Sodium plating and stripping from Na- b " -alumina ceramics beyond 1000 mA/cm 2 Materials Today Energy

Dendrite formation limits the cycle life of lithium and sodium metal anodes and remains a major challenge for their integration into next-generation batteries, even when replacing the liquid electrolyte by a solid electrolyte. Voids forming in solid metal anodes at the interface to a solid electrolyte on stripping cause current constrictions on plating and promote dendrite formation. Recent studies showed that alkali metal creep is the primary mechanism for replenishing the voids at room temperature. Here, we investigate plating and stripping of liquid sodium metal from a carbon-coated ceramic Na- b " -alumina electrolyte at 250 (cid:1) C, thereby eliminating creep-related mass transport limitations. We demonstrate extremely high current densities of up to 2600 mA/cm 2 and cumulative plating capacities of > 10 Ah/cm 2 at 1000 mA/cm 2 without dendrite formation. Our results demonstrate that liquid metal anodes can be paired with solid electrolytes, providing a practical solution to suppress dendrite formation at high current densities. © 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license offering signiﬁcantly enhanced energy density. Dendrites do not only form in battery cells with liquid electrolytes but also an analogous phenomenon is observed in cells with solid electrolytes, causing premature cell failure. During discharge, stripping from lithium and sodium metal anodes results in the formation of voids at the interface to the solid electrolyte, which cause current constrictions and promote dendrite formation. Here, we conﬁrm experimentally that avoiding the formation of such voids by operating the anode at temperatures higher than its melting temperature enables and plating at unprecedented current densities. At an extremely high current density of 1000 mA/cm , we are able to and plate a of metal anodes,


a b s t r a c t
Dendrite formation limits the cycle life of lithium and sodium metal anodes and remains a major challenge for their integration into next-generation batteries, even when replacing the liquid electrolyte by a solid electrolyte. Voids forming in solid metal anodes at the interface to a solid electrolyte on stripping cause current constrictions on plating and promote dendrite formation. Recent studies showed that alkali metal creep is the primary mechanism for replenishing the voids at room temperature. Here, we investigate plating and stripping of liquid sodium metal from a carbon-coated ceramic Na-b"-alumina electrolyte at 250 C, thereby eliminating creep-related mass transport limitations. We demonstrate extremely high current densities of up to 2600 mA/cm 2 and cumulative plating capacities of >10 Ah/cm 2 at 1000 mA/cm 2 without dendrite formation. Our results demonstrate that liquid metal anodes can be paired with solid electrolytes, providing a practical solution to suppress dendrite formation at high current densities.

Context and scale
Dendrite formation remains the major challenge to enable next-generation batteries, with lithium and sodium metal anodes offering significantly enhanced energy density. Dendrites do not only form in battery cells with liquid electrolytes but also an analogous phenomenon is observed in cells with solid electrolytes, causing premature cell failure. During discharge, stripping from lithium and sodium metal anodes results in the formation of voids at the interface to the solid electrolyte, which cause current constrictions and promote dendrite formation. Here, we confirm experimentally that avoiding the formation of such voids by operating the anode at temperatures higher than its melting temperature enables stripping and plating at unprecedented current densities. At an extremely high current density of 1000 mA/cm 2 , we are able to strip and plate a cumulative capacity of >10 Ah/cm 2 , corresponding to the US Department of Energy fast charging target for metal anodes, without any signs of dendrite formation. Thus, controlling the mechanical properties of metal anodes holds the key to reach this target also at room temperature.

Introduction
Lithium and sodium metal anodes are considered the 'holy grail' for lithium and sodium batteries as they provide the lowest chemical potential and high volumetric capacity (À3.04 V vs standard hydrogen electrode (SHE) and 2062 mAh/cm 3 for lithium and À2.71 V vs SHE and 1129 mAh/cm 3 for sodium at room temperature) [1e5]. However, the realization of reversible plating and stripping without dendrite formation over many cycles remains a major challenge for both liquid and solid electrolytes [6,7]. The US Department of Energy currently sets the fast charging goal for lithium metal anodes at a cumulatively cycled capacity of 10 Ah/ cm 2 at a cycling rate of 10 mA/cm 2 , ideally with no excess of (non-cycled) lithium in the cell [8]. So far, most studies have focused on plating and stripping at room temperature.
Much progress has been made in recent years in controlling the interface between lithium and sodium metal anodes and the electrolyte. In particular, it was shown that the critical current density for dendrite formation with solid electrolytes scales with interfacial resistance [9e11]. The latter can be reduced by thermal treatment of the electrolyte surface or by applying interfacial coatings [12e19]. However, even when the interfacial resistance becomes negligible (<10 Ucm 2 ), it is not yet possible to sustain a cycling rate of 10 mA/cm 2 over an extended period without dendrite formation. Some of us recently compared plating and stripping of lithium vs sodium metal at temperatures lower than their melting temperature from ceramic electrolytes under otherwise identical experimental conditions and observed that the critical current density for sodium is about one order of magnitude higher than for lithium, leading to the hypothesis that the mechanical properties of lithium and sodium are governing dendrite formation [10]. Based on the pressure dependence of stripping, Kasemchainan et al [20] and Wang et al [21] concluded that lithium and sodium metal creep (or cold flow) behavior rather than lithium and sodium diffusion is the primary mechanism for replenishing the voids forming in lithium and sodium metal anodes at the interface to the solid electrolyte interface on stripping. Avoiding the formation of such voids during stripping is key to suppress dendrite formation during plating.
To circumvent creep-related mass transport limitations, we investigate plating and stripping of sodium at temperatures higher than its melting temperature from the archetypical Na-b''-alumina solid electrolyte. Carbon coatings are applied to prevent dewetting of liquid sodium from the sodiophobic Na-b''-alumina surface. We demonstrate that plating and stripping is possible at extremely high current densities of 2600 mA/cm 2 and cumulative plated capacities of 10 Ah/cm 2 at 250 C without dendrite formation. Thus, using a liquid sodium metal anode as in commercial sodium-sulfur (NaS) and sodium-nickel chloride (Na-NiCl 2 ) batteries is very effective in preventing dendrite formation. Furthermore, we perform open-cell experiments to assess the amount of sodium stored within the porous carbon coating and determine the retrievable amount of sodium during stripping. We show that these coatings efficiently supply sodium to the Na-b''-alumina surface, minimizing the excess of active material in the electrode. We further characterize how coating thickness and operating temperature affect the rate performance of the liquid sodium metal anode.  functions: (1) it acts as a porous electrode that contacts the electrolyte in the absence of sodium, (2) it prevents dewetting of liquid sodium from the sodiophobic Na-b"-alumina surface, and (3) it stores and supplies liquid sodium in and from its pores during cell operation at elevated temperatures. Such coatings are also used in high-temperature Na-NiCl 2 batteries to ensure sodium coverage of the entire electrolyte area at all states of charge [22]. Crosssectional scanning electron microscopy (SEM) images of the carbon coating reveal a porous structure, comprising carbon particles with a typical size of 0.1 mm (Fig. 1B). Depending on the number of spray passages applied, average coating thicknesses of approximately 50 and 200 mm with a relative thickness variation of ±15%

Results and discussion
were obtained ( Fig. 1C and D). From the weight difference before and after coating application, the film thickness from SEM images, and assuming a bulk density of 2.5 g/cm 3 (40 wt% carbon and 60 wt % sodium hexametaphosphate), a coating porosity of~80% is deduced. A schematic of the symmetrical cell design is shown in Fig. 2A and Fig. S1. The cell consists of the carbon-coated Na-b"-alumina solid electrolyte disk, sandwiched between two sodium foils and cylindrical nickel current collectors. Fig. 2B compares the electronic and ionic conductivity of the cell components at 250 C. The electronic conductivity of the carbon coating of 1.2 S/cm is only one order of magnitude higher than the ionic conductivity of the Na-b''alumina electrolyte (0.12 S/cm, [26]), whereas the conductivity of liquid sodium and the nickel current collector are four to five orders of magnitude higher [24,25]. While the cell components ( Fig. 2A) are generally coupled in series, the conductivity of the carbon coating is only of relevance in the absence of sodium. When liquid sodium starts to fill its pores on plating, the internal resistance of the cell is dominated by the resistance of the Na-b''-alumina electrolyte and passive cell components.
We first cycled the cell galvanostatically at a temperature of 250 C, transferring 25 mAh/cm 2 per half-cycle. The current density was increased stepwise from 200 mA/cm 2 to 2600 mA/cm 2 for each plating/stripping cycle (Fig. 2C). Owing to the fast charge transfer kinetics at the interface between the Na-b''-alumina electrolyte and the sodium metal electrodes, the voltage in Fig. 2D remains linearly dependent on the current density up to the highest value of 2600 mA/cm 2 . This exceeds practical current densities for batteries by at least one order of magnitude (e.g. in a Na-NiCl 2 cell with 40 Ah and 200 cm 2 electrolyte area, the current density during 1C pulse operation reaches 200 mA/cm 2 [27], and for lithium-ion batteries, the current density rarely exceeds 10 mA/cm 2 ) [8]. Using Ohm's law, we extract a cell resistance of 1.3 Ucm 2 . Taking into account the ionic conductivity of the Na-b''-alumina electrolyte (Fig. 2B), it contributes about 0.8 Ucm 2 , whereas the remaining contribution stems mainly from the connectors and connection cables. Information on cell impedance evolution after cell assembly and cycling can be found in Fig. S2A (supplemental information). Note that high current densities of, e.g. 2600 mA/cm 2 generate a considerable amount of Joule heating (56 W per cm 3 of the electrolyte). Consequently, the cell temperature was stabilized for each measurement to 250 C to ensure that the cell resistance can be extracted from the graph's slope in Fig. 2D. Stabilizing the cell thermally at even higher current densities was increasingly difficult. Nevertheless, the cycling data ( Fig. 2C) indicate no signs of dendrite formation or short circuits, even at current densities of 2600 mA/cm 2 . Fig. 2E shows a magnified view of the cycle at 2600 mA/cm 2 . At such high current density, we initially observe a slight increase in cell voltage, which then stabilizes after the first 5 s (  behavior is not seen for half-cycles at negative currents. We ascribe this to insufficient sodium supply on stripping from the top-side compartment, caused by sodium loss within the cell (e.g. squeezed out behind the nickel current collector). Although managing the substantial volume changes of the sodium anode in our test cell is challenging, technical solutions based on metal shims are available in commercial Na-NiCl 2 batteries [22]. The entire cycling of the symmetric cell over 1400 cycles, including run-in cycles and end of life (due to voltage limitations), is shown in Fig. S2B (supplemental information). The evolution of galvanostatic overpotential traces during plating and stripping in symmetric cells was discussed in detail in the study by Wood et al [28] for solid lithium metal electrodes operated near room temperature, which are prone to dendrite formation at current densities on the order of 10 mA/cm 2 . Cells with solid lithium (and sodium) metal electrodes tend to exhibit a  much more pronounced 'peaking' behavior at the start of each half-cycle, which was associated with the initial activation barrier for nucleation on the cathode [28e32]. However, we do not observe such behavior for liquid sodium electrodes with a graphite coating, indicating that dendrite formation does not play a role, even at current densities of 2600 mA/cm 2 or cumulative plating capacities of 10 Ah/cm 2 .
To investigate sodium filling into and extraction from the porous carbon coating in more detail, we used an open cell design with optical access to the top-side carbon coating (Fig. 3A, white arrow). A nickel ring electrode replaced the top-side cylindrical current collector, and the top-side sodium foil was eliminated to allow visual access by a camera. Owing to the high conductivity of the porous carbon coating of 1.2 S/cm at 250 C (Fig. 2B), the current distribution remains relatively uniform across the surface, especially once the pores are filled with liquid sodium metal. Thermography on the open cell was conducted to confirm accurate temperature readings of the built-in thermocouple and even temperature distribution across the entire electrolyte surface in the open configuration (Fig. S3). Fig. 3B illustrates the cycling protocol applied to the open cells. At the beginning of each cycle, the topside carbon coating is filled by plating 25 mAh/cm 2 of liquid sodium, corresponding to a nominal (dense) liquid sodium metal thickness of 242 mm, taking into account the volumetric capacity of liquid sodium metal of 1035 mAh/cm 3 at 250 C [33]. Plating is performed at a current density of 30 mA/cm 2 . Sodium extraction from the carbon coating is evaluated at constant current (CC) densities (10 mA/cm 2 to 600 mA/cm 2 ) until a cell potential of 2.5 V is reached. Subsequently, the cell is soaked at a constant voltage (CV) of 2.5 V until the current density falls below 0.1 mA/cm 2 .
In Fig. 4, we show examples of top-view photographs of different open cell experiments taken during the sodium stripping process at 100 mA/cm 2 (see Fig. S4 in the supplemental information for enlarged views of Fig. 4J, N). For each experiment, we depict the situation after initial plating of 25 mAh/cm 2 , at the end of CC stripping, and at the end of CV stripping, as per the corresponding electrochemical data. After the initial plating, excess sodium is contained in large drops on top of the coating (red arrows). We visually identify when these large sodium drops are consumed on stripping (Fig. 4B, F, J, N), which we refer to as saturation of the carbon coating. The remaining capacity at saturation thus corresponds to the amount of sodium that can be stored in the pores of the carbon coating and in the carbon itself.
In the first column (Fig. 4AeD), we illustrate the phobicity of the Na-b"-alumina surface to liquid sodium metal at 250 C. In this cell, the 50-mm-thick carbon coating on the top side was applied only to a 4-mm-wide outer ring of the Na-b"-alumina disk, leaving the center uncoated. After plating of 25 mAh/cm 2 , corresponding to a nominal sodium metal thickness of 377 mm on the coated ring area, the liquid sodium metal remains confined to the coated area of the Na-b"-alumina disk. Owing to poor wetting of liquid sodium metal on Na-b"-alumina, the uncoated center remains free from sodium (Fig. 4A). The large drop adopts a convex shape toward the center to minimize its contact area with the uncoated, sodiophobic Na-b"- alumina surface. Instead, wetting of the nickel ring electrode is favored, as indicated by the spreading of the large drop along the ring electrode. The second and third column (Fig. 4EeH, IeL) show the corresponding cell with 50-mm-and 200-mm-thick carbon coatings, covering the entire Na-b"-alumina disk surface. Here, plating of 25 mAh/cm 2 corresponds to a nominal sodium metal thickness of 242 mm. In all experiments at 250 C, small sodium droplets in and on the carbon coating lead to a metallic shine of the electrode after plating (Fig. 4A, E, I). This is in contrast to the situation at 140 C depicted in the fourth column (Fig. 4MeP), where the carbon coating appears mainly black. Excess sodium contained in large drops after plating of 25 mAh/cm 2 features a smaller footprint at 140 C than at 250 C, indicating reduced wetting of the carbon coating by liquid sodium at the lower temperature. Despite the different visual appearance at 250 C and 140 C, the 200-mm-thick carbon coating contains a similar amount of sodium at saturation (11 mAh/cm 2 and 9.5 mAh/cm 2 , respectively, Fig. 4J, N). The capacity at saturation scales with thickness, but also with the area of the coating (2 mAh/cm 2 , Fig. 4B, and 2.5 mAh/cm 2 , Fig. 4F). Thus, the carbon coating holds 5 mAh/cm 2 of liquid sodium per 100-mm coating thickness in all cases shown in Fig. 4. Taking into account the volumetric capacity of liquid sodium metal [33], this corresponds to a filling of the porous carbon coating of~65 vol%.
On continued galvanostatic stripping, sodium is gradually removed from the carbon coating. At the end of the CC step, the metallic shine of sodium is partially visible through the 50-mm coating at 250 C (Fig. 4C, G), whereas the 200-mm coating appears black (Fig. 4K, O). For the thinner coating, brighter areas remain where large drops were located before (Fig. 4C, G), indicating inhomogeneous transport through the carbon coating. However, sodium is extracted also from these regions during the CV step (Fig. 4D, H). Additional top-view recordings linking visual appearance with CC capacity and cycle time are available as supplemental information for the different coating thicknesses (Video 1) and operating temperatures (Video 2).
Supplementary data related to this article can be found online at https://doi.org/10.1016/j.mtener.2020.100515 Fig. 5 summarizes the electrochemical results obtained with the open cell during stripping at 250 C as a function of coating thickness and current density (see Fig. S5 for an overview on cell voltage and capacity evolution with time). A schematic of the open cell is shown in Fig. 5A. Fig. 5B shows the coulombic efficiency of the open cell with thin and thick coatings for CC stripping at current densities ranging from 10 to 600 mA/cm 2 , followed by CV soaking at 2.5 V until the current drops to 0.1 mA/cm 2 . The maximum variation over five consecutive cycles is given for each current density by the error bars. At low stripping current densities of 10 mA/cm 2 , the coulombic efficiency is 99.2%. We attribute this to slow, irreversible chemical oxidation of the highly reactive liquid sodium surface by residues in the argon glove box atmosphere during the long cycle time of 2.8 h. For higher stripping current densities of 300 and 600 mA/cm 2 resulting in much shorter cycle times, an excellent coulombic efficiency of 99.8% is achieved, demonstrating that the carbon coating is an effective solution for supplying liquid sodium metal to the sodiophobic Na-b''-alumina surface. Comparison of the results of the 50-and 200-mm-thick coating in Fig. 5B shows that the coating thickness does not significantly affect the coulombic efficiency. In Fig. 5C, we display the capacity retrieved from the carbon coating at CC only (not including the CV capacity, see Fig. S5 for detailed stripping curves and coulombic efficiency without CV soaking). Comparison of the results of the 50-and 200-mm-thick coating reveals that the capacity retrieved during the CC step is strongly dependent on the carbon coating thickness. Especially at high CC densities, up to 1.64 mAh/cm 2 , more capacity is retrieved with the 50mm-thick coating than with the 200-mm-thick coating. We ascribe this to the increased flow resistance of liquid sodium in the thicker porous coating. Specifically, at 600 mA/cm 2 , 96% of the capacity is retrieved under CC stripping with the 50-mm-thick coating, whereas only 90% of the capacity is accessible with the 200-mm-thick coating. Fig. 5D compares the CC and CV stripping times, t, for the two coating thicknesses at different CC stripping rates. The CC stripping time is linearly dependent on the CC stripping rate. The CC stripping time for the 200-mm coating is slightly shorter than for the 50mm coating owing to the lower CC capacity extracted with the thicker coating (see Fig. 5C). The CV stripping time is not significantly affected by the CC stripping current density, although the capacity extracted at CV differs significantly at high CC (compare with Fig. 5B). The 50-mm-thick coating accelerates the CV step by a factor of 3 compared with the 200-mm-thick coating because of reduced flow resistance in the porous coating.
In Fig. 6, we investigate the effect of temperature on the stripping behavior with the 200-mm-thick coating. Coulombic efficiency and CC retrievable capacity shown in Fig. 6A and B are affected only slightly by temperature. However, the CV stripping time shown in Fig. 6C increases by almost a factor of 4, when the temperature is reduced from 250 to 140 C. We ascribe the increase in CV stripping time to the 20% increase in dynamic viscosity of liquid sodium when reducing the temperature to 140 C [34], resulting in an increase in flow resistance in the porous coating. Detailed stripping curves for 10, 30, and 100 mA/cm 2 can be found in Fig. S6. Fig. 6D depicts schematically the situation at 250 and 140 C. At 250 C, numerous small sodium droplets cover the top of the carbon coating, some of which coalesce into a large, millimeter-sized drop with good wetting to the coating. In contrast, at 140 C, few small sodium droplets are visible on the top of the carbon coating, and the plated sodium appears as a single large drop with poor wetting to the coating. At both temperatures, the large drop is consumed during CC stripping. At saturation, small sodium droplets remain visible on top of the coating at 250 C, whereas very few are visible at 140 C. Despite the different surface appearance, sodium filling within the coating and CC capacity is similar. Extraction of sodium remaining in the pores of the carbon coating and the carbon itself during the CV stripping step requires more time at 140 C owing to the increase in flow resistance resulting from the increase in sodium viscosity.

Conclusion
In summary, we investigated the plating and stripping behavior of liquid sodium from ceramic Na-b''-alumina electrolytes in symmetric closed and asymmetric open-cell configurations. Critical for the operation of liquid sodium electrodes is the porous carbon coating applied to the Na-b''-alumina electrolyte. Besides acting as a porous electrode providing electrons to the Na-b''-alumina surface, the porous carbon coating serves as a reservoir to store and supply liquid sodium to the Na-b''-alumina surface and allows wetting of the sodiophobic Na-b"-alumina surface. On the one hand, the storage capacity increases with the coating thickness (approximately 5 mAh/cm 2 per 100-mm coating thickness). On the other hand, thinner carbon coatings enable higher CC current rates and reduce the CV soaking time by minimizing the flow resistance of liquid sodium in the porous carbon coating.
For practical applications, the sodium in the porous carbon coating should not be depleted completely during cycling to avoid additional run-in cycles. Our results show that for a 50-mm-thick carbon coating that can hold up to 2.5 mAh/cm 2 of liquid sodium in its pores at saturation, CC stripping at 600 mAh/cm 2 can be maintained down to a filling level of 1 mAh/cm 2 . For a 25 mAh/cm 2 sodium metal anode, this corresponds to a small sodium excess of 4%. When considering a sodium metal anode with 5 mAh/cm 2 , which is an areal charge density typical for a lithium-ion battery, the excess is on the order of 20%, which is still a very low value compared with other alkali metal anodes reported in the literature [8]. Reducing the temperature from 250 C to 140 C mainly prolongs the CV soaking time, but does not affect the CC stripping capacity and time substantially.
No dendrite formation is observed for liquid sodium metal anodes up to current densities of 2600 mAh/cm 2 and 10 Ah/cm 2 cumulative plated capacity at 250 C. This indicates that dendrite formation is less related to the properties of the Na-b''-alumina electrolyte, but rather to the properties of the sodium metal anode. Thus, battery operation at increased temperatures is an efficient means enabling enhanced charge and discharge rates. For room-temperature applications, controlling the mechanical properties of solid alkali metal anodes, e.g. by alloying with another component, may hold the key for suppressing dendrites and enabling the commercialization of solid alkali metal anodes in next-generation batteries.

Experimental section
Na-b"-alumina disks with a thickness of 1 mm (symmetric cell) or 1.7 mm (open cells) and 35-mm diameter were sintered at 1600 C for 5 min from powders following the method used by Bay et al [23] and joined to two a-alumina rings using a glass paste. To prevent dewetting of liquid sodium from the Na-b"-alumina surface, a porous coating consisting of carbon black particles (7 wt%) and sodium hexametaphosphate (11 wt%) was applied by spray coating (2e8 passages) from a dispersion in isopropanol (55 wt%) and water (27 wt%) followed by drying in air at 280 C. The coating thickness varied between 50 and 200 mm on the top electrode and was fixed to 50 mm at the bottom electrode. SEM images were taken using a Hitachi S4800. The electronic conductivity of the carbon coatings was measured using a Keithley 2000 multimeter and 10mm-wide platinum electrodes sputtered onto the Na-b"-alumina disks at a distance of 2, 5, and 15 mm using the transfer length method on a hot plate at 250 C. The conductivity of the Na-b"alumina disks was measured by electrochemical impedance spectroscopy in four-point probe configuration inside a tube furnace using a Zahner IM6 impedance analyzer [23]. Electrochemical cells with spring-loaded cylindrical nickel current collectors and an active area of 3.14 cm 2 were assembled in an argon-filled glove box based on the ceramic Na-b"-alumina electrolyte disk and a-alumina rings (Fig. S1). A type K thermocouple was inserted at the interface between a-alumina and Na-b"-alumina for accurate cell temperature determination. Sodium metal foils with a purity of 99.9% and an areal mass of 0.16 g/cm 2 (0.5 g) were inserted between the Nab"-alumina and the current collectors.
To monitor sodium plating and stripping visually using a camera (DNT Digi Micro Scale, 1920 Â 1020 pixels, 10-s picture interval), an open cell was assembled by replacing the cylindrical top current collector by a nickel ring electrode. Thermography was conducted using a Hotfind LX camera. In a separate calibration experiment, the radiative emissivity of the carbon coating was determined to be 0.85 at a temperature of 250 C. In this calibration experiment, the surface temperature of a carbon-coated hot plate was accurately measured using a type K thermocouple, while simultaneously, thermographic images of the hot plate surface were taken. The radiative emissivity of the coating was finally determined by internally adjusting the emissivity settings of the thermography camera until the surface temperature measurements matched between the thermocouple measurement and the thermography measurement. Cell heating for all experiments was realized via a resistive coil heater wrapped around the cell, while cell temperature was controlled via a feedforward controller.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.