Exceptional Superionic Conductivity in Disordered Sodium Decahydro‐closo‐decaborate

Na2B10H10 exhibits exceptional superionic conductivity above ca. 360 K (e.g., ca. 0.01 S cm−1 at 383 K) concomitant with its transition from an ordered monoclinic structure to a face-centered-cubic arrangement of orientationally disordered B10H102− anions harboring a vacancy-rich Na+ cation sublattice. This discovery represents a major advancement for solid-state Na+ fast-ion conduction at technologically relevant device temperatures. Recently, complex hydride salts undergoing solid-state, entropy-driven, order–disorder transitions have been shown to exhibit impressive fast-ion conduction properties as a result of the appearance of vacancy-rich cation sublattices within networks of highly mobile, reorientationally disordered polyanions.1 Initial interest has mainly focused on the light-metal Li and Na salts possessing tetrahydroborate (BH4−) polyanions (see Figure 1), such as LiBH4 and Na2BH4NH2, as well as related derivative materials.2-5 Very recently, it was discovered that Li and Na salts possessing the larger icosahedral dodecahydro-closo-dodecaborate (B12H122−) anions (see Figure 1) also undergo order–disorder phase transitions,6, 7 with Na2B12H12 exhibiting disorder-induced superionic conductivity (approaching 0.1 S cm−1) above around 480 K.8 This conductivity rivals that of traditional ceramic materials, Na β”-alumina solid electrolyte (BASE), and Na superionic conductor (NASICON), as well as more recent Na3PS4-based glass ceramic electrolytes, all currently of considerable interest for use in Na-ion batteries.9-11 The pronounced superionicity and relatively low conduction barrier for disordered Na2B12H12 are probably due, in part, to the larger size (and roughly spherical shape) of the B12H122− anions compared with the substantially smaller BH4− anions (see Figure 1) present in other investigated fast-ion conductors.1 These types of large polyanion compounds represent a potentially fertile area for discovering new materials with superionic conductivities, but with lower transition temperatures. Here we report our finding that sodium decahydro-closo-decaborate (Na2B10H10), a related sodium salt containing large, ellipsoidal-shaped, B10H102− anions (see Figure 1), forms a disordered, face-centered-cubic (fcc) phase above ca. 360 K, possessing a vacancy-rich Na+ cation sublattice. This cation sublattice is highly mobile within the spacious corridors formed by the large B10H102− anions and exhibits remarkable superionic conductivity (e.g., σ ≈ 0.01 S cm−1 at 383 K) to substantially lower temperatures than for Na2B12H12. This conductivity is more than an order of magnitude higher than that of all other solid-state Na-based complex-hydride materials investigated to date in this temperature region.1 As such, this discovery represents a major advancement in the field of solid-state Na+ fast-ion conduction at technologically relevant device temperatures. Figure 2 shows differential scanning calorimetry (DSC) results for Na2B10H10 after several heating/cooling cycles. There is a clear reversible transition, first observed by Bonnetot et al.,12 with minor hysteresis. After a number of cycles to 500 K, the hysteresis decreases and the onset temperature stabilizes near 360 K upon both heating and cooling. A slow attenuation of the DSC features begins to occur upon repeated cycling to temperatures near 600 K. Figure 3 shows the neutron powder diffraction (NPD) result for a partially deuterated Na211B10H10 sample at 410 K, above the phase transition. Table S1 in the Supporting Information lists the corresponding structural parameters. The Rietveld-refined model confirms transformation from the known, low-T, ordered monoclinic structure13 (not shown) to a high-T disordered structure with Na+ cations partially occupying a variety of interstitial sites within an fcc lattice of orientationally disordered anions. Neutron-scattering Fourier difference maps initially suggested that the broad distribution of cation positions and the inclusion of the three most intense positions were ultimately necessary to attain a good model fit to the data. Furthermore, the refinements suggested extensive anion orientational disorder, which could be represented in various ways via multiple B and H positions. The model reflects the simplest representation of six approximately superimposed anion orientations, each equally probable. In particular, the six possible anion orientations are aligned in pairs with their long axes oriented along any one of the three orthogonal crystallographic axes. The members of each pair are azimuthally offset from each other by 45° about their long axes. Each B and H position in the structure is 1/3 occupied, being shared by two of the six possible orientations. Anion dynamical behavior was probed by neutron elastic-scattering fixed-window scans (FWSs)14 of Na211B10H10. The results in Figure 4 suggest that a dramatic change in B10H102− anion reorientational mobility occurs upon phase transformation. In particular, the high elastic neutron counts in the low-T ordered phase suggest anion reorientational jump frequencies less than 108 s−1, whereas the roughly 80% lower counts in the high-T disordered phase suggest an orders-of-magnitude enhancement in jump frequencies to greater than 1010 s−1. This is reminiscent of the FWS behavior observed for Na2B12H12.7, 15 Indeed, the inset in Figure 4 showing a quasielastic neutron scattering (QENS) spectrum for the disordered phase at 375 K indicates a quasielastic component with a Lorentzian linewidth of about 41(1) μeV FWHM, which reflects a jump correlation frequency on the order of 3 × 1010 s−1. The ratio of elastic and total scattering intensities (which is also consistent with the ratio of FWS neutron counts in the disordered and ordered phases from Figure 4) suggests that, besides the B10H102− reorientational jumps around the long axis, two-fold anion flips leading to exchanges of apical H atom positions are also occurring. We also probed the Na+ dynamical behavior in Na2B10H10 by 23Na NMR measurements. Figure 5 shows the 23Na spin-lattice relaxation rate R1 at the resonance frequency of ω/2π = 23 MHz as a function of T−1. The general features of the behavior of R1 for Na2B10H10 resemble those for Na2B12H12 at the phase transition.16 Here, R1 exhibits a jump accompanied by a change in sign of its temperature dependence. Such behavior indicates that the transition from the ordered to the disordered phase is accompanied by an abrupt increase in the Na+ jump rate τd−1. The expected R1(T) maximum is ‘folded’; i.e., because of the abrupt increase in τd−1 at the phase transition, there is a jump directly from the low-T slope to the high-T slope of the R1(T) peak. On the low-T slope, R1 is proportional to τd−1; on the high-T slope, R1 is proportional to τd. From the two slopes, we obtain activation energies for Na+ jumps in the ordered and disordered phases of 750(20) meV and 190(10) meV, respectively. Although we cannot reliably determine the absolute values of τd−1 due to the ‘folded’ nature of the R1(T) peak, the data do allow us to conclude that the Na+ jump rate exceeds ω ≈ 1.5 × 108 s−1 just above the phase transition. Moreover, the very small 23Na NMR linewidth (0.2 kHz full-width half-maximum (FWHM)) observed in the disordered phase confirms that the Na+ cations are undergoing long-range diffusion. To characterize the Na2B10H10 conduction behavior, we carried out AC impedance measurements between 295 K and 423 K using a pressed disk of polycrystalline Na2B10H10 powder with either gold or (similarly performing) molybdenum foil contacts. Cross-sectional SEM images of the pelletized sample in Figure S1 of the Supporting Information confirm that intimate contacts among the particles were achieved. Na2B10H10 can be easily pelletized without further sintering, as has been reported for LiBH4 and other complex hydrides.1,5 The results are shown in Figure 6 for the gold contacts. The inset shows typical complex impedance plots at various temperatures. They consist of an arc in the high-frequency region and a spike in the low-frequency region due to contributions from the bulk/grain boundaries and the electrode, respectively. The results suggest that Na2B10H10, similar to Na2B12H12,8 behaves like a typical ionic conductor. The temperature dependence of the conductivity indicates dramatic superionic conductivity above the hysteretic order–disorder transition near 373 K, rising two orders of magnitude higher than that in the low-T phase. This behavior is superior to that of other investigated complex hydride materials. Indeed, the conductivity exhibits a value of about 0.01 S cm−1 at 383 K, which is about 25× greater than that of Na2BH4NH2 (4 × 10−4 S cm−1).4 A conductivity above 0.1 S cm−1 at 500 K is suggested by extrapolation of the lower-T data. The activation energy for conduction is evaluated to be 0.47 eV, which is higher than that reported for Na2B12H12 (0.21 eV)8 but lower than those reported for Na2BH4NH2 (0.61 eV)4, 8 and the high-T phase of LiBH4 (0.53 eV).2 Nevertheless, any rationalization of the differences based solely on anion size is complicated by the substantial differences in structure and conduction pathways. It should be noted that the NMR-derived activation energies discussed above reflect the average microscopic barriers for all cation diffusional jumps between neighboring sites within the Na2B10H10 lattice, some of which may have little effect on the macroscopic conductivity barrier. In contrast, the latter barrier is more reflective of an overall rate-limiting step, such as a particular type of cation jump within the material required to maintain conduction pathways or cation transport, e.g., across grain boundaries (although such grain-boundary bottlenecks are believed to be small in the present system). The superionic conductivity of disordered Na2B10H10 is consistent with the relatively small activation energy for Na+ diffusion within the liquid-like cation sublattice. Again, similar to disordered Na2B12H12,8 the overly large size and spheroidal shape of the polyanions result in less restrictive interstitial pathways and, hence, reduced Na+ diffusional bottlenecks between the various cation sites within the close-packed anion sublattice. As for the other disordered complex hydrides, the reorientationally mobile anions associated with superionic Na2B10H10 may also lower the cation diffusional barrier by providing a dynamically cooperative environment for cation jumps within the voids of the anion sublattice. Indeed, at least an order-of-magnitude-higher anion reorientational jump rate compared with the Na+ diffusional jump rate provides a dynamic environment where the anions can behave as ‘lubricants’ for cation diffusive motions. A comparison of the relative sizes of the B10H102− and B12H122− anions in Figure 1 indicates a similar maximum dimension for each anion. In fact, the lattice constants for the disordered fcc Na2B10H10 and body-centered cubic (bcc) Na2B12H12 structures7 indicate that both disordered anions possess similar spherical packing radii of ≈3.5 Å. This makes the small 190 meV activation energy for Na+ diffusion in Na2B10H10 particularly noteworthy, since it is less than half that of Na2B12H12 (410 meV).16 Such a difference may be the result of the different natures of the diffusion saddle points inherent within the fcc and bcc structures, but may also signal a local geometric advantage that the less-spherical B10H102− anions have over their more spherical relatives. In particular, within a cubic structure, one might expect each of the more-ellipsoidal B10H102− anions to take up slightly less space in directions perpendicular to their long axes than the more spherical B12H122− anions. On a local level, this would allow more free space between anions for cation diffusion. The QENS results are consistent with a locally ellipsoidal anion, by suggesting that each anion retains a particular orientation of its long axis over at least a nanosecond timescale. Much needs to be done to provide a better understanding of the superionic properties of this new class of conducting materials. For example, since Na2B10H10 possesses a much lower order–disorder phase-transition temperature than Na2B12H12, one might think that the lighter-metal analogue, Li2B10H10, would also possess a lower order–disorder phase-transition temperature than Li2B12H12,6, 7 thus enhancing the stability of the disordered, fast-ion-conducting structure. Our Li2B10H10 DSC measurements suggest the contrary. Rather, Li2B10H10 appears to possess a slightly higher transition temperature than Li2B12H12 does, leading to an unstable disordered structure, which makes it unsuitable as a solid-state, Li+-conducting electrolyte. However, the addition of other anions or cations to Li2B10H10 and to Na2B10H10 may lead to hybrid materials displaying even lower transition temperatures than seen here for pure Na2B10H10. We are currently pursuing such potentially favorable modifications. For all these disordered materials, a more thorough understanding of the relationship of structural disorder and anion reorientational mobility to cation diffusion and conductivity will benefit from future first-principles molecular-dynamics calculations,1, 17, 18 which may in turn lead to a more rational pathway to develop improved modified materials. In conclusion, the discovery of very high superionic conductivity in Na2B10H10 that persists to temperatures as low as 360 K is a marked improvement over Na2B12H12 and other complex hydrides. Although hygroscopic, Na2B10H10 remains air-stable at room temperature with no noticeable degradation in its diffraction pattern up to at least 500 K. DSC indicates that the compound decomposes/polymerizes with some mass loss of presumably H2 at around 850 K. Preliminary cyclic voltammetry measurements indicate that ordered Na2B10H10 is electrochemically stable up to at least 4 V at 353 K and disordered Na2B10H10 up to at least 5 V at 393 K (see Figure S2 in the Supporting Information). Its favorable properties and high conductivity warrant a serious investigation of Na2B10H10's applicability to next-generation solid-state Na-ion battery technologies. Based on these results, successful future searches for related materials with even better cation conductivity properties may be enhanced by the inclusion of similar- or even larger-sized polyanions compared with B10H102−. Synthesis: 11Boron-enriched Na211B10H10 (and partially deuterated Na211B10H10) was synthesized as follows: the triethylammonium salt (Et3NH)2[11B10H10] was synthesized via reaction of 11B10H14 (Katchem19 and triethylamine in para-xylene at reflux. The crude product was recrystallized from water/EtOH and dried in vacuum (10 mTorr) at room temperature for 16 h. The (Et3NH)2[11B10H10] was then converted into the corresponding acid (H3O)2[11B10H10] by ion exchange using an Amberlite resin in H+-form. Aqueous Na211B10H10 was prepared by neutralization of (H3O)2[11B10H10] with 0.1 m NaOH until a pH value of 7 was reached. The solvent was removed on a rotary evaporator at 323 K. Unlabeled Na2B10H10 was synthesized using a similar approach. The resulting hydrated materials were dried under vacuum at 433 K for 16 h. For the partially deuterated sample, a single exchange treatment was performed by dissolution and stirring for 3 h of 1 g of Na211B10H10 in 20 mL of D2O slightly acidified by adding 50 μL of a saturated solution of deuterochloric (DCl) acid in D2O. The resulting dried sample had a D:H ratio of only 27:73 as determined from refinement of the 20 K NPD pattern, yet led to some reduction of the incoherent neutron scattering background from the lighter H isotope. Measurement Details: DSC measurements were made with a Netzsch (STA 449 F1 Jupiter) TGA-DSC under He flow with Al sample pans. The neutron-scattering measurements were performed at the National Institute of Standards and Technology Center for Neutron Research. NPD patterns were measured on the BT-1 High-Resolution Powder Diffractometer using the Ge(311) monochromator at a neutron wavelength of 2.077 Å. Horizontal divergences of 60′, 20′, and 7′ of arc were used for the in-pile, monochromatic-beam, and diffracted-beam collimators, respectively. The sample was contained in a 6 mm-diameter V can inside a He closed-cycle refrigerator. FWSs were measured on the High-Flux Backscattering Spectrometer using 6.27 Å wavelength neutrons, with a resolution of 0.8 μeV FWHM. QENS spectra were collected at 270 K (resolution measurement) and 375 K on the Disk Chopper Spectrometer using 4.08 Å wavelength neutrons with a resolution of 79 μeV FWHM. 23Na NMR measurements were performed on the pulse spectrometer described earlier16 at the frequency ω/2π = 23 MHz. The nuclear spin–lattice relaxation rates were measured using the saturation–recovery method. NMR spectra were recorded by Fourier transforming the solid-echo signals. Ionic conductivities were determined in heating and cooling runs repeatedly in the temperature range between 303 K and 423 K by the AC complex-impedance method using an NF FRA5097 frequency response analyzer over a frequency range of 10 Hz to 10 MHz. All the measurements were performed under Ar. The powder sample was pressed into a pellet of 8 mm in diameter and 2 mm in thickness without sintering. The pellet density was about 1.17 g cm−3, which is more than 95% of the density calculated from the lattice parameters. Au or Mo foils were used as electrodes and were mechanically fixed on both faces of the pellet. The resistances of the sample were obtained by least-square fittings of a single arc in the high-frequency range using equivalent circuits of a parallel combination of a resistance and a capacitance. At high temperature, since only a spike caused by the electrode contribution was observed, the resistance values were calculated from the intercept of the spike. The cross-section of the pelletized sample was examined by scanning electron microscopy (SEM) (JEOL JSM6009). Cyclic voltammetry measurements were conducted at scan rates of 5 mV s−1 using a potentiostat/galvanostat (Princeton VersaSTAT4) with a Mo disk as the working electrode and counter/reference electrodes of Na or Na–In at 353 K and 393 K, respectively. Finally, for all the figures, standard uncertainties are commensurate with the observed scatter in the data, if not explicitly designated by the vertical error bars. This work was performed, in part, in collaboration between members of IEA HIA Task 32-Hydrogen-based Energy Storage. The authors gratefully acknowledge support from DOE EERE through Grant Nos. DE-EE0002978 and DE-AC04–94AL85000; the Russian Foundation for Basic Research under Grant No. 12–03–00078; the U.S. Civilian Research & Development Foundation (CRDF Global) under Award No. RUP1–7076-EK-12; the National Science Foundation (NSF) under Cooperative Agreement No. OISE-9531011; the Integrated Materials Research Center for the Low-Carbon Society (LC-IMR), Tohoku University; the Advanced Low Carbon Technology Research and Development Program (ALCA) from the Japan Science and Technology Agency (JST); and JSPS KAKENHI under Grant Nos. 25220911 and 26820311. This work utilized facilities supported in part by the NSF under Agreement No. DMR-0944772. The authors also thank Dr. Nina Verdal for with the QENS measurements. As a to our authors and this provides by the Such materials are and may be for but are not or support from than should be to the The is not for the or of any by the than should be to the corresponding for the