 model was used an example to illustrate the calculation steps. First, a unitcell was built, and then relaxed fully. Second, the relaxed unitcell was expanded to a 1 3 2 3 1 supercell. For samples containing single or coupled oxygen vacancy, we constructed and fully optimized the 1 3 2 3 1 supercell, where one La substituted for Y1 and O1 or O1 and O2 (six-type oxygen sites) atom(s) near the W atom neighboring La was removed. Third, the defect formation energy (E formation ) of all La Yk 1 V O(ij) models were calculated. Finally, when the E formation of one model was closest to the average value of the fifteen models in every type of La Yk , the DOS and energy band structure were computed. In order to overcome the bandgap underestimation drawback of density function theory (DFT) calculation, the generalized gradient approximation (GGA) 1 U method was applied in all calculations. Through a series of tests, the optimal U values for O, Y, La and W were found to be 4.5, 0.0, 0.0 and 9.9 eV 40 .