J. Phys. Chem. B 1998, 102, 9033-9038
9033
Structural Details of Aqueous Attack on a Phosphate Glass by 1H/31P Cross-Polarization NMR R. M. Wenslow† and K. T. Mueller* Department of Chemistry, 152 DaVey Laboratory, The PennsylVania State UniVersity, UniVersity Park, PennsylVania 16802-6300 ReceiVed: June 1, 1998; In Final Form: August 13, 1998
One- and two-dimensional heteronuclear correlation solid-state nuclear magnetic resonance experiments allow a direct study of the chemistry of water incorporation into a simple phosphate glass network. One main pathway for water incorporation into phosphate glass networks is through the formation of P-OH groups, and previous 1H magic-angle-spinning NMR investigations of proton environments in silicate and phosphate glasses suggest that hydroxyl groups are involved in multiple hydrogen-bonding motifs with backbone units within the glass networks. The presence of multiple proton environments presents more complex structural issues for aqueous incorporation into glass systems, which have yet to be explained unambiguously. The spectral editing capabilities of two-dimensional cross-polarization NMR identify specific subunits in modified sodium phosphate glasses, and strong evidence is presented for the formation of intramolecular as well as weaker intermolecular hydrogen bonds. Using a combination of one- and two-dimensional 1H/31P crosspolarization NMR studies, a total of five distinct phosphate units are identified in a near-metaphosphate glass after aqueous treatment.
Introduction Structural studies investigating the reactions of water with oxide glass networks have utilized both infrared (IR) spectroscopy and solid-state nuclear magnetic resonance (NMR) to determine distinct proton environments as well as estimates of O-H‚‚‚O bond distances.1-4 1H magic-angle-spinning (MAS) NMR experiments on silicate glasses1 identified proton environments that are separated into two major chemical shift ranges depending on the strength of hydrogen bonding in the system. The first 1H isotropic chemical shift range (0-10 ppm referenced to TMS) contains proton resonances from species such as molecular water, and this group is characterized by relatively weak hydrogen bonding. The second range (11-16 ppm from TMS) contains resonances from protons in stronger hydrogenbonding environments. Proton environments from both chemical shift ranges have been identified in silicate1 and phosphate2 glasses, with the most likely origin for stronger hydrogen bonding being interactions between a hydroxyl group and a terminal oxygen bonded to a backbone atom. The structural ramifications of aqueous attack on glasses are extremely important for understanding the complex relationships between glass properties and glass composition. In sodium phosphate systems, this is especially true for near-metaphosphate (Na/P ) 1.0) and ultraphosphate (Na/P < 1) glasses due to the significant amount of water uptake in these regions of the sodium phosphate glass-forming system. Solid-state NMR spectroscopic investigations of aqueous attack on alkali-metal phosphate glasses have been limited, with early NMR investigations identifying either distinct proton environments using 1H MAS2 or specific phosphate tetrahedra using 31P MAS.3,4 The * To whom correspondence should be addressed. E-mail:
[email protected]. Fax: (814) 863-8403. † Current address: Merck & Company, 126 E. Lincoln Avenue, Rahway, NJ 07065.
major limitations associated with these MAS investigations have been the lack of sufficient spectral resolution brought about by both the extreme complexity of the amorphous sample and the insufficient MAS rotor frequencies. Recently, one-dimensional 1H/31P cross-polarization (CPMAS) experiments have been combined with the results from 1H and 31P MAS experiments, as well as spin-lattice and spin-spin relaxation time measurements, to provide the most comprehensive picture to date for the local structure of ultraphosphate glass systems containing large numbers of hydroxyl groups.5 In this current study, we combine high-speed MAS NMR experiments (at rotor frequencies up to 12 kHz) with the spectral editing capabilities of oneand two-dimensional heteronuclear correlation NMR experiments to study aqueous attack upon a near-metaphosphate glass system, with the subsequent identification of five different phosphate species and a correlation of three of these species with specific 1H NMR resonances. In previous work, Kohn and co-workers1 monitored proton environments in hydrous silicate glasses using 1H MAS NMR. Four separate proton resonances were resolved in a silica glass containing 8.7 mol % H2O. Also, in hydrous alkali-metal and alkaline-earth-metal disilicate glasses, strongly hydrogen-bonded Si-OH groups were identified. The strong hydrogen-bonded proton environment produced a resonance at high frequencies (11-17 ppm), and the most likely structural motif for this species is a proton in a hydroxyl species bridged between a Si-OH group and terminal oxygens (-O-Si) attached either to the same silicon unit (intramolecular) or neighboring silica tetrahedra (intermolecular). In studies of phosphate glass systems, Hosono and coworkers2 utilized 1H MAS NMR and IR spectroscopic investigations for the characterization of aqueous attack. The glasses analyzed in this investigation were both a barium and a lead phosphate glass containing 2 and 7 mol % water, respectively. The IR absorption spectra for both glasses displayed bands due
10.1021/jp9824252 CCC: $15.00 © 1998 American Chemical Society Published on Web 10/16/1998
9034 J. Phys. Chem. B, Vol. 102, No. 45, 1998 to hydroxyl stretching modes at approximately 2750 and 2250 cm-1 with comparable intensity. These two peaks are located at significantly lower wavenumbers than silanol stretching modes in SiO2 glass (3670 cm-1). The 1H MAS spectra for both phosphate glasses displayed a multicomponent resonance centered at approximately 13 ppm. This resonance was assumed to be composed of two components which could be deconvoluted into Gaussian bands by using a nonlinear least-squares fitting program. The peaks of the two Gaussian components were located at 15 and 11 ppm, respectively. The 1H resonance at 15 ppm was assigned to strongly hydrogen-bonded protons, similar to the hydrogen-bonded proton in hydrous alkali-metal silicate glasses examined by Kohn.1 MAS NMR investigations of 31P nuclei in phosphate glasses by Brow and co-workers4 concluded that water enters into the phosphate glass network, causing depolymerization similar to that caused by the addition of alkali-metal oxide modifiers. In 31P MAS NMR spectra, distinct resonances for phosphate tetrahedra containing protons in the second coordination sphere were not evident. Instead, it was assumed that protonated phosphate sites would contribute resonances at approximately the same frequency as their alkali-metal phosphate counterparts. With this approximation, phosphate site distributions were predicted based on the structural depolymerization model of Van Wazer.6 In a separate investigation, however, Hartmann and co-workers3 were able to resolve two distinct phosphate resonances when performing 31P MAS on a sodium metaphosphate (Na/P ) 1) glass. A distinct resonance at -8.6 ppm was assigned to a Q1(1H) phosphorus site (a Q1 phosphate tetrahedron with one hydroxyl group). The assignment of this resonance was supported by direct measurement of the chemical shift parameters using an analysis of spinning side-band intensities.7 The first resolution of a peak assignable to a Q1(1H) unit in a metaphosphate glass presents a more complex picture for aqueous attack than suggested by previous studies. The presence of spectral crowding or accidental degeneracies limits the confidence placed on full spectral assignments of resonances from complex glass systems. Further information about the local interactions of 1H and 31P nuclei can be obtained using heteronuclear correlation NMR methods.8 Quite recently, zinc ultraphosphate glasses containing large amounts of hydroxyl groups were studied by Mercier et al. using 1H/31P CPMAS methods.5 Bridging (Q2) phosphate sites bonded to either protons or zinc cations were identified by variable-contact-time CPMAS experiments. Three different phosphate species and at least three types of proton species were characterized. Measurements of spin-lattice and spin-spin relaxation times argued for a homogeneous distribution of the protons which are intimately connected to the Q2 phosphate species. A great deal of additional information, including direct spectral assignments, is available using multidimensional NMR correlation methods. We report initial studies of the fate of water incorporated into a near-metaphosphate glass by extending CPMAS measurements into a second spectral dimension. Preliminary assignments of specific 31P resonances are made with one-dimensional MAS and CPMAS experiments, but it is demonstrated that the use of two-dimensional NMR allows increased spectral resolution. A minority phosphate species, with a resonance initially unresolved in either one-dimensional NMR experiment, is uncovered using two-dimensional CPMAS methods. The new resonance is assigned as an orthophosphate species with an attached hydrogen atom exhibiting strong intramolecular hydrogen bonding.
Wenslow and Mueller
Figure 1. One-dimensional 1H MAS spectrum for a sodium phosphate glass, after allowing further degradation and depolymerization through attack by water. The MAS spectrum was obtained with 1024 onedimensional scans, spinning at 12 kHz, with pulse lengths of 5 µs corresponding to approximately 30° tip angles. A recycle delay of 1 s was used, and Lorentzian broadening of 100 Hz was applied.
Experimental Section A near-metaphosphate glass was prepared using a standard melt/quenching technique.9 NaH2PO4 (obtained from Aldrich and used as received) was placed in a platinum crucible and heated to 800 °C, where the melt was held for 1 h. The melt was then quenched in mineral oil. The phosphate melt lost phosphorus during the melt hold time, and elemental analysis revealed a Na/P ratio of 1.05. The glass was shown to be amorphous using powder X-ray diffraction analysis. Water (60 µL) was added to approximately 400 mg of glass as a drop of water placed onto a glass pellet taken from storage in mineral oil. The glass/water mixture was then crushed and mixed and immediately transferred to a dry 4.0-mm MAS rotor. The hydrated glass was also characterized by powder X-ray diffraction analysis, and no sharp peaks from crystalline material were observed. All solid-state NMR experiments were performed on a Chemagnetics CMX-300 NMR spectrometer (7.05-T magnetic field) using an Otsuka Electronics/Chemagnetics doubleresonance MAS probe. The 31P and 1H resonance frequencies are 120.376 and 297.368 MHz, respectively, at this magnetic field strength. Recycle delays for the 1H and 31P onedimensional MAS experiments were 1 and 60 s, respectively, while recycle delays in the 1H/31P CPMAS experiments were 0.5-1.0 s. All 1H spectra are referenced to TMS using an external methanol standard, while all 31P shifts are referenced to 85% H3PO4. Rotor frequencies were maintained at 12 kHz in all experiments with an active-feedback spin-rate controller. Results One-Dimensional MAS Experiments. The 1H MAS NMR spectrum of the “hydrated” glass is displayed in Figure 1. The 1H resonance at 1.2 ppm is assigned to protons from the mineral oil in which the glass was quenched due to both the small fullwidth at half-maximum (fwhm) of this resonance and the absence of the peak in the 1H MAS spectra obtained from similar phosphate glasses not quenched in mineral oil. 1H resonances at low frequencies (2-7 ppm) are assumed to be molecular water, which is relatively weakly bound to the phosphate glass network since these resonances significantly decrease in intensity when 1H MAS spectra are obtained from the water-treated glass after it has been heated to 140 °C. The remaining distinguishable 1H MAS resonances (11.2 and 15.7 ppm from TMS) are within the 1H isotropic chemical shift range for protons presumably involved in strong hydrogen bonding in silicates1 and as further studied for phosphate glasses by Hosono2 and Mercier.5 Based on a combination of crystallographic data for hydrogen-bond lengths in crystalline compounds joined with
Aqueous Attack on a Phosphate Glass
Figure 2. One-dimensional 31P MAS spectra for a water-modified sodium phosphate glass. (a) The entire 31P MAS spectrum for a modified phosphate glass is displayed. Spinning side bands in the spectrum are labeled with an asterisk. (b) An expanded portion of the same spectrum. Sixty-four scans were averaged, with pulse lengths of 5 µs corresponding to approximately 30° tip angles, and a recycle delay of 60 s. Lorentzian broadening of 100 Hz was applied. 1H
NMR isotropic chemical shift data, an estimated O-H‚‚‚O distance of approximately 2.5 Å is obtained for resonances in this chemical shift range.10 In silicates, this estimate suggests that the protons within this range of shifts may be involved in Si-OH‚‚‚-O-Si bonding arrangements, with both groups possibly attached to the same silicon atom. This theory was also supported by data from Raman spectroscopy.11 The high-speed 31P MAS spectrum (Figure 2) of the watermodified phosphate glass displays multiple and distinct isotropic 31P resonances. This phosphate glass contains at least three different phosphate environments, accounting for distinct resonances at 1.9, -6.3, and -18.9 ppm. The resonance at 1.9 ppm is within the chemical shift range for a Q1 phosphate species determined by Brow.4 The major 31P resonance at -18.9 ppm is assigned to a Q2 phosphate tetrahedron, which is the main component of all metaphosphate glasses. The resonance at -6.3 ppm is outside of the chemical shift range for the three basic units commonly found in phosphate glasses (Q1, Q2, and Q3). The only reported occurrence of a phosphorus resonance for an alkali-metal phosphate glass in this chemical shift range was for the proposed Q1(1H) site observed in the 31P MAS spectra of a sodium metaphosphate glass analyzed by Hartmann and co-workers.3 Therefore, we make a tentative assignment of this resonance arising from the presence of a substantial number of Q1(1H) tetrahedra. One-Dimensional 1H/31P CPMAS Experiments. Onedimensional 1H/31P CPMAS spectra, obtained at two different contact times, are shown in Figure 3, where they are also compared to the 31P MAS spectrum. In the spectrum obtained with a 0.2-ms contact time (Figure 3b), two 31P resonances appear at isotropic shifts of -5.3 and -18.9 ppm. The resonance at -5.3 ppm has a fwhm of approximately 10 ppm and is asymmetric. The shift and asymmetry of the resonance at -5.3 ppm is possibly due to the presence of a third resonance which is unresolvable in either the one-dimensional 31P MAS or one-dimensional 1H/31P CPMAS spectra. Comparing the short-contact-time 1H/31P CPMAS spectrum to the isotropic resonances obtained from 31P MAS, it appears that the resonance at 1.9 ppm has diminished significantly in intensity, while the broader resonance at -6.3 ppm in the 31P MAS spectrum has shifted to -5.3 ppm in the 1H/31P CPMAS spectrum. The resonance at -18.9 ppm in the MAS spectrum has also shifted
J. Phys. Chem. B, Vol. 102, No. 45, 1998 9035
Figure 3. (a) 31P MAS spectrum from a modified phosphate glass compared to one-dimensional 1H/31P CPMAS spectra acquired at (b) 0.2-ms and (c) 7.0-ms contact times. The 1H/31P CPMAS spectra were acquired with 4096 scans and recycle delays of 500 ms. Lorentzian broadening of 100 Hz was applied.
slightly to -18.4 ppm in the CPMAS spectrum of Figure 3b. The disappearance of the resonance at 1.9 ppm in the 1H/31P CPMAS spectrum supports the assignment of this peak to a Q1 phosphate species influenced only by sodium cations in its second coordination sphere. Also apparent when comparing 1H/31P CPMAS spectra to the isotropic 31P spectra is the drastic alteration of relative intensities for the two resonances. In the 31P MAS spectrum, the resonance at -18.9 ppm accounts for approximately 87% of the intensity for all three resonances, while the resonance at -6.3 ppm accounts for only 7% of the combined intensities. In the 1H/31P CPMAS using a 0.2-ms contact time, however, the lower frequency resonance now accounts for only 32% of the total intensity. CPMAS experiments using a much longer contact time (7 ms, with the data shown in Figure 3c) reveal an increase in relative intensity of the lower frequency resonance, but it still has not gained the same proportion as it occupied in the 31P MAS spectrum of Figure 3a. To monitor the different polarization-transfer dynamics for the phosphorus resonances, more in-depth variable-contact-time CPMAS experiments were performed. In variable-contact-time experiments, a faster transfer rate is evidence of larger dipolar coupling between protons and the 31P contained within phosphate tetrahedra, argued to be caused by either more protons on average surrounding the P site or shorter 1H to 31P internuclear distances. In the water-modified glass, the asymmetric resonance centered at -5.3 ppm in the 1H/31P CPMAS spectra possessed faster cross-polarization rates compared with the 31P resonance at -18.4 ppm. This was also evident when the resonance was separated into two components centered at -2.4 and -5.8 ppm (vide infra), although full deconvolution of this asymmetric resonance was hindered by the lack of site resolution. This indicates that the proton environment surrounding the 31P nuclei contributing to the resonance at -18.4 ppm is significantly different than the environments surrounding nuclei contributing to the remaining CPMAS resonances, and we attribute this difference to longer 1H/31P internuclear distances.
9036 J. Phys. Chem. B, Vol. 102, No. 45, 1998
Wenslow and Mueller
Figure 4. Two major forms of hydrogen bonding proposed to occur in the hydrated phosphate glass, as discussed in the text.
Figure 6. 1H and 31P MAS subspectra from the two-dimensional CPMAS spectrum (a-e), displaying specific frequency-selected 1H and 31 P signals. The 1H/31P CPMAS spectrum obtained with a contact time of 0.2 ms is shown for reference.
Figure 5. Two-dimensional 1H/31P CPMAS correlation spectrum from the modified phosphate glass, displaying correlations between the resonances in the MAS spectra of 1H and 31P nuclei. Contours are placed from 8 to 90% of the maximum intensity in nine equal steps. The dwell time in the first (1H) spectral dimension was 50 µs, while the dwell time in the second (31P) dimension was 16.67 µs. A total of 2048 scans were acquired for each of 128 experiments with an array of 1H evolution times before the cross-polarization step. Line broadenings of 100 Hz were applied in each frequency dimension.
One possible source for shorter 1H- to 31P-internuclear distances for the components of the resonance centered at around -5.3 ppm would be the presence of intramolecular hydrogen bonding (as shown in Figure 4) or uneven sharing of a proton in intermolecular hydrogen bonding. The O‚‚‚H-O distance, and hence the 1H- to 31P-internuclear distance, could be longer for one member of the pair of 31P nuclei involved in intermolecular hydrogen bonding, depending on the affinity for protons of each phosphate group. A second internuclear hydrogen-bonding arrangement could also be imagined, where the proton is shared between chains. The exact placement of the proton would be affected by the relative charges of the associated oxygen species, and the exact form of the potential well could even contain a number of minima. Although the differences in cross-polarization rates for the different 31P resonances may be due to hydroxyl protons being involved in different hydrogen-bonding schemes, final confirmation of local bonding environments can only be made by determining which phosphate species are connected with each type of proton detected in the 1H MAS spectrum. Two-Dimensional 1H/31P CPMAS Experiments. A direct correlation of proton resonances with distinct phosphate units is obtained using two-dimensional versions of CPMAS experiments.13 A two-dimensional correlation plot of proton and phosphorus resonances displays cross-peak intensity when heteronuclei have exchanged magnetization via the crosspolarization process. The result from a two-dimensional 1H/ 31P CPMAS experiment performed with a 0.2-ms contact time on the water-treated glass is displayed in Figure 5. The twodimensional contour plot shows only two proton resonances, at 10.9 and 15.8 ppm, and three distinct correlations: (A) a 1H resonance at 10.9 ppm spatially correlated to a 31P resonance at -18.4 ppm; (B) a 1H resonance at 10.9 ppm
spatially correlated to a 31P resonance at -5.8 ppm; (C) a 1H resonance at 15.8 ppm spatially correlated to a 31P resonance at -2.4 ppm. Notably, only one 31P resonance (-2.4 ppm) is correlated through space to the 1H resonance at 15.8 ppm. The short contact time in these experiments was chosen to select only for short distances in the CP experiment. 31P MAS slices correlated to the distinct isotropic 1H resonances (15.8 and 10.9 ppm) are displayed in Figure 6a and 6b, respectively. The 31P MAS slice in Figure 6a contains a predominant resonance at -2.4 ppm (fwhm ) 7 ppm) and a weak secondary resonance at approximately -19 ppm, while the slice in Figure 6b contains resonances at -5.8 ppm (fwhm ) 5 ppm) and -18.4 ppm. From these slices, it is evident that the asymmetry (and shift in frequency) associated with the one-dimensional 1H/31P CPMAS resonance at -5.3 ppm (fwhm ) 10 ppm) is due to the presence of two distinct phosphorus sites: one at -2.4 ppm coupled strongly to the 1H resonance at 15.8 ppm and one at -5.8 ppm coupled strongly to the 1H resonance at 10.9 ppm. It is only through the spectral editing capabilities of two-dimensional CPMAS that the minor 31P resonance at -2.4 ppm is fully resolved. Two-dimensional 1H/31P CPMAS experiments performed at longer contact times (not shown) display the total connectivity of all 1H resonance to all 31P resonances due to the increased internuclear distances accessed with longer contact time CPMAS experiments. Discussion Utilizing the full set of one- and two-dimensional NMR results, a simple model for the attack of water on a nearmetaphosphate glass is now presented. As prepared, the glass was quenched in mineral oil to retard any attack from atmospheric or other water. A quantity of water was then added, approximately equimolar with the number of bridging phosphate units (0.003-0.004 mol). Ideally, a metaphosphate glass is composed of Q2 phosphate tetrahedra in an infinitely long polymeric chain, although it could also contain sets of rings of various lengths and tortuosities. Glasses slightly above the metaphosphate composition, as is the case for the glass studied here (Na/P ) 1.05 as determined by elemental analysis), are conventionally viewed as composed of Q2 phosphate chains, varying in length, with Q1 phosphate tetrahedra at each end of the chains.6 Therefore, the amount of water was in excess of the number of Q1 phosphate units present, although not all water reacted directly with the glass. The 1H MAS spectrum of Figure 1 verifies the presence of molecular water. Two-dimensional CPMAS experiments also demonstrate that this water is not
Aqueous Attack on a Phosphate Glass
Figure 7. Two simple pathways for aqueous attack on a finite-length phosphate chain containing both Q1- and Q2-type phosphate species.
strongly bound to the phosphate network, due to the absence of any CP signals containing a molecular water peak. Water attack upon the phosphate chains in the dry glass can occur theoretically by two distinct pathways as outlined in Figure 7. If water attacks through pathway 1 (in the middle of a phosphate chain), two Q1(1H) units will form. In crystalline phosphate compounds, Q1(1H) units possess isotropic chemical shifts for 31P in the range of -7 to -20 ppm.3,14 For example, a Q1(1H) group in a crystalline sodium phosphate, Na2H2P2O7, possesses an isotropic chemical shift of -8.0 ( 0.3 ppm.3 If the water molecule attacks the phosphate chain through a second avenue (pathway 2 in Figure 7), this attack will produce a Q1(1H) phosphate unit along with the formation of a Q0(1H) species. In crystalline phosphates, Q0(1H) phosphate units appear in a chemical shift range of 5 to -7 ppm.3 The presence of hydrogen-bonded protons is evident through the isotropic chemical shift values of the two higher frequency proton resonances observed in the 1H MAS spectrum (Figure 1). To identify the origin of this hydrogen bonding, as well as the structural source of the two proton resonances, it is necessary to explore the possibilities for hydrogen bonding in the two protonated species (Q1(1H) and Q0(1H)) formed through the depolymerizing action of water. The Q1(1H) species, formed through either of the two aqueous attack pathways of Figure 7, is still “connected” to the phosphate backbone through its remaining bridging oxygen atom. The phosphate chain contains predominantly Q2 bridging units, each containing two terminal oxygens with their associated negative charges. Also, the Q1(1H) unit contains two formally negatively charged terminal oxygens. Therefore, the terminal hydroxyl group associated with the Q1(1H) unit has the possibility of hydrogen bonding to either the terminal oxygens on a nearby Q2 unit, the remaining terminal oxygens on the same phosphate tetrahedron, or possibly a Q1 species at the end of another phosphate chain. In the case of intermolecular bonding to a Q2 species, a hydroxyl proton will be correlated and dipole-coupled to 31P nuclei both within the Q1(1H) unit itself and within Q2 units from the backbone. For an intramolecular bond, the dominant coupling should be with 31P nuclei in the Q1(1H) unit only, and short-contact-time two-dimensional CPMAS experiments should reveal simple resonances. For the case of two Q1 units sharing a proton in an intermolecular hydrogen bond, either the chemical shifts of the Q1 species will be very different (with the proton shared unequally) or the chemical shift values should be comparable and the average bond distances very similar. In this case,
J. Phys. Chem. B, Vol. 102, No. 45, 1998 9037 however, a strongly bonded proton (as evidenced by larger chemical shift values) would lead to steric and/or electrostatic instability in the association of two chain ends. Therefore, a Q1(1H)-Q1(1H) connection is considered highly improbable and inconsistent with the full set of one-dimensional and twodimensional NMR data. In the contour plot from the two-dimensional 1H/31P CPMAS experiment performed at 0.2-ms contact time (Figure 5), it is evident that the proton resonance at approximately 11 ppm is indeed correlated to two phosphorus resonances, one at -18.4 ppm (Q2 backbone unit) and the other resonance centered at -5.8 ppm (previously assigned to Q1(1H)). The maxima of the two resonances occur at exactly the same 1H frequency, and the resonance at -18.4 ppm is definitely shifted from the predominant Q2 resonance observed in the 31P MAS spectrum. The 1H chemical shift of the resonance (11.2 ppm) is indicative of a hydrogen-bonded proton, and from variable-contact-time experiments performed on this glass, it was shown that the rate of polarization transfer to the Q2 unit is slower than the rate obtained for transfer to the higher frequency resonances. Comparing 1H-31P dipole couplings involving single protons from the terminal hydroxyl group on a Q1(1H) unit to phosphorus atoms on either a Q2 backbone unit or a phosphorus atom from the same Q1(1H) unit, it is evident that the dipolar coupling should be stronger when involving protons on the same unit since the 1H-31P internuclear distance will be shorter in this interaction. However, the two-dimensional NMR data show a proton resonance at 11.2 ppm clearly connected to two 31P resonances. This is explained by assuming the negatively charged terminal oxygens on a Q2 tetrahedron will attract the proton attached to the Q1(1H) unit and slightly lengthen the H-P internuclear distance (where the phosphorus atom is located on the Q1(1H) unit). The distance from the proton to the 31P within the Q2 unit is longer, thus leading to a slower rate of CP magnetization transfer. The 31P resonance for the Q2 associated in this way with a proton is also slightly shifted relative to the bulk Q2 species. However, the difference in shift for the Q2 site upon association with a proton is not nearly as great as that for a Q1(1H) unit when compared to the unprotonated Q1 species, a shift of -8 ppm in this glass. Therefore, the 31P resonance at -5.8 ppm is assigned to a Q1(1H) unit connected to a phosphate backbone by a lone bridging oxygen, with a weak hydrogen bond to a Q2 unit. Whether this second Q2 phosphate unit is within the same chain is not clear from the NMR data. In this context, a “weaker” hydrogen bond is characterized by a smaller net transfer of the hydrogen atom. In summary, this assignment is supported by the 31P isotropic chemical shift values, the isotropic shift value of the proton to which these specific sites are correlated, relative cross-polarization rates obtained from variable-contact-time 1H/31P CPMAS experiments, and the appearance of two different resonances correlated strongly with the exact same proton resonance frequency in the two-dimensional CPMAS map. A well-resolved 31P resonance at -2.4 ppm, which is hidden in the one-dimensional 31P MAS spectrum, is revealed only through the spectral editing capabilities of two-dimensional 1H/ 31P CPMAS. From the contour plot of the two-dimensional 1H/31P CPMAS experiment (Figure 5), it is noted that this 31P resonance is connected only to the 1H resonance at 15.8 ppm. Therefore, two very similar 31P species (at least lying within the chemical shift range of the resonance) must be connected through a strong hydrogen bond. The chemical shift of the 31P resonance is not unambiguous evidence for a confident assignment, as it lies in a region not fully explored in phosphate
9038 J. Phys. Chem. B, Vol. 102, No. 45, 1998 glasses. Likely candidates are indeed the Q1(1H) site discussed earlier, as well as the Q0(1H) sites. In crystalline compounds, a resonance at -2.4 ppm falls within the range of Q0(1H) sites, and a Q1(1H)-Q1(1H) connectivity was discussed above and considered unlikely. The second avenue shown above for aqueous attack on phosphate chains (pathway 2 in Figure 7) predicts the formation of Q0(1H) units as well as a Q1(1H) species. The Q0(1H) unit will not be covalently connected to the phosphate backbone due to the absence of bridging oxygens. With the lack of terminal oxygens from nearby Q2 units in the backbone, the closest remaining terminal species are those oxygens on the orthophosphate itself. Intramolecular hydrogen bonding associated with a Q0(1H) unit would account for the faster cross-polarization rates associated with this site compared to the Q2 phosphate site, and therefore, we propose that this resonance originates from a Q0(1H) unit with intramolecular hydrogen bonding. The hydrogen bonding is different in the proposed Q0(1H) unit than the bonding in the Q1(1H) unit due to the nonparticipation of the hydroxyl group from the Q0(1H) unit in bonding with the phosphate backbone. This distinct difference in hydrogen-bonding environments for hydroxyl groups is proposed as the explanation for the difference in isotropic chemical shifts for the two different proton resonances. The hydroxyl proton on the Q0(1H) unit is considered to participate in stronger hydrogen bonding, leading to a higher frequency resonance for the proton. This reasoning leads to the assignment of the 1H MAS resonance at 15.8 ppm to a proton involved entirely in intramolecular hydrogen bonding within the Q0(1H) unit. Once again, the 1H MAS resonance at 11.2 ppm is assigned to a proton involved in weaker intermolecular hydrogen bonding. The solid-state NMR results leading to these assignments provide the first direct spectroscopic evidence of intramolecular hydrogen bonding in a glass of any composition and demonstrate the increased structural information available from the application of an array of related, but complimentary, NMR methods. Conclusions One-dimensional 1H and 31P MAS NMR experiments have been combined with one- and two-dimensional CPMAS connectivity experiments to monitor aqueous attack on a nearmetaphosphate glass. The results from these experiments provide an improved understanding of the structural results of aqueous attack, revealing a more complex chemistry than suggested by past investigations. The water-modified glass contained the following five distinguishable phosphate units, with the associated 31P isotropic chemical shifts given in
Wenslow and Mueller parentheses: (1) a Q1 phosphate site with only sodium nextnearest-neighbors (δiso ) 1.9 ppm); (2) a Q2 phosphate site with only sodium next-nearest-neighbors (δiso ) -18.9 ppm); (3) a Q0(1H) unit in which the hydroxyl group is involved in strong intramolecular hydrogen bonding (δiso ) -2.4 ppm); (4) a Q1(1H) unit in which the hydroxyl group is predominantly involved in weaker intermolecular hydrogen bonding (δiso ) -5.8 ppm); (5) a Q2 phosphate site which is the next neighbor to a Q1(1H) unit such that weaker intermolecular hydrogen bonding can occur (δiso ) -18.4 ppm). A simple structural and chemical model was discussed, consistent with the formation or existence of all five phosphate units. Acknowledgment. This report is based upon work supported by the National Science Foundation under Grant No. DMR9458053 and an Arnold and Mabel Beckman Foundation Young Investigator Award to K.T.M. Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for partial support of this research. The NMR spectrometer and probe used in this research were upgraded and obtained, respectively, through the assistance of the National Science Foundation under Grant No. DMR9413674. We also thank an anonymous reviewer for helpful comments. References and Notes (1) Kohn, S. C.; Dupree, R.; Smith, M. E. Nature 1989, 337, 539541. (2) Hosono, H.; Abe, Y.; Deguchi, K. J. Non-Cryst. Solids 1992, 142, 103-107. (3) Hartmann, P.; Vogel, J.; Schnabel, B. J. Magn. Reson., Ser. A 1994, 111, 110-114. (4) Brow, R.; Kirkpatrick, R.; Turner, G. J. Non-Cryst. Solids 1990, 116, 39-45. (5) Mercier, C.; Montagne, L.; Sfihi, H.; Palavit, G.; Boivin, J. C.; Legrand, A. P. J. Non-Cryst. Solids 1998, 224, 163-172. (6) Van Wazer, J. R. Phosphorous and its Compounds; Interscience: New York, 1958; Vol. 1. (7) Herzfeld, J.; Berger, A. E. J. Chem. Phys. 1980, 73, 6021-6030. (8) Wenslow, R. M.; Fiske, K.; Mueller, K. T. Solid-State Nuclear Magnetic Resonance of Inorganic Materials; ACS Symposium Series; American Chemical Society: Washington, DC, 1998, in press. (9) Elliott, S. R. Physics of Amorphous Materials; Longman: New York, 1984. (10) Eckert, H.; Yesinowski, J. P.; Silver, L. A.; Stolper, E. M. J. Phys. Chem. 1988, 92, 2055-2064. (11) McMillan, P. F.; Remerle, R. L. Am. Mineral. 1986, 71, 772-778. (12) Mehring, M. Principles of High-Resolution NMR in Solids, 2nd ed.; Springer-Verlag: Berlin, 1983. (13) Caravatti, P.; Bodenhausen, G.; Ernst, R. R. Chem. Phys. Lett. 1982, 89, 363-367. (14) Larbot, P. A.; Durand, J.; Norbert, A.; Cot, L. Acta Crystallogr. Sect. C 1983, 39, 6.
