3. Results and Discussion
1H and 11B nuclear magnetic resonance results. The evolution of the
1H NMR spectra for Na
3(BH
4)(B
12H
12) with temperature is shown in
Figure 2. The spectra exhibit a considerable narrowing with increasing temperature, which can be attributed to a partial averaging of dipole-dipole interactions of
1H spins due to jump motion of H atoms.
Figure 3 shows the temperature dependence of the
1H line width Δ
H (full width at half-maximum). It should be noted that at high temperatures, Δ
H does not drop to very small values, remaining on a plateau of approximately 10 kHz. This feature (typical of all the studied hydroborates [
17,
19,
22]) indicates the
localized nature of the motion of H atoms, as can be expected for anion reorientations. In contrast to the long-range translational diffusion, the localized motion leads to only partial averaging of the dipole-dipole interactions between moving nuclear spins.
Quantitative information on the H jump rates
can be obtained from measurements of the
1H spin-lattice relaxation rate
.
Figure 4 shows the measured
1H spin-lattice relaxation rates at two resonance frequencies
ω/2
π as functions of the inverse temperature. As can be seen from this figure,
exhibits two frequency-dependent peaks near 240 K and 130 K. Such
peaks are typical of many hydroborate systems [
17,
19,
22]; they occur at the temperatures, at which the jump rates
of anion reorientations become approximately equal to the resonance frequency
ω (~10
8 s
-1). The observation of two
maxima indicates a coexistence of two types of reorientational motion with different characteristic rates, as can be expected in the case of two types of the anions.
To distinguish between reorientations of the [BH
4]
− and [B
12H
12]
2− anions, we have used the behavior of the
11B spin-lattice relaxation rate
. For
, we may also expect the peaks related to the reorientational motion; however, the amplitudes of these peaks should differ significantly. The amplitude of the
peak due to B
12H
12 reorientations is determined by strong fluctuations of the electric quadrupole interaction of
11B nuclei [
22]. This amplitude should be considerably higher than the amplitude of the
peak due to BH
4 reorientations, since for the latter the quadrupole interaction is relatively unimportant, because of the special position of B atom in the center of the nearly regular tetrahedron. Indeed, a reorientational jump of BH
4 tetrahedron does not change a configuration of H atoms surrounding B atom, and the electric field gradient at B site remains nearly the same. The difference between the amplitudes of the
peaks due to B
12H
12 and BH
4 reorientations is consistent with the experimental results for
closo-borates and borohydrides [
17,
22,
23].
The results of the
11B spin-lattice relaxation measurements for Na
3(BH
4)(B
12H
12) are shown in
Figure 5. It should be noted that at
T < 200 K the recovery of the
11B nuclear magnetization deviates from a single-exponential behavior. As in the case of K
3(BH
4)(B
12H
12) [
16], such deviations may be related to nonzero electric quadrupole moment of
11B nuclei [
24] and to the presence of several well-separated inequivalent
11B nuclei relaxing with different rates. The relaxation curves at
T < 200 K can be satisfactorily approximated by sums of two exponential components. The
11B spin-lattice relaxation rates shown in
Figure 5 correspond to the single exponent at
T > 200 K and to the faster (dominant) exponential component at
T < 200 K.
As can be seen from
Figure 5,
exhibits two peaks in the same temperature ranges as the corresponding
peaks. On the basis of the observed amplitudes of the
peaks, we can conclude that the low-temperature peak originates from BH
4 reorientations, and the high-temperature one is due to B
12H
12 reorientations. Note that the
1H and
11B relaxation results resemble those found for K
3(BH
4)(B
12H
12) [
16]; however, the relaxation-rate peaks for the Na-based system (near 240 K and near 130 K) are shifted to considerably lower temperatures with respect to those for the K-based system (near 390 K and near 200 K, respectively [
16]). These results indicate higher reorientational mobility of both [BH
4]
− and [B
12H
12]
2− anions in the Na-based compound. Such a difference may be related to the smaller size of Na
+ cations.
For parametrization of the proton spin-lattice relaxation data, we have used the model based on two independent reorientational processes with the H jump rates
(
i = 1, 2). Similar model have been previously employed for K
3(BH
4)(B
12H
12) [
16]. We assume that
i = 1 corresponds to the faster process. According to the standard theory of nuclear spin-lattice relaxation due to the motionally-modulated dipole-dipole interaction [
25], in the limit of slow motion (
ωτi » 1),
is proportional to
ω-2τi-1, and in the limit of fast motion (
ωτi « 1),
is proportional to
τi, being frequency-independent. If the temperature dependence of both H jump rates follows the Arrhenius law,
with the activation energy
Eai for the
ith type of motion, for each of the peaks, the plot of ln
vs.
T -1 is expected to be linear in the limits of both slow and fast motion with the slopes of −
Eai/
kB and
Eai/
kB, respectively. The behavior of the proton spin-lattice relaxation rate shown in
Figure 4 exhibits some deviations from the predictions of the standard theory. First, for both relaxation rate peaks, the high-
T slope appears to be steeper than the low-
T one. Second, at the low-
T slope of each of the peaks, the frequency dependence of
is weaker than the predicted
ω-2 dependence. These features suggest the presence of a certain distribution of the H jump rates [
26]. The simplest approach to introducing such a distribution is based on using a Gaussian distribution of the activation energies [
26]. The details of the two-peak model [
27] used for analysis of the proton spin-lattice relaxation data for Na
3(BH
4)(B
12H
12) are presented in the Supplementary Information. The model parameters are the average activation energies
, the distribution widths (dispersions) Δ
Eai, the pre-exponential factors
τ0i, and the amplitude factors Δ
Mi determined by the strength of the fluctuating part of dipole-dipole interaction between nuclear spins for the
ith type of motion. These parameters have been varied to find the best fit of the model to the experimental
data at two resonance frequencies
simultaneously. The results of this simultaneous fit over the temperature range of 98 – 298 K are shown by black solid curves in
Figure 4; the corresponding parameters are
= 159(5) meV, Δ
Ea1 = 14(2) meV,
τ01 = 5.9(3) × 10
-15 s, Δ
M1 = 8.2(2) × 10
9 s
-2 (for the faster process of BH
4 reorientations), and
= 319(4) meV, Δ
Ea2 = 26(3) meV,
τ02 = 1.2(2) × 10
-15 s, Δ
M2 = 3.8(2) × 10
9 s
-2 (for the slower process of B
12H
12 reorientations). Comparison of these results with those obtained for K
3(BH
4)(B
12H
12) [
16] indicates that the average activation energies for both reorientational processes in Na
3(BH
4)(B
12H
12) are considerably lower than the corresponding activation energies in K
3(BH
4)(B
12H
12) (236 meV for BH
4 reorientations and 594 meV for B
12H
12 reorientations [
16]). This is consistent with the higher reorientational mobility of both anions in Na
3(BH
4)(B
12H
12).
23Na nuclear magnetic resonance results. Information on the cation (Na
+) dynamics can be obtained from
23Na NMR measurements. The
23Na NMR spectra have been studied in the low magnetic field (at the resonance frequency of 23 MHz) over a wide temperature range (8 – 552 K). The
23Na spin-lattice relaxation rate measurements (which require better signal-to-noise ratios) have been performed in the high field (at the resonance frequency of 132 MHz) above room temperature. Representative shapes of the
23Na NMR spectra at three temperatures are shown in
Figure S1 of the Supplementary Information.
Figure 6 shows the temperature dependence of the
23Na NMR line width Δ
Na (full width at half-maximum).
As the temperature increases, the line width becomes smaller due to motional averaging of the local magnetic and electric fields at Na sites. As can be seen from
Figure 6, the temperature dependence of Δ
Na exhibits two characteristic “steps”. The minor “step” near 100 K can be ascribed to the excitation of the reorientational motion of BH
4 groups inducing fluctuations of the
1H –
23Na dipole-dipole interaction. The major “step” near 300 K can be attributed to diffusive motion of Na
+ cations themselves, since this motion averages out all dipole-dipole and quadrupole interactions of
23Na nuclei. Indeed, in contrast to the case of Δ
H, the
23Na line width at high temperatures is very small (~ 1 kHz), which indicates that Na
+ cations participate in the
long-range diffusion. At the temperature of the main “step”, the diffusive jump rate
is expected [
25] to become nearly equal to the “rigid lattice” line width (~ 10
4 s
-1).
To probe the diffusive motion in the range of higher jump rates, we can use the
23Na spin-lattice relaxation measurements, since for these measurements, the characteristic frequency scale is determined by the resonance frequency
ω (~ 10
8 – 10
9 s
-1).
Figure 7 shows the measured
23Na spin-lattice relaxation rate
as a function of the inverse temperature.
As can be seen from this figure, above 420 K,
exhibits a significant increase, reaching a maximum near 530 K. The measured
values in the region of the maximum are much higher than those expected for the
23Na –
1H dipole-dipole interaction; therefore, the
23Na spin-lattice relaxation is dominated by fluctuations of the quadrupole interaction resulting from Na
+ jumps. This is a typical feature of all the studied sodium hydroborates [
5,
17,
22]. At the temperature of the
maximum, the Na
+ jump rate is expected to be nearly equal to the resonance frequency, i. e.,
(530 K) ≈ 8 × 10
8 s
-1. The activation energy for Na
+ diffusion estimated from the slope of the
peak is 390 meV. For comparison, the activation energy derived from the ionic conductivity data in the range of 273 – 468 K is 340 meV [
15].
Possible diffusion pathways were analyzed on the basis of the structure of Na
3(BH
4)(B
12H
12) [
15]. According to this analysis, Na
+ cations jump between the tetrahedral interstitial sites forming layers perpendicular to the
b axis, i. e., the diffusion is expected to be quasi-two-dimensional [
15]. The distance between the nearest-neighbor Na sites is approximately 2.30 Å. Neglecting any correlations in diffusive jump motion, the tracer diffusion coefficient of Na
+ ions for the case of two-dimensional diffusion can be estimated as
D =
L2/4
τd, where
L is the elementary jump length. Taking the distance between the nearest-neighbor Na sites as an estimate of
L, and using
τd at the temperature of the
maximum, we obtain
D(530 K) ≈ 1.1 × 10
-7 cm
2/s. This value is close to that found from the direct pulsed-field-gradient (PFG) NMR measurements of Li
+ diffusivity in LiLa(BH
4)
3Cl (
D(403 K) = 1.13 × 10
-7 cm
2/s [
28]). However, the estimated diffusivity in Na
3(BH
4)(B
12H
12) at 530 K is much lower than the record Na
+ diffusivity found from the PFG-NMR measurements for the mixed-anion Na
2(CB
9H
10)(CB
11H
12) (
D(403 K) = 8.7 × 10
-6 cm
2/s [
29]).
Following Matsuo et al. [
3], for comparison of the NMR diffusivity data with the ionic conductivity
σ, we can use the Nernst-Einstein equation,
where
n is the number of charge carriers per unit volume, and
Ze is the electrical charge of the carrier. It should be noted that this equation assumes uncorrelated diffusion. Using the above estimate of
D(530 K) and the value of
n = 8.8 × 10
21 cm
-3 found on the basis of the structural data [
15], we obtain from Eq. (2) that
σ(530 K) = 4.3 × 10
-3 S/cm. Experimentally, the ionic conductivity was measured up to 473 K, and its value at this temperature was 1.5 × 10
-2 S/cm [
15]. Taking into account the Arrhenius-type behavior of the measured
σ [
15], its extrapolation to 530 K gives
σext(530 K) = 5.0 × 10
-2 S/cm. Thus, the ionic conductivity estimated from the NMR diffusivity data appears to be nearly an order of magnitude lower than the corresponding experimental (extrapolated) value. Similar situation has been well documented in the case of the mixed-anion Li(CB
9H
10)
0.7(CB
11H
12)
0.3 system [
7] and the disordered LiCB
9H
10 phase [
28], where the measured conductivities are considerably higher than those derived from the PFG-NMR diffusivity data on the basis of the Nernst-Einstein equation. Such effects have been attributed to strong correlations between the diffusing ions (concerted diffusion) [
30,
31,
32].