Validity of Valence Estimation of Dopants in Glasses using XANES Analysis

X-ray absorption near edge structure (XANES) measurement is one of the most powerful tools for the evaluation of a cation valence state. XANES measurement is sometimes the only available technique for the evaluation of the valence state of a dopant cation, which often occurs in phosphor materials. The validity of the core excitation process should be examined as a basis for understanding the applicability of this technique. Here, we demonstrate the validity of valence estimation of tin in oxide glasses, using Sn K-edge and L-edge XANES spectra, and compare the results with 119Sn Mössbauer analysis. The results of Sn K-edge XANES spectra analysis reveal that this approach cannot evaluate the actual valence state. On the contrary, in LII-edge absorption whose transition is 2p1/2-d, the change of the white line corresponds to the change of the valence state of tin, which is calculated from the 119Sn Mössbauer spectra. Among several analytical approaches, valence evaluation using the peak area, such as the absorption edge energy E 0 at the fractions of the edge step or E 0 at the zero of the second derivative, is better. The observed findings suggest that the valence state of a heavy element in amorphous materials should be discussed using several different definitions with error bars, even though L-edge XANES analyses are used.

in a wide k range, which is quite different from the L-edge analysis, in which the EXAFS region is restricted by each L-edge. However, the observed change in the absorption edge energy E 0 of the K-edge may be too small relative to the resolution of the measurement to determine the origin, especially in heavier cations in amorphous materials [15][16][17] . This is due to the ambiguity of the s-p transition in the K-edge absorption, due to the heavy atom effect. Although we can measure both K-and L-edge XAFS spectra, there is no clear metric for determining the difference in the valence estimation between the K-edge and L-edge, especially in glass materials. Considering the accuracy of each XAFS measurement, alternative methods should be considered in the evaluation.
In this report, we focus on valence estimation via conventional XANES analysis using a Sn target element. We selected Sn because the valence state can be also evaluated using 119 Sn Mössbauer spectroscopy 16,[18][19][20][21] . Although Mössbauer spectroscopy is a powerful analysis method for estimating the coordination state using isomer shift, the necessary radiation sources for 119 Sn Mössbauer spectroscopy are not readily available. Recently, our group has demonstrated the photoluminescence of RE-free glass phosphors containing the Sn 2+ centre, which is an ns 2 -type emission centre that exhibits parity-allowed excitation ( 1 S 0 → 1 P 1 ) 16,17,[22][23][24][25][26][27][28][29][30][31] . The Sn 2+ centre in oxide glasses exhibit a high UV-excited emission, comparable to that of crystal phosphors such as MgWO 4 16,17,26-31 . However, in Sr-containing materials, since the absorption of 119 Sn Mössbauer γ-rays of the Sr cation is larger than that of the lighter cations, it is difficult to estimate the Sn 2+ /Sn 4+ ratio in Sr-containing materials using 119 Sn Mössbauer spectroscopy 17 . Since the Sr cation is often used as a key component in various phosphors 32,33 , the ability to examine their valence states via an alternate approach such as XANES, is potentially important.
In this study, we measured the valence state of Sn using both K-and L-edge XAFS analyses, as well as 119 Sn Mössbauer analysis. Our aim was to examine the relationship between the valence state of tin calculated using 119 Sn Mössbauer spectroscopy, Sn K-and L-edge XANES analyses. Glasses with different Sn 2+ /Sn 4+ ratios were prepared by tuning both the preparation atmospheres, viz. Ar and air, and the starting materials, viz. SnO and SnO 2 . Based on previous reports on Sn-doped oxide glasses 16,30 , two kinds of base glasses were selected: 1SnO α -60ZnO-40P 2 O 5 and 1SnO β -60ZnO-40B 2 O 3 (in molar ratio), denoted by SZP and SZB, respectively.
The SZP and SZB glasses were colourless, transparent and independent of the melting atmosphere. Figure 1 shows the glass transition temperature T g of the SZP glasses with different Sn concentrations, melted in Ar and air. Although T g decreases with an increase in the amount of Sn in both cases, the rate at which they decrease varies. The Sn-doped ZnO-P 2 O 5 glasses melted in Ar, produced a steeper slope compared with glasses melted in air. It has been reported that the Sn 2+ species induce a larger decrease in T g than the Sn 4+ species 34,35 . In other words, the value of T g may reflect the valence state of Sn, and the greater the Sn 4+ ratio, the greater the increase in T g . Since this difference suggests the oxidation of the SnO species, we can conclude that several Sn 2+ species are oxidized during air melting. Figure 2a shows the 119 Sn Mössbauer spectra at room temperature for the SZP glasses melted in Ar and air atmospheres. The starting material for tin was Sn(II)O. The peak at 2-4 mm s −1 corresponds to the Sn 2+ species, whereas the peak at 0 mm s −1 corresponds to the Sn 4+ species [18][19][20] . This figure indicates that most of the Sn in the glass melted in air, exists as Sn 4+ , whereas Sn 4+ was not observed in the glass melted in the Ar atmosphere. After peak deconvolution, the amounts of Sn 2+ in the SZP glass melted in air and in Ar were calculated as 14% and ~100%, respectively, ignoring the difference in the recoilless fraction between the Sn 4+ and Sn 2+ sites. This suggests that some of the Sn 2+ species were oxidized into Sn 4+ during air melting, which is also indicated by the change in T g values. Figure 2b shows the Sn K-edge XANES spectra of the SZP glasses melted in the Ar and air atmospheres. The spectra of SnO and SnO 2 are also shown for comparison. Since a higher absorption edge indicates a higher oxidation state of the cation, we take the absorption edge energy E 0 , to be the energy at the zero-crossing of the 2nd derivative. The E 0 of the SZP glass prepared in air is lower than that of the glass melted in Ar. This suggests that the amount of Sn 2+ in the former is higher than that of the latter. In order to check for any inconsistency between the Sn 2+ ratio estimated from E 0 and that calculated from the Mössbauer spectra, we prepared several glass samples that were melted in Ar and air.  Figure 1 shows the Sn K-edge XANES spectra of SZP glasses with different Sn concentrations, melted in Ar and air, with Sn-foil, SnO, and SnO 2 also shown for reference. Supplementary Table 1 lists the ΔE 0 values obtained by subtracting the E 0 value of the Sn-foil (Suppl. Figure 1) from the SZP glasses containing different amounts of Sn. From these data, it is observed that the E 0 of the SZP glass prepared in air is lower than that of SnO. This indicates that there is a difference between the real valence state of Sn obtained from the 119 Sn Mössbauer spectra, and the evaluated valence state of Sn from the K-edge XANES spectra. Since the measurement resolution is ΔE/E ~6 × 10 −5 , a difference of less than 1.75 eV is insignificant. Therefore, a quantitative analysis of the Sn 2+ ratio from the E 0 of K-edge XANES spectra will be difficult. However, the peak height of the steep peak near an absorption edge, also-called the 'white line' , is sometimes used for the evaluation of a valence state 35,36 . If we use the white line height located at approximately 29.2 keV, the peak height of the SZP glass prepared in Ar is comparable to that prepared in air, which suggests that the valence estimation using the K-edge peak height of the white line will also be difficult. Based on the analysis of the 119 Sn Mössbauer spectra (Fig. 2a), we can conclude that it is difficult to determine the valence state of tin using the E 0 values or the peak height of the white line calculated from K-edge XANES analysis. Figure 2c shows the Sn L II -edge XANES spectra of the SZP glasses, along with the spectra for SnO and SnO 2 . The spectrum of the SZP glass prepared in Ar is similar to that of SnO, while the spectrum of the glass prepared in air is similar to that of SnO 2 , indicating that each preparation atmosphere has a clear effect on the L II -edge XANES spectra. It is notable that the peak heights of these glasses are lower than those of the references. Since the observed white line is affected by the coordination symmetry, Sn 2+ in the SZP glass has a more disordered structure compared with the SnO crystal. This suggests that the valence state of tin can be evaluated from the L II -edge XANES spectra, which is indicative of the 2p 1/2 -d energy transition. In contrast to the K-edge XANES spectra whose s-p transition path becomes more obscure with an increase in the atomic number, the observed difference originates from the local coordination state. Considering the fraction of Sn cations compared to the total cation count (1/141), we assume that there is no structural interaction between the Sn cations, and that they are homogeneously dispersed as isolated cations in the zinc phosphate network. Therefore, we conclude that L II -edge XANES spectra are suitable for the evaluation of the valence state of tin (dopant) in amorphous glasses.
We then prepared different SZP and SZB glasses using the starting material SnO 2 in Ar and air atmospheres, in order to tailor different Sn 2+ /Sn 4+ ratios. Figure 3a and b show the 119 Sn Mössbauer spectra of the SZB and SZP glasses prepared under different conditions. The isomer shifts of Sn 2+ and Sn 4+ in SZB glasses are different from those in SZP glasses, which is due to difference in the local coordination states. These Mössbauer spectra confirm that the valence state of tin in the SZP glass is different from that in the SZB glass, even though the preparation conditions are the same. The Sn L II -edge XANES spectra of each composition are shown in Fig. 3c and d. The white line intensity of the SZP glass prepared in Ar, whose Sn 2+ ratio is almost 100%, is the lowest among these glasses. Conversely, the white line intensity of the SZB glass prepared in air is the highest. In addition, the peak energy of the white line shifts to the higher-energy side upon air melting, which corresponds to an energy shift in the absorption edge due to oxidation. Therefore, the white line intensities are correlated with the Sn 4+ ratios calculated from Fig. 3a and b. Figure 4a shows the Sn L II -edge XANES spectra of the SZB and SZP glasses prepared in Ar (solid lines) and in air (dashed lines). The spectra of SnO and SnO 2 are also depicted as references. For evaluation of the valence state of cations using the XANES technique, several definitions are conventionally used: E 0 energy at the fractions of the edge step, E 0 energy at the zero of the second derivative, and the peak area of a species. Figure 4b shows the relationship between E 0 , which is defined as a fraction of the edge step, and the Sn 4+ /(Sn 2+ +Sn 4+ ) ratio of the glasses. Although the 119 Sn Mössbauer spectra suggest the existence of Sn 2+ and/or Sn 4+ , a ZP glass exhibits a lower E 0 energy than SnO, whereas a ZB glass exhibits a higher E 0 energy than SnO 2 . This clearly indicates that a simple signal convolution of SnO and SnO 2 is unadoptable for evaluation of Sn cation in glass materials. Such a difference is observed in the E 0 energy at the zero of the second derivative of each sample. Supplementary  Figure 2 shows the relationship between E 0 , which are defined as the zero of the second derivative, and the Sn 4+ / (Sn 2+ +Sn 4+ ) ratio of the glasses. The deviation from the linearity between the valence state and E 0 energy is observed to increase. The valence estimation using the zero of the second derivative, is therefore worse compared to that using the E 0 energy at the fractions of the edge step.
Considering the aforementioned results, we use the peak area for evaluation of valence state of tin. As previously indicated, the Sn 2+ ratios in the SZP glasses are almost 100%, which was confirmed by 119 Sn Mössbauer spectroscopy (see Fig. 3d). Using the SZP glass melted in Ar as a standard, the spectral changes from the Sn 2+   Figure 5a shows the differential L II -edge Sn XANES spectra of the SZP and SZB glasses, which were prepared under different conditions. These spectra were obtained after subtraction of the normalized XAFS spectra of the SZP glass prepared in Ar.
Supplementary Figure 3 shows the L II -edge Sn XANES spectra of the SZP glasses and the differential spectra. Although these peak energies are different because of the local coordination state of the Sn cation, the differential peak height can be fitted with a Gaussian function. Figure 5b shows the relationship between the differential peak height at the white line peak, and the Sn 4+ /(Sn 2+ +Sn 4+ ) ratio in the SZP and SZB glasses. A linear relationship that appears to be independent of the glass composition was observed. Considering the precision of the 119 Sn Mössbauer analysis (±2%), we can conclude that these spectra are adequate for the evaluation of the Sn 2+ /Sn 4+ ratio. Supplementary Figure 4 shows the relationship between the peak height of the pre-edge region (~4.165 keV) and the Sn 4+ /(Sn 2+ +Sn 4+ ) ratio of the Sn-doped glasses. In this region, the relationship is non-linear, although they are correlated, and we can conclude that the Sn L II -edge measurement is effective in evaluating the valence state of tin.
It has been reported that the L-edge XANES of heavy elements is useful in quantifying their valence states [8][9][10][11][12] , and we have also demonstrated valence estimation using L III -edge XANES 37 . However, as previously indicated, there is no available report on the difference in valence estimation between K-edge and L-edge analyses in glass materials. Since it is expected that this difference will be affected by the glass system, i.e. the local coordination state of the cation and electrons, we emphasize that the present approach will contribute to a deeper understanding of the local coordination of useful activators in materials science.
In summary, we have examined the correlation between the valence state of Sn in oxide glasses using 119 Sn Mössbauer spectra as well as Sn K-edge and L II -edge XANES analyses. We found that it is difficult to evaluate the valence state of Sn using the K-edge XANES analysis because of an obscure s-p transition. In addition, it is also difficult to determine the valence state from the E 0 value in the L II -edge XANES analysis. Conversely, it was determined that the peak height of the white line in L II -edge XANES is an indicator for the local coordination state, which is confirmed by the 119 Sn Mössbauer spectra. Although Sn 2+ exists in SZP glasses, the white line is broadened compared with the standard SnO, suggesting that the coordination state of Sn 2+ is not equal, but similar to that of Sn 2+ in SnO crystals. The peak energy of the white line shift also depends on the actual Sn 2+ ratio in the glasses, whereas the peak height is independent of the chemical composition of the host glass. Here, we have demonstrated that a valence estimation strongly depends on the estimation approach, even for L-edge analysis. Most industrial glass plates made using the "float method" contain tin at the surface, and the valence states affect the physical properties of the industrial products. The valence state of tin in transparent conducting films is also of significance. Therefore, the present findings are noteworthy, particularly for materials containing tin as a key element. Although the present data are only concerned with tin, we wish to emphasize that our findings are adaptable to other heavy metal cation-doped materials 15 , which is important for a deeper understanding of materials science.

Experimental Section
Sample Preparation. The Sn-doped 60ZnO-40P 2 O 5 (SZP) and 60ZnO-40B 2 O 3 (SZB) glasses were prepared according to a conventional melt-quenching method by employing a platinum crucible 38,39 . The mixture of ZnO and (NH 4 ) 2 HPO 4 was initially calcined at 800 °C for 3 h using a Pt crucible in the ambient atmosphere. After treatment, the calcined matrix was mixed with SnO and/or SnO 2 and melted in an electric furnace at 1100 °C for 30 min in ambient or Ar (5 N) atmosphere. In the case of inert melting, the mixture was set in the atmosphere-controlled electric furnace at room temperature. It took 2 h to heat up from r.t. to 1100 °C, and the temperature was fixed at this value for 30 min. Before initiating the heating, an Ar purge process was performed in the furnace tube. The air in the tube was removed using a vacuum pump, and subsequently purged using 5 N Ar gas. This purging process was performed three times. The glass melt was quenched on a stainless-steel plate at 200 °C and then annealed at T g , which was measured by a differential thermal analysis (DTA) for 1 h.
Characterization. T g was determined using a DTA system operating at a heating rate of 10 °C/min, using a TG8120 (Rigaku, Japan). The 119 Sn Mössbauer spectra, i.e. the absorption spectra of the γ-rays by the 119 Sn nuclei in the samples, were measured using conventional transmission geometry using a Ca 119m SnO 3 source at room temperature. The energy of the γ-rays from the source were modulated by the Doppler effect using a velocity transducer with a constant acceleration mode, and the abscissae of the spectra were identified with the units of the Doppler velocity, as in the literature 20 . The valence states of the Sn atoms, which were detected as the peak positions in the 119 Sn Mössbauer spectra 20 , were deduced by fitting the measured spectra using the commercial software Normos (made by R. A. Brand, commercially available from WissEl GmbH).
The Sn K-edge (29.3 keV) and L II -edge (4.17 keV) of the XAFS spectra were measured at the BL01B1 beamline of the SPring-8 (Hyogo, Japan). The storage ring energy was operated at 8 GeV, with a typical current of 100 mA. The Sn K-edge XAFS measurements were carried out using a Si (311) double-crystal monochromator in the transmission mode (Quick Scan method). Conversely, the Sn L II -edge XAFS measurements were carried out using a Si (111) double-crystal monochromator in the fluorescence mode using 19-SSD at r.t. The XAFS data of Sn-foil, SnO, and SnO 2 were also collected under the same conditions.