Preferential Binding of Fluorine to Aluminum in High Peralkaline

J. Phys. Chem. B 2007, 111, 10413-10420

10413

Preferential Binding of Fluorine to Aluminum in High Peralkaline Aluminosilicate Glasses N. G. Karpukhina,† U. Werner-Zwanziger,† J. W. Zwanziger,*,† and A. A. Kiprianov‡ Deparment of Chemistry and Institute for Research in Materials, Dalhousie UniVersity, Halifax, NS, Canada B3H 4J3, and Department of Physical Chemistry, St.-Petersburg State UniVersity, UniVersitetskiy pr. 26, PetrodVorets, St.-Petersburg, 198504, Russia ReceiVed: May 14, 2007; In Final Form: June 27, 2007

For two series of fluoride-containing aluminosilicate glasses of high peralkaline type, we apply 27Al, 19F, 29 Si, and 23Na NMR spectroscopy to understand the structural changes introduced by the addition of alkali fluorides. Adding fluoride in concentrations above the solubility limit causes crystallization of different phases in sodium and potassium glasses despite identical composition. However, the NMR spectra reveal that the structural evolution of the precrystallized states is similar in both series. In particular, fluorine coordinates exclusively to alkaline cations and aluminum. No indication of direct bonding with silicon was found from 19 F f 29Si cross-polarization experiments. In contrast to other glass systems, double resonance experiments in these peralkaline systems show that halide addition produces at most a minor fraction of tetrahedral aluminum containing fluorine in its coordination sphere. Instead, the fluorine addition prior to crystallization converts up to about 20% of the initial tetrahedral aluminum (1 mol % in absolute units) to 5- and 6-fold coordinated aluminum. A minor portion of five-coordinated aluminum groups is considered as the intermediate to the growing fraction of octahedral aluminum in the silicate matrix. The initialization of the crystallization process is correlated with the saturation of the silicate matrix by octahedral aluminum clusters segregating out under further doping by fluoride. It is suggested that the formation of the nonframework Al-F bonds is responsible for structural relaxation, reflected by the reduction of the glass transition temperature.

1. Introduction Fluorided silicate glasses find technological application in the manufacture of machinable, heat-resistant glass ceramics as well as optical materials. However, their structure-property relationships are complex as compared to silicate glasses. Early work in the field assumed preferential bonding between silicon and fluorine, but recent 19F magic angle spinning nuclear magnetic resonance (MAS NMR) studies of multicomponent silicate glasses demonstrated that fluoride readily forms chemical bonds with components other than silicon.1-6 These reports showed that complexes of the fluoride ion with glass modifiers are more abundant than structural units involving direct Si-F bonds. In glasses with several modifiers, fluoride bonding with the highest field strength ion is most prevalent.1,3 Evidently, Si-F bonds dominate only at high temperatures. Recently, Kiczenski and Stebbins showed that in glasses with typical network modifiers the abundance of Si-F bonding grows with increasing fictive temperature.7 Similarly, fluorided aluminoborates show significant fractions of B-F bonding.8,9 It appears that aluminosilicate glasses are unusual in that it is the fraction of Al-F bonding that increases with fictive temperature, although the existence of Si-F bonding in these systems could not be definitely ruled out. The work described above together with the results of refs 10 and 11 indicate the strong selective interaction of the halide component with aluminum in a silicate glass network. The presence of [AlF5]2- and [AlF6]3- groups in sodium fluoroaluminosilicate glass was initially proposed based on 27* Corresponding author. E-mail address: [email protected] (J.W. Zwanziger) † Dalhousie University. ‡ St.-Petersburg State University.

Al NMR.12 Schaller et al.13 confirmed the formation of only six-coordinated aluminum groups in fluoride-doped albite glass measured at a comparatively low field of 7.1 T. Subsequently, it was found that the fluoride-containing glasses of compositions close to albite showed at most a small fraction of 5-fold coordinate aluminum, which was difficult to resolve by 27Al MAS NMR even at 14.1 T.14 However, these findings relate only to compositions that are equimolar or nearly equimolar in alumina and alkali content.7,12-14 The question of selectivity of the halide interaction in glasses with relatively small content of aluminum as compared to alkali, also known as peralkaline compositions, is yet to be clarified. Understanding the structureproperty relationships in this composition range is important because of the significantly lower viscosity and, hence, easier processability as compared to the eqiumolar composition range. The interpretations of structure-property relationships in high peralkaline aluminosilicate glasses are so far based on the conventional assumption that aluminum remains tetrahedral with addition of halogen. Thus, it is important to clarify the coordination state of aluminum for these systems as well as the nature of a mixed oxyfluoride character of its environment. In the solidified oxide-cryolite melts, the existence of oxofluoroaluminates has been excluded.15 Here, we present the results of a multinuclear solid-state NMR study of the high peralkaline glass families, that is, glasses with very low aluminum to alkali ratios. In contrast to equimolar aluminum/alkali compositions, fluoride doping in high peralkaline aluminosilicate glasses does not necessarily correlate with a reduction of viscosity and glass transition temperature. The crystallization processes catalyzed by fluorine compete with their fluxing action on the melt leading to complicated structureproperty-composition relationships. The purpose of the present

10.1021/jp073687j CCC: $37.00 © 2007 American Chemical Society Published on Web 08/11/2007

10414 J. Phys. Chem. B, Vol. 111, No. 35, 2007

Karpukhina et al.

TABLE 1: Glass Compositions Studieda sample

NaF/KF

Na2O/K2O

Al2O3

SiO2

NaAl-2404F0 NaAl-2404F2 NaAl-2404F4 NaAl-2404F6 KAl-2505F0 KAl-2505F1 KAl-2505F2 KAl-2505F3 KAl-2505F6 KAl-2505F9

0.0 3.6 7.4 11.1 0.0 2.6 5.1 8.3 15.8 22.3

24.0 21.8 19.4 17.2 25.0 23.4 21.8 19.8 15.1 11.1

4.0 3.9 3.9 3.8 5.0 4.9 4.9 4.8 4.6 4.4

72.0 70.7 69.4 68.0 70.0 69.1 68.2 67.1 64.5 62.2

a

crystalline phase

NaF

K3AlF6 K3AlF6

Components are Given in mol %.

work is to gain knowledge of the short-range structure for the key components of the glass in order to explain the complicated effects of fluoride on physicochemical properties in these particular systems.10 Fluoride was added gradually up to the point where crystallization processes began. We tried to correlate concentration limits of fluoride solubility in our systems as well as the identity of the crystallized phases with structurally specific features dictated by aluminum cations. We found that in high peralkaline compositions, fluorine is preferentially associated with high-coordinate aluminum through both fluoride and oxyfluoride aluminate species in contrast to the less preferential speciation and tetrahedral coordination observed in equimolar compositions. The short-range structure of the glasses determined by NMR is contrasted to measured physicochemical characteristics. 2. Experimental Methods 2.1. Sample Preparation. Glasses were made from reagent grade Na2CO3, K2CO3, NaF, and SiO2 and dried before mixing. The dried alumina powder was prepared for batching from spectroscopically pure Al2O3. Reagent grade hydrous potassium fluoride KF‚2H2O was heated to 700 °C for dehydration and used immediately for batch preparation. Glasses were synthesized in the temperature range 1550-1600 °C for 2 h in an open platinum crucible in air. Homogenization of the melt induced by stirring was followed by quenching onto a steel form at room temperature and immediate annealing at approximately the glass transition temperature Tg. Glass compositions are represented in Table 1 as batched, with the exception of fluorine which was analyzed potentiometrically and given as recalculated mol % of sodium or potassium fluoride. Glasses of the compositions with the highest content of halide addition were typically nontransparent because of the precipitated fluorine-containing crystalline phases. These phases were identified by X-ray diffraction as NaF in the sodium glasses and the potassium analogue of cryolite, K3AlF6 or 3KF‚ AlF3, in the potassium glasses.11 The commercially available crystalline compounds NaF, KF, AlF3, Na3AlF6, K3AlF6, Na2SiF6, and K2SiF6 were also studied (reagent grade, SigmaAldrich, USA). 2.2. NMR Measurements. All NMR spectra in this study were acquired on powdered glass and crystalline samples at room temperature on a Bruker Avance NMR spectrometer operating at 16.4 T corresponding to 700 MHz proton Larmor frequency. 19F NMR measurements were performed at 658.9 MHz using 2.5 mm ZrO2 rotors and 25.0 kHz sample spinning. To reduce artifacts from electronic ringing, a Hahn-echo sequence was used with 2.65 µs 90° pulse, 36.03 µs echo delay, and 10.5 s recycle delay for the potassium glasses and 350 s

for the sodium glasses. The center transition frequency in Teflon was used as the secondary chemical shift reference of -121.18 ppm. Data were processed with line broadening of 100 Hz. One-dimensional 27Al spectra were obtained at 182.5 MHz in a 4 mm Bruker probe after a single pulse of 1.1 µs, which gave a 20° tip angle in a 0.1 M solution of AlCl3. six hundred transients with 0.5 s delay were acquired in these experiments at multiple spinning rates of 7.0, 9.0, and 10.0 kHz. Data were background corrected by subtraction of the empty rotor signal. 27Al triple quantum magic angle spinning (3QMAS) spectra were obtained in a 2.5 mm Bruker probe spinning at 25.0 kHz using a split t1 three-pulse sequence with full echo acquisition.16 The triple quantum excitation pulse of 3.9 and 1 µs conversion pulse, both with 100 kHz fields, were followed by a 180° selective pulse of 11 µs at approximately 15 kHz field strength. The delay between the second and third pulses was rotor synchronized and set to 60 rotor periods to allow for full echo buildup. A recycle delay of 100 ms was used, validated by relaxation time measurements. The spectral width in the MAS dimension was set to 50 kHz, and 4800 scans were accumulated for each of 100 t1 experiments for KAl-2505F9 and 19 200 scans per 40 t1 experiments for NaAl-2404F6 with 12 dummy scans and a 48-multiple-phase cycle. 3QMAS spectra were processed using the Bruker standard software XWIN-NMR. Fourier transformation along the t2 domain was done with a line broadening factor of 50 Hz. 3QMAS spectra were referenced against 0 ppm signal in AlCl3 solution in accordance with the convention Ck(1) of ref 17. The frequency offset in the indirect dimension Ω1 was taken as -17/ 31 multiplied by the value of the offset in the direct dimension (Ω1 ) -(17/31)Ω2). The values of isotropic chemical shift and quadrupole product were estimated from the overlaid grids of constant isotropic chemical shift and second-order quadrupole corrections calculated according to the formulas of their contribution to the shift in the MQMAS spectra.17 The 19F f 27Al cross-polarization (CP) MAS NMR and 27Al{19F} rotational echo double resonance (REDOR) experiments were initially optimized on commercial polycrystalline AlF3. The CP measurements were done at 15.0 kHz spinning with a 90° pulse of 2.65 µs on 19F, SPINAL-64 decoupling,18 and recycle delays identical to the direct 19F experiments. The contact time of 100 µs used during the optimization on AlF3 was increased to 600 µs for the glass samples. The REDOR sequence used approximately 15 kHz radio frequency field strength for selective excitation of the 27Al central transition with an initial pulse of 6 µs. The echo delay on 27Al was set to of 60 µs corresponding to one rotor period with a spinning speed of 16.666 kHz. The 180° pulse on 19F was 5.6 µs. Each spectrum was acquired with 2048 scans using 1 s recycle delays. The 27Al spectra obtained from the echoes acquired with and without pulsing on 19F were integrated, and the REDOR ratios were evaluated (Table 3). The integration error was estimated from the spectral parameters of the signals, that is, signal-to-noise, width at half-height, and resolution in Hz/pt. In addition, an average systematic error of 5% arising from the processing and integration of the spectra was taken into account. Both uncertainties were propagated into the REDOR ratio. The 23Na NMR spectra were obtained at 185.2 MHz Larmor frequency in a 4 mm probe using an excitation pulse of 0.5 µs, collecting 16 free induction decays with relaxation delay of 11 s and a spinning rate of 14.0 kHz. The chemical shift was referenced to 1 M NaCl at 0 ppm. All sodium spectra were processed with 20 Hz line broadening.

Fluorine to Aluminum Binding

J. Phys. Chem. B, Vol. 111, No. 35, 2007 10415

Figure 1. 19F MAS NMR spectra for (a) potassium aluminosilicate glasses and (b) sodium aluminosilicate glasses as a function of fluorine content. Spinning side bands are marked by an asterisk.

Figure 2. (a) 27Al MAS NMR spectra for the potassium aluminosilicate glasses as a function of fluorine content taken at 182.5 MHz of Larmor frequency. Spinning side bands are marked by an asterisk. (b) 19F f 27Al CP spectra for potassium aluminosilicate glasses as a function of fluorine content. 29Si NMR measurements were performed at 139.1 MHz Larmor frequency in a 4 mm probe at 7.0 kHz MAS. A 2 µs pulse was used with a typical recycle delay 100-150 s. The 19F f 29Si cross-polarization MAS NMR experiments were optimized on commercial polycrystalline Na2SiF6. The CP measurements were done at 10.0 kHz MAS with a 90° pulse of 4.5 µs on 19F, SPINAL-64 decoupling, and contact time varied in the range 0.5-5 ms. Recycle delays of 5 and 150 s for potassium and sodium samples, respectively, were used. Kaolin at -91.34 ppm was used as a secondary reference for the silicon chemical shift. All 29Si spectra were processed with 50 Hz exponential line broadening.

3. Results The 19F MAS NMR spectra of the potassium and sodium glasses with different fluoride amounts are shown in Figure 1. Several polycrystalline fluorides and fluoride-related compounds were used as potential reference materials that could assist the assignment in glass samples, including NaF, KF, Na3AlF6, K3AlF6, AlF3, Na2SiF6, and K2SiF6. Results in accord with literature were obtained.1,6,13,19

The results from one-dimensional 27Al MAS NMR experiments are summarized in Figures 2a and 3a. The evolution of the spectra as a function of increased fluorine content is clearly resolved for both glass families. The results of 19F f 27Al CP experiments are shown beside the corresponding single pulse excitation spectrum of each glass series (Figures 2b and 3b). The difference spectra of the REDOR experiments obtained by subtracting the echo spectrum with fluorine pulses from the ones without fluorine excitation for each sample are presented in Figure 4. The 3QMAS NMR spectra are presented in Figure 5a for KAl-2505F9 and Figure 5b for NaAl-2404F6. As seen from the figures, three well-separated features were observed in both cases. The values of isotropic chemical shift and quadrupole product estimated from the overlaid grids are given in Table 2. Splitting of the signal into two resonances from octahedral aluminum observed in conventional MAS spectra was noted for KAl-2505F9 (Figure 5a, inset) in addition to sinc artefacts due to truncated free induction decays. Figure 5c shows the 3QMAS spectrum of polycrystalline K3AlF6 with three closely situated signals around 0 ppm. The 23Na MAS NMR data for sodium glasses NaAl-2404Fx are shown in Figure 6.


Karpukhina et al.

Figure 3. (a) 27Al MAS NMR spectra for the sodium aluminosilicate glasses as a function of fluorine content taken at 182.5 MHz of Larmor frequency. Spinning side bands are marked by an asterisk. (b) 19F f 27Al CP spectra for sodium aluminosilicate glasses as a function of fluorine content.

Figure 4. 27Al{19F} MAS REDOR difference spectra for (a) potassium aluminosilicate glasses and (b) sodium aluminosilicate glasses as a function of fluorine content, obtained by subtraction of the spectra with fluorine pulses from the ones without fluorine excitation.

Figure 7 shows the results of 29Si MAS measurements for potassium and sodium compositions. With increase in fluorine concentration, the broad asymmetric 29Si lines shift upfield by 5-6 ppm from the parent glass compositions (-90.8 ppm for NaAl-2404F0, -92.7 ppm for KAl-2505F0) and change shape in both the sodium and potassium glasses. 4. Discussion 4.1. Spectral Assignments. 4.1.1. Fluorine-19 Resonances. The 19F NMR spectra of the noncrystalline potassium glass samples showed broad 19F resonances at about -140 ppm with prominent shoulders at -170 ppm (Figure 1a). Upon crystallization (as detected by X-ray diffraction) the broad feature at -140 ppm splits into two relatively sharp resonances, the dominant one at about -157 ppm and another one at -133 ppm. A shoulder at -171 ppm was observed in the crystallized samples too. The feature at -157 ppm matches that observed in K3AlF6, both by us and by others,19 an assignment in agreement with the features observed from the X-ray diffraction of these samples. Both Si-F and K-F fluoride environments could cause 19F signals at around -133 ppm. For example, fluorine in K2SiF6 resonates at -134 ppm. However, as

discussed below, Si-F bonding appears to be negligible in these glasses. Therefore, the signal at -133 ppm is most likely due to K-F similar to that in KF. The shoulder at about -171 ppm, observed in all samples, can be produced not only by AlF3-like fragments with corner-shared [AlF6]3- units, that is, bridging fluorine but also by a variety of aluminum fluoride and oxyfluoride units as demonstrated earlier.20 The broad band at -140 ppm observed before crystallization could belong to fluorine next to the aluminum, according to the calculations of ref 20, or next to the potassium (as close to the resonance in KF). Again, since the signal observed in the high-frequency region could be potentially produced by units with Si-F bonds, we refer here to the later discussion of 29Si NMR resonance. The 19F NMR spectra of the sodium glasses (Figure 1b) showed sharp resonances at about -220 ppm upon crystallization, which match that of NaF, an assignment in accordance with the X-ray data on this sample. The noncrystallized glasses show a broad feature in this chemical shift range suggesting that they too contain coordinations similar to NaF. All sodium samples also showed a broad resonance near -175 ppm similar to the -171 ppm resonance seen in the potassium glasses, although broader in sodium glasses, and is assigned to aluminum-



Figure 6. 23Na MAS NMR spectra for sodium aluminosilicate glasses as a function of fluorine content. Spinning side bands are marked by an asterisk.

Figure 5. 27Al 3QMAS NMR spectra for samples (a) KAl-2505F9, (b) NaAl-2404F6, and (c) K3AlF6. Highest contours correspond to 95% maximum intensity. The inset in (a) zooms into the 6-fold coordinated aluminum resonance.

centered units. There was no signal at -190 ppm, which would be assigned to Na3AlF6 in the sodium glasses. 4.1.2. Aluminum-27 Resonances. In both the sodium and the potassium fluorine-free glasses, the 27Al NMR signal at 58 ppm is assigned to tetrahedral aluminum coordination (Figures 2a and 3a). The addition of fluorine is correlated with new resonances at -1.2 to -1.7 ppm, assigned to octahedral aluminum coordination, and at 23-24 ppm. Similar resonances were observed by Kohn et al.12 in jadeite-NaF glass of the composition Na3Al2Si4O12F and were assigned to [AlF6]3- and [AlF5]2- species, respectively, although the presence of [AlF4]units as well as mixed oxyfluoride polyhedra was additionally

mentioned. Such sites have been identified directly in molten sodium fluoroaluminates.21 The 3QMAS experiment was used to investigate these resonances in more detail in the sodium and potassium glasses with maximum fluorine content (Figure 5a,b). The analysis of the spectral parameters extracted for each resonance (Table 2) shows that the site at about 23-24 ppm (position in MAS spectra) has a large quadrupole interaction but that it is not dramatically larger than the other sites. On the other hand, its chemical shift is about 25-30 ppm. Both findings are in agreement with data at higher magnetic field on five-coordinate aluminum in the fluorine-doped sodium enriched albite.14 Therefore, this site can be confidently assigned to 5-fold coordinate aluminum rather than a highly distorted 4-fold coordinate site, the other reasonable alternative. Comparison with ab initio work22 suggests that this site can have either pure fluoride surroundings as in [AlF5]2- or mixed oxyfluoride coordination, for instance, [AlF3O2]2- groups. The latter seems more likely due to a wide distribution of chemical shift and quadrupole coupling parameters observed for the site Al(5) (Table 2). At the same time, the presence of some [AlO5]2fraction cannot be completely rejected based merely on the lower chemical shift range that is known for typical 5-fold [AlO5]2site in alkali-free systems.23 Also of note is the fact that the quadrupole interaction in the nominally octahedral site at about 0 ppm is substantial for this kind of site, suggesting disorder in the coordination shell possibly in the form of mixed [AlOnF6-n]3coordination. Upon crystallization, the 0 ppm resonance in the potassium glass shows several resolved sites similar to what is seen in polycrystalline K3AlF6 (Figure 5a and ref 24). Prior to crystallization, however, the octahedral sites may also have mixed oxyfluoride coordination. Further assignment is provided by the double resonance experiments. The 19F f 27Al cross-polarization experiments at short contact times are sensitive to F-Al bonds, due to throughspace dipole dipole coupling, and due to the aluminum chemical shift dispersion can be used to distinguish different fluorinated aluminum-centered polyhedra. The CP results in Figures 2b and 3b demonstrated that in the present case only octahedral fluorinated aluminum was detected. No CP signals were observed from tetrahedral fluorinated aluminum or pentacoordinate aluminum even under a range of contact times. However, given the low abundance of this fraction observed in regular


Karpukhina et al.

Figure 7. 29Si MAS NMR spectra for (a) potassium aluminosilicate glasses and (b) sodium aluminosilicate glasses as a function of fluorine content. Spinning side bands are marked with an asterisk.

TABLE 2: Ranges of the Quadrupole Product PQ ) CQ

x1+η2/3and the Isotropic Chemical Shift Estimated for 90 and 70% Intensities of Each Signal in 3QMAS Spectra Shown in Figure 5a,b 90% intensity sample

site

KAl-2505F9

Al(4) Al(5) Al(6) Al(4) Al(5) Al(6)

NaAl-2404F6

27Al

δiso CS,

ppm

57-59 25-28 -2-0 59-62 27-30 0-2

PQ, MHz 2-3 2-4 0-1 2-4 2-4 1-3

70% intensity δiso CS,

ppm

55-61 25-28 -2-0 57-64 26-32 -1-3

PQ, MHz 1-4 2-4 0-2 1-5 0-5 0-4

resonance spectra (Figures 2a and 3a) and the possibility of insufficient spin-locking of the aluminum species due to the offset dependence and the large quadrupole coupling constants determined from the 3QMAS experiments (Table 2), the signalto-noise ratio of the collected CP spectra is not sufficient to rule out the presence of fluorine in the five-coordinate aluminum units. To clarify this under conditions independent of the magnitude of the quadrupole coupling, we performed 27Al{19F} REDOR experiments to detect aluminum sites with short-range Al-F dipolar interactions. The measured REDOR ratios ∆S/ S0, that is, the fractions of the signal reduction from the unperturbed echo signal, are given in Table 3 for the three aluminum coordinations. In both glass series, the biggest REDOR differences were obtained for the octahedral aluminum sites (Figures 3 and 4). However, the difference spectra in Figures 3 and 4 show a small fraction of fluorinated tetrahedral and pentahedral aluminum sites as well. Quantitatively, the confidence level in the absolute REDOR fractions is small (Table 3). Also, even though we have used the shortest experimentally possible echo delay, the impact of Al-F contacts from next nearest coordination spheres may also contribute to the small REDOR effect. Still, the existence of a visually distinguishable, though small, REDOR effect on the tetrahedral and pentacoordinated aluminum sites indicates a minor presence of halogen close by. Overall, the similarity of the aluminum MAS spectra in the sodium and potassium series indicates that in these peralkaline compositions the nature of the alkali cation is not significant in the aluminum speciation, which proceeds from tetrahedral, when no fluorine is present, to the formation of penta- and octahedral coordination as fluorine is added.

TABLE 3: REDOR Ratio ∆S/S0 of Each Aluminum Fraction Calculated from Experimental Dataa. sample

Al(4)

Al(5)

Al(6)

KAl-2505F1 KAl-2505F2 KAl-2505F3 KAl-2505F6 KAl-2505F9 NaAl-2404F2 NaAl-2404F4 NaAl-2404F6 AlF3 K3AlF6

0.011 ( 0.071 0.020 ( 0.070 0.001 ( 0.072 0.015 ( 0.071 0.018 ( 0.071 0.006 ( 0.071 0.018 ( 0.070 0.012 ( 0.071

0.126 ( 0.249 0.167 ( 0.137 0.076 ( 0.126 0.057 ( 0.167 0.137 ( 0.087 0.058 ( 0.234 0.112 ( 0.137 0.045 ( 0.186

0.236 ( 0.475 0.317 ( 0.135 0.274 ( 0.103 0.194 ( 0.066 0.125 ( 0.064 0.194 ( 0.281 0.335 ( 0.077 0.316 ( 0.101 0.415 ( 0.043 0.129 ( 0.062

a The errors were estimated from both processing and resolution of the signals in the spectra.

4.1.3. Sodium-23 Resonances. The 23Na NMR spectrum (Figure 6) of the fluorine-free sample shows a broad resonance at -5.7 ppm. Upon introduction of fluorine, a shoulder appears at roughly 4 ppm that evolves into a strong peak in NaAl2404F6 at 5.1 ppm corresponding to the crystalline phase of NaF as determined by X-rays. This chemical shift is in agreement with literature on other glass ceramic systems and shows that the sodium fluoride crystals are under stress, imposed by the glass matrix.25 This characterization is also consistent with the 19F assignments outlined above. In addition, the bulk signal shifts slightly to lower frequency reaching finally -7.1 ppm. This may indicate the presence of some fluorine in the coordination sphere or a slight increase in oxygen coordination. 4.1.4. Silicon-29 Resonances. Both series of 29Si NMR spectra (Figure 7) show a broad 29Si resonance centered around -90 to -100 ppm, typical of tetrahedral SiO4 coordination. These resonances do not resolve into component peaks, but they do clearly shift upfield as fluorine is added. The shifts of the signals as compared to the parent glass compositions (-90.8 ppm for NaAl-2404F0, -92.7 ppm for KAl-2505F0) are in accordance with the trends of chemical shift values known for similar glass compositions.26 These shifts and the asymmetric nature of the peaks may be either due to increasing polymerization of the network as Q3 and Q2 units are converted to Q4 or due to the loss of aluminum in the next coordination sphere of Q4 units.23 Both interpretations seem reasonable; since we keep the concentration of alkali cations constant, we effectively lower the concentration of alkali oxides as we add alkali fluorides. The lower concentration of alkali oxides causes a decrease in


Figure 8. Average coordination number of aluminum calculated from 27 Al MAS NMR spectra as a function of atomic percentage of fluorine in glasses. The vertical line indicates the onset of crystallization.

the concentration of nonbridging oxygen in the whole system. Because the fluoride anions could simply exchange the nonbridging oxygens in the silicon bonding environment, the shift in nonbridging oxygen to fluoride concentration would not have an impact on the silicon polymerization. However, since the fluorine atoms bond preferentially to aluminum, by increasing the aluminum coordination number (as we have shown above), the amount of nonbridging oxygen is effectively reduced for the silicon subsystem thereby causing higher polymerization. Our second interpretation, namely, the removal of aluminum from the second coordination sphere, also seems possible, again based on the fluorine preference to bond with aluminum while simultaneously increasing the aluminum coordination. In a purely oxide system, the removal of aluminum from the silicon bonding network would require creation of aluminum oxide clusters or aluminum terminated by nonbridging oxygens. In the present case of mixed anions, we find preferential Al-F, as opposed to Al-NBO, bonding. In summary, both interpretations rely on the fluorine selectivity to aluminum. To see whether fluorine coordinates to silicon, we looked very carefully for Si-F bonding. This search was performed using 19F f 29Si cross-polarization measurements under conditions that had been carefully optimized with Na2SiF6. While the Si-F interaction was easily detected in this model compound, no such signals were found in any of the glasses even after long contact times and many scans. Therefore, we believe that in these peralkaline samples, the Si-F bonding is negligible and fluorine binds preferentially to aluminum in higher coordination. 4.2. Structural Model. All of the NMR data combined show that the addition of fluorine to the glasses impacts the aluminum environment the most. The average coordination number of aluminum, as calculated from the peak integral areas in 27Al MAS spectra, is shown in Figure 8. Within the solubility limit, the coordination number increases linearly with fluorine content independently of the identity of the alkaline component due to the appearance of five- and six-coordinated aluminum in fluoride-doped glasses. The REDOR ratios ∆S/S0 (Table 3) and the 19F f 27Al CP experiments show that the fraction of fluorinated tetrahedral aluminum in both glass series is significantly low and is not influenced by the increase of the fluoride content. These remarkable results indicate that the 4-fold alumina units containing fluorine in their coordination sphere are by no means the dominant species as a result of halidealuminum interaction in our peralkaline systems. This finding is in contrast to the typical assumption that fluorine bonds with aluminum in tetrahedral units.4,5,10,11 Certainly, the dominant


Figure 9. Fluoride content dependence of the fractions of aluminum groups with coordination states five and six for sodium and potassium glass series. The vertical line indicates the onset of crystallization.

fraction bound with fluorine is the octahedral aluminum in both glass systems though a small fraction of fluorinated pentahedral aluminum is noticeable (Table 3). Thus, fluorine selectively forms bonds with aluminum but simultaneously increases the aluminum coordination number. Figure 9 demonstrates the changes in each of the highcoordinated aluminum fractions as a function of fluoride content in the glasses. Adding only the high-coordinated aluminum fractions (Figure 9), it is evident that fluoride modifies no more than 20% of the aluminum structural units within the solubility limit of our glasses. The percentage of aluminum in five- and six-coordination states estimated by Kohn et al.12 for Na3Al2Si4O12F glass was less than 7%. However, the composition investigated there contained three times more aluminum (15 vs 23 mol % of total sodium), which indicates that the absolute fraction of aluminum transferred by fluorine to the high coordinations was still about the same. Thus, in general it could be concluded that only a small portion of initially tetrahedral aluminum is involved in interaction with fluoride. Perhaps that is the reason that in fluoride-containing glasses of equimolar compositions such as albite type (NaAlSi3O8), the portion of high-coordinated aluminum species is very low (

Preferential Binding of Fluorine to Aluminum in High Peralkaline

Recommend Documents