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

Oneofthekeyrequirementsoflarge-scalegrid-storagesystemsisdevelopmentofinexpensiveandsafebatteries.Sodium-ionbatteriesusingearth-abundantFeorTibasedcathodesandanodeswouldbeidealcandidatesforsuchstoragesystems.Herein,anewphaseofNa-richandallFePrussianBlueAnalogue,monoclinicNa 2 Fe 2 (CN) 6 .2H 2 O, 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 NaBF 4 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 Na 2 Ti 3 O 7 (cid:2)(cid:3) Na 3-x Ti 3 O 7 pathway of Na 2 Ti 3 O 7 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. ﬁeldemissionscanningelectronmicroscopy(FESEM)measurements,aJEOLJSM-7000Fmodelwasusedandoperatedat15kVand20mAwhiletheenergy-dispersiveX-rayspectroscopy(EDX)wereobtainedonaJEOLJED-2300FEnergyDispersiveSpectrometer.Fouriertrans-form infrared spectroscopy (FTIR) 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 K α (1486.7 eV). The binding energy of C1s was taken as 284.8 for calibration purposes. X-ray diffraction (XRD) patterns were acquired with a Bruker AXS D8 ADVANCE powder diffractometer using Cu K α radiation source in the 2 θ 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 structural model of and of reﬁned. Variable temperature XRD patterns were obtained on a Bruker X-ray diffractometer equipped Oven-Chamber. The mea- surements were conducted in either atmospheric air or high vacuum − 2 mbar) as indicated.

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][9][10][11][12][13][14][15][16][17][18][19][20][21][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 Na x M 1 [M 2 (CN) 6 ] 1-y y .nH 2 O with 0 ≤ x ≤ 2 and 0 ≤ y < 1. Here, M 1 and M 2 are transition-metal ions strongly covalently bridged by cyano (C≡N − ) ligands with M 1 octahedrally coordinated to N, M 2 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 M 1 and M 2 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][25][26][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][29][30][31][32][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][35][36][37] Till date, there have been only two reports which have displayed a value of 1.6 ≤ x ≤ 2 in their general formula Na x M 1 M 2 (CN) 6 .nH 2 O, particularly with both M 1 and M 2 as Fe. Guo et al. synthesized a Na richer Na 1.63 Fe 1.89 (CN) 6 with a rhombohedral 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 Na 1.63 Fe 1.89 (CN) 6 . Goodenough et al. recently reported three variations of Na rich PBAs-monoclinic M-Na 2-δ Mn[Fe(CN) 6 ].1.87H 2 O, rhombohedral R-Na 2-δ Mn[Fe(CN) 6 ] and rhombohedral R-Na 1.92 Fe[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][40][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 P2 1 /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-Na 2-δ Mn[Fe(CN) 6 ].1.87H 2 O. We will discuss many facets about the sodium storage characteristics of M-Na 2 Fe 2 (CN) 6 .2H 2 O, 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O and R-Na 2 Fe 2 (CN) 6 cathodes paired with either graphite or the Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 pathway as anodes with this non-flammable electrolyte resulting in inexpensive and stable NIBs suitable for large-scale grid-storage applications.

Experimental
Material synthesis.-In a typical synthesis, 5 mmoles of Na 4 Fe(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 spectroscopy (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 N 2 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 N 2 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 photoelectron spectroscopy (XPS) was measured on the as-synthesized powder on a Kratos Analytical Axis Ultra DVD using monochromated Al Kα (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 Kα radiation source in the 2θ 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-Na 2-δ Mn[Fe(CN) 6 ].1.87H 2 O was used 40 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 assynthesized 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 H 2 O and O 2 < 5 ppm. For the 4.3-2.0 V cycling with carbonate based electrolytes, 1 M NaClO 4 (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 NaPF 6 (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 NaBF 4 in tetraethylene glycol dimethyl ether (TEGDME or tetraglyme) electrolyte, the electrolyte was prepared using both NaBF 4 (purity 98%) and tetraglyme (purity ≥ 99%) obtained from Sigma Aldrich and were used as received without further purification. 1 M NaPF 6 in EC:DMC, dimethyl carbonate, (1:1 v/v) was purchased from Kishida chemicals. 0.6 M NaPF 6 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). Na 2 Ti 3 O 7 /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 Na 2 Ti 3 O 7 /C will be published elsewhere. The Na 2 Ti 3 O 7 /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-Na 2 Ti 3 O 7 : in-situ and ex-situ carbon: CMC = 76:19:5. For the M-Na 2 Fe 2 (CN) 6 .2H 2 O//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-Na 2 Fe 2 (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-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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-Na 2 Fe 2 (CN) 6 //graphite full cell, the discharge 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 2p 3/2 edge. f) XRD plot with Rietveld refinement using a monoclinic structural model (space group P2 1 /n). The inset depicts the higher angles of 70-140 • 2θ for clarity. The vertical green ticks indicate the expected positions of the Bragg reflections. A reasonable fitting was obtained with reliable refinement factors (R wp = 7.42%, R Bragg = 4.621%, χ 2 = 3.39%, and R exp = 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-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 full cell, owing to the flat Na + ion insertion plateau of the Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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-Na 2 Fe 2 (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-Na 2 Fe 2 (CN) 6 .2H 2 O, special electrodes were made with just M-Na 2 Fe 2 (CN) 6 .2H 2 O 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 Arfilled 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.

Results and Discussion
Material synthesis and characterization.-The as-synthesized material displayed homogenous particle size of cubic shape with dimensions below 3 μm (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 CO 2 , 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 Fe 2+ in such PBAs. 36 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 Na 2 Fe 2 (CN) 6 .2H 2 O highlighting the Na rich nature and lack of vacancies resulting from our synthesis protocol. This stoichiometry implies both Fe atoms should exist as Fe 2+ . For a second confirmation of this fact (apart from FTIR), XPS analysis was carried out to track the position of the Fe 2p 3/2 edge. The as-synthesized material displayed predominantly a single peak at 708.6 eV which is consistent with the Fe 2+ oxidation state, as depicted in Figure 1e. 42 During the curve fitting, a minor Fe 3+ peak was revealed at 710.3 eV. Since XPS is a surface analysis technique limited to the first 10 nm, the presence of the Fe 3+ 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 Fe 2+ state. This also explains why no Fe 3+ 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 Na 1.92 Fe 2 (CN) 6 . 41 To determine the crystal structure of the as-synthesized Na 2 Fe 2 (CN) 6 .2H 2 O, 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 Na 2 Mn[Fe(CN) 6 ].1.87H 2 O and monoclinic Na 2 Mn 2 (CN) 6 .2H 2 O with the space group P2 1 /n. 27,40 In fact, the structural model proposed for M-Na 2-δ Mn[Fe(CN) 6 ].1.87H 2 O resulted in satisfactory fitting for the as-synthesized material with favorably low refinement reliability factors (R wp = 7.42%, R Bragg = 4.621%, χ 2 = 3.39%, and R exp = 2.19%) with similar lattice parameters: a = 10.45983 (56) Å, b = 7.51295 (42) Å, c = 7.27153 (48) Å and β = 92.7379 (33) • (refer to Table S2 for the atomic coordinates). 40 Henceforth, the assynthesized material shall be referred as M-Na 2 Fe 2 (CN) 6 .2H 2 O. A literature review of Na and all Fe PBAs with the general formula Na x Fe 2 (CN) 6 .nH 2 O revealed a clear trend: most of such syntheses relied on a mild acidic environment to breakdown the Na 4 Fe(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][29][30][31][32][33][34][35][36][37][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 Na 2 Fe 2 (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-Na 2 Fe 2 (CN) 6 .2H 2 O, composite electrodes were fabricated with KB as the conductive additive and CMC as binder.
As M-Na 2 Fe 2 (CN) 6 .2H 2 O 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 waterbased binders could potentially reduce electrode processing costs by an order of magnitude. 43 The first charge-discharge galvanostatic cycle of the M-Na 2 Fe 2 (CN) 6 .2H 2 O 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 NaClO 4 in EC:PC) which displays an electrochemical stability window till at least 4.4-4.5 V vs Na/Na + , 44 the M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O upon charging is stopped just before the start of the 4.0 V plateau (if cycled until 3.9 V vs Na/Na + ), the galvanostatic 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 gridstorage applications.
Sodium storage mechanism.-Structural changes during cycling.-To understand the cause of the excellent sodium storage performance for M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 • 2θ into a single peak, signaling the symmetry transition from monoclinic to cubic;  6 .2H 2 O 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 NaClO 4 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 NaPF 6 in EC:PC based. the cubic phase of PBAs is widely reported for Na poorer versions of Na x Fe 2 (CN) 6 in the literature (0 ≤ x ≤ 1.6). [28][29][30][31][32][33][34][35][36][37][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 2θ 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 • 2θ 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-Na 2 Fe 2 (CN) 6 .2H 2 O but with slight differences in intensity particularly to the peaks at 23.7 and 24.5 • 2θ, 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-Na 2-δ Mn[Fe(CN) 6 ].1.87H 2 O where its two mole sodium cycling resulted in a mixture of monoclinic and rhombohedral 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-Na 2 Fe 2 (CN) 6 .2H 2 O, 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2-δ Mn[Fe(CN) 6 ].1.87H 2 O. 40 This release of structural water was also reflected in the DSC curves of M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 + ).  6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O was charged to 4.3 V. c) DSC curves on charged M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Fe 2+ and LS-Fe 2+ in the galvanostatic lower and upper charge/discharge voltage plateaus of M-Na 2 Fe 2 (CN) 6 .2H 2 O, 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 • 2θ for M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O renders the Fe 2+ coordinated to N (Fe 2+ -NC) in the high spin configuration (HS-Fe 2+ ) and the Fe 2+ coordinated to C (Fe 2+ -CN) in the low spin configuration (LS-Fe 2+ ). 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][46][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 Na 0.75 Fe 2.08 (CN) 6 , indicating that one of the Fe 2+ has been oxidized to Fe 3+ . 31 On further charge to 4.3 V, the cyanide band shifts to a higher wavenumber of 2086  6 .2H 2 O powder in ambient air at various periods as indicated. c) Air stability of R-Na 2 Fe 2 (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 Fe 3+ Fe 3+ (CN) 6 , indicating that at 4.3 V, both Fe 2+ have been oxidized to Fe 3+ . 48 Furthermore, another minor peak at 2171 cm −1 was detected for the cathode charged to 4.3 V. In previous reports of K x Fe 2 (CN) 6 and Na x Fe 2 (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 Fe 3+ bonded to the C of C≡N − anion. 25,[45][46][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-Fe 2+ to LS-Fe 3+ while the lower voltage plateau (3.0-3.3 V) is due to the corresponding processes between HS-Fe 2+ and HS-Fe 3+ for M-Na 2 Fe 2 (CN) 6 .2H 2 O. These results are consistent with theoretical calculations 33 and with other PBAs reported in NIBs where the HS-Fe 2+ is active in the lower voltage regions while the LS-Fe 2+ is responsible for the higher voltage plateaus. 41 Thermal and air stability.-TGA-MS data for M-Na 2 Fe 2 (CN) 6 .2H 2 O (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-Na 2 Fe 2 (CN) 6 .2H 2 O 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 Na 2 Fe 2 (CN) 6 with the space group R3 recently reported by Goodenough et al. (abbreviated as R-Na 2 Fe 2 (CN) 6 henceforth). 41 It is interesting to note that while thermal dehydration changed the structure from monoclinic to rhombohedral, 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-Na 2 Fe 2 (CN) 6 .2H 2 O lost its structural water completely at 100 • C to form once again R-Na 2 Fe 2 (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-Na 2 Fe 2 (CN) 6 .2H 2 O in particular, are heated. Hence, if a M-Na 2 Fe 2 (CN) 6 .2H 2 O electrode is purposefully heated above 100-200 • C in vacuum/air/inert atmosphere, it can be made to easily transform to R-Na 2 Fe 2 (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-Fe 2+ and the LS-Fe 2+ , respectively. 41 Indeed, when a R-Na 2 Fe 2 (CN) 6 electrode was prepared by simply heating an already fabricated M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O electrodes lies in the ambient air stability of these two phases. When left to ambient air, the M-Na 2 Fe 2 (CN) 6 .2H 2 O 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 • 2θ peaks and a corresponding quite significant peak shift of the 34.3 • 2θ peak (refer to Figure 4b).The original M-Na 2 Fe 2 (CN) 6 .2H 2 O 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 Na x Fe 2 (CN) 6 .nH 2 O with the space group Fm-3m, though with a greatly suppressed intensity of the 38.8 • 2θ peak compared to some of the previous reports. [28][29][30][31][32] However, more detailed investigation into this phase was not attempted and is left for future studies. These XRD results indicate that M-Na 2 Fe 2 (CN) 6 .2H 2 O 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 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 Na 2 Fe 2 (CN) 6 with a novel, non-aqueous and non-flammable NIB liquid electrolyte; glymebased 1 M NaBF 4 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 NaBF 4 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 α-NaMnO 2 cathode with this electrolyte between 4.0-1.7 V vs Na/Na + . 53 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, α-NaMnO 2 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][56][57][58][59] What is rare, on the other hand, is for glyme-based electrolytes to display high voltage stability in NIBs. This 1 M NaBF 4 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 NaClO 4 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 NaBF 4 in tetraglyme rather surprisingly appeared to not decompose completely above 4.75 V vs Na/Na + , as occurred for 1 M NaClO 4 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-Na 2 Fe 2 (CN) 6 .2H 2 O cathode using this electrolyte was found to be almost identical with that using the standard NaPF 6 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 NaBF 4 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 α-NaMnO 2 cathode using 1 M NaBF 4 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 NaBF 4 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 pathway at C/2 rate with 1 M NaBF 4 in tetraglyme. For comparison, the first cycle at C/2 of this pathway with a traditional carbonate based electrolyte (1 M NaClO 4 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 NaBF 4 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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 Na 3-x Ti 3 O 7 pathway in a Na half cell with 1 M NaBF 4 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 NaClO 4 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 NaBF 4 in tetraglyme electrolyte, the Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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 NaBF 4 in tetraglyme electrolyte. cell utilizing this pathway as anode and 1 M NaBF 4 in tetraglyme as electrolyte, as a correspondingly lighter cathode would need to be used. Moreover, as seen from Figure 6b and Figure 6c, the Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 pathway in 1 M NaBF 4 in tetraglyme will be revealed in a separate publication. These preliminary cycling results indicate that the Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 pathway or graphite as possible anode using the 1 M NaBF 4 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 NaBF 4 in tetraglyme as an NIB electrolyte is its inherent non-flammable nature.  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 NaPF 6 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 NaClO 4 in tetraglyme, was also recently shown to function extremely well with a high voltage Na 0.7 CoO 2 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 NaClO 4 , 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 NaBF 4 in tetraglyme, registering no major thermal events till a relatively high temperature of 273 • C. In contrast, NaPF 6 in diglyme and NaPF 6 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 NaBF 4 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 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 NaBF 4 in tetraglyme electrolyte could still catch fire in severe circumstances. However, the much greater thermal stability of 1 M NaBF 4 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-Na 2 Fe 2 (CN) 6 .2H 2 O and its excellent compatibility with 1 M NaBF 4 in tetraglyme electrolyte, we sought to firstly fabricate a full cell with graphite anode as proof-ofconcept. Figure 7a presents a representative cycle of such a M-Na 2 Fe 2 (CN) 6 .2H 2 O//graphite full cell using the non-flammable 1 M NaBF 4 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-Na 2 Fe 2 (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-Na 2 Fe 2 (CN) 6 cycled very stably over 500 cycles with high coulombic efficiency (see Figure S8b and Figure S8c). Such a R-Na 2 Fe 2 (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-Na 2 Fe 2 (CN) 6 .2H 2 O//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-Na 2 Fe 2 (CN) 6 as the cathode and the Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 pathway as the anode. Similar to the case Figure 8. Galvanostatic cycling of a R-Na 2 Fe 2 (CN) 6 //graphite full cell at C/4 with 1 M NaBF 4 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-Na 2 Fe 2 (CN) 6 .2H 2 O//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.
) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 207.241.231.83 Downloaded on 2018-07-20 to IP Figure 9. Galvanostatic cycling of an R-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 full cell with 1 M NaBF 4 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 pathway anode with 1 M NaBF 4 in tetraglyme electrolyte (refer to Figure 6a). of R-Na 2 Fe 2 (CN) 6 //graphite full cell, the lower charge plateau of R-Na 2 Fe 2 (CN) 6 was used to compensate for the lower coulombic efficiencies in the R-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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.

Conclusions
The results described in this text relate firstly with the sodium storage characteristics of a newly discovered Na rich all Fe hexacyanoferrate phase with a monoclinic symmetry. This material, M-Na 2 Fe 2 (CN) 6 .2H 2 O, 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-Na 2 Fe 2 (CN) 6 .2H 2 O cathode were demonstrated to arise due to the redox activities of the high spin Fe 2+ and the low spin Fe 2+ , respectively. The as-synthesized M-Na 2 Fe 2 (CN) 6 .2H 2 O 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-Na 2 Fe 2 (CN) 6 .2H 2 O phase could be thermally dehydrated in air/inert atmospheres or in vacuum to form the rhombohedral R-Na 2 Fe 2 (CN) 6 phase. Finally, we discussed a practical way of using the high capacity but moisture sensitive R-Na 2 Fe 2 (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 NaBF 4 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 Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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-Na 2 Fe 2 (CN) 6 .2H 2 O//graphite, a R-Na 2 Fe 2 (CN) 6 //graphite and a R-Na 2 Fe 2 (CN) 6 //Na 2 Ti 3 O 7 Na 3-x Ti 3 O 7 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.