Rechargeable Na-CO2 Batteries Starting from Cathode of Na2CO3 and Carbon Nanotubes

Na-CO2 batteries have attracted significant attentions due to their high energy density and effective utilization of greenhouse gas CO2. However, all reported Na-CO2 batteries employ excessive preloaded metal Na, which will lead to safety issues such as dendrite formation and short circuit. In addition, the charging mechanism of reported Na-CO2 batteries is not very clear. Here we report the Na-CO2 batteries, starting from the cathode of cheap Na2CO3 and multiwalled carbon nanotubes (CNTs). Due to the effective electron transfer and high reactivity, the decomposition of Na2CO3 and CNTs could take place under 3.8 V. The charging mechanism of 2Na2CO3 + C → 4Na + 3CO2 without any side reactions is revealed by in/ex situ techniques such as Raman, gas chromatograph, and optical microscope. Dendrite-free Na can quantitatively deposit on the Super P/Al anode because of large specific surface area and low nucleation barrier of the anode for Na plating. The batteries could deliver an energy density of 183 Wh kg−1 (based on the whole mass of the pouch-type batteries, 4 g) with stable cycling performance. This work reveals that safe rechargeable Na-CO2 batteries could be constructed by cheap Na2CO3 and multiwalled carbon nanotubes.


Introduction
Rechargeable Na-CO 2 batteries have attracted a lot of interests because they can reuse the greenhouse gas CO 2 to realize energy storage and conversion [1][2][3][4]. The reported rechargeable Na-CO 2 batteries with liquid or quasi-solid state electrolyte employed metal Na and CO 2 as the reactants. The discharge reaction can be described by 4Na + 3CO 2 → 2Na 2 CO 3 + C [3,4]. However, Na-CO 2 batteries still suffer from safety issues and self-discharge due to the use of excessive preloaded metal Na as anode. Metal Na would encounter dendrite formation or surface cracks during cycling process due to the uneven deposition of Na, leading to short circuit [5][6][7]. By contrast with metal Na, Na 2 CO 3 and carbon are more stable, safer, and easier to get [8]. More importantly, the generated metal Na by the charging decomposition of Na 2 CO 3 and carbon is stoichiometric. The batteries with stoichiometric metal Na will be safer than that with excessive preloaded metal Na. Therefore, designing dendrite-free rechargeable Na-CO 2 batteries without preloaded metal Na is of great significance.
Due to its insulating properties, the decomposition of Na 2 CO 3 will result in large polarization and inevitable side reactions such as electrolyte decomposition [9][10][11]. Thus, to develop rechargeable Na-CO 2 batteries starting with Na 2 CO 3 and carbon, the main issue is how to realize the decomposition of Na 2 CO 3 and carbon under proper voltage range. To ensure fast kinetics such as effective electron transfer during the decomposition process, Na 2 CO 3 should be evenly intertwined with highly conductive carbon and the particle size should be small [12][13][14][15][16][17]. Meanwhile, the reactivity of carbon materials also has a great influence on the decomposition process [18][19][20][21]. Furthermore, the decomposition mechanism of Na 2 CO 3 and carbon in previous works is not very clear [3,4]. It is also important to find out the source of CO 2 in the charge process. In addition, to ensure Na metal is produced during charging and the generated stoichiometric metal Na does not form dendrite is also challenging [22][23][24].
Here we report Na-CO 2 batteries with Na-free architecture, which operate in pure CO 2 atmosphere (1 atm) (Figure 1(a)). The cathode is composed of Na 2 CO 3 and CNTs (multiwalled carbon nanotubes) with defects, where stick-shaped Na 2 CO 3 (about 200-300 nm in length) are intertwined with CNTs closely and uniformly. The close and uniform structure is beneficial for promoting the decomposition of Na 2 CO 3 and CNTs in proper voltage range (< 3.8 V).  The anode is Super P/Al, which can effectively reduce the generation of Na dendrite and increase the Coulombic efficiency of the batteries because of the large specific surface area and low nucleation barrier for Na plating. The optimized electrolyte is 1 M NaClO 4 -tetraethylene glycol dimethyl ether (TEGDME) in view of its high voltage window (> 4.8 V versus Na + /Na) and high ionic conductivity (0.18 S m −1 ) ( Figure  S1). By means of various characterizations such as in situ Raman, gas chromatograph (GC), in situ optical microscopy, and nuclear magnetic resonance (NMR), we demonstrate that the charging reaction is 2Na 2 CO 3 + C → 4Na + 3CO 2 without any side reactions. After optimization, the dendritefree Na-CO 2 batteries show a relatively long cycling life (100 cycles) with cut-off capacity of 0.3 mAh cm −2 . Remarkably, the pouch-type Na-CO 2 batteries starting from Na 2 CO 3 and CNTs deliver a total energy density of 183 Wh kg −1 (based on the whole mass of the pouch-type batteries, 4 g).

Preparation of Na 2 CO 3 /CNTs Composite.
To ensure the decomposition of cathode (Na 2 CO 3 and carbon) under low voltage range, there are two key points which should be addressed properly. One is the combination status of Na 2 CO 3 and carbon. The other is the carbon source, which acts as the conductive base and the reactant to participate in the charging reaction.
To guarantee uniform and close combination of Na 2 CO 3 and carbon, we prepared the composites of Na 2 CO 3 and carbon by a facile dissolution-crystallization method ( Figure  S2) [25]. The commercial CNTs have a porous cross-linking structure with the wall thickness of about 5 nm and spacing layers of 0.34 nm (Figures 1(b) and 1(c)). The outer walls of CNTs offer spaces for Na 2 CO 3 to nucleate and anchor.
Meanwhile, the inner walls act as conductive networks to promote electron transport easily [26,27]. After the dissolutioncrystallization process, we got the Na 2 CO 3 /CNTs composite, in which the stick-shaped Na 2 CO 3 (about 200-300 nm in length) are intertwined with CNTs closely and uniformly ( Figure 1(d)). Through the HRTEM images and the selected area electron diffraction (SAED), we can see the (201) and (310) planes of Na 2 CO 3 and multiwall structure of CNTs in the composite (Figures 1(e) and 1(f)). In addition, we find that Na 2 CO 3 crystal only grows on the outer walls of CNTs ( Figure S3). The reason for this phenomenon is that the CNTs have less polarity and belong to superhydrophobic materials [28,29]. There are some defects with hydrophilic groups (-OH, -COOH) in the outer wall of the CNTs, which is conducive to the nucleation and crystallization of watersoluble Na 2 CO 3 on the outer walls of CNTs [30,31]. However, the hydrophobicity of the inner walls prevents the water from entering.
We also used the same fabricating method to explore another four kinds of carbon materials, including reduced graphene oxide, Super P, single-walled carbon nanotubes, and double-walled carbon nanotubes. We finally chose multiwalled carbon nanotubes as the carbon source due to its high electrical conductivity, cross-linking structure, and suitable defects (optimization details can be seen in Supporting Information, Figures S4-6).

Decomposition of Na 2 CO 3 /CNTs Cathode.
After obtaining the Na 2 CO 3 /CNTs composite, we investigate the decomposition issue of Na 2 CO 3 /CNTs cathode by using Super P/Al as anode (the detailed optimization of anode can be seen in Figures S7-9). In this part, we first optimized the content of Na 2 CO 3 in the cathode and then revealed the charging mechanism thoroughly. The powder X-ray diffraction (XRD) patterns of pure CNTs, Na 2 CO 3 , and as-prepared Na 2 CO 3 /CNTs cathodes with different mass ratios are shown in Figure S10A. All diffraction peaks of pure Na 2 CO 3 are well indexed with the standard XRD pattern of Na 2 CO 3 (PDF no. . In the XRD pattern of CNTs, a broad peak at 26.4 ∘ is assigned to the (002) crystal face. When the mass percentage of Na 2 CO 3 in the cathodes is low (10, 20, and 30 wt%), there are no apparent diffraction peaks of Na 2 CO 3 , indicating that Na 2 CO 3 are coated by CNTs. When the mass percentage of Na 2 CO 3 further increases (40, 50, and 60 wt%), diffraction peaks of Na 2 CO 3 emerge and gradually increase, which suggests that the grain size of Na 2 CO 3 becomes large [32]. Due to the insulting properties of Na 2 CO 3 , charge transfer resistance of Na 2 CO 3 /CNTs cathodes becomes larger and larger when the mass percentage of Na 2 CO 3 gradually increases ( Figure S10B). When the mass percentage of Na 2 CO 3 increases to 60 wt%, there is a jump in the impedance value, indicating that the content of Na 2 CO 3 is too much to form uniform mixture. Then, we explore the decomposition behavior of cathodes with different percentage of Na 2 CO 3 by linear sweep voltammetry (LSV, Figure 2(a)). The results show that the decomposition voltages reduce first and then increase with the increased content of Na 2 CO 3 in cathodes (the inset in Figure 2(a)). This result could be attributed to the following reasons. First, when there are only pure CNTs or Na 2 CO 3 , the charge reaction lacks another reactant, leading to a high charging voltage and resulting in the decomposition of the electrolyte. Note that we added 90 wt% titanium powders as conductive additive when we investigated the decomposition of pure Na 2 CO 3 ( Figure S11). Second, most of Na 2 CO 3 are coated with CNTs when the content of Na 2 CO 3 is low, so the electrolyte is difficult to spread to the surface of Na 2 CO 3 . When the amount of Na 2 CO 3 increases to high value (60 wt%), the large particles of Na 2 CO 3 cannot mix well with CNTs, resulting in poor conductivity. Therefore, a minimum decomposition voltage would be achieved when the content of Na 2 CO 3 is intermediate. Here, a minimum decomposition voltage of 3.72 V was achieved when the content of Na 2 CO 3 is 50 wt%, which should be originated from the best threephase interface and the most uniform mixing degree with this content. In addition, the charge tests of cathodes were carried out, and the results are consistent with the LSV curves ( Figure  S12).
The interaction between Na 2 CO 3 and CNTs in cathode is confirmed by X-ray photoelectron spectroscopy (XPS) ( Figure S13). Compared with pure Na 2 CO 3 , the C1s spectra of CO 3 2− in cathode shifted to the lower binding energy, indicating that the C atoms on the CNTs interact with the O atoms on the CO 3 2− [33]. Meanwhile, the O atoms in the CO 3 2− also reduce the strength of C-C on the CNTs, resulting in a new C-C peak at the 283.47 eV [34].
After optimizing the Na 2 CO 3 /CNTs cathode, we studied the charging mechanism through several methods. We first designed an in situ Raman coin cell with a 2 mm observation hole in the cathode cap to collect Raman signals from cathode in real time ( Figure S14). Thirteen points were selected on charge profiles. As shown in Figure 2(b), the main peaks at 1080 and 700 cm −1 corresponding to Na 2 CO 3 decrease little by little during charging process. In addition, the intensity of G and D band corresponding to CNTs also gradually decreases. The results indicate that the cathode can decompose in charging process.
Moreover, we employed gas chromatography (GC) measurements to monitor the gas generation during the charging process in pure argon atmosphere (1 atm) (Figure 2(c)). The results show that CO 2 generates throughout the charging process. The retention time of CO 2 in GC column is 6 minutes and no other new gases were detected. The practical evolution concentration of CO 2 is 3.39 × 10 4 ppm, which corresponds to 7.845 × 10 −5 mol. This value is less than the theoretical one with 3.63 × 10 4 ppm (8.393 × 10 −5 mol, and detailed calculations can be found in the Supporting Information). The difference can probably be attributed to the reasons that CNTs has strong adsorption effect on CO 2 . In addition, CO 2 also show certain solubility in TEGDME solvent [4,35]. In order to prove that the CO 2 is not produced by side reactions, we confirmed the components of electrolyte after charge by 1 H and 13 C NMR spectroscopy. The 1 H and 13 C NMR spectroscopy of the electrolyte have not changed after charge ( Figure S15), indicating that the electrolyte did not decompose during the charging process.
We also characterized the morphological features of the cathode after charge by SEM ( Figure S16). The results show that the surface of cathode becomes porous and the stick Na 2 CO 3 disappears at the end of the charge process. The specific surface area of the cathode also increased from 27.498 m 2 g −1 to 215.52 m 2 g −1 after charge process ( Figure  S17). In the pristine cathode, Na 2 CO 3 closely combined with carbon nanotubes (Figures S4A and S4B), resulting in a dense electrode. After charging, all the Na 2 CO 3 are decomposed, and the original position of Na 2 CO 3 is vacated. In addition, the carbon nanotube skeleton is well maintained ( Figure S16). Therefore, the cathode becomes loose and porous, resulting in a significant increase in specific surface area. Furthermore, the charge transfer resistance (diameter of the semicircle) of the batteries also significantly reduced after full charge ( Figure S18), indicating that the nonconductive Na 2 CO 3 decomposed.

Na Generation.
Combining the Super P/Al anode and optimized Na 2 CO 3 /CNTs cathode, we construct full batteries to characterize the generation of Na. SEM images of the pristine Super P/Al anode show that Super P with diameter of 30 nm are coated on the surface of Al (Figure 3(a) and Figure  S19). Compared with the pristine anode, the electrode surface is covered by silver-white charging products after charging to 1 mAh (Figure 3(b)). XRD pattern of the charging products is shown in Figure 3 Figure S20) that the metal Na has a smooth surface and disperses evenly, indicating that Super P plays a role in dispersing current density. The yield of Na is directly proportional to the charging capacity (1-3 mAh, Figure S21). When the charging capacity increased to 3 mAh, the surface of Super P was covered with a layer of Na.
In order to further observe the process of Na generation more visually, we designed an in situ optical microscopy setup to monitor the change of anode during charge process in real time (Figure 3(d)). The in situ optical microscopy setup includes an Au anode, a Na 2 CO 3 /CNTs cathode, a seal ring for sealing electrolyte, and an optics lens. We collected optical microscope images of the Au anode during the charge process at the sweep rate of 0.2 mV s −1 ( Figure S22). Figure 3(e) shows a series of light-field optical microscope images, which successively present the morphology evolution of the Au electrode. The pristine Au electrode is clean and flat before charge voltage of 3.7 V (0 s). With the charging time going on, the surface of Au electrode gradually becomes dark (60 s). The XRD patterns ( Figure S23) of Au electrode prove that the dark materials are metal Na. The entire surface of the Au electrode was covered by generated Na after 300 s. With further increase of charging depth, more Na deposited and covered the electrode layer by layer (480s, 720s, and 1020s). A video of Na deposition process can be seen in Movie S1. Moreover, we measured the surface height of Au electrode before and after charge through atomic force microscope (AFM) (Figure 3(f)). After charging for 2040 s, the height of Au electrode increases to 200 nm, indicating the thickness of deposited metal Na layer. These results demonstrate that the charging process accords with the following equation: 2Na 2 CO 3 + C → 4Na + 3CO 2 .

Electrochemical Performance.
After revealing the mechanism, we evaluate the electrochemical performance of the Na-CO 2 batteries without preloaded metal Na in CO 2 atmosphere (1 atm). The batteries firstly charged to produce Na. The voltage profile displays a flat plateau before charging to 4.6 mAh cm −2 ( Figure S24), indicating that Na 2 CO 3 and CNTs can decompose at low voltage. At the end of the charge process, the voltage curve starts to go up, demonstrating that all the Na 2 CO 3 have almost broken down. The batteries show a maximal capacity of 5 mAh cm −2 , which is very close to the theoretical value (calculation details can be seen in Supplementary Materials). Subsequently, we evaluate cycling performance of the batteries at different current densities with the limited capacity at 0.3 mAh cm −2 . At a current density of 0.05 mA cm −2 , the charge potential of the batteries was maintained below 4 V after 100 cycles (Figure 4(a)). The results demonstrate an impressive stability of Na-CO 2 batteries. As the current density increased to 0.10 and 0.15 mA cm −2 , the batteries were capable of discharging/charging for over 50 cycles ( Figures  S25A and S25B). The excellent performance of the batteries shows that the Na generated by decomposition of Na 2 CO 3 and CNTs can be used effectively. The surface morphology of Na coated Super P/Al anode after 50 cycles was studied by SEM ( Figure S26). It can be seen that the Na covered surface of Super P/Al anode remains smooth and no Na dendrite is discovered, which suggests a uniform deposition/stripping process of Na on the surface of Super P/Al anode.
The morphological features of discharge products at different current densities were observed by SEM. We find that the current densities could affect the morphologies of the discharge products. As shown in Figure 4(b), round-shaped discharge products (∼ 100 nm) form when the current density is 0.05 mA cm −2 . As the current densities rise (0.10 and 0.15 mA cm −2 ), the discharge products become larger with rectangle shape (∼ 200 nm) at 0.10 mA cm −2 and nanorod shape (∼ 500 nm) at 0.15 mA cm −2 ( Figures S27A and S27B). The changes in the morphology of the discharge products are similar to previous report [4]. Although the morphology of discharge products is different, we prove that Na 2 CO 3 appear in all discharge products at different current densities through Raman characterization (the insets of Figure 4(b), Figures S27A and S27B). According to our previous work, the other discharge product is amorphous carbon that is mixed with sodium carbonate [3,4].
We further fabricate pouch-type Na-CO 2 batteries to assess their potential applications in energy storage. The pouch-type batteries are composed of Super P/Al (anode), Na 2 CO 3 /CNTs (cathode), and Celgard (separator) soaked in electrolyte (Figure 4(c) and Figure S28). During the charging process, the sodium ions are removed from Na 2 CO 3 and in situ reduced to metal Na on the Super P/Al anode. When the charging process finished, the open circuit voltage of the batteries is 2.35 V (Figure 4(d)), which is consistent with the voltage of the common Na-CO 2 batteries with preloaded metal Na. The pouch-type Na-CO 2 batteries can light a LED bulb with 1 watt (Figure 4(e)). Movie S2 is a video of the process that bulb is lit up. With cut-off voltage at 1.9 V, the initial discharge capacity of 350 mAh can be obtained at a current of 10 mA ( Figure S29), corresponding to an energy density of 183 Wh kg −1 based on the whole mass of the pouchtype batteries (4 g). The energy density of the pouch-type Na-CO 2 batteries is higher than that of commercial Li-ion batteries (150-180 Wh kg −1 ) [36]. The pouch-type Na-CO 2 batteries exhibit stable cycling performance with an average discharge voltage of 2.13 V (Figure 4(f)). The results indicate the possibility of practical applications of Na-CO 2 batteries which start from Na 2 CO 3 and CNTs.

Discussion
In conclusion, we have successfully constructed dendritefree rechargeable Na-CO 2 batteries without preloaded metal Na. The batteries operate in pure CO 2 atmosphere, consisting of Na 2 CO 3 /CNTs cathode, Super P/Al anode, and NaClO 4 /TEGDME electrolyte. The uniform Na 2 CO 3 /CNTs cathode with 50 wt% Na 2 CO 3 could decompose under 3.8 V during the charge process. The Super P/Al anode is beneficial for dendrite-free Na plating/stripping (0.5 mAh cm −2 , 100 cycles), showing higher and steadier Coulombic efficiency (98%) than that of pure Al anode. Combination of in situ Raman, GC, in situ optical microscopy, XRD, and NMR demonstrates that Na generate in the Super P/Al anode, and CO 2 release in Na 2 CO 3 /CNTs cathode without any side reactions during initial charge process. The charging reaction could be described as 2Na 2 CO 3 + C → 4Na + 3CO 2 . Due to effective use of the generated Na, the coin-type batteries are capable of charging/discharging for 100 cycles with a cut-off capacity of 0.3 mAh cm −2 . Moreover, the pouch-type batteries deliver a large energy density of 183 Wh kg −1 . Our work paves the way to construct safe and cheap Na-CO 2 batteries without preloaded metal Na for practical energy storage applications.

Na 2 CO 3 /CNTs Cathode Preparation.
In a typical preparation process, 50 mg commercial Na 2 CO 3 and 40 mg CNTs dissolved in a mixed solvent including 4 mL ethanol and 4 mL water (1:1, v/v) under sonication for 2 hours. Then the solution was transferred to a culture dish and ultrasonic heated until solvent evaporation at 80 ∘ C. During the solvent volatilization process, Na 2 CO 3 recrystallized and deposited on the surface of CNTs. After full evaporation of solvent, black Na 2 CO 3 /CNTs composite powder was obtained. The powder was characterized by XRD, XPS, SEM, and TEM. Then we rolled piece with Na 2 CO 3 /CNTs composite powder (90 wt%) and PTFE solution as binder (10 wt%), followed by drying in a vacuum oven at 80 ∘ C for 12 hours. Every piece weighs about 10∼30 mg with a diameter of 14 mm. The other cathodes were prepared in the same way.

Super P/Al Anode Preparation.
Conductive carbon black (TIMCAL Super P li) and polyvinylidene fluoride (PVDF) (9:1) were mixed with N-methylpyrrolidone (NMP). Slurry was then doctor bladed onto Al foil, followed by drying in a vacuum oven at 100 ∘ C for 10 hours to obtain Super P/Al anode with Super P mass loading of ∼0.2 mg/cm 2 .

Batteries Assembly.
In order to optimize the anode, the Na-Al or Na-Super P/Al batteries were assembled by stacking a metal sodium (12 mm in diameter), Celgard separator (16 mm in diameter) soaked with 60 L of electrolyte, and a Al foil or Super P/Al electrodes (14 mm in diameter) successively in a CR2032 coin-type battery. The full cells were also assembled in CR2032 coin cells with Super P/Al anode, Celgard separator soaked with 80 L of electrolyte, and the as-prepared Na 2 CO 3 /CNTs cathode. A hole is drilled in the cathode shell (4 mm in diameter) so that the CO 2 can easily and quickly pass the cathode. Then, the battery is placed in a bottle filled with pure CO 2 atmosphere. Pouch-type full cells are made up of one plastic case (8 × 9 cm 2 ), a Super P/Al anode (7 × 6.8 cm 2 , 0.25 g), a piece of Celgard (8 × 7 cm 2 ) with 2 mL electrolyte, and a Na 2 CO 3 /CNTs cathode (6 × 5.1 cm 2 , 1.6 g). The total mass of this pouch-type battery is 4 g. The pouch-type full cells run in a glove box with pure CO 2 atmosphere.

Data Availability
All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials.
Additional data related to this paper may be requested from the authors.

Conflicts of Interest
The authors declare no competing financial interests. Figure S1. The electrochemical stability window and ionic conductivity of electrolyte. Figure S2. Schematic diagram of the fabrication process of Na 2 CO 3 /CNTs composites. Figure S3. TEM images of Na 2 CO 3 /CNTs cathode. Figure  S4. The morphology characterization of various cathodes with different carbon sources and the comparision of charge voltage. Figure S5. Characterization of raw Na 2 CO 3 . Figure  S6. The comparision of different carbon nanotubes. Figure  S7. The optimized anode. Figure S8. Comparison of the sodium nucleation overpotential for Super P/Al and bare Al current collectors. Figure S9. SEM images of Na deposition (3 mAh) at different anodes. Figure S10. The optimization of Na 2 CO 3 /CNTs composites. Figure S11. EIS of mixed materials consisting of Na 2 CO 3 and titanium powder (mass ratio of 1: 9). Figure S12. The charge profiles of pure CNTs, Na 2 CO 3 /CNTs, and pure Na 2 CO 3 with titanium powder at current density of 0.1 mA cm −2 . Figure S13. The Xray photoelectron spectroscopy (XPS) of pure CNTs, pure Na 2 CO 3 and Na 2 CO 3 /CNTs composites cathode of 50 wt% Na 2 CO 3 content. Figure S14. The in-situ Raman battery. Figure S15. 1 H NMR and 13 C NMR spectra of electrolyte before and after charge. Figure S16. SEM images of Na 2 CO 3 /CNTs cathode after charge. Figure S17. The specific surface area (BET) of cathode. Figure S18. EIS of the battery before and after charge. Figure S19. SEM images of pristine Super P/Al anode. Figure S20. SEM images of Super P/Al electrode after charge. Figure S21. The photographs of Super P/Al anode with different charging capacity (0-3 mAh). Figure S22. A representative LSV curve of Na deposition in the in-situ tests with Na 2 CO 3 /CNTs as working electrode and Au as counter electrode. Figure S23. XRD patterns of the deposited Na on the Au electrode. Figure S24. The full charge profile of Na-CO 2 batteries with 5 mAh cm −2 at 0.1 mA cm −2 . Figure  S25. The cycling stability of Na-CO 2 batteries with a cutoff capacity of 0.3 mAh cm −2 at different current densities.