Rechargeable Sodium All-Solid-State Battery.

A reversible plating/stripping of a dendrite-free metallic-sodium anode with a reduced anode/ceramic interfacial resistance is created by a thin interfacial interlayer formed in situ or by the introduction of a dry polymer film. Wetting of the sodium on the interfacial interlayer suppresses dendrite formation and growth at different discharge/charge C-rates. All-solid-state batteries were obtained with a high cycling stability and Coulombic efficiency at 65 °C.

Introduction

Rechargeable Na-ion batteries have a cost advantage for large-scale stationary storage of electric power generated by solar and/or wind power owing to the widespread availability of Na in the oceans.[1−5] To meet the requirements of high energy density, safety, and long cycle life, it is necessary to replace the hard-carbon anode of the Na-ion battery by a metallic-sodium anode that can be plated/stripped reversibly without dendrites from a solid electrolyte; the electrolyte should have an energy gap large enough that the electrolyte is not reduced beyond a thin surface layer that allows a low impedance for Na+ transfer between the anode and electrolyte.[6−13] The solid electrolyte needs to be wetted by metallic sodium and to have a Na+ conductivity σNa ≳ 10–4 S cm–1 at the operating temperature, which would preferably be room temperature. To date, ceramic Na+ electrolytes, Na super ionic conductor (NASICON) with the Na1+3Zr2(P1–SiO4)3 structure, have reached a σNa ≃ 10–3 S cm–1 at T ≳ 65 °C, but they have been plagued by a large interfacial impedance; and anode dendrite formation is followed by rapid penetration of grain boundaries (Figure a).[14,15] Moreover, the volume changes of the electrodes over a charge/discharge cycle have prevented retention over many cycles of a solid–solid electrode–electrolyte interface. The interface problems of an alkali-metal/ceramic contact over many charge–discharge cycles may be largely overcome by the existence of a stable interface interstitial interlayer that is wetted by the anode and is a conductor of the working ion.[16,17] This observation has motivated an exploration of two avenues for creating the interface interstitial layer for a rechargeable sodium anode with a NASICON ceramic,[18,19] Na1+3Zr2(P1–SiO4)3 with x ≈ 2/3: (1) in situ formation of a Na+-conductive thin layer and (2) introduction of a dry Na+-conductive thin polymer layer. As illustrated in Figure b, the grain boundaries of the ceramic do not contact directly the sodium anode with an interfacial interlayer that undergoes homogeneous wetting by the sodium anode.

Figure 1

Contact model of ceramic pellet solid electrolyte and sodium metal with (a) a poor wetting ability ceramic pellet and (b) a good wetting ability artificial interlayer during the plating of sodium.

Results and Discussion

The ceramic pellets of Na3Zr2(PO4)(SiO4)2 were prepared by spark-plasma sintering (SPS) as previously reported;[18,20] the XRD spectra of the pellets verified single-phase formation of the NASICON structure. Scanning electron microscopy (SEM) images of the NASICON pellets are shown in Figure S1; numerous nanospheres of the ceramic were homogeneously melted and bonded together to form a solid substrate; no evidence of cracks or interparticle pores was observed in the thin pellets.

Wetting with an Interlayer Formed in Situ

NASICON Na3Zr2Si2PO12 gradually reacts with molten sodium at temperatures over 300 °C.[21,22] However, a Na3Zr2Si2PO12 membrane is quite stable toward sodium metal at lower temperatures. After heating of the sodium metal on the surface of the NASICON pellet at 175 °C for 5 h, no obvious change can be observed (Figure a). At 175 °C, liquid sodium metal stays on the NASICON pellet surface as a bead, indicating that the sodium metal has a stronger bonding with itself than with the surface of a NASICON pellet. On the other hand, a black layer on the surface of NASICON was formed when the pellet and sodium metal were heated at 380 °C for 30 min (Figure b and Figure S2). Moreover, the sodium metal spread out over the surface of the NASICON pellet, indicating that the black interlayer is wetted by the sodium metal.

Figure 2

Optical photos of sodium metal on a NASICON pellet (a) at 175 °C and (b) at 380 °C. XPS data of (c) Zr, (d) Si, and (e) P before and after reaction with Na metal at 380 °C for 0.5 h. (f) The XRD spectra of the Na3Zr2Si2PO12 NASICON (black line), H-NASICON (red line), and heat treated NASICON powder with Na metal for 2 h (blue line) and then heated at 950 °C for 2 h in air (wine line).

X-ray photoelectron spectroscopy (XPS) of the 380 °C heat-treated NASICON (H-NASICON) pellets in the presence of sodium metal exhibited an obvious shift to lower energy of the elements of Zr, Si, and P (Figures c–2e). The Zr4+ 3d3/2 and 3d5/2 peaks at 185.02 and 182.63 eV in Na3Zr2Si2PO12 shifted, respectively, to 184.08 and 181.67 eV on the H-NASICON. The Si 2p peaks shifted from 102.02 to 101.00 eV. The P 2p peak at 133.3 eV shifted to 132.4 eV. The only exception was the oxygen; no significant O2– 2p binding energy difference was detected in H-NASICON (Figure S3). These XPS peak shifts on P, Si, and Zr indicate that the surface of the H-NASICON was partially reduced by the sodium metal to form an in situ interlayer. The XRD pattern of the H-NASICON with the black layer did not show extra signals compared with the Na3Zr2Si2PO12 (red line in Figure f). To get more information on the black species, the NASICON powder was treated with sodium metal at 380 °C for 2 h and XRD was then taken. Much weaker NASICON diffraction signals were detected without an additional signal of a new phase although all the powder became black-brown (blue line in Figure f), indicating that the black species is amorphous. In addition, we found that the black-brown powder returned white and the NASICON phase was recovered by heating the H-NASICON samples at 950 °C for 2 h in air (wine line in Figure f). These phenomena verified that the H-NASICON was partially reduced by sodium metal at high temperature.

In order to investigate the electrochemical performance of the NASICON pellet as a solid electrolyte for Na-ion batteries, typical 2032 coin cells were assembled in a glovebox. The impedance spectra of the Na/NASICON/Na and Na/H-NASICON/Na cells were first studied and are presented in Figure a; compared with the spectrum of the Na/NASICON/Na, the resistance of the Na/H-NASICON/Na is obviously lowered from 4000 to 400 ohm cm–2, which can be attributed to the better wetting and adhesion across the Na/H-NASICON interface.

Figure 3

(a) The impedance plots of Na/NASICON/Na and Na/H-NASICON/Na symmetric cell at 65 °C. (b) CV curve of a Na/H-NASICON/gold-foil at a scanning rate of 0.5 mV s–1. (c) Cycling stability test of the Na/H-NASICON/Na symmetric cells at 65 °C. (d, e) Surface SEM images of sodium metal anode after cycling in the Na/H-NASICON/Na cell.

The electrochemical stability window of the solid electrolyte was then examined in a cell of Na/H-NASICON/gold-foil. The positive scan of the cyclic voltammetry (CV) curves in Figure b shows that the electrolyte did not experience obvious oxidation until 5 V, which indicates that the electrolyte membrane could be stable under 5 V. This electrochemical working window is capable of satisfying the requirement of most of the sodium-ion cathodes since they are generally operated at lower than 4.2 V. On the negative scan, the H-NASICON electrolytes experienced a symmetric lithium plating–stripping curve at −0.5–0.5 V; the sodium plating–stripping curves were similar, indicating a good electrochemical efficiency.

The electrochemical stability and whether anode dendrites form at the electrolyte interface were then evaluated in sodium/sodium symmetric cells by periodically charging for 1 h and discharging for 1 h. Figure c and S4 show the time-dependent voltage profiles for Na/Na symmetric cells with the H-NASICON and NASICON pellet electrolyte under a constant current density of 0.15 mA cm–2 and then increased to 0.25 mA cm–2. As shown in Figure c, the cell with the H-NASICON delivered stable sodium plating–stripping cycles for up to 550 h; the voltage did not show any significant increase/decrease, indicating that either the dendrite growth was blocked or dendrite formation was suppressed in the Na/H-NASICON interface during long-term cycling. On the contrary, the Na/NASICON/Na symmetric cells short-circuited in only 1 h under a current density of 0.15 mA cm–2 owing to a rapid dendrite formation and penetration of the grain boundaries (Figure S4), similar to other lithium ceramic electrolytes such as the garnet Li7La3Zr2O12 (LLZO).[14,15] Interestingly, we found that the ceramic pellets synthesized by SPS showed an even faster short-circuit in the Na/Na symmetric cells than the symmetric cells with pellets synthesized by a conventional sintering process in a regular heating oven (Figures S5 and S6), which can be attributed to the formation of narrower grain boundaries in pellets synthesized by SPS. Since the sodium tends to plate on the grain boundaries where the electric field is enhanced locally during charge, the narrower grain boundaries only allow the growth of thinner dendrites and therefore lead to a more rapid dendrite penetration. Figures d and 3e give the SEM images of sodium metal after cycling in a Na/H-NASICON/Na cell; uniform sodium plating can be observed without obvious sodium dendrite formation although the sodium-metal surface showed some cracks after cycling owing to the volume expansion during plating.[23,24] On the contrary, uneven sodium plating and obvious dendrite growth were clearly observed on the surface of the sodium metal after cycling in a Na/NASICON/Na cell (Figure S7). These phenomena indicate that the interlayer formed in situ between sodium metal and ceramic pellets not only effectively lowers the interfacial resistance but also successfully suppresses dendrite formation owing to a better wetting of the interlayer; wetting by the sodium metal gives a more uniform sodium flux across the interface.

The H-NASICON pellet electrolyte was then tested in a NaTi2(PO4)3/H-NASICON/Na all-solid-state cell since NaTi2(PO4)3 has been shown to have a stable capacity during long-term sodiation–desodiation in both liqiud[25] and polymer[26] electrolytes. The NaTi2(PO4)3 cathode membranes were prepared with cross-linked poly(ethylene glycol) methyl ether acrylate (CPMEA) as a Na+-conducting binder and carbon black as the electron conductor. The CPMEA consists of an electrochemically inert polyacrylate framework and soft oligo-ethylene oxide pendants, which realizes an acceptable solid-state ionic conductivity through a swing of the pendant groups in the polymer framework.[16] The ionic conductivity of the CPMEA shows a gradual increase with increasing temperature, reaching 0.7 × 10–4 S cm–1 at 65 °C owing to the increasing movement of the oligo-ethylene oxide pendants in the polymer network (Figure S8). Figures a and 4b show, respectively, the charge/discharge voltage profiles and the cycling performance of a cell at different rates at 65 °C. The one flat plateau at 2.1 V and a capacity of around 110 mAh g–1 at 0.2 C are the characteristic behavior of NaTi2(PO4)3 electrodes, indicating that the H-NASICON can function effectively as a solid electrolyte in an all-solid-state sodium battery. In long cycling, the discharge capacity is generally around 110 mAh g–1 at 0.2 C rate during the first 25 cycles and is slightly higher than 75 mAh g–1 at 1 C during the following 35 cycles. When the C rate was switched to 0.5 C, a capacity of 94 mAh g–1 could be recovered. It is remarkable that the Coulombic efficiency is always kept at 99.8 ± 0.2% at all three C-rates after the initial formation cycles, indicating high sodium plating/stripping efficiency and good electrochemical stability across the Na/H-NASICON interface. More importantly, this long cycling stability also verifies the good capability of dendrite suppression at a Na/H-NASICON interface since the cell should have short-circuited if a dendrite had penetrated across the H-NASICON pellet.

Figure 4

(a) Charge and discharge voltage profiles and (b) cycling performance of NaTi2(PO4)3/Na cell with a H-NASICON pellet electrolyte at different C-rates and 65 °C.

Wetting with a Dry Polymer Interlayer

We have reported a sandwich electrolyte of polymer/ceramic-pellet/polymer to plate a dendrite-free lithium anode.[16] The polymer used in our previous report was CPMEA, the same polymer used in this study. The polymer interlayer between lithium and ceramic was wetted well by the lithium metal, which homogenized the Li-ion flux across the interface. An all-solid-state Li/CPMEA/ceramic/CPMEA/LiFePO4 battery with a notably high Coulombic efficiency of 99.8%–100% over 600 cycles was obtained. To investigate further whether this sandwich electrolyte is effective in a sodium-based all-solid-state battery, a CPMEA/NASICON/CPMEA sandwich electrolyte was fabricated with the previous method. Figure a shows the impedance spectra of a Na/CPMEA/NASICON/CPMEA/Na cell, which is also obviously lower than that of the Na/NASICON/Na, indicating that the polymer inerlayer improves the ionic conductivity across the solid Na/NASICON interface. The resistance is a little higher than that of the Na/H-NASICON/Na, which can be attributed to the internal resistance of the polymer layer and a new interfacial impedance across CPMEA/NASICON interfaces. The symmetric cell of Na/CPMEA/NASICON/CPMEA/Na was then tested to evaluated the polymer stability and dendrite suppression performance on repeated sodium plating and stripping, as is shown in Figure b. The cell gave a stable voltage profile for up to 380 h under a constant current density of 0.20 mA cm–2 after an obvious voltage decrease during the initial 2 h owing to an interfacial wetting process, indicating that the sodium dendrite was also suppressed by the polymer interlayer and the Na/CPMEA/NASICON interface is stable during repeated sodium plating–stripping cycles.

Figure 5

(a) The impedance spectra of a Na/Na cell with a CPMEA/NASICON/CPMEA electrolyte. (b) Voltage profiles of a symmetric Na/Na cell with a CPMEA/NASICON/CPMEA sandwich electroyte cycled at 65 °C. (c) Charge–discharge voltage profiles and (d) cycling performance of a NaTi2(PO4)3/Na cell with CPMEA/NASICON as the electrolyte at 0.2 C and 65 °C.

The stability of the solid sandwich electrolyte was then further evaluated in NaTi2(PO4)3/Na cells with a CPMEA/NASICON double layer electrolyte since the cathode films were prepared with the polymer electrolyte as the binder. As shown in Figures c and 5d, a stable capacity of around 102 mAh g–1 was obtained for 70 cycles at 0.2 C and 65 °C; the Coulombic efficiency remained at 99.7 ± 0.3% during this cycling, indicating good stability of the Na/CPMEA/NASICON interfaces and capability of dendrite suppression during the charge/discharge process, which are comparable to the Na/H-NASICON interface.

A large interface resistance and severe dendrite formation across an alkali-metal/ceramic interface are two critical factors that have restricted the application of ceramic electrolytes in all-solid-state batteries. If the untreated NASICON pellets are employed, the sodium does not wet the ceramic electrolyte and, therefore, uniform Na+ flux across the interface is not established. As a result, the Na+ ions plate preferentially on the grain boundaries where the Na+ flux is locally enhanced under an electronic field and dendrites nucleate and grow there during a charge. Since the grain boundaries are usually interconnected, the dendrites can penetrate the ceramic in a short time.[14,15] With an interface interlayer, a homogeneous ceramic/interlayer/sodium interface gives a more uniform Na-flux across the interface since the sodium metal wets better the interlayer than the ceramic. Moreover, the anode–interlayer bonding restricts expansion/contraction of the anode during plating/stripping to be perpendicular to the interface to give the interface a long cycle life.

We have presented here two strategies to fabricate an interface interlayer between a Na metal and a ceramic NASICON electrolyte: heat treated NASICON in the presence of Na metal and a polymer/NASICON/polymer sandwich. Better wetting of the electrolyte by the sodium anode gives a uniform Na+ flux across the interface, which not only dramatically lowers the interface resistance but also successfully suppresses unwanted dendrite formation. A good cycling performance and high efficiency of all-solid-state sodium cells were demonstrated with two types of strategies for good electrochemical stability and high sodium plating/stripping efficiency across a Na/ceramic interface.