Nanoscale insights into doping behavior, particle size and surface effects in trivalent metal doped SnO2

Despite considerable research, the location of an aliovalent dopant into SnO2 nanoparticles is far to be clarified. The aim of the present study on trivalent lanthanide doped SnO2 is to differentiate between substitutional versus interstitial and surface versus bulk doping, delineate the bulk and surface defects induced by doping and establish an intrinsic dopant distribution. We evidence for the first time a complex distribution of intrinsic nature composed of substitutional isolated, substitutional associates with defects as well as surface centers. Such multi-modal distribution is revealed for Eu and Sm, while Pr, Tb and Dy appear to be distributed mostly on the SnO2 surface. Like the previously reported case of Eu, Sm displays a long-lived luminescence decaying in the hundreds of ms scale which is likely related to a selective interaction between the traps and the substitutional isolated center. Analyzing the time-gated luminescence, we conclude that the local lattice environment of the lattice Sn is not affected by the particle size, being remarkably similar in the ~2 and 20 nm particles. The photocatalytic measurements employed as a probe tool confirm the conclusions from the luminescence measurements concerning the nature of defects and the temperature induced migration of lanthanide dopants.

Selection of a luminescence approach that is sensitive more to the local structure is expected to expose better the dopant location and distribution in SnO 2 . The trivalent Eu Ln metal with its characteristic orange/red emission is known to be highly sensitive to its local environment 23 . This explains why the Eu represents the most frequently investigated luminescent dopant for SnO 2 with first reports dating back four decades ago 24,25 . The information extracted from Eu luminescence can reveal whether the dopants segregate on the surface of nanosized SnO 2 or/and they enter the oxide lattice via interstitial or substitutional way. In case of substitutional doping, substitution of Sn(4+) by bulkier Eu(3+) leads to both strain and electric effects due to the mismatch of both ionic radii (0.076 nm compared to 0.095 nm, in six-fold coordination 26 ) and valence. The oxygen vacancies arising from charge-compensation can, in principle, locate either in the close vicinity or distant to Eu dopants leading to Eu-defects associates or substitutional isolated center, respectively. There are several reports that describe the emission shape of Eu as being dominated by relatively strong, three magnetic dipole (MD) emission lines around 590 nm, corresponding to 5 D 0 -7 F 1 transition, in good agreement with the selection rules for the inversion low symmetry (C 2h ) at Sn sites. It is also established that SnO 2 host acts as an efficient antenna that absorbs the UV excitation and subsequently transfers the excitation energy to the substitutional isolated Eu, compensating thus for the intrinsic low absorptions of the parity forbidden f -f transitions 27,28 . Recently, the co-existence of uniform and exponential distributions in trap depths along with the thermal activation mechanism were suggested to explain a unique afterglow decay behavior in Eu -SnO 2 27 . It is established that, as the particle size decreases, the increasing number of surface and interface atoms generates stress/strain and concomitant structural perturbations 28 . By use of luminescence spectroscopy, a strongly distorted Eu environment has been suggested to occur in SnO 2 nanoparticles with dimension close to the exciton Bohr radius (around 2 nm) compared to greater sized nanoparticles 29 . In addition, our literature survey evidences that a complex Eu related emission with both broad and narrow spectral features accompanies the characteristic emission of substitutional isolated Eu centers. The shape of this emission varies strongly with Eu concentration, synthesis approaches, thermal treatment and excitation conditions used in luminescence measurements [28][29][30][31][32][33][34] . The origin of this emission is undecided in the literature, being tentatively related to either Eu -defect associates, impurity phases (such as Eu 2 O 3 or Eu 2 Sn 2 O 7 ) or some unknown phase. In all, the literature data evidence a complex emission of still unknown origin for the Eu-SnO 2 system. Elucidation of the specific location of the aliovalent dopant into the lattice is crucial since this influence significantly the functional properties of doped SnO 2 nanomaterials. As an example, for a different significant semiconductor, anatase TiO 2 , it has been established that only Ln that enter substitutional in the TiO 2 lattice contribute to the change of electronic structure and light absorption efficiency of the host 35 . For this host, substitutional doping mechanism was advanced at least for Nd, Sm, Eu, Er lanthanide ions 36 , whilst for Eu and Sm several lattice sites were identified already by site selective luminescence spectroscopy [36][37][38] .
To date, besides Eu, only Er Ln has been demonstrated by luminescence studies to substitute for the Sn lattice sites in the inversion C 2h sites as an isolated substitutional dopant 39 . Indeed, Chen et al. 38 measured an ultranarrow long-lived emission around 1500 nm which corresponds to the 4 I 13/2 -4 I 15/2 transition having a strong magnetic dipole component 40 . For Nd, substitutional doping was suggested by luminescence and near edge X-ray absorption fine structure (NEXAFS) measurements 41 ; however, the observance of a relatively strong 4 F 3/2 emission of Nd in the range of 800 to 1400 nm which is predominantly of electric dipolar nature sustains more a substitutional doping in association with a nearby defect. In such case, the association of the lanthanide with a defect in the nearest neighbour position remove the inversion symmetry around the dopant and thus the electric dipole transitions become allowed and their intensity dominate the spectra. For Sm 42,43 , Tb 44 or Dy 45,46 doped SnO 2 , the reported emission shapes deviate from the ones expected for an inversion local symmetry [47][48][49] , and thus, the isolated substitutional doping route remain uncertain.
It can be thus concluded that a critical question to be answered when studying doped SnO 2 nanoparticles is How do we know if SnO 2 nanoparticles have been successfully doped? In this context, our study on the trivalent Ln doped SnO 2 is aiming to differentiate between substitutional versus interstitial and surface versus bulk doping, delineate the bulk and surface defects induced by doping and establish an intrinsic dopant distribution. The experimental confirmation of successful doping of semiconductor nanoparticles with luminescent lanthanide activators is usually done in literature by a combination of the ex situ luminescence, X-ray diffraction (XRD), Diffuse Reflectance in the Ultraviolet/Visible (DR-UV/Vis) Raman, and Transmission Electron (TEM) or Scanning Electron (SEM) microscopy methods and techniques. Since for SnO 2 , reports show that the distribution of various metal dopants changes dramatically with temperature, from ordered, lattice sites to disordered, likely surface regions 50 , we use the in situ XRD, Raman and luminescence to provide additional insight into the dynamic evolution of dopant environment during thermal treatment 51 . In addition to conventional bulk doping by rapid hydrothermal or sol-gel methods, the impregnation of Ln on pre-calcined SnO 2 nanoparticles was tested to differentiate between the bulk and surface defects. Photocatalytic oxidation of phenol was used as a reaction tool to further differentiate between the natures of Ln induced defects.

Results and Discussion
Overview of the texture, morphology and structure properties. As synthetized pure SnO 2 consists of well-defined 2-4 nm sized nanoparticles with a near-spherical shape ( Figure S1) aggregated into rather uniform structure with wormhole mesoporosity, originating from the diffusion controlled hydrolysis-polycondensation mechanism 52 . With thermal treatment at 400 °C, more strongly intergrown structure of 5-6 nm particles are evidenced along with sharp decrease of the surface area and increase of the pore size which continues with the increase of temperature at 700 and 1000 °C (Table 1). Calcination at 700 °C led to a bimodal pore size distribution combining mesopores and macropores whilst further increase at 1000 °C induced formation of a compact nonporous oxide.
The purity of Ln (Eu, Sm) doped SnO 2 samples was confirmed by X-ray fluorescence analysis ( Figure S1). Doping SnO 2 with 1% Ln did not cause any change in these textural properties and overall morphology. Generally, similar texture, morphology and structure properties were observed for Eu or Sm doped SnO 2 ; therefore, we describe Eu-SnO 2 in more detail. The ex situ XRD patterns of SnO 2 and Eu-SnO 2 calcined at 400, 700 and 1000 °C are presented in Fig. 1 (see also Figure S3 for the in situ with temperature XRD patterns of Eu-SnO 2 ).
The diffraction angles at 26.3°, 33.8° and 51.8° from the diffractograms illustrated in Fig. 1, can be assigned to (110), (101) and (211) planes of the tetragonal SnO 2 (PDF card 00-041-1445). In the limit of instrumental resolution, impurity phases, such as Eu 2 O 3 or pyrochlore Eu 2 Sn 2 O 7 were not detected. Broadening of XRD patterns with Eu doping together with the crystallite size and lattice parameters values listed in Table 2 demonstrate that Eu doping reduce the crystallite growth in large agreement with literature.
Similar effect was reported for other lanthanide ions and attributed to a solute drag and lattice distortion caused by the substitutional rather than surface doping 53,54 . It is also worth noting that all crystallite sizes of the calcined samples exceed the quantum confinement region set by the exciton Bohr radius of SnO 2 at 2.7 nm 55 . Figure 2 shows the Raman spectra of SnO 2 and Eu-SnO 2 samples. The as -synthetized pure sample presents a broad and intense band at 570 cm −1 which is typically attributed to surface effects along with weak band around 760 cm −1 corresponding to B 2g mode 56 . Samples calcined at 400, 700 and 1000 °C show in addition modes at 270-302, 474-510 (E g ), 626-641(A 1g ) and 680-700 and 768-782 cm −1 (B 2g ) in good agreement with literature 57 . Along with the increase of the temperature, a shift of the A 1g and B 2g lines towards higher wavenumbers and of the E g line to lower wavenumbers is observed, associated to the increase of the particle size 56 . It should be remarked that doping with Eu is leading both to a stronger intensity of the broad bands around 500-570 cm −1 and decrease of the particle size as shown by the XRD patterns ( Fig. 1, Table 2) which confirm the surface nature of this phonon mode.
DRIFTs spectra of SnO 2 and Eu-SnO 2 exhibit the typical profile for SnO 2 58 (see Figure S4 and associated text). Pure samples show that the water content as well as hydroxyl absorption bands progressively disappear as the calcination temperature is increased whilst the presence of Eu appears to preserve more of the hydroxyl defects even at 1000 °C. The DR/UV-Vis diffuse reflectance spectra of SnO 2 and Eu-SnO 2 show a broad absorption band in the UV region ( Figure S5). The shifts detected with the increase of the temperature may be attributed to the XRD observed changes in crystallinity and particle size ( Fig. 1 and Table 2). The band-gaps (Eg) calculated by plotting (F(R)•hν) 2 against hν 18 (see also Experimental Section) are narrower than the value of 3.6 eV stated in literature 9, 10 likely because of high concentration of oxygen vacancies/defects 59 . Further calcination at 1000 °C leads to the decrease of the band-gap up to 2.9 eV for pure SnO 2 which narrowed to 2.7 eV for Eu-SnO 2 .
In situ short to long range investigations in the mild calcination regime. To get deeper insight into how the particle size, surface features, defects and order/disorder effects influence dynamically the local structure around the trivalent dopant, we further analyzed the Raman, luminescence and XRD data obtained by in situ measurements using Eu as a structural probe. In situ luminescence spectra were extracted from the in situ Raman spectra measured during heating and cooling cycles RT-500 °C-RT (RT stands for room-temperature, around 25 °C) 60 while the in situ XRD patterns of Eu-SnO 2 ( Figure S3) were measured during both heating and cooling cycles (RT-1000 °C-RT). As shown in Fig. 3, the position, width and intensity of the Raman modes of Eu-SnO 2 change continuously with temperature during heating cycle. Up to 250 °C the Raman spectra of Eu-SnO 2 are dominated by a broad band around 570 cm −1 which relates to the small particle size/surface effects 56,61 or to amorphous nature of the particles 62 . The average particle size is estimated to increase slightly from 2-4 nm before calcination to 6 nm at 500 °C ( Figure S3).
Besides this broad band (labelled as S 1 in ref. 56) another broad phonon band is localized around 480 cm −1 (labelled as S 2 in ref. 56). In the region of S 2 band, the E g mode around 470-480 cm −1 considered to be the most sensitive to oxygen vacancies, is expected to arise but its intensity is too low to be accounted for in the fitting 57 . Both S 1 and S 2 bands decrease in intensity with increase of particle size at the expense of A 1g mode at 620-626 cm −1 . The two Raman modes were both assigned to surface effects, more specifically to surface layer (around 1 nm as width) formation of a nonstoichiometric SnO 2 with different symmetries than SnO 2 . However, the bands seem to differentiate in what concerns their evolution with temperature: while before calcination their relative intensities are rather similar, at 500 °C the S 2 band exceed in intensity by a factor of four that of S 1 band, equaling again after cooling down to RT. During cooling cycle, a gradual shift towards red, from 620 to 633 cm −1 is measured for A 1g band parallels the slight increase of the particle size ( Figure S3). As concern the in situ luminescence, this evidences broad shapes characteristic of surface Eu, dominated by strong emission around 612 nm which shifts and broadness continuously during heating and to less extent, during cooling cycles. To provide more insight into Eu distribution during the thermal treatment, the middle panel of Fig. 3 zoom the emission around 5 D 0 -7 F 0 transition around 579 nm. This J = 0 − J = 0 transition is well known to be informative (in some conditions 23 ) on the number of distinct Eu sites. Only single band assigned to average single Eu center could be detected with shape being progressively broadened and blue shifted from 578, 5 to 577, 4 nm during heating and then red shifted and slightly narrowed during cooling cycle. The luminescence evolution with temperature appears to Sample S BET (m 2 /g) Pore size (nm) be associated to either S 1 or S 2 or both phonon modes. The broadening of Eu emission during heating means an increased disorder around its nearest environment, likely induced by defects in the surface layer. Further series of experiments using pulsed laser excitation above 300 nm and long delay (time-gate mode) on the as-synthetized and calcined samples at 400 °C show a much weaker and narrower emission dominated by three 5 D 0 -7 F 1 lines at 588 nm/593 nm/599 nm assigned to Eu substituting for Sn(4+) in C 2h inversion sites 24 . In all, the in situ Raman and in situ luminescence results confirm that in the absence of calcination most Eu reside on the surface of ca 2 nm sized SnO 2 nanoparticles with only few Sn inner lattice sites being substituted by Eu. Mild calcination up to 500 °C lead to a significant reduction of surface effects while the redistribution of dopants along with generation of defects induced by weak incorporation process remain limited to the surface layer of ~2-6 nm sized particles.

Substitutional doping, interaction with bulk defects and surface effects at high calcination temperatures.
To date, literature generally describes a complex Eu related emission in SnO 2 with both broad and narrow spectral features that changes strongly varies with Eu concentration, synthesis approaches, thermal treatment and excitation conditions used in luminescence measurements 28-32, 34, 63 . Here, we have extensively analyzed several tens of emission spectra of Eu-SnO 2 calcined in air at 700 and 1000 °C at 80 and 300 K using various excitation wavelengths spanning the UV to Vis range and gate/time delays after the laser pulse ( Figure S6). The specific aim of these investigations was to derive an intrinsic distribution of Eu described in terms of substitutional/interstitial doping, interaction with induced defects and surface effects. By using a simple deconvolution procedure of the emission spectra 64 backed also by the analysis of the excitation spectra and emission decays, at least five Sn centers, labelled as I to V were clearly differentiated (Fig. 5). The emission of center labelled as I is exclusively excited via SnO 2 host as shown by the excitation spectrum measured around 588 nm (band at 290 nm). Its shape is characterized by relative strong three 5 D 0 -7 F 1 lines around 590 nm being readily assigned to Eu substituting for the lattice Sn sites (with inversion C 2h local symmetry) in line with previous reports 24 . In SnO 2 , the C 2h inversion symmetry around Sn cation is shaped by the location of each of the two Sn in the unit cell of SnO 2 amidst six oxygen atoms which approximately from the corners of a regular octahedron 65 (Fig. 4). For Eu substituting for Sn with no defect nearby, the emission should be dominated by the allowed MD emission as the electric dipole (ED) transitions are forbidden in the inversion symmetry.
In opposition to this center I, a much broader emission, excited exclusively by the f -f absorptions of Eu is observed and assigned to surface type Eu center, labelled as center V.
Besides the substitutional isolated C 2h (I) and surface centers (V), at least three more Eu centers, labelled as II-IV (or non-C 2h centers) could be differentiated for the first time (Fig. 5). The characteristic excitation spectra measured around their emissions around 606-612 nm, show a superposition of broad UV absorption bands around 300-320 nm along with narrow f-f absorptions of Eu. In the excitation spectra, the presence of O 2− Eu 3+ charge -transfer band (typically observed between 250-270 nm 66 ) may be also masked on the high-energy side of broad UV absorption of SnO 2 host. There is also a non -negligible interference from the 290 nm band characteristic of C 2h center (I) due to incomplete spectral separation. The average lifetimes of non -C 2h centers (1 ÷ 1.5 ms) are longer than that corresponding to surface center, V (0.7 ms), which is an expected result. The surface      activators have unsaturated low symmetry coordinations being also subjected to efficient non-radiative energy transfer to surface defects, mainly OH groups still present at 700 and even 1000 °C ( Figure S4b). Though the emission lifetime of C 2h Eu is expected to be large due to its forbidden MD nature, it exceeds significantly the values of few ms measured for other hosts with inversion symmetry, such as Y 2 O 3 (S 6 /C 3i ) 47 or CeO 2 (O h ) 48 .
Indeed, as shown in Fig. 5, the emission decay of C 2h center displays an exceptional long decay time of 130 ms as estimated from the tail fitting at 80 K. Such a remarkable long-lived emission has recently been assimilated to an atypical persistent emission process that is thermally activated below (or above) 220 K via the co-existence of uniform and exponential distributions in trap depths 27 .
To check on the possibility that part of II-IV emissions relates to impurity phases such as Eu 2 O 3 or Eu stannate oxide (Eu 2 Sn 2 O 7 ), likely induced by the reportedly low solubility limit of Ln in tin oxide lattice, (below 1%), we compared the luminescence of calcined (700/1000 °C) Eu 2 O 3 and Eu-SnO 2 measured in identical conditions. There is obviously no resemblance between this emission (included as cyan and orange spectra in Fig. 5) and those of the II-IV centers. Next, the absence in the Raman spectra of the strongest characteristic phonon mode of stannate pyrochlore (around 510 cm −1 67 ) along with the strongly dissimilar emission exhibited by Eu 2 Sn 2 O 7 68-70 grown by us using a sol-gel method 71 (also included in Fig. 5 as orange spectrum) to Eu centers II-IV dismiss the occurrence of this phase in our samples.
Further, we asked whether these Eu II-IV centers were of intrinsic nature or resulted from the procedure of preparation or interaction with specific impurities. Comparison with an additional Eu-SnO 2 sample grown by a simple sol-gel method give identical emission shapes to those obtained for samples grown by rapid hydrothermal approach. Besides, according to our extensive literature survey on the luminescence of Eu-SnO 2 published over the last two decades, part of the emission related to centers II-IV can be observed as strongly non-separated, overlapped emission, though discernable in Eu-SnO 2 grown by sol-gel 30,31,34,72 glass-ceramic waveguides 32 , hydrolysis 73 , impregnation and decomposition cycle method 63 , etc. Taken together, the emission shapes and narrowness (few cm −1 , comparable to that of C 2h center) suggest that centers II-IV are intrinsic to SnO 2 , resulting from the association of substitutional Eu with nearby bulk defects, most probably oxygen vacancy. Such defects arise from the charge-imbalance on substitution of tetravalent Sn by the trivalent Eu which in the Kroger-Vink 74 notation is written as: x g 0 2 ( ) O 00 In a further set of experiments, we have surface doped SnO 2 with Eu by impregnation of pre -calcined pure SnO 2 at 400 and 1000 °C followed by calcination at 400 and 1000 °C. The XRD patterns, Raman spectra and DR-UV/Vis as well as particle size, lattice constants and estimated band-gap for these additional samples are summarized in Figs 1 and 2, Figures S5 and S6 and Table 2. Despite weak incorporation expected for the pre-calcination-impregnation-calcination procedures, the impregnated samples do show some emission of the substitutional isolated C 2h center (I) (Figure not shown) embedded into a much broader emission like that of the surface center V. However, the notable result is that no emission can be associated to any of the defect centers (II-IV) giving thus a further confirmation that these centers are induced by Eu substituting for lattice Sn in close association with bulk defects.
Although similar emission properties were measured for Eu-SnO 2 calcined at 700 and 1000 °C, there are some differences in the details. At higher calcination temperature, both weaker emission and appearance of a distinct distorted emission ( Figure S7) suggest a migration of the substitutional Eu towards the nanoparticles surface where they may further associate with adsorbed OH defects 75 . Indeed, the DRIFTS spectra in Figure S4a,b confirm that the surface of the particles is terminated by adsorbed OH (bands centered at 3450 and 1640 cm −1 ) that persist even at 1000 °C and their contribution is exceeding that measured for pure SnO 2 .
The distribution found above for Eu is confirmed by a similar analysis applied to Sm. Compared to Eu, Sm was less frequently investigated as luminescent dopant for SnO 2 48, 66, 67 . Luminescence studies revealed an effective sensitization of Sm luminescence via host absorption 67 similar to Eu case and also some dependency of the emission shape on the excitation wavelength. Figure 6 summarizes the analysis of luminescence properties of Sm-SnO 2 calcined at 700/1000 °C that gives similar results for the employed methods (rapid hydrothermal and sol-gel). Under above bandgap excitation (280 nm) the emission of Sm (labelled as center I) displays narrow (~0.7 nm) and relative strong 4 G 5/2 -6 H 5/2 emission around 568 nm which is mostly of MD nature 76 . A weak intensity is observed for the 4 G 5/2 -6 H 7/2 transition at 606 and 621 nm which is of mixed magnetic and electric dipole nature 76 while no emission could be detected around 670 nm corresponding to the 4 G 5/2 -6 H 7/2 transition which is mostly of electric dipolar nature (ED). In all, the emission shape is consistent with the substitution of Sn by Sm in the inversion C 2h symmetry sites with no association with a nearby defect. Our results provide a first evidence of Sm emission assignable to the substitutional isolated C 2h center (equivalent to center I of Eu). The emission shape is quite similar to that of Sm doped into other hosts with inversion symmetry sites, such as Y 2 O 3 (S 6 /C 3i ) 76 and CeO 2 (O h ) 48 recently reported by some of us. In addition, similar to Eu center I 27 (Fig. 5), the Sm center I displays also a remarkable long lived emission decaying on hundreds of ms scale with an estimated decay time from the tail fitting of 80 ms. It appears thus that that the long-lived emission is not Eu specific, given that it belongs to Ln substituting for Sn in the C 2h sites (substitutional isolated center). It was previously suggested 27 that the persistent luminescence of Eu in SnO 2 may be associated with the co-existence of uniform and exponential distributions in trap depths. It should be noted that most of the persistent luminescence displayed by the lanthanide based phosphors is associated with ED and not to MD transitions 27,66 .
In addition to the emission assignable to center I, we could separate by use of a similar procedure used for Eu-SnO 2 , at least three additional centers characterized by narrow (~0.7 nm) and shorter-lived emission (average lifetime around 1 ms) and labelled as centers II-IV in Fig. 6. The separation of the excitation spectra and decays of centers II-IV of Sm was however more difficult than in the case of Eu, therefore we have included only one excitation spectrum in Fig. 6. The average lifetimes of Sm centers II-IV is roughly estimated between 0.5 and 1ms. A broad and strongly distorted emission was observed and readily assigned to Sm surface center (corresponding center V of Eu). In addition to centers I and V, corresponding to isolated substitutional and surface centers of Eu and Sm, respectively, we can tentatively suggest a common nature also for centers IV of Eu and Sm. Both centers IV of Eu and Sm display a relative strong intensity of MD transition ( 5 D 0 -7 F 1 for Eu and 4 G 5/2 -6 H 5/2 for Sm) suggesting that the coordination around Eu and Sm centers is slightly deviated from the inversion C 2h symmetry. Taken together, the luminescence properties of Eu and Sm in SnO 2 are consistent with a common distribution, irrespective of the synthesis method, that includes substitutional isolated, associates with defects and surface centers.
We have also investigated the luminescence of Pr, Tb and Dy-SnO 2 samples in the as-synthetized state and after calcination at 400 and 700 °C. Selected XRD patterns and DR-UV/Vis spectra as well as luminescence spectra of Pr, Tb and Dy-SnO 2 are gathered in Figures S8 and S9, respectively. To the best of our knowledge, there is no literature report on Pr emission in SnO 2 , while for Tb 44 or Dy 45,46 doped SnO 2 the emissions appear to be related to surface location rather than substitutional doping. As a general observation, we note that the intensities of the Pr, Tb and Dy related emissions were poor, precluding advanced investigation. The emissions relate possibly to 1 D 2 level of Pr and definitely to 5 D 4 or 4 F 9/2 levels of Tb or Dy. To decide whether the observed emission may relate to isolated substitutional doping, we included, in Figure S9 the emission spectra of Tb and Dy in the inversion sites of Y 2 O 3 (S 6 /C 3i ) and CeO 2 (O h ) reported recently by some of us 47,48 . It can be observed that the emission shapes of Tb and Dy depart strongly from those expected for an inversion symmetry. Besides, the emissions are broad (~13 nm for Dy and ~9 nm for Tb) and short-lived (up to few hundreds of microsec). We therefore concluded that, that in contrast to Eu and Sm, Pr, Tb and Dy in SnO 2 cannot substitute for Sn in the inversion lattice sites, preferring instead highly distorted environments, likely connected with the SnO 2 surface.
Particle size effect on the local structure. It is well established that, as the particle size decreases, the increasing number of surface and interface atoms generates stress/strain and concomitant structural perturbations 28 . The comparison of luminescence properties of small (mean particle size comparable to the bulk exciton Bohr radius) and large sized Eu-SnO 2 nanoparticles led to the conclusion that the Eu environment is strongly distorted in the small SnO 2 while it is crystalline in the larger nanoparticles 29 . We re-asses here the particle size effect on the local structure selecting Eu-SnO 2 particle sizes around 2 nm (as-synthetized) and 20 nm (calcined at 1000 °C). Figure 7 compares the steady-state emission spectra of the two Eu-SnO 2 nanoparticles (excited above bandgap around 300 nm) that evidence a strongly distorted emission for the smaller sized nanoparticles (magenta spectrum). This may be used as a spectroscopic evidence that the local environment gets more distorted with decreasing particle size. However, using the time gated (delayed) mode, a narrow, long-lived type emission with orders of magnitude less intensity than the broad emission could be extracted with the 2 nm sized particles. Except for the slightly different widths, the delayed emission spectrum measured with the 2 nm particle is quite similar to that measured with the 20 nm particle. Both emissions correspond to the substitutional isolated Eu center (center I) with similar values of emission peaks (588.24, 593.04 and 599.08 nm) and intensity ratio (1/0.76/0.59) suggesting similar atomic scale environments of the inner lattice Sn, irrespective of the particle size. The reported distortions likely arises from the generic "surface effect" which is responsible for a broad and short-lived luminescence 48 observed with wide range of doped nanoparticles, irrespective of host and luminescent activator.
We also highlight here that the advantage of using time-gated luminescence as a local technique over long-range XRD and medium range Raman techniques in detecting weak crystallinity features is obvious. In the Figs 1b and 2b only broad, non-crystalline patterns and surface phonon modes are observed for the as -synthetized Eu-SnO 2 making difficult to assess the substitutional doping effect.
Surface versus bulk doping in the photocatalytic behavior. Recent literature shows that the Ln dopants increase the photocatalytic activity of SnO 2 in the photo-oxidation of phenol under UV or solar simulated light 5 . Our photocatalytic measurements used here as an additional probe shows that pure, doped and impregnated Eu-SnO 2 exhibit catalytic activity in the oxidation of phenol (Fig. 8a-c, Figure S11) irrespective of the calcination temperature and UV versus Vis irradiation range (see also Figure S1 for the spectral excitation profiles). The conversion decreased with the increase of the temperature in opposition to reports on non-structured catalysts 15 being, however in line with textural analysis (surface and pore blockage by carbonaceous products and collapse of the structure by calcination (Table 1) and DR-UV-Vis measurements ( Figure S5). Decrease of the conversion efficiency with calcination temperature (reflected in the decrease of the surface area, increase of the crystallite sizes (Table 1) and slight increase in the band gap (Table 2)) are due to modification of defects population in both pure and Eu doped/impregnated SnO 2 . The advantage of the materials structuration is demonstrated by the higher catalytic conversions of doped Eu-SnO 2 ( Figure S11b) and the smaller photocatalytic conversions on the impregnated Eu-SnO 2 ( Figure S11c) compared to pure SnO 2 . Although Eu presence does not affect strongly the band gap (Table 2), its dispersion in the bulk is generating a higher photoactivity compared to pure and impregnated samples. The luminescence investigations evidenced a migration of the lattice Eu towards the surface with increase of calcination temperature from 700 to 1000 °C while the impregnated samples exhibit mainly the surface defects (surface center V, Fig. 5) that along with a slightly wider band gap is reflected in a lower activity compared with the doped samples. Although photocatalysis is a surface process, the photons can penetrate at a certain depth in the nanoparticle where the presence of bulk defects may be beneficial. The difference of only 0.2 eV in the band-gap is not enough to explain the increased conversion with doped Eu-SnO 2 compared to pure SnO 2 . Moreover, the impregnated Eu-SnO 2 exhibits a higher band-gap than pure SnO 2 or doped Eu-SnO 2 . On the other side, the activity of the impregnated samples increases with the increase in the temperature (pre-calcination and subsequent calcination) ( Figure S11c), that is in agreement with previous reports 15 . The activity under Vis irradiation (allowed by the spectral profile illustrated in Figure S10) exceeds by almost 300 and 50% those of doped and pure SnO 2 for the same (pre)calcination temperature of 400 °C. At 1000 °C, the Vis activity is comparatively weaker than at 400 °C which arises likely for the compromise of larger band gap (3.2 compared to 2.6 eV, Table 2) and reduction of bulk defects over surface ones (as Eu incorporation into the lattice is hindered by the higher pre-calcination temperature).
Figure 8d-f shows the evolution of the selectivity towards hydroquinone (HQ), catechol (Cat) and 1, 4-benzoquinone (BQ) intermediates 77,78 Under UV irradiation, the reaction with pure SnO 2 advanced producing both BQ and Cat (Fig. 8d) with no Cat being produced after one hour of irradiation. In concordance to the above comments concerning the decrease in the population of the defects and changes in the band gap, the sample calcined at 400 °C afforded HQ and BQ in the first hour of irradiation, Cat being produced only after two hours, the sample calcined at 700 °C, after one hour afforded mainly HQ, and calcination at 1000 °C led to HQ as major product (100% selectivity after one hour), Cat not being produced. In all the cases, the plots suggest a consecutive/ parallel oxidation process (Fig. 9). Doping SnO 2 with Eu ( Fig. 8e) changed the selectivity predominantly in the favor of HQ (see samples calcined at 700 and 1000 °C, respectively) following the same mechanism and preserving the effect of the calcination. However, the parallel reaction is slower in this case, with no Cat being produced. For the impregnated Eu-SnO 2 , a small amount of Cat is also detected (Fig. 8f) which is lower for the sample treated at 1000 °C than 400 °C following the same trend as pure SnO 2 .
For both pure and Eu-SnO 2 (doped or impregnated) samples, under Vis irradiation, the reaction stopped at hydroquinone (HQ) with complete selectivity (not shown). As the results showed, the presence of surface and bulk defects is beneficial for the photocatalytic performance of the samples. The neutral oxygen vacancies (Vo) in the structure of SnO 2 that are produced by elimination of water during the Sn(OH) 4 -SnO 2 transition 15 . Although the insertion of Eu is supposed to enhance the population of the defects, except for the samples calcined at high temperatures, the band-gap remained basically unchanged. According to the equations included in Fig. 9, phenol can be oxidized by the O 2 − species and OH radicals generated on the defected sites under the light irradiation. In all, the results described above suggest that the correlation of luminescence and photocatalytic properties, being both defects -related, can be used as an additional tool for improving the identification of bulk over surface defects in trivalent metal doped SnO 2 . Generally, the substitution of metal cations in oxide lattice by metal ions with different valence states disrupts the chemical bonds in the host oxide generating superficial active centers represented either by the oxygen atoms near the dopant or the dopant itself 79 . Low valence dopants such as the lanthanide ions, activate the near oxygen atoms, thus the doped host oxide is a better oxidant than the undoped one. The effect of higher valence dopants depends on the nature of the host oxide 79 . For irreducible oxides (e.g. Ta 2 O 5 ), the host surface oxygen atoms are less active and the oxidation takes place merely by the surface adsorbed and activated O 2 . When the host oxide is reducible (e.g. TiO 2 , CeO 2 , SnO 2 ), the dopant metal will donate electrons to the host cations (for the case of SnO 2 , reducing their valence as Sn4+ → Sn2+). For SnO 2 , in addition to the generation of active sites, the electronic control of energy levels in dopant clusters is involved in the activation of certain reactions by dopants 80 . On the other hand, the particle size of the doped metal oxide becomes smaller than the pure host oxide, as consequence of restriction of the motion of crystallites at the interaction on the boundaries between host and dopant crystallites 81 , as it can be seen in Table 2.

Conclusions
We have investigated pure and trivalent lanthanide doped SnO 2 during and at various calcination temperatures up to 1000 °C, by use of a suite of in situ/ex situ techniques and methods. It is found that, before calcination most Eu resides on nanoparticle surface while mild calcination up to 500 °C lead to a significant reduction of surface effects. Besides, the redistribution of dopants along with induced defects remain limited to the surface layer of 2-6 nm sized particles. At higher calcination temperatures (≥700 °C) a complex distribution as substitutional isolated, associates with bulk defects and surface centers was evidenced for both Eu and Sm. The manifestation of a remarkable long-lived luminescence decaying on the hundreds of ms scale for both ions suggested that such phenomenon is not lanthanide specific, being tentatively associated to interaction between traps and the substitutional isolated Ln center. In contrast to Eu and Sm, Pr, Tb and Dy lanthanide dopants distributed mostly on the surface of SnO 2 nanoparticles as judged by their characteristic emission. As both luminescence and photocatalytic properties are defects -related, their correlation proved effective in distinguishing the surface over bulk defects and confirming the temperature induced migration of dopants from the lattice sites towards the surface.

Materials and Methods
Synthesis. The synthesis of SnO 2 nanoparticles was carried out by a rapid hydrothermal and sol-gel methods. In the first approach, a quick introduction of one gram of solid tin tetraisopropoxide isopropanol solvate (Multivalent Co UK) into 50 ml of boiling Milli-Q water with subsequent refluxing of the obtained milky dispersion for 30 min is used 16 . The mixture was then left for precipitation overnight, the supernatant was removed by decantation and the solid residue was dried at room temperature in air overnight. The Ln (Sm, Eu) doped SnO 2 nanopowders were produced by the same procedure using instead of pure water a solution of stoichiometric amount of Ln(NO 3 ) 2 in 50 ml of water. Ln (1 wt%) was also deposed on pre-calcined SnO 2 (at 400 and 1000 °C, respectively) via the wetness impregnation with an aqueous solution of Ln(NO 3 ) 3 •6H 2 O (Alfa Aesar) followed by drying overnight at room temperature and calcination at 400 and 1000 °C, respectively. In the sol-gel method, Ln(Eu, Sm, Pr, Tb and Dy, 1 wt%) doped SnO 2 nanopowders were prepared by dissolving of 2 g stannous chloride dihydrate (SnCl 2 .2H2O) and the appropriate amount of dopant Ln(NO 3 ) 3 •6H 2 O in 100 ml distilled water. After complete dissolution, ammonia solution was added to the above solution by drop wise under stirring. The suspensions were left overnight for ageing, centrifuged and then dried at room temperature for 24 hours and 60 °C for 4 hours. The samples were calcined in air at the desired temperature (400, 700 and 1000 °C) with a rate of 10 °C/ min.
Characterization. Microbeam X-ray fluorescence (micro-XRF) spectrometry was performed on a custom-made instrument with an X-ray tube: Oxford Instruments, Apogee 5011, Mo target, focus spot ∼40 µm, max. high voltage −50 kV, max current −1 mA, Amptek X-123 complete X-Ray spectrometer with Si-PIN detector. The key element of the micro-XRF instrument is an X-ray policapillary minilens (IfG-Institute for Scientific Instruments) which provides a focal spot size on the sample of 15-20 µm. X-Ray Diffraction (XRD) measurements were recorded on a Shimadzu XRD-7000 diffractometer using Cu Kα radiation (λ = 1.5418 Å, 40 kV, 40 mA) at a scanning speed of 2 degrees min −1 in the 10-90 degrees 2θ range. In-situ experiments were carried out in a cell accessory in the 50-1000 °C temperature range. The heating rate was 10 °C/min and the samples were kept for 10 min at the set temperature before collecting the diffractograms. The Raman spectra were acquired in the extended spectral region from 150 to 4000 cm −1 . Raman analysis was carried out with a Horiba JobinYvon -Labram HR UV-Visible-NIR Raman Microscope Spectrometer, at 514 nm. In situ Raman spectra were recorded using the same apparatus equipped with an in-situ cell, until 600 °C. The heating rate was 10 °C/min and the sample was kept for 10 min at each temperature before collecting the spectra. Diffuse reflectance optical (DR-UV-Vis) spectra were recorded at room temperature on a Analytik Jena Specord 250 spectrophotometer with an integrating sphere for reflectance measurements and MgO as the reflectance standard. DR-UV-Vis spectra of the materials were recorded in reflectance units and were transformed in Kubelka-Munk remission function F(R). The estimation of the band-gap energy was made using the E = hc/λ (eV) formula, where h is the Planck constant (4.135 • 10 −15 eVs), c is the speed of the light (3 • 10 8 ms −1 ), and λ the absorption threshold value for each sample determined from the spectra. The value of E g (ie the band gap of the semiconductor) was derived from the spectra by plotting (F(R) • hν) 2 against hν 18 . Diffuse Reflectance Fourier Transform Infrared (DRIFT) spectra were measured with a Thermo Electron Nicolet 4700 FTIR spectrometer with a Smart Accessory for diffuse reflectance measurements. The IR spectra were scanned in the region of 4000-400 cm −1 at the resolution of 4 cm −1 . The final spectra are an accumulation of 400 scans. Transmission electron microscopy (TEM) investigations were made using a Tecnai G 2 T20 microscope (FEI TM , Holland) with an accelerating voltage of 200 kV. Scanning electron microscopy (SEM) studies were carried out with Hitachi TM 1000 tabletop environmental microscope equipped with μ-DeX EDS detector system. Luminescence measurements. The photoluminescence (PL) measurements were carried out using a Fluoromax 4 spectrofluorometer (Horiba) operated in both the fluorescence and the phosphorescence mode. The repetition rate of the xenon flash lamp was 25 Hz and 1 Hz for persistent luminescence decay, the integration window varied between 0.1 and 0.5 s, the delay after flash varied between 0.03 and 500 ms, and up to 30 flashes were accumulated per data point. For excitation spectra, the slits were varied from 5 to 29 nm while in emission the slits were varied from 1 nm to 3 nm. The PL decays were measured by using the "decay by delay" feature of the phosphorescence mode. The average decay time was calculated as integrated area of normalized decay. Time resolved emission spectra were recorded using a wavelength tunable (from 210 to 2300 nm) NT340 Series EKSPLA OPO (Optical Parametric Oscillator) operated at 10 Hz as excitation light source. The tunable wavelength laser has a narrow linewidth <4 cm −1 with scanning step varying from 0.05 to 0.1 nm. As detection system, an intensified CCD (iCCD) camera (Andor DH720) coupled to a spectrograph (Shamrock 303i, Andor) was used. The time-resolved emission spectra were collected in the spectral range of 450 nm < λ em < 850 nm using the box car technique. The luminescence measurements were recorded at low-temperature (80 K) by use of FL-1013 Liquid Nitrogen (LN) dewar assembly (Horiba) and at 300 K. Data availability statement. The datasets generated during and/or analyzed during the current study are available from the corresponding author (CT) on reasonable request.