Structural Changes above the Glass Transition and Crystallization in

Jun 9, 2007 - Structural Changes above the Glass Transition and Crystallization in Aluminophosphate Glasses: An in Situ High-Temperature MAS NMR ...
0 downloads 0 Views 233KB Size
J. Phys. Chem. B 2007, 111, 7529-7534

7529

Structural Changes above the Glass Transition and Crystallization in Aluminophosphate Glasses: An in Situ High-Temperature MAS NMR Study Leo van Wu1 llen,* Sebastian Wegner, and Gregory Tricot Institut fu¨r Physikalische Chemie, Westfa¨lische Wilhelms-UniVersita¨t Mu¨nster, Corrensstrasse 30-36, D-48149 Mu¨nster, Germany ReceiVed: January 26, 2007; In Final Form: April 14, 2007

We present an in situ high-temperature nuclear magnetic resonance study on the structural changes in aluminophosphate glasses occurring in the temperature range between the glass transition temperature Tg and the crystallization temperature Tc, Tg < T < Tc. Decisive changes in the network organization between Tg and Tc in potassium aluminophosphate glasses in the compositional range 50K2O-xAl2O3-(50 - x)P2O5 with 2.5 < x < 20 could be monitored for the first time employing 1D 31P- and 27Al-MAS NMR. Accompanying ex situ NMR experiments (31P-RFDR NMR and 31P-{27Al} CP-HETCOR NMR) on devitrified samples were performed at room temperature to further characterize the phases formed during the crystallization process. The structural role of boronswhich is known to inhibit the crystallization process in these aluminophosphate glassesson short and intermediate length scales was analyzed employing 11B-MQMAS, 11B-{27Al} TRAPDOR and 11B-{31P} REDOR NMR spectroscopy.

Introduction Phosphate based glasses find a wide range of applications including biomaterials,1 laser hosts,2 nuclear waste storages,3 anti-oxidation coatings, or metal-to-metal seals.4 To overcome the poor chemical durability and resistance to moisture attack, alumina is often incorporated into the phosphate glass network.5 In some of the above applications, the glasses are exposed to temperatures above the glass transition temperature, thus possibly entailing changes in the phosphate network organization and hence performance of the material. The evolution of the network structure with temperature therefore constitutes an important issue of material performance in phosphate based glasses. Two strategies can be applied to elucidate the structural changes occurring in a glass network at high temperatures. In the ex situ approach, the glass sample is first annealed at a given temperature; then, the structural evolution is monitored in subsequent experiments, performed at ambient temperature. It is thus clear that utilizing this approach only allows tracing permanent non-reversible changes induced by the annealing process. In the in situ approach, the changes in the network organization during the annealing process are monitored at the annealing temperature as they happen, therefore offering the opportunity to elucidate reversible changes in the network organization and to characterize the intrinsic details of the crystallization process. To date, however, only sparse in situ data is available on phosphate based glasses. To our knowledge, only a single in situ high-temperature X-ray diffraction study6 and one in situ Raman study7 have been reported to characterize the structural changes upon annealing in these glass systems. Solid-state nuclear magnetic resonance (NMR) has long been established as a very powerful tool for the characterization of the structure of amorphous solids. With the advent of MAS (magic-angle spinning) NMR probes which can be operated up to temperatures of 900 °C,8,9 in situ MAS NMR studies of the reversible and nonreversible changes in the network organization of amorphous solids at high temperature have become

accessible.10-12 In this report, we have employed this approach to analyze the changes in the network organization of aluminophosphate glasses occurring in the temperature range between the glass transition temperature Tg and crystallization temperature Tc utilizing 27Al- and 31P-MAS NMR studies in the temperature range RT < T < 650 °C. In addition, the phases formed during the crystallization process were characterized ex situ employing 2D 31P-RFDR NMR (Radio Frequency Driven Recoupling) and 2D 31P-{27Al} CP-HETCOR (HETeronuclear CORrelation) experiments. The addition of small amounts of B2O3 to aluminophosphate glasses has been shown to significantly reduce the tendency of crystallization in these systems.13 Using a model glass containing 5% B2O3, the structural changes on short and intermediate length scales upon boron incorporation into the aluminophospate network were studied employing ex situ NMR experiments, i.e., 11B-MQMAS (Multiple Quantum), 11B-{27Al} TRAPDOR NMR (Transfer of Populations in Double Resonance), and 11B{31P} REDOR NMR (Rotational Echo Double Resonance) spectroscopy to obtain a rationale for the inhibition of the crystallization process in the modified glasses. Experimental Transparent and colorless samples in the glass system 50 K2O-xAl2O3-(50 - x) P2O5 were prepared employing the standard melt quenching method. Reagent grade K2CO3, Al(OH)3 and (NH4)2HPO4 in appropriate quantities (for 10 g of glass) were thoroughly grounded and slowly heated (1 °C/min) in a Pt crucible up to 600 °C to remove H2O, NH3 and CO2. Then, the mixture was heated to 900-1300 °C for 20 min, depending on composition with subsequent quenching between two copper plates. Since the weight loss (∆m/m) always proved to be less than 3%, glass compositions given in the text correspond to the batch compositions. The glasses are labeled using the notation KAlP_x, with x denoting the alumina content. For example, a sample containing 10% Al2O3 (50 K2O-10 Al2O3-40 P2O5) will be referred to as KAlP_10. For the boron

10.1021/jp070692e CCC: $37.00 © 2007 American Chemical Society Published on Web 06/09/2007

7530 J. Phys. Chem. B, Vol. 111, No. 26, 2007

van Wu¨llen et al.

TABLE 1: Melting Temperatures (Tm), Weight Loss (∆m/m), Glass Transition Temperatures (Tg), and Crystallization Temperatures (Tc) for the KAlP_x (50 K2Ox Al2O3- (50-x) P2O5) Glasses KAlP_x

Tm (°C)

∆m/m (%)

Tg (°C)

Tc (°C)

2.5 5 7.5 10 12.5 15 17.5 20

900 1000 1000 1000 1050 1100 1150 1300

2.5 3.0 0.8 2.1 2.0 1.2 0.8 1.2

278 303 331 349 350 349 a 310

410 468 467 460 466 a 460

a

Not determined.

containing glass sample, 47.5 K2O-9.5 Al2O3-38 P2O5-5 B2O3, KAlP_10B, the appropriate amount of B2O3 (dried at 200 °C for 3 d) was added to a corresponding mixture of K2CO3, Al(OH)3 and (NH4)2HPO4 and subsequently treated as described above. The devitrified samples used for the ex situ studies were obtained applying a thermal treatment at 450 - 600 °C for 10 h to the KAlP_x samples and are referred to as KAlP_x dev. Thermal analysis on the glass samples was performed employing a heating rate of 10 °C/min to determine the glass transition (Tg) and crystallization (Tc) temperatures with an estimated error of (10 °C. Table 1 summarizes the melting temperatures (Tm), weight losses (∆m/m), as well as Tg and Tc for the glasses studied in this work. The high-temperature in situ NMR experiments were performed employing a Doty HT-MAS probe. The 31P- and 27AlMAS NMR experiments were performed at 4.7 T for 31P (ν0 ) 81.1 MHz) and at 7.04 T for 27Al (ν0 ) 78.2 MHz). MAS was applied at 4.5 kHz using dry N2 for sample spinning. The magic angle proved to be stable over the entire temperature range used in this study (RT < T < 650 °C). A radiofrequency field (rf) of 27.7 kHz corresponding to a π/2-pulse length of 9 µs was used for the acquisition of the 31P-MAS NMR spectra with 2432 transients and a repetition time of 90 s to record the signal. Longer repetition times did not change the appearance of the spectra. 27Al-MAS NMR spectra were obtained with 1024 to 2048 accumulations, a recycle delay of 0.5 s, a rf field of 30 kHz (obtained on a liquid sample) and a π/6 pulse angle. The 31P and 27Al chemical shifts are referenced to 85% H PO and 3 4 1M Al(NO3)3 respectively. RFDR NMR experiments14 were performed at room temperature on the devitrified samples to study spatial proximity between the different phosphorus sites. In this sequence, the homonuclear dipolar interaction is reintroduced through the application of several rotor-synchronized π-pulses. The 2D 31PRFDR NMR spectrum was acquired at 9.4 T at a frequency of 161.9 MHz, employing a mixing time of 32 ms, a MAS frequency of 12.5 kHz, and a rf amplitude of 100 kHz. The 1024 × 512 data points were collected employing a rotor synchronized t1-increment using TPPI and a recycle delay of 40 s, following presaturation by a saturation comb. CP-HETCOR experiments15,16 were performed at room temperature on the devitrified samples to probe the spatial proximity between phosphate and aluminate species. Magnetization transfer from a quadrupolar nucleus to a 1/2 spin has been studied in detail by several authors.17-19 In essence, very low rf amplitudes for the contact pulse on the quadrupolar nucleus have to be used to ascertain reasonable spin-lock efficiency. The 2D 31P-{27Al} CP-HETCOR NMR experiment was performed at 9.4 T and a 4 mm triple channel probe (spinning frequency of 8 kHz) employing a rf amplitude of 73.5 kHz (determined on

Figure 1. In situ high-temperature 31P-MAS NMR spectra (at 4.7 T) for samples KAlP_5 (Tg ) 303 °C, Tc ) 410 °C) (a) and KAlP_10 (Tg ) 349 °C, Tc ) 467 °C) (b) for the indicated temperatures.

a liquid) for the first π/2 pulse on 27Al. The optimized rf amplitudes used for the cross-polarization process were determined to 8 kHz and 6.5 kHz for 31P and 27Al, respectively; the contact time was set to 5 ms. The 2048 × 200 data points were acquired using TPPI with a 10 µs t1 increment. Each slice was recorded with 2048 transients and a relaxation delay of 0.5 s. The 11B-{27Al} TRAPDOR NMR experiments were performed employing a DSX-500 Bruker spectrometer with resonance frequencies of 160.5 and 130.3 MHz for 11B and 27Al, respectively, with ν (11B) ) 33.3 kHz and ν (27Al) ) RF RF 62.5 kHz. MAS was performed at 3 kHz. The 11B-{31P} REDOR NMR experiment was performed on a Bruker DSX 400 spectrometer employing rf amplitudes of 66 kHz for 11B and 45 kHz for 31P, respectively. The experiments allow an estimation of the heteronuclear dipolar interaction between 11B and 27Al or 11B and 31P, respectively and thus enable us to characterize the second coordination sphere around a central boron atom within the amorphous network. Results and Discussion In Situ Experiments. The 31P-MAS NMR spectra for two representative glass samples (KAlP_5 and KAlP_10) from ambient temperature to 550 °C are plotted in Figure 1. The room-temperature spectra for both glass compositions exhibit broad overlapping resonances characteristic of a mixed aluminophosphate network. A detailed solid-state NMR study, including CP-HETCOR, REAPDOR, and J-RESolved NMR experiments,20 enabled the deconvolution of the observed signals into five different lines. The chemical shift values indicate the presence of pure phosphate species (Q2 at -19 ppm and Q1 sites at 0 ppm) as well as phosphate species connected to one or more aluminate species. The signals at -8, -12, and -16 ppm could be assigned to Q1 sites connected to one AlO6 unit (signal at -8 ppm), one AlO4 unit (signal at -12 ppm), and two aluminate species (signal at -16 ppm), respectively.20 Upon heating, no significant spectral changes can be identified for temperatures below Tg. The observed minor evolution in the chemical shift values can be ascribed to a drift in the effective B0-field strength with temperature. For temperatures between Tg and Tc, however, distinct changes in the 31P-MAS spectra can be observed. For KAlP_5 (Figure 1a), the resonance at -12 ppm considerably increases in intensity at 350 °C (60 °C below the crystallization temperature). Finally, for temperatures higher than Tc, six narrow resonances can be distinguished in the chemical shift range of -8 to -20 ppm.

Crystallization in Aluminophosphate Glasses

J. Phys. Chem. B, Vol. 111, No. 26, 2007 7531

Figure 2. In situ high-temperature 27Al-MAS NMR spectra (at 7.05 T) for samples KAlP_10 (a) and KAlP_15 (b) for the indicated temperatures.

Figure 3. In situ high-temperature 27Al-MAS NMR spectra for KAlP_10 during heating (a) and cooling (b) for the indicated temperatures.

Each signal is accompanied by numerous spinning sidebands (not shown), indicating that the system is still in the solid state without any extended exchange between the individual species. The narrow resonances are therefore assigned to the presence of crystalline phase(s) in the sample induced by the devitrification process. The evolution of the 31P-MAS spectra for the KAlP_10 glass sample follows the same trend. Around 400 °C a broadening of the three different resonances can be observed, whereas at 500 °C four different narrow signals arise, again indicating the crystallization of the material. 27Al-MAS NMR spectra were recorded from room temperature to 550 °C for all the samples. The spectra for glasses KAlP_10 and _15 are collected in Figure 2. The three resonances identified in the room-temperature spectra at approximately 40, 10, and -18 ppm indicate the presence of aluminum in 4-, 5-, and 6-fold coordination by oxygen. In accordance with the trends observed for sodium aluminophosphate glasses, the fraction of four coordinated aluminum, AlO4, increases with increasing alumina content.21 As already noticed for the 31P-MAS NMR spectra, no distinct spectral changes in the 27Al-MAS spectra can be identified in the temperature range below the glass transition Tg; the relative proportion between the three contributions remains constant. Clear changes however do occur in the temperature range between Tg and Tc on both samples. In both cases the amount of tetrahedral aluminum present in the glass increases at the expense of the fraction of five and six coordinated aluminum. Then, after octahedral aluminum is almost entirely converted to tetrahedral species, the glasses crystallize, now exhibiting narrow resonances at approximately 40 ppm. For KAlP_10, an additional signal centered at approximately -30 ppm can be identified in the 27Al-MAS NMR spectrum at T ) 550 °C. Unfortunately, we were not yet able to assign this signal. The main spectral changes in the temperature range Tg < T < Tc, a decrease in the aluminum coordination from 6- to 4-fold, accompanied by a considerable increase in the intensity of the 31P-MAS NMR signal around -12 ppm indicating the presence of phosphate polyhedra connected to AlO4 units, have been found in all studied glass samples. In a second set of experiments, the reversibility of the spectral changes between Tg and Tc was monitored using a heating/ cooling cycle. Figure 3 presents the 27Al-MAS NMR spectra for the KAlP_10 glass in the temperature range from 300 to 440 °C. The spectra recorded during heating and cooling are compiled in Figure 3a and 3b, respectively. The most important

observation to be taken from this figure is the complete reversibility of the gradual change of aluminum coordination from 6- to 4-fold upon heating and from fourfold to sixfold during cooling. Thus, increasing the temperature favors the formation of the lower-coordinate aluminate species. Since from 27Al-{31P} REDOR NMR experiments at ambient temperature it is known that the AlO4, AlO5, and AlO6 units are fully connected to phosphate species (i.e., to 4, 5, and 6 phosphate tetrahedra, respectively),20 the gradual decrease in the Alcoordination number necessarily imparts the breaking of P-OAl bonds. This is corroborated by the observed increase in the fraction of phosphate species connected to AlO4 units (vide supra). This local demixing of structural units may pave the way to the phase separation and crystallization observed at temperatures above Tc. Such a decrease in the average Al coordination with increasing temperature has been observed by Sen et al.22 in sodium boroaluminate glasses using ex situ NMR techniques on samples prepared with different fictive temperatures. We note that the observed reversible structural changes associated with the coordination change AlVI T AlIV are manifestations of structural relaxation processes in the glasses above Tg which contribute to the configurational heat capacity 23-25 The or entropy Sconf, related by dSconf ) (Cconf Cconf p p /T) dT. VI contribution of the Al speciation reaction Al T AlIV to the configurational heat capacity may be calculated from the enthalpy of the above speciation reaction employing the van’t Hoff equation25 ∆H ) -R{(ln K(Tb) - ln K(Ta))/(1/Tb - 1/Ta)}, with K(Tb) and K(Ta) being the equilibrium constants for the above reaction at temperatures Tb and Ta. A rough estimate of the relative fractions of the four- and six-coordinated aluminate species (by integration of the respective signals of the spectra in Figure 3) at T ) 673 K (AlIV, 0.41; AlVI, 0.38) and T ) 613 K (AlIV, 0.33; AlVI, 0.48) produces a reaction enthalpy ∆H ) 25 ( 10 kJ mol-1. From this, the contribution of the Alcoordination change to the overall configurational heat capacity can be estimated to ∼0.07 J g-1 K-1.26,27 Cconf p Ex Situ Experiments. To further examine the nature of the crystallization process and the chemical nature of the crystalline phases formed during the devitrification, ex situ experiments at ambient temperature were applied to annealed samples (10 h at 600 °C). The X-ray diffraction pattern (not shown) indicate the presence of a potassium metaphosphate (KPO3; PDF-35 0819) in the devitrified low Al2O3 glasses (x < 12.5) and potassium pyrophosphate (K4P2O7; PDF-21 0676) at higher Al2O3 content. In addition, further diffraction peaks can clearly

7532 J. Phys. Chem. B, Vol. 111, No. 26, 2007

van Wu¨llen et al.

Figure 4. Comparison between in situ and ex situ 31P-MAS NMR and 27Al-MAS NMR spectra of the annealed sample (KAlP_5 dev). Chemical shift deviation between ex situ and in situ spectra are due to a drift in the effective B0-field strength with temperature.

Figure 5. 2D 31P-RFDR NMR (a) and 2D 31P-{27Al }CP-HETCOR NMR (b) spectra for sample KAlP_5 dev. The RFDR spectrum was recorded employing a mixing time of 32 ms, and the CP-HETCOR was acquired using a contact time of 5 ms.

be identified, but not be assigned to any known crystalline phase in the system K-Al-P-O. The ex situ 1D 31P- and 27Al-MAS NMR spectra are collected in Figure 4 for glass KAlP_5, together with the corresponding in situ spectra obtained at 600 °C. The 31P-MAS NMR spectrum (Figure 4a) exhibits several resonances from 0 to -15 ppm and two signals at -18 and -20 ppm. The two latter can be assigned to the two different phosphorus positions present in KPO3 and thus confirm its presence in the devitrified sample.28 As stated above, the minor difference in the peak positions of the in situ and ex situ spectra can be ascribed to the slight drift of B0 with temperature. Although the resolution in the in situ spectra is distinctively lower as compared to the ex situ spectra, it is clear that both are composed of the same signals, even if the absolute intensities of the different signals do not match. The 27Al-MAS NMR spectra of both measurements, ex situ and in situ, are indicative of the presence of tetrahedral AlO4 moieties as the dominant Al coordination in the sample (Figure 4b). With the help of a 2D 31P-RFDR NMR experiment we checked for the presence of individual, separated phases in the crystallized material. With the employed mixing time of 32 ms, it can be assumed that only the homonuclear dipolar coupling between phosphorus sites within one microcrystallite (and hence within the same crystalline phase) contributes to off diagonal intensity in the 2D RFDR NMR spectrum. Thus, the number of individual phases present may be obtained from this experiment. In the spectrum (Figure 5a), two different groups of signals can be identified. The signals at -18 and -20 ppm exhibit off diagonal intensity among each other. The correlation between these two signals indicates that both belong to the same structure (KPO3). The six remaining signals at -1, -3, -7.5, -8, -12, and -12.5 ppm constitute the second group of signals

Figure 6. Compilation of 27Al-MAS NMR (a) and 31P-MAS NMR spectra (b) for samples KAlP_10 and KAlP_10B before and after annealing. The addition of 5% B2O3 clearly stabilizes the glassy network, as indicated by the lack of any change in the spectra of the boron containing sample after annealing.

exhibiting off diagonal intensity only within this group. No offdiagonal intensity is observed between peaks from different groups which confirms the existence of two different phases within the material. Having identified the existence of two different phosphate containing phases (KPO3 and a second unknown phase), the 31P-{27Al} CP-HETCOR NMR experiment can help to address the question whether all the aluminum present in the sample is contained in this second phase. Again, we make the assumption that a correlation signal, indicating spatial proximity between the relevant aluminum and phosphorus sites, is only possible if these are within the same phase. The 31P-{27Al} CP-HETCOR NMR spectrum is displayed in Figure 5b together with the 1D 31P- and 27Al-MAS NMR spectra as projections. As expected, no correlation signal between the 31P-MAS NMR signals at -18 and -20 ppm (assigned to KPO3) and any aluminum sites can be observed. On the other hand, all six 31P-MAS NMR signals from the second group exhibit a clear correlation to the 27AlMAS NMR signals. In addition, all 27Al signals show a correlation to at least one of the six 31P signals. From this, we may conclude that the unknown phase constitutes a mixed aluminophosphate network and that no phosphate-free aluminate phase exists in the sample. Boron-Containing Samples. The addition of small amounts of B2O3 to the base aluminophosphate glasses has been shown to reduce the crystallization tendency of the glasses.13 This is confirmed by a comparison of the 27Al- and 31P-MAS NMR spectra (ex situ) of the boron-containing and boron-free glass samples before the annealing process to those after annealing

Crystallization in Aluminophosphate Glasses

J. Phys. Chem. B, Vol. 111, No. 26, 2007 7533

Figure 7. (a) 11B-MQMAS NMR spectrum together with the projections onto the isotropic axis (F1) and the MAS axis (F2) for glass KAlP_10B. (b) 11B-MAS NMR spectra for sample KAlP_10B before and after annealing at 450 °C.

Figure 8. 11B-{31P}REDOR NMR data for sample KAlP_10B: (a) Top spectrum, rotor-synchronized 11B-MAS spin echo spectrum (τ ) 1.2 ms) defining the reference intensity S0 for the individual signals; middle spectrum, 11B-{31P}-REDOR NMR; bottom spectrum, difference of the two above spectra. (b) Normalized difference intensities (S0 - S)/S0, obtained at different dipolar evolution times τ; full circles, signal at -0.6 ppm; full triangles, signal at -2.5 ppm. The solid and dotted lines represent results from simulations assuming a B-P two-spin system (dotted line) and a B-P2 three-spin system, respectively. A dipolar coupling of 784 Hz, corresponding to the average distance of 2.71 Å (average B-P distance in Na5B2P3O13),31 was used for the calculations.

at 450 °C (cf. Figure 6). As borne out by the spectra, the base glass completely crystallizes under the annealing conditions (15 min at 450 °C), whereas in the boron containing sample, no change can be observed in the 27Al- or 31P-MAS spectra upon annealing, indicating an intact amorphous network. Boron enters the network as BO4- species, as obvious from the 11B-(MQ)MAS spectra compiled in Figure 7. Two signals at -0.6 ppm (CQ ) 0.78 MHz) and -2.5 ppm (CQ ) 0.43 MHz) can be identified. The boron incorporation entails a considerable reduction of the amount of five- and six-coordinated aluminum species present in the glass, in accordance with the generally accepted decrease of the mean Al coordination number with increasing O/P ratio, which in the present case is increasing from 3.5 in the base glass KAlP_10 to 3.70 in the boron containing sample KAlP_10B.20,21 Further, a reduction in the amount of pure Q20 phosphate units can be observed. From these observations, the structural role of the BO4 units within the aluminophosphate network seems to be closely related to that of the AlO4 species. This assumption is confirmed by an exploration of the second coordination sphere of a central boron

atom employing 11B-{31P} REDOR or 11B-{27Al}-TRAPDOR NMR spectroscopy. Whereas the 11B-{27Al} TRAPDOR NMR results indicate only a marginal B-Al dipolar coupling from which the presence of direct B-O-Al connectivity can be excluded (data not shown), the results of the 11B-{31P} REDOR NMR experiment (cf. Figure 8) present unambiguous proof for direct B-O-P connectivity. For the simulationssperformed employing the SIMPSON software29swe assumed a B-P distance of 2.71 Å and a B-P two spin interaction (dotted line in Figure 8b) or a B-P2 three spin interaction (solid line), respectively. From this, the presence of B(1P) and/or B(2P) species, i.e., BO4 units connected to 1 or 2 phosphate species, as found, e.g., in silver borophosphate glasses30 can be inferred. Thus a possible explanation for the increased resistance of the boron containing glasses toward crystallization is the presence of the identified B-O-P linkages which have to be cleaved in addition to the P-O-Al linkages prior to the formation of the crystalline pure phosphate and mixed aluminophosphate phases.

7534 J. Phys. Chem. B, Vol. 111, No. 26, 2007 Conclusion The presented in situ and ex situ NMR data allow for new insights into the structural evolution of the amorphous network in aluminophosphate glasses in the temperature range between glass transition and crystallization temperature. The most important observation is a gradual decrease in the aluminum coordination together with an increase in the fraction of phosphate tetrahedra connected to AlO4 units with increasing temperature, necessarily accompanied by the breaking of P-OAl bonds. These structural changes however prove to be reversible as long as the crystallization temperature is not reached. The results indicate a structural relaxation process characterized by a decrease in the average Al coordination number, accompanied by the breaking of Al-O-P bonds. This local demixing of structural units is possibly paving the way for the phase separation and crystallization, observed at T > Tc. Above Tc, two different crystalline phases are developed during devitrification: pure potassium phosphate (KPO3 or K4P2O7; depending on composition) and one unknown aluminophosphate compound. In the boron-containing glasses, boron enters the network as tetrahedral BO4 units, connected predominantly to phosphate species. Thus, B-O-P bond breaking has to be accomplished in addition to the P-O-Al bond breaking for the crystallization of the pure potassium phosphate and aluminophosphate phases, thus reducing the crystallization tendency. Acknowledgment. Financial support from the Deutsche Forschungsgemeinschaft is gratefully acknowledged. References and Notes (1) Vogel, J.; Holand, W.; Naumann, K.; Gummel, J. J. Non-Cryst. Solids 1986, 80, 34. (2) Weber, M. J. J. Non-Cryst. Solids 1990, 123, 208. (3) Donald, I. W.; Metcalfe, B. L. J. Non Cryst. Solids 2004, 348, 118. (4) Wilder, J. A. J. Non-Cryst. Solids 1980, 38/39, 879. (5) Kreidl, N. J.; Weyl, W. A. J. Am. Ceram. Soc. 1941, 24 (11), 372.

van Wu¨llen et al. (6) Dias, A. G.; Skakle, J. M. S.; Gibson, I. R.; Lopes, M. A.; Santos, J. D. J. Non-Cryst. Solids 2005, 351, 810. (7) Mysen, B. O. Contrib. Mineral. Petrol. 1998, 133, 38. (8) van Wu¨llen, L.; Schwering, G.; Naumann, E.; Jansen, M. Solid State Nucl. Magn. Reson. 2004, 26, 84. (9) Stebbins, J. F.; Farnan, I.; Williams, E. H.; Roux, J. Phys. Chem. Minerals 1989, 16, 763. (10) Stebbins, J. F.; Sen, S. J. Non-Cryst. Solids. 1998, 224, 80. (11) Stebbins, J. F.; Sen, S.; George, A. M. J. Non-Cryst. Solids 1995, 192+193, 298. (12) Stebbins, J. F.; Sen, S.; Farnan, I. Am. Mineral. 1995, 80, 861. (13) Donald, I. W.; Metcalfe, B. L.; Fong, S. K.; Gerrad, L. A. J. NonCryst. Solids 2006, 352, 2993. (14) Benett, A. E.; Ok, J. K.; Griffin, R. G.; Vega, S. J. Chem. Phys. 1992, 96, 8624. (15) Caravatti, P.; Bodenhausen, G.; Ernst, R. Chem. Phys. Lett. 1982, 89 (5), 363. (16) Hartmann, S. R.; Hahn, E. L. Phys. ReV. 1962, 128, 2042. (17) de Paul, S.; Ernst, M.; Shore, J.; Stebbins, J. F.; Pines, A. J. Phys. Chem. B 1997, 101, 3240. (18) Vega, A. J. Solid State Nucl. Magn. Reson. 1996, 1, 17. (19) Fyfe, C. A.; Grondey, H.; Mueller, K. T.; Wong-Moon, K. C.; Markus, T. J. Am. Chem. Soc. 1992, 114, 5876. (20) Wegner, S.; Tricot, G.; van Wu¨llen, L. Manuscript in preparation. (21) Brow, R. K.; Kirkpatrick, R. J.; Turner, G. L. J. Am. Ceram. Soc. 1993, 76 (4), 919. (22) Sen, S.; Xu, Z.; Stebbins, J. F. J. Non-Cryst. Solids 1998, 226, 29. (23) Richet, P.; Bottinga, Y. Rheology and Configurational Entropy of Silicate Melts; Stebbins, J. F., McMillan, P. F., Dingwell, D. B., Eds.; Structure, Dynamics and Properties of Silicate Melts, ReV. Mineral. 1995, 32, 67. (24) Stebbins, J. F. Dynamics and Structure of Silicate and Oxide Melts: Nuclear Magnetic Resonance Studies; Stebbins, J. F., McMillan, P. F., Dingwell, D. B., Eds.; Structure, Dynamics and Properties of Silicate Melts, ReV. Mineral. 1995, 32, 191. (25) Dubinsky, E. V.; Stebbins, J. F. Am. Mineral. 2006, 91, 753. (26) Brandriss, M. E.; Stebbins, J. F. Geochim. Cosmochim. Acta 1988, 52, 2659. (27) Stebbins, J. F.; Ellsworth, S. E. J. Am. Ceram. Soc. 1996, 79, 2247. (28) Grimmer, A. R.; Haubenreisser, U. Chem. Phys. Letters 1983, 99, 487. (29) Bak, M.; Rasmussen, J. T.; Nielsen, N. C. J. Magn. Reson. 2000, 147, 296. (30) Elbers, S.; Strojek, W.; Koudelka, L.; Eckert, H. Solid State Nucl. Magn. Reson. 2005, 27, 65. (31) Hauf, C.; Friederich, T.; Kniep, R. Z. Kristallogr. 1995, 210, 446.