Scholarly article on topic 'Monoclinic Sodium Iron Hexacyanoferrate Cathode and Non-Flammable Glyme-Based Electrolyte for Inexpensive Sodium-Ion Batteries'

Monoclinic Sodium Iron Hexacyanoferrate Cathode and Non-Flammable Glyme-Based Electrolyte for Inexpensive Sodium-Ion Batteries Academic research paper on "Chemical sciences"

Share paper
OECD Field of science

Academic research paper on topic "Monoclinic Sodium Iron Hexacyanoferrate Cathode and Non-Flammable Glyme-Based Electrolyte for Inexpensive Sodium-Ion Batteries"


Monoclinic Sodium Iron Hexacyanoferrate Cathode and Non-Flammable Glyme-Based Electrolyte for Inexpensive Sodium-Ion Batteries

Ashish Rudola,a b Kang Du,a and Palani Balaya®' *'z

aDepartment of Mechanical Engineering, National University of Singapore, 117576, Singapore b Department of Materials Science and Engineering, National University of Singapore, 117575, Singapore

One of the key requirements of large-scale grid-storage systems is development of inexpensive and safe batteries. Sodium-ion batteries using earth-abundant Fe or Ti based cathodes and anodes would be ideal candidates for such storage systems. Herein, a new phase of Na-rich and all Fe Prussian Blue Analogue, monoclinic Na2Fe2(CN)6.2H2O, is reported as a potential cathode for such grid-storage sodium-ion batteries. This water-insoluble and air-stable cathode can deliver 85 mAh g-1 at an average discharge voltage of 3 V vs Na/Na+ with excellent cycle life (3,000 cycles). Many facets about its sodium storage characteristics are discussed with particular emphasis on the role of interstitial water on the sodium storage performance and its conversion to the dehydrated rhombohedral phase. Its compatibility with a newly developed non-flammable glyme-based liquid electrolyte, 1M NaBF4 in tetraglyme, is also disclosed along with general electrochemical and thermal characterization of this electrolyte for sodium-ion battery application. Finally, three different types of full cells are revealed with either monoclinic or rhombohedral phase as cathode and graphite or the recently reported Na2Ti3O7 ^ Na3-xTi3O7 pathway of Na2Ti3O7 as anode. Full cell energy densities of 70-90 Wh kg-1 (using cumulative cathode and anode weights) could be obtained without any pre-cycling steps. This new cathode and safe electrolyte may hold great promise toward development of inexpensive, non-flammable and highly stable grid-storage sodium-ion batteries. © The Author(s) 2017. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY,, which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0701706jes] All rights reserved. | (pDI

Manuscript submitted January 20, 2017; revised manuscript received March 6, 2017. Published March 25, 2017.

One of the compelling requirements for mitigating CO2 emission is to generate our primary electricity from clean and renewable energy sources such as solar or wind. This action will have a greater impact in curbing global warming than widespread use of electric vehicles (EVs): if EVs are charged from electricity originating from power plants relying on fossil fuels, then the cumulative CO2 emission (upon considering emissions from the dirty power plants) will be higher than those from using traditional gasoline-powered vehicles.1 The urgent need of the hour, hence, is the mass adoption of renewable power plants. However, the intermittent energy generation from such renewable sources needs to be addressed if they are to ever become commercially viable. Large-scale electrochemical energy storage (EES) devices, or grid-storage batteries, are the most convenient and practically relevant means for handling this intermittent energy generation from such power plants.2,3 Combined costs of the power plant and storage systems would be the deciding factor for market penetration of renewable energy sources into our electricity grid. Apart from low cost, the other most desirable performance metrics for such EES batteries are long cycle life (a few thousand cycles), high efficiency and a high degree of safety.2,3 Gravimmetric and volumetric energy densities of such batteries are only a secondary factor, as the footprint and weight are not major concerns.

Sodium-ion batteries (NIBs) are very attractive for such large-scale EES applications owing to the expected cost reductions arising from the globally abundant Na resources.4,5 However, to truly reduce costs further, it is imperative that cathodes and anodes in NIBs targeted for grid-storage batteries also use other earth-abundant elemental resources, such as Fe, Mn and Ti.1,6 Costs can further be decreased if the electrode materials display air and water stability, scalable syntheses are employed preferably in water medium without the need for additional calcinations/firing steps at high temperatures, environmentally safe, non-toxic and inexpensive chemicals are used throughout the production process and sodium storage occurs in the battery with good thermal and chemical stability at different charged/discharged states at room temperature. These factors would ultimately reduce the costs associated with battery manufacturing, production, maintenance and management processes which can significantly affect

* Electrochemical Society Member. zE-mail:

battery costs over its lifetime.1,6,7 However, these requirements are quite stringent. Despite such harsh demands, there have been a few promising NIB electrode materials reported which meet most of the above requirements for grid-storage batteries.8-22 Among them, there is a class of cathodes belonging to the Prussian Blue Analogue (PBA) family which is very appealing due to its reliance on Fe and/or Mn as the redox active centers and possession of high sodium storage capacities (theoretical capacity limit as high as 170.8 mAh g-1 assuming two mole sodium storage per mole of material) at relatively high voltages.23 The general formula for PBAs relevant for NIBs is NaxM1[M2(CN)6]1-yDy.nH2O with 0 < x < 2and0 < y < 1. Here, M1 and M2 are transition-metal ions strongly covalently bridged by cyano (C=N-) ligands with M1 octahedrally coordinated to N, M2 octahedrally coordinated to C and □ refers to vacancies which may arise during the synthesis. The crystal structure of these compounds consists generally of either perfect or slightly distorted cubes with M1 and M2 situated at the corners, bridged by the cyano ligands. This arrangement leaves eight sub-cube units within each unit cell where alkali ions such as Na+ and/or interstitial water molecules may reside.24-27 Most of the PBAs which have been reported to store sodium have less Na content in the as-synthesized compounds (0 < x < 1) demonstrating a cubic structure with a space group Fm-3m.28-33 This may become a disadvantage in a full cell formulation against a suitable anode as a full cell relies on the cathode supplying Na during the first charging (sodium extraction from cathode and insertion into the anode). An x = 1 value would mean that the practically achieved capacity of such a cathode would be half of its theoretical capacity (corresponding to the capacity for an analogous cathode with x = 2) in a practical full cell assuming no loss of Na due to surface passivation and 100% coulombic efficiency for both cathode and anode. Recently, some papers have reported successful synthesis of Na and Fe containing PBAs with slightly richer Na content (1.5 < x < 1.6). Interestingly, for these PBAs, the cubic structure is still maintained.34-37 Till date, there have been only two reports which have displayed a value of 1.6 < x < 2 in their general formula NaxM1M2(CN)6.nH2O, particularly with both M1 and M2 as Fe. Guo et al. synthesized a Na richer Na163Fe189(CN)6 with arhombohedral crystal structure capable of delivering 153 mAh g-1 during its first charge (Na extraction process).38 However, the authors did not provide details about the space group or the water content in the structure of Na163Fe189(CN)6. Goodenough et al. recently reported three variations of Na rich PBAs- monoclinic

M-Na2.8Mn[Fe(CN)6].1.87H2O, rhombohedral R-Na2-8Mn[Fe(CN)6] and rhombohedral R-Na192Fe[Fe(CN)6]- demonstrating high capacity of 150-160 mAh g-1 at an attractive voltage of 3-3.6 V vs Na/Na+.39-41 They demonstrated the critical effect of interstitial water in influencing the structure of the Mn-Fe PBA; water expulsion was shown to raise the symmetry from monoclinic P21/n space group to rhombohedral R3 along with a flattening of the voltage profiles.40 Curiously, the monoclinic equivalent of an all Fe based PBA is not currently known.

In this contribution, we will reveal the existence of a Na rich all Fe analogue of M-Na2-8Mn[Fe(CN)6].1.87H2O. We will discuss many facets about the sodium storage characteristics of M-Na2Fe2(CN)6.2H2O, along with the effect of interstitial water on its electrochemical performance and its conversion to the corresponding rhombohedral phase. Keeping its ultimate utilization in large-scale EES in mind, we will argue why the air-stable and water-insoluble M-Na2Fe2(CN)6.2H2O cathode reported here could be very attractive for such applications. In this context, we will also reveal for the first time a non-flammable liquid electrolyte for NIB application which works very well for low voltage anodes as well as, crucially, for high voltage cathodes. Finally, we will demonstrate completely new full cells of the M-Na2Fe2(CN)6.2H2O and R-Na2Fe2(CN)6 cathodes paired with either graphite or the Na2Ti3O7 ^ Na3-xTi3O7 pathway as anodes with this non-flammable electrolyte resulting in inexpensive and stable NIBs suitable for large-scale grid-storage applications.


Material synthesis.—In a typical synthesis, 5 mmoles of Na4Fe(CN)6 and 22.5 mmoles of ascorbic acid were added to 100 ml of Milli-Q water in a round bottom flask. The flask was immersed in a silicone oil bath which was kept at 140°C. The solution was stirred for 4 hours while being refluxed such that the reflux temperature of the solution, measured by a thermometer dipped into the solution, was around 107°C (solution displayed vigorous bubbling throughout). The flask was then taken out of the oil bath and allowed to cool to room temperature, whereupon a white precipitate was obtained below a yellow coloured solution. The precipitate could be recovered by either centrifugation or filtration (the precipitate retrieval method did not alter phase purity). During this process, the white precipitate acquired a faint cyan tinge. The precipitate was then dried at 70°C in air for 3 h resulting in the final as-synthesized compound.

As-synthesized material characterization.—For Na:Fe molar ratio determination by inductively coupled plasma optical emission spec-troscopy (ICP-OES), a Perkin Elmer Optima 5300 DV instrument was used while for C and N measurements by CHN elemental analysis, an Elementar vario MICRO Cube elemental analyzer was used. For both ICP and CHN experiments, measurements were repeated and yielded consistent results. Water content was measured by thermogravimetric analysis, TGA (TA instrument; model 2960), where the measurements were obtained till 450°C in N2 atmosphere at 10°C/min ramp rate. TGA-MS (mass spectrometry) measurements were taken on a TGA-MSMettler-Toledo TGA/DSC coupled with Pifzer mass spectrometry instrument in N2 atmosphere till 400°C at 10°C/min ramp rate. For field emission scanning electron microscopy (FESEM) measurements, a JEOL JSM-7000F model was used and operated at 15 kV and 20 mA while the energy-dispersive X-ray spectroscopy (EDX) were obtained on a JEOL JED-2300F Energy Dispersive Spectrometer. Fourier transform infrared spectroscopy (FTIR) was conducted on a Variant 3100 (Excalibur Series) instrument in transmission mode. X-ray photoelec-tron spectroscopy (XPS) was measured on the as-synthesized powder on a Kratos Analytical Axis Ultra DVD using monochromated Al Ka (1486.7 eV). The binding energy of C1s was taken as 284.8 eV for calibration purposes. X-ray diffraction (XRD) patterns were acquired with a Bruker AXS D8 ADVANCE powder diffractometer using Cu Ka radiation source in the 29 range of 10-140° and operated at 25 mA and 40 kV. Rietveld refinement was conducted using the TOPAS academic version 4.2 software. For Rietveld refinement, the struc-

tural model of M-Na2-8Mn[Fe(CN)6].1.87H2O was used40 with Fe atom in place of Mn atom and with the occupancy of Fe-C and Na freely refined. Variable temperature XRD patterns were obtained on a Bruker D8 Advance powder X-ray diffractometer equipped with an Anton Paar HTK1200 High-Temperature Oven-Chamber. The measurements were conducted in either atmospheric air or high vacuum (10-2 mbar) as indicated.

Electrode preparation, cell assembly and electrochemical

evaluation.—Composite electrodes were made with the as-synthesized material as the active material, Ketjen Black (KB) (Lion Corporation) as the conductive additive and sodium salt of carboxymethyl cellulose, CMC (Alfa Aesar), as the binder in the weight ratio 85:10:5. In order to make the slurry, CMC was first dissolved in Milli-Q water to which a hand ground mixture of M-Na2Fe2(CN)6.2H2O and KB were added. After stirring at 1200 rpm for 2 h, the slurry was coated on Al foil with the doctor blade technique and then dried overnight at 120°C under 1 mbar vacuum. Upon drying, the coated electrode was pressed by a twin roller at a pressure of 37 psi. Electrodes were hence punched with an active material loading between 3-4 mg cm-2. Coin cells of 2016 type (MTI Corporation) were fabricated with such electrodes as the working electrode and Na metal (Merck) as the counter and reference electrodes with a glass fiber (Whatman, grade GF/A) as a separator layer. Prior to cell assembly, the electrodes were dried at 120°C in 1 mbar vacuum and brought inside an Ar filled glove box (MBraun, Germany) with H2O and O2 < 5 ppm. For the 4.3-2.0 V cycling with carbonate based electrolytes, 1 M NaClO4 (Alfa Aesar, 98+%, anhydrous) in ethylene carbonate, EC (Alfa Aesar): propylene carbonate, PC (Sigma Aldrich) in a 1:1 volume ratio prepared in house with no further purification, was used. For the 3.9-2.0 V cycling with carbonate based electrolytes, the electrolyte used was 0.6 M NaPF6 (Alfa Aesar, purity 99+%) in EC:PC in a 1:1 volume ratio also prepared in house with no further purification. For rate performance studies, 5 volume % of fluoroethylene carbonate, FEC (Sigma Aldrich, 99%) was added while for long term cycling studies, 2.5 volume % of FEC and 2.5 volume % vinylene carbonate, VC (Alfa Aesar, 97%) were added. For all the experiments conducted with 1 M NaBF4 in tetraethylene glycol dimethyl ether (TEGDME or tetraglyme) electrolyte, the electrolyte was prepared using both NaBF4 (purity 98%) and tetraglyme (purity > 99%) obtained from Sigma Aldrich and were used as received without further purification. 1 M NaPF6 in EC:DMC, dimethyl carbonate, (1:1 v/v) was purchased from Kishida chemicals. 0.6 M NaPF6 in diethylene glycol dimethyl ether, diglyme (Sigma Aldrich, 99.5%, anhydrous), was prepared in house with no further purification. The coin cells were cycled in a computer controlled Arbin battery tester (model BT2000, USA) at room temperature.

Full cell evaluation.—Graphite (MCMB graphite, model TB-17, from MTI) was used to make the graphite slurry with CMC as the binder in the weight ratio 95:5 (no external conductive additive was used). Na2Ti3O7/C was synthesized by a scaled-up modified version of the synthesis reported in our previous report with water as the solution medium with an in-situ C content of about 14 weight %.14 Further synthesis and material characterization details of the as-prepared Na2Ti3O7/C will be published elsewhere. The Na2Ti3O7/C slurry was prepared with Super P carbon black (as conductive additive) and CMC as binder in the weight ratio 90:5:5 such that the final weights in the slurry were as follows- Na2Ti3O7: in-situ and ex-situ carbon: CMC = 76:19:5. For the M-Na2Fe2(CN)6.2H2O//graphite full cell, the weight ratio of active material in the anode to cathode was 0.68:1 (excess cathode was used to compensate for initial coulombic inefficiencies). For the R-Na2Fe2(CN)6//graphite full cell, the anode to cathode (active material) weight ratio was 1.05:1 (much more excess cathode was taken than actually required due to reasons mentioned below) while for the R-Na2Fe2(CN)6//Na2Ti3O7 ^ Na3-xTi3O7 full cell, the anode to cathode (active material) weight ratio was 0.95:1. All full cells were straightaway assembled without any pre-cycling of cathodes or anodes. For the R-Na2Fe2(CN)6//graphite full cell, the discharge

Figure 1. Characterization of the as-synthesized powder. a) FESEM image and b) TGA plot in N2 atmosphere. c) TGA-MS plot also in N2 atmosphere clearly showing that water loss from the structure is responsible for the weight loss above 200° C. The slight increase in temperature recorded in the TGA-MS experiment with respect to that observed in the TGA (about 15°C) is believed to be arising from the different instrument used for TGA-MS. d) FTIR in transmission mode and e) the fitted XPS curve zoomed in to the Fe 2p3/2 edge. f) XRD plot with Rietveld refinement using a monoclinic structural model (space group P21/n). The inset depicts the higher angles of 70-140°20 for clarity. The vertical green ticks indicate the expected positions of the Bragg reflections. A reasonable fitting was obtained with reliable refinement factors (Rwp = 7.42%, RBragg = 4.621%, x2 = 3.39%, and Rexp = 2.19%) validating the monoclinic structural model.

was controlled by time rather than voltage while the charge cutoff voltage was 3.3 V. For the R-Na2Fe2(CN)6//Na2Ti3O7 ^ Na3-xTi3O7 full cell, owing to the flat Na+ ion insertion plateau of the Na2Ti3O7 ^ Na3-xTi3O7 pathway, the upper cutoff voltage was dynamically increased by small increments in the initial cycles to compensate for the slight increase of the cathode potential per cycle due to the lower coulombic efficiencies in the initial cycles. By utilizing such modified cycling protocols for both the full cells utilizing R-Na2 Fe2 (CN)6 as the cathode, the slightly lower coulombic efficiencies in the initial cycles inherently led to voltage slippage to higher potentials for the cathode such that it eventually cycled within its upper charge-discharge plateaus.

Ex-situ XRD, FTIR and DSC measurements.—For ex-situ XRD measurements at various states of charge and discharge, the electrodes were cycled at C/9 to the appropriate charge/discharge state. The cells were then opened in the glove box, the electrodes retrieved and all XRD patterns reported were obtained within 30 s-5 min air exposure. For the ex-situ FTIR and differential scanning calorimetry (DSC) measurements of charged/discharged M-Na2Fe2(CN)6.2H2O, special electrodes were made with just M-Na2Fe2(CN)6.2H2O and KB in the weight ratio 90:10. No binder was used so as to eliminate its contribution to the FTIR/DSC spectra. After hand-grinding, the homogenously mixed powders were stirred in Milli-Q water and then hand coated on Al foils. After cycling to the required state of charge/discharge, the cells were opened in the glove box and the electrodes were washed 20 times with anhydrous PC to remove any electrolyte salt. The washed electrodes were then dried in 1 mbar vacuum for 16 h and were hence scratched. For FTIR measurement, the scratched powders were packed into Ar filled vials which were opened just prior to FTIR measurements. The air exposure time for FTIR measurements was about 5

min. However, air exposure was not a concern for the charged samples as they were found to be air-stable. For DSC measurements, the scratched powders were sealed in aluminum capsules inside the Ar-filled glove box itself. The DSC measurements were hence performed on a TA Instrument 2920 at 10°C/min ramp rate. For DSC measurements on electrolytes, the same procedure was adopted. Hence, no air exposure occurred during these DSC measurements.

Electrolyte flammability tests.—A standard protocol was used to assess the flammability of all reported electrolytes. A precise 400 |xL of each electrolyte was taken in sealed Ar-filled vials from the glove box. They were then transferred to a fume hood in ambient air. Each electrolyte was poured onto a coin cell casing such that it completely filled the case. Then, an open flame was made to touch the surface of the electrolyte with t = 0 s as the moment the flame touched the electrolyte. The open flame was made to continuously touch the electrolyte until it caught fire. The 1 M NaBF4 in tetraglyme electrolyte did not catch fire for 1 complete minute of continuous open flame exposure.

Results and Discussion

Material synthesis and characterization.—The as-synthesized material displayed homogenous particle size of cubic shape with dimensions below 3 |im (refer to Figure 1a). ICP-OES and EDX revealed a Na:Fe molar ratio of 1:1 confirming the sodium rich nature of the as-synthesized material. Measured amounts of C and N were almost identical to expected quantities corresponding to the (CN)6 backbone (refer to Table S1 in the supplemental material). TGA revealed no significant weight loss below 180°C and about 10 wt% loss between 180-245°C (see Figure 1b). To ascertain the cause of this

weight loss, a TGA-MS experiment was undertaken and the obtained curves are presented in Figure 1c. As can be seen, the weight change event above 200°C was solely due to water loss from the structure, with negligible release of HCN and CO2, indicating that the backbone of the structure was intact. The presence of water was also confirmed by FTIR. The FTIR spectrum, displayed in Figure 1d, showed two sharp peaks at 1619 and 2071 cm-1 and two broad peaks around 3445 and 3611 cm-1. The peak at 2071 cm-1 corresponds to the cyanide stretching vibration band coordinated to Fe2+ in such PBAs. The other peaks at 1619, 3445 and 3611 cm-1 are attributable to structural water present, with the 1619 cm-1 peak corresponding to the O-H bending band and the broader peaks at high wavenumbers to the O-H stretching bands.25,36 The weight loss of about 10 wt% indicates that about 2 moles of water were present per mole of the material. Hence, based on ICP-OES, EDX, CHN and TGA analyses, the stoichiometry of this material could be stated as Na2Fe2(CN)6.2H2O highlighting the Na rich nature and lack of vacancies resulting from our synthesis protocol. This stoichiometry implies both Fe atoms should exist as Fe2+. For a second confirmation of this fact (apart from FTIR), XPS analysis was carried out to track the position of the Fe 2p3/2 edge. The as-synthesized material displayed predominantly a single peak at 708.6 eV which is consistent with the Fe2+ oxidation state, as depicted in Figure 1e.42 During the curve fitting, a minor Fe3+ peak was revealed at 710.3 eV. Since XPS is a surface analysis technique limited to the first 10 nm, the presence of the Fe3+ peak indicates a very slight Na loss from the surface occurring mainly during the precipitate retrieval step (either filtration or centrifugation) during the synthesis (refer to the Experimental section for synthesis details), while the bulk of the material remained in the Fe2+ state. This also explains why no Fe3+ peaks were observed in the FTIR spectrum as FTIR is a bulk characterization technique. In fact, the color change of the as-synthesized material mirrored this hypothesis: the precipitate while settled in the solution was white in color, however, after filtration/centrifugation, it acquired a faint cyan tinge, similar to the observation reported for the rhombohedral Prussian white Na192Fe2(CN)6.41

To determine the crystal structure of the as-synthesized Na2Fe2(CN)6.2H2O, Rietveld refinement of the powder XRD pattern was performed and shown in Figure 1f. The XRD plot looked almost identical to that reported for the monoclinic Na2Mn[Fe(CN)6].1.87H2O and monoclinic Na2Mn2(CN)6.2H2O with the space group P21/n.27,40 In fact, the structural model proposed for M-Na2-8Mn[Fe(CN)6].1.87H2O resulted in satisfactory fitting for the as-synthesized material with favorably low refinement reliability factors (Rwp = 7.42%, RBragg = 4.621%, x2 = 3.39%, and Rexp = 2.19%) with similar lattice parameters: a = 10.45983 (56) A, b = 7.51295 (42) А, с = 7.27153 (48) A and в = 92.7379 (33)° (refer to Table S2 for the atomic coordinates).40 Henceforth, the as-synthesized material shall be referred as M-Na2Fe2(CN)6.2H2O. A literature review of Na and all Fe PBAs with the general formula NaxFe2(CN)6.nH2O revealed a clear trend: most of such syntheses relied on a mild acidic environment to breakdown the Na4Fe(CN)6 precursor at low temperatures between 25 to 80°C. Under those conditions, the value of x was fixed between 0 and 1.7 (Na poorer PBAs).28-38 Goodenough et al. synthesized Na rich PBAs with x ^ 2 at an elevated temperature of 140°C in a pressure based hydrothermal reactor.41 However, we used a reflux based synthesis conducted at 140°C at ambient pressure such that the solution temperature was around 107°C during reflux. Hence, it appears that the combined action of vigorous reflux temperature and elimination of pressure during the synthesis resulted in the M phase of Na2Fe2(CN)6 rather than the R phase. As will be shown later, direct synthesis of the M phase using our reflux based approach could be advantageous as not only it can always be converted to the R phase by thermal heating, the synthesis itself is expected to be cheaper and easily scalable, being atmospheric pressure based.

Sodium storage performance.—To study the sodium storage performance of M-Na2Fe2(CN)6.2H2O, composite electrodes were fabricated with KB as the conductive additive and CMC as binder.

As M-Na2Fe2(CN)6.2H2O was found to be water-insoluble (refer to Figure S1), water was used as the slurry preparation medium, thus eliminating the need for toxic and costly n-methylpyrrolidone (NMP) as the binder solvent, which is the traditional solvent used for the most widely known polyvinylidene fluoride (PVDF) binder employed for lithium-ion battery (LIB) cathodes/anodes. The use of environmentally safe and non-toxic CMC, along with water as the medium, is expected to reduce the electrode processing cost for this material. In fact, a recent study on LIBs has suggested that switching to water-based binders could potentially reduce electrode processing costs by an order of magnitude.43 The first charge-discharge galvanostatic cycle of the M-Na2Fe2(CN)6.2H2O cathode against Na metal as the counter and reference electrodes (half cell configuration) in a coin cell between 4.3-2.0 V vs Na/Na+ is depicted in Figure 2a. With a standard NIB electrolyte (1M NaClO4 in EC:PC) which displays an electrochemical stability window till at least 4.4-4.5 V vs Na/Na+,44 the M-Na2Fe2(CN)6.2H2O cathode could deliver a capacity of 170.9 mAh g-1 during the first charge, in excess of its theoretical capacity of 153.2 mAh g-1 (corresponding to two mole sodium storage per mole of material), indicating that apparently more than two moles of sodium were extracted per mole of cathode. In the first discharge (Na insertion into the cathode), about 150 mAh g-1 was obtained, confirming that two moles of sodium were inserted back into the cathode, per mole of material. The cause of the excess charge capacity will be discussed in the following section. The Na rich nature of M-Na2Fe2(CN)6.2H2O was further evidenced when the cathode was discharged first rather than charged: discharging first resulted in negligible capacity, as indicated in Figure 2a. The cyclability of M-Na2Fe2(CN)6.2H2O in the voltage window 4.3-2.0 V was found to be quite poor, with capacity retention of just 77% of its initial discharge capacity in 10 cycles and 44% in 500 cycles at C/1.8 rate. The voltage profiles also changed significantly during cycling, with almost a complete loss of the chargedischarge plateau above 4.0 V within a few cycles. These facts indicate that the structure of M-Na2Fe2(CN)6.2H2O changes when it is forced to store two moles of sodium per mole of the material (details discussed in the following section). In stark contrast, if the cycling of M-Na2Fe2(CN)6.2H2O upon charging is stopped just before the start of the 4.0 V plateau (if cycled until 3.9 V vs Na/Na+), the galvano-static profile of this material looked very different, as presented in Figure 2b. Within this voltage window (3.9-2.0 V vs Na/Na+), the M-Na2Fe2(CN)6.2H2O cathode could deliver 84 mAh g-1 at a C/4.5 rate which indicates that 1.1 mole of sodium was stored per mole of material (76.6 mAh g-1 corresponds to the theoretical capacity of one mole sodium storage) at an acceptable average discharge voltage of 3.03 V. As shown in Figure 2b and Figure 2c, almost 100% capacity retention was observed at a slow C/4.5 (4.5 h) or fast 5.6 C (10.7 min) discharge with very stable capacities for all rates. For even faster response times such as 11.1 C rate (5.4 min discharge), the as-synthesized material could still deliver 74 mAh g-1, which was 88% of the delivered capacity at slower rates. It should be mentioned that polarization became significant only at 11.1 C as shown in Figure 2b, which indicates potential possibility for high power applications if paired with an equally responsive anode. More importantly, the cycle life of M-Na2Fe2(CN)6.2H2O within the 3.9-2.0 V voltage window was found to be dramatically improved with capacity retention of 82% and 67% observed after 2,000 and 3,000 cycles, respectively (see Figure 2d), with a highly stable coulombic efficiency above 99% throughout cycling. These results are indeed quite appealing for grid-storage applications.

Sodium storage mechanism.—Structural changes during cycling.—To understand the cause of the excellent sodium storage performance for M-Na2Fe2(CN)6.2H2O within the 3.9-2.0 V voltage window and not within 4.3-2.0 V, insights were obtained by ex-situ XRD, DSC and FTIR at different states of charge/discharge. Figure 3a presents the ex-situ XRD plots for M-Na2Fe2(CN)6.2H2O within the 3.9-2.0 V window. As the charging cycle proceeded, sodium extraction led to a merger of the peaks at 23.7 and 24.5°29 into a single peak, signaling the symmetry transition from monoclinic to cubic;

Figure 2. Galvanostatic cycling of M-Na2Fe2(CN)6.2H2O vs Na metal in a half cell. a) C/1.8 cycling within 4.3-2.0 V window (charged first). The effect of discharging first is shown as well by the red curve - a negligible capacity was obtained reflecting the Na rich nature of the material. Electrolyte used was 1 M NaClO4 in EC:PC (1:1 v/v). b) Cycling within the 3.9-2.0 V window corresponding to slightly in excess of one mole Na storage per mole of material. The discharge profiles at various rates are shown, with the charging cycle being conducted at C/4.5 rate. c) The corresponding discharge capacity values vs cycle number at various rates. d) Long term cycling of M-Na2Fe2(CN)6.2H2O within the 3.9-2.0 V window at 2.2 C rate over 3,000 cycles. For the 3.9-2.0 V window, the electrolyte was 0.6 M NaPF6 in EC:PC based.

the cubic phase of PBAs is widely reported for Na poorer versions of NaxFe2(CN)6 in the literature (0 < x < 1.6).28-38 During this transition, the voltage profile was quite flat (see Figure S2), in accordance with the co-existence of these two phases dictated by the Gibbs Phase Rule. On further charging, sodium extraction caused the cubic peaks to shift to higher 29 values, indicating the expected volume contraction owing to the removal of sodium from the structure. This continuous shift, indicating a solid-solution reaction mechanism, is the reason for the sloping voltage profiles as shown in Figure S2. During the discharge process, Na insertion caused the structure to first expand in volume while still maintaining the cubic symmetry before lowering the symmetry back to monoclinic at the later stages of discharge, confirming a smooth structural transition in the course of cycling. When cycled between 4.3-2.0 V such that both moles of sodium were extracted per mole of M-Na2Fe2(CN)6.2H2O during charging, the ex-situ XRD plot for the fully charged cathode at 4.3 V displayed greatly intensified peaks (see the peaks at 17.4 and 35.1°29 in Figure S3a) with respect to that charged to 3.9 V, without any peak shift. Upon discharge down to 2.0 V, the XRD pattern was similar to that of M-Na2Fe2(CN)6.2H2O but with slight differences in intensity particularly to the peaks at 23.7 and 24.5°29, indicating probably slight structural distortions in the course of cycling between 4.3-2.0 V. These observations are different from those reported for M-Na2-8Mn[Fe(CN)6].1.87H2O where its two mole sodium cycling resulted in a mixture of monoclinic and rhom-bohedral phases at the end of the first discharge as a consequence of the structural water loss during its cycling.40

In order to gain a deeper understanding of the galvanostatic cycling effects on the structural water of M-Na2Fe2(CN)6.2H2O, the cause of the poor cycling stability within the 4.3-2.0 V window and the origin of the 4.0 V charge plateau, ex-situ FTIR experiment was conducted at selected states of charge/discharge. Firstly, the obtained FTIR plots within the high wavenumber region are shown in Figure 3b. The O-H stretching bands at 3445 cm-1 were preserved within the 3.9-2.0 V window, indicating that the structural water present in M-Na2Fe2(CN)6.2H2O remained within the crystal structure during cycling in this restricted voltage window. On the other hand, when charged to 4.3 V, the O-H band significantly reduced, indicating release of most of its structural water during sodium extraction after the first charge itself, consistent with observations made for M-Na2-8Mn[Fe(CN)6].1.87H2O.40 This release of structural water was also reflected in the DSC curves of M-Na2Fe2(CN)6.2H2O when charged to 3.9 and 4.3 V (refer to Figure 3c). While both the 3.9 and 4.3 V charged samples exhibited broad exothermic reactions around 198°C, the DSC curve of the 3.9 V sample displayed an endothermic peak at 233°C concomitant with the loss of its structural water caused by heating during the DSC experiment. Such an endothermic reaction was not observed for the 4.3 V charged sample further confirming that M-Na2Fe2(CN)6.2H2O lost its structural water during the 4.0 V charge plateau upon cycling. This loss of structural water during charging to 4.3 V must contribute to the excess first charge capacity observed in Figure 2a due to the likely side reaction of released water molecules with the electrolyte at that high potential range (4.0-4.3 V vs Na/Na+).

Figure 3. Structural transformations occurring in M-Na2Fe2(CN)6.2H2O during sodium storage. a) Ex-situ XRD plots at various points as indicated during charging and discharging within the 3.9-2.0 V window illustrating the mixture of two-phase and solid-solution reaction mechanisms. b) Ex-situ FTIR at selected states of charge and discharge in the high wavenumber region, highlighting that the structural water was preserved during the 3.9-2.0 V cycling but released if M-Na2Fe2(CN)6.2H2O was charged to 4.3 V. c) DSC curves on charged M-Na2Fe2(CN)6.2H2O to 3.9 V and 4.3 V illustrating the presence of an endothermic peak for the 3.9 V sample which was due to water loss from the structure triggered by heating during the DSC experiment. For the 4.3 V sample, no such endothermic reaction was observed indicating that water loss had already occurred during galvanostatic charging to 4.3 V. The DSC measurements were conducted in Ar atmosphere without any air exposure to the samples. d) Ex-situ FTIR plots zoomed to the cyanide stretching frequencies at corresponding points to that shown in panel b). The FTIR spectra clarify the involvement of HS-Fe2+ and LS-Fe2+ in the galvanostatic lower and upper charge/discharge voltage plateaus of M-Na2Fe2(CN)6.2H2O, respectively.

Moreover, this structural water loss probably distorted the structure after the first charge itself, explaining the slight differences in the intensities of the peaks at 23.7 and 24.5°29 for M-Na2Fe2(CN)6.2H2O discharged to 2.0 V after being charged to 4.3 V (shown in Figure S3a). Furthermore, this combined action of two mole sodium storage per mole of M-Na2Fe2(CN)6.2H2O during cycling between 4.3-2.0 V and water loss brought about significant structural collapse and distortion after repeated cycling, as indicated by the greatly suppressed peaks along with slight peak shifts of the ex-situ XRD pattern of M-Na2Fe2(CN)6.2H2O after 500 cycles between 4.3-2.0 V (shown in Figure S3b). Such structural collapse and distortion explains the poor cycling stability within this 4.3-2.0 V voltage window and the changing voltage profiles, respectively (shown in Figure 2a). Alternatively, the poor cycling stability and changing voltage profiles could also occur due to adverse changes to the surface of the M-Na2Fe2(CN)6.2H2O particles. Additional high resolution transmission electron microscopy and FESEM studies on cycled electrodes would help in explaining such observations further. Conversely, retention of the structural water for the 3.9-2.0 V cycled M-Na2Fe2(CN)6.2H2O not only keeps the structure intact as evidenced by its excellent cycle life, but also seems to be beneficial from a safety point of view as the endothermic reac-

tion at 233°C for the charged sample to 3.9 V may arrest any thermal runaway type scenarios. This inherent in-built "thermal safety fuse" mechanism is quite appealing considering its ultimate application in large-scale grid-storage batteries.

Redox mechanism.—The crystal structure of M-Na2Fe2(CN)6.2H2O renders the Fe2+ coordinated to N (Fe2+-NC) in the high spin configuration (HS-Fe2+) and the Fe2+ coordinated to C (Fe2+-CN) in the low spin configuration (LS-Fe2+).41 To ascertain which Fe partakes in the low (3.0-3.3 V) and high (4.0 V) voltage plateaus witnessed during galvanostatic cycling, ex-situ FTIR measurements were conducted at relevant states of charge and discharge to follow the cyanide stretching vibration band as it is a sensitive indicator of the Fe oxidation state in such PBAs.25,45-47 Within the 3.9-2.0 V window, the cyanide stretching bands shift from 2072 cm-1 for the pristine state to 2078 cm-1 for the cathode charged to 3.9 V and then it shifts back to 2072 cm-1 when discharged to 2.0 V (see Figure 3d). The stretching band at 2078 cm-1 agrees well with the band at 2076 cm-1 reported for Na0.75Fe2.08(CN)6, indicating that one of the Fe2+ has been oxidized to Fe3+.31 On further charge to 4.3 V, the cyanide band shifts to a higher wavenumber of 2086

Figure 4. Thermal dehydration of the M phase in ambient air and the ambient air stabilities of the M and R phases. a) Variable temperature XRD patterns of M-Na2Fe2(CN)6.2H2O at various temperatures in ambient air atmosphere, showing the conversion to the R phase upon the loss of structural water caused by heating. b) Air stability of M-Na2Fe2(CN)6.2H2O powder in ambient air at various periods as indicated. c) Air stability of R-Na2Fe2(CN)6 electrode (formed by heating M electrode above 240°C in inert Ar atmosphere) when exposed to ambient air.

cm 1 in good agreement with the band at 2090 cm 1 observed for Fe3+Fe3+(CN)6, indicating that at 4.3 V, both Fe2+ have been oxidized to Fe3+.48 Furthermore, another minor peak at 2171 cm-1 was detected for the cathode charged to 4.3 V. In previous reports of KxFe2(CN)6 and NaxFe2(CN)6, such a relatively weak peak at 2171 cm-1 vs the 2085 cm-1 peak was attributed to a splitting of the CN stretching bands and assigned to Fe3+ bonded to the C of C=N- anion.25,45-47,49 Rather expectedly, it appears that the high voltage plateau above 4.0 V vs Na/Na+ is caused due to the oxidation/reduction of LS-Fe2+ to LS-Fe3+ while the lower voltage plateau (3.0-3.3 V) is due to the corresponding processes between HS-Fe2+ and HS-Fe3+ for M-Na2F^(CN)6.2H2O. These results are consistent with theoretical calculations33 and with other PBAs reported in NIBs where the HS-Fe2+ is active in the lower voltage regions while the LS-Fe2+ is responsible for the higher voltage plateaus.41

Thermal and air stability.—TGA-MS data for M-Na2Fe2(CN)6.2H2O (refer to Figure 1c) indicated that the structural water in this compound can be removed by heating this material above 200° C. In order to investigate the concomitant changes to the structure (if any), XRD patterns were obtained as a function of temperature and the results are presented in Figure 4a. Upon heating in ambient air, the XRD peaks of M-Na2Fe2(CN)6.2H2O began to decrease with a new set of XRD peaks appearing as the temperature reached 200°C. Further heating caused these new peaks to grow at the expense of those of the pristine phase and by 250°C, only the peaks of the new phase existed, signaling a thermally induced phase transformation above 200° C. In fact, the peaks of this new phase correspond to the rhombohedral phase of Na2Fe2(CN)6 with the space group R3 recently reported by Goodenough et al. (abbreviated as R-Na2Fe2(CN)6 henceforth).41 It is interesting to note that while thermal dehydration changed the structure from monoclinic to rhom-bohedral, electrochemical dehydration on the other hand (caused by galvanostatic charging to 4.3 V which releases the structural water), resulted in structure change from monoclinic to cubic (refer to Figure S3a) most probably due to the concomitant Na loss that also occurred during charging (Na loss did not occur during thermal dehydration). In a high vacuum environment of 10-2 mbar, on the other hand, M-Na2Fe2(CN)6.2H2O lost its structural water completely at 100°C to form once again R-Na2Fe2(CN)6 (see Figure S4a). But, when heated in a milder 1 mbar vacuum, this conversion from M to R phase did not occur even till 120°C, as shown in Figure S4b. These results indicate the importance of the atmosphere when Na rich PBAs are handled. From a practical consideration, all electrodes are typically heated in vacuum prior to fabrication. It is imperative that

the degree of vacuum must be taken into account when Na rich PBAs in general, and the M-Na2Fe2(CN)6.2H2O in particular, are heated. Hence, if a M-Na2Fe2(CN)6.2H2O electrode is purposefully heated above 100-200°C in vacuum/air/inert atmosphere, it can be made to easily transform to R-Na2Fe2(CN)6. The advantage of the R phase is that it can deliver higher capacity (theoretical capacity of 170.8 mAh g-1) within a narrow voltage window of 3.0 and 3.4 V: it displays two very flat charge-discharge plateaus centered at 3.1 and 3.3 V in accordance with the HS-Fe2+ and the LS-Fe2+, respectively.41 Indeed, when a R-Na2Fe2(CN)6 electrode was prepared by simply heating an already fabricated M-Na2Fe2(CN)6.2H2O electrode above 240°C in inert Ar atmosphere and then cycled in a sodium battery, a high capacity of 164 mAh g-1 close to its theoretical value was obtained with the charge/discharge plateaus consistent with the previous report (refer to Figure S4c).41 Please note that the shifting minor voltage step observed during discharge (as indicated by the arrows in Figure S4c) is due to the voltage step phenomenon caused by an increased polarization of the sodium counter electrode in EC:PC based solutions during the discharge cycle in a half cell.50

The advantage of the method proposed in this article of using the high capacity R phase by heating already fabricated M-Na2Fe2(CN)6.2H2O electrodes lies in the ambient air stability of these two phases. When left to ambient air, the M-Na2Fe2(CN)6.2H2O phase was found to be stable up to 5 days of complete air exposure (in ambient air and not in a dry room) as seen from Figure 4b. Eventually, another phase started appearing within 2 weeks as indicated by the appearance of a new XRD peak at the expense of the 23.7 and 24.5° 29 peaks and a corresponding quite significant peak shift of the 34.3°29 peak (refer to Figure 4b).The original M-Na2Fe2(CN)6.2H2O phase was lost completely within 5 weeks. The XRD pattern of this new phase is very similar to that reported for the cubic phase of NaxFe2(CN)6.nH2O with the space group Fm-3m, though with a greatly suppressed intensity of the 38.8°29 peak compared to some of the previous reports.28-32 However, more detailed investigation into this phase was not attempted and is left for future studies. These XRD results indicate that M-Na2Fe2(CN)6.2H2O is a metastable phase in ambient air. Under inert atmosphere, this phase was very stable for months (see Figure S5). The R phase, on the other hand, was found to be extremely air sensitive: the phase was lost within 25 min of ambient air exposure as shown in Figure 4c. It should be stated that in a dry room atmosphere, the R phase was reported to be stable for at least 20 h.41 Hence, it can be concluded that the R phase is actually moisture sensitive rather than air sensitive. It follows that water-based binders cannot obviously be used with the R phase if electrodes are fabricated directly with it. Therefore, the advantage of using the method proposed here for fabricating NIBs with the R phase (viz. converting an already

Figure 5. Electrochemical and thermal stability characteristics of the 1 M NaBF4 in tetraglyme electrolyte. a) CV curves of the first cycle against an Al electrode in a Na half cell with 1 M NaBF4 in tetraglyme and 1 M NaClO4 in EC:PC electrolyte solutions. Flammability tests conducted in ambient air by touching an open flame on b) 1 M NaBF4 in tetraglyme, c) commercially available 1 M NaPF6 in EC:DMC and d) 0.6 M NaPF6 in diglyme electrolyte solutions. "t = 0 s" represents the moment when the open flame first touched various electrolyte solutions. e) DSC heating curves of various electrolytes in Ar atmosphere highlighting the superior thermal stability of 1 M NaBF4 in tetraglyme electrolyte solution.

fabricated electrode with the M phase using water-based binder into the R phase by a simple heating step in a dry room condition) is that it may enable use of water-based binders for the R phase as well. This could significantly reduce processing costs for the R phase.

Thermally stable and non-flammable electrolyte.—In this contribution, we also reveal the compatibility of selected low voltage anodes and the relatively high voltage M and R phases of Na2Fe2(CN)6 with a novel, non-aqueous and non-flammable NIB liquid electrolyte; glyme-based 1 M NaBF4 in tetraglyme. We were encouraged by a report on shale oil processing using this electrolyte which stated an acceptable sodium ionic conductivity of 1.3 mS cm-1,51 placing its ionic conductivity just above the threshold value of 1 mS cm-1 required for battery application.52 1 M NaBF4 in tetraglyme, has, to the best of our knowledge, been reported just twice as an NIB electrolyte. In the first report, Kim et al. studied the sodium storage performance of a-NaMnO2 cathode with this electrolyte between 4.0-1.7 V vs Na/Na+. While they demonstrated relatively stable cycling for that cathode in 20 cycles, the stable coulombic efficiency was very poor, being about 82%. Such a low efficiency for a cathode would imply that a full cell fabricated with a sodium deficient anode, a-NaMnO2 cathode and this electrolyte would essentially fail within just 5-10 cycles. In a second, very recently published report, some of the same authors demonstrated very stable cycling for the low voltage Sn anode vs Na metal with this electrolyte.54 However, it is widely known in the NIB field that other glyme-based electrolytes are excellent candidates with low voltage NIB anodes.55-59 What is rare, on the other hand, is for glyme-based electrolytes to display high voltage stability in NIBs. This 1 M NaBF4 in tetraglyme electrolyte has excellent sodium storage characteristics for both high voltage cathodes as well as low voltage anodes. This is firstly gauged by the cyclic voltammetry (CV) curves of an Al foil against Na metal with this electrolyte solution (refer to Figure 5a). For comparison, the CV curves of a widely used NIB electrolyte for low voltage anodes and high voltage cathodes, 1 M NaClO4 in EC:PC, are also included.44 It can be seen that, while the reductive stability was very similar at low voltages for both electrolytes, the 1 M NaBF4 in tetraglyme rather surprisingly appeared to not decompose completely above 4.75 V vs Na/Na+, as occurred for 1 M NaClO4 in EC:PC. At the very least, it appeared that the electrolyte proposed here may be a good choice even with high voltage cathodes operating around 4.0 V vs Na/Na+. Encouragingly, the sodium storage performance of the M-Na2Fe2(CN)6.2H2O cathode using this electrolyte was found to be almost identical with that using the standard NaPF6 in EC:PC based electrolytes (compare Figure 2b and Figure 2c with Figure S6a and Figure S6b, respectively). The only difference observed was a slightly

lower coulombic efficiency in the initial few cycles with 1 M NaBF4 in tetraglyme: the efficiency would typically increase from 95% in the first cycle to 98% within 5 cycles, before eventually registering a stable value above 99% in the ensuing cycles (refer to Figure S6c). It is not clear why Kim et al. observed such poor coulombic efficiencies for a-NaMnO2 cathode using 1 M NaBF4 in tetraglyme electrolyte. Our results seem to indicate that this electrolyte has good stability at high voltages and this should be generally the case with different NIB cathodes (owing to the CV results shown in Figure 5a, which is not specific to any cathode material).

To gauge the sodium storage performance of this electrolyte at low voltages, two different anodes were chosen. Firstly, graphite was used as it has been recently demonstrated that it is capable of storing sodium in glyme-based electrolytes by forming ternary graphite intercalation compounds through solvent co-intercalation.57,58 Apart from demonstrating an attractive capacity of 100 mAh g-1 at an average voltage of roughly 0.95 V vs Na/Na+ in sodium batteries, graphite is extremely inexpensive and has established itself as a benchmark LIB anode over the past three decades. In NIBs, graphite delivers very little capacity below 0.3 V vs Na/Na+, hence, this should help in alleviating any Na plating concerns, unlike the case for hard carbon anode in NIBs and lithium storage in graphite anode of LIBs, which demonstrate the majority of their storage capacity below 0.1 V vs Na/Na+ and Li/Li+ respectively.44,58 With 1 M NaBF4 in tetraglyme electrolyte, graphite could perform equally well compared to other glyme-based electrolytes, delivering 90 mAh g-1 with essentially no capacity drop from C/5 to 5C along with excellent stability over 200 cycles at C/2 (see Figure S7). The compatibility of this electrolyte was also tested with an insertion based transition metal oxide anode. We chose the recently discovered Na2Ti3O7 ^ Na3-xTi3O7 pathway (0 < x < 1) as it is the lowest voltage non-carbon based anode known currently in NIBs, demonstrating a flat charge plateau at 0.2 V vs Na/Na+ with a moderately high capacity of about 80 mAh g-1.14 Figure 6a presents the first galvanostatic cycle of the Na2Ti3O7 ^ Na3-xTi3O7 pathway at C/2 rate with 1 M NaBF4 in tetraglyme. For comparison, the first cycle at C/2 of this pathway with a traditional carbonate based electrolyte (1 M NaClO4 in EC:PC) is also included. A charge capacity close to 74 mAh g-1 could be obtained for both cases; however, there was a stark difference in the first cycle coulombic efficiency. With 1 M NaBF4 in tetraglyme electrolyte, the first cycle coulombic efficiency was 73% while with the traditional carbonate based electrolyte, it was only 33%. This indicates that by simply switching the electrolyte for the Na2Ti3O7 ^ Na3-xTi3O7 pathway, a huge amount of first cycle irreversible capacity loss amounting to114 mAh g-1 can be saved. This will undoubtedly help in boosting the energy density of a full

Figure 6. Sodium storage performance of the Na2Ti3O7 ^ Na3-xTi3O7 pathway in a Na half cell with 1 M NaBF4 in tetraglyme electrolyte. a) The first galvanostatic cycle at C/2 rate highlighting the 0.2 V charge plateau of this pathway as well as the higher first cycle coulombic efficiency of 73% with this electrolyte as opposed to a very low coulombic efficiency of just 33% when a traditional carbonate based electrolyte, such as 1 M NaClO4 in EC:PC (1:1 v/v), was used. Rate performance of this pathway with the b) charge profiles at various rates from C/5 to 40 C with the discharging cycle at C/5 rate and c) the corresponding charge capacity obtained as a function of cycle number. Interestingly, even with the 1 M NaBF4 in tetraglyme electrolyte, the Na2Ti3O7 ^ Na3-xTi3O7 sodium storage pathway is extremely stable at all rates with little polarization till 20 C. d) Stable cycling with high coulombic efficiency of this pathway at C/2 rate with the non-flammable 1 M NaBF4 in tetraglyme electrolyte.

cell utilizing this pathway as anode and 1 M NaBF4 in tetraglyme as electrolyte, as a correspondingly lighter cathode would need to be used. Moreover, as seen from Figure 6b and Figure 6c, the Na2Ti3O7 ^ Na3-xTi3O7 pathway demonstrated equally impressive rate performance characteristics even with this new electrolyte: it could still deliver 65 mAh g-1 at a very fast 40 C rate (90 s response). The polarization became significant only at 40 C as the charge plateau was still observed at a favorably low 0.47 V for the cycling at 20 C indicating potential for high power densities if this anode is used in a full cell. Furthermore, it also displayed very stable cycling at C/2 along with a stable coulombic efficiency above 99% (refer to Figure 6d). Further details such as passivation layer etc. during sodium storage of the Na2Ti3O7 ^ Na3-xTi3O7 pathway in 1 M NaBF4 in tetraglyme will be revealed in a separate publication. These preliminary cycling results indicate that the Na2Ti3O7 ^ Na3-xTi3O7 pathway or graphite as possible anode using the 1 M NaBF4 in tetraglyme electrolyte could lead to moderate energy density and inexpensive full cells if paired with an appropriate cathode.

Apart from its favorable electrochemical traits, perhaps the most important aspect of 1 M NaBF4 in tetraglyme as an NIB electrolyte is its inherent non-flammable nature. Its non-flammability was gauged by simply lighting the electrolyte on fire with the help of an open flame. As seen from Figure 5b, it did not catch fire even after 1 min of continuous open flame exposure. In contrast, some of the standard carbonate based NIB electrolytes, such as commercially available 1

M NaPF6 in EC:DMC, caught fire within 2 s of open flame exposure under identical experimental conditions (refer to Figure 5c). It should be mentioned that there are two other glyme-based electrolytes that have been reported to function very well for both low voltage anodes and high voltage cathodes. 1 M NaPF6 in diglyme is an excellent electrolyte, performing very well with high voltage cathodes and low voltage anodes.57 However, the solvent diglyme is highly flammable. In fact, this electrolyte too caught fire within 5 s of open flame exposure as shown in Figure 5d, highlighting its inherent safety concerns. The other glyme-based electrolyte, 1 M NaClO4 in tetraglyme, was also recently shown to function extremely well with a high voltage Nao.7CoO2 cathode cycled between 3.8-2.0 V vs Na/Na+ and with graphite as the anode.60 However, owing to the explosive nature of NaClO4, its use in a practical NIB would not be appealing. These flammability testing results were complemented by DSC data on the electrolytes. As seen from Figure 5e, the DSC heating curves of these electrolytes revealed significant thermal stability for 1M NaBF4 in tetraglyme, registering no major thermal events till a relatively high temperature of 273°C. In contrast, NaPF6 in diglyme and NaPF6 in EC:DMC displayed significant thermal events at much lower temperatures of 116 and 135°C respectively, suggesting poorer thermal stability of these electrolytes. The enhanced thermal safety of 1 M NaBF4 in tetraglyme undoubtedly stems from the higher flash point of the solvent tetraglyme (141°C), in contrast with the much lower flash points of diglyme (57°C) and DMC (18°C).52,61 Despite these

Figure 7. Galvanostatic cycling of a M-Na2Fe2(CN)6.2H2O//graphite full cell at C/1.5 using the 1 M NaBF4 in tetraglyme as electrolyte. A representative cycling curve is shown in panel a) while the capacity retention and coulombic efficiency over 500 cycles is depicted in panel b). Please note that no pre-cycling of cathode or anode vs Na in half cells was conducted prior to full cell fabrication. The energy density values take into account the combined weight of the active materials in the cathode as well as the anode.

flammability test results in ambient air conditions, it should be remembered that there could be vapor pressure buildup in a sealed battery in extreme cases due to the limited volume available. Hence, such a sealed NIB even with 1 M NaBF4 in tetraglyme electrolyte could still catch fire in severe circumstances. However, the much greater thermal stability of 1 M NaBF4 in tetraglyme compared with the other electrolytes mentioned above would certainly provide a much larger buffer before such a catastrophic event, thus ensuring enhanced safety of the NIB.

Safe and inexpensive NIBs.—Encouraged by the good sodium storage characteristics of M-Na2Fe2(CN)6.2H2O and its excellent compatibility with 1 M NaBF4 in tetraglyme electrolyte, we sought to firstly fabricate a full cell with graphite anode as proof-of-concept. Figure 7a presents a representative cycle of such a M-Na2Fe2(CN)6.2H2O//graphite full cell using the non-flammable 1 M NaBF4 in tetraglyme electrolyte. Based on the cumulative active material weights of the cathode and anode, such a full cell was able to deliver an energy density of about 68 Wh kg-1 at an average voltage of 1.94 V. More importantly, it demonstrated quite stable cycling: capacity retention of 70% could be obtained after 500 cycles at C/1.5 rate with a stable coulombic efficiency above 99.5% (refer to Figure

7b). It should be noted that no pre-cycling of the cathode or anode was conducted in half cells prior to full cell fabrication as often reported in the literature to deal with the low coulombic efficiencies of the cathodes/anodes in the initial cycles: such pre-cycling approaches may boost the energy density of a full cell, but may be cumbersome from a commercial point of view.

A full cell was also fabricated using the R-Na2Fe2(CN)6 cathode such that it was made to cycle within its upper charge-discharge plateaus. In this arrangement, the lower charge plateau was used to compensate for the slightly lower coulombic efficiencies of the cathode and anode in the initial cycles (refer to Figure S8a). In a half cell configuration, the upper charge-discharge plateau of R-Na2Fe2(CN)6 cycled very stably over 500 cycles with high coulombic efficiency (see Figure S8b and Figure S8c). Such a R-Na2Fe2(CN)6//graphite full cell delivered a higher energy density of 79 Wh kg-1 at an average voltage of 2.32 V as depicted in Figure 8a. As presented in Figure 8b, this full cell displayed even more stable cycling than the M-Na2Fe2(CN)6.2H2O//graphite full cell with essentially no capacity fade in 300 cycles at C/4 cycling rate with high coulombic efficiency. Finally, another full cell was attempted with the upper charge/discharge plateaus of R-Na2Fe2(CN)6 as the cathode and the Na2Ti3O7 ^ Na3-xTi3O7 pathway as the anode. Similar to the case

Figure 8. Galvanostatic cycling of a R-Na2Fe2 (CN)6 //graphite full cell at C/4 with 1 M NaBF4 in tetraglyme as electrolyte. A representative cycling curve is shown in panel a) while the capacity retention and coulombic efficiency over 300 cycles is presented in panel b) illustrating the outstanding stability of such a full cell with a higher energy density than the M-Na2Fe2(CN)6.2H2O//graphite full cell shown in Figure 7. Once again, no pre-cycling of cathode or anode vs Na in half cells was conducted prior to full cell fabrication and the energy density values take into account the combined weight of the active materials in the cathode as well as the anode.

Figure 9. Galvanostatic cycling of an R-Na2Fe2(CN)6//Na2Ti3Ö7 ^ Na3.xTi3Ö7 full cell with 1 M NaBF4 in tetraglyme electrolyte. A representative C/1.5 cycling curve of the full cell illustrating the high discharge plateau at 3 V resulting in an attractive energy density of 88.4 Wh kg-1 (based on cumulative weights of the active materials in the cathode and anode). The capacity retention of such a full cell over 40 cycles at C/1.5 rate is also depicted. As with the graphite full cells, no pre-cycling of cathode or anode in a half cell was conducted prior to full cell assembly. Such a moderately high energy density was possible in large part due to the high first cycle coulombic efficiency of the Na2Ti3Ö7 ^ Na3-xTi3Ö7 pathway anode with 1 M NaBF4 in tetraglyme electrolyte (refer to Figure 6a).

of R-Na2Fe2(CN)6//graphite full cell, the lower charge plateau of R-Na2Fe2(CN)6 was used to compensate for the lower coulombic efficiencies in the R-Na2Fe2(CN)6//Na2Ti3O7 ^ Na3-xTi3O7 full cell which primarily arose due to the lower first and second cycle coulombic efficiencies of the anode (73 and 91%, respectively, as shown in Figure 6d). The cycling profile of a R-Na2Fe2(CN)6//Na2Ti3O7 ^ Na3-xTi3O7 full cell is presented in Figure 9 at C/1.5 rate. The full cell was able to deliver an impressive energy density of 88.4 Wh kg-1 (based on the active material weights in the cathode and anode) at a moderately high average discharge voltage of 2.53 V. From this figure, the beneficial effect of the 0.2 V charge plateau of the Na2Ti3O7 ^ Na3-xTi3O7 pathway is abundantly clear: the full cell displayed a flat discharge plateau from 3.1-3.0 V accounting for nearly 50% of the discharge capacity. Furthermore, this full cell delivered quite stable cycling over 40 cycles at C/1.5 rate, still retaining about 83% of its initial capacity. These preliminary results are quite encouraging and it is expected that with further optimization, such a R-Na2Fe2(CN)6//Na2Ti3O7 ^ Na3-xTi3O7 full cell can deliver above 100 Wh kg-1 with stable cycling analogous to their respective half cells. These reports on above three full cells relying on earth-abundant and inexpensive cathodes and anodes (with little material and synthesis related costs) and the ability to use water-based CMC as binder for all electrodes (translating to less electrode manufacturing costs), would certainly be attractive for grid-storage applications due to their expected low costs, long cycle lives, moderate energy densities and high efficiencies. Furthermore, the inherent safety of such full cells championed by the non-flammable nature of the electrolyte is perhaps their most appealing aspect for large-scale EES applications.


The results described in this text relate firstly with the sodium storage characteristics of a newly discovered Na rich all Fe hexa-cyanoferrate phase with a monoclinic symmetry. This material, M-Na2Fe2(CN)6.2H2O, is shown to cycle in a very stable fashion as a NIB cathode for over 3,000 cycles along with an excellent high rate performance up to 11 C with little capacity drop, if slightly more than

one mole of sodium is extracted per mole of material by limiting the voltage window. Under this constraint on voltage window, we have shown that the structural water present in the material remains intact within the crystal structure and this fact aids in its cycling stability. If two moles of sodium were extracted per mole of material, then the structural water was released which was shown to have a detrimental effect on its cycle life and thermal stability of the charged state. The low and high voltage charge/discharge plateaus of the M-Na2Fe2(CN)6.2H2O cathode were demonstrated to arise due to the redox activities of the high spin Fe2+ and the low spin Fe2+, respectively. The as-synthesized M-Na2Fe2(CN)6.2H2O material was found to be air-stable for at least 5 days in ambient air and also water-insoluble, enabling the use of water-based binders. The M-Na2Fe2(CN)6.2H2O phase could be thermally dehydrated in air/inert atmospheres or in vacuum to form the rhombohedral R-Na2Fe2(CN)6 phase. Finally, we discussed a practical way of using the high capacity but moisture sensitive R-Na2Fe2(CN)6 cathode in NIB application by converting an electrode previously fabricated with the M phase into the R phase by a simple heating process.

In this contribution, we also revealed a new NIB electrolyte, 1 M NaBF4 in tetraglyme, which was shown to be non-flammable, significantly more thermally stable than some current popular NIB electrolytes and also elicited excellent performance from both low voltage anodes as well as high voltage cathodes for the first time. It displayed the added advantage of being compatible with the inexpensive and abundant NIB anode graphite. Furthermore, this electrolyte led to a drastic improvement in the first cycle coulombic efficiency of a promising anode, the Na2Ti3O7 ^ Na3-xTi3O7 pathway, thus partially solving a critical bottleneck associated with this as well as other low voltage NIB anodes. We believe that this electrolyte has most of the requirements to potentially become the next state-of-the-art electrolyte in the field of NIBs and sodium batteries. With this electrolyte, three different types of full cells were also demonstrated without any pre-cycling of the anode or cathode in half cells prior to the full cell assembly, as is often done in the literature but may or may not be practically viable. Despite this, energy density values of 68 Wh kg-1 (based on both cathode and anode active material weights) at an average voltage of 1.94 V, 79 Wh kg-1 at 2.32 V and 88.4 Wh kg-1 at 2.53 V could be obtained from a M-Na2Fe2(CN)6.2H2O//graphite, a R-Na2Fe2(CN)6//graphite and a R-Na2Fe2(CN)6//Na2Ti3O7 ^ Na3-xTi3O7 full cell with stable cycling. Such preliminary results are quite promising and point favorably to their use in large-scale grid-storage NIBs especially when considering the expected low material costs, synthesis and processing/manufacturing related costs associated with each of these cathodes and anodes, the non-flammability of the electrolyte thus ensuring safety, the excellent cycling stabilities, high efficiencies and moderate energy densities.


The authors thank the following agencies for funding: EMA (WBS No. R-265-000-568-279), MoE (WBS No. R-265-000-510-112) and NUS (WBS No. R-261-510-001-646). K.D. thanks the China Scholarship Council.


1. D. Larcher and J. M. Tarascon,Nat. Chem., 7(1), 19 (2015).

2. Z. Yang, J. Zhang, M. C. W. Kintner-Meyer, X. Lu, D. Choi, J. P. Lemmon, and J. Liu, Chem. Rev., 111(5), 3577 (2011).

3. B. Dunn, H. Kamath, and J.-M. Tarascon, Science, 334(6058), 928 (2011).

4. D. Kundu, E. Talaie, V. Duffort, and L. F. Nazar, Angew. Chem. Int. Ed., 54(11), 3431


5. N. Yabuuchi, K. Kubota, M. Dahbi, and S. Komaba, Chem. Rev., 114(23), 11636 (2014).

6. E. J. Berg, C. Villevieille, D. Streich, S. Trabesinger, and P. Novak, J. Electrochem. Soc., 162(14), A2468 (2015).

7. J. Peters, D. Buchholz, S. Passerini, and M. Weil, Energy Environ. Sci., 9(5), 1744


8. L. Mu, S. Xu, Y. Li, Y.-S. Hu, H. Li, L. Chen, and X. Huang, Adv. Mater., 27(43), 6928 (2015).

9. M. Keller, D. Buchholz, and S. Passerini, Adv. Energy Mater., 6(3), 1501555 (2016).

10. H. Kim, G. Yoon, I. Park, K.-Y. Park, B. Lee, J. Kim, Y.-U. Park, S.-K. Jung, H.-D. Lim, D. Ahn, S. Lee, and K. Kang, Energy Environ. Sci., 8(11), 3325 (2015).

11. P. Barpanda, G. Liu, C. D. Ling, M. Tamaru, M. Avdeev, S.-C. Chung, Y. Yamada, and A. Yamada, Chem. Mater., 25(17), 3480 (2013).

12. H. Kim, I. Park, S. Lee, H. Kim, K.-Y. Park, Y.-U. Park, H. Kim, J. Kim, H.-D. Lim, W.-S. Yoon, and K. Kang, Chem. Mater., 25(18), 3614 (2013).

13. A. Rudola, K. Saravanan, S. Devaraj, H. Gong, and P. Balaya, Chem. Commun., 49(67), 7451 (2013).

14. A. Rudola, N. Sharma, and P. Balaya, Electrochem. Commun., 61, 10 (2015).

15. A. Rudola, K. Saravanan, C. W. Mason, and P. Balaya, J. Mater. Chem. A, 1(7), 2653


16. J. Ni, S. Fu, C. Wu, Y. Zhao, J. Maier, Y. Yu, and L. Li, Adv. Energy Mater., 6(11), 1502568 (2016).

17. J. Xu, C. Ma, M. Balasubramanian, and Y. S. Meng, Chem. Commun., 50(83), 12564


18. H. Xiong, M. D. Slater, M. Balasubramanian, C. S. Johnson, and T. Rajh, J. Phys. Chem. Lett., 2(20), 2560 (2011).

19. B. Mortemard de Boisse, D. Carlier, M. Guignard, L. Bourgeois, and C. Delmas, Inorg. Chem., 53(20), 11197 (2014).

20. M. Shirpour, J. Cabana, and M. Doeff, Energy Environ. Sci., 6(8), 2538 (2013).

21. D. Seshadri, M. Shirpour, and M. Doeff, J. Electrochem. Soc., 162(1), A52 (2015).

22. H. He, D. Sun, Q. Zhang, F. Fu, Y. Tang, J. Guo, M. Shao, and H. Wang, ACS Applied Materials & Interfaces, 9(7), 6093 (2017).

23. Y. Lu, L. Wang, J. Cheng, and J. B. Goodenough, Chem. Commun., 48(52), 6544 (2012).

24. M. Pasta, C. D. Wessells, N. Liu, J. Nelson, M. T. McDowell, R. A. Huggins, M. F. Toney, and Y. Cui, Nat. Commun., 5 (2014).

25. X. Sun, V. Duffort, and L. F. Nazar, Adv. Sci., 3, 1600044 (2016).

26. J. Keggin and F. Miles, Nature, 137(7), 577 (1936).

27. H.-W. Lee, R. Y. Wang, M. Pasta, S. Woo Lee, N. Liu, and Y. Cui, Nat. Commun., 5 (2014).

28. Y. You, X.-L. Wu, Y.-X. Yin, and Y.-G. Guo, Energy Environ. Sci., 7(5), 1643 (2014).

29. S. J. R. Prabakar, J. Jeong, andM. Pyo, RSCAdv., 5(47), 37545 (2015).

30. A. J. Fernandez-Ropero, M. J. Piernas-Muñoz, E. Castillo-Martínez, T. Rojo, and M. Casas-Cabanas, Electrochim. Acta, 210, 352 (2016).

31. M. J. Piernas-Munoz, E. Castillo-Martínez, J. L. Gomez-Camer, and T. Rojo, Elec-trochim. Acta, 200, 123 (2016).

32. D. Yang, J. Xu, X.-Z. Liao, H. Wang, Y.-S. He, and Z.-F. Ma, Chem. Commun., 51(38), 8181 (2015).

33. Y. Jiang, S. Yu, B. Wang, Y. Li, W. Sun, Y. Lu, M. Yan, B. Song, and S. Dou, Adv. Funct. Mater., 26, 5315 (2016).

34. W.-J. Li, S.-L. Chou, J.-Z. Wang, Y.-M. Kang, J.-L. Wang, Y. Liu, Q.-F. Gu, H.-K. Liu, and S.-X. Dou, Chem. Mater, 27(6), 1997 (2015).

35. Y. Liu, Y. Qiao, W. Zhang, Z. Li, X. Ji, L. Miao, L. Yuan, X. Hu, and Y. Huang, Nano Energy, 12,386 (2015).

36. M.J. Piernas-Munoz, E. Castillo-Martinez, O. Bondarchuk, M. Armand, and T. Rojo, J. Power Sources, 324, 766 (2016).

37. Y. Tang, W. Zhang, L. Xue, X. Ding, T. Wang, X. Liu, J. Liu, X. Li, and Y. Huang, J. Mater. Chem. A, 4(16), 6036 (2016).

38. Y. You, X. Yu, Y. Yin, K.-W. Nam, and Y.-G. Guo, Nano Res., 8(1), 117 (2015).

39. L. Wang, Y. Lu, J. Liu, M. Xu, J. Cheng, D. Zhang, and J. B. Goodenough, Angew. Chem. Int. Ed., 52(7), 1964 (2013).

40. J. Song, L. Wang, Y. Lu, J. Liu, B. Guo, P. Xiao, J.-J. Lee, X.-Q. Yang, G. Henkelman, and J. B. Goodenough, J. Am. Chem. Soc., 137(7), 2658 (2015).

41. L. Wang, J. Song, R. Qiao, L. A. Wray, M. A. Hossain, Y.-D. Chuang, W. Yang, Y. Lu, D. Evans, J.-J. Lee, S. Vail, X. Zhao, M. Nishijima, S. Kakimoto, and J. B. Goodenough, J. Am. Chem. Soc., 137(7), 2548 (2015).

42. M. Datta and A. Datta, J. Phys. Chem., 94(21), 8203 (1990).

43. D. L. Wood III, J. Li, and C. Daniel, J. Power Sources, 275, 234 (2015).

44. A. Ponrouch, E. Marchante, M. Courty, J.-M. Tarascon, and M. R. Palacin, Energy Environ. Sci., 5(9), 8572 (2012).

45. J. Childers, A. Crumbliss, P. Lugg, R. Palmer, N. Morosoff, and D. Patel, J. Phys. Colloques, 44(C6), C6-C286 (1983).

46. P. J. Kulesza, M. A. Malik, A. Denca, and J. Strojek, Anal. Chem., 68(14), 2442 (1996).

47. A. L. Crumbliss, P. S. Lugg, J. W. Childers, and R. A. Palmer, J. Phys. Chem., 89(3), 482 (1985).

48. X. Wu, W. Deng, J. Qian, Y. Cao, X. Ai, and H. Yang, J. Mater. Chem. A, 1(35), 10130 (2013).

49. S. N. Ghosh, J. Inorg. Nucl. Chem., 36(11), 2465 (1974).

50. A. Rudola, D. Aurbach, and P. Balaya, Electrochem. Commun., 46, 56 (2014).

51. J. Gordon, in "Post Retort, Pre Hydro-treat Upgrading of Shale Oil, Quarterly Report: April 2010-June 2010". Ceramatec, Inc., 2010.

52. A. Ponrouch, D. Monti, A. Boschin, B. Steen, P. Johansson, and M. R. Palacin, J. Mater. Chem. A, 3(1), 22 (2015).

53. I.-H. Jo, H.-S. Ryu, D.-G. Gu, J.-S. Park, I.-S. Ahn, H.-J. Ahn, T.-H. Nam, and K.-W. Kim, Mater. Res. Bull., 58, 74 (2014).

54. C. Kim, K.-Y. Lee, I. Kim, J. Park, G. Cho, K.-W. Kim, J.-H. Ahn, and H.-J. Ahn, J. Power Sources, 317, 153 (2016).

55. Z. W. Seh, J. Sun, Y. Sun, and Y. Cui, ACS Cent. Sci., 1(8), 449 (2015).

56. Y. Zhu, L. Suo, T. Gao, X. Fan, F. Han, and C. Wang, Electrochem. Commun., 54, 18 (2015).

57. H. Kim, J. Hong, Y.-U. Park, J. Kim, I. Hwang, and K. Kang, Adv. Funct. Mater., 25(4), 534 (2015).

58. B. Jache and P. Adelhelm, Angew. Chem. Int. Ed., 53(38), 10169 (2014).

59. B. Jache, J. O. Binder, T. Abe, and P. Adelhelm, Phys. Chem. Chem. Phys., 18(21), 14299 (2016).

60. I. Hasa, X. Dou, D. Buchholz, Y. Shao-Horn, J. Hassoun, S. Passerini, and B. Scrosati, J. Power Sources, 310, 26 (2016).

61. S. Tang and H. Zhao, RSCAdvances, 4(22), 11251 (2014).