- Email: [email protected]
Redox profile of the glass surface
Journal of Non-Crystalline Solids 357 (2011) 3200–3206
Contents lists available at ScienceDirect
Journal of Non-Crystalline Solids j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j n o n c r y s o l
Redox profile of the glass surface A.-M. Flank a, P. Lagarde a,⁎, J. Jupille b, H. Montigaud c a b c
Synchrotron SOLEIL, l'Orme des Merisiers, BP 48 91192 Gif/Yvette cedex, France INSP, CNRS et Université Paris 6, Campus de Boucicaut, 140 rue de Lourmel 75015 Paris, France Saint-Gobain Recherche 39, quai Lucien Lefranc, BP 135 93303 Aubervilliers Cedex, France
a r t i c l e
i n f o
Article history: Received 16 February 2011 Received in revised form 30 March 2011 Available online 3 June 2011 Keywords: Glass; Surface of the glass; Redox profiles; Float glass
a b s t r a c t The redox profiles of tin, iron and sulfur at the float glass surface were determined on purposely cut samples by Electron Probe Micro Analysis (EPMA), X-ray Fluorescence mapping and X-ray Absorption Spectroscopy (XAS) at the micron scale. Going inward from the surface to the bulk, it was observed that (features do not depend on the glass thickness (holding time) though they extend over depths that vary from ca. 25 to 50 μm): (i) after a diffusion-driven decrease and prior to vanishing, the tin concentration passes through a local maximum (the tin hump), where stannous ions, which dominate the shallow layers, switch to stannic ions; (ii) the iron concentration decreases, passes through a minimum at the tin hump where iron is in the more reduced form (lowest Fe 3+/Fe 2+ ratio); it then increases and, after a hump which appears as a chemical echo of the tin hump, it reaches the bulk value; (iii) the concentration of sulfur increases up to reach the bulk concentration beyond the tin hump region. In a mixed S 6+/S2− form at the surface, sulfur is only in sulfate form in the bulk. In the case under study, the iron concentration is much too low to balance the redox reaction Sn 2+ → Sn4+ that occurs at the tin hump. Sulfur is shown to play the role usually attributed to iron, according to the reaction 4Sn
2þ
6þ
4þ
þ S →4Sn
2−
þS
The occurrence of that reaction is supported by the appearance of sulfide S 2− in the tin hump region with an appropriate concentration profile of a much stronger S2−/S 6+ ratio on the tin side than on the atmosphere side of the float glass. The conclusions drawn herein likely apply to the many cases in which the glass composition is similar as that encountered herein. © 2011 Elsevier B.V. All rights reserved.
1. Introduction The first step of the float process consists in spreading the glass melt onto liquid tin to take advantage of the balance between surface and interface tensions to produce flat glass sheets. Those conditions, that are quite suitable to produce glazings for common use, are not adapted regarding new applications which raise the level of demand required for the production of float glass, such as the manufacture of supports for electronic display elements that needs nearly atomically smooth substrates [1] and the control of the nanomechanical properties of glass surfaces [2,3]. The control at the microscopic level of the glass surface requires a detailed understanding of the redox reactions involved in the float process. The two faces of the glass ribbon are different. To prevent the oxidation, the whole part of the process which involves the tin bath is
⁎ Corresponding author. E-mail address: [email protected] (P. Lagarde). 0022-3093/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2011.03.046
run in reducing conditions by circulating a gas mixture comprising nitrogen and hydrogen. Due to such an environment, the chemical composition of the skin of float glass differs from that of bulk over a few tens of microns with different behaviors on the atmosphere side and on the tin side. Complex concentration profiles result from interdiffusion and redox reactions at the interface between tin and glass melt [1]. Researches focus on the concentrations and profile of iron species because that element is stressed to be central [1,3–6] in the formation of the tin satellite peak that appears at a depth of ~5–20 μm from the glass surface. The so-called tin hump was first identified by Sieger [3]. It has been then demonstrated by Mossbauer spectroscopy that the shallower part of the tin profile mostly involves Sn2+ ions and that the tin hump is dominated by Sn4+ ions [5,7,8]. On the basis of a simulation, Wang et al. suggest that the tin hump arises from an easy diffusion of Sn4+[9]. At variance, Cook and Cooper [1] argue that the diffusion of stannic ions that act as glass modifiers is much easier than that of stannous ions that behave as glass formers. These authors and Frischat et al. [5,6] assign the occurrence of the tin hump to redox reactions. They assume that the formation of stannic ions within the
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
glass is favored by the presence of ferric ions via an ion-exchange process: 2þ
Sn
3þ
4þ
þ 2Fe →Sn
2þ
þ 2Fe
ð1Þ
A maximum in iron concentration is also observed at the extentend of the tin profile. It is associated to the inward diffusion of the then produced ferrous ions [1]. There is also some suggestion that the formation of Sn 4+ could involve sulfur [10,4,5], although no evidence was ever provided. To date, direct measurements of the redox profile of the glass surface are lacking. The objective of the present work is to determine concentration profiles of tin, sulfur and iron at the vicinity of a float glass surface by examining cuts of glass plates by microanalysis. Electron Probe Micro Analysis (EPMA) was used to determine quantitatively the concentration as a function of the depth. Micro X-ray Fluorescence mapping and X-ray Absorption micro-Spectroscopy (XAS) allowed a characterization of the oxidation states and their spatial distribution. These techniques have a comparable resolution of about 1 μm. Therefore, they were suited to analyze concentration along the profiles under study that extend over tens of microns. Moreover X-ray spectroscopy is of relevance to determine the chemical states of tin, sulfur and iron in glasses [11]. It has been chosen to work on common float samples with iron (Fe2O3) and sulfur (SO3) bulk concentrations of 0.1 and 0.3 wt.%, respectively. It will appear during the discussion that this choice allows to pinpoint the respective contributions of these two species to the redox profile of the float glass surface.
2. Experimental Samples were float glass plates with different thicknesses: 3, 4, 6 and 10 mm. Glass surfaces were analyzed by sampling both the bath side and the atmosphere side. In order to study concentration profiles, the bath sides of two glass plates were glued together (Fig. 1), and this cross-section sample was prepared by grinding with SiC abrasive papers and then polishing with diamond (3 and 1 μm size). A quantitative analysis of the concentration was achieved on the 10 mm thick sample by a Cameca SX50 Electron Probe Micro Analyser employing four different wavelength-dispersive spectrometers, at Saint-Gobain Recherche. Such elemental analysis was realized on the bath surface and the bulk (from the cross section) after plating them with a thin graphite conducting layer. The experiments were performed with an accelerating voltage of 15 kV and a beam intensity of 10 or 150 nA according to the element. Five different spots were analyzed in every case. The composition of this glass studied here is shown in Table 1, together with that of the glass CNF 0.2 studied already by Farges et al. [12] (see below).
Sample raster
Sn or atmosphere faces
Fluorescence
X-rays (3 x 3 µm2)
Image
Fig. 1. Geometry of the experiments.
Glue
(4 µm)
3201
Table 1 Composition of the float glass used herein (weight per cent of oxides). Error bars are maximized by the last digit. Comparison with the CNF 0.2 composition analyzed by Farges et al. [12].
This work CNF 0.2
SiO2
Na20
MgO
CaO
Al2O3
Fe2O3
SO3
71.9 72.1
14.0 13.8
3.9 –
8.7 13.9
0.7 –
0.09 0.2
0.26 –
X-ray experiments have been done at the LUCIA beamline [13], installed on the Swiss synchrotron facility SLS (Villigen). The monochromator is equipped with a pair of Si(111), and the spot size was set to 3 × 3 μm 2 at all energies of interest. By using the two classical techniques of the total electron yield (TEY) and fluorescence yield (FY), we could probe two different depths into the sample, typically 0.5 μm in TEY and 5 μm by FY. The TEY signal is obtained by a measure of the sample drain current, while the FY is detected by a monoelement (10 mm 2) silicon drift diode. The analysis of the oxydation state of the different elements was be done comparing the near edge spectra (XANES) to those of reference compounds whose valency is well known. XANES of sulfur in several well characterized compounds were purposely recorded. Tin spectra were compared with published data [14]. Finally, the oxydation state of iron was determined by an analysis of the pre-peaks [12,15,16] as explained below. To avoid any artifact from the self-absorption of the fluorescence signal, the common surface of the glass was set horizontal, and this common surface could be precisely localized by examining the silicon mapping since this element was expected to exhibit a constant concentration throughout the sample (Fig. 1). Given the flux on the sample of about 5 × 10 10 photons/s, we could be concerned by possible radiation damage or photoreduction processes, mostly on sulfur which is well known to be a very sensitive element. In order to check that, we did successive experiments on the same point. In the case of tin and iron, the spectra remains identical and therefore we believe that radiation damages should be of a minor importance in that case. At the contrary, and this will be discussed in Section 3.4, some of the sulfur features could be the result of the irradiation, as it has been also pointed out by Wilke et. al. [17].
3. Results 3.1. Mapping of the different elements The concentration profiles at the surface of the 10 mm thick float glass plate were first examined by EPMA. Tin penetrates in the glass over a depth of ≈40 μm. Going inward, its concentration decreases steeply prior to show a buried peak — the so-called tin hump — at 10–20 μm below the surface. The concentration of iron is peaking at the extreme surface. Then, after a depleted zone with a minimum just before the tin hump, it increases to peak again as the Sn concentration vanishes, prior to reach its bulk value. The sulfur concentration, which is at its lowest level at the glass surface, increases progressively to level off at a depth of ≈15 μm which coincides with the Sn hump. Its bulk value is about three times higher than the surface value. Sn, Fe and S profiles that are observed herein are consistent with the commonly accepted picture [1,3]. Fig. 2 shows the images obtained on the 10 mm thick sample for the different elements analyzed in this study, with two Sn faces stuck together as described previously. The silicon and iron maps have been obtained with an incident energy of 7130 eV, the tin map has been made at 3990 eV in order to avoid the contribution of calcium present in the glass, and the sulfur one has been recorded at 2500 eV in order to maximize the fluorescence signal. On the silicon map, which must be uniform, we see the signature of the glue (blue and green zones) which amounts about
3202
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
Fig. 2. Top: Cartographies of the silicon, tin, iron and sulfur concentrations, as recorded at the Si K-edge, S K-edge, Fe K-edge and Sn L-edge, respectively for the 10 mm thick glass. The dashed area corresponds to the glue between the two faces. The depth is indicated on the vertical scale. Bottom: Concentration of the different elements as a function of the distance from the surface, deduced from line profile analysis of the above maps (open symbols) and superimposed with the electron microprobe results (filled symbols).
8 μm and this zone must be excluded from the concentration analysis of the other elements. The concentrations of Sn, Fe and S were analyzed as a function of depth on each side of the junction between the two glass plates, and the results are also shown in Fig. 2. The values of concentrations at a given depth are normalized by the silicon concentration recorded at the same depth. This silicon concentration appears to be constant throughout the sample which is true within a few percent [3,18]. The variations of the other elements appear fully in line with the features
already revealed by microprobe analysis. The buried tin hump is observed at around 10 μm from the surface of the 10 mm sample. Iron concentration peaks at the surface, goes through a flat minimum ca 15 μm and then increases progressively till it shows a maximum as the tin concentration vanishes. Sulfur concentration is at minimum at the surface. It increases continuously in the inward direction up to reach its bulk value. Fig. 2 shows also the results of the electron microprobe. In the absence of absolute calibration, the concentration profiles that are derived from microcartography are rescaled with respect to the
Fig. 3. The same analysis for the 3 mm thick glass as in Fig. 2. The evolution of the concentrations with the distance to the surface shows that, in this case, the iron bump appears closer to the surface. Note that the concentration scales have no absolute meaning and are different from one element to another one.
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
3203
3.2.1. Tin A series of Sn L edge spectra, collected every 2 μm from the surface (excluding the glue region) is shown in Fig. 4. The maximum is seen to shift from 3954 eV for XAS spectra recorded at the sample surface to 3961 eV for spectra recorded at depth higher than 5–10 μm. Apart from a shift of 4 eV attributed to a different monochromator calibration, and a smoothing of the overall structures due to the glassy state of our samples, those spectra are in line with those recorded by Liu et al. on reference compounds [14]. In particular the 7 eV blue shift of the main maximum is attributed to the switch from SnO to SnO2. The oxidation of Sn2+ to Sn4+ between 5 and 10 μm coincides with the onset of the tin hump (seen in Fig. 2). The present data support the widely accepted picture of a strong contribution of Sn4+ to the buried tin peak [1,4–6,19]. However, the relative concentration of stannous and stannic ions likely depends on the float sample under study. Following a Mössbauer analysis of tin species at a float glass surface, Williams et al. reported a percentage contribution of Sn4+ close to zero at the vicinity of the glass surface and of about 1/3 in the hump region [7]. By means of cathodoluminescence and measuring the refractive index profile, Townsend et al. found also a maximum in Sn4+ concentration within the glass at the interface with the bath [19].
3.2.2. Iron Iron is expected to exhibit different valence states throughout the surface region of the glass. At the Fe K edge absorption spectrum, the degree of oxidation of iron and its speciation are given by the pre-edge features and the fine structure of the XANES [12,15,16]. The energy separation between features associated with Fe2+ and Fe3+ is about 1.5 keV [12,15,16,20]. Despite this sizeable energy shift, the identification of the iron species is uneasy because, beyond the spectral shape, the intensity of the pre-edge features depends strongly on the local structural environment. Losses in centrosymmetry result in an increase of the intensity so that contributions from Fe ions in tetrahedral position are about four times more intense than contributions from ions in octahedral sites [12,15,16]. Fe K edge spectra recorded through the surface region by XAS microscopy are shown in Fig. 5. To improve the signal to noise ratio and to avoid as much as possible radiation effects (see below) the as recorded Fe K edge spectra were added over three zones which correspond to the very surface (before the tin hump, 0 to 8 μm), the tin hump (8 to 12 μm) and the final iron hump (15 to 30 μm), respectively. A zoom of the regions of the pre-edge is presented in Fig. 5. The composition of the float glass that is studied herein is very close to the CNF glasses that were analyzed by Farges et al., in particular the CNF 0.2 sample [12], the chemical composition of which is given in Table 1. The pre-edge profile recorded around the tin hump region shows an increase of the 2+ component with respect to the two other spectra, which indicates a higher Fe2+ content around 10 μm from the surface. Moreover, in the XAS spectrum associated to the 10 μm region, the edge itself is located at a lower energy than that recorded outside this region, which is also a signature of a lower valence state. All these features favor a higher Fe 2+/Fe3+ ratio in the tin hump region relative to deeper regions. A characteristic of the aforementioned CNF glass is that ferrous and ferric ions have a similar average coordination number of 4.6. That means that the composition of the silicate glass has hardly any effect on the coordination number of iron (either Fe 2+ or Fe 3+) to the composition of the silicate-based glass since strong concentrations of either alumina, or magnesium oxide or potassium oxide result in marginal change in the coordination number [12]. In our spectra, the XANES of surface, tin hump and bulk regions show quite similar profiles, regarding both the main white line at 7130 eV, which stems from medium to long range order, and the first EXAFS maximum, which is related to the Fe–O distance. That the XANES remains the same here while the pre-edge shows a change in the valence state of iron is therefore consistent with the findings relative to the coordination number of iron in CNF glass. Nevertheless, there is a
Fig. 4. Sn L-edge XAS spectra recorded throughout the surface region at various depths, as indicated at the right. The spectra have been normalized to 1 at 4000 eV and vertically shifted for clarity. Shifts in energy associated with depth are evident.
Fig. 5. Fe K-edge XAS spectra (normalized to 1) as a function of depth: spectra corresponding to the three zones described in the text: the very surface (before the tin hump, average 4 μm), the tin hump (10 μm) and the final iron hump (22 μm); the insert shows a zoom of the pre-edge spectra.
surface concentration, as determined by microprobe analysis, prior to being compared to microprobe data. The agreement between these two independent sets of profiles is very good, including the relative concentrations and features such as the buried tin hump, the minimum in iron concentration, and the progressive increase in sulfur content at increasing distance from the surface. A similar analysis is shown in Fig. 3 for the 3 mm glass. Due to a shorter holding time on the tin bath at high temperature, the tin and iron bumps appear closer to the surface in this thinner sample, as expected [4]. Sulfur hardly shows any significant variation, consistently with the fact that it already reaches its bulk concentration at the tin bump (5 μm in the present case).
3.2. Chemical states analyzed by XANES All XANES spectra were normalized by setting (i) the pre-edge domain to zero and (ii) the absorption far from the edge to one. This normalization ‘per atom’ allowed to follow the evolution of the valence states of the element.
3204
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
Fig. 6. S K-edge XAS spectra (normalized to 1) as a function of depth on the 10 mm sample. The spectra have been vertically shifted for clarity. Insert: the S6+/S2− ratio as a function of the distance from the surface for two glass thicknesses, showing that S6+ increases more rapidly for the thin sample. On the right hand side are shown the spectra of three reference compounds.
definite difference in the amplitude of the so-called ‘white line’, which could mean some change on the local order close to the surface, but we can hardly be more quantitative. The prepeaks were analyzed quantitatively in a now classical way: the background was modeled by an arctg function, and two Voigt functions simulate the two pre-peak components. The width and the shape of these Voigts have been kept constant, while their position and amplitude were the free parameters of the fits. Actually, the energy position of the peaks does not move from one fit to another (7113.1 and 7114.8 eV) while their amplitude ratio gives the relative amount of Fe 3+/Fe 2+. Within the three zones defined above, values of 1.8, 1.1 and 1.9 were obtained for Fe 3+/Fe 2+ at distances from the surface of 4, 10 and 22 μm respectively. In agreement with the general expectation, the more reduced state of iron around 10 μm is associated with the so-called tin hump were Sn 4+ dominates [6]. 3.2.3. Sulfur Sulfur concentration is much lower in surface than in bulk (Fig. 2). It increases monotonically as the distance from the surface increases till it reaches the bulk value at a depth of ~ 25–30 μm. A comparison with the reference spectra that are presented on the right hand side in Fig. 6, leads to assign the features at 2473 eV and 2482 eV of the XANES spectra shown in Fig. 6 to fingerprints of sulfide and sulfate, respectively. Actually, while the intense white line at about 2482 eV is an undisputable fingerprint of S 6+, the signature of S 2− is more ambiguous since, for instance, the main peak in the S K edge of ZnS is lying at a higher energy than the peak observed in the S K-edge spectrum of CuFeS2[21]. Another possible valence state of sulfur corresponding to the feature at 2473 eV could be S 0, with sulfur in the form of S–S units, but this appears very unlikely here. The intensities of the 2473 eV and 2482 eV peaks were used to determine the relative abundance of reduced and oxidized sulfur. The spectra were modeled by a sum of four Lorentzians (at 2473.2, 2477.4, 2481.6 and 2497 eV) superimposed to an arctg function centered at 2486 eV to represent the step at the S K-edge. The evolution of the ratio of the S 2− and S6+ species was then extracted as a function of the distance from the glass surface. In the inset of Fig. 6, it appears that (i) sulfur is more in its reduced form at the vicinity of the surface than in bulk; (ii) the S6+/S2− ratio increases monotonously as the distance from the surface increases (and as the sulfur content increases too); (iii) sulfur is almost completely in the form of S6+ beyond the tin hump (that means for a depth greater than 20–25 μm in the case of the 10 mm thick glass plate, and greater than
10 μm for the 3 mm thick glass), as sulfur reaches its bulk concentration. This explains the more rapid increase of sulfate concentration in the 3 mm than in the 10 mm thick sample. Note that the values plotted in the inset of Fig. 6 only show trends and are not a measure of absolute concentrations.
3.3. Dependence on the thickness and on the side In Fig. 7 are plotted the sulfur XANES for different glass thicknesses and for both sides (tin and atmosphere). Since we used FY, the depth sampled was similar in both cases (≈5 μm). The sulfide contribution is more intense on the tin side than on the atmosphere side and the relative content of S 6+ is much higher on the atmosphere side. Therefore, sulfide is more abundant at the surface of the tin side than at the atmosphere side and sulfate is much closer to the surface on the atmosphere side than on the bath side. In addition, the fully oxidized species S 6+ is much present at the surface of the thinner glasses (3 and 4 mm thick) than the thicker ones (6 and 10 mm thick).
Fig. 7. The sulfur spectra taken by FY on the atmosphere side and the tin side for different glass thicknesses: 3 mm (red), 4 mm(blue), 6 mm (green) and 10 mm (black). Spectra of the atmosphere side have been vertically shifted.
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
3205
Fig. 8. Comparison of three successive spectra taken on the same point. Left: at the sulfur edge, the second scan shows the signature of S4+, which then remains constant on the third run. Right: at the iron pre-edge, the second spectrum shows a change in the Fe2+/Fe3+ ratio, which also remains constant on the third scan.
3.4. Radiation effects Radiation effects were accounted for during the whole series of X-ray experiments that are presented herein. However, their description required the knowledge of the spectra associated with the different elements under study and could not be made before. By comparison with standards, the feature at about 2478 eV is assigned to S 4+ species [17,21]. In the case of silicate glass, it was proved to result from X-ray radiation damage [17]. To check this point, the same sulfur XANES spectrum was recorded on three different points of the glass surface with the focused beam, and then two more scans were performed on each of those three points (Fig. 8, left). The first three experiments (made on a pristine surface) were identical with only a small signature of S 4+. The two other sets were identical with 2478 eV peaks of similar amplitude. At the same time, a slight decrease in absorption was observed between S 2− and S 4+ as well as a very slight increase of the weight of S 4+ and S 6+. It was concluded that: (i) S 4+ is not constitutive of the glass but it is an effect of the microfocused beam irradiation — this in agreement with Wilke et. al. [17] — and (ii) its concentration does not increase with the irradiation but remains constant after the first XAS spectrum has been recorded. Actually, the above analysis as a function of the distance to the surface of the XANES spectra of sulfur (Fig. 6) shows also that this S 4+ component remains constant. As a consequence, the radiation damage is not going to affect the conclusions on the speciation of the sulfur along the surface concentration profile. Pre-edge spectra of iron that were recorded after similar irradiations as in the study of the sulfur edge are shown on the right hand side of Fig. 8. The first scan exhibits a signature of Fe 2+ not present in the two next ones, while the extended spectra (not shown in the figure) are identical. This observation proves that the oxidation state of iron is perturbed by X-ray radiations. Finally, we could not detect any radiation damage in the absorption spectrum of tin. Therefore, during the analysis of sulfur and iron edges (Section 3), exposure of given spots were limited in time to avoid damages. Spectra shown in the present work were obtained by adding data recorded at locations very close to each other but different.
the observation of energy shifts in the XAS features of the Sn L-edge (Fig. 4) indicates that this fraction is significant. Therefore the formation of stannic ions needs another redox reaction than the formation of ferrous ions. The much higher sulfide/sulfate content that is observed on the tin side than on the atmosphere side (Fig. 7) proves that the formation of sulfide at the vicinity of the glass surface is linked to the presence of tin, as if S 6+ ions were ‘consumed’ by Sn2+ on the glass side which is in contact with the bath. A suggestion is to involve sulfur: 2þ
4Sn
6þ
4þ
þ S →4Sn
2−
þS
ð2Þ
The examination of the data relative to the S K edge XAS spectra (Figs. 2 and 6) supports the assumption. Going inward from the surface, the sulfide/sulfate ratio is about constant till a depth of 10 μm. It then decrease continuously (inset of Fig. 6). The product of the sulfur concentration (extracted from Fig. 2, bottom) by the ratio S 2−/S 6+ (extracted from the inset of Fig. 6) gives the amount of sulfur in the sulfide state. This quantity is plotted in Fig. 9 versus the distance to the surface and superimposed to the tin concentration (from Fig. 2). We see that this amount, which is given here is arbitrary units, exhibits a maximum at the position of the tin hump. The sulfide profile is quite consistent with the above suggestion of a glass surface in which the tin hump is dominated by Sn4+ ions resulting from reaction (2). This appears as a direct proof of the chemical role of the sulfur in the change of the valence of tin from 2+ to 4+ according to the reaction (2). Finally, sulfur and tin contents are such that the picture is realistic. In the tin hump region, the concentrations of SnO2
4. Discussion At the depth corresponding to the tin hump, the concentration of tin oxide is about 1.3 wt.%. The concentration of iron in this region, which only amounts to 0.05%, is much too low to balance the redox reaction Sn 2+ → Sn 4+ + 2e − which is at the origin of the formation of the hump [1,5,6] since the reaction (1) given in the course of the introduction would require an iron content of 1.4 wt.% to occur. Even if the reaction that takes place within the tin hump region only involves a fraction of the total tin content which is present at that depth [7,8],
Fig. 9. Comparison between the profile of S2− (black circles) and the tin concentration (red markers) as a function of the distance to the surface in the case of the 10 mm thick sample, which shows that a maximum of S2− coincides with the tin hump.
3206
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
and SO3 are 1.3wt.% and 0.20–0.25wt.%, respectively, so that the concentration of SO3 of 0.17 wt.% which is required to fully transform stannic ions into stannous ions is clearly available. Therefore, regarding the redox reaction which gives rise to the tin hump, it is concluded that a process similar to that described by Frischat et al. [5,6] by involving iron ions (reaction (1)) takes place with sulfur ions (reaction (2)). From the point of view of the oxydo-reduction of tin, iron and sulfur play a similar role. However, their behavior within the glass matrix is quite different. Because of the low iron concentration, the redox reaction (1) is marginal in the formation of stannic ions. It is nevertheless central in the iron concentration profile since it governs the iron redox. Whenever the major part of the stannic ions arises from the reduction of sulfur, the reduction of ferric ions into ferrous ions in the tin hump region is quite effective till the extent-end of the tin profile (Fig. 5). Ferrous ions are glass modifiers that diffuse inward from the tin hump as a part of the compensation for the diffusion of bulk cations toward the interface between tin and glass melt [1]. In the oxidizing environment that prevails in the bulk of glass beyond the depth of penetration of tin, ferrous ions transform into ferric ions which, as the Fe 3+/Fe 2+ ratio increases, act as glass formers [20]. The chemistry that accompanies the inward diffusion of Fe 2+ results in the iron hump [1] which is seen on the concentration profiles and on the images recorded by X-ray microscopy (Fig. 3). Owing to the origin of the Fe 2+ ions, the iron hump appears as a chemical echo of the tin hump. In the opposite direction, the ferric form observed in the surface region of the glass (Fig. 5) is somewhat unexpected since the presence of stannous ions should favor the reduced ferrous form [4,5]. However, the chemistry at the extreme surface of the glass depends on the environmental history of the material which is out of the scope of the present work. The above study has been mostly performed on the 10 mm glass plate because concentration profiles were more easily observed on those samples (Fig. 2) than on thinner samples (Fig. 3). However, the redox equilibria were shown to be independent of the thickness of the glass plates (compare Fig. 3 to Fig. 2). Moreover, the composition of float glass under study is quite common. Therefore, the conclusions drawn herein likely apply to the many cases in which the iron content is too low to balance the redox profile of the tin on the bath side of the glass.
modifier) that are formed in the tin hump region and their transformation into glass former Fe 3+ ions as the tin concentration vanishes is at the origin of the iron hump which thus appears as a chemical echo of the tin hump. - sulfur concentration increases till it levels off when it reaches its bulk value beyond the tin hump. Sulfur is in both sulfide S 2− and sulfate S 6+ form at the surface. As the depth increases, the relative sulfate content increases. It becomes the unique form in bulk. The redox reaction involving tin and iron which is commonly associated to the tin hump region was evident by a strong enrichment in Fe 2+. However, in the glass under study, the iron concentration is much too low to balance the oxidation of Sn 2+ into Sn 4+. On the other hand, the amount of sulfur with a 2− valence exhibits a maximum at the tin hump depth around 10 μm where the valence of this element changes from 2+ to 4+. This appears therefore as a strong evidence that sulfur plays the role which is usually attributed to iron according to: 2þ
4Sn
The redox profiles of tin, iron and sulfur at the float glass surface were studied on purposely cut samples by EPMA, to determine concentrations, and micro X-ray analysis to characterize the state of oxidation. The common resolution of ~1 μm of the two methods was well suited to analyze concentration profiles that extend over tens of microns. Analysis focused on 10 mm thick glass plates to make observation easier. Going inward from the surface to the bulk of the glass, the main characteristics of the concentration profiles are: - after a diffusion-driven decrease, the tin concentration shows a local maximum at a depth of 10 μm, the so-called tin hump, and then vanishes at a depth of ≈40 μm. XAS analysis shows that stannous ions, that dominate in the shallow layers of the glass, switch to stannic ions in the tin hump region. - the iron concentration decreases, passes through a minimum in the tin hump region, prior to progressively increasing and, after a hump which appears as the extent-end of the tin profile, reaches its bulk value. The inward diffusion of the Fe 2+ ions (glass
4þ
2−
þS
These conclusions are stressed to be quite general since the composition of the float glass that was examined herein is quite commonly encountered. Acknowledgments We acknowledge friendly and enlightening discussion with MarieHélène Chopinet and Patrick Garnier. We are grateful to the machine and beamline groups at SLS-PSI whose outstanding efforts have made these experiments possible. References [1] [2] [3] [4] [5] [6] [7] [8]
5. Conclusion
6þ
þ S →4Sn
[9] [10] [11] [12] [13]
[14] [15] [16] [17] [18] [19] [20] [21]
G.B. Cook, R.F. Cooper, J. Non-Cryst. Solids 249 (1999) 210. J.A. Howell, J.R. Hellmann, C.L. Muhlstein, J. Non-Cryst. Solids 354 (2008) 1891. J.S. Sieger, J. Non-Cryst. Solids 19 (1975) 213. Y. Hayashi, K. Matsumoto, M. Kudo, J. Non-Cryst. Solids 282 (2001) 188. G.H. Frischat, C. Müller-Fildebrandt, D. Moseler, G. Heide, J. Non-Cryst. Solids 283 (2001) 246. G.H. Frischat, C.R. Chim. 5 (2002) 759. K.F.E. Williams, C.E. Johnson, J. Greengrass, B.P. Tilley, D. Gelder, J.A. Johnson, J. Non-Cryst. Solids 211 (1997) 164. K.F.E. Williams, C.E. Johnson, O. Nikolov, M.F. Thomas, J.A. Johnson, J. Greengrass, J. Non-Cryst. Solids 242 (1998) 183. T.-J. Wang, H. Zhang, G. Zhang, T. Yuan, J. Non-Cryst. Solids 271 (2000) 126. T.-J. Wang, Glass Technol. 38 (1997) 104. P. Lagarde, A.-M. Flank, J. Jupille, H. Montigaud, J. Phys. Conf. Ser. 190 (2009) 012079. F. Farges, Y. Lefrère, S. Rossano, A. Berthereau, G. Calas, G.E. Brown Jr., J. Non-Cryst. Solids 344 (2004) 176. A.-M. Flank, G. Cauchon, P. Lagarde, S. Bac, M. Janousch, R. Wetter, J.-M. Dubuisson, M. Idir, F. Langlois, T. Moreno, D. Vantelon, Nucl. Instrum. Methods Phys. Res. Sect. B 246 (2006) 269. Z. Liu, K. Handab, K. Kaibuchi, Y. Tanaka, Jun Kawai, J. Electron. Spectrosc. Relat. Phenom. 135 (2004) 155. M. Wilke, F. Farges, P.-E. Petit, G.E. Brown Jr., F. Martin, Am. Mineral. 86 (2001) 714. M. Wilke, F. Farges, G.M. Partzsch, C. Schmidt, H. Behrens, Am. Mineral. 92 (2007) 44. M. Wilke, P. Jugo, K. Klimm, J. Susini, R. Botcharnikov, S.C. Kohn, M. Janousch, Am. Mineral. 93 (2008) 235. F. Lamouroux, N. Can, P.D. Townsend, B.W. Farmery, D.E. Hole, J. Non-Cryst. Solids 212 (1997) 232. P.D. Townsend, N. Can, P.J. Chandler, B.W. Farmery, R. Lopez-Heredero, A. Peto, L. Salvin, D. Underdown, B. Yang, J. Non-Cryst. Solids 223 (1998) 73. V. Magnien, D.R. Neuville, L. Cormier, B.O. Mysen, V. Briois, S. Belin, O. Pinet, P. Richet, Chem. Geol. 213 (2004) 253. D.A. McKeown, I.S. Muler, H. Gan, I.L. Pegg, W.C. Stolte, J. Non-Cryst. Solids 333 (2004) 74.
Contents lists available at ScienceDirect
Journal of Non-Crystalline Solids j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j n o n c r y s o l
Redox profile of the glass surface A.-M. Flank a, P. Lagarde a,⁎, J. Jupille b, H. Montigaud c a b c
Synchrotron SOLEIL, l'Orme des Merisiers, BP 48 91192 Gif/Yvette cedex, France INSP, CNRS et Université Paris 6, Campus de Boucicaut, 140 rue de Lourmel 75015 Paris, France Saint-Gobain Recherche 39, quai Lucien Lefranc, BP 135 93303 Aubervilliers Cedex, France
a r t i c l e
i n f o
Article history: Received 16 February 2011 Received in revised form 30 March 2011 Available online 3 June 2011 Keywords: Glass; Surface of the glass; Redox profiles; Float glass
a b s t r a c t The redox profiles of tin, iron and sulfur at the float glass surface were determined on purposely cut samples by Electron Probe Micro Analysis (EPMA), X-ray Fluorescence mapping and X-ray Absorption Spectroscopy (XAS) at the micron scale. Going inward from the surface to the bulk, it was observed that (features do not depend on the glass thickness (holding time) though they extend over depths that vary from ca. 25 to 50 μm): (i) after a diffusion-driven decrease and prior to vanishing, the tin concentration passes through a local maximum (the tin hump), where stannous ions, which dominate the shallow layers, switch to stannic ions; (ii) the iron concentration decreases, passes through a minimum at the tin hump where iron is in the more reduced form (lowest Fe 3+/Fe 2+ ratio); it then increases and, after a hump which appears as a chemical echo of the tin hump, it reaches the bulk value; (iii) the concentration of sulfur increases up to reach the bulk concentration beyond the tin hump region. In a mixed S 6+/S2− form at the surface, sulfur is only in sulfate form in the bulk. In the case under study, the iron concentration is much too low to balance the redox reaction Sn 2+ → Sn4+ that occurs at the tin hump. Sulfur is shown to play the role usually attributed to iron, according to the reaction 4Sn
2þ
6þ
4þ
þ S →4Sn
2−
þS
The occurrence of that reaction is supported by the appearance of sulfide S 2− in the tin hump region with an appropriate concentration profile of a much stronger S2−/S 6+ ratio on the tin side than on the atmosphere side of the float glass. The conclusions drawn herein likely apply to the many cases in which the glass composition is similar as that encountered herein. © 2011 Elsevier B.V. All rights reserved.
1. Introduction The first step of the float process consists in spreading the glass melt onto liquid tin to take advantage of the balance between surface and interface tensions to produce flat glass sheets. Those conditions, that are quite suitable to produce glazings for common use, are not adapted regarding new applications which raise the level of demand required for the production of float glass, such as the manufacture of supports for electronic display elements that needs nearly atomically smooth substrates [1] and the control of the nanomechanical properties of glass surfaces [2,3]. The control at the microscopic level of the glass surface requires a detailed understanding of the redox reactions involved in the float process. The two faces of the glass ribbon are different. To prevent the oxidation, the whole part of the process which involves the tin bath is
⁎ Corresponding author. E-mail address: [email protected] (P. Lagarde). 0022-3093/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2011.03.046
run in reducing conditions by circulating a gas mixture comprising nitrogen and hydrogen. Due to such an environment, the chemical composition of the skin of float glass differs from that of bulk over a few tens of microns with different behaviors on the atmosphere side and on the tin side. Complex concentration profiles result from interdiffusion and redox reactions at the interface between tin and glass melt [1]. Researches focus on the concentrations and profile of iron species because that element is stressed to be central [1,3–6] in the formation of the tin satellite peak that appears at a depth of ~5–20 μm from the glass surface. The so-called tin hump was first identified by Sieger [3]. It has been then demonstrated by Mossbauer spectroscopy that the shallower part of the tin profile mostly involves Sn2+ ions and that the tin hump is dominated by Sn4+ ions [5,7,8]. On the basis of a simulation, Wang et al. suggest that the tin hump arises from an easy diffusion of Sn4+[9]. At variance, Cook and Cooper [1] argue that the diffusion of stannic ions that act as glass modifiers is much easier than that of stannous ions that behave as glass formers. These authors and Frischat et al. [5,6] assign the occurrence of the tin hump to redox reactions. They assume that the formation of stannic ions within the
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
glass is favored by the presence of ferric ions via an ion-exchange process: 2þ
Sn
3þ
4þ
þ 2Fe →Sn
2þ
þ 2Fe
ð1Þ
A maximum in iron concentration is also observed at the extentend of the tin profile. It is associated to the inward diffusion of the then produced ferrous ions [1]. There is also some suggestion that the formation of Sn 4+ could involve sulfur [10,4,5], although no evidence was ever provided. To date, direct measurements of the redox profile of the glass surface are lacking. The objective of the present work is to determine concentration profiles of tin, sulfur and iron at the vicinity of a float glass surface by examining cuts of glass plates by microanalysis. Electron Probe Micro Analysis (EPMA) was used to determine quantitatively the concentration as a function of the depth. Micro X-ray Fluorescence mapping and X-ray Absorption micro-Spectroscopy (XAS) allowed a characterization of the oxidation states and their spatial distribution. These techniques have a comparable resolution of about 1 μm. Therefore, they were suited to analyze concentration along the profiles under study that extend over tens of microns. Moreover X-ray spectroscopy is of relevance to determine the chemical states of tin, sulfur and iron in glasses [11]. It has been chosen to work on common float samples with iron (Fe2O3) and sulfur (SO3) bulk concentrations of 0.1 and 0.3 wt.%, respectively. It will appear during the discussion that this choice allows to pinpoint the respective contributions of these two species to the redox profile of the float glass surface.
2. Experimental Samples were float glass plates with different thicknesses: 3, 4, 6 and 10 mm. Glass surfaces were analyzed by sampling both the bath side and the atmosphere side. In order to study concentration profiles, the bath sides of two glass plates were glued together (Fig. 1), and this cross-section sample was prepared by grinding with SiC abrasive papers and then polishing with diamond (3 and 1 μm size). A quantitative analysis of the concentration was achieved on the 10 mm thick sample by a Cameca SX50 Electron Probe Micro Analyser employing four different wavelength-dispersive spectrometers, at Saint-Gobain Recherche. Such elemental analysis was realized on the bath surface and the bulk (from the cross section) after plating them with a thin graphite conducting layer. The experiments were performed with an accelerating voltage of 15 kV and a beam intensity of 10 or 150 nA according to the element. Five different spots were analyzed in every case. The composition of this glass studied here is shown in Table 1, together with that of the glass CNF 0.2 studied already by Farges et al. [12] (see below).
Sample raster
Sn or atmosphere faces
Fluorescence
X-rays (3 x 3 µm2)
Image
Fig. 1. Geometry of the experiments.
Glue
(4 µm)
3201
Table 1 Composition of the float glass used herein (weight per cent of oxides). Error bars are maximized by the last digit. Comparison with the CNF 0.2 composition analyzed by Farges et al. [12].
This work CNF 0.2
SiO2
Na20
MgO
CaO
Al2O3
Fe2O3
SO3
71.9 72.1
14.0 13.8
3.9 –
8.7 13.9
0.7 –
0.09 0.2
0.26 –
X-ray experiments have been done at the LUCIA beamline [13], installed on the Swiss synchrotron facility SLS (Villigen). The monochromator is equipped with a pair of Si(111), and the spot size was set to 3 × 3 μm 2 at all energies of interest. By using the two classical techniques of the total electron yield (TEY) and fluorescence yield (FY), we could probe two different depths into the sample, typically 0.5 μm in TEY and 5 μm by FY. The TEY signal is obtained by a measure of the sample drain current, while the FY is detected by a monoelement (10 mm 2) silicon drift diode. The analysis of the oxydation state of the different elements was be done comparing the near edge spectra (XANES) to those of reference compounds whose valency is well known. XANES of sulfur in several well characterized compounds were purposely recorded. Tin spectra were compared with published data [14]. Finally, the oxydation state of iron was determined by an analysis of the pre-peaks [12,15,16] as explained below. To avoid any artifact from the self-absorption of the fluorescence signal, the common surface of the glass was set horizontal, and this common surface could be precisely localized by examining the silicon mapping since this element was expected to exhibit a constant concentration throughout the sample (Fig. 1). Given the flux on the sample of about 5 × 10 10 photons/s, we could be concerned by possible radiation damage or photoreduction processes, mostly on sulfur which is well known to be a very sensitive element. In order to check that, we did successive experiments on the same point. In the case of tin and iron, the spectra remains identical and therefore we believe that radiation damages should be of a minor importance in that case. At the contrary, and this will be discussed in Section 3.4, some of the sulfur features could be the result of the irradiation, as it has been also pointed out by Wilke et. al. [17].
3. Results 3.1. Mapping of the different elements The concentration profiles at the surface of the 10 mm thick float glass plate were first examined by EPMA. Tin penetrates in the glass over a depth of ≈40 μm. Going inward, its concentration decreases steeply prior to show a buried peak — the so-called tin hump — at 10–20 μm below the surface. The concentration of iron is peaking at the extreme surface. Then, after a depleted zone with a minimum just before the tin hump, it increases to peak again as the Sn concentration vanishes, prior to reach its bulk value. The sulfur concentration, which is at its lowest level at the glass surface, increases progressively to level off at a depth of ≈15 μm which coincides with the Sn hump. Its bulk value is about three times higher than the surface value. Sn, Fe and S profiles that are observed herein are consistent with the commonly accepted picture [1,3]. Fig. 2 shows the images obtained on the 10 mm thick sample for the different elements analyzed in this study, with two Sn faces stuck together as described previously. The silicon and iron maps have been obtained with an incident energy of 7130 eV, the tin map has been made at 3990 eV in order to avoid the contribution of calcium present in the glass, and the sulfur one has been recorded at 2500 eV in order to maximize the fluorescence signal. On the silicon map, which must be uniform, we see the signature of the glue (blue and green zones) which amounts about
3202
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
Fig. 2. Top: Cartographies of the silicon, tin, iron and sulfur concentrations, as recorded at the Si K-edge, S K-edge, Fe K-edge and Sn L-edge, respectively for the 10 mm thick glass. The dashed area corresponds to the glue between the two faces. The depth is indicated on the vertical scale. Bottom: Concentration of the different elements as a function of the distance from the surface, deduced from line profile analysis of the above maps (open symbols) and superimposed with the electron microprobe results (filled symbols).
8 μm and this zone must be excluded from the concentration analysis of the other elements. The concentrations of Sn, Fe and S were analyzed as a function of depth on each side of the junction between the two glass plates, and the results are also shown in Fig. 2. The values of concentrations at a given depth are normalized by the silicon concentration recorded at the same depth. This silicon concentration appears to be constant throughout the sample which is true within a few percent [3,18]. The variations of the other elements appear fully in line with the features
already revealed by microprobe analysis. The buried tin hump is observed at around 10 μm from the surface of the 10 mm sample. Iron concentration peaks at the surface, goes through a flat minimum ca 15 μm and then increases progressively till it shows a maximum as the tin concentration vanishes. Sulfur concentration is at minimum at the surface. It increases continuously in the inward direction up to reach its bulk value. Fig. 2 shows also the results of the electron microprobe. In the absence of absolute calibration, the concentration profiles that are derived from microcartography are rescaled with respect to the
Fig. 3. The same analysis for the 3 mm thick glass as in Fig. 2. The evolution of the concentrations with the distance to the surface shows that, in this case, the iron bump appears closer to the surface. Note that the concentration scales have no absolute meaning and are different from one element to another one.
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
3203
3.2.1. Tin A series of Sn L edge spectra, collected every 2 μm from the surface (excluding the glue region) is shown in Fig. 4. The maximum is seen to shift from 3954 eV for XAS spectra recorded at the sample surface to 3961 eV for spectra recorded at depth higher than 5–10 μm. Apart from a shift of 4 eV attributed to a different monochromator calibration, and a smoothing of the overall structures due to the glassy state of our samples, those spectra are in line with those recorded by Liu et al. on reference compounds [14]. In particular the 7 eV blue shift of the main maximum is attributed to the switch from SnO to SnO2. The oxidation of Sn2+ to Sn4+ between 5 and 10 μm coincides with the onset of the tin hump (seen in Fig. 2). The present data support the widely accepted picture of a strong contribution of Sn4+ to the buried tin peak [1,4–6,19]. However, the relative concentration of stannous and stannic ions likely depends on the float sample under study. Following a Mössbauer analysis of tin species at a float glass surface, Williams et al. reported a percentage contribution of Sn4+ close to zero at the vicinity of the glass surface and of about 1/3 in the hump region [7]. By means of cathodoluminescence and measuring the refractive index profile, Townsend et al. found also a maximum in Sn4+ concentration within the glass at the interface with the bath [19].
3.2.2. Iron Iron is expected to exhibit different valence states throughout the surface region of the glass. At the Fe K edge absorption spectrum, the degree of oxidation of iron and its speciation are given by the pre-edge features and the fine structure of the XANES [12,15,16]. The energy separation between features associated with Fe2+ and Fe3+ is about 1.5 keV [12,15,16,20]. Despite this sizeable energy shift, the identification of the iron species is uneasy because, beyond the spectral shape, the intensity of the pre-edge features depends strongly on the local structural environment. Losses in centrosymmetry result in an increase of the intensity so that contributions from Fe ions in tetrahedral position are about four times more intense than contributions from ions in octahedral sites [12,15,16]. Fe K edge spectra recorded through the surface region by XAS microscopy are shown in Fig. 5. To improve the signal to noise ratio and to avoid as much as possible radiation effects (see below) the as recorded Fe K edge spectra were added over three zones which correspond to the very surface (before the tin hump, 0 to 8 μm), the tin hump (8 to 12 μm) and the final iron hump (15 to 30 μm), respectively. A zoom of the regions of the pre-edge is presented in Fig. 5. The composition of the float glass that is studied herein is very close to the CNF glasses that were analyzed by Farges et al., in particular the CNF 0.2 sample [12], the chemical composition of which is given in Table 1. The pre-edge profile recorded around the tin hump region shows an increase of the 2+ component with respect to the two other spectra, which indicates a higher Fe2+ content around 10 μm from the surface. Moreover, in the XAS spectrum associated to the 10 μm region, the edge itself is located at a lower energy than that recorded outside this region, which is also a signature of a lower valence state. All these features favor a higher Fe 2+/Fe3+ ratio in the tin hump region relative to deeper regions. A characteristic of the aforementioned CNF glass is that ferrous and ferric ions have a similar average coordination number of 4.6. That means that the composition of the silicate glass has hardly any effect on the coordination number of iron (either Fe 2+ or Fe 3+) to the composition of the silicate-based glass since strong concentrations of either alumina, or magnesium oxide or potassium oxide result in marginal change in the coordination number [12]. In our spectra, the XANES of surface, tin hump and bulk regions show quite similar profiles, regarding both the main white line at 7130 eV, which stems from medium to long range order, and the first EXAFS maximum, which is related to the Fe–O distance. That the XANES remains the same here while the pre-edge shows a change in the valence state of iron is therefore consistent with the findings relative to the coordination number of iron in CNF glass. Nevertheless, there is a
Fig. 4. Sn L-edge XAS spectra recorded throughout the surface region at various depths, as indicated at the right. The spectra have been normalized to 1 at 4000 eV and vertically shifted for clarity. Shifts in energy associated with depth are evident.
Fig. 5. Fe K-edge XAS spectra (normalized to 1) as a function of depth: spectra corresponding to the three zones described in the text: the very surface (before the tin hump, average 4 μm), the tin hump (10 μm) and the final iron hump (22 μm); the insert shows a zoom of the pre-edge spectra.
surface concentration, as determined by microprobe analysis, prior to being compared to microprobe data. The agreement between these two independent sets of profiles is very good, including the relative concentrations and features such as the buried tin hump, the minimum in iron concentration, and the progressive increase in sulfur content at increasing distance from the surface. A similar analysis is shown in Fig. 3 for the 3 mm glass. Due to a shorter holding time on the tin bath at high temperature, the tin and iron bumps appear closer to the surface in this thinner sample, as expected [4]. Sulfur hardly shows any significant variation, consistently with the fact that it already reaches its bulk concentration at the tin bump (5 μm in the present case).
3.2. Chemical states analyzed by XANES All XANES spectra were normalized by setting (i) the pre-edge domain to zero and (ii) the absorption far from the edge to one. This normalization ‘per atom’ allowed to follow the evolution of the valence states of the element.
3204
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
Fig. 6. S K-edge XAS spectra (normalized to 1) as a function of depth on the 10 mm sample. The spectra have been vertically shifted for clarity. Insert: the S6+/S2− ratio as a function of the distance from the surface for two glass thicknesses, showing that S6+ increases more rapidly for the thin sample. On the right hand side are shown the spectra of three reference compounds.
definite difference in the amplitude of the so-called ‘white line’, which could mean some change on the local order close to the surface, but we can hardly be more quantitative. The prepeaks were analyzed quantitatively in a now classical way: the background was modeled by an arctg function, and two Voigt functions simulate the two pre-peak components. The width and the shape of these Voigts have been kept constant, while their position and amplitude were the free parameters of the fits. Actually, the energy position of the peaks does not move from one fit to another (7113.1 and 7114.8 eV) while their amplitude ratio gives the relative amount of Fe 3+/Fe 2+. Within the three zones defined above, values of 1.8, 1.1 and 1.9 were obtained for Fe 3+/Fe 2+ at distances from the surface of 4, 10 and 22 μm respectively. In agreement with the general expectation, the more reduced state of iron around 10 μm is associated with the so-called tin hump were Sn 4+ dominates [6]. 3.2.3. Sulfur Sulfur concentration is much lower in surface than in bulk (Fig. 2). It increases monotonically as the distance from the surface increases till it reaches the bulk value at a depth of ~ 25–30 μm. A comparison with the reference spectra that are presented on the right hand side in Fig. 6, leads to assign the features at 2473 eV and 2482 eV of the XANES spectra shown in Fig. 6 to fingerprints of sulfide and sulfate, respectively. Actually, while the intense white line at about 2482 eV is an undisputable fingerprint of S 6+, the signature of S 2− is more ambiguous since, for instance, the main peak in the S K edge of ZnS is lying at a higher energy than the peak observed in the S K-edge spectrum of CuFeS2[21]. Another possible valence state of sulfur corresponding to the feature at 2473 eV could be S 0, with sulfur in the form of S–S units, but this appears very unlikely here. The intensities of the 2473 eV and 2482 eV peaks were used to determine the relative abundance of reduced and oxidized sulfur. The spectra were modeled by a sum of four Lorentzians (at 2473.2, 2477.4, 2481.6 and 2497 eV) superimposed to an arctg function centered at 2486 eV to represent the step at the S K-edge. The evolution of the ratio of the S 2− and S6+ species was then extracted as a function of the distance from the glass surface. In the inset of Fig. 6, it appears that (i) sulfur is more in its reduced form at the vicinity of the surface than in bulk; (ii) the S6+/S2− ratio increases monotonously as the distance from the surface increases (and as the sulfur content increases too); (iii) sulfur is almost completely in the form of S6+ beyond the tin hump (that means for a depth greater than 20–25 μm in the case of the 10 mm thick glass plate, and greater than
10 μm for the 3 mm thick glass), as sulfur reaches its bulk concentration. This explains the more rapid increase of sulfate concentration in the 3 mm than in the 10 mm thick sample. Note that the values plotted in the inset of Fig. 6 only show trends and are not a measure of absolute concentrations.
3.3. Dependence on the thickness and on the side In Fig. 7 are plotted the sulfur XANES for different glass thicknesses and for both sides (tin and atmosphere). Since we used FY, the depth sampled was similar in both cases (≈5 μm). The sulfide contribution is more intense on the tin side than on the atmosphere side and the relative content of S 6+ is much higher on the atmosphere side. Therefore, sulfide is more abundant at the surface of the tin side than at the atmosphere side and sulfate is much closer to the surface on the atmosphere side than on the bath side. In addition, the fully oxidized species S 6+ is much present at the surface of the thinner glasses (3 and 4 mm thick) than the thicker ones (6 and 10 mm thick).
Fig. 7. The sulfur spectra taken by FY on the atmosphere side and the tin side for different glass thicknesses: 3 mm (red), 4 mm(blue), 6 mm (green) and 10 mm (black). Spectra of the atmosphere side have been vertically shifted.
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
3205
Fig. 8. Comparison of three successive spectra taken on the same point. Left: at the sulfur edge, the second scan shows the signature of S4+, which then remains constant on the third run. Right: at the iron pre-edge, the second spectrum shows a change in the Fe2+/Fe3+ ratio, which also remains constant on the third scan.
3.4. Radiation effects Radiation effects were accounted for during the whole series of X-ray experiments that are presented herein. However, their description required the knowledge of the spectra associated with the different elements under study and could not be made before. By comparison with standards, the feature at about 2478 eV is assigned to S 4+ species [17,21]. In the case of silicate glass, it was proved to result from X-ray radiation damage [17]. To check this point, the same sulfur XANES spectrum was recorded on three different points of the glass surface with the focused beam, and then two more scans were performed on each of those three points (Fig. 8, left). The first three experiments (made on a pristine surface) were identical with only a small signature of S 4+. The two other sets were identical with 2478 eV peaks of similar amplitude. At the same time, a slight decrease in absorption was observed between S 2− and S 4+ as well as a very slight increase of the weight of S 4+ and S 6+. It was concluded that: (i) S 4+ is not constitutive of the glass but it is an effect of the microfocused beam irradiation — this in agreement with Wilke et. al. [17] — and (ii) its concentration does not increase with the irradiation but remains constant after the first XAS spectrum has been recorded. Actually, the above analysis as a function of the distance to the surface of the XANES spectra of sulfur (Fig. 6) shows also that this S 4+ component remains constant. As a consequence, the radiation damage is not going to affect the conclusions on the speciation of the sulfur along the surface concentration profile. Pre-edge spectra of iron that were recorded after similar irradiations as in the study of the sulfur edge are shown on the right hand side of Fig. 8. The first scan exhibits a signature of Fe 2+ not present in the two next ones, while the extended spectra (not shown in the figure) are identical. This observation proves that the oxidation state of iron is perturbed by X-ray radiations. Finally, we could not detect any radiation damage in the absorption spectrum of tin. Therefore, during the analysis of sulfur and iron edges (Section 3), exposure of given spots were limited in time to avoid damages. Spectra shown in the present work were obtained by adding data recorded at locations very close to each other but different.
the observation of energy shifts in the XAS features of the Sn L-edge (Fig. 4) indicates that this fraction is significant. Therefore the formation of stannic ions needs another redox reaction than the formation of ferrous ions. The much higher sulfide/sulfate content that is observed on the tin side than on the atmosphere side (Fig. 7) proves that the formation of sulfide at the vicinity of the glass surface is linked to the presence of tin, as if S 6+ ions were ‘consumed’ by Sn2+ on the glass side which is in contact with the bath. A suggestion is to involve sulfur: 2þ
4Sn
6þ
4þ
þ S →4Sn
2−
þS
ð2Þ
The examination of the data relative to the S K edge XAS spectra (Figs. 2 and 6) supports the assumption. Going inward from the surface, the sulfide/sulfate ratio is about constant till a depth of 10 μm. It then decrease continuously (inset of Fig. 6). The product of the sulfur concentration (extracted from Fig. 2, bottom) by the ratio S 2−/S 6+ (extracted from the inset of Fig. 6) gives the amount of sulfur in the sulfide state. This quantity is plotted in Fig. 9 versus the distance to the surface and superimposed to the tin concentration (from Fig. 2). We see that this amount, which is given here is arbitrary units, exhibits a maximum at the position of the tin hump. The sulfide profile is quite consistent with the above suggestion of a glass surface in which the tin hump is dominated by Sn4+ ions resulting from reaction (2). This appears as a direct proof of the chemical role of the sulfur in the change of the valence of tin from 2+ to 4+ according to the reaction (2). Finally, sulfur and tin contents are such that the picture is realistic. In the tin hump region, the concentrations of SnO2
4. Discussion At the depth corresponding to the tin hump, the concentration of tin oxide is about 1.3 wt.%. The concentration of iron in this region, which only amounts to 0.05%, is much too low to balance the redox reaction Sn 2+ → Sn 4+ + 2e − which is at the origin of the formation of the hump [1,5,6] since the reaction (1) given in the course of the introduction would require an iron content of 1.4 wt.% to occur. Even if the reaction that takes place within the tin hump region only involves a fraction of the total tin content which is present at that depth [7,8],
Fig. 9. Comparison between the profile of S2− (black circles) and the tin concentration (red markers) as a function of the distance to the surface in the case of the 10 mm thick sample, which shows that a maximum of S2− coincides with the tin hump.
3206
A.-M. Flank et al. / Journal of Non-Crystalline Solids 357 (2011) 3200–3206
and SO3 are 1.3wt.% and 0.20–0.25wt.%, respectively, so that the concentration of SO3 of 0.17 wt.% which is required to fully transform stannic ions into stannous ions is clearly available. Therefore, regarding the redox reaction which gives rise to the tin hump, it is concluded that a process similar to that described by Frischat et al. [5,6] by involving iron ions (reaction (1)) takes place with sulfur ions (reaction (2)). From the point of view of the oxydo-reduction of tin, iron and sulfur play a similar role. However, their behavior within the glass matrix is quite different. Because of the low iron concentration, the redox reaction (1) is marginal in the formation of stannic ions. It is nevertheless central in the iron concentration profile since it governs the iron redox. Whenever the major part of the stannic ions arises from the reduction of sulfur, the reduction of ferric ions into ferrous ions in the tin hump region is quite effective till the extent-end of the tin profile (Fig. 5). Ferrous ions are glass modifiers that diffuse inward from the tin hump as a part of the compensation for the diffusion of bulk cations toward the interface between tin and glass melt [1]. In the oxidizing environment that prevails in the bulk of glass beyond the depth of penetration of tin, ferrous ions transform into ferric ions which, as the Fe 3+/Fe 2+ ratio increases, act as glass formers [20]. The chemistry that accompanies the inward diffusion of Fe 2+ results in the iron hump [1] which is seen on the concentration profiles and on the images recorded by X-ray microscopy (Fig. 3). Owing to the origin of the Fe 2+ ions, the iron hump appears as a chemical echo of the tin hump. In the opposite direction, the ferric form observed in the surface region of the glass (Fig. 5) is somewhat unexpected since the presence of stannous ions should favor the reduced ferrous form [4,5]. However, the chemistry at the extreme surface of the glass depends on the environmental history of the material which is out of the scope of the present work. The above study has been mostly performed on the 10 mm glass plate because concentration profiles were more easily observed on those samples (Fig. 2) than on thinner samples (Fig. 3). However, the redox equilibria were shown to be independent of the thickness of the glass plates (compare Fig. 3 to Fig. 2). Moreover, the composition of float glass under study is quite common. Therefore, the conclusions drawn herein likely apply to the many cases in which the iron content is too low to balance the redox profile of the tin on the bath side of the glass.
modifier) that are formed in the tin hump region and their transformation into glass former Fe 3+ ions as the tin concentration vanishes is at the origin of the iron hump which thus appears as a chemical echo of the tin hump. - sulfur concentration increases till it levels off when it reaches its bulk value beyond the tin hump. Sulfur is in both sulfide S 2− and sulfate S 6+ form at the surface. As the depth increases, the relative sulfate content increases. It becomes the unique form in bulk. The redox reaction involving tin and iron which is commonly associated to the tin hump region was evident by a strong enrichment in Fe 2+. However, in the glass under study, the iron concentration is much too low to balance the oxidation of Sn 2+ into Sn 4+. On the other hand, the amount of sulfur with a 2− valence exhibits a maximum at the tin hump depth around 10 μm where the valence of this element changes from 2+ to 4+. This appears therefore as a strong evidence that sulfur plays the role which is usually attributed to iron according to: 2þ
4Sn
The redox profiles of tin, iron and sulfur at the float glass surface were studied on purposely cut samples by EPMA, to determine concentrations, and micro X-ray analysis to characterize the state of oxidation. The common resolution of ~1 μm of the two methods was well suited to analyze concentration profiles that extend over tens of microns. Analysis focused on 10 mm thick glass plates to make observation easier. Going inward from the surface to the bulk of the glass, the main characteristics of the concentration profiles are: - after a diffusion-driven decrease, the tin concentration shows a local maximum at a depth of 10 μm, the so-called tin hump, and then vanishes at a depth of ≈40 μm. XAS analysis shows that stannous ions, that dominate in the shallow layers of the glass, switch to stannic ions in the tin hump region. - the iron concentration decreases, passes through a minimum in the tin hump region, prior to progressively increasing and, after a hump which appears as the extent-end of the tin profile, reaches its bulk value. The inward diffusion of the Fe 2+ ions (glass
4þ
2−
þS
These conclusions are stressed to be quite general since the composition of the float glass that was examined herein is quite commonly encountered. Acknowledgments We acknowledge friendly and enlightening discussion with MarieHélène Chopinet and Patrick Garnier. We are grateful to the machine and beamline groups at SLS-PSI whose outstanding efforts have made these experiments possible. References [1] [2] [3] [4] [5] [6] [7] [8]
5. Conclusion
6þ
þ S →4Sn
[9] [10] [11] [12] [13]
[14] [15] [16] [17] [18] [19] [20] [21]
G.B. Cook, R.F. Cooper, J. Non-Cryst. Solids 249 (1999) 210. J.A. Howell, J.R. Hellmann, C.L. Muhlstein, J. Non-Cryst. Solids 354 (2008) 1891. J.S. Sieger, J. Non-Cryst. Solids 19 (1975) 213. Y. Hayashi, K. Matsumoto, M. Kudo, J. Non-Cryst. Solids 282 (2001) 188. G.H. Frischat, C. Müller-Fildebrandt, D. Moseler, G. Heide, J. Non-Cryst. Solids 283 (2001) 246. G.H. Frischat, C.R. Chim. 5 (2002) 759. K.F.E. Williams, C.E. Johnson, J. Greengrass, B.P. Tilley, D. Gelder, J.A. Johnson, J. Non-Cryst. Solids 211 (1997) 164. K.F.E. Williams, C.E. Johnson, O. Nikolov, M.F. Thomas, J.A. Johnson, J. Greengrass, J. Non-Cryst. Solids 242 (1998) 183. T.-J. Wang, H. Zhang, G. Zhang, T. Yuan, J. Non-Cryst. Solids 271 (2000) 126. T.-J. Wang, Glass Technol. 38 (1997) 104. P. Lagarde, A.-M. Flank, J. Jupille, H. Montigaud, J. Phys. Conf. Ser. 190 (2009) 012079. F. Farges, Y. Lefrère, S. Rossano, A. Berthereau, G. Calas, G.E. Brown Jr., J. Non-Cryst. Solids 344 (2004) 176. A.-M. Flank, G. Cauchon, P. Lagarde, S. Bac, M. Janousch, R. Wetter, J.-M. Dubuisson, M. Idir, F. Langlois, T. Moreno, D. Vantelon, Nucl. Instrum. Methods Phys. Res. Sect. B 246 (2006) 269. Z. Liu, K. Handab, K. Kaibuchi, Y. Tanaka, Jun Kawai, J. Electron. Spectrosc. Relat. Phenom. 135 (2004) 155. M. Wilke, F. Farges, P.-E. Petit, G.E. Brown Jr., F. Martin, Am. Mineral. 86 (2001) 714. M. Wilke, F. Farges, G.M. Partzsch, C. Schmidt, H. Behrens, Am. Mineral. 92 (2007) 44. M. Wilke, P. Jugo, K. Klimm, J. Susini, R. Botcharnikov, S.C. Kohn, M. Janousch, Am. Mineral. 93 (2008) 235. F. Lamouroux, N. Can, P.D. Townsend, B.W. Farmery, D.E. Hole, J. Non-Cryst. Solids 212 (1997) 232. P.D. Townsend, N. Can, P.J. Chandler, B.W. Farmery, R. Lopez-Heredero, A. Peto, L. Salvin, D. Underdown, B. Yang, J. Non-Cryst. Solids 223 (1998) 73. V. Magnien, D.R. Neuville, L. Cormier, B.O. Mysen, V. Briois, S. Belin, O. Pinet, P. Richet, Chem. Geol. 213 (2004) 253. D.A. McKeown, I.S. Muler, H. Gan, I.L. Pegg, W.C. Stolte, J. Non-Cryst. Solids 333 (2004) 74.