Exafs and Nexafs Studies of Cation Environments in Oxide Glasses

X-ray absorption spectroscopic studies of cation environments in oxide glasses are selectively reviewed. New results are presented on K and Yb environments in silicate glasses and on Fe in silicate melts at temperatures up to 1173 " K. INTRODUCTION Recent investigations of oxide glasses and crystalline model compounds by x-ray absorption spectroscopy (XAS and the acronyms EXAFS for extended fine structure and NE)(AFS or XANES for near-edge structure) have provided useful and sometimes unique information on the structural environments of network-forming and network-modifying cations. Synchrotron-based XAS is well suited for studying the local structural environment and bonding of cations in amorphous materials that cannot be probed directly by standard spectroscopic or scattering methods (e.g. Na, Mg, K, Ca, Zr) or which are present in small concentrations (e.g. transition metals, rare earth and actinide elements a t 100 2000 ppm). To date, several dozen XAS studies of cations in oxide glasses have reported information on local coordination environments of network formers such as Si, Al, Ga, and Ge and of network modifiers such as Na, K, Ca, Ti, V, Fe, Zn, Zr, Pb, and U. This work has provided new insights about oxide glass structure and structure-property-composition relationships which shed light on processes such as homogeneous nucleation, viscous flow, cation diffusion, and corrosion behavior of glasses. This paper presents a brief overview of recent XAS studies of silicate and oxide glasses which is selective because of space limitations. Also reported are new results on the structural environments of K and Yb in silicate glasses under ambient conditions and of Fe in silicate melts at temperatures up to 1173" K. XAS STUDIES OF SILICATE AND OXIDE GLASSES: AN OVERVIEW EXAFS spectroscopy has found wide application to the study of amorphous materials because of its sensitivity to short-range order and its element specificity. Several reviews of synchrotron XAS studies of oxide glasses have been published since 1979 (1-9). The discussion below will highlight only a portion of this work. Although of great utility in structural studies of glasses, EXAFS analysis of amorphous materials is limited in important ways (10-16). A major limitation occurs when the distribution of interatomic distances about an absorber is asymmetric and when this effect is not explicitly included in the model used to extract distance and coordination number information from the EXAFS spectra. This effect has not been a major problem when the absorber is strongly bonded to four or six oxygen ligands, but i t can be a significant problem when an absorber is weakly bonded to more than six ligands. In the latter case, EXAFS contributions of the longer bonds in a distorted p* lyhedron can be lost, and the coordination number and interatomic distance from W S analysis may be less than the correct values. This topic will be considered only briefly in the examples cited below. Another major limitation of EXAFS analysis, and one of the reasons for its sensitivity to short-range order, is the loss of low k data due to multiple-scattering effects and the consequent loss of information about more distant atomic shells around the absorber. Network-Forming Elements One of the first EXAFS studies of an oxide glass using synchrotron radiation was that of Sayers et al. in 1972 (17) on the network former Ge in GeO, glass. The results of this and later DCAFS studies of vitreous GeO, (18-20) are consistent with structural models of GeO, glass from x-ray and neutron scattering studies (see 21, 22), i.e. a relatively well-ordered, three-dimensional network of corner-shared GeO, tetrahedra [d(Ge-0) = 1.74 A] with a relatively narrow distribution of Ge-0-Ge angles centered a t 130'. EXAFS studies of GeOiNa,O glasses (19) produced slightly longer Ge-0 distances (up to 1.77 A) which were interpreted as evidence for some Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:19868125 C8-662 JOURNAL DE PHYSIQUE six-coordinated Ge. An alternative explanation is that the bonding of Na to the bridging oxygen of some Ge-OGe linkages in these glasses causes a lengthening of Ge-0 bonds, without effecting a change in Ge coordination number. The EXAFS study of Ge and Ga in glasses of composition NaGaSiaO,, NAGe,O,, and NaGaGeaO, (23) was also interpreted as indicating small amounts of GeO, and GaO, octahedra in these glasses, based on slightly longer than expected Ge-0 and Ga-0 distances in comparison to those in GeO, (a-quartz) and pGa,O,, respectively. Again, this interpretation is not necessary because Na bonded to bridging oxygens should cause a lengthening of Ge-0 and Ga-0 bonds in adjoining GeO, and GaO, tetrahedra relative to these other structure types. A Ga K-EXAFS study of NaGaSiO, glasses synthesized at 1 and 25 kbar pressure (24) showed only fourcoordinated Ga. These EXAFS studies are consistent with or can be reinterpreted in terms of Ge and G a acting as network formers in the above glass compositions. The only evidence for some six-coordinated Ge in oxide glasses comes from Cox and McMillan's EXAFS study of Li,O-GeO, glasses (I) where up to 25% of the Ge is thought to occur as GeO, units based on two-shell fits giving Ge-0 distances of 1.71 (GeO,) and 1.84 A (GeOJ. These studies also illustrate one of the major advantages of the EXAFS method in studying amorphous materials relative to the more widely used x-ray or neutron scattering methods, i.e. the ability of EXAFS analysis to distinguish between similar bond lengths involving different elements that would normally be unresolved in x-ray radial distribution functions (e.g. Ge-0 vs. Al-0 and Ga-0 vs. Si-0 in amorphous N d G e , O , and NaGaSi,O,, respectively) (see e.g. 23). Si and Al are more common network-forming elements than Ge and Ga in geochemically or technologically important oxide glasses. However, they are more difficult to study by XAS because a high-vacuum environment and photoemission methods are generally required due to the long wavelength of the x-rays. One of the first XAS studies of Si in oxide glasses was carried out by Greaves e t al. (25) for the compositions SiO,, Na,Si,O,, and Na,CaSi,O,, using transmission methods on blown, thin glass films. Fits of the EXAFS data yielded Si-0 and Si-Si distances of 1.61 and 3.17 A respectively, and an Si-0-Si angle of 160°, which are typical of corner-shared SiO, tetrahedra; the firstshell DebyeWaller factors for the glasses are the same as for a-quartz, indicating little static disorder around Si in these glasses. NEXAFS and EXAFS photoelectron yield studies of A1 in silicate glasses from the Na,O-Al,O,-Si0, system (26,27) showed that Al remains four-coordinated with increasing Al/Na ratio a t least up to a ratio of 1.6. The A1 near-edge spectra of several crystalline model compounds with Al in fouror six-coordinated sites and of one model compound with Al in both types of sites, show features that are clearly identifiable with the different Al sites (26). Al NJXAFS spectra of the glasses show only a single K-edge maximum a t 1566 eV, typical of NO, tetrahedra. The A1-0 distances of 1.77 A in all glass samples are consistent with EXAFS-derived coordination numbers near 4.0. The small first-shell Debye-Waller factors for the glasses indicate little static disorder around Al. These results help constrain structural models for viscous flow and cation diffusion in sodium aluminosilicate melts. Transition Elements One of the major emphases of EXAFS studies of oxide glass structure has been determination of the local coordination environment of transition metals (6,9,28-47). The first such EXAFS study was by Brown et al. (28) on Fe in glasses of composition KFeSi,O,, NaFeSi,O,, and NaFeo,,Alo,$i,O,. Using phase shift and amplitude information from chemically similar crystalline model compounds (e.g. NaFeSi,O,), Fe-0 distances of 1.88, 1.88, and 1.92 A, respectively, were derived for these glasses. These values are consistent with FeS+ in tetrahedral coordination (Tea+) as the predominant Fe species, indicating that Fe acts as a network former in these glasses. NEXAFS spectra show an intense pre-edge feature (Is to 3d transition) for each glass, indicative of tetrahedral Fe. In contrast, F e N E W S of the crystalline model compound (NaFeSi,O, acmite), with only Te" , shows a very weak pre-edge feature (see ref. 48 for additional examples). A 6 T e Mossbauer study of the same glass samples also indicated Tea+, with less than 5% TVFe2+. Other EXAFS/NEXAFS studies of Mn, Fe, Co, and Ni in oxide glasses (29-33) give similar information about these elements. The high-resolution Fe K-NEXAFS study of several oxide glasses by Calas and Petiau (33) is a good example of the use of near-edge fine structure to determine oxidation state and site geometry of transition elements. TiO, is an important additive to silicate glasses because i t lowers thermal expansion and serves as a nucleating agent. XAS studies of Ti in TiOiSiO, glasses prepared by flame hydrolysis (34,35) showed that the coordination environment of Ti* varies with TiO, content. At less than 0.5 wt% TiO,, only WTi is present, whereas just above 0.5 wt%, T i is dominant. W T i / T i increases with increasing TiO, content up to about 15 wt% where TiO, crystallizes. Tbe intensity of the Ti pre-edge feature is sensitive to site geometry and was used in this work to estimate WTi /Vi . In contrast, an XAS study of TiOiSiO, sol-gel glasses with TiO, contents ranging from 4.5 to 19 wt% (36) found only T i ' + and suggested that increases in the intensity of the Ti pre-edge feature are caused by shortening of Ti-0 bonds and ordering of TiO, units. An interesting neutron diffraction and XAS study of the Ti environment in a K,TiSi,O, glass prepared by standard melt quenching methods (37) indicates the presence of TiO, square-based pyramids. Thus, it is possible that Ti can adopt a t least three different coordination environments in oxide glasses depending on composition and possibly on glass preparation method, although the oxidation state was found to be 4+ in each of these studies. The few XAS studies of vanadium in oxide glasses (38-41) indicate that, like Ti, V can exist in more than one type of local environment. In contrast with Ti, the oxidation state of V can vary with the redox state of the melt. At the 1000 ppm level in SiO, glass, V* and V6f were found to be tetrahedrally coordinated (38). In vanadium phosphate glasses, VO, square pyramids involving both V4+ and V6+ are suggested from NEXAFS and EXAFS spectra (39). In amorphous V,O, (40,41) V is thought to exist in distorted VO, bipyramids arranged in layers, based on modelling of the NEXAFS spectra using multiple-scattering theory. One of the more unique applications of XAS to oxide glasses is the study of nucleating elements such as Ti, Zr, and Zn in heat-treated alurninosilicate glasses (1, 9, 4244). Dumas and Petiau (43) succeeded in detecting transition-state complexes involving ZrO, octahedra sharing an edge with Si04 tetrahedra in a heat-treated cordierite-composition glass. Although they did not detect any transition-state complexes involving Ti or Zn in similar heat-treated cordierite glasses (44), they found that the local environment of T i changes progressively from tetrahedral to octahedral with heat treatment and that A12Ti05 is the first crystalline phase containing Ti. In contrast, Zn remains tetrahedrally coordinated throughout the heat-treatment and nucleation process in these glasses. Zr EXAFS studies have also been carried out on sodium aluminosilicate glasses containing Zr at 2000ppm (45) and in sodium zirconium silicate glasses (46). In both studies Zr was found to occur predominantly in octahedral coordination. Alkali and Alkaline Earth Elements Na, K, and Ca are the most important network modifiers in oxide glasses and melts of geochemical and techne logical importance and have large effects on physical properties. The structural environments of these cations in glasses are difficult to study directly and quantitatively by any other means except XAS. However, because these cations are typically coordinated by more than six nearest-neighbor oxygens and are weakly bonded relative to network-formers and transition elements in crystalline and amorphous oxides, caution must be used in EXAFS studies of them. The problem of asymmetric distance distributions in the first shell about these cations can be significant. Because of the low energy of the Na K-edge (-1057 eV), high-vacuum, photoelectron yield methods are usually employed for Na EXAFS work. EXAFS studies of K and Ca can be done on non-vacuum beam lines, but fluorescence methods and a He beam path are generally used to maximize S/N because of the relatively low K-edge energies (3608 and 4050 eV, respectively). The environment of Na in several silicate glass compositions has been studied by EXAFS methods (25,48-50). The work of Greaves e t a/. (25,48,49) yielded Na-0 distances of 2.3 to 2.4 A and Na coordination numbers of 5 to 6 in Na.$iO, and Na2CaSi50, glasses. These distances are shorter and the coordination numbers are larger than those observed in crystalline Na$i,O, and NaSiO, (2.40 A, CN= 5) by x-ray diffraction methods. It was concluded that the Na environments in these glasses are different because of bulk compositional differences. The Na total electron yield study of McKeown e t a[. (50) considered glasses in the system Na,O-Al,O,-SiO,, including the composition Na.$i,O,. In all glasses studied, Na-0 distances average 2.60 A and Na coordination numbers range from 5 to 7.5, depending on composition. Comparison of the two independent data sets for Na,Si,O, glass shows a significant difference in chi-functions and an 0.3 A difference in Na-0 distance from EXAFS fitting. It is suggested that the different glass preparation methods (blown thin film vs. melt quenching) may be responsible for these differences. Na total electron yield EXAFS study of crystalline KNa&l,Si,O, (nepheline) (50) produced an Na-0 distance of 2.57 A and a Na coordination number of 5.5, compared with 2.63 A and 8.0 from x-ray refinement (51). The closest five oxygens have an average Na-0 distance of 2.56 A from the x-ray work. This comparison suggests that EXAF'S contributions from the outer three oxygens surrounding Na a t distances of 2.69, 2.72, and 2.84 A have been lost. This is not a surprising result in light of the highly distorted Na site in nepheline. The quality of the EXAFS data for nepheline does not justify the fitting of a second shell of oxygens to account for the missing oxygen ligands. The first-shell Debye-Waller factors for the glasses are 213 to 113 less than that of crystalline nepheline, suggesting less static disorder around Na and a more regular Na environment in the glasses. The network-modifier Ca has been studied by XAS in several glasses including CaAl,Si,O, (52,54) CaMgSi,O, (52,54), CaTiSiO, (52), -CaSiO, (53), and Ca&ig&l2Si,O, (55), as well as in model compounds. Ca KN E W S spectra (52,53) show a rich diversity of fine structure with variations in Ca environment. Attempts to model the Ca near-edge structure by multiple-scattering calculations (53) did not produce good matches with exC8-664 JOURNAL DE PHYSIQUE periment. The detailed Ca EXAFS study of CA2Si20, (anorthite) and CaMgSi,06 (diopside) glasses and crystals (54) derived Ca-0 distances of 2.60 and 2.64 A for anorthite crystal and glass, respectively, and 2.50 and 2.63 A for diopside crystal and glass, respectively, using multiple-shell fits. Ca is coordinated by 8 and 7 oxygens in anorthite and diopside, respectively, and was modeled with these coordination numbers in the glasses. The EXAFS-derived bond lengths of the crystals are within 0.1 A of the crystal Ca-0 bond lengths from x-ray diffraction. This study suggests a significant asymmetry of Ca-0 distances in the glasses and crystals, producing Ca-0 distances 0.13 to 0.23 A too short if a single-Gaussian fit is used. It also helps explain the higher density of diopside crystal relative to the glass. A similar fitting procedure was used in the Ca EXAFS study of a model basaltic glass (55), producing an average Ca-0 distance of 2.48 A and a Ca coordination number of 9. Potassium is a considerably larger network-modifying cation than Na or Ca and typically occupies 6-12 coordinated sites in crystalline compounds. The near-edge structure of K has been studied in a number of silicate glasses and crystalline model compounds of various compositions (56,57). Figure 1 compares K NEXAFS spectra for the crystalline compounds KZA14[Si$U2]020(OH)4 (muscovite, CN = 12), Na3KM4Si401, (nepheline, CN = Q), KAlSi,O, (orthoclase, CN = 9), KAlSi20, (leucite, CN = 6) and four mixed-alkali silicate glasses along the join NaAISiaO, (Ab,,) KAISiaO, (Or,). The lack of similarity in near-edge structure between orthoclase crystal and glass suggests that the K environment in these compounds is different. Instead, the K environment in these glasses is similar to that in leucite, based on the similarity of near-edge structures. Using leucite as a model compound for phase and amplitude information, EXAFS spectra of these glasses were fit to models yielding K-0 distances of 3.00 to 3.06 A and K coordination numbers of 8.9 to 10.4. Plots of these values vs. composition (Fig. 2) show distinct maxima for the Ab,Or, composition. This new result (57) provides the first directlymeasured structural constraint on models of the mixed-alkali effect in these glasses. Fig. 1. Comparison of potassium K-NEXAFS for model compounds and glasses. Fig. 2. Variation of K-0 distance and K CN in potassium glasses as a function of orthoclase content. Rare Earth and Actinide Elements The structural environments of P b (58), the rareearths La, Gd, and Yb (59), and the actinides U (6,60,61) and Tc (62) in oxide glasses have been studied by XAS methods. P b Lm EXAFS measurements on PbO-PbCl, glasses (58) were interpreted as indicating PbO,Cl, clusters. U in borosilicate glasses of importance in nuclearwaste storage was found to occur as U6+ (6). Reflection EXAFS methods were used to study the corrosion resistance of similar U-containing glasses (61) and showed an increase of U in the surface region of these glasses following water corrosion. The structural environment of Yb has been studied in a series of glasses in the system Nap-AI,OiSi0, (59). The results of EXAFS measurments on the Yb Lm-edge indicate that in albite (NaAlSi,OJ glass the Yb-0 distance is 2.14 A with a CN of -8. In the depolymerized glasses studied, a peralkaline composition (Naa,@S$O,) and sodium trisilicate (Na>i307), the Yb-0 distance and CN increased to 2.22 A and -10, respectively. The raw and backtransformed chi functions for these glasses in Figures 3 and 4 clearly show these siniilarities and differences. The similarity of the distances in the partially depolymerized and more highly depclymerized glasses suggests that the Yb environment does not change gradually as the structure is depolymerized, but abruptly at some composition between the albite and peralkaline compositions. The effect of fluorine on the local structure around Yb was also studied. Two wt% F added to the albite glass increased the Yb-0 distance to 2.21 A with a CN of -12. However, there is no evidence for Yb-F bonds in the glass. Thus the major effect of F on the local Yb environment appears to be due only to the depolymerizing effect of F on the aluminosilicate network structure. 4 6 8 10 12 4 6 8 10 12 wove vector 1;-3 Wave Vector Fig. 3. kS ~ ( k ) functions for albite (bottom), sodium, trisilicate (middle) and albite f F (top) glasses. Fig. 4. Backtransformed chi functions for albite (solid), sodium trisilicate (dashed) and albite + F (dotted) glasses. Note the large phase and amplitude differences between polymerized and depolymerized glasses. F e in Silicate Mel t s The structural environment of cations in an oxide melt is not necessarily the same as in the quenched melt, although it is often assumed to be. In order to study this problem directly for geochemically relevant melts, we have made the first E X A F S / N E W S measurements on Fe-containing silicate melts and crystalline model compounds at high temperatures (63). The data were collected using fluorescence detection methods and an evacuable furnace assembly consisting of a Ta-wound resistance heater and BN sample boat. A selection of Fe KN E W S spectra for some of these compounds at different temperatures is shown in Figures 5 and 6. The melt compositions studied were NaFeSis08 and v e S i , 0 8 in which the oxidation state and structural environments of Fe were unknown. Comparison of the near-edge spectra of the melts with model compounds indicates that Fe in the two melts is predominantly Fe2+ in tetrahedral coordination. The Fe K-NEXAFS spectrum of FeAI,O, (hercynite) with Y e 2 + best matches the near-edge spectra of the glasses and melts. The background-subtracted EXAFS spectra of the Na-Fe and K-Fe silicate melts a t 1123 ' and 1173" K, respectively, are shown in Figure 7 together with the spectrum of hercynite collected at 1073 " K. The data quality is surprisingly good for this temperature and the conditions under which the experiments were performed. The relatively close correspondence of features in the three spectra is consistent with the conclusion made above from NEXAFS spectra. Further EXAFS work on melts at high temperatures is underway. CONCLUSIONS XAS studies of cation environments in oxide glasses during the past 10 years have provided important structural information complementary to that from x-ray and neutron scattering and other spectroscopic methods. In certain cases, the information provided by XAS is unique. One of the major inferences that can be drawn from EXAFS/NEXAFS work on network modifiers in oxide glasses is that these cations occupy less distorted sites in the glasses than in chemically similar crystalline phases, where local structural environments are typically dictated by the network-forming cations. There is a clear need for improvements in EX4FS data-fitting procedures for asymmetric distance distributions such as those involving higly-coordinated and weakly-bonded networkmodifying cations In this regard, the new method of Stern e t al. (64) appears promising. Also crucial to further advances in XAS studies of oxide glasses are improved multiple-scattering models with which NEXAFS spectra can be synthesized for different local environments. Recent attempts in this area ( e . ~ ; 65) illustrate this need. Results from the first application of XAS to structural studies of molten oxides are promising. C8-666 JOURNAL DE PHYSIQUE N A ~ F E S ~ O ~ M E L T ~ ~ Z ~ ' K Na2F~Sl3O8 Guss-873°K N A ~ F E S I ~ O ~ GLASS-29SoK I I , I I I I . 1 1 1 1 1 1 , ,