Rechargeable Na/Cl2 and Li/Cl2 batteries

Lithium-ion batteries (LIBs) are widely used in applications ranging from electric vehicles to wearable devices. Before the invention of secondary LIBs, the primary lithium-thionyl chloride (Li-SOCl2) battery was developed in the 1970s using SOCl2 as the catholyte, lithium metal as the anode and amorphous carbon as the cathode1–7. This battery discharges by lithium oxidation and catholyte reduction to sulfur, sulfur dioxide and lithium chloride, is well known for its high energy density and is widely used in real-world applications; however, it has not been made rechargeable since its invention8–13. Here we show that with a highly microporous carbon positive electrode, a starting electrolyte composed of aluminium chloride in SOCl2 with fluoride-based additives, and either sodium or lithium as the negative electrode, we can produce a rechargeable Na/Cl2 or Li/Cl2 battery operating via redox between mainly Cl2/Cl− in the micropores of carbon and Na/Na+ or Li/Li+ redox on the sodium or lithium metal. The reversible Cl2/NaCl or Cl2/LiCl redox in the microporous carbon affords rechargeability at the positive electrode side and the thin alkali-fluoride-doped alkali-chloride solid electrolyte interface stabilizes the negative electrode, both are critical to secondary alkali-metal/Cl2 batteries. Rechargeable Na/Cl2 and Li/Cl2 batteries are produced with a microporous carbon positive electrode, aluminium chloride in thionyl chloride as the electrolyte, and either sodium or lithium as the negative electrode.

Lithium-ion batteries (LIBs) are widely used in applications ranging from electric vehicles to wearable devices. Before the invention of secondary LIBs, the primary lithium-thionyl chloride (Li-SOCl 2 ) battery was developed in the 1970s using SOCl 2 as the catholyte, lithium metal as the anode and amorphous carbon as the cathode [1][2][3][4][5][6][7] . This battery discharges by lithium oxidation and catholyte reduction to sulfur, sulfur dioxide and lithium chloride, is well known for its high energy density and is widely used in real-world applications; however, it has not been made rechargeable since its invention [8][9][10][11][12][13] . Here we show that with a highly microporous carbon positive electrode, a starting electrolyte composed of aluminium chloride in SOCl 2 with fluoride-based additives, and either sodium or lithium as the negative electrode, we can produce a rechargeable Na/Cl 2 or Li/Cl 2 battery operating via redox between mainly Cl 2 /Cl − in the micropores of carbon and Na/Na + or Li/Li + redox on the sodium or lithium metal. The reversible Cl 2 /NaCl or Cl 2 /LiCl redox in the microporous carbon affords rechargeability at the positive electrode side and the thin alkali-fluoride-doped alkali-chloride solid electrolyte interface stabilizes the negative electrode, both are critical to secondary alkali-metal/Cl 2 batteries.
Devising new battery concepts with high gravimetric, volumetric and areal capacity and high energy density is important to meet society's growing demand for energy storage. Different rechargeable batteries have been developed, including lithium-ion batteries (LIBs), sodium-ion batteries (SIBs) and aluminium-ion batteries (AIBs) [14][15][16][17][18][19][20][21][22][23] . The primary lithium-thionyl chloride (Li-SOCl 2 ) battery is well known for its high energy density and wide use in professional electronics, military, utility metering and GPS tracking applications, but lacks rechargeability 10,11 . The battery performs a single discharge through Li anode oxidation and catholyte SOCl 2 reduction into sulfur (S), sulfur dioxide (SO 2 ) and a chloride ion (Cl − ) [1][2][3][4][5][6][7][8][9] . The Cl − ions react with Li + stripped from the Li anode to form LiCl deposited on the carbon surface until passivation. This primary battery can deliver a high specific capacity of about 2,300 mAh g −1 and a high energy density of up to 710 Wh kg −1 in a single discharge 8,9 . Sodium (Na) batteries have been actively pursued as an alternative to Li batteries due to the low standard electrode potential of Na (only about 0.34 V higher than Li), much lower price and equal promise for high-energy-density batteries 24,25 . There have been no reports of a Na-SOCl 2 primary battery as in the Li primary battery case, let alone rechargeable Na batteries in an SOCl 2 -based electrolyte.
Here we report a Na/Cl 2 battery using amorphous carbon nanospheres (aCNS) as the cathode and aluminium chloride (AlCl 3 ) in SOCl 2 as the main components in the starting electrolyte. The battery operates/cycles with a 3.5-V discharge voltage and up to 1,200 mAh g −1 capacity (based on the aCNS mass throughout this paper unless otherwise specified) over more than 200 cycles, with a Coulombic efficiency and energy efficiency (ratio of energy discharged over charging energy input per cycle) of greater than 99% and greater than 90%, respectively.
The battery delivered a first discharge capacity of about 2,800 mAh g −1 with an average discharge voltage of about 3.2 V. To our initial surprise, the battery could be cycled reversibly at a specific capacity of 1,200 mAh g −1 with a discharge voltage of about 3.55 V and an average Coulombic efficiency greater than 99% (up to 1,860 mAh g −1 cycling capacity with a lower Coulombic efficiency). The battery's first discharge led to NaCl formation on the aCNS positive electrode resembling LiCl in the Li-SOCl 2 primary battery. The carbon microstructure on the positive side and the fluoride-doped NaCl solid electrolyte interface (SEI) on sodium were found to be critical to subsequent reversible battery cycling, with the redox between NaCl and Cl 2 as the dominant reaction that contributed to the main reversible capacity of the battery. The same concept also led to a rechargeable Li/Cl 2 battery.
We constructed a battery using Na metal as the negative electrode and packed aCNS with a polytetrafluoroethylene (PTFE) binder in a nickel (Ni) foam as the positive electrode in a coin cell (Methods, Supplementary Fig. 2). The starting electrolyte was 4 M AlCl 3 dissolved in SOCl 2 mixed with 2 wt% sodium trifluoromethanesulfonimide (NaTFSI) and 2 wt% sodium bis(fluorosulfonyl)imide (NaFSI) additives (Fig. 1a). The as-made battery was first discharged to 2 V, exhibiting a capacity of about 2,810 mAh g −1 and two plateaus at about 3.47 V and 3.27 V (Fig. 1c), corresponding to Na discharge to NaCl to first Article neutralize the acidic electrolyte and then deposition of NaCl on the aCNS electrode, respectively (Supplementary Text 2). Through the second plateau, the discharged NaCl was deposited onto the pores of the aCNS in the positive electrode and on the surfaces of the nanospheres (Fig. 1e inset, Extended Data Fig. 1a, Supplementary Text 2, 3), which was confirmed by X-ray diffraction (XRD), scanning electron microscopy (SEM) (Fig. 1e, Extended Data Fig. 1a) and a large increase in the electrochemical impedance of the cell (Extended Data Fig. 2a). Mass spectrometry revealed the formation of SO 2 (Fig. 1d, see Supplementary Text 1 for all mass spectrometry experiments and results in this study), which was highly soluble in SOCl 2 without pressurizing the cell.
When re-charging the battery after the first discharge, Na was deposited on the Na electrode and the NaCl deposited on the aCNS electrode was oxidized (at about 3.83 V) (Fig. 2a) to form Cl 2 residing in the large volume of pores in the aCNS electrode (Extended Data Table 1). Oxidation of NaCl was confirmed by ex situ X-ray photoelectron spectroscopy (XPS) showing decreases in Na and Cl elements on the aCNS (Fig. 2b), removal of NaCl crystallites on the aCNS positive electrode to expose the underlying carbon nanospheres in SEM imaging (Fig. 2b inset and Extended Data Fig. 1b), and a decrease in XRD peaks of NaCl on the electrode (Fig. 2c). The charging voltage immediately spiked to about 4.16 V and then decreased to about 3.83 V, suggesting anodic removal of insulating NaCl to expose the aCNS (Fig. 2b inset, Extended Data  Fig. 2b, c) to facilitate further charging/oxidation of NaCl in the pores of the aCNS electrode. SEM imaging showed exposure of the aCNS in the gaps of the NaCl microcrystal coating from the first discharge (Extended Data Fig. 1b, indicated by the square). However, not all of the surface NaCl layer was oxidizable, with parts remaining regardless of the re-charging capacity. Instead of oxidizing the remaining NaCl (loosely bound to aCNS), towards the end of charging, a higher-charging-voltage plateau (about 3.91 V) was observed (Fig. 2a), attributed to oxidation of SOCl 2 in the electrolyte over the exposed carbon nanospheres to form SCl 2 , S 2 Cl 2 and SO 2 Cl 2 (refs. 11,28,29 ).
Mass spectrometry of the species in opened batteries (Supplementary Text 1) after cycling 21 times at a stable Coulombic efficiency showed that the main discharge plateau of about 3.55 V was attributed to Cl 2 reduction. This was based on the detected Cl 2 species (excluding fragments from other molecules, Supplementary Fig. 3a) decreasing to about 0% in the fully discharged state from about 100% in the charged state (Supplementary Text 1). Reduction of molecular Cl 2 species generated and trapped in the aCNS contributed to the majority of the discharge capacity during cycling, and was responsible for the main charge/discharge 3.83 V/3.55 V plateaus with an overall battery reaction of (Supplementary Text 2): Na + 1/2Cl ↔ NaCl.   The two small discharge plateaus at about 3.69 V and about 3.18 V were attributed to reduction of S 2 Cl 2 /SCl 2 and SO 2 Cl 2 (formed at the end of the charging), respectively (Fig. 2a). When the battery was held at open circuit in a charged state over longer times (up to five days), we observed that the main discharge plateau at about 3.55 V decreased in capacity while the lower discharge plateau at about 3.18 V extended ( Fig. 2d). Mass spectrometry showed that the detected Cl 2 in the battery decreased in proportion to the capacity of the roughly 3.55 V plateau, together with a decrease in SO 2 , increase in SO 2 Cl 2 and longer SO 2 Cl 2 discharge plateau at about 3.18 V (Fig. 2e,  . b, Atomic percentages of Na and Cl from XPS survey spectra recorded on the aCNS cathode after the battery was charged to different capacities. Inset: SEM imaging of the cathode charged to 600 mAh g −1 with most of the NaCl removed and revealing the underlying carbon nanosphere. Error bars were calculated based on the XPS survey spectra recorded at different positions on the aCNS cathode. c, XRD spectra (normalized to Ni current collector) of aCNS cathodes when batteries in discharged state were charged to various capacities. NaCl was increasingly oxidized/removed from the cathode. d, Charge-discharge curve of a Na/Cl 2 battery, with the discharge curves recorded after the battery was held at open circuit for different numbers of days in the fully charged state. e, f, Percentage changes for the parameters indicated versus different battery retention times in the open-circuit charged state before discharging (Cl 2 , SO 2 , and 3.55-V plateau in e, SO 2 , SO 2 Cl 2 and 3.18-V plateau in f). DC in the label indicates discharged state of the battery. g, Charge-discharge curves (red) of a Na/Cl 2 battery recorded after discharging the battery post-five-days retention in the charged state. h, Cycling performance of the Na/Cl 2 battery with different retention cycles at 500 mAh g −1 (150 mA g −1 ). The loading of the aCNS was about 4.5 mg cm −2 .
Article a decreased Cl 2 reduction plateau at about 3.55 V and an increased SO 2 Cl 2 discharge plateau at about 3.18 V (Fig. 2d). The reduction of SO 2 Cl 2 for the lower discharge plateau at roughly 3.18 V was (Supplementary Text 2): 2Na + SO Cl + 2e → SO + 2NaCl Open-circuit holding of the battery for days slowly shortened the higher discharge plateau at about 3.55 V, but the battery discharge capacity was about 99.9% retained with the average discharge voltage remaining high, more than 3.2 V. The 3.55-V plateau was immediately restored in subsequent cycles (Fig. 2g, h). Mass spectrometry data also suggested that during battery cycling, SCl 2 and S 2 Cl 2 were involved in the small, highest charge-voltage (about 3.91 V, due to SOCl 2 oxidation) and discharge-voltage (about 3.69 V, SCl 2 and S 2 Cl 2 reduction) plateaus ( Supplementary Fig. 3a, Supplementary Text 1, 2). On discharge, part of the NaCl produced reacted with AlCl 4 − •SOCl + in the electrolyte to re-generate SOCl 2 oxidized in the charging step, which was important to electrolyte re-generation and the rechargeability of the Na/Cl 2 battery (Supplementary Text 2) 11 .
The Na/Cl 2 battery cycled for more than 200 cycles at a set specific capacity of 500 mAh g −1 at 150 mA g −1 current (Extended Data Fig. 3a). The Na/Cl 2 batteries cycled with a discharge capacity of up to about 1,800 mAh g −1 , but at about 90% Coulombic efficiency (Fig. 3a, b) due to non-oxidizable loosely bound NaCl microcrystals on the aCNS (Fig. 3c, d, Extended Data Fig. 4). The Na/Cl 2 battery cycled with a Coulombic efficiency greater than 99% at a reversible capacity of about 1,200 mAh g −1 (Fig. 3e, Extended Data Fig. 3b, c) due to NaCl in the pores of the aCNS. The charge-discharge polarization voltage decreased discernibly as the charging capacity increased (Fig. 3f), indicating reduced impedance as more NaCl was oxidized/removed in the pores of the carbon nanospheres. The energy efficiency of the Na/ Cl 2 battery reached 92.4% (150 mA g −1 ) and 94.2% (100 mA g −1 ) owing to small polarizations. . a, Na/Cl 2 battery cycling at 1,500 mAh g −1 . The electrolyte was 4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI. The battery was able to cycle with a Coulombic efficiency of about 95-96%. The loading of the battery was about 3.5 mg cm −2 . b, Chargedischarge curve of Na/Cl 2 battery at 1,500 mAh g −1 and 1,860 mAh g −1 with 100 mA g −1 current. The electrolyte was 4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI. The loading of the battery was about 3.5 mg cm −2 . About 1,250 mAh g −1 capacity was contributed by NaCl/Cl 2 redox. c, Na 1s spectrum of the aCNS after charging to 1,860 mAh g −1 . A strong Na 1s peak corresponding to NaCl was observed. d, Cl 2p spectrum of the aCNS after charging to 1,860 mAh g −1 .
A strong Cl 2p peak corresponding to NaCl was observed. e, Cycling performance of Na/Cl 2 battery when the charging capacity was 1,200 mAh g −1 (75 mA g −1 and 100 mA g −1 ). The loading of the aCNS was about 2.6 mg cm −2 . f, Charge-discharge curves of a Na/Cl 2 battery when the charging capacities were varied from 375 mAh g −1 to 1,200 mAh g −1 (150 mA g −1 ). The overpotential of the charge-discharge slightly decreased as the cycling capacity increased (about 350 mV for 375 mAh g −1 cycling versus about 190 mV for 1,200 mAh g −1 cycling). The battery was cycling stably at each capacity for a few cycles before the cycling capacity was increased by increasing the charging time of the battery. The electrolyte used in e, f was 4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI.
The Na/Cl 2 battery showed high cyclability at 500 mAh g −1 at 1.2 C rate (600 mA g −1 , 1.39 mA cm −2 Na) ( Supplementary Fig. 4). We also observed interestingly that charging could be done much faster than discharging with a charging rate up to 3.9 C at 500 mAh g −1 (5.63 mA cm −2 Na; in about 15 min) and 0.5 C (1.39 mA cm −2 Na) at 1,200 mAh g −1 over many cycles at a high Coulombic efficiency of more than 99%, with only a slight increase in overpotential (defined as the voltage difference between the main charging and discharging plateaus) (Extended Data Fig. 3d-f).
Importantly, throughout cycling of hundreds of Na/Cl 2 battery coin cells over a period of about three years (discharge cut-off voltage as low as 0.1 V at room temperature), we did not encounter any safety problems under all battery operating conditions, including discharging to various degrees ( Supplementary Fig. 5). No pressurizing problems were found due to strong solvation of SO 2 and Cl 2 species by SOCl 2 , SO 2 Cl 2 and NaAlCl 4 in the electrolyte.
We investigated various electrolyte additives (no additive, NaFSI, NaFSI + NaTFSI, sodium hexafluorophosphate (NaPF 6 ) and fluoroethylene carbonate (FEC)) and found that the mixed 2 wt% NaFSI and 2 wt% NaTFSI afforded the best cycling performance (Fig. 4a, b, Extended Data Fig. 5a, b). The main component in the SEI layer was NaCl, formed as soon as the Na electrode was in contact with the electrolyte regardless of the additives added (Extended Data Figs. 5c, d, 6a, d). As the NaCl layer was impermeable to Na + , the Na/Na + redox through the SEI layer for reversible Na deposition and stripping was attributed to cracks and void regions of the SEI that were sufficiently thin, NaF enriched and Na ion permeable 30,31 . XPS and SEM showed that fluoride-containing additives in the electrolyte indeed afforded voids on our Na anode, with the mixture of NaFSI and NaTFSI additives giving the highest fluoride contents (NaF and -CF 3 -) in the SEI (Extended Data Figs. 5c, e, 6b). Over battery cycling, the fluoride content on the Na surface decreased with increased NaCl (Extended Data Figs. 5c, g, h, 6a, e, f), and with the electrolyte containing the optimal additives, the NaCl crystallites formed on the Na anode were the smallest in size with the highest number of voids, corroborating the reversible Na + /Na redox and the longest battery cycle life (Extended Data Figs. 6g, 7, Supplementary Text 5). Our observation was consistent with the SEI on the alkali-metal anode being more robust when both FSI − and TFSI − anions were present, as TFSI − was less reactive and reacted with Na slower than FSI − , allowing a more uniform and robust SEI to form on the alkali-metal anode 14,32,33 .
The aCNS used for the positive electrode were key to the rechargeable Na/Cl 2 battery due to the high surface area (3,167.82 m 2 g −1 ) and high porosity (2.49 cm 3 g −1 ), especially the high microporosity (1.33 cm 3 g −1 ) (Extended Data Table 1). We compared several widely used amorphous carbon materials including acetylene black (AB) and ketjenblack carbon black (KJ) as the positive electrode (Fig. 4c). The AB material showed the lowest surface area and pore volume (Extended Data Table 1), giving the lowest first discharge capacity and cycle life (Fig. 4c). Compared with aCNS, the KJ material exhibited a lower surface area but larger pore volume (micropores + mesopores = 3.09 cm 3 g −1 ) (Extended Data Table 1), affording a higher first discharge capacity (about 3,250 mAh g −1 versus about 2,810 mAh g −1 ). The increase in discharge capacity with the increase in pore volume (KJ > aCNS > AB) in the positive electrode suggested that the high discharge capacity of the battery was due to sustained NaCl discharge product filling the abundant micropores and mesopores a b c d e f Article in the electrode (not due to surface NaCl coating; see Supplementary Text 3 for detail) 8,9,34,35 . On re-charging, regions of the surface NaCl coating were oxidized/removed to expose the underlying nanospheres (see SEM image in Fig. 2b inset, Extended Data Fig. 1b) with a large impedance decrease (Extended Data Fig. 2b, c), allowing oxidation of NaCl residing in the pores of the nanospheres to release the majority of the capacity stored in the lower plateau of the first discharge.
The aCNS positive electrode afforded Na/Cl 2 cells with Coulombic efficiency and cycling stability (more than 200 cycles) (Extended Data Fig. 3a) superior to cells with a KJ (about 50 cycles) (Fig. 4c) or AB (about 20 cycles) (Fig. 4c) positive electrode. Despite a lower total pore volume than KJ, the aCNS cycled more stably than KJ, pointing to the importance of the more than 60-fold higher micropore volume of the aCNS (about 53.4% micropores in aCNS and about 1.33 cm 3 g −1 versus only about 0.7% micropores in KJ and about 0.021 cm 3 g −1 ) (Extended Data Table 1) to Na/Cl 2 battery cycle life. The much greater micropore volume in the aCNS probably stabilized battery cycling by better retaining Cl 2 , and preventing excessive oxidizers in the electrolyte and anode corrosion (Supplementary Fig. 6). Developing carbon materials with further improved micropore volume could further boost the capacity and cycling stability of secondary Na/ Cl 2 batteries.
We employed the Na/Cl 2 battery to light up a light-emitting diode (LED) that required an operating voltage of 3.0-3.2 V. The current measured through the LED was about 12.03 mA with a high current density of 6.14 mA cm −2 of Na, equivalent to a discharge rate of 1,563.35 mA g −1 (based on aCNS mass) (Fig. 4d). Although the Na/Cl 2 battery is promising in terms of voltage, specific capacity, cycle life and capacity retention compared with various Na metal anode batteries (Supplementary Table 1), optimization and engineering are needed for real-world use 14,[36][37][38][39][40][41] . We explored reducing the amount of electrolyte and using thin separators down to 60 μm (Extended Data Fig. 8). The batteries cycled well with increased gravimetric/ volumetric energy density when the electrolyte volume was lowered to 100 μl and 50 μl, respectively.
Lastly, we extended the Na/Cl 2 battery concept to rechargeable Li/Cl 2 batteries by pairing the aCNS positive electrode with a Li metal as the negative electrode in electrolytes composed of 1-4 M AlCl 3 in SOCl 2 with 2 wt% LiFSI/LiTFSI (Na was focused in this work due to chronological order of the research). The battery delivered about 3,309 mAh g −1 first discharge capacity and was cyclable at 500-1,200 mAh g −1 (150 mA g −1 and 100 mA g −1 currents) with a charging voltage of about 3.80 V and a discharging voltage of about 3.6 V (Fig. 4e, f, Extended Data Fig. 9). Despite the similarity, the differences between the Li/Cl 2 and the Na/Cl 2 batteries warrants further investigation. In terms of practical applications, the Li-metal batteries could be more advantageous due to the higher processability and lower reactivity of Li metal than Na metal.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-021-03757-z.

Synthesis of aCNS
Deionized water (50 ml) and ethanol (20 ml, >99.9 %, J. T. Baker) were mixed uniformly at room temperature. Triblock copolymer, F-127 (0.25 g, PEO106-PPO70-PEO106, weight-averaged molecular mass M W : 12,600, Aldrich), was then added in the mixture and stirred for about 10 min. After F-127 dissolved completely, ammonia solution (0.5 g, 25%, Choneye) was then added to the solution and stirred for about 30 min, followed by the addition of resorcinol (0.5 g, 99%, Alfa Aesar) to the solution. Finally, formaldehyde solution (0.763 g, 37 wt%, Aldrich) was added gradually into the solution and stirred for 24 h at room temperature. The solution was centrifuged at 14,900 r.p.m. to separate the solid and liquid. The solid was dried at 100 °C in an oven and heated at 350 o C for 2 h in N 2 to remove the template. The carbonization process was conducted at 800 °C for 4 h in N 2 followed by the activation process using CO 2 at 1,000 °C for 45 min.

Fabrication of aCNS electrode
Ninety per cent by weight of aCNS and 10% by weight of PTFE (60% aqueous dispersion, FuelCellStore) were mixed in 100% ethanol (Fisher Scientific). The mixture was sonicated for 2 h until the aCNS were uniformly dispersed in ethanol. Ni foam substrate was cut into a circular shape with diameter of 1.5 cm using a compact precision disk cutter (MTI, MSK-T-07). The circular Ni foam substrate was sonicated in 100% ethanol for 15 min and dried in an 80 °C oven until all the ethanol evaporated. The weight of the Ni foam substrate was measured and then hovered over a hot plate. The aCNS, PTFE and ethanol mixture was then slowly dropped (180 μl each time) onto the Ni foam. Between each drop, we waited for approximately 4 min to allow all the ethanol from the previous drop to fully evaporate. This process was repeated and stopped until the loading of the aCNS on Ni foam substrate was desirable (for lower-and higher-loading aCNS electrodes, the loading was 2-3 mg cm −2 and 4-5 mg cm −2 , respectively). The electrodes were then dried in an 80 °C oven overnight. After drying, the electrode was pressed using a spaghetti roller and the final weight of the electrode was measured. After calculating the weight of aCNS-that is, the final weight of the electrode minus the initial weight of the Ni foam times 90%-the electrode was ready to be used in a battery.

Electrolyte making
The electrolyte was made inside an argon-filled glovebox. NaFSI (TCI Chemical) and NaTFSI (Alfa Aesar) were dried at 100 °C vacuum oven overnight before use and stored in an argon-filled glovebox. Thionyl chloride (purified, Spectrum catalogue number TH138) was used without any further purification. The appropriate amount of thionyl chloride liquid was added into a 20-ml scintillation vial (Fisher Scientific) and its weight was measured. Aluminium chloride (4 M, Fluka, 99%, anhydrous, granular) was weighed and added to the thionyl chloride and stirred until all the aluminium chloride was fully dissolved. Then the appropriate amount of NaFSI and NaTFSI (2 wt% of the total weight of aluminium chloride and thionyl chloride) were added to the solution and stirred until both NaFSI and NaTFSI completely dissolved, after which the electrolyte was ready to be used. The electrolyte for Li/Cl 2 battery was made similarly to the electrolyte for Na/Cl 2 battery by replacing NaFSI and NaTFSI with LiFSI and LiTFSI (TCI Chemical).

Battery making
All batteries were made inside an argon-filled glovebox. Sodium-metal block (Sigma Aldrich) was dried using kimwipes (Kimberly-Clark Professional Kimtech Science) to remove the mineral oil on the surface. A razor blade was then used to cut all sides of the Na block to expose the shiny Na metal. The Na-metal block was then placed inside a zip lock bag and pressed using a scintillation vial to make a flat thin Na foil. The Na foil was then pasted onto the spacer in a coin cell. Any extra Na was removed so that the Na foil had the exact shape as the spacer and could be used as the negative electrode. aCNS loaded on Ni foam were used as the positive electrode. Two layers of quartz fibre filters (Sterlitech, Advantec, QR-100) were used as the separators and were dried in 120 °C vacuum oven overnight before each use. The aCNS positive electrode was put in the middle of the SS316 positive coin cell case. Two layers of QR-100 separators were then put on top of the aCNS positive electrode. One-hundred and fifty microlitres of the electrolyte (4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI) was then added to wet the QR-100 separators. The Na negative electrode on the spacer was then put on top of the separators, with Na foil directly facing the aCNS positive electrode. One piece of spring was put on top of the spacer. Lastly, one layer of PTFE foil (Mill-Rose) was put on top of the spring and underneath the SS316 negative coin cell case to prevent corrosion from the electrolyte. After all the components of the coin cell were put together, the coin cell was pressed using a digital pressure controlled electric crimper (MTI, MSK-160E) with the pressure reading set to 9.23. Then the coin cell was taken out the glovebox, and a layer of silicone (GE Sealants & Adhesives 281 Advanced Silicone) was applied to the edge of the coin cell where the two cases sealed, as an extra layer of protection to prevent water and air from leaking into the battery. After the silicone was cured, the battery was tested using a battery tester from Neware, BTS80, version 17.
To prepare the Li negative electrode for the Li/Cl 2 battery, Li-metal foil (Sigma Aldrich) was polished using a file. Then the shiny Li metal was pasted onto the spacer and used as the negative electrode. The separator used for Li/Cl 2 battery was one layer of quartz fibre filter (Sterlitech Advantec, QR-200). Everything else in assembling the Li/ Cl 2 battery was the same as for assembling the Na/Cl 2 battery.

Electrochemical impedance spectroscopy
The electrochemical impedance spectroscopy (EIS) of the battery was measured using a potentiostat/galvanostat (model CHI 760D, CH Instruments). The working electrode was connected to the aCNS positive electrode, and the counter and reference electrodes were connected to the Na negative electrode. The initial voltage of the measurement was set to be the open-circuit potential of the battery at the time of the measurement. The high frequency was 1 × 10 5 Hz and the low frequency was 0.01 Hz. The amplitude of the measurement was 0.005 V.
Scanning electron microscopy SEM imaging was conducted using a Hitachi/S-4800 SEM instrument. To conduct SEM imaging on the aCNS, aCNS powder was first stuck on the sample stage of SEM using double-sided conductive carbon adhesive tape and the stage was then loaded into the SEM chamber for measurement. To conduct SEM imaging on electrodes in the actual battery, the battery was first opened inside an argon-filled glovebox. The electrodes were taken out from the opened battery and transferred into the argon-filled antechamber of the glovebox. The electrodes were vacuumed and dried inside the antechamber for approximately 3 h to remove any electrolyte trapped in them. After drying, the electrodes were transferred back into the glovebox and ready to be characterized. The samples were stuck onto the SEM sample stage using double-sided conductive carbon adhesive tape and introduced into the SEM chamber for measurement. The sample was observed by SEM with 15-kV acceleration voltage of an electron beam at a pressure of 10 −7 torr. A magnification of 200,000 could be achieved.

Transmission electron microscopy
Transmission electron microscopy (TEM) imaging was conducted on a FEI EO Tecnai F20 G2 MAT S-TWIN field TEM. To prepare samples for TEM imaging, 0.02 g aCNS was dispersed in 10 ml deionized water in a 20-ml scintillation vial (Fisher Scientific). The mixture was sonicated for 30 min until a uniform dispersion of aCNS was achieved. After sonication, one drop of the mixture was dropped onto a copper TEM grid using glass dropping pipette. The grid was then placed inside a 100 °C oven for three days. After drying, the copper TEM grid with aCNS sample was introduced into the TEM instrument operating at 200 kV for measurement.

XPS experiments
XPS measurements were conducted at the Stanford Nano Shared Facilities, Stanford University, and the XPS instrument used was PHI VersProbe 1. To conduct XPS on Na immersed in different solutions, the sample preparation was done inside an argon-filled glovebox. Na foil was prepared in the same way as for the Na electrode in the battery ('Battery making'). After immersion in the appropriate solution, the Na foil was taken out from the solution and any liquid remaining on the surface was dried using kimwipes (Kimberly-Clark Professional Kimtech Science). The antechamber of the glovebox was refilled with argon and the sample was transferred into the antechamber, in which the sample was vacuumed dried. After drying, the sample was transferred into the glovebox and was ready to be characterized by XPS. To conduct XPS on electrodes from battery, the sample preparation was the same as that for SEM imaging. After sample preparation, the sample was clamped onto the XPS stage and was transferred into the main chamber of the XPS instrument for measurement. All the spectra reported were the spectra obtained after 20-nm argon ion sputtering to remove any possible surface contamination during sample handling.

X-ray diffraction
XRD was conducted on an X-ray diffraction system (Rigaku Miniflex 600 Benchtop) with Cu Kα radiation. The aCNS powder was put on the XRD sample stage and a razor blade was used to press the powder until a flat surface was obtained and the powder was uniformly and firmly distributed over the sample stage. Any extra powder was carefully removed from the sample stage. The sample stage was then transferred into the centre of the XRD instrument for measurement. The start angle and the stop angle were set to be 5° and 90°, respectively, with a scan speed of 3° min −1 . To conduct XRD measurements of electrodes from the battery, the sample preparation was the same as that for SEM imaging, and XRD was performed after the samples were transferred out from glovebox into the XRD instrument.

Brunauer-Emmett-Teller surface area and porosity
Brunauer-Emmett-Teller (BET) surface area and pore volume were measured by a 2020 Accelerated Surface Area and Porosimetry System from Micromeritics. Before each measurement, the appropriate amount of carbon (about 0.14 g) was weighed and placed in the instrument for degas at 350 °C. After degassing, the weight of the carbon was measured again and this weight was input into the software for final surface area and porosity analysis. In the final analysis, the evacuation time was set to be 6 h and the dose amount was set to be 10 cm 3 g −1 standard temperature and pressure. After the measurement was done by the instrument, the surface area and porosity were reported.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Source data are provided with this paper. a, SEM images of aCNS through the first discharge (from 950 mAh g −1 , to 2,100 mAh g −1 , and then full discharge) of the Na/Cl 2 battery and atomic percentage of C, Na and Cl at these stages measured by SEM/energy dispersive X-ray spectroscopy (EDS) mapping (right bar graph). As discharge continued, more and more NaCl was formed on the aCNS and the discharge stopped when NaCl passivated the aCNS. Some of the NaCl formed was very large in size (tens of micrometres). b, SEM images of aCNS when the Na/Cl 2 battery was re-charged to different capacities (375 mAh g −1 , 600 mAh g −1 , 900 mAh g −1 ) and the atomic percentage of C, Na and Cl at these stages measured by SEM/EDS mapping (right bar graph). As charging increased, more and more NaCl was removed from the aCNS, exposing the nanospheres underlying the NaCl coating. The active sites of the battery (the sites at which oxidation reactions happened) were in the gaps in the NaCl microcrystal coating that remained intact during battery operations. c, SEM images of aCNS when the Na/Cl 2 battery was charged to 900 mAh g −1 then discharged to different capacities (375 mAh g −1 , 600 mAh g −1 , 900 mAh g −1 ) and atomic percentage of C, Na and Cl at these stages measured by SEM/EDS mapping (right bar graph). As discharge increased, more and more NaCl formed on the aCNS. When the battery was fully discharged, all the nanospheres were covered and passivated by the NaCl. To take these SEM images, batteries stopping at the designated states were opened inside an argon-filled glovebox and the electrodes were first dried under vacuum, then taken out of the glovebox and transferred into an SEM instrument for the measurements. See Methods for details.

Extended Data Fig. 2 | EIS of Na/Cl 2 battery with acidic 4 M AlCl 3 in SOCl 2 + 2 wt%
NaFSI + 2 wt% NaTFSI as the electrolyte through its first discharge and re-charging and first discharge curve of Na/Cl 2 battery using neutral 4 M AlCl 3 + 4 M NaCl in SOCl 2 as the electrolyte. a, Impedance measurements at six points along the curve of first discharge of the battery when acidic 4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI was used as the electrolyte. b, Charging curve of the Na/Cl 2 battery when the charging capacity was 500 mAh g −1 . Each spike along the curve was a point at which battery charging was stopped for EIS measurements and then allowed to continue to charge. c, Impedance measurements of the Na/Cl 2 battery at different charging capacities tracing the charging curve in b. As charging started, the impedance of the battery rapidly decreased due to removal of NaCl in the coating layer on the positive electrode. d, First discharge curve of Na/Cl 2 battery when neutral 4 M AlCl 3 + 4 M NaCl in SOCl 2 was used as the electrolyte. Only one discharge plateau was observed in neutral electrolyte case. Fig. 3 | Cycling performance of Na/Cl 2 battery at different capacities. a, Cycling performance of a Na/Cl 2 battery at 500 mAh g −1 (150 mA g −1 ). The battery was kept at open circuit in a discharged state for two weeks. We found that simply aging the battery in the discharged state for days could improve the battery's cycle life, probably due to the slower formation of a more uniform SEI layer on the electrode. The loading of aCNS was about 4.5 mg cm −2 . b, Na/Cl 2 battery cycling at 1,200 mAh g −1 . The electrolyte was 4 M AlCl 3 in SOCl 2 + 1 wt% NaFSI + 1 wt% NaTFSI. c, Na/Cl 2 battery cycling at 1,200 mAh g −1 . The electrolyte was 4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI. Both of the batteries in b, c were first cycling at 500 mAh g −1 (150 mA g −1 ) for 15 cycles and the cycling capacity was gradually increased to 1,200 mAh g −1 with 150 mA g −1 and 100 mA g −1 currents. The loading of both batteries was about 2.6 mg cm −2 . d, Cycling performance of Na/Cl 2 battery as the charging current increased from 0.3 C (150 mA g −1 ) up to 3.9 C (1,950 mA/g −1 ) with 0.3 C (150 mA g −1 ) increased for every five cycles. The discharge current was kept at 0.3 C (150 mA g −1 ). The loading of aCNS was about 3 mg cm 2 . e, Cycling performance of Na/Cl 2 battery at 1,200 mAh g −1 with charging current increased to 0.5 C (600 mA g −1 ) and discharging current kept at 0.08 C (100 mA g −1 ). Cycles 1-3: 0.0625 C (75 mA g −1 ), cycles 4 and 5: 0.08 C (100 mA g −1 ) for battery stabilization. The loading of the battery was about 3 mg cm −2 . f, Typical charge-discharge curves of Na/Cl 2 battery at 1,200 mAh g −1 . Black curve: 0.5 C (600 mA g −1 ) charging, 0.08 C (100 mA g −1 ) discharging. Red curve: 0.08 C (100 mA g −1 ) charging and discharging. Only a slight increase in overpotential (about 182 mV at 0.08 C versus about 298 mV at 0.5 C) was observed. The loading of the battery was about 3 mg cm −2 .

Extended Data
Article Extended Data Fig. 4 | SEM images of aCNS after charging to 1,860 mAh g −1 .
Left image: the nanospheres in aCNS were readily observed as NaCl depositing on the surface of aCNS were oxidized. Middle and right images: NaCl microcrystals that were either loosely deposited on top of the nanospheres clusters (not inside the nanospheres, middle image) or deposited in the gaps between the aCNS clusters (right image) were not oxidizable, and could not contribute to the battery's rechargeable capacity. Middle and right images have different magnifications. Fig. 5 | See next page for caption.

Article
Extended Data Fig. 5 | Na/Cl 2 battery performances when 2 wt% FEC and 2 wt% NaPF 6 were used as the electrolyte additives and XPS of Na metal immersing in electrolytes with different additives (2 wt% NaFSI + 2 wt% NaTFSI, 2 wt% NaPF 6 , and 2 wt% FEC) and after battery cycling. a, Na/Cl 2 battery cycling performance at 500 mAh g −1 , 150 mA g −1 when 4 M AlCl 3 in SOCl 2 + 2 wt% FEC was used as the electrolyte. The battery behaved poorly and died after cycle 9. b, Na/Cl 2 battery cycling performance at 1,200 mAh g −1 , 100 mA g −1 when 4 M AlCl 3 in SOCl 2 + 2 wt% NaPF 6 was used as the electrolyte. The battery showed worse cycling performance than when 2 wt% NaFSI + 2 wt% NaTFSI was used as the electrolyte additive. c, Atomic percentage of different elements, calculated from XPS survey spectrum, on the Na metal after immersing in 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaFSI + 2 wt% NaTFSI, 2 wt% NaPF 6 and 2 wt% FEC). d, Cl 2p spectrum of Na metal after immersing in 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaPF 6 and 2 wt% FEC). e, F 1s spectrum of Na metal after immersing in 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaPF 6 and 2 wt% FEC). f, S 2p spectrum of Na metal after immersing in 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaPF 6 and 2 wt% FEC). g, Atomic percentage of different elements, calculated from XPS survey spectrum, on the Na electrode after cycling in batteries using 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaFSI + 2 wt% NaTFSI, 2 wt% NaPF 6 and 2 wt% FEC) as the electrolyte. h, F 1s spectrum of Na electrode after cycling in batteries using 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaPF 6 and 2 wt% FEC) as the electrolyte. The batteries using 2 wt% NaFSI + 2 wt% NaTFSI and 2 wt% NaPF 6 as the electrolyte additives in g, h were stopped at cycle 21. The battery using 2 wt% FEC as the electrolyte additive was stopped at cycle 9 when the battery died. Fig. 6 | Characterizations of Na anode immersed and cycled in 4 M AlCl 3 in SOCl 2 with and without 2 wt% NaFSI/NaTFSI, and chargedischarge curves of the normal battery versus decayed battery. a, Atomic percentage of different elements on the Na metal when immersed in 4 M AlCl 3 in SOCl 2 with and without 2 wt% NaFSI/NaTFSI as additives. b, F 1s spectrum of Na immersed in 4 M AlCl 3 in SOCl 2 with/without additives. c, S 2p spectrum of Na immersed in 4 M AlCl 3 in SOCl 2 with/without additives. d, Cl 2p spectrum of Na immersed in 4 M AlCl 3 in SOCl 2 with/without additives. e, Atomic percentage of different elements on the Na metal after cycling for 21 cycles in Na/Cl 2 battery when 4 M AlCl 3 in SOCl 2 with and without 2 wt% NaFSI/NaTFSI as additives were used as the electrolyte. f, F 1s spectrum of Na cycled in Na/Cl 2 battery when 4 M AlCl 3 in SOCl 2 with/without additives were used as the electrolyte. g, SEM images of Na anode from actual Na/Cl 2 battery in charged state (top images) and when lost cycling capability (bottom images). Note that in the case of battery without fluoride additive, the Na anode surface was coated by more densely packed NaCl particles, eventually leading to the loss of re-chargeability. h, Charge-discharge curves of the battery at normal state and after the battery started to decay. Extended Data Fig. 7 | SEM images of Na electrodes after cycling in batteries using 4 M AlCl 3 in SOCl 2 with different additives (2 wt% NaFSI + 2 wt% NaTFSI, 2 wt% NaPF 6 , and 2 wt% FEC) as the electrolytes. Top row: SEM images of Na electrode after cycling in battery using 4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI as the electrolyte. The SEI layer contained loosely packed, square-shaped NaCl crystals and abundant voids still present in the SEI (indicated by circles). Middle row: SEM images of Na electrode after cycling in battery using 4 M AlCl 3 in SOCl 2 + 2 wt% NaPF 6 as the electrolyte. The SEI layer contained closely packed, square-shaped NaCl crystals that were grown on top of a uniform layer of NaCl crystals. Such morphology made ions penetrations much less efficient. Bottom row: SEM images of Na electrode after cycling in battery using 4 M AlCl 3 in SOCl 2 + 2 wt% FEC as the electrolyte. The SEI layer was made of very large NaCl crystals (tens of micrometres in size) packed together. Such morphology made ions penetrations only possible via the small cracks between these crystals and the least efficient. The batteries using 2 wt% NaFSI + 2 wt% NaTFSI and 2 wt% NaPF 6 as the electrolyte additives were both stopped at cycle 21. The battery using 2 wt% FEC as the electrolyte additive was stopped at cycle 9 when the battery died. Fig. 8 | Na/Cl 2 battery cycling performance using less electrolyte (4 M AlCl 3 in SOCl 2 + 2 wt% NaFSI + 2 wt% NaTFSI) and thinner separators down to 60 μm. a, Na/Cl 2 battery cycling performance at 500 mAh g −1 using 100 μl electrolyte with one layer of QR-100 separator. The loading of the battery was about 5 mg cm −2 . b, Na/Cl 2 battery cycling performance at 500 mAh g −1 using 75 μl electrolyte with one layer of QR-100 separator. The loading of the battery was about 5 mg cm −2 . c, Na/Cl 2 battery cycling performance at 500 mAh g −1 using 50 μl electrolyte with one layer of 60-μm glass fibre separator. The loading of the battery was about 5 mg cm −2 . d, Charge-discharge curve of Na/Cl 2 battery at 500 mAh g −1 using 50 μl electrolyte. e, Na/Cl 2 battery cycling performance at 1,200 mAh g −1 using 100 μl electrolyte with one layer of QR-100 separator. The loading of the battery was about 3.6 mg cm −2 . f, Charge-discharge curve of Na/Cl 2 battery at 1,200 mAh g −1 using 100 μl electrolyte.