Site Discrimination in Mixed-Alkali Glasses Studied by Cross

Cation−cation interactions are thought to play a significant role in shaping the nonlinear compositional dependence of ionic conductivity, known as ...
0 downloads 0 Views 259KB Size
Download PDF
J. Phys. Chem. B 2006, 110, 14253-14261

14253

Site Discrimination in Mixed-Alkali Glasses Studied by Cross-Polarization NMR Stefan Peter Puls and Hellmut Eckert* Institut fu¨r Physikalische Chemie, Westfa¨lische Wilhelms-UniVersita¨t Mu¨nster, Corrensstrasse 30, D-48149 Mu¨nster, Germany ReceiVed: April 11, 2006; In Final Form: May 15, 2006

Cation-cation interactions are thought to play a significant role in shaping the nonlinear compositional dependence of ionic conductivity, known as the mixed-alkali effect (MAE) in glassy solid electrolytes. For providing a structural rationale of this effect, the discrimination of various cation sites in mixed-alkali glasses is of interest. In the present study, cross-polarization (CP) experiments have been applied to glasses in the system [(Li2O)x(Na2O)1-x]0.3[B2O3]0.7 to discriminate between alkali ions by virtue of different heteronuclear 7 Li-23Na dipole-dipole coupling strengths. Cross-polarization studies involving two types of quadrupolar nuclei (both 7Li and 23Na have a spin-quantum number I ) 3/2) are complicated by spin state mixing under radio frequency irradiation and magic-angle spinning (MAS). Therefore careful validation and optimization protocols are reported for the model compound LiNaSO4 prior to conducting the measurements on the glassy samples. 23Na f 7Li CP/MAS NMR spectra have been obtained on glasses containing the Na+ ions as the dilute species. They reveal that those lithium species interacting particularly strongly with sodium ions have the same average 7Li chemical shift as the entire lithium population; the symmetrical situation applies to the 23 Na nuclei at the sodium rich end of the composition range. On the other hand, a clear site discrimination is afforded by temperature-dependent static 23Na f 7Li CP experiments, indicating that the Li ions that are most strongly interacting with sodium ions are strongly immobilized. This finding provides the first direct experimental evidence for the proposed secondary mismatch concept invoked for explaining the strong MAE in the dilute foreign ion limit.

Introduction The structural properties of the mobile ions and their spatial distribution in ion conducting glassy electrolytes have attracted considerable interest during the last few years. Detailed knowledge of the cation environment in these glasses may help to understand the microscopic mechanisms of ionic transport. Among these materials, glasses containing two different types of cations have gained considerable interest because of their unusual properties concerning the ionic mobility, namely, the so-called mixed-alkali effect (MAE). In these glasses, a dramatic decrease of ionic conductivity is observed, if one type of the mobile cation is replaced by its homologue at a constant overall cation content.1-4 One of the most prominent interpretations in this context is the dynamic structure model proposed by Bunde et al.,5-7 attributing the MAE to site mismatch between the two unlike cations. On the basis of EXAFS results the mobile ions are thought to possess their own distinct environments in the glass structure leading to preferred diffusion pathways for each type of ion. As a consequence, the path connectivities, which depend only on the relative cation content, determine the mobility of both cation types and their respective contributions to ionic conductivity (see top of Figure 1). This model is supported by molecular dynamics simulations8-11 as well as tracer diffusion experiments.12,13 However, some questions in the context of the mixed-alkali effect are still not fully resolved. In particular, the idea of individual, noninterfering pathways for the cations cannot explain the very strong immobilization effects especially in the dilute foreign ion substitution limit.14,15 In addition to these experiments, trends of the 23Na chemical shift in mixed-alkali silicate16-18 and borate glasses19-21 clearly * Corresponding author. E-mail: [email protected].

Figure 1. Top: schematic illustrating the primary mismatch concept for explaining the mixed-alkali effect. Arrows indicate path connectivity. Bottom: schematic illustrating the secondary mismatch concept for explaining cation immobilization via cation-cation interactions in the dilute foreign ion substitution limit. Arrows indicate path connectivity between A and A′ sites.

show that the local environment of the sodium ions changes continuously upon substitution of sodium with a homologue, a finding that also supports the idea of site modification. Following the experimental results of Xue and Stebbins22 and the theoretical calculations by Tossell,23 these chemical shift trends suggest

10.1021/jp062251l CCC: $33.50 © 2006 American Chemical Society Published on Web 06/30/2006

14254 J. Phys. Chem. B, Vol. 110, No. 29, 2006

Puls and Eckert

that the sodium sites shrink upon substitution with potassium (i.e., a larger cation), whereas they expand upon substitution with lithium (i.e., a smaller cation). Inspection of a large body of chemical shift trends in various mixed-alkali glasses16,17,20,24,25 suggests that these structural readjustments are rather universal: substitution of an alkali ion A by a larger homologue B always leads to a slight compression of the A site and an expansion of the B site, as compared to the situation in the corresponding single-alkali glass. On the basis of this evidence for site modification, Ratai et al.21 suggested an additional (secondary) mismatch effect as a fundamental principle underlying the MAE, which seems particularly relevant in the dilute foreign ion limit: each foreign ion B acts not only by interrupting transfer paths between the well-matched A sites by mere occupancy but also by distorting regular A sites nearby into A′ sites via specific cation-cation interactions. As a result, each B ion has the effect of immobilizing multiple A ions in its immediate vicinity (see Figure 1, bottom). So far, however, no direct evidence for this secondary mismatch effect has been published. To this end, one needs to be able to detect those ions in the direct vicinity of a foreign ion in a selective way and show that they are different from the remainder of the ions constituting the majority species. In the present study, we address this task by using the technique of cross-polarization, which utilizes the magnetic dipole-dipole interaction mechanism: magnetization originating from the nuclei belonging to the dilute ion species can be transferred selectively to ions of the other species that are located in their immediate vicinity, while further remote nuclei go undetected. The present study applies this approach to a series of lithium sodium borate glasses of composition [(Li2O)x(Na2O)1-x]0.3[B2O3]0.7. Using 23Na f 7Li and 7Li f 23Na cross-polarization, we will directly investigate the local environments of those cations that are affected by this site perturbation and probe their specific dynamics. As this particular application of CP/MAS, involving two quadrupolar spin species, has never been reported before, the methodological aspects will be developed and discussed on a crystalline model compound before the approach is transferred to the mixed-alkali glass system of the present study. Concepts and Methodology. The principal technique used in this study for selectively detecting Li nuclei near Na ions and vice versa is cross-polarization (CP) with and without magic-angle spinning (MAS). The standard pulse sequence is shown in Figure 2, top and middle. This method relies on magnetization transfer between spin-locked nuclear spin systems based on a difference in spin temperature, facilitated by magnetic dipole-dipole interactions. To this end, the precession frequencies of the two spin species must be identical in the rotating frame, which can be accomplished by adjusting the spin-lock amplitudes accordingly (Hartmann-Hahn matching condition). As discussed previously,26-28 for weakly dipolar coupled spin systems under MAS, various Hartmann-Hahn matching conditions can be realized, given by

νeff(I) - νeff(S) ) nνr

(1)

νeff(I) + νeff(S) ) nνr

(2)

Figure 2. Pulse sequences employed in this paper. Top: spin-lock pulse sequence. Following a 90° preparation pulse, the magnetization is spin-locked in the rotating frame, by application of a radio frequency field that oscillates in phase with the transverse magnetization. During the detection period t2, the signal amplitude is measured as a function of irradiation length. Middle: CP, CP/MAS pulse sequence: while the nuclei of type S are spin-locked as described in panel a, a contact pulse is applied to the nuclei of type I such that the Hartmann-Hahn matching condition is fulfilled. This generates transverse magnetization associated with the I nuclei, which is then collected during the detection period t2. Bottom: z-filtered TQ/MAS pulse sequence. For a detailed description, see text.

of quadrupolar nuclei, the effective precession (nutation) frequencies can adopt values between

νeff(I) ) ν1I and (I + 1/2)ν1I

(3a)

νeff(S) ) ν1S and (S + 1/2)ν1S

(3b)

where ν1I ) γIB1I and ν1S ) γSB1S are the liquid-state nutation frequencies of the nuclei with the spin quantum numbers I and S, respectively. The first Hartmann-Hahn condition (eq 1) represents polarization transfer between the two spins I and S via an energy conserving flip-flop process (zero-quantum CP), whereas the second condition (eq 2) corresponds to a polarization transfer associated with a nonenergy conserving flip-flip or flop-flop process, where the energy is balanced by the mechanical rotation of the sample (double-quantum CP). If CP/MAS experiments are conducted on two quadrupolar nuclear spin species (as in the present case), their spin-locking behavior poses severe complications. As shown by Vega,29,30 rf-irradiation during the contact time under MAS conditions induces spin state mixing, which seriously interferes with the ability to spin-lock transverse magnetization. Following selective excitation of the central |1/2〉 T |-1/2〉 coherence, the spinlock efficiency strongly depends on the adiabicity parameter

R)

νeff2 νQνr

(4)

or

Here, νeff(I) and νeff(S) represent the effective precession frequencies of spins I and S (which are determined by the chosen radio frequency amplitudes), νr denotes the MAS spinning frequency, and n can take integer values of 1 or 2. In the case

where νQ is the quadrupolar frequency and νr is the MAS spinning frequency. Effective spin-locking is possible either in the limit R 〉〉 1 (high amplitude of the spin-lock field, adiabatic regime) or R 〈〈 1 (low amplitude, sudden regime). Thus, the challenge of such CP/MAS experiments consists of finding a Hartmann-Hahn matching condition (1 or 2) with rf-amplitudes that enable both nuclear species to be effectively

Site Discrimination in Mixed-Alkali Glasses

J. Phys. Chem. B, Vol. 110, No. 29, 2006 14255

TABLE 1: Sample Compositions, Glass Transition Temperatures, and NMR Results of the Glasses under Study

a M1 M2 M3 M4 M5 b a

x

Tg (K) ((2 K)

δiso(23Na) (ppm) ((0.30 ppm)

SOQE(23Na) (MHz) ((0.05 MHz)

0 0.05 0.10 0.50 0.90 0.95 1

760 742 737 720 756 763 771

-3.0 -3.1 -3.2 -4.3 -5.1 -5.3

2.4 2.40 2.35 2.30 2.30 2.20

Data are taken from Epping et al.16

b

δ(7Li) (ppm) ((0.05 ppm)

CQ(7Li) (MHz) (( 0.02 MHz)

0.18 0.17 0.07 -0.03 -0.04 -0.06

0.30 0.30 0.32 0.32 0.32 0.33

Data are taken from Ratai et al.18

spin-locked simultaneously. It is therefore essential to characterize the spin-lock behavior of these nuclei prior to recording CP/ MAS NMR spectra. In general, previous CP/MAS experiments involving two quadrupolar nuclei31,32 have turned out to be most efficiently performed in the sudden regime under low power level conditions. In particular, the heteronuclear double-quantum CP33,34 has been shown to result in very efficient magnetization transfer between two types of quadrupolar nuclei. To optimize the spin-lock behavior of 7Li and 23Na, detailed knowledge of the average strength of the nuclear electric quadrupolar coupling is beneficial. In the case of 23Na NMR, two-dimensional triple-quantum NMR (TQ/MAS NMR) can be used.35,36 The bottom panel of Figure 2 shows the pulse sequence employed. The technique exploits the fact that the forbidden |3/2〉 T |-3/2〉 triple-quantum coherence has no intrinsic anisotropy caused by first-order quadrupolar coupling. After excitation of the triple-quantum coherence with the first pulse, it can evolve for the duration of the evolution time t1. Subsequently, a second pulse stops this evolution. The third pulse, usually applied at a low rf-field amplitude, finally produces single-quantum coherences associated with observable magnetization. Incrementing the evolution time t1 followed by 2-D data processing allows a correlation between the standard MAS NMR spectrum (affected by second-order quadrupolar broadening) and a spectrum governed by isotropic quadrupolar and chemical shifts. In this paper, we use TQ/MAS NMR to measure the average 23Na isotropic chemical shifts and nuclear electric quadrupolar coupling parameters. In contrast, the 7Li quadrupolar coupling parameters have been estimated from the span of the MAS spinning sideband pattern generated by the |1/2〉 T |3/2〉 and |-1/2〉 T |-3/2〉 satellite coherences37-39 (satellite transition spectroscopy, SATRAS). Experimental Procedures Sample Preparation and Characterization. Homogeneous and transparent glasses in the system [(Li2O)x(Na2O)1-x]0.3[B2O3]0.7 (x ) 0.05, 0.10, 0.50, 0.90, 0.95) were prepared from Li2CO3 and Na2CO3 (MERCK, 99.999%), as well as B2O3 (MERCK, 99.995%). Finely ground mixtures of the starting materials (10 g batches) were melted in a Pt crucible up to 980 °C and quenched to room temperature by removing the crucibles from the furnace and immersing them in liquid nitrogen. Table 1 summarizes the compositions and glass transition temperatures measured on a NETZSCH DSC 204 thermal analyzer, using a heating rate of 10 K/min. None of the glasses showed any indication of phase separation. NMR Spectroscopy. Standard 7Li and 23Na nuclear magnetic resonance experiments were carried out at 194.34 and 132.29 MHz, respectively, using a Bruker DSX 500 spectrometer at magic-angle spinning frequencies between 10 and 15 kHz. Chemical shifts are reported relative to 1 M aqueous solutions of LiCl and NaCl.

Figure 3. Representative 7Li-SATRAS spectra recorded at 10 kHz spinning frequency. Top: crystalline LiNaSO4. Bottom: glass of composition [(Li2O)0.9(Na2O)0.1)]0.3(B2O3)0.7. 7Li-SATRAS spectra were obtained under the following conditions: spinning speed 10 kHz; pulse length 1.5 µs; and recycle delay 15 s. The rf-field strength of the excitation pulse corresponded to a nutation frequency of 150 kHz in aqueous LiCl solution. All the 23Na TQ/MAS spectra were obtained at a spinning rate of 12 kHz applying the z-filtering sequence.39 The rf-field strengths of the first two hard pulses and the third soft pulse corresponded to nutation frequencies of approximately 120 and 12 kHz, respectively, for liquid NaCl(aq). The optimized pulse widths were determined to be 4.5, 2.1, and 10 µs for the three consecutive pulses. A 24-step phase cycle suppressed all unwanted coherences and selected only the TQ coherences. Typically, 960 transients were accumulated for each t1 incre-

14256 J. Phys. Chem. B, Vol. 110, No. 29, 2006

Puls and Eckert

TABLE 2: Hartmann-Hahn Matching Conditions Observed for CP/MAS Experiments on LiNaSO4a H-H matches in Figure 6

υr (kHz)

υeff, Na (kHz)

υeff, Li (kHz)

RNa

RLi

H-H condition

a b c d a′ b′ c′ d′ Ab Bb Cb Db

10 10 10 10 14 14 14 14 10 10 10 10

20.8 18.0 11.4 9.6 30.5 25.8 16.2 11.7 20.1 17.5 11.2 9.4

1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8

0.0733 0.0549 0.0220 0.0156 0.1126 0.0806 0.0318 0.0166 0.0337 0.0255 0.0105 0.0074

0.0041 0.0041 0.0041 0.0041 0.0029 0.0029 0.0029 0.0029 0.0020 0.0020 0.0020 0.0020

υeff,Na - υeff,Li ) 2υr υeff,Na + υeff,Li ) 2υr υeff,Na - υeff,Li ) 1υr υeff,Na + υeff,Li ) 1υr υeff,Na - υeff,Li ) 2υr υeff,Na + υeff,Li ) 2υr υeff,Na - υeff,Li ) 1υr υeff,Na + υeff,Li ) 1υr υeff,Na - υeff,Li ) 2υr υeff,Na + υeff,Li ) 2υr υeff,Na - υeff,Li ) 1υr υeff,Na + υeff,Li ) 1υr

See also Figure 7. b A-D: Hartmann-Hahn matching conditions observed for a glass with composition (Li2O)0.5(Na2O)0.5))0.3(B2O3)0.7: matching spectra are not shown in Figure 7). a

Figure 4. Compositional dependence of chemical shifts in mixedalkali borate glasses of composition [(Li2O)x(Na2O)1-x]0.3[B2O3]0.7. Top: 23Na chemical shift as a function of relative lithium content. Bottom: 7Li chemical shift as a function of relative lithium content.

ment, and a total of 128 increments was done at steps of 5.21 µs. A relaxation delay of 1 s was employed. Quadrature detection in the F1 dimension was achieved by the hypercomplex approach. The sheared spectra were analyzed by projecting each individual site onto the F1 and F2 axes. Chemical shifts and quadrupolar shifts were obtained according to published procedures.35 7Li and 23Na spin-lock experiments as well as the 23Na f 7Li and 7Li f 23Na CP/MAS experiments were measured on a Bruker DSX 500 spectrometer, equipped with a 4 mm Bruker

Figure 5. 7Li NMR spin-locking experiments conducted on LiNaSO4. Top: intensity of the spin-locked 7Li MAS NMR signal as a function of the rf-field amplitude at a constant spin-lock time (10 ms); νr ) 10 kHz. The maxima in spin-lock intensity a and b correspond to effective nutation frequencies of νeff(7Li) ) 109 and 1.8 kHz, respectively. Bottom: intensity of spin-locked 7Li magnetization as a function of spin-lock time at two different rf-field amplitudes. Closed circles: νeff(7Li) ) 1.8 kHz and closed squares: νeff(7Li) ) 109 kHz; νr ) 10 kHz.

double resonance probe. Typical initial 90° pulse lengths were 3 and 3.5 µs for 7Li and 23Na, respectively, the rf-field amplitudes of the contact pulses in these experiments corresponded to nutation frequencies between 1.8 and 109 kHz, and contact times between 4 and 10 ms were used. Typically, 256 FIDs (in LiNaSO4) and 2000-4000 FIDs (in glasses) were accumulated at a relaxation delay of 1 s for 23Na f 7Li CP spectra and 10s for 7Li f 23Na CP spectra. More experimental details are summarized in Table 2. Control experiments (with

Site Discrimination in Mixed-Alkali Glasses

Figure 6. 23Na NMR spin-locking experiments conducted on LiNaSO4. Top: intensity of the spin-locked 23Na MAS NMR signal as a function of the rf-field amplitude at constant spin-lock time (10 ms); νr ) 10 kHz. The maximum in spin-lock intensity a corresponds to an effective nutation frequency of νeff(23Na) ) 4.2 kHz. Bottom: intensity of spinlocked 23Na magnetization as a function of spin-lock time; νeff(23Na) ) 4.2 kHz and νr ) 10 kHz.

the rf-power of the spin-lock pulse applied to the source nuclei set to zero) resulted in zero signal intensity for the detected nuclei under these conditions, indicating that the CP signals truly originate from heteronuclear cross-relaxation. Results and Data Interpretation Single Resonance and TQ/MAS Experiments on Glassy Samples. Table 1 summarizes the compositional dependences of the 23Na and 7Li chemical shift values and quadrupolar coupling parameters extracted from 23Na TQ/MAS and 7Li SATRAS measurements. In addition, Figure 4 shows the compositional trends of the 7Li and 23Na chemical shift values. Obviously, the 23Na chemical shift decreases with increasing lithium content, while the 7Li resonance shifts to higher frequencies with increasing sodium content. These findings are in line with those obtained by Ratai et al.21 for a Li-Na borate glass series with the same total composition but with somewhat different relative mixed-alkali compositions. However, for the 23Na chemical shifts, there is a systematic offset of 3.5 ppm between both studies. We suspect that this discrepancy arises from a referencing error made in the previous study. As previously discussed, the chemical shift trends observed in Figure 4 reflect subtle changes in the size of the alkali sites as compared to the situation in the single-alkali endmembers, forming the basis of the secondary mismatch hypothesis.21 MAS Spin-Locking and Cross-Polarization CP/MAS Experiments on LiNaSO4. Because of the previously men-

J. Phys. Chem. B, Vol. 110, No. 29, 2006 14257

Figure 7. 23Na f 7Li and 7Li f 23Na CP/MAS experiments on LiNaSO4 establishing Hartmann-Hahn matching conditions. Top: intensity of 23Na f 7Li CP/MAS spectra as a function of rf-field amplitude at a constant contact time of 5 ms, the minima and maxima are Hartmann-Hahn matches (for details, see Table 2); νr ) 10 kHz. Middle: intensity of 23Na f 7Li CP/MAS spectra as a function of rffield amplitude at constant contact time of 5 ms, the minima and maxima are Hartmann-Hahn matches (for details, see Table 2); νr ) 14 kHz. Bottom: intensity of 7Li f 23Na CP/MAS spectra as a function of rf-field amplitude at constant contact time of 5 ms, the minima and maxima are Hartmann-Hahn matches (for details, see Table 2); νr ) 10 kHz.

tioned complications for cross-polarization experiments involving two species of quadrupolar nuclei, a crystalline model compound (LiNaSO4) has been investigated prior to any experiments on glasses. To find optimum experimental conditions, the spin-lock behavior of 7Li and 23Na has been studied in detail. Figure 5 shows the dependence of the spin-locked 7Li magnetization as a function of rf-field amplitude. Obviously, efficient spin-locking is possible both in the adiabatic as well as in the sudden limit, as expected for 7Li with its rather weak quadrupolar interaction (CQ ≈ 0.16 MHz) in LiNaSO4 (Figure 3). The maxima in intensity a and b appear at rf-amplitudes corresponding to effective nutation frequencies of νeff(Li) ) 109.0 kHz (adiabatic regime) and 1.8 kHz (sudden regime), respectively. Figure 5, bottom, shows that at these rf-amplitudes, spin-locking is possible for sufficiently long times required for efficient cross-polarization (t > 1 ms). For 23Na with its relatively strong quadrupolar interaction (CQ ) 1.2 MHz), a quite different behavior is observed. Spin locking for sufficiently long times can only be achieved in the sudden regime. Figure 6 shows that at a rotor frequency of 10 kHz maximum spin lock efficiency is observed at an rf-amplitude corresponding to νeff(Na) ) 4.2 kHz.

14258 J. Phys. Chem. B, Vol. 110, No. 29, 2006

Puls and Eckert

Figure 8. 23Na f 7Li and 7Li f 23Na CP/MAS experiments on mixed-alkali borate glasses. All experiments were carried out at condition D in Table 2. Top left: (x ) 0.95) 7Li single-pulse spectrum (upper trace) and 23Na f 7Li CP/MAS spectra at 10 and 4 ms contact time (lower traces). Top right: (x ) 0.95) 7Li single-pulse spectrum (upper trace) and 23Na f 7Li CP/MAS spectrum at 200 K and 4 ms contact time (lower trace). Middle left: (x ) 0.90) 7Li single-pulse spectrum (upper trace) and 23Na f 7Li CP/MAS spectra at 10 and 4 ms contact time (lower traces). Middle right: (x ) 0.10) 23Na single-pulse spectrum (upper trace) and 7Li f 23Na CP/MAS spectra at 10 and 4 ms contact time (lower traces). Bottom left: (x ) 0.05) 23Na single-pulse spectrum (upper trace) and 7Li f 23Na CP/MAS spectra at 10 and 4 ms contact time (lower traces). Bottom right: (x ) 0.05) 23Na single-pulse spectrum (upper trace) and 7Li f 23Na CP/MAS spectrum at 200 K and 4 ms contact time (lower trace).

In view of the 23Na and 7Li spin-lock behavior observed, it is obviously most promising to focus on cross-polarization in the low-power limit, where effective spin-locking is possible and selective excitation conditions are met for both the nuclear species involved. Figure 7 presents the results of CP/MAS experiments conducted at 10 kHz (top) and 14 kHz (middle) spinning frequencies to detect the match conditions, showing the dependence of the CP signal on the amplitude of the CP-

pulse on 23Na, while the amplitude for the 7Li channel was kept constant at a value corresponding to νeff(Li) ) 1.8 kHz. According to eqs 1 and 2, zero-quantum and double-quantum Hartmann-Hahn matches could be identified, as both positive and negative signals occur in the spectra. On the basis of the 23Na effective nutation frequencies, all Hartmann-Hahn matches can be assigned to either zero- or double-quantum matches. Figure 7, bottom shows the corresponding result of the 7Li f

Site Discrimination in Mixed-Alkali Glasses

J. Phys. Chem. B, Vol. 110, No. 29, 2006 14259

Figure 9. Static 7Li NMR (left) and 23Na f 7Li CP spectra (right) for glass with composition ((Li2O)0.9(Na2O)0.1))0.3(B2O3)0.7, measured at three different temperatures. 23Na

CP/MAS experiments at 10 kHz spinning frequency. Table 2 summarizes all Hartmann-Hahn matching conditions identified, along with the corresponding experimental conditions. Obviously, the observed results represent a tradeoff between the different spin-lock requirements for 7Li and 23Na, which are characterized by quadrupolar interactions of different orders of magnitude. Similar to CP/MAS experiments previously done on 11B and 27Al,31-33 the only viable Hartmann-Hahn conditions corresponding to the sudden regime involve very low power levels for both nuclei. Meier26 showed that under magicangle spinning, the Hartmann-Hahn matches for the heteronuclear zero- or double-quantum CP may partially overlap and have different signs. As a consequence, the Hartmann-Hahn spinning sideband profiles may significantly deviate from that theoretically predicted for pure zero- or double-quantum CP. Exactly this behavior is observed for LiNaSO4 as can be seen from the patterns in Figures 7. In particular, the strong dependence of the match conditions on the MAS spinning frequency again underlines the necessity for optimizing the CP experiment on a model compound prior to conducting it on glassy samples. In the latter case, a distribution of quadrupolar coupling constants makes it even more difficult to find the optimum conditions. CP and CP/MAS Experiments on Glassy Samples. On the basis of the results on LiNaSO4, the Hartmann-Hahn conditions were reoptimized on a glass with composition [(Li2O)0.5(Na2O)0.5]0.3[B2O3]0.7. As the quadrupolar coupling parameters of the glasses are rather similar to those in LiNaSO4, only small differences to the model compound were observed (see also bottom of Table 2). Figure 8 represents all CP/MAS spectra for the glass series with composition [(Li2O)x(Na2O)1-x]0.3[B2O3]0.7. The contact times were varied between 4 and 10 ms. A comparison of 7Li single resonance spectra with 23Na f 7Li CP/MAS spectra is shown for lithium rich glasses (x ) 0.95 and 0.9), while for sodium rich glasses (x ) 0.1 and 0.05), a comparison of 23Na single resonance spectra with 7Li f 23Na CP/MAS spectra is depicted. As discussed in more detail next, no obvious differences between single-pulse spectra (representative of the entire population of the majority cation species) and CP/MAS spectra

(representative only of those cations of the majority species, which are in direct proximity to a foreign cation) are visible. This behavior was found to be the same for all contact times used; thus, there is no evidence for any spin-diffusion effects interfering with site discrimination. To exclude the potential interference of cation dynamics, experiments at 200 K were carried out for glasses containing 5 and 95 mol % lithium oxide. No differences to corresponding spectra obtained at room temperature could be observed. From all these experiments, it appears that the local environments cannot be discriminated on the basis of their chemical shifts. Figure 9 shows the temperature-dependent static 7Li singlepulse spectra as compared to corresponding static 23Na f 7Li CP spectra for glass of composition ((Li2O)0.9(Na2O)0.1))0.3(B2O3)0.7. It can be seen on the left side of Figure 9 that the 7Li single resonance spectra are clearly motionally narrowed at temperatures above room temperature. At 393 K, the value of the static line-width is smaller by a factor of 0.6 as compared to the value at 200 and 293 K. The right side of Figure 9 shows the corresponding static 23Na f 7Li CP spectra. Note that they are much broader than the single resonance spectra, suggesting that lithium ions in direct proximity to sodium ions are significantly different. From the shape of the CP spectra, it seems that still some of those lithium ions detected in the singlepulse spectra make a small contribution to the signal; therefore, all CP spectra were deconvoluted in two components: a component taken from a fit of the single pulse (SP) spectra corresponding to the lithium majority species and a second, broad Gaussian component reflecting only those lithium ions in the immediate neighborhood of sodium ions. Using both spectral contributions, it is then possible to fit the single-pulse spectra, allowing the fraction of immobilized Li ions to be estimated at approximately 10-15%. All the fitting parameters are summarized in Table 3. The width of the broad component turns out to be temperature-independent, whereas the sharper component narrows significantly as the temperature is increased. This finding supports the assumption that the broad static component dominating the CP spectra reflects a species that is significantly immobilized in comparison to the average Li species detected via single-pulse NMR.

14260 J. Phys. Chem. B, Vol. 110, No. 29, 2006

Puls and Eckert

TABLE 3: Fitting Parameters (chemical shift of peak maximum, δ, full width at half maximum, fwhm, and fraction of Gaussian vs. Lorentzian character, G) for Static Single Resonance and CP Spectra Obtained for the Glass with Composition ((Li2O)0.9(Na2O)0.1))0.3(B2O3)0.7 narrow component

broad component

T (K)

δ (ppm) ((0.5 ppm)

fwhm (ppm) ((5 ppm)

G

200

0.3

48

0.58

293

0.3

48

0.57

393

0.3

32

0.64

area (%) ((1%)

δ (ppm) ((0.5 ppm)

fwhm (ppm) ((5 ppm)

G

19 (CP) 85 (SP) 20 (CP) 88 (SP) 30 (CP) 89 (SP)

-0.1

117

1

-0.1

116

1

-0.1

115

1

Discussion and Conclusion As stated in the Introduction, two fundamental mechanisms of site mismatch are thought to be responsible for the immobilization of the mobile ions in mixed-alkali glasses:21 a primary site mismatch between the two unlike cations leading to so-called preferred diffusion paths for both cation species and a secondary site mismatch resulting from cation-cation interactions. Even though this secondary mismatch has been postulated on the basis of NMR chemical shift trends,16-18,21 the latter do not provide any direct evidence. Indeed, they reflect only average properties over the entire cation population for each sample. In this respect, the present study provides some fundamentally new insight, as the cross-polarization experiment allows the selective detection of only those ions that are in the immediate vicinity of a foreign ion. Thus, in the present study, magnetization is transferred selectively from the dilute foreign cation species only to those cations of the majority species, which are in the immediate vicinity (A′ sites) of a foreign B cation. A comparison with the standard single-pulse spectra (representative for all A and A′ sites) allows us then to address the question as to whether the A and A′ sites are any different: as illustrated by Figure 8, the A and A′ sites turn out to have, within the limits of experimental precision, identical chemical shifts at a given glass composition. Considering the fact that the overall effect shown in Figure 4 is rather small to begin with, we conclude that the chemical shift effect caused by the proximity of one foreign ion is too small to be resolved. Apparently, the isotropic chemical shift distribution in these glasses is dominated by other sources of dispersion (distribution of interatomic distances, bond angles, and coordination numbers), which are unrelated to cation-cation proximities. However, from the comparison of temperature-dependent static single resonance spectra with static 23Na f 7Li CP spectra, a rather different picture emerges. The possibility of deconvoluting the static CP spectra into two components, namely, a component that is definitely motionally narrowed and a broader component with a nearly temperature-independent width, strongly supports the idea that the mobility of lithium sites in the immediate proximity to sodium ions is significantly reduced as compared to the remainder of the lithium population. This fact provides rather strong direct evidence for the secondary mismatch concept. Thus, the immobilization effect observed for the lithium ions in the immediate vicinity of sodium can account for the strong MAE observed particularly in the dilute foreign ion regime. In summary, we have been able to demonstrate that heteronuclear cross-polarization spectroscopy between two quadrupolar nuclear species is a well-suited method for studying cation mismatch effects in mixed-alkali glasses. By magnetization transfer from a dilute foreign cation species to only nearby cations of the majority species, we were able to show that the latter ones are significantly immobilized as compared to the

area (%) ((1%) 81 (CP) 15 (SP) 80 (CP) 12 (SP) 70 (CP) 11 (SP)

remaining majority cations. These results provide direct evidence for the previously postulated secondary mismatch concept. Besides these findings, the systematic study of the crosspolarization experiment on crystalline LiNaSO4 displays that in the case of quadrupolar nuclei, this experiment requires a detailed investigation of the nutation and the spin-lock behavior of both nuclear species involved. Once these kinds of experiments have been validated and optimized, they should represent a powerful new approach for discriminating between different sites and establishing spatial proximities in many other applications beyond the realm of mixed-alkali glasses. Acknowledgment. This work has been funded by the DFG Sonderforschungsbereich Program SFB 458. S.P.P. also appreciates support by a personal stipend from the NRW Graduate School of Chemistry. References and Notes (1) Isard, J. O. J. Non-Cryst. Solids 1969, 1, 235. (2) Day, D. E. J. Non-Cryst. Solids 1976, 21, 343. (3) Tomozawa, M.; Yoshiyagawa, M. Glastech. Ber. 1983, 56, 939. (4) Ingram, M. D. Phys. Chem. Glasses 1987, 28, 215. (5) Bunde, A.; Ingram, M. D.; Maass, P.; Ngai, K. L. J. Non-Cryst. Solids 1991, 131-133, 1109. (6) Bunde, A.; Ingram, M. D.; Maass, P. J. Non-Cryst. Solids 1994, 172-174, 1222. (7) Bunde, A.; Ingram, M. D.; Russ, S. Phys. Chem. Chem. Phys. 2004, 6, 3663. (8) Habasaki, J. J. Non-Cryst. Solids 1995, 183, 12. (9) Balasubramanian, S.; Rao, K. J. J. Phys. Chem. 1993, 97, 8835. (10) Huang, C.; Cormack, A. N. J. Mater. Chem. 1992, 2, 281. (11) Cormack, A. N.; Cao, Y. Mol. Eng. 1996, 6, 183. (12) Jain, H.; Peterson, N. L.; Dowing, H. L. J. Non-Cryst. Solids 1983, 55, 283. (13) Day, D. E. J. Non-Cryst. Solids 1976, 21, 343. (14) Ingram, M. D.; Moynihan, C. T.; Lesikar, C. T. J. Non-Cryst. Solids 1980, 38-39, 371. (15) Ingram, M. D.; MacKenzie, M. A.; Muller, W.; Torge, M. Solid State Ionics 1990, 40-41, 671. (16) Gee, B.; Janssen, M.; Eckert, H. J. Non-Cryst. Solids 1997, 215, 41. (17) Hater, W.; Mu¨ller-Warmuth, W.; Meier, M.; Frischat, G. H. J. NonCryst. Solids 1989, 113, 210. (18) Dupree, R.; Holland, D.; Williams, D. S. J. Phys., Colloq. 1985, 8, 119. (19) Epping, J. D. Ph.D. Thesis, Mu¨nster, 2004. (20) Ratai, E.; Janssen, M.; Eckert, H. Solid State Ionics 1998, 105, 25. (21) Ratai, E.; Chan, J. C. C.; Eckert, H. Phys. Chem. Chem. Phys. 2002, 4, 3198. (22) Xue, X.; Stebbins, J. F. Phys. Chem. Miner. 1993, 20, 297. (23) Tossell, J. A. Phys. Chem. Miner. 1999, 27, 70. (24) Ali, F.; Chadwick, A. V.; Greaves, G. N.; Jermy, M. C.; Ngai, K. L.; Smith, M. E. Solid State Nucl. Magn. Reson. 1995, 5, 133. (25) Dupree, R.; Holland, D.; McMillan, P. W.; Pettifer, R. F. J. NonCryst. Solids 1984, 68, 399. (26) Meier, B. H. Chem. Phys. Lett. 1992, 188, 201. (27) Stejskal, E. O.; Schaefer, J.; Waugh, J. S. J. Magn. Reson. 1977, 28, 105. (28) Marks, D.; Vega, S. J. Magn. Reson. 1996, 118A, 57. (29) Vega, A. J. J. Magn. Reson. 1992, 96, 50. (30) Vega, A. J. Solid State Nucl. Magn. Reson. 1992, 1, 17.

Site Discrimination in Mixed-Alkali Glasses (31) van Wu¨llen, L.; Zu¨chner, L.; Mu¨ller-Warmuth, W.; Eckert, H. Solid State Nucl. Magn. Reson. 1996, 6, 203. (32) Chan, J. C. C.; Bertmer, M.; Eckert, H. Chem. Phys. Lett. 1998, 292, 154. (33) Chan, J. C. C.; Bertmer, M.; Eckert, H. J. Am. Chem. Soc. 1999, 121, 5238. (34) Eastman, M. A. J. Magn. Reson. 1999, 139, 98. (35) Medek, A.; Harwood, J. S.; Frydman, L. J. Am. Chem. Soc. 1995, 117, 12779. (36) Medek, A.; Frydman, L. J. Braz. Chem. Soc. 1999, 10, 263.

J. Phys. Chem. B, Vol. 110, No. 29, 2006 14261 (37) van Wu¨llen, L.; Mu¨ller-Warmuth, W. Solid State Nucl. Magn. Reson. 1993, 2, 279. (38) Ja¨ger, C.; Mu¨ller-Warmuth, W.; Mundus, C.; van Wu¨llen, L. J. Non-Cryst. Solids 1992, 49, 209. (39) Ja¨ger, C. NMR 1994, 31, 13. (40) Amoureux, J. P.; Fernandez, C.; Steuernagel, S. J. Magn. Reson. 1996, A123, 116. (41) Fernandez, C.; Amoureux, J. P.; Chezeau, J. M.; Delmotte, L.; Kessler, H. Microporous Mater. 1996, 6, 33.