Sodium Ion Diffusion in Nasicon (Na3Zr2Si2PO12) Solid Electrolytes

Sep 26, 2016 - Molten sodium (Na) based batteries (molten Na batteries) implementing Na–S or Na–metal halide (i.e., Na–NiCl2 or Na–FeCl2) chem...
2 downloads 10 Views 5MB Size
Research Article www.acsami.org

Sodium Ion Diffusion in Nasicon (Na3Zr2Si2PO12) Solid Electrolytes: Effects of Excess Sodium Heetaek Park,† Keeyoung Jung,‡ Marjan Nezafati,§ Chang-Soo Kim,*,§ and Byoungwoo Kang*,† †

Department of Materials Science and Engineering, Pohang University of Science and Technology (POSTECH), Po-hang, Gyeongbuk 790-784, South Korea ‡ Energy Storage Materials Research Center, Research Institute of Industrial Science and Technology (RIST), Pohang, Gyeongbuk 790-330, South Korea § Materials Science and Engineering Department, University of WisconsinMilwaukee, Milwaukee, Wisconsin 53211, United States S Supporting Information *

ABSTRACT: The Na superionic conductor (aka Nasicon, Na1+xZr2SixP3−xO12, where 0 ≤ x ≤ 3) is one of the promising solid electrolyte materials used in advanced molten Na-based secondary batteries that typically operate at high temperature (over ∼270 °C). Nasicon provides a 3D diffusion network allowing the transport of the active Na-ion species (i.e., ionic conductor) while blocking the conduction of electrons (i.e., electronic insulator) between the anode and cathode compartments of cells. In this work, the standard Nasicon (Na3Zr2Si2PO12, bare sample) and 10 at% Na-excess Nasicon (Na3.3Zr2Si2PO12, Na-excess sample) solid electrolytes were synthesized using a solid-state sintering technique to elucidate the Na diffusion mechanism (i.e., grain diffusion or grain boundary diffusion) and the impacts of adding excess Na at relatively low and high temperatures. The structural, thermal, and ionic transport characterizations were conducted using various experimental tools including X-ray diffraction (XRD), differential scanning calorimetry (DSC), scanning electron microscopy (SEM), and electrochemical impedance spectroscopy (EIS). In addition, an ab initio atomistic modeling study was carried out to computationally examine the detailed microstructures of Nasicon materials, as well as to support the experimental observations. Through this combination work comprising experimental and computational investigations, we show that the predominant mechanisms of Na-ion transport in the Nasicon structure are the grain boundary and the grain diffusion at low and high temperatures, respectively. Also, it was found that adding 10 at% excess Na could give rise to a substantial increase in the total conductivity (e.g., ∼1.2 × 10−1 S/cm at 300 °C) of Nasicon electrolytes resulting from the enlargement of the bottleneck areas in the Na diffusion channels of polycrystalline grains. KEYWORDS: molten Na battery, solid electrolyte, Nasicon, ionic conductivity, high temperature, Na excess

1. INTRODUCTION

solid electrolyte disk. The Nasicon (Na1+xZr2SixP3−xO12, where 0 ≤ x ≤ 3, Na superionic conductor) ceramic compounds are one of the candidate solid electrolyte materials for these molten Na secondary cells, featured by their 3D network channels comprising (Si/P)O4 tetrahedra that share the pocket corners with ZrO6 octahedra for Na ion diffusion, which can potentially present a high ionic conductivity.1,6−8 The synthesis procedure of Nasicon compounds could be relatively simpler compared with other types of solid electrolyte such as β/β″-Al2O3.1,9 The Nasicon compounds can form either a rhombohedral (R3̅ c) or a monoclinic (C2/c) crystallographic structure depending on its compositions and ambient temperatures. Most of the prior studies have focused on the Nasicon materials with the composition range of 1.8 < x < 2.4 (x in the Na1+xZr2SixP3−xO12 stoichiometry), as the Na3Zr2Si2PO12

Molten sodium (Na) based batteries (molten Na batteries) implementing Na−S or Na−metal halide (i.e., Na−NiCl2 or Na−FeCl2) chemistries are considered as one of the most attractive secondary cells for advanced grid-scale energy storage systems (ESS) because of their high energy density, long discharge time, long cell life, low manufacturing cost, and superior capacity retention capabilities.1−5 For facile transport of Na ions between anode and cathode, the molten Na batteries are typically operated at relatively high temperatures (over ∼300 °C and ∼270 °C for Na−S and Na−metal halide chemistries, respectively). The anodic and cathodic electrodes in these cells are physically and electronically separated by a thin ceramic solid electrolyte that enables ionic diffusion and migration upon charge/discharge. For a deeper penetration to the application of molten Na batteries, developing superior solid electrolyte materials that exhibit high Na ion conductivities (σion) is indispensable. Figure 1 provides an example of molten Na battery cross-section structure containing a thin © 2016 American Chemical Society

Received: August 9, 2016 Accepted: September 26, 2016 Published: September 26, 2016 27814

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

advanced Nasicon solid electrolytes for molten Na battery applications.

2. EXPERIMENTAL SECTION Sample Preparation. The standard Nasicon (“bare sample”, Na3Zr2Si2PO12 stoichiometry) and the 10 at% Na-excess Nasicon (“Na-excess sample”, Na3.3Zr2Si2PO12 stoichiometry) specimens were synthesized using a solid-state reaction process. The bare sample was prepared by mixing a stoichiometric combination of Na2CO3, ZrO2, SiO2, and NH4H2PO4, and for the Na-excess sample, additional Na2CO3 was added controlling the ratios of (Na/Zr/Si/P = 3.3/2/2/ 1) in the starting mixture. The raw materials were mixed for 24 h by ball-milling with 3, 5, and 10 mm ZrO2 balls with ethanol solvent at 300 rpm. Next, the ball-milled mixture was preheated at 600 °C for 4 h and then calcined at 1150 °C for 4 h under air environment to remove any volatile species in the synthesized compounds. The calcined samples were pulverized for 2 h by a planetary ball-mill at 500 rpm with 1 mm ZrO2 balls with ethanol solvent in a ZrO2 container. The pulverized particles were ground and were then put into a mold and pressed at ∼300 MPa to form 12.9 mm pellet samples. The pellet samples were finally sintered at 1100 °C for 10 h under the air environment. The 20 at% Na-excess sample (Na3.6Zr2Si2PO12, stoichiometry) was also synthesized and the experimental results of this 20 at% Na-excess sample are included in Supporting Information (see Figure S4, Table S3, and Table S4 in Supporting Information). Structural and Thermal Characterization. The crystal structure of the powder samples was characterized using a synchrotron powder X-ray diffraction (XRD, 9B HRPD beamline at Pohang Light Source II (PLS-II), South Korea). Synchrotron XRD spectra of the samples have been recorded with six base detectors using a step size of 0.02°, exposure time of 6 s, and 2θ angle range of 10−60°. MDI Jade 6 software was utilized to detect the impurities in the synthesized samples, and the Rietveld refinement with the X’pert HighScore Plus software program was used to analyze the XRD patterns. The size and morphology of the particulates in the synthesized pellet samples have been identified using a BV field emission scanning electron microscopy (SEM, XL30S FEG, Philips Electron Optics). The SEM images were obtained at a vacuum level of below 10−5 mbar with an accelerating voltage of 5 kV. Differential scanning calorimetry (DSC, SDT Q600, TA Instruments) testing was performed for the bare sample in the temperature range of 25−300 °C with synthetic air flow (50 mL/min). To extract the peak temperature (Tp) associated with the monoclinic− rhombohedral phase transformation, the measured DSC curves of the samples have been corrected by the sapphire calibration and the blank test.17 The blank curve was obtained by heating the empty crucibles (Pt). A ramping rate of 10 °C/min was used in all DSC experiments. Electrochemical Test. After deposition of Pt blocking electrodes with a thickness of ∼100 nm on the top and the bottom surfaces of pellets by sputtering, an electrochemical ac impedance spectroscopy (EIS) testing has been conducted in the frequency (f) range of 1−106 Hz. For high temperature EIS test (above 100 °C), the temperature was controlled by a box furnace (Ajeon Heating Industrial Co., South Korea). σion of the samples was obtained from the intercept point of the Nyquist plot and the real axis at high f. In an effort to distinguish the total σion from the grain diffusion and the grain boundary diffusion in the polycrystalline samples, EIS testing at various low temperatures (−10, −20, −30, and −40 °C) was also performed maintaining constant temperature and humidity values. DFT Computation. The density-functional theory (DFT) calculations have been conducted using 4 formula units of standard monoclinic 4(Na3Zr2Si2P1O12) system with 75% occupancy of Na sites. The most stable structure was selected among different configurations with different Na2 and Na3 site occupancies. For the Na-excess samples, 16.67 at% excess Na (i.e., two more Na atoms in the 4 formula unit) was added to the thermodynamically stable system to track any structural changes due to the presence of excess Na. Note that due to the stoichiometric ratios among the constituent elements, 16.67 at% excess Na, instead of 10 at%, was added to the standard bare Nasicon system. For all of the computations, the Dmol3 module

Figure 1. Schematics of the cross-section of a molten Na battery (Na− S chemistry) and Nasicon structures.

(Na1+xZr2SixP3−xO12, where x = 2) structure has shown the highest ionic conductivity (σion).1,10−12 With increasing temperature, the stable monoclinic phase of standard Nasicon (i.e., Na3Zr2Si2PO12 stoichiometry) at room temperature transforms to the rhombohedral phase at ∼150 °C.11−13 With the standard Na3Zr2Si2PO12 stoichiometry, there are four Na sites in 1 formula unit of both monoclinic and rhombohedral crystals. The four Na sites are composed of one Na1 and three Na2 sites in the rhombohedral crystal, and these three Na2 sites in rhombohedral structure are further split into one Na2 and two Na3 sites in the monoclinic crystal.14 Out of these four available Na sites, three Na sites are in principle occupied (i.e., 75% Na site occupancy) in the standard Na3Zr2Si2PO12 stoichiometry.10,12 To develop Nasicon electrolytes with enhanced σion, most of the previous studies have been dedicated to test the effects of doping on the octahedral/tetrahedral site or adopting new sintering processes such as spark plasma sintering (SPS).1,6,15,16 It has been reported that certain elements such as Nb, Ta, and V can lower Ea, and some dopants such as Y and Al can reduce the porosity of sintered body. In contrast to the previous works, to improve σion of the Nasicon solid electrolytes, we present a study to explore the impacts of adding excess Na (i.e., higher Na site occupancy than 75%) to the standard Nasicon compounds combining solid-state synthesis, structural and electrochemical characterizations, and in silico density-functional theory (DFT) computations. In the structural point of view, it is speculated that adding excess Na could give rise to a counter effect, increasing the absolute concentration of Na-ions and decreasing the vacant Na sites in the Nasicon structure, thereby increasing and decreasing the resultant σion of electrolytes, respectively. σion of Nasicon electrolytes containing excess-Na will be influenced by these complicated factors: absolute amounts of Na ions, available hopping sties, and the mobility of effective Na ions. In conjunction with the effects of adding excess Na to the standard Nasicon compound to increase σion, we attempted to reveal the predominant Na-ion transport mechanisms (i.e., intergranular or transgranular diffusion) at relatively low and high temperatures in the polycrystalline Nasicon, as these diffusion mechanisms have not been clearly understood previously. The knowledge on the excess-Na effects and the diffusion mechanism of Na ions obtained in this work will provide useful insights for developing 27815

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

Figure 2. (a) Synchrotron XRD patterns obtained from the bare and Na-excess samples, (b) heat capacity (Cp) curve of the bare sample calibrated from DSC measurement, and SEM images of the two sintered pellets, (c) bare and (d) Na-excess samples. implemented in the Materials Studio software package (2016 version, BIOVIA Inc.)18,19 was used. The initial 2 × 2 × 1 supercell structures were geometry optimized to attain respective equilibrium standard bare structures. In conducting the DFT computations, the exchange− correlation energy with PW91 functional in the GGA scheme was applied to the system.20 We used the double-numerical plus dfunctions quality basis set with all electron core treatment, Fermi smearing equal to 0.01 Ha (1 Ha = 27.2114 eV), and a real space cutoff value of 4.6 Å to improve the computational performance. The self-consisting iteration tolerance of 10−5 was selected, and the density of mixing with the value of 0.05 for both charge and spin was used. The k-point mesh for these calculations was set to (1 × 1 × 2) and the convergence tolerance was defined as 2 × 10−5 Ha for energy, 4 × 10−3 Ha/Å for force gradient, and 5 × 10−3 Å for displacement, respectively.

excess sample. Larger amounts of the impurities were observed by adding 20 at% excess Na (see Figure S4a and Table S3). As the two samples were prepared under the identical experimental condition, it is likely that the formation of Na3PO4 impurity phase could be a consequence of the presence of excess Na rather than the decomposition of Nasicon phase at high temperature. In Figure 2b, we show the heat capacity (Cp) curve from the bare sample calibrated through DSC experiments. The characteristic feature of this DSC curve is generally consistent with the observations in the previous literature;11,13,24 an endothermic peak showing a monoclinic− rhombohedral transition is detected with a maximum peak position of ∼157 °C. The peak position and the temperature range of the second order phase transition can be varied depending on the sample preparation method, sintering conditions, and scan conditions.6,25 The DSC measurement curve for the Na-excess sample exhibited a similar trend including the peak phase transition temperature of ∼157 °C (see Figure S3). On the basis of the DSC measurement curves for both samples, it was also found that the thermal stability has not noticeably changed after adding excess Na. In Figure 2c and Figure 2d, we present the SEM images from the two samples to show the sizes and the morphologies of the sintered particulates. The particle sizes of the two samples (∼500 nm) are similar to each other partly because they have been subject to the planetary ball-milling process before sintering. These two samples also exhibited similar morphologies of particles, i.e., irregular faceted surfaces. Therefore, it would be safe to assume that the difference in the ion conduction behaviors, if any, in the two samples would not be majorly attributed to the differences in the microstructural particle size or morphology of

3. RESULTS AND DISCUSSION Structural Features. Figure 2a shows the XRD patterns measured from the bare and Na-excess samples at room temperature. In this study, the crystal structure proposed in ref 14 was used as a reference to analyze the structural features of the synthesized samples. The detailed results of Rietveld refinement analysis of the bare and Na-excess samples are included in Supporting Information (see Figure S1, Figure S2, Table S1, and Table S2). Both samples majorly contained the monoclinic Nasicon phase with minor monoclinic ZrO2 as an impurity phase. The compositions (wt %) of the Nasicon and the impurity phases were 95.5 and 4.5 wt % for the bare sample and 91.5 and 8.5 wt % for the Na-excess sample, respectively. ZrO2 may be formed either by the loss of Na and P at high temperature or by the low reactivity of monoclinic ZrO2 during the solid-state synthesis process.21−23 As indicated in Figure 2a, a minimal trace of Na3PO4 phase was observed only in the Na27816

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces the synthesized pellets. The relative pellet densities were measured as 3.00 g/cm3 (10.1% porosity) and 2.95 g/cm3 (12.0% porosity) for the bare and Na-excess samples, respectively. With this similar porosity between the two sintered samples, we also assume that the porosity would minimally affect the Na ion transport phenomena in the two samples, as reported in ref 26.26 Table 1 lists the details of the structural features of the two samples acquired from the XRD refinement analysis. In the Table 1. Unit Cell Parameters for the Monoclinic Bare and the Na-Excess Nasicon Systems Obtained by XRD Refinements lattice parameter (Å) bare Na-excess

a

b

c

β (deg)

volume (Å3)

15.6487 15.6671

9.0517 9.0632

9.2167 9.2146

123.7531 123.8471

1085.4621 1086.6781

Figure 3. Four different types of bottlenecks (A−D) in Na conducting pathways in the monoclinic Nasicon structure.

Table 2. Summary of the Bottleneck Sizes in the Bare and the Na-Excess Samples Measured by the XRD Refinements

table, β is the lattice angle between x- and z-axes of monoclinic unit cell. It was calculated that although the differences between the bare and Na-excess samples were not large, the unit cell volume of the Na-excess specimen is slightly greater (i.e., ∼ 0.11% increase) than that of the bare sample with larger a and c lattice parameter values. Because of the content of impurities including ZrO2 and Na3PO4 increases by adding excess Na, the Na-excess sample might be deficient in the Zr and P elements in the bulk Nasicon structure; therefore, the change of the lattice parameters would be stemming from the change of bulk composition. For both of the monoclinic and rhombohedral structures, the Na-ions must be transported through triangular “bottleneck” areas consist of three O atoms in the (Si/P)O4 tetrahedron and ZrO6 octahedron pockets.27,28 Because the Na ion diffusion energy barrier (Ea) within the electrolytes would be strongly influenced by the sizes of these bottleneck areas that can in turn directly determine the ion transport kinetics of Nasicon,29 it is vital to quantitatively understand the relationship between Ea, the bottleneck size/area, and the resultant σion of the Nasicon electrolytes. As addressed in the Introduction section, the standard monoclinic structure of Nasicon has three inequivalent crystallographic Na sites per formula unit, one Na1, one Na2, and two Na3 sites. The diffusion pathways for Na ions are formed between Na1 and either Na2 or Na3 sites as shown in Figure 3.14 Therefore, four distinguishable bottleneck areas (two in the Na1-Na2 and two in the Na1-Na3 channels) are formed. In Figure 3, the sky blue pockets, the light purple pockets, the yellow pockets, the red atoms, and the dark purple atoms represent ZrO 6 octahedra, SiO 4 tetrahedra, PO 4 tetrahedra, O atoms, and Na atoms, respectively. Because a higher content of secondary phases and a change in the cell volume/lattice parameter for the Na-excess sample (see Figure 2a and Table 1) could affect the bottleneck sizes, we estimated the bottleneck sizes in the two samples; these bottleneck areas were computed based on the O−O bond lengths of the corresponding triangles obtained from XRD refinement analysis (Tables S1 and S2).15,29 In Table 2, we summarized the experimentally calculated results of the bottleneck areas from the bare and Na-excess samples. Here, the bottleneck regions are designated by the symbols “A−D” shown in Figure 3. The bottleneck areas “A, B” and “C, D” correspond to the triangles positioned along the Na1-Na2 and the Na1-Na3 channels, respectively. The centers of bottlenecks “A” and “C” triangles

bottleneck area (Å2) Na1-Na2 pathway bare Na-excess

Na1-Na3 pathway

A

B

C

D

6.2719 6.0773

5.0648 5.2007

5.9995 5.7602

5.6488 5.7338

are located closer to the original Na1 site, whereas the centers of bottlenecks “B” and “D” triangles are located closer to the Na2 and Na3 sites, respectively. Because Ea would be primarily determined by the smallest bottleneck,15,29 it appears that the bottleneck “B” in the Na1-Na2 pathway is likely the limiting factor for conduction of Na-ions. Note that bottleneck “B” in the Na-excess sample is larger (∼2.68% increase) than that in the bare sample, which can expedite the Na ion transport in the conduction channel by reducing Ea of the Na-excess sample. In the standard monoclinic Nasicon structure (Na3Zr2Si2PO12) with 75% Na site occupancy in a formula unit, there are four missing Na sites in the 4 formula units. Because it has been reported that Na1 sites are thermodynamically more stable and are likely to be completely filled and that other Na atoms are randomly distributed over the available Na2 and Na3 sites,30 five distinguishable atomistic structures are in theory probably depending on the relative occupancy for the Na2 and Na3 sites (i.e., 0%, 25%, 50%, 75%, and 100% occupancy for the Na2 sites) in the bare sample DFT models with 4 formula units. Figure 4a−c illustrates the atomistic structures of these probable standard monoclinic bare sample models in the DFT computations. Here, three candidate structures are shown, (a) 0Na2-8Na3 (0% Na2 site occupancy), (b) 2Na2-6Na3 (50% Na2 site occupancy), and (c) 4Na2-4Na3 (100% Na2 site occupancy) per 4 formula units. In the figure, the same color scheme as in Figure 3 to represent the oxide pockets and Na and O atoms was used. On the basis of these models, we first calculated the lattice parameters, unit cell volumes, and the formation energies of these structures using the ab initio DFT computation technique. The calculated values based on these distinct structures are listed in Table 3. The formation energies are provided on the basis of 1 formula unit. As seen in the table, the formation energies of the three candidate structures are all fairly negative (−262.7 to −263.3 kJ/mol) indicating that they are all thermodynamically stable with reference to the completely separated structure and, 27817

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

Figure 4. Atomistic structures of the bare (a) 0Na2-8Na3, (b) 2Na2-6Na3, (c) 4Na2-4Na3 and the Na-excess (d) 4Na2-6Na3, (e) 2Na2-8Na3 sample models employed in the DFT computations.

Table 3. Unit Cell Parameters for the Monoclinic Bare and the Na-Excess Nasicon Systems Obtained by DFT Calculations lattice parameter (Å) bare

Na-excess

0Na2-8Na3 2Na2-6Na3 4Na2-4Na3 4Na2-6Na3 2Na2-8Na3

a

b

c

β (deg)

volume (Å3)

formation energy (eV/unit)

15.8800 15.6921 15.9348 16.1277 16.0905

9.2662 9.3519 9.1995 9.3355 9.3279

9.2878 9.2044 9.2582 9.2052 9.2330

125.3640 122.8539 124.9868 125.5903 125.9257

1114.5114 1134.7102 1111.9160 1127.0427 1122.1780

−54.4791 −54.6063 −54.5286 −54.8032 −54.7711

therefore, they could be all observable in the experimental structures. Of these models, the formation energy of 2Na2-6Na3 with 50% Na2 site occupancy was the lowest manifesting the most stable atomistic configuration. For Na-excess sample models, as described before, we added two extra Na atoms to the standard monoclinic structures assuming Na3.5Zr2Si2P1O12 stoichiometry (16.67 at% excess Na) instead of Na3.3Zr2Si2PO12 (10 at% excess Na) stoichiometry due to the finite integer numbers in the atomistic configuration. The construction of these Na-excess models was realized by adding two extra Na atoms at the available Na2 or Na3 sites. Then, there are two candidate Na-excess models that are feasible, 4Na2-6Na3 (100% Na2 site occupancy) and 2Na2-8Na3 (50% Na2 site occupancy). These model structures are visualized in Figure 4d and Figure 4e, and the computationally calculated results for lattice parameters, unit cell volumes, and formation energies are summarized in Table 3. The formation energies of Na-excess models were predicted as slightly lower than the bare model (from −264.1 to −264.3 kJ/mol), indicating that adding extra Na atoms would decrease or at least would not increase the

thermodynamic stability of the Nasicon compounds. When the features of the calculated structures are compared with the experimental characterizations, they are in general comparable, but the unit cell volumes are slightly larger; the average cell volumes of the bare and the Na-excess models calculated by DFT computations are 1120.3792 and 1124.6104 Å3, which are 3.2% and 3.4% larger than the experimental observation, respectively. If the Na2 and Na3 sites are randomly filled as reported in ref 30, these average unit cell volumes would represent the volumes of experimentally synthesized structures, and the computed results appear to reproduce the general trend of XRD analysis found in Table 1; by adding excess Na, the unit cell volume slightly increases. On the basis of the geometry optimized structures using DFT computations, the bottleneck areas were then calculated as shown in Table 4. The bottleneck area values have been averaged from all available Na transport channels in the model structures. For instance, for 0Na2-8Na3 model structure, there are 4 and 0 possible Na1-Na2 and Na1-Na3 pathways, respectively, in the 4 formula unit cell structures. In the table, 27818

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

the bottleneck areas of “A, C, and D” are also in good agreement with the XRD refinement analysis results; the bottleneck sizes of “A”, “C”, and “D” are decreased, decreased, and increased, respectively, as more Na are added to the monoclinic Nasicon structure. Therefore, it is again ensured that the limiting factor of the Na transport could come from the smaller size of bottleneck “B” that is closer to the Na2 site in the moving pathway of Na1-Na2 channel, and this bottleneck area is increased by supplying extra Na atoms especially when the atomic configurations are altered from 2Na2-6Na3 to 2Na28Na3 in the 4 formula unit model. Electrochemical Performances. In Figure 5, we show the Nyquist plots from EIS (electrochemical impedance spectroscopy) for the (a) bare and (b) Na-excess samples. In a Nyquist plot, the equivalent circuits of polycrystalline materials are typically constructed by a series of parallel resistor-capacitor (RC) elements for grain resistance (RG), grain boundary resistance (RGB) of polycrystals, and a capacitor representing the blocking electrodes. Because the signals of RG and RGB will appear at high and low frequencies (f), respectively, the Nyquist plot of a polycrystalline sample routinely shows two distinct semicircles due to the contribution from two RC circuits, i.e., grain and grain boundary. However, in the current study, both of the two samples displayed only one semicircle at room temperature as seen in Figure 5a and Figure 5b. This is because the Nasicon polycrystalline grain has an extremely low capacitance at room temperature that requires very high f (≫1

Table 4. Summary of the Bottleneck Sizes Calculated by DFT Computations for the Bare and the Na-Excess Samples bottleneck area (Å2) Na1-Na2 pathway bare

Na-excess

0Na2-8Na3 2Na2-6Na3 4Na2-4Na3 4Na2-6Na3 2Na2-8Na3

Na1-Na3 pathway

A

B

A

B

5.8525 6.5747 N/A N/A 5.7776

6.0567 4.9922 N/A N/A 5.4086

N/A 6.2688 5.8084 5.5879 N/A

N/A 6.0148 6.1469 6.3355 N/A

the Na ion transport channel designations “A−D” are identical to the ones given in Figure 3 and Table 2. From the analysis of the computational results, the smallest bottleneck area was identified as the “B” route in the Na1-Na2 pathway of 2Na26Na3 structure. When excess Na atoms are added to the Na3 site in this structure (2Na2-8Na3), the area of the bottleneck “B” increases from 4.9922 to 5.4086 Å2. Therefore, it is likely that the role of this 2Na2-6Na3 structure is important because it can provide both of the diffusion channels, Na1-Na2 and Na1Na3. Although the amounts of excess Na are different for the Nasicon electrolyte structures of computational models (i.e., 16.67 at% excess Na) and experimental samples (i.e., 10 at% excess Na), the prediction results using ab initio DFT computations are consistent with the experimental measurement data. Additionally, other general trends for the changes in

Figure 5. Nyquist plots at various temperatures measured from the (a) bare and (b) Na-excess samples, and (c) Nyquist and (d) Bode plots at −40 °C measured from the bare sample. 27819

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

Figure 6. Measured (solid symbols) conductivity at low and extrapolated (open symbols) conductivity at high temperatures for the (a) bare and (b) Na-excess samples.

method from the measured σG and σGB values at −40 to −10 °C. Table 5 lists the extracted Ea values along with the pre-

MHz) for RG.31 When the total resistance (Rtot) is compared, Rtot for the Na-excess sample was lower than that for the bare sample in the entire temperature range tested. Additionally, it was observed that in both samples, the semicircle in the Nyquist plot disappears as the testing temperature increases, indicating that the contribution from RGB to Rtot is reduced with increasing temperature. Because a quantitative analysis with regard to the individual contributions from RG and RGB at higher temperatures (>100 °C, the green and blue symbols in Figure 5a and Figure 5b) is not possible, the samples have been subject to the EIS measurement at low temperatures (i.e., lower than room temperature); in this low temperature regime, the resonance frequencies of grain and grain boundary would present low but distinct values. Figure 5 also provides the (c) Nyquist and (d) Bode plots obtained from the bare sample at −40 °C. The numeric numbers in Figure 5c indicate log(f). As shown in Figure 5c, two semicircles with tails were observed at −40 °C in the Nyquist plot. Also, from the phase (φ) plot in Figure 5d, it is clear that the RC circuits that correspond to RG and RGB appear in the frequency ranges of ∼105−106 Hz and ∼102−105 Hz, respectively. After confirmation that the low temperature impedance test (−40 °C) can reveal the difference between the grain and the grain boundary contributions, the EIS measurements were carried out at the range of the temperatures from −40 to −10 °C to separate out the grain and grain boundary contributions to the total σion of samples. Conductivity values from grain (σG) and grain boundary (σGB) in both samples at these low temperatures were quantitatively evaluated based on the measured impedance and the geometrical factors including the diameter and the thickness of pellets, and they are given using the solid symbols in Figure 6. From the figure, one can easily notice that the dominant factor affecting the total conductivity at low temperatures is σGB rather than σG in both samples. Furthermore, it was found that the Na-excess sample exhibits higher total σion compared with the bare sample. If we assume that σG and σGB follow the classical Arrheniustype behavior (σ =

( σT ) exp(− kTE ), 0

a

Table 5. List of Activation Energies and Pre-Exponential Factors in the Arrhenius Expression for the Bare and NaExcess Samples bare

grain grain boundary

Ea (eV) 0.28 0.39

Na-excess

ln(σ0) (S K cm−1)

Ea (eV)

ln(σ0) (S K cm−1)

9.333 13.222

0.24 0.43

8.321 16.330

exponential factor (σ0) from σG and σGB variations at low temperatures. From the table, in both samples, a larger Ea was identified for σGB compared with σG. Using the σ0 and Ea values for each case given in Table 5, we then extrapolated σG and σGB extending the temperatures above room temperature. These extrapolated σG and σGB values are also presented in Figure 6 with open symbols for the (a) bare and (b) Na-excess samples. From the figure, it is clearly seen that σG (lower Ea) increases more slowly than σGB (higher Ea) with increasing temperature for both samples, implying that the primary factor determining total σion shifts from σGB at low temperature to σG at high temperature. When this switching temperature is examined, it deems that the σG contribution becomes predominant beyond the temperature of ∼100 °C. Therefore, one can readily see that maximization of σG is vital to increase total σion of Nasicon electrolyte above ∼100 °C, which covers the nominal operation temperature (above ∼270 °C) of Na−S or Na−metal halide battery systems. As σ0 contains charge carrier concentration Nq2 k

( )n (1 − n )l v , where γ is the proportion-

term (σ0 = γ

2

v

v

0

ality constant that depends on the number of dimensions where the charge carrier can move, N is the number of normal sites per unit volume, q is charge of charge carriers, nv is the fraction of vacant sites, l is the hopping distance, v0 is a vibrational frequency to overcome the potential barrier, and nv is charge carrier concentration), the σ0 values listed in Table 5 could be correlated with the intrinsic charge carrier concentration. As given in the table, σ0 of grain boundary was much greater than that of grain for both samples, signifying that the intrinsic

where σ0 is the pre-

exponential intrinsic conductivity factor independent of temperature, k is the Boltzmann constant, and T is the prescribed temperature, respectively) with a constant activation energy (Ea), Ea can be extracted using a linear regression 27820

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

Figure 7. Extrapolated and measured total ionic conductivity for the (a) bare and (b) Na-excess samples and (c) comparison of the total ionic conductivity for the bare and the Na-excess samples measured in the current study with the previous studies.

contribution independent of temperature in σGB is larger than that in σG. Additionally, it was found that the Na-excess sample showed higher σ0 for σGB compared with the bare sample, which might be attributed to the presence of excess Na. This tells that some of the added extra Na would take the position in the grain boundary area of the Na-excess sample. Further, it is thought that the excess Na might cause the formation of minor Na3PO4 impurity detected in the XRD scan near the grain boundary area of polycrystalline Nasicon. The Na-excess sample will then contain excess Si relative to the standard stoichiometric Na3Zr2Si2PO12 structure. Such excess Si can lead to the increase of the unit cell volume (as shown in Table 1) because the bond strength of Si4+−O2− bonds are weaker than that of P5+−O2−32 with a greater bottleneck area and a lower Ea value. In contrast to the grain boundary case, σ0 for the Naexcess sample is lower than that for the bare sample in the grain. This is because, although the absolute amount of Na in the polycrystalline grain of the Na-excess sample is increased by adding 10 at% Na, the “effective” Na-ion concentration can be reduced as the number of vacancy sites available for Na-ion conduction is decreased. Note that for the standard Na3Zr2Si2PO12 structure, the Na occupancy is 75% (i.e., 25% available vacancy sties), but for the Na-excess sample, the Na occupancy can be increased as much as 82.5% (i.e., 17.5% available vacancy sites), if all of the excess Na is positioned in

the interior region of the polycrystalline grains. It is unlikely that all of the added Na atoms would occupy the Na sites inside the grain of polycrystals as discussed before, and some of the added Na atoms may have contributed to form impurity phases; however, at least a significant portion of the extra Na would fill the Na vacancy sites of the original standard structure. Therefore, the added Na will locate partly in the grain boundary and partly in the grain structure serving to increase σ0 by increasing the absolute charge carrier concentrations and to decrease σ0 by decreasing the effective charge carrier concentrations in the respective areas. In Figure 6, we show measured (solid symbols) along with extrapolated (open symbols) σion at low and high temperatures, respectively, for the (a) bare and (b) Na-excess samples. When Figure 6a and Figure 6b are contrasted with each other, the Naexcess sample showed increased conductivity in both aspects, i.e., σG and σGB. As addressed previously, in the Na-excess sample, the increase of σG resulted from the decrease of Ea (0.28−0.24 eV) with a larger bottleneck size despite lower σ0, whereas the increase of σGB can come from higher σ0 in the grain boundary area. The results presented in Figure 6a and Figure 6b clearly confirm that the impacts of a larger bottleneck size are much more dominant than those of smaller σ0 in the Na-excess sample at high temperature. As shown in Table 2, the smallest bottleneck area related to the Na1-Na2 site diffusion 27821

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces

conditions. However, total σion of the Na-excess sample is distinctly higher. When the operation temperatures of molten Na battery are taken into consideration (over ∼270 °C), the total σion value for the Na-excess sample synthesized in the present work can reach ∼1.2 × 10−1 S/cm, which could be viewed as one of the best conductive solid electrolyte compounds for Na−S or Na−metal halide applications at this temperature.9,36 Finally, assuming that the pre-exponential factor (σ0 in Table 5) is identical for both monoclinic and rhombohedral structures and that Rtot is determined only by RG at high temperature where the rhombohedral phase is stable, as provided in Table 6,

was enlarged from 5.0648 to 5.2007 Å2 by adding 10 at% excess Na. Previously, it was reported that the area of the smallest bottleneck is the most critical factor that determines Ea and the conductivity of Nasicon electrolytes and that a decrease of bottleneck area by ∼0.1 Å2 (∼2% decrease) corresponds to an increase of Ea by ∼0.05 eV.29 When 20 at% excess Na is added, the trend of reducing Ea and σ0 of grain is consistent (see Table S4 in Supporting Information), which clearly confirms the role of excess Na. Therefore, considering that the total σion is approximated only from σG at high temperature (higher than ∼100 °C), the primary factor to increase the total conductivity of the Na-excess sample above ∼100 °C is likely the enlargement of the smallest bottleneck size and the consequential decrease of Ea in the polycrystalline Nasicon grains. This finding is particularly important for the Na ion transport behaviors of Nasicon electrolyte at high temperature because typical molten Na batteries are operated over ∼270 °C. To confirm the adequacy of the extrapolated ionic conductivities at high temperatures shown in Figure 6, σion of the two samples was directly measured at various temperatures using EIS. The results of the σion measurements are included in Figure 7 for the (a) bare and (b) Na-excess samples. As shown in the results, the difference between the σion values obtained from extrapolation and direct measurements is not significant at the temperature ranges below ∼150 °C. However, where the temperature is greater than ∼150 °C, extrapolated σion exhibits relatively larger deviation from the experimental value. Such deviation could be associated with the phase transition behavior of Nasicon near ∼150 °C; although the monoclinic− rhombohedral phase transition will take place in a continuous fashion with its second order nature, when the temperature is higher than ∼150 °C, it is likely that the samples are mostly transformed to rhombohedral phase possibly with a lower Ea.6,11,33 However, for extrapolated σion, a constant value of Ea was assumed based on the monoclinic structure of Nasicon in the entire range of high temperatures. This can explain the underestimation of extrapolated σion depicted in Figure 7 for both samples. Note that although the measured σion shows some positive deviations from extrapolated σion, it is postulated that the higher σion of the Na-excess sample in the rhombohedral phase provided in Figure 7b would result from the enlargement of bottleneck area, as the concentrations of the ionic charge carrier are identical in both phases. In Figure 7c, the results of total σion of the bare and Naexcess samples above 100 °C are compared with reference to the previously reported data.34,35 The measured total σion in the current work clearly shows that adding excess Na promotes the ion conduction over the temperature range above 100 °C. The increase of σion was also observed in the 20 at% Na-excess sample with respect to σion of the bare sample; however, σion of the 20 at% Na-excess sample was smaller than σion of the 10 at% Na-excess sample (see Figure S4b in Supporting Information). With incorporating 10 at% excess Na, the relative increase of the total σion was found to increase by approximately 183% and 92% at 100 and 300 °C, respectively. Because the monoclinic− rhombohedral phase transition of Nasicon structure continuously occurs in the temperature range of approximately 100− 200 °C, total σion of the Na-excess sample will present higher values than that of the bare sample both for the monoclinic and the rhombohedral crystal structures. In the figure, compared with the prior literature values, total σion of the bare sample in the current work is positioned between the upper and the lower bounds, probably due to the different sample synthesis

Table 6. Calculated Activation Energy (Ea) Values for the Bare and the Na-Excess Nasicon Systems with Monoclinic and Rhombohedral Structures bare Na-excess

Ea, monoclinic (eV)

Ea, rhombohedral (eV)

0.28 0.24

0.21 0.18

the Ea values for the Na transport through grains in the Naexcess samples have been again calculated based on the direct measurement data applying a linear regression analysis. From Table 6, it is evident that for both samples Ea decreases at the high temperature ranges (rhombohedral structure) because the rhombohedral phase will present lower Ea primarily due to its higher structural symmetry.6,11,33 Additionally, compared with the bare sample, the Na-excess sample at this rhombohedral phase temperature shows higher conductivity, likely because of a higher degree of reduction in Ea (from 0.24 to 0.18 eV). From these results, therefore, it could be again thought that the kinetic limiting factor of the smallest bottleneck size in the rhombohedral phase is likely to be much increased by incorporating excess Na in the Nasicon system. The amounts of impurity phases such as ZrO2 and Na3PO4 are increased by adding excess Na in the Nasicon structure; however, these impurity phases are poor Na conductors; for example, the monoclinic ZrO2 is an oxygen ionic conductor but its oxygen ionic conductivity is extremely low (∼2.0 × 10−6 S/cm at 600 °C with an Ea of ∼0.40 eV),37 and the Na3PO4 phase is a Na ionic conductor but its σion is also very low, reported as 10−3− 10−5 S/cm at 100−300 °C.38,39 Furthermore, we showed that the improvement of σion is mainly derived from the grain transport rather than the grain boundary transport at high temperatures. Considering that these impurity phases typically affect the conduction behavior through grain boundary, therefore, the measured σion values for the Na-excess sample especially at the molten Na battery operation temperature would not be strongly influenced by the presence of such impurity phases in the Nasicon electrolytes.

4. CONCLUSION In the present work, we studied the impacts of adding excess Na to the standard Nasicon ionic conductor structure along with the Na ion transport mechanisms. The bare and Na-excess Nasicon electrolyte samples were fabricated using a simple solid-state processing route. After sample fabrication, the structural and the Na ion transport properties have been systematically examined using various characterization tools including XRD, DSC, SEM, and EIS. The following lists the important findings attained through the current study. 27822

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces



• By systematically separating out the grain (σG) and the grain boundary (σGB) contributions to the total ionic conductivity (σion) based on the quantification of the intrinsic conductivity (σ0) and the activation energy (Ea) for Na conduction at low temperatures, we found that the dominant factor determining σion of the Nasicon structure is switched from σGB at low temperature (100 °C). • Incorporation of excess Na in the Nasicon system can significantly improve σG by increasing the smallest bottleneck size in the Na conduction channels at both low and high temperatures. σion value measured from the Na-excess sample at 300 °C in the current work was recorded as ∼1.2 × 10−1 S/cm, which is approximately 92% higher than that from the bare sample. We demonstrated that adding excess Na could alter the bulk structures of synthesized Nasicon; for example, in the monoclinic Na-excess sample, the smallest bottleneck area (bottleneck “B” in the Na1-Na2 channel) was increased by ∼2.7%, which could be substantial for more expedited movement of conducting Na ions. These experimental results for structural changes have been confirmed by the ab initio DFT computations, which also predicts the smallest bottleneck (in Na1-Na2 diffusion channels) size increase with added excess Na. • The effects of Na ion concentration on σion were also investigated by analyzing σG and σGB contributions at low temperatures. Considering σ0 in the grain, it was found that excess Na may adversely affect the conductivity by reducing the effective charge carrier concentration in the grain. On the other hand, it was experimentally and computationally shown that excess Na can increase the mobility of Na-ion by increasing the smallest bottleneck size. Therefore, the significant improvement of conduction property attained from the Na-excess sample is a consequence of the increased mobility of Na ions rather than the increased charge carrier concentration in the Nasicon crystals.

Energy Technology Evaluation and Planning (KETEP), financial resource from the Ministry of Trade, Industry & Energy, Republic of Korea (Grants 20128510010070 and 20158510050010), and Center of Futuristic Material-Systems of Brain Korea 21 Project (Grant F14SN02D1707).



(1) Hueso, K. B.; Armand, M.; Rojo, T. High Temperature Sodium Batteries: Status, Challenges and Future Trends. Energy Environ. Sci. 2013, 6 (3), 734−749. (2) Li, G.; Lu, X.; Kim, J. Y.; Meinhardt, K. D.; Chang, H. J.; Canfield, N. L.; Sprenkle, V. L. Advanced Intermediate Temperature Sodium-Nickel Chloride Batteries with Ultra-High Energy Density. Nat. Commun. 2016, 7, 10683. (3) Jung, K.; Lee, S.; Park, Y.-C.; Kim, C.-S. Finite Element Analysis Study on the Thermomechanical Stability of Thermal Compression Bonding (TCB) Joints in Tubular Sodium Sulfur Cells. J. Power Sources 2014, 250, 1−14. (4) Jung, K.; Lee, S.; Kim, G.; Kim, C.-S. Stress Analyses for the Glass Joints of Contemporary Sodium Sulfur Batteries. J. Power Sources 2014, 269, 773−782. (5) Sun, T.; Li, Z.-j.; Wang, H.-g.; Bao, D.; Meng, F.-l.; Zhang, X.-b. A Biodegradable Polydopamine-Derived Electrode Material for HighCapacity and Long-Life Lithium-Ion and Sodium-Ion Batteries. Angew. Chem. 2016, 128 (36), 10820−10824. (6) Lalère, F.; Leriche, J. B.; Courty, M.; Boulineau, S.; Viallet, V.; Masquelier, C.; Seznec, V. An All-Solid State NASICON Sodium Battery Operating at 200 °C. J. Power Sources 2014, 247, 975−980. (7) Kim, J.; Jo, S. H.; Bhavaraju, S.; Eccleston, A.; Kang, S. O. Low Temperature Performance of Sodium−Nickel Chloride Batteries with NaSICON Solid Electrolyte. J. Electroanal. Chem. 2015, 759, 201−206. (8) Noguchi, Y.; Kobayashi, E.; Plashnitsa, L. S.; Okada, S.; Yamaki, J.-i. Fabrication and Performances of All Solid-State Symmetric Sodium Battery Based on NASICON-Related Compounds. Electrochim. Acta 2013, 101, 59−65. (9) Lu, X.; Xia, G.; Lemmon, J. P.; Yang, Z. Advanced Materials for Sodium-Beta Alumina Batteries: Status, Challenges and Perspectives. J. Power Sources 2010, 195 (9), 2431−2442. (10) Yde-Andersen, S.; Lundsgaard, J. S.; Møller, L.; Engell, J. Properties of Nasicon Electrolytes Prepared from Alkoxide Derived Gels: Ionic Conductivity, Durability in Molten Sodium and Strength Test Data. Solid State Ionics 1984, 14 (1), 73−79. (11) von Alpen, U.; Bell, M. F.; Wichelhaus, W. Phase Transition in Nasicon (Na3Zr2Si2PO12). Mater. Res. Bull. 1979, 14 (10), 1317−1322. (12) Goodenough, J. B.; Hong, H. Y.-P.; Kafalas, J. A. Fast Na+-Ion Transport in Skeleton Structures. Mater. Res. Bull. 1976, 11 (2), 203− 220. (13) Ignaszak, A.; Pasierb, P.; Gajerski, R.; Komornicki, S. Synthesis and Properties of Nasicon-Type Materials. Thermochim. Acta 2005, 426 (1−2), 7−14. (14) Boilot, J. P.; Collin, G.; Colomban, P. Crystal Structure of the True Nasicon: Na3Zr2Si2PO12. Mater. Res. Bull. 1987, 22 (5), 669− 676. (15) Guin, M.; Tietz, F. Survey of the Transport Properties of Sodium Superionic Conductor Materials for Use in Sodium Batteries. J. Power Sources 2015, 273, 1056−1064. (16) Lee, J.-S.; Chang, C.-M.; Lee, Y. I.; Lee, J.-H.; Hong, S.-H. Spark Plasma Sintering (SPS) of NASICON Ceramics. J. Am. Ceram. Soc. 2004, 87 (2), 305−307. (17) Shaw, T. L.; Carrol, J. C. Application of Baseline Correction Techniques to the “Ratio Method” of DSC Specific Heat Determination. Int. J. Thermophys. 1998, 19 (6), 1671−1680. (18) Delley, B. An All-Electron Numerical Method for Solving the Local Density Functional for Polyatomic Molecules. J. Chem. Phys. 1990, 92 (1), 508−517. (19) Delley, B. From Molecules to Solids with the DMol[sup 3] Approach. J. Chem. Phys. 2000, 113 (18), 7756−7764.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.6b09992. Results of Rietveld refinement of XRD patterns of the bare and Na-excess Nasicon samples (PDF)



REFERENCES

AUTHOR INFORMATION

Corresponding Authors

*C.-S.K.: phone, +1-414-229-3085; e-mail, [email protected]. *B.K.: phone, +82 54 279 2154; e-mail, [email protected]. Author Contributions

H. Park carried out most of the experiments, and M. Nezafati performed DFT calculations. K. Jung discussed the experiments and data analysis parts. H. Park, C.-S. Kim, and B. Kang cowrote the paper. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by the International Collaborative Energy Technology R&D Program of the Korea Institute of 27823

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824

Research Article

ACS Applied Materials & Interfaces (20) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868. (21) Gordon, R. S.; Miller, G. R.; McEntire, B. J.; Beck, E. D.; Rasmussen, J. R. Fabrication and Characterization of Nasicon Electrolytes. Solid State Ionics 1981, 3−4, 243−248. (22) Krok, F. Influence of Sintering Conditions on Chemical Composition of NASICON. Solid State Ionics 1987, 24 (1), 21−28. (23) Fuentes, R. O.; Figueiredo, F. M.; Marques, F. M. B.; Franco, J. I. Processing and Electrical Properties of NASICON Prepared from Yttria-Doped Zirconia Precursors. J. Eur. Ceram. Soc. 2001, 21 (6), 737−743. (24) Feist, T.; Davies, P. K.; Vogel, E. The Energetics of Phase Transitions in the System Na1+XZr2SiXP3‑XO12, 1.9 ≤ X ≤ 2.5. Thermochim. Acta 1986, 106 (15), 57−61. (25) Boilot, J.-P.; Colomban, P. Nasicon: Amorphous to Crystalline Compound. Solid State Ionics 1986, 18, 974−980. (26) Gasmi, N.; Gharbi, N.; Zarrouk, H.; Barboux, P.; Morineau, R.; Livage, J. Comparison of Different Synthesis Methods for Nasicon Ceramics. J. Sol-Gel Sci. Technol. 1995, 4 (3), 231−237. (27) Kumar, P. P.; Yashonath, S. Structure, Conductivity, and Ionic Motion in Na1+xZr2SixP3‑xO12: A Simulation Study. J. Phys. Chem. B 2002, 106, 7081−7089. (28) Kohler, H.; Schulz, H. NASICON Solid Electrolytes Part II - XRay Diffraction Experiments on Sodium-Zirconium-Phosphate Single Crystals at 295K and at 993K. Mater. Res. Bull. 1986, 21 (1), 23−31. (29) Losilla, E. R.; Aranda, M. A. G.; Bruque, S.; París, M. A.; Sanz, J.; West, A. R. Understanding Na Mobility in NASICON Materials: A Rietveld, 23Na and 31P MAS NMR, and Impedance Study. Chem. Mater. 1998, 10 (2), 665−673. (30) Hong, H. Y.-P. Crystal Structures and Crystal Chemistry in the System Na1+xZr2SixP3−xO12. Mater. Res. Bull. 1976, 11 (2), 173−182. (31) Bayard, M. L.; Barna, G. G. A Complex Impedance Analysis of the Ionic Conductivity of Na1+xZr2SixP3−xO12 Ceramics. J. Electroanal. Chem. Interfacial Electrochem. 1978, 91 (2), 201−209. (32) Boilot, J. P.; Collin, G.; Colomban, P. Relation Structure-Fast Ion Conduction in the NASICON Solid Solution. J. Solid State Chem. 1988, 73 (1), 160−171. (33) Colomban, P. Orientational Disorder, Glass/Crystal Transition and Superionic Conductivity in Nasicon. Solid State Ionics 1986, 21 (2), 97−115. (34) Bohnke, O.; Ronchetti, S.; Mazza, D. Conductivity Measurements on Nasicon and Nasicon-Modified Materials. Solid State Ionics 1999, 122 (1−4), 127−136. (35) Hayashi, K.; Shima, K.; Sugiyama, F. A Mixed Aqueous/Aprotic Sodium/Air Cell Using a NASICON Ceramic Separator. J. Electrochem. Soc. 2013, 160 (9), A1467−A1472. (36) Hayashi, A.; Noi, K.; Sakuda, A.; Tatsumisago, M. Superionic Glass-Ceramic Electrolytes for Room-Temperature Rechargeable Sodium Batteries. Nat. Commun. 2012, 3, 856. (37) Nasrallah, M. M.; Douglass, D. L. Ionic and Electronic Conductivity in Y2O3-Doped Monoclinic ZrO2. J. Electrochem. Soc. 1974, 121 (2), 255−262. (38) Hruschka, H.; Lissel, E.; Jansen, M. Na-Ion Conduction in the Solid Solutions of Na3PO4/Na2SO4 and Na3AlF6/Na2SO4. Solid State Ionics 1988, 28−30, 159−162. (39) Ghule, A.; Baskaran, N.; Murugan, R.; Chang, H. Phase Transformation Studies of Na3PO4 by Thermo-Raman and Conductivity Measurements. Solid State Ionics 2003, 161 (3−4), 291−299.

27824

DOI: 10.1021/acsami.6b09992 ACS Appl. Mater. Interfaces 2016, 8, 27814−27824