Non-bridging oxygen sites in barium borosilicate glasses: results from 11B and 17O NMR Peidong Zhao, Scott Kroeker, Jonathan F. Stebbins * Department of Geological and Environmental Sciences, Stanford University, Stanford, CA 94305-2115, USA Received 5 April 2000 Abstract The concentration and structural role of non-bridging oxygen (NBO), defined as oxygen bonded to one framework cation, play a major role in the physical properties of glasses and liquids. In this study, the distribution of NBO connected to either silicon and boron was investigated in a borosilicate glass system using 17O magic-angle spinning (MAS) and triple quantum (3QMAS) nuclear magnetic resonance. In order to obtain better separation of di€erent 17O NBO resonances, barium borosilicate glasses were chosen and measured by MAS and 3QMAS. The two-dimensional 3QMAS spectra show clear resolution between peaks for Si–NBO and B–NBO, allowing their relative proportions to be at least roughly quantified for the first time. The NMR parameters for these sites were also determined and are con- sistent with previous studies of binary barium borate and silicate systems. Site populations are significantly di€erent from predictions of conventional models based on alkali borosilicates, which suggests a more disordered nature for these barium borosilicate systems. In order to explain the anomalously high NBO concentrations, we hypothesize that a few percent of the oxygens may have three network cation neighbors, forming ÔtriclustersÕ. Ó 2000 Elsevier Science B.V. All rights reserved. 1. Introduction Borosilicate glasses are used in a wide range of technological applications, from chemical con- tainers and piping to fiber composites and radio- active waste storage. As in any oxide glass, one of the most important aspects of the atomic-scale structure is the concentration and distribution of oxide ions that are bonded to only a single net- work-forming cation (Ônon-bridgingÕ oxygens or NBO) instead of two network formers (ÔbridgingÕ oxygens or BO). Higher NBO contents generally lower the viscosity of the precursor glass melt but contribute to ease of corrosion; higher BO con- tents often lead to harder, more durable glasses that must be melted and worked at higher temperature. The structure of boron-containing oxide glasses has been studied for decades, and detailed models of their structure have been developed, in large part from spectroscopic studies, particularly using NMR. The most complete models have been based on ÔwidelineÕ, 11B NMR on static (non-spinning) samples, such as that for sodium borosilicates by Dell and Bray [1]. These have been tested and expanded by numerous more recent studies using high resolution, magic-angle spinning (MAS) NMR Journal of Non-Crystalline Solids 276 (2000) 122–131 www.elsevier.com/locate/jnoncrysol * Corresponding author. Tel.: +1-650 723 1140; fax: +1-650 725 2199. E-mail address:
[email protected] (J.F. Steb- bins). 0022-3093/00/$ - see front matter Ó 2000 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 3 0 9 3 ( 0 0 ) 0 0 2 9 0 - 8 on 11B,17O,27Al,29Si and other nuclides [2]. 17O NMR, either MAS or the higher resolution method of Ôtriple quantumÕ MAS (3QMAS), provides a particularly useful complement to information on cation environments, and can clearly distinguish a number of the critical oxygen species, for example, the bridging oxygens Si–O–B, Si–O–Si, Al–O–Al and B–O–B as well as NBO [3–6]. In a recent pa- per, we applied these methods to directly charac- terize NBO sites in a number of binary Na- and Ba-borate glasses [7], following up previous work on NBO sites in silicate glasses [6,8,9]. Here we address the question of whether NBO in ternary borosilicates are preferentially bonded to Si sites, as predicted by most conventional models, or whether they are more randomly dis- tributed among B and Si sites. As a starting point, we use the model of Dell and Bray for sodium borosilicates [1]. Based essentially on NMR studies of boron coordination number and the fraction of three-coordinated boron sites that are Ôsymmetri- calÕ (all BO) vs ÔasymmetricalÕ (one or more NBO), the model explicitly predicts NBO distribution on borate and silicate sites as a function of composi- tion. The latter is expressed as the variables R, the molar ratio of modifier oxide to B2O3 and K, the ratio of SiO2 to B2O3. Considering a series of stoichiometric reactions among structural species as modifier oxide is added, the model suggests compositional regions in which no NBO should be present (I and II in Fig. 1), and regions where NBO should be bonded only to Si (III in Fig. 1) or Si and B (IV and V). Because of the probability of the overlap of the 17O NMR peaks for NBO and BO in the sodium borosilicate system, we have chosen to study the corresponding barium system, as the larger, higher field strength cation is known to shift NBO resonances to much higher chemical shift values [6,10,11]. 2. Experimental procedure 2.1. Sample preparation Barium borosilicate samples were synthesized from barium carbonate and 17O-enriched B2O3 and SiO2 as described previously [7,12]. These re- agents were carefully mixed and transferred to a platinum crucible, and heated to 700°C for decarbonation. The temperature was then raised to 1100°C for 3 h to melt the mixture, followed by quenching the crucible in cold water and grinding into a coarse powder. A flow of argon was main- tained throughout the heating process to avoid the exchange of 17O with air. Co3O4 (0.1 wt%) was added to the samples to reduce the spin-lattice relaxation time. No crystals were found in the re- sulting glasses by microscopic observation at 400�. Samples were stored in vacuum desiccators even though no hydration was found after an ex- posure to ambient air of several days. Samples were labeled in the same manner as for sodium borosilicates studied previously (see Table 1) [4,7]. Glass compositions were chosen to be well outside the known liquid–liquid immiscible region [12], and to lie in regions where conventional models [1] predict significantly di€erent speciation. The dif- ferences (less than 1%) between nominal and ob- served weight losses during the glass syntheses indicate that volatility of these samples was not significant, and thus compositions should be close to nominal. The compositions of samples are summarized in Table 1, and plotted in Fig. 1. For comparison, a binary barium borate (BB46) and a barium silicate glass (BS55) are also included in Table 1 and Fig. 1. Fig. 1. Compositions of the samples studied here (dots), shown in mole% in the BaO–B2O3–SiO2 system. Solid lines show compositional regions as defined by the structural model of Dell and Bray [1] for the corresponding Na-containing system. P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 123 2.2. NMR spectroscopy The 17O NMR spectra were measured at 14.1 T and 9.4 T, using Varian Inova 600 and modified Varian VXR-400S spectrometers at Larmor fre- quencies of 81.31 and 54.22 MHz, respectively. A 3.2 mm Varian/Chemagnetics MAS probe and a 5 mm Doty scientific MAS probe were used to ac- quire spectra, with sample spinning rates of 16 and 12 kHz, respectively. The recycle delays for both MAS and 3QMAS spectra at both fields were optimized for signal-to-noise ratio and were usu- ally 0.5–1 s. At 9.4 T, MAS spectra were obtained with one-pulse and 90–180° Hahn-echo pulse se- quences, respectively. In one-pulse experiments, a short, high-power single pulse (15° tip angle) with a dead time of 10–20 ls was applied [13]. In the Hahn-echo experiments, the delay between the two pulses was synchronized with the spinning speed to avoid rotational artifacts and incomplete refocus- ing of the echo. 3QMAS spectra were measured at both fields using a shifted-echo pulse sequence described previously [14,15]. The first and second pulses (excitation and reconversion pulses), with radiofrequency (rf) power of 70 kHz at 9.4 T and 120 kHz at 14.1 T, were optimized to give the highest intensity of the 3QMAS signal. The typical lengths of the first and second pulses were around 6 and 2 ls at 9.4 T, and 2.8 and 0.8 ls at 14.1 T. An echo was formed by the use of a third, selective soft pulse and the delay between the second and third pulse was set to the integral multiple of the rotor period. All 17O NMR data were processed using the computer program RMN (P.J. Grandi- netti, Ohio State University). The MAS spectra were apodized with 100–150 Hz line broadening. For the 3QMAS spectra, apodizations were 100–150 Hz for the MAS dimension and 10–30 Hz for the 3QMAS dimension. A shear transforma- tion was applied during processing of all of the 3QMAS spectra. All two-dimensional spectra were plotted with 18 contour lines between 10% and 100% of the maximum intensity, with extra con- tour lines at 6% and 8% added to accentuate the relatively low intensity of the NBO peaks in samples with lower modifier contents. All 17O NMR spectra were referenced to external, 20% 17O-enriched H2O. In glasses, NMR peaks in 3QMAS spectra contain signals from many similar sites with ranges in NMR parameters such as the quadrupolar coupling constant, CQ, and the isotropic chemical shift, diso. As described previously [3,4,15], mean values for diso and the quadrupolar product PQ (PQ � CQ) can be estimated from the centers of gravity of the peak in the MAS and 3QMAS di- mensions by solving two linear-independent equations, and are reported in Table 2. 11B NMR spectra were acquired at 14.1 T (192.4 MHz) using a 3.2 mm Varian/Chemagnetics MAS probe. Single pulse acquisition was used with spinning rate of 15 kHz and short rf pulses 0.2 ls, chosen not to exceed one-sixth of the Ôsolution 90°Õ pulse duration [16,17], as calibrated using a 1 M aqueous solution of boric acid. That this pulse length did not jeopardize the accurate quantitation of boron sites populations with di€erent quadru- polar properties was verified by characterizing the nutation behavior of each solid sample. A relax- ation delay of 1 s was used. Spectra are refer- enced to the conventional standard, neat BF3�O(CH2CH3)2, which resonates at )19.6 ppm relative to the observed secondary standard of 1 M boric acid [18]. To obtain accurate values of the site popula- tions giving rise to the central transition peak intensities measured by integration of the base- line-corrected spectra, it is necessary to account explicitly for both the contribution of satellite transition magnetization to the observed center- band, and the distribution of central transition magnetization into spinning sidebands. Calculated Table 1 Nominal compositions (in mole%) of the 17O-enriched glass samples. R is the molar ratio BaO/B2O3, and K is the ratio SiO2/B2O3 Samples BaO B2O3 SiO2 R K BBS252 24 52 24 0.46 0.46 BBS352 28 48 24 0.58 0.50 BBS433 40 30 30 1.33 1.00 BB46 40 60 – 0.67 – BS55a 45 – 55 – – a This barium silicate glass was studied previously, and the synthesis route and properties were described in detail elsewhere [6]. 124 P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 correction factors [19] were thus applied to the raw intensity data, resulting in small but signifi- cant changes. Trigonal boron CQ values used in this procedure were derived from typical litera- ture value for BO3 sites (CQ ˆ 2:6 MHz) [1] obtained from lineshape fitting, whereas those for BO4 species were estimated from the extent of satellite transition spinning sidebands (CQ ˆ 500 kHz). 3. Results 3.1. 11B and 17O MAS spectra A typical 11B MAS NMR spectrum acquired at 14.1 T is shown in Fig. 2. There are two well- resolved peaks like those that can be observed for most borate and borosilicate glasses at this mag- netic field. The broader peak with larger CQ is due to the trigonal boron (BO3), while the sharp peak is from tetrahedral boron (BO4). The trigonal boron consists of ÔsymmetricÕ boron (sites with three BO) and ÔasymmetricÕ boron (sites with one or two NBO), which are not resolved for this series of glasses. Compared with spectra of sodium bo- rosilicate glasses [1,3,4,7], the tetrahedral boron peak is broader and the trigonal boron peak be- comes nearly featureless, suggesting more disor- dered boron local environments in barium borosilicate glasses. The proportions of trigonal and tetrahedral boron obtained as described above are listed in Table 3. The 17O MAS spectra of three barium borosi- licate glasses, measured at 9.4 T using the Hahn echo pulse sequence, are shown in Fig. 3. For comparison, the MAS spectrum of the binary glass sample (BB46) acquired at the same magnetic field is also shown [7]. In each, two well-resolved peaks were observed. By analogy to the Ba–silicate and Fig. 2. A typical 11B MAS spectrum of barium borosilicate glasses at 14.1 T. In this and other one-dimensional spectra (Figs. 3 and 4), vertical scale (not shown) is signal intensity in arbitrary units. Horizontal scale is relative frequency in parts per million (ppm). Table 2 17O NMR parameters for NBO in glass samples BS55, BB46 and BBS433 derived from 3QMAS spectra measured at 14.1 T a Samples dMAS; ppm …�2† d3QMAS; ppm …�1† diso ppm …�2† PQ, MHz (�0:2) BS55b Ba–O–Si 156 )90 161 2.3 BB46 Ba–O–B 183 )111 196 3.7 BBS433 Ba–O–Si 153 )88 158 2.3 Ba–O–B 185 )112 197 3.6 a dMAS is the center of gravity in the MAS dimension and d3QMAS is that in the 3QMAS isotropic dimension. diso and PQ are the isotropic chemical shift and quadrupolar product, respectively. 17O NMR parameters for NBO obtained from BBS352 and BBS252 are the same as those from BBS433. b 17O NMR parameters are calculated from the 3QMAS spectrum at 9.4 T reported previously [6]. P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 125 Ba–borate glasses studied previously [8], the peak centered at 158 ppm was assigned to NBO sites, which in the borosilicates may include one or both of Ba–O–Si and Ba–O–B. The large peak centered at about 43 ppm in these spectra is quite broad, and is assigned to BO sites. This peak may contain unresolved contributions from several di€erent types of bridging oxygen (Si–O–Si, B–O–B and Fig. 3. 17O MAS spectra of glass samples measured at 9.4 T. Peaks marked by * are spinning sidebands. ÔBOÕ and ÔNBOÕ mark the peaks for bridging and non-bridging oxygens. Spectra are normalized to same maximum intensity. Table 3 Comparison between NMR results for boron and oxygen speciation and predictions of Dell and Bray model [1] a Samples B% NBO% Oxygens in TriclusterN3 N4 B–O–Ba Si–O–Ba total BBS433 Model 58.3 41.7 16.4 12.6 29.0 MAS 45 55 – – 26.0 1.3% 3QMAS – – 13:8� 5 16.3 30.1 – BBS352 Model 46.9 53.1 0 2.3 2.3 – MAS 48 52 – – 6.8 4.0% 3QMAS – – 2.8 2.3 5.1 – BBS252 Model 53.8 46.2 0 0 0 – MAS 52 48 – – 4.8 5.6% 3QMAS – – 2.5 1.6 4.1 – a Uncertainties are 1% unless otherwise noted. The fractions of oxygen in hypothesized tricluster sites are also shown. 126 P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 Si–O–B) sites, where boron may be 3- or 4-coor- dinated. 3.2. 17O 3QMAS spectra For 17O in borosilicate glasses, resolution better than MAS can be achieved by using 3QMAS spectroscopy [20]. Total projections in the isotro- pic dimension of 3QMAS spectra at 9.4 T for bridging and non-bridging oxygen regions are shown in Fig. 4. Distinct peaks or shoulders for di€erent types of bridging and non-bridging oxy- gen species can be seen, especially in samples BBS352 and BBS252, even though these peaks are not completely resolved. From sample BBS252 to BBS433, the intensities of the Si–O–Si and Si–O–B peaks increase, a trend consistent with the in- creased Si/B ratios (see the following for peak as- signment). Generally, in 3QMAS experiments, sites with large CQ values have low eciency of excitation and reconversion and thus have reduced relative intensities. The relative populations of non-bridging Ba–O–B sites may thus be somewhat underestimated, because they have higher CQ val- ues than Ba–O–Si sites (Table 2). In any case, significant populations of the two types of NBO sites are present in all three glasses. The 17O 3QMAS spectra of BB46 and BBS433 acquired at 14.1 T are shown in Fig. 5. Two well- resolved peaks were observed in the binary glass BB46, which, according to its composition and previous study, were assigned to bridging B–O–B and non-bridging Ba–O–B sites [6,7]. The 3QMAS spectrum of BBS433 has three major peaks cen- tered at )56, )88 and )118 ppm in the isotropic dimension. The peak at )56 ppm is very broad (ranging from )20 to )85 ppm), covering the range of the observed positions for Si–O–Si ()28 ppm), Si–O–B ()43) and B–O–B ()52 to )65 ppm) sites [3,4], and thus may contain contributions from all three. Compared with the well-resolved bridging oxygen peaks in sodium borosilicate glasses [4], the 17O NMR peaks for each type of bridging site in barium borosilicate glass are broader and more overlapped with one another because of the presence of the higher-field strength modifier cation. This indicates more disordered environments around the BO, consistent with similar observations for alkaline earth silicate and borate glasses [6,7]. Based on the comparison of the 3QMAS peak positions for Ba–O–B in BB46 and for Ba–O–Si in BS55 (indicated in Fig. 5), the assignment of peaks at )88 and )118 ppm to Ba–O–B and Ba–O–Si sites is straightforward (see Table 2). 3QMAS spectra of BBS433, BBS352 and BBS252 glasses acquired at 9.4 T are shown in Fig. 6. Three peaks, centered at )67, )88 and )118 ppm, are observed in these samples and are again assigned to BO and two types of NBO, Ba–O–B and Ba–O–Si, respectively. Fig. 4. Projections of 3QMAS spectra of BBS433, BBS352 and BBS252 at 9.4 T. The left side of the figure shows the BO region and the right side shows the NBO region. P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 127 It is interesting to note that sample BBS252, predicted to have no NBO by Dell and BrayÕs model for compositional region 1 (Fig. 1), in fact has a considerable concentration of these species, observed both in 3QMAS and MAS spectra. Even at this low modification level, the peaks centered at )88 and )119 ppm in the isotropic dimension of the 3QMAS spectrum also clearly substantiate the formation of NBO connected to both silicon and boron sites. In compositional region 3, represented by sample BBS352, the Dell and Bray model pre- dicts that all NBO should be associated with the silicon network only. However, again the 3QMAS Fig. 5. 17O 3QMAS spectra for glass samples BBS433 and BB46 acquired at 14.1 T. Here and elsewhere Ba–O–B and Ba–O–Si represent the NBO at boron and silicon, respectively. The black square …j† shows the position of the Ba–O–Si peak based on the previous 3QMAS spectrum of BS55 [6]. Hori- zontal scales here and in Fig. 6 are relative frequencies in ppm and contour vertical scale (see text) shows signal intensity in arbitrary units. 0 100 200 0 -100 100 200 BBS252 BBS352 -25 -50 -75 -100 -125 -150 0 100 200 BBS433 Isotropic Dimension(ppm) M AS D im en sio n(p pm ) BO BO Ba-O-B BO Ba-O-Si Ba-O-Si Ba-O-Si Ba-O-B Ba-O-B Fig. 6. 17O 3QMAS spectra for glass samples BBS433, BBS352 and BBS252 acquired at 9.4 T. The solid circles represent the estimated position of the triclusters resonances. 128 P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 spectrum (Fig. 6) demonstrates that both Ba–O–B and Ba–O–Si exist. From the Dell and Bray model, sample BBS433, in compositional region 4 (see Fig. 1), should have both NBO Ba–O–B and Ba–O–Si sites. The existence of two overlapped peaks cen- tered at )88 and )118 ppm in the isotropic di- mension is qualitatively consistent with this prediction. 3.3. Site populations The relative proportions of the total NBO sites were measured by integration of single pulse 17O MAS spectra and are listed in Table 3, along with the site populations predicted from Dell and BrayÕs model. Although the 3QMAS spectra are not inherently quantitative due to the dependence of the eciency of excitation and reconversion on CQ, the populations of species from 3QMAS spectra are also listed for comparison and agree fairly well with the MAS data. The population of NBO measured by MAS in BBS433 (26%) is slightly di€erent from the Dell and Bray model prediction (29%) but this is probably not signifi- cant in the given experimental error. In contrast, relatively large di€erences between the Dell and Bray model prediction and the MAS data were observed for both BBS252 and BBS352. Accord- ing to the model, NBO in BBS352 should only occur in the silicon part of the network with a concentration of 2.3%, compared with the ob- served value of 7%. Sample BBS252 has an even lower R value …Ba=B < 0:5†. In this compositional region, the ternary system is predicted to behave like the silica-free binary system irrespective of the SiO2 concentration, in that the BaO acts to create BO4 only and no NBO should be produced. However the MAS spectrum of BBS252 shows a considerable concentration of NBO (5% of total oxygen), which is analogous to the similar finding in alkali borosilicates by Raman spectroscopy [21]. The 3QMAS spectra further confirm the existence of both Ba–O–B and Ba–O–Si in BBS352 and BBS252 glasses. The proportions of trigonal and tetrahedral boron obtained from 11B NMR (N3 and N4) in BBS352 and BBS252 agree well with Dell and BrayÕs prediction, while the N3 and N4 of BBS433 are quite di€erent from the model. The reasons for the large di€erence is not clear, and is perhaps related to the field strength of Ba2‡ when com- pared to Na‡. In the conventional model of borate and borosilicate glasses, the addition of one mole of BaO should create a total of two moles of BO4 groups and/or NBO. When the observed propor- tions of NBO are considered, however, it seems that more NBO are present than allowed by this stoichiometry. This unexpected finding suggests that the presence of some other sort of oxygen site, as discussed in the following. 4. Discussion 4.1. 17O NMR parameters for NBO The 17O NMR parameters of Ba–O–B and Ba– O–Si sites in selected barium borosilicate glasses were calculated from their positions in the 3QMAS spectra and are summarized in Table 2. The good agreement of these data among the binary com- positions (BS55, BB46) and ternary glass BBS433 corroborates the assignment of the two types of NBO sites. Compared with NBO sites in sodium borate and borosilicate glasses, the isotropic chemical shift …diso† for Ba–O–B in barium bo- rosilicate is, as expected, shifted to much higher frequency, and is consistent with the same trend found in barium borates and oxides, as well as the other studies of the e€ects of cation field strength [6,7,11]. The relatively larger CQ value of Ba–O–B compared to Ba–O–Si suggests a more distorted charge distribution, perhaps related to the shorter bond to the network cation [7]. 4.2. Bridging oxygens Some information about network cation or- dering in barium borosilicate glasses can be ob- tained from 17O 3QMAS peaks for BO, although the increased disorder in these systems apparently decreases the resolution among Si–O–Si, Si–O–B and B–O–B peaks relative to that seen in the Na system [3,4,7]. Some existing models of borosilicate glass structure suggest the presence of P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 129 silica- and boron-rich regions [1,22]. In a glass such as BBS252, this would imply a relatively low in- tensity for the Si–O–B peak. From 3QMAS studies at 9.4 T of sodium borate and borosilicate glasses [4], B–O–B sites should have a peak centered at )73 ppm in the sotropic dimension with a full width of about 25 ppm. The Si–O–Si peak should be centered at )44 ppm with a similar or smaller width. If the Si–O–B sites are low in concentration because of intermediate-range network cation or- dering, a gap between the Si–O–Si and B–O–B peaks would be expected because the separation between Si–O–Si and B–O–B peaks is 30 ppm. However, 3QMAS spectra show no evidence of such a gap. The peak found in this region is un- derstood to be from Si–O–B linkages, implying that Si–B ordering is not prominent in BBS252. The same conclusion holds for BS352 and BBS433. In fact, a clear shoulder attributable to Si–O–B is visible in the isotropic projections for several samples (Fig. 4). The presence of both types of NBO peaks (Ba–O–B and Ba–O–Si) at three modification levels may also suggest a relatively high degree of mixing between B and Si networks. 4.3. NBO distribution Dell and BrayÕs model of glass structure was primarily based on observed boron speciation in sodium borosilicates [1,23–25]. The replacement of the alkali cation by Ba2‡ could cause significant local structural changes due to its higher field- strength and greater radius, and thus the barium borosilicate system may not necessarily show the same structural variations with composition. For example, in crystalline barium metaborate, the NBO interacts with three Ba2‡ instead of five Na‡ in sodium metaborate [26,27]. Therefore, the dif- ferences reported here between model predictions and observed NBO distribution could be the result of compositional e€ects and not due to inaccuracy in the model. Both 17O MAS and 3QMAS spectra provide clear evidence for the existence of both Ba–O–B and Ba–O–Si at three modification levels. The occurrences of Ba–O–B in BBS352 and Ba–O–B and Ba–O–Si in BBS252 do not match the pre- dictions of the Dell and Bray model for the cor- responding sodium-bearing system. In general, the addition of alkali or alkaline earth oxide triggers three competing processes: the conversion of tri- gonal to tetrahedral boron, and the formation of NBO connected to boron and/or silicon cations. Up to some level of modifier, the first process usually dominates, indicating its energetic favor- ability, while the numbers of NBO on both boron and silicon are minimized. In barium-containing glasses, as opposed to more well-studied alkali borosilicates, the higher field strength of the Ba2‡ cation promotes the formation of NBO and thus these species appear at a lower modifier content. Although most of the BaO is still used to convert BO3 to BO4 groups, the population of NBO is increased, perhaps leading to the observed in- compatibility with of the model prediction. 4.4. Triclusters The proportions of BO4 and NBO listed in Table 3 are obtained from integration of the 11B and 17O MAS spectra. The amount of oxygen needed to form BO4 and NBO is more than that which BaO could provide if one mole of BaO could create two moles of BO4 or of NBO. For example, in one mole of BBS352, 0.28 moles of BaO can create a total 0.56 moles of BO4 and/or NBO, but we determined a total of 0.65 moles of BO4 and NBO using the relation N4 �MB ‡ NNBO �MO; …1† where N4 is the fraction of boron in tetrahedral sites and NNBO is the fraction of oxygen present as NBO. MB and MO are the numbers moles of total boron and oxygen atoms in the sample. These extra BO4 or NBO must thus be accompanied by other types of oxygen sites to account for the ox- ygen stoichiometry requirement. An analogous situation has recently been discussed for calcium and sodium aluminosilicate glasses and liquids. Anomalies in trends of melt viscosity with com- position [28,29], and direct 17O NMR observations [8] both indicated the presence of a few percent of NBO even at fully ÔpolymerizedÕ compositions, where conventional models suggest that all oxygens should be bridging. These could be 130 P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 compensated by the formation of AlO5 or AlO6 groups, but available spectroscopic evidence indi- cates that these species are either absent or too low in concentration to explain the observed NBO concentrations. The likely alternative was sug- gested to be oxygens with three Al and/or Si tet- rahedral neighbors, designated as ÔtriclustersÕ. For the borosilicates described here, the unavailability of coordination states higher than 4 for boron suggests that triclusters of 3-tetrahedra such as (BO3)2O(SiO3) are a possible explanation for the observed anomalous NBO contents. Triclusters involving trigonal borate groups, or two SiO4 groups, are probably unlikely, as the oxygen would be even more overbonded than for (BO3)2O(SiO3). The tricluster oxygen in a group of three BO4 tetrahedra would be less overbonded, but the relatively short B–O distances and B–B repulsion might preclude such sites. This is con- sistent with the observed NBO contents for binary borate glasses, [7,22] for which the data can be rationalized without postulating the presence of triclusters. Xue and Kanzaki [30] have calculated the CQ and diso of aluminum- and silicon-con- taining triclusters. Based on the trends of their results and previous NMR parameters determined for BB46, we can estimate the parameters for bo- ron-containing triclusters. Unfortunately, the pre- dicted peaks for such species may be close to the broad BO peaks in both MAS and 3QMAS spec- tra (Fig. 6), thus making the observation of tricl- usters dicult by 17O NMR. Future experiments, such as those at higher field or involving double resonance techniques, may be necessary to directly observe such species, if present. Acknowledgements This work was supported by the US National Science Foundation, grant DMR-9802072. S. Kroeker acknowledges the Natural Sciences and Engineering Research Council of Canada for a postdoctoral fellowship. References [1] W.J. Dell, P.J. Bray, S.Z. Xiao, J. Non-Cryst. Solids 58 (1983) 1. [2] J.F. Stebbins, in: M. Duer (Ed.), Solid State NMR: Theory and Applications, Blackwell, Oxford, in press. [3] S. Wang, J.F. Stebbins, J. Non-Cryst. Solids 231 (1998) 286. [4] S. Wang, J.F. Stebbins, J. Am. Ceram. Soc. 82 (1999) 1519. [5] J.F. Stebbins, S.K. Lee, J.V. Oglesby, Am. Mineral. 84 (1999) 983. [6] J.F. Stebbins, J.V. Oglesby, Z. Xu, Am. Mineral. 82 (1997) 1116. [7] J.F. Stebbins, P. Zhao, S. Kroeker, Solid State NMR 16 (2000) 9. [8] J.F. Stebbins, Z. Xu, Nature 390 (1997) 60. [9] P. Florian, K.E. Vermillion, P.J. Grandinetti, I. Farnan, J.F. Stebbins, J. Am. Chem. Soc. 118 (1996) 3493. [10] H.K.C. Timken, S.E. Schramm, R.J. Kirkpatrick, E. Oldfield, J. Phys. Chem. 91 (1987) 1054. [11] G.L. Turner, S.E. Chung, E. Oldfield, J. Magn. Reson. 64 (1985) 316. [12] E.H. Hamilton, G.W. Cleek, O.H. Grauer, J. Am. Ceram. Soc. 41 (1958) 209. [13] Z. Xu, J.F. Stebbins, Solid State NMR 11 (1998) 243. [14] D. Massiot et al., Solid State NMR 6 (1996) 73. [15] J.H. Baltisberger, Z. Xu, J.F. Stebbins, S. Wang, A. Pines, J. Am. Chem. Soc. 118 (1996) 7209. [16] A. Samoson, Chem. Phys. Let. 119 (1985) 29. [17] A. Samoson, E. Lippmaa, Phys. Rev. B 28 (1983) 6567. [18] M.J.S. Dewar, R. Jones, J. Am. Chem. Soc. 89 (1967) 2410. [19] D. Massiot, C. Bessada, J.P. Coutures, F. Taulelle, J. Magn. Reson. 90 (1990) 231. [20] L. Frydman, J.S. Harwood, J. Am. Chem. Soc. 117 (1995) 5367. [21] E.I. Kamitsos, M.A. Karakassides, G.D. Chryssikos, Phys. Chem. Glasses 30 (1989) 229. [22] B.C. Bunker, D.R. Tallant, R.J. Kirkpatrick, G.L. Turner, Phys. Chem. Glasses 31 (1990) 30. [23] P.J. Bray, J.G. O’Keefe, Phys. Chem. Glasses 4 (1963) 37. [24] Y.H. Yun, P.J. Bray, J. Non-Cryst. Solids 27 (1978) 363. [25] Y.H. Yun, S.A. Feller, P.J. Bray, J. Non-Cryst. Solids 33 (1979) 273. [26] M. Marezio, H.A. Plettinger, W.H. Zachariasen, Acta. Cryst. 16 (1963) 594. [27] A.D. Mighell, A. Perlo€, S. Block, Acta. Crystallogr. 20 (1965) 819. [28] M.J. Toplis, D.B. Dingwell, Eos, Trans. Am. Geophys. Union 77 (1996) F848. [29] M.J. Toplis, D.B. Dingwell, T. Lenci, Geochim. Cosmo- chim. Acta 61 (1997) 2605. [30] X. Xue, M. Kanzaki, J. Phys. Chem. B 103 (1999) 10816. P. Zhao et al. / Journal of Non-Crystalline Solids 276 (2000) 122–131 131