Structural Studies of NaPO3−MoO3 Glasses by Solid-State Nuclear

J. Phys. Chem. B 2007, 111, 10109-10117

10109

Structural Studies of NaPO3-MoO3 Glasses by Solid-State Nuclear Magnetic Resonance and Raman Spectroscopy Silvia H. Santagneli, Carla C. de Araujo,† Wenzel Strojek, and Hellmut Eckert* Institut fu¨r Physikalische Chemie, Westfa¨lische Wilhelms-UniVersita¨t Mu¨nster, Corrensstrasse 30, D-48149 Mu¨nster, Germany

Gae1 l Poirier, Sidney J. L. Ribeiro, and Younes Messaddeq Departamento de Quı´mica Geral e Inorgaˆ nica, Instituto de Quı´mica-UNESP, Araraquara-SP, Brazil, CEP 14800-200 ReceiVed: April 13, 2007; In Final Form: June 29, 2007

Vitreous samples were prepared in the (100 - x)% NaPO3-x% MoO3 (0 e x e 70) glass-forming system by a modified melt method that allowed good optical quality samples to be obtained. The structural evolution of the vitreous network was monitored as a function of composition by differential scanning calorimetry (DSC), Fourier transform infrared spectroscopy (FT-IR), Raman scattering, and solid-state nuclear magnetic resonance (NMR) for 31P, 23Na, and 95Mo nuclei. Addition of MoO3 to the NaPO3 glass melt leads to a pronounced increase in the glass transition temperatures up to x ) 45, suggesting a significant increase in network connectivity. For this same composition range, vibrational spectra suggest that the Mo6+ ions are bonded to some nonbridging oxygen atoms (Mo-O- or Mo)O bonded species). Mo-O-Mo bond formation occurs only at MoO3 contents exceeding x ) 45. 31P magic-angle spinning (MAS) NMR spectra, supported by twodimensional J-resolved spectroscopy, allow a clear distinction between species having two, one, and zero P-O-P linkages. These sites are denoted as Q(2)2Mo, Q(2)1Mo, and Q(2)0Mo, respectively. For x < 0.45, the populations of these sites can be described along the lines of a binary model, according to which each unit of MoO3 converts two Q(2)nMo sites into two Q(2)(n+1)Mo sites (n ) 0, 1). This structural model is consistent with the presence of tetrahedral Mo()O)2(O1/2)2 environments. Indeed, 95Mo NMR data suggest that the majority of the molybdenum species are four-coordinated. However, the presence of additional six-coordinate molybdenum in the MAS NMR spectra indicates that the structure of these glasses may be more complicated and may additionally involve sharing of network modifier oxide between the network formers phosphorus and molybdenum. This latter hypothesis is further supported by 23Na{31P} rotational echo double resonance (REDOR) data, which clearly reveal that the magnetic dipole-dipole interactions between 31P and 23Na are increasingly diminished with increasing molybdenum content. The partial transfer of modifier from the phosphate to the molybdate network former implies a partial repolymerization of the phosphate species, resulting in the formation of Q(3)nMo species and accounting for the observed increase in the glass transition temperature with increasing MoO3 content that is observed in the composition range 0 e x e 45. Glasses with MoO3 contents beyond x ) 45 show decreased thermal and crystallization stability. Their structure is characterized by isolated phosphate species [most likely of the P(OMo)4 type] and molybdenum oxide clusters with a large extent of Mo-O-Mo connectivity.

Introduction Transition metal ion-containing phosphate glasses display several interesting physicochemical properties that lead to many different applications of these glasses in materials science. In fact, the incorporation of transition metal ions in phosphate glasses is generally known to improve the chemical resistance against atmospheric moisture1,2 as well as the thermal stability against devitrification.1-6 Moreover, the electronic properties, ionic conductivities, and optical absorption in the visible are modified to a large extent.7-17 In particular, tungsten oxidesodium phosphate glasses with high WO3 concentrations were * Author to whom correspondence should be addressed: phone 49-2518329161; fax 49-251-8329159; e-mail [email protected]. † Present address: Max Planck Institute for Solid State Research, 70569 Stuttgart, Germany.

intensively investigated for their optical properties such as upconversion phenomena when doped with Tm3+ 18,19 or twophoton nonlinear optical absorption.20 In addition, a new volumetric photochromic effect was observed in samples with high levels of WO3 incorporation under exposure to visible laser light.21 These characteristics are strongly related to the structural peculiarities of the vitreous network. In fact, structural studies of these tungsten phosphate glasses indicated the formation of clusters containing octahedrally coordinated tungsten species22,23 in the sample most highly concentrated in WO3, and it has been strongly suggested that these highly polarizable clusters are responsible for the nonlinear and photochromic properties. Molybdenum phosphate glasses are also well-known and have been shown to be particularly interesting for their ionic and electronic transport properties.24-27 Structural investigations have suggested that the effect of molybdenum incorporation into

10.1021/jp072883n CCC: $37.00 © 2007 American Chemical Society Published on Web 08/07/2007

10110 J. Phys. Chem. B, Vol. 111, No. 34, 2007 phosphate glasses is essentially similar to that of tungsten and results in the depolymerization of the phosphate chains. Thus, it is assumed that the vitreous network is mainly built from corner-sharing [MoO6/2] octahedral and phosphate tetrahedral units.28-33 The study of the optical properties has been limited by the poor transparency of the glasses in the visible and nearinfrared regions, generally attributed to the absorption of Mo5+ species,34-38 which create color centers due to oxygen vacancies.39,40 In this work, we report on vitreous samples in the binary system NaPO3-MoO3. A modified glass synthesis method was developed in order to obtain transparent samples with high MoO3 concentrations. Structural changes as a function of MoO3 concentration are probed by Fourier transform infrared spectroscopy (FT-IR), Raman scattering, and multinuclear solidstate NMR spectroscopy. Based on the experimental evidence, a comprehensive structural model is developed, which rationalizes the compositional dependence of the macroscopic physical properties of these glasses. Experimental Section Materials and Methods. Vitreous samples in the binary system (100 - x)% NaPO3-x% MoO3 (0 < x e 70) were prepared by conventional melt-cooling with molybdenum oxide from Alfa, 99+%, and sodium polyphosphate from Acros, 99+%. The powdered starting materials were mixed and heated at 400 °C for 1 h to remove water and adsorbed gases. For 0 < x e 15 samples, the batches were melted in a platinum crucible at 900 °C and the resulting liquid was kept at this temperature for 60 min to ensure homogenization and fining. Finally, melts were quenched in brass molds preheated at a temperature around 20 °C below Tg. For samples with 15 < x e 40, melting was also performed at 900 °C, and from that temperature, melts were slowly cooled at 1 °C/min to room temperature. For samples with 40 < x e 70, melting was performed at 1000 °C. From that temperature, slow cooling at 1 °C/min to room temperature was applied. The amorphous state was checked by X-ray powder diffraction. Differential scanning calorimetry (DSC) measurements were conducted with a Netzsch DSC-200 apparatus, using a heating rate of 10 K min-1. FT-IR spectra were obtained in a Nicolet Avatar 320 spectrometer. A diamond attenuated total reflectance (ATR) accessory was used. Raman scattering measurements were obtained with a Jobin Yvon Horiba HR800 instrument, operating with a Nd/YAG laser at 532 nm. 31P and 23Na NMR experiments were carried out at room temperature on a Bruker DSX-500 spectrometer, with a 4 mm magic-angle spinning (MAS) NMR probe operated at a spinning speed of 15 kHz. 31P NMR measurements were carried out at 202.5 MHz, with 90° pulses of 3.6 µs length and a recycle delay of 65 s. Chemical shifts are reported relative to 85% H3PO4. Signal deconvolutions into Gaussian components were done with the DMFIT software package.41 23Na triple-quantum MAS NMR42 data were obtained at 132.3 MHz, by the three-pulse sequence (zero-quantum filtering) method43 under conditions similar to those previously reported.44 The first two hard pulses were 4.3 and 1.5 µs in length, respectively, and transmitted at radio frequency amplitudes corresponding to a nutation frequency of 110 kHz. The soft detection pulse was 10 µs in length (nutation frequency 10 kHz). Spinning speed was 15 kHz and sampling in the t1 dimension was done with a dwell time of 17.9 µs (1/4 rotor period). Depending on sodium content, 120288 scans were taken per t1 increment. 95Mo NMR measurements were carried out at 32.6 MHz, with a 7 mm MAS NMR probe operated at a spinning frequency of 4 kHz. Data were

Santagneli et al. acquired with a rotor-synchronized spin-echo sequence, using solid 90° and 180° pulses of 2.0 and 4.0 µs length, respectively. A recycle delay of 1 s was used, which was shown to yield representative spectra. Chemical shifts are reported relative to a 2 M Na2MoO4 aqueous solution (pH ) 11). Different phosphate sites were identified from the different connectivity patterns of each of the Qn sites, through the usage of through-bond 31P-O-31P spin-spin coupling, applying twodimensional (2D) homonuclear J-resolved MAS NMR spectroscopy.45,46 In this experiment, the modulation of a rotorsynchronized spin echo due to the homonuclear indirect 2J(P-P) coupling is recorded. Fourier transformation of the twodimensional time domain data set yields a 2D spectrum exhibiting J-resolution in the F1 dimension. The applicability of this method to phosphate glasses has been shown recently47 and was applied by us in a previous study of NaPO3-WO3 glasses.48 The data were measured at 162.0 MHz on a Bruker DSX 400 spectrometer. The spin echo sequence including a z-filter was used, incorporating the 32-step phase cycle reported in ref 45. The π/2 pulse lengths were around 1.5 µs, and a z-filter delay of one rotor period was used at a MAS frequency of 14 kHz. Depending on the sample, echoes were recorded for evolution times of up to 80 ms, corresponding to 50 t1 increments. For the 2D spectra, the States method49 was used to obtain pure absorption-phase spectra and the t1 increment was rotor-synchronized to minimize spinning side bands in the F1 dimension. For each t1 slice, 32 scans were recorded with a recycle delay of 50 s. The 2D time domain data were apodized with a Gaussian function prior to Fourier transformation. To quantify the strength of 23Na-31P interactions, 23Na{31P} rotational echo double resonance (REDOR) experiments were conducted on representative samples, with the standard sequence of Gullion and Schaefer,50 augmented by the compensation scheme introduced in ref 51. Optimum π pulse lengths for the decoupling channel were set by maximizing the REDOR difference signal ∆S at a chosen dephasing time. The experiments were conducted on a Bruker DSX 500 spectrometer with a special double resonance probe (frequencies 202.5 and 132.3 MHz). The π-pulse lengths of 31P and 23Na were 5.3-6.0 and 5.7 µs, respectively. Phase cycling according to the XY4 scheme52 was used for the 31P pulses and only a third of the 4 mm MAS rotor was filled in order to optimize radio frequency homogeneity. Normalized difference signal intensities ∆S/S0 [corresponding to the signal amplitudes without (S0) and with (S) 31P pulsed irradiation] were plotted as a function of dipolar evolution time N·Tr, and second-moment values M2(23Na{31P}) were extracted from the parabolic fits to the initial decay regime (0 < ∆S/S0 < 0.2) by the procedures described in ref 51. Results, Data Analysis, and Interpretation Figure 1 shows DSC curves obtained for the prepared glasses together with the curve obtained for the precursor NaPO3. For this last sample the glass transition temperature (Tg) and the onset of crystallization (Tx) were observed at 290 and 325 °C, in agreement with the literature. With the addition of MoO3 a continuous, almost linear increase in Tg is observed, up to 414 °C observed for the sample with x ) 45. A remarkable increase in the thermal stability can be inferred from the complete absence of a crystallization peak for the samples with 15 e x e 45 over a temperature range up to 900 °C [as confirmed by additional differential thermal analysis (DTA) experiments, not shown]. For MoO3 contents higher than x ) 45, Tg is observed to decrease down to 352 °C observed for the sample with composition x ) 70 (see Table 1 and Figure 2). A

Structural Studies of NaPO3-MoO3 Glasses

J. Phys. Chem. B, Vol. 111, No. 34, 2007 10111

Figure 2. Compositional dependence of Tg in (NaPO3)100-x(MoO3)x glasses.

Figure1. DSCcurvesforglasssamplesinthesystem(NaPO3)100-x(MoO3)x.

TABLE 1: Molar Compositions and Tg and Tx Values of the NaPO3-MoO3 Glasses sample composition (% NaPO3-% MoO3)

Tg ((2 °C)

Tx ((2 °C)

100-0 95-05 90-10 85-15 80-20 75-25 70-30 65-35 60-40 55-45 50-50 45-55 40-60 35-65 30-70

290 296 306 318 336 353 374 395 409 414 408 401 387 369 352

325 403 434

580 531 468 476 475 443

typical crystallization event is observed for the samples with 45 e x e 70 with the relatively sharp exothermic peak shifting to lower temperatures as the concentration of MoO3 increases. Table 1 summarizes the values observed for the characteristic temperatures. Figure 2 strongly suggests that the structural transformation principles are different for glasses with high and low Mo content. Figures 3 and 4 show the FT-IR and the Raman spectra of this glass series. The data can be interpreted on the basis of previous publications on sodium phosphate53 and NaPO3WO3 glasses.48 For pure NaPO3 glass as well as glasses with low MoO3 content, the FTIR spectra display bands at 1255 cm-1 assigned to asymmetric stretching and at 1160 cm-1 (shoulder) assigned to symmetric stretching of the PO2- metaphosphate units. In addition, bands at 850 and 770 cm-1 are observed, which are assigned to asymmetric and symmetric stretching

Figure3. FT-IRspectraofglasssamplesinthesystem(NaPO3)100-x(MoO3)x.

modes of P-O-P linkages, respectively, and the O-P-O bending vibration is found at around 490 cm-1. With increasing MoO3 content the band observed at 1255 cm-1 is observed to broaden and to shift gradually toward 1200 cm-1; in conjunction with the NMR data, we attribute this observation to changes in the terminal P-O bond stretching vibrations of the PO2- group caused by the replacement of P-O-P by P-O-Mo linkages. For samples with 60 and 70 mol % MoO3, the absorption bands related to the PO2- and P-O-P stretching modes are no longer observed. In these samples, a new band that can be assigned to the symmetric stretching vibration of MoO3 octahedral

10112 J. Phys. Chem. B, Vol. 111, No. 34, 2007

Figure 4. Raman spectra of (NaPO3)100-x(MoO3)x glasses.

(Mo-O-Mo) appears around 890 cm-1.54 Little information can be extracted about the structural features of the molybdenum species. Figure 4 shows Raman spectra obtained for the glasses and crystalline reference compounds. The spectrum of glassy NaPO3 is dominated by two main bands at 1160 and 680 cm-1 assigned to the (PO2-) symmetric stretching vibration associated with the terminal oxygen species and the symmetric P-O-P bond

Santagneli et al. stretching vibration, respectively. With the MoO3 addition, a new band at 943 cm-1 with a shoulder at 911 cm-1 appears. As the MoO3 content increases, the relative intensities of the 1160 and 680 cm-1 bands decrease. The bands at 943 and 911 cm-1 can be most likely associated with terminal oxygen atoms (Mo)O or M-O-) bond vibrations associated with either four-, five-, or six-coordinate Mo atoms.55,56 At MoO3 contents exceeding x ) 40, the low-frequency component of this doublet diminishes gradually, while the high-frequency component shifts toward higher frequencies. This latter spectral feature persists up to the highest molybdenum oxide concentrations (x ) 70), suggesting that some terminal molybdenum-oxygen bonds are present within the entire glass-forming range. For glass compositions above x ) 45, a new broad band around 840 cm-1 appears, which gains dramatically in intensity and width at higher molybdenum oxide contents. On the basis of the Raman spectrum of pure molybdenum oxide, we assign this band to Mo-O single bond stretching vibrations within Mo-O-Mo bonded units. Figure 5 shows the 31P MAS NMR spectra. The spectrum of pure NaPO3 glass is dominated by a resonance near -20 ppm, which can be attributed to metaphosphate-type chain (Q(2)-type) units. Upon MoO3 addition, a new signal near -7.1 ppm appears, the intensity of which increases with increasing MoO3 content. As the MoO3 content increases beyond x ) 35, this signal broadens further due to the appearance of a new contribution near 0 ppm. The intensity of this component seems to increase, becoming the dominant component in the spectra obtained for the samples with x > 50. Further information concerning the connectivity patterns of the P atoms contributing to the resolvable MAS NMR signals are available from the J-resolved 2D 31P NMR spectra shown for representative samples in Figure 6. The 2D spectrum of the glass with x ) 20 (Figure 6a) shows three distinct phosphorus environments. The resonance near -20 ppm is correlated with a regular J-coupling triplet, consistent with the presence of either unaltered metaphosphate-type Q(2) sites or newly formed Q(3)1Mo sites. The resonance centered at -7.1 ppm is associated with a J-coupling doublet, consistent with the formation of a Q(2)1Mo

Figure 5. Solid state 31P MAS NMR spectra of (NaPO3)100-x(MoO3)x glasses. Spinning side bands are indicated by asterisks.

Structural Studies of NaPO3-MoO3 Glasses

J. Phys. Chem. B, Vol. 111, No. 34, 2007 10113

Figure 6. Two-dimensional J-resolved spectra of (NaPO3)100-x(MoO3)x glasses. J-coupling doublets resolved at specific chemical shifts are plotted beneath. (a) x ) 20; (b) x ) 40; (c) x ) 60.

10114 J. Phys. Chem. B, Vol. 111, No. 34, 2007

Santagneli et al. TABLE 3: 23Na NMR Spectral Parameters and Second Moments Derived from 31P{23Na} and 23Na{31P} REDOR Data of NaPO3-MoO3 Glasses x 0 10 20 30 40 60

Figure 7. Species concentrations derived from peak deconvolutions of the 31P MAS NMR spectra. Dashed curves are theoretical predictions based on the binary model.

TABLE 2: 31P NMR Spectral Deconvolutions and Chemical Shift of NaPO3-MoO3 Glasses site Q(2)2Mo x 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

δ (ppm) ( 0.5

3.3 0 -0.9 -2.0 -2.4 -2.8 -3.3 -3.3 -3.3

area (%) ( 2%

1 6 23 40 57 77 92 100 100

site Q(2)1Mo

site Q(2)0Mo

δ (ppm) ( 0.5

area (%) ( 2%

δ (ppm) ( 0.5

area (%) ( 2%

-7.1 -7.1 -7.4 -7.7 -7.8 -8.0 -8.9 -8.2 -8.7 -9.3 -10.5 -13.5

0 14 21 33 48 58 73 80 71 58 43 23 8

-20.6 -19.7 -19.8 -19.8 -19.8 -19.8 -19.8 -20.3 -19.2 -20.5

100 86 79 67 52 42 26 14 6 2

unit forming a single P-O-P- and a P-O-Mo linkage each. The third identified resonance centered near 0 ppm is associated with a singlet, consistent with the formation of a Q(2)2Mo unit. Both the Q(2)1Mo and Q(2)2Mo sites are also visible in the 2D spectrum of the glass with x ) 40 (Figure 6b). Figure 6c shows the 2D J-resolved spectrum of the sample with x ) 60. While this spectrum does not give any positive experimental evidence for indirect homonuclear 2J (31P-31P) coupling, we cannot exclude the possibility that small signal contributions of such doublet or triplet species are masked by the strong singlet resonance. Table 2 and Figure 7 show the results obtained from the deconvolution of the MAS spectra when the information from the two-dimensional measurements is taken into account as fitting constraints. The Q(2)0Mo component contribution is observed to decrease almost linearly with increasing x, disappearing for samples with x > 35. The contribution of the component Q(2)1Mo increases up to the composition x ) 35 and then decreases again at higher Mo contents. The contribution of the third broad peak assigned to Q(2)2Mo starts to appear for the sample with x ) 30 and continues to grow until it dominates the spectrum for samples with x > 60. Only one broad signal can be observed in this concentration range, making the approach of simple Gaussian peak fitting somewhat questionable. Table 2 illustrates further that at x ) 45 and beyond, the chemical shift attributed to the Q(2)1Mo component is quite a bit different from that measured at lower Mo contents. In addition,

δiso (ppm) ( 0.5

SOQE (MHz) ( 0.2

-3.4

2.1

-8.1

1.9

-14.3 -17.6

2.0 1.8

M2 (23Na{31P}) [106 (rad/s)2] ( 10% 3.8 3.7 3.1 2.7 1.9

the singlet initially observed at 3.3 ppm in the sample containing 30 mol % MoO3 has shifted to -2.5 ppm. All these observations suggest that the peak deconvolution scheme used for the low MoO3 samples may not be applicable here, and rather, new structures (other than Q(2)2Mo units) are being formed in this compositional domain. For example, the formation of local structures based on P(OMo)4 groups, as they are known for phosphorus molybdenum heteropolyacid salts,57,58 seems possible for x > 40. The clear detection of Mo-O-Mo bonding in the Raman spectra points toward the same conclusion. The formation of new types of structures is further supported by the 23Na NMR data. Table 3 indicates that the 23Na isotropic chemical shifts measured by TQMAS for these glasses are significantly more negative than those measured for glassy NaPO3. Further evidence for this conclusion comes from the 23Na{31P} REDOR results, which are summarized in Figure 8. As already observed for NaPO3-WO3 glasses, the M2 values characterizing the strength of the 23Na-31P dipole-dipole coupling decrease with increasing MoO3 content, suggesting that the contribution of phosphate to the second coordination sphere of Na diminishes with increasing x and the contribution of Mo becomes increasingly significant. Complementary insights into the glass structure can be obtained by 95Mo NMR. Figure 9 shows the spectra obtained on the glasses containing 30 and 50 mol % MoO3. Two main broad resonances are observed at 0 and at around -500 ppm. The large widths (which exceed the spinning frequency significantly) reflect wide distributions of chemical shift and nuclear electric quadrupolar coupling parameters as often observed in disordered systems. On the basis of chemical shift trends observed in the literature,59,60 we tentatively assign these resonances to MoO4 and MoO6 units, respectively. Discussion Based on the 31P MAS NMR peak assignments and the deconvolutions summarized in Table 2, simple scenarios describing the structural transformation arising from the addition of MoO3 to NaPO3 glass can be discussed. To facilitate this discussion, it is practical to consider two distinct regions within the glass forming range: the region of low to medium MoO3 contents (0 e x e 40) and the high-MoO3 region (x > 40). Low to Medium MoO3 Contents. In this concentration region, Raman spectroscopy detects no Mo-O-Mo linkages, indicating that Mo is exclusively participating in P-O-Mo bond formation. If it is assumed that the peak near -20 ppm represents exclusively Q(2)0Mo units, the relative signal areas lead to the conclusion that, for each MoO3 unit added to the glass, approximately two Q(2)0Mo units are transformed into Q(2)1Mo groups. This result suggests that the majority of the molybdenum species are four-coordinated units of the type Mo()O)2(O1/2)2, forming two bridging oxygen links to phosphorus, in rough agreement with the 95Mo NMR spectra. In principle, the process of P-O-Mo interlinking operative in this composition region

Structural Studies of NaPO3-MoO3 Glasses

J. Phys. Chem. B, Vol. 111, No. 34, 2007 10115

Figure 8. 23Na{31P} REDOR experiments obtained on representative (NaPO3)100-x(MoO3)x glasses. Solid curves are parabolic fits yielding the second moment values listed in Table 3.

Figure 9. Solid-state 95Mo MAS NMR spectra of representative (NaPO3)100-x(MoO3)x glasses.

could produce both Q(2)1Mo and Q(2)2Mo units. However, there is actually no evidence for the formation of Q(2)2Mo units for x < 30. This result suggests that a simple “binary” structural transformation model, in which the MoO3 component first converts all the Q(2)0Mo polyphosphate chains into Q(2)1Mo units before the latter are converted into Q(2)2Mo units at higher MoO3 contents, may provide an appropriate description. This model

predicts that the Q(2)0Mo units will disappear at x ) 33, while the fraction of Q(2)1Mo units is maximized at this composition. As x is further increased, the concentration of Q(2)1Mo should decrease, resulting in zero for x ) 50, whereas Q(2)2Mo should be the only observable component. Figure 7 contrasts the predictions from this binary model with the experimental data, documenting excellent agreement, particularly for samples with x < 30. This result implies a special stability of the Q(2)1Mo units, which prevents bond randomization and formation of Q(2)2Mo (and more Q(2)0Mo units) in this concentration range. Nevertheless, some of this randomization does seem to take place at higher MoO3 contents, leading to the observation of both Q(2)2Mo and Q(2)0Mo units at compositions near x ) 30 and 35. Despite the overall good agreement of the data with the binary model, it is clear that the latter can provide only an approximate description. First of all, the 95Mo NMR spectra observed for the x ) 30 sample suggest that significant amounts of highercoordinated Mo species are present, implying a conversion rate larger than two. Indeed a higher conversion rate can be rationalized, if the -20 ppm resonance contains a contribution from Q(3)1Mo units. The formation of such uncharged phosphate units implies a partial redirection of the sodium network modifier species toward the MoO3 component, producing some anionic molybdenum species. Support for such a process comes from the REDOR data, which reveal that the strength of the 23Na31P dipolar interactions diminishes significantly with increasing

10116 J. Phys. Chem. B, Vol. 111, No. 34, 2007 x, suggesting that the molybdenum species do become increasingly important in the second coordination sphere of the 23Na nuclei. High MoO3 Contents. As x increases beyond the value of 40, the experimental phosphorus speciations deviate significantly from the binary model prediction. Specifically, the concentration of the Q(2)1Mo units tends to be higher while the concentration of Q(2)2Mo units tends to be lower than predicted. At the same time, the Raman data indicate the formation of Mo-O-Mo bonding in this concentration range. As mentioned above, the applicability of the deconvolution model is uncertain in this concentration range, and different chemical shifts observed for the individual line shape components suggest the formation of other structures (possibly Q(3)2Mo, Q(3)3Mo and Q(4)4Mo [i.e., P(OMo)4 groups]) in this region. Since neither of these phosphate species is anionic, no particular proximity to the Na+ ions is expected for such structures. As a matter of fact, the low M2(31P-23Na) values measured by REDOR in this composition region and the substantially different 23Na chemical shifts found here are very well consistent with the idea that a significant part of the modifier in these glasses is used to create new anionic molybdenum species containing nonbridging oxygen atoms. Conclusions In summary, we have developed a comprehensive structural model for glasses in the system NaPO3-MoO3, based on complementary NMR and vibrational spectroscopic experiments. The results provide a structural rationale for the pronounced compositional dependence of Tg shown in Figure 2: addition of MoO3 to NaPO3 glass produces P-O-Mo connectivities, which have special stability. They initially insert between the different metaphosphate chains and, in addition, form crosslinks via Q(3)1Mo branching groups, ultimately transforming the one-dimensional chain structure into a three-dimensional network based on interlinked PO4 tetrahedra and MoO4/6 polyhedra. While the 31P MAS NMR data can be explained on the basis of a simple binary model describing the successive transformation of Q(2)0Mo into Q(2)1Mo and then into Q(2)2Mo units, there is evidence that this picture is only a rough approximation. Rather, 23Na and 95Mo NMR data suggest that some of the MoO 3 component introduced into the glass becomes anionic, resulting in an increased degree of polymerization of the phosphate species, producing Q(3)nMo units in addition. As a result, a remarkably stable glass structure is formed, whose Tg is greatly increased and which resists recrystallization under thermal analysis heating conditions. For samples beyond x ) 45, the glass transition temperatures decrease again, and the glasses become more prone to crystallization. These latter trends appear to be directly correlated with the appearance of isolated phosphate tetrahedral units connected to Mo as well as clusters involved in Mo-O-Mo bonding. Acknowledgment. This work was funded by the Deutsche Forschungsgemeinschaft and the joint PROBRAL programme sponsored by the Deutscher Akademischer Austauschdienst (DAAD) and CAPES. Personal stipends by the NRW Graduate School of Chemistry (to C.C.d.A.) and the Fonds der Chemischen Industrie (to W.S.) are most gratefully acknowledged. Brazilian agencies FAPESP and CNPq are also acknowledged. References and Notes (1) Al-Hawery, A. S. J. Phys. Chem. Solids. 1997, 58, 1325. (2) Rothermel, J. J. J. Am. Ceram. Soc. 1949, 32, 153.

Santagneli et al. (3) Mansingh, A.; Dhawan, A.; Tandon, R. P. J. Non-Cryst. Solids 1978, 27, 309. (4) Mansingh, A.; Tandon, R. P. Phys. ReV. B 1980, 21, 4829. (5) Yafan, Q.; Li, H. J. Non-Cryst. Solids 1986, 80, 527. (6) Boudlich, D.; Bih, L.; El Hassane Archidi, M.; Haddad, M.; Yacoui, A.; Nadiri, A.; Elouadi, B. J. Am. Ceram. Soc. 2002, 85, 623. (7) Chiodelli, G.; Magistris, A.; Villa, M.; Bjorkstam, J. J. Non-Cryst. Solids 1982, 51, 143. (8) Selvaraj, U.; Rao, K. J. J. Non-Cryst. Solids 1985, 72, 315. (9) Muthupari, S.; Prabakar, S.; Rao, K. J. J. Phys. Chem 1994, 98, 2646. (10) Linsey, A. S.; Owen, A. E.; Hayatee, F. M. J. Non-Cryst. Solids 1970, 4, 208. (11) Damodaran, K. V.; Rao, K. J. J. Am. Ceram. Soc. 1985, 72, 533. (12) Chowdari, B. V. R.; Tan, K. L.; Chia, W. T.; Gopalakrishna, R. Solid State Ionics 1990, 40/41, 684. (13) Patel, N. D.; Bridge, B. Phys. Chem. Glasses 1983, 24, 30; (14) Biha, L.; El Omaric, M.; Reau, J. M., Haddad, M.; Boudlich, D.; Yacoubi, A.; Nadiri, A. Solid State Ionics 2000, 132, 7185. (15) Morgan, S. H.; Magruder, R. H. J. Am. Ceram. Soc. 1990, 73, 753. (16) Magistris, A.; Chiodelli, G.; Duclot, M. Solid State Ionics 1983, 9, 61 (17) Muthupari, S.; Rao, K. J. Chem. Phys. Lett. 1994, 223, 133. (18) Poirier, G.; Messaddeq, Y.; Ribeiro, S. J. L.; Poulain, M. J. NonCryst. Solids 2005, 351, 293. (19) Poirier, G.; Jerez, V. A.; de Arau´jo, C. B.; Messaddeq, Y.; Ribeiro, S. J. L.; Poulain, M. J. Appl. Phys. 2003, 93, 1493. (20) Poirier, G.; de Arau´jo, C. B.; Messaddeq, Y.; Ribeiro S. J. L.; Poulain, M. J. Appl. Phys. 2002, 91, 10221. (21) Poirier, G.; Nalin, M.; Messaddeq, Y.; Ribeiro, S. J. L. Brazilian Pending Patent PI 0502711-0, 2005. (22) Poirier, G.; Messaddeq, Y.; Ribeiro, S. J. L.; Poulain, M. J. Solid State Chem. 2005, 178, 1533. (23) Poirier, G.; Michalowicz, A.; Messaddeq, Y.; Ribeiro, S. J. L. J. Non-Cryst. Solids 2005, 351, 3644. (24) Bih, L.; El Omari, M.; Reau, J. M.; Haddad, M.; Boudlich, D.; Yacoubi, A.; Nadiri, A. Solid State Ionics 2000, 132, 71. (25) Bih, L.; El Omari, M.; Reau, J. M.; Nadiri, A.; Yacoubi, A.; Haddad, M. Mater. Lett. 2001, 50, 308. (26) Mogus-Milankovic, A.; Santic, A.; Karabulut, M.; Day, D. E. J. Non-Cryst. Solids 2003, 330, 128. (27) Bih, L.; Omari, M. E.; Reau, J. M.; Haddad, M.; Boudlich, D.; Yacoubi, A.; Nadiri, A. Solid State Ionics 2000, 132, 71. (28) Muthupari, S.; Rao, K. J. J. Phys. Chem. Solids 1996, 57, 553. (29) Yu, X.; Day, D. E.; Long, G. J.; Brow, R. K. J. Non-Cryst. Solids 1997, 215, 21. (30) Brow, R. K.; Arens, C. M.; Yu, X.; Day, D. E. Phys. Chem. Glasses 1994, 35, 132. (31) Mogus-Milankovic, A.; Pivac, B.; Furic, K.; Day, D. E. Phys. Chem. Glasses 1997, 38, 74. (32) Mogus-Milankovic, A.; Rajic, M.; Drasner, A.; Trojko, R.; Day, D. E. Phys. Chem. Glasses 1998, 39, 70. (33) Selvaraj, U.; Rao, K. J. J. Non-Cryst. Solids 1985, 72, 315. (34) Parke, S.; Watson, A. I. Phys. Chem. Glasses 1969, 10, 37. (35) Wood, D. L.; Ferguson, J.; Knox, K.; Dillon, J. P. J. Chem. Phys. 1963, 39, 890. (36) Johnson, K. E.; Palmer, R. Spectrochim. Acta 1965, 21,1697. (37) Berretz, M.; Holt, S. L. J. Inorg. Nucl. Chem. 1074, 36, 49. (38) Sreedhar, B.; Sairam, M.; Chattopadhyay, D. K.; Kojima, K. Mater. Chem. Phys. 2005, 92, 492. (39) Deb, S. K.; Chopoorian, J. A. J. Appl. Phys. 1966, 37, 4818. (40) Siokou, A.; Leftheriotis, G.; Papaefthimiou, S.; Yanoulis, P Surf. Sci. 2001, 482-485, 294. (41) Massiot, D.; Fayon, F.; Capron, M.; King, I.; Le Calve, S.; Alonso, B.; Durand, J. O.; Bujoli, B.; Gan, Z.; Hoatson, G. Magn. Reson. Chem. 2002, 40, 70. (42) Medek, A.; Harwood, J. S.; Frydman, L. J. Am. Chem. Soc. 1995, 117, 12779. (43) Amoureux, J. P.; Fernandez, C.; Steuernagel, S. J. Magn. Reson. A 1996, 123, 116. (44) Zu¨chner, L.; Chan, J. C. C.; Mu¨ller-Warmuth, W.; Eckert, H. J. Phys. Chem. 1998, 112, 4495. (45) Brown, S. P.; Pe´rez-Torralba, M.; Sanz, D.; Claramunt, R. M.; Emsley, L. Chem. Commun. 2002, 1852. (46) Duma, L.; Lai, W. C.; Carravetta, M.; Emsley, L.; Brown, S. P.; Levitt, M. H. Chem. Phys. Chem. 2004, 5, 815. (47) Fayon, F.; King, I. J.; Harris, R. K.; Evans, J. S. O.; Massiot, D. C. R. Chim. 2004, 7, 351. (48) Araujo, C.; Zhang, L.; Eckert, H.; Poirier, G.; Ribeiro, S. J. L.; Messaddeq, Y. J. Mater. Chem. 2006, 16, 3277. (49) States, D. J.; Haberkorn, R. A.; Ruben, D. J. J. Magn. Reson. 1982, 48, 286. (50) Gullion, T.; Schaefer, S. J. Magn. Reson. 1989, 81, 196

Structural Studies of NaPO3-MoO3 Glasses (51) Chan, J. C. C.; Eckert, H. J. Magn. Reson. 2000, 147, 170. (52) Gullion, T. Magn. Reson. ReV. 1997, 17, 83. (53) Hudgens, J. J.; Brow, R. K.; Tallant, D. R.; Martin, S. W. J. NonCryst. Solids 1998, 223, 21. (54) Dong, W.; Dunn, B. J. Mater. Chem. 1998, 8, 65. (55) Du, X.; Dong, L.; Li, C.; Liang, Y.; Chen, Y. Langmuir 1999, 15, 1693.

J. Phys. Chem. B, Vol. 111, No. 34, 2007 10117 (56) Maass, J.; Ahrens, H.; Frijbel, P.; Barner, K. Solid State Commun. 1993, 87, 567. (57) Gatehouse, B. M.; Jozsa, A. J. Acta Crystallogr. 1993, C39, 658. (58) Kierkegaard, P.; Asbrink, S. Acta Chem. Scand. 1964, 18, 2329. (59) Edwards, J. C.; Adams, R. D.; Ellis, P. D. J. Am. Chem. Soc. 1990, 112, 8349. (60) Machida, N.; Eckert, H. Solid State Ionics 1998, 107, 255.