Comparison of KNN and LNN

To gain additional insight into the local structure and disorder in doped NaNbO3,

LNN samples were also synthesised and characterised using solid-state NMR. It is worth noting here that only samples of LNN-3 and LNN-4 can be included to serve as a comparison to the KNN analogues - this is due to very different phase behaviour being present where x > 0.04, as discussed in detail in Chapter 5. Using a combination of solid-state NMR and DFT, there are four principal differences between the data obtained for samples of KNN and LNN: (i) the presence of an additional resonance in MQMAS NMR spectra of KNN (not observed for LNN); (ii) a distribution of chemical shifts for KNN; (iii) the presence of two additional resonances in 23Na DOR NMR spectra of KNN, and (iv) differences in the energies of structures from DFT when considering preferential site substitution of Li/K for Na. Each of these observations is discussed in more detail below.

In contrast to the observation of a third resonance in the 23Na MQMAS spectra of KNN-3, 4 and 5, 23Na MQMAS spectra of LNN-3 and LNN-4 contain only two resonances, as shown in Figs. 4.4 and 4.5. This suggests that any additional polymorph or clustering present in the KNN samples is not present in those in LNN. This is best exemplified at 20.0 T (Fig. 4.4) where the spectrum of KNN-4 clearly shows three resonances projected on the δ1 axis, whereas only two are observed for LNN-4.

When analysing 23Na MQMAS spectra (14.1 T) of KNN-2, 3, 4, 5 and 8, it is clear that the resonance attributed to site Na1 (<δ1> ≈ –10 ppm) lies along a gradient, compared to the lower ridge which lies parallel to the δ2 axis, as expected from a sheared spectrum. For KNN-5, this gradient is 2.12, where the value arises from

Δδ1/Δδ2. For a I = 3/2 nucleus, a gradient of 17/8 (or 2.125) corresponds to a distribution of chemical shifts and –5/4 (or –1.25) is indicative of a distribution of quadrupolar parameters. This suggests that the upper ridge is primarily affected by a distribution of chemical shifts, whereas the lower ridge is notably less affected, as it

lies parallel to the δ2 axis. Neither resonance is broadened along a gradient close to –

5/4, indicating that neither site is heavily affected by a distribution of quadrupolar

couplings. By comparison, the 23Na MQMAS spectrum of LNN-3 at 14.1 T, shown in

Fig. 4.5, does not show similar results. The lower ridge lies parallel to the δ1 axis and

the upper ridge is not noticeably broadened along any other gradient. This suggests

that samples with small amounts of Li+ _{substitution are less affected by distributions}

of chemical shifts compared to those with K+ _{substitution, perhaps due to smaller}

changes in the structure. Complementary 23Na DOR NMR spectra were acquired for

samples of KNN-4, 5 and 8 and LNN-3 and 4. The DOR spectrum of KNN-4 shows 4 resonances, while the spectrum of KNN-5 shows 3 (Fig. 4.6). In addition to the two

principal resonances assigned to phase Q (δ≈ –3.7 and –5.3 ppm), a small resonance

at –2.3 ppm can be seen for KNN-4, and a shoulder at –6.2 ppm. By determining the

δ1 line positions in the MQMAS spectra at two different fields, it is possible to predict the corresponding shifts of resonances in the DOR spectra, as outlined in Section

3.3.3. From the 23Na MQMAS spectra shown previously, the predicted 23Na DOR

shift can be calculated, giving a value of –0.41 ppm, where δiso = –0.33 ppm and δQ =

0.20 ppm at 20.0 T. This value, however, does not correspond to any of the

resonances in the 23_{Na DOR spectra, shown in Fig. 4.6. Unfortunately, the region}

immediately around the predicted DOR shift has a spinning sideband present. Faster spinning speeds would have moved the sidebands away from this area, although

difficulties in spinning stably at higher speeds prevented this. 23Na DOR NMR spectra

were also obtained for LNN-3 and LNN-4, for comparison to the spectra of KNN, as

shown in Fig. 4.7. Similar to the 23Na MQMAS spectra of these phases, only two

resonances are observed - both assigned to phase Q.

Figure 4.6: 23_{Na (20.0 T) DOR NMR spectra of (a) KNN-4 and (b) KNN-5. Inner} rotor spinning speeds of 1400 and 1650 Hz were used for KNN-4 and 5, respectively. All experiments were repeated with a second spinning speed to determine which peaks were spinning sidebands, denoted *. Q denotes resonances corresponding to phase Q and a + denotes additional resonances.

(a) (b) −15 −10 −5 5 0 ppm 5 0 −5 −10 −15 ppm * * * * * * * _* δ (ppm) δ (ppm)

Figure 4.7: 23Na (20.0 T) DOR NMR spectra of (a) LNN-3 and (b) LNN-4. Inner rotor spinning speeds of 1600 and 1400 Hz were used for LNN-3 and 4, respectively. All experiments were repeated with a second spinning speed to determine which peaks were spinning sidebands, denoted *.

−15 −10 −5 5 0 ppm 5 0 −5 −10 −15 ppm (a) (b) * * * * _{* * * *} –15 δ (ppm) δ (ppm) Q Q Q Q + +

The presence of the ‘additional’ resonances in the KNN spectra may suggest the formation of polar clusters/regions or the presence of an additional polymorph, neither of which there is evidence for in the spectra of LNN. In addition, a 23Na DOR NMR spectrum of KNN-8 was also acquired, as shown in Fig. 4.8. As the spectrum shows, the isotropic peaks are now heavily broadened, compared to KNN-4 and 5, indicative of an increase in the amount of disorder present. However, two distinct isotropic peaks can be seen, attributed to the two crystallographically-distinct Na sites in phase Q. There is good agreement between the DOR shifts of these resonances for KNN-4, 5 and 8, with small differences most likely due to subtle changes in the local structure caused by increased levels of K+ substitution. However, due to the breadth of the resonances it cannot be unambiguously stated that the other, low intensity resonances are not present, i.e., they may be overlapped with this signal.

* * * *

−10 −5

5 0 –15ppm

_{δ (ppm)}

Figure 4.8: 23Na (20.0 T) DOR NMR spectrum of KNN-8, obtained with an inner rotor spinning speed of 1400 Hz. Spinning sidebands are marked with a *.