Introduction

Carbon capture and utilization (CCU) is gaining increasing attention in the field of CO2 reduction1,2, global warming mitigation3,4, and potentially future Mars migration5,6. Varieties of CCU technologies that can convert CO2 into value-added chemicals, such as methane dry reforming7,8, hydrogenation9,10, electrochemical reduction11,12, and photocatalytic reduction13,14, have been developed. In the past decade, a kind of energy storage device of Li–CO2 battery was proposed, offering an attractive tactic to utilize CO2 and produce electrical energy15,16,17. A typical Li–CO2 battery is composed of a lithium metal anode separated by an aprotic electrolyte from a porous CO2 cathode. The typical reversible reaction at the cathode involves the reduction of CO2 to form Li2CO3 and carbon on discharge, and the process is reversed on charge (Eq. 1). The thermodynamic equilibrium potential and specific energy are calculated to be about 2.80 V and 1876 Wh kg−1, respectively. The high theoretical specific energy far exceeds that of commercial Li-ion batteries, making it a potentially disruptive technology for energy storage18,19,20.

$$4{{{{{{\rm{Li}}}}}}}^{+}+3{{{{{{\rm{CO}}}}}}}_{2}+4{e}^{-}\leftrightarrow 2{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}+{{{{{{\rm{C}}}}}}},\, {E}^{0}=2.80\,{{{{{\rm{V}}}}}}$$
(1)

However, practical Li–CO2 batteries usually present discharge voltages of around 2.5 V, sometimes even lower than 2.0 V in previous reports21,22,23. Generally, the quality of electrical energy is determined by the voltage supplied. An output voltage lower than 3.0 V leads to a low-quality electrical energy24. Obviously, the actual output voltage of Li–CO2 batteries is far lower than the theoretical value which is not high enough. Apart from the thermodynamic information of the reaction, the voltage that the battery can provide depends on the catalytic activity of catalysts and the transport properties of charge and mass in bulk and between phase boundaries.

On this basis, much efforts on solid catalysts have been exerted in raising the discharge voltage and reducing the charge voltage. The reported catalysts include carbon allotropes25,26, noble metals26,27,28, and transition metal oxides29,30. Although they can remarkably reduce the charging overpotential, they have minimal effect in increasing the discharge voltage. It should be explained here that the catalytic characteristics of solid catalysts bring the difficulty of raising battery discharge voltage. As illustrated in Fig. 1a, four phases are involved in the electrochemical reduction of CO2. Specifically, three solid phases contain solid catalysts, Li2CO3, and carbon on the cathode surface. The liquid phase includes Li ions and dissolved CO2 in electrolyte. CO2 reduction during discharge occurs at the catalyst/electrolyte interface. The effect of catalytic reaction partially depends on the catalytic surface area of the solid catalyst particles on which CO2 is reduced. The sluggish kinetics of charge-transfer and mass-transport across multiphase interfaces aggravate the large voltage hysteresis. What’s more, active sites of solid catalysts are occupied by insulating and insoluble Li2CO3 products, leading to their invalidation31,32.

Fig. 1: Schematic comparison of solid catalysts-mediated and liquid catalysts-mediated discharge reactions in Li–CO2 batteries.
figure 1

a In the presence of solid catalysts, CO2 is reduced to carbon and Li2CO3. Four phases are involved, including three solid phases on the cathode surface (solid catalysts, Li2CO3, and carbon) and liquid electrolyte. With the use of liquid catalysts, RM molecules capture dissolved CO2 in electrolyte, and then the formed RM–CO2 species are reduced to corresponding products and original RM. Here, the products are Li2C2O4. Only two phases, namely solid products and species in liquid electrolyte, are involved. b Discharge voltage of the battery using liquid catalyst depends on the redox potential of RM/RM+ couple. This condition allows to adjust the output voltage from 2.80 V (CO2-to-Li2CO3 conversion) to above 3.0 V by selecting and designing RM molecules.

Full size image

It is worth mentioning that discharge products also affect the charging performance of the next cycle. In accordance with previous reports, electrochemical decomposition of Li2CO3 itself usually occurs irreversibly during the charging process of Li–CO2 batteries (Eq. 2)33,34. In this case, the charging potential is higher than 4.5 V and will result in the low energy efficiency. Besides, highly active intermediates of superoxide radicals (\({{{{{\mathrm{O}}}}}}_{2}^{ \cdot - }\))33 or singlet oxygen (1O2)34 cause severe parasitic reactions. The gradual accumulation of irreversible byproducts threatens the stability of batteries. Thus, a catalyst designed with the strategy of Li2CO3-free pathway might be a good choice.

$$2{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{CO}}}}}}}_{3}\to 2{{{{{{\rm{CO}}}}}}}_{2}+{{{{{{\rm{O}}}}}}}_{2}^{\cdot -}({\,\!}^{1}{{{{{\rm{O}}}}}}_{2})+4{{{{{{\rm{Li}}}}}}}^{+}+3{e}^{-}$$
(2)

On the basis of the above discussion, liquid catalyst (or redox mediator, RM) rather than solid one can reduce the number of phases involved in the CO2 reduction process, which is effective to reduce the discharge overpotential. The reported liquid catalysts, including 2,5-ditert-butyl-1,4-benzoquinone35, 2-ethoxyethylamine36, and tris(2,2′-bipyridyl)-dichloro-ruthenium(II)37, can promote the discharge potential to a certain extent. However, batteries involving these catalysts still follow the Li2CO3 pathway. It might be better for the battery to discharge without taking the Li2CO3 path by liquid catalysts. As depicted in Fig. 1a, only two phases are involved. Li ions, dissolved CO2, and catalytic RMs are mixing at the molecular level in the liquid phase of electrolyte. Typical discharge process contains two steps. RM molecules capture CO2 to form RM–CO2 species first. The newly formed molecules then gain electrons at the cathode and are reduced to original RM and corresponding products (such as Li2C2O4). The liquid catalyst has more full contact with CO2 at the molecular level in the liquid phase, which effectively improves the reaction kinetics. The electrochemical redox process of the RM–CO2 at the electrode replaces the directly electrochemical reduction of CO2. This process allows to adjust the battery’s output voltage up to above 3.0 V by selecting and designing RM molecules (Fig. 1b).

Apart from the abovementioned conventional liquid catalysts, some soluble metal complexes can also catalyze the electrochemical reduction of CO2 to oxalate chemicals38. This condition inspires us to introduce the catalytic effect of metal complexes into the design of Li–CO2 batteries. Moreover, Li2C2O4 as an electrochemical product can take the battery out of the troublesome Li2CO3 pathway. Herein, we introduce a binuclear copper(I) complex (denoted as Cu(I) RM) as the liquid catalyst in Li–CO2 batteries and study the battery performance, including the discharge potential, capacity, and cycle performance, in detail. In addition, we use a variety of spectroscopic analysis techniques, such as Raman and differential electrochemical mass spectrometry (DEMS), to explore the Li2CO3-free path experienced by the discharge process of the cathode. Furthermore, we employ an additional catalyst containing Ru nanoparticles to reduce the charge overpotential synergistically. This study increases the output voltage of Li–CO2 batteries to more than 3.0 V, which strongly promotes the practical application of this electric energy storage system.

Results and discussion

Structure characterization of Cu(I) RM

The binuclear Cu(I) RM was synthesized by the reaction of disulfide ligand with two equivalents of [Cu(CH3CN)4]BF4 in dry acetonitrile (MeCN)39 (details of the preparation procedure are provided in the Methods section). 1H nuclear magnetic resonance spectroscopy (1H NMR) and electrospray ionization mass spectrometry (ESI-MS) were performed first to verify the molecular structure of the prepared ligand. As shown in Supplementary Fig. 1a, peaks at 8.53, 7.68, 7.58, 7.19, 3.80, 3.22, 2.59, and 1.25 can be observed, corresponding to H3, H5, H6, H4, H7, H10, H9, and H11 of the ligand, respectively. Besides, the signal at the mass/charge (m/z) ratio of 545.25 matches well with that calculated for [C30H37N6S2]+ of the ligand (Supplementary Fig. 1b). A yellow solution obtained was analyzed by ESI-MS after mixing the ligand with [Cu(CH3CN)4]BF4 in MeCN. A prominent signal at the m/z ratio of 335.05 in Supplementary Fig. 2 is consistent with that of [Cu2C30H36N6S2]2+, confirming the successful synthesis of the target complex.

Electrochemical measurements of Cu(I) RM

Electrochemical experiments were undertaken in a typical three-electrode cell with various atmospheres to investigate the catalytic effect of Cu(I) RM on CO2 complexation and electron transfer. Electrolyte preparation and cell assembly are described in the Methods section. As shown in Fig. 2a and Supplementary Fig. 3, two peaks appear in the cathodic region near 2.68 V (Ec,1) and 2.99 V (Ec,2) under Ar atmosphere. The specific redox potentials of active centers in complex molecules are affected by the structure and property of ligands38. Consequently, it is reasonable to assign the two cathodic peaks of Ec,1 and Ec,2 to the electron transfer of Cu active centers in Cu(I) RM, as depicted in Eqs. 3 and 4, respectively40. Meanwhile, two anodic peaks at 2.90 V (Ea,1) and 3.29 V (Ea,2) are rationally ascribed to reverse reactions (Fig. 2a and Supplementary Fig. 3). When CO2 is introduced, the cell containing Cu(I) RM displays two similar cathodic peaks to that operated in Ar, depending on the electron transfer of Cu active centers in complexes. While the cathodic peak of Ec,2 moves to a more positive potential of 3.05 V (Ec,3), suggesting that the reduction reaction of Cu(II) RM to Cu(I) RM is affected by the presence of CO2. Besides, an additional anodic peak at 4.11 V (Ea,3) corresponds to the oxidation of reduction products, which will be investigated subsequently. In comparison, the CV profile shows no sign of reaction in LiClO4/MeCN electrolyte under CO2 atmosphere, which is consistent with previous reports35.

$${{{{{\rm{Cu}}}}}}({\rm I})\,{{{{{\rm{RM}}}}}}+2{e}^{-}\to {{{{{\rm{Cu}}}}}}(0)\,{{{{{\rm{RM}}}}}}$$
(3)
$${{{{{\rm{Cu}}}}}}({\rm I}{\rm I})\,{{{{{\rm{RM}}}}}}+2{e}^{-}\to {{{{{\rm{Cu}}}}}}({\rm I})\,{{{{{\rm{RM}}}}}}$$
(4)
Fig. 2: Electrochemical performance, cathode characterization, and CO2 quantification of Li–CO2 batteries with Cu(I) RM-based electrolyte.
figure 2

a CV curves for cells with various electrolytes and atmospheres: 0.5 mM Cu(I) RM in 0.1 M LiClO4/MeCN under Ar (dark gray), 0.5 mM Cu(I) RM in 0.1 M LiClO4/MeCN under CO2 (red), and 0.1 M LiClO4/MeCN under CO2 (blue). Scan rate is 20 mV s−1. b Galvanostatic discharge-charge curves for Super P carbon cathodes in various electrolytes and atmospheres: 0.5 mM Cu(I) RM in 0.1 M LiClO4/MeCN under Ar (dark gray), 0.5 mM Cu(I) RM in 0.1 M LiClO4/MeCN under CO2 (red), and 1 M LiTFSI/G4 under CO2 (blue). Current density is 100 mA g−1. c Cyclic performance of the Li–CO2 battery with Cu(I) RM-based electrolyte under a fixed specific capacity of 1000 mAh g−1 at a current density of 100 mA g−1. d Raman spectra of Super P carbon cathodes at different reaction stages in Li–CO2 batteries containing Cu(I) RM. e SEM and f TEM images of the discharged Super P carbon cathode. Corresponding SAED pattern is displayed in the inset of f. g FTIR spectra of Super P carbon cathodes at different reaction stages in Li–CO2 batteries containing Cu(I) RM. h Rate of CO2 consumption as a function of time after the injection of Cu(I) RM-based electrolyte. Inset of h shows the schematic of reaction device. Specifically, 1 mL of Cu(I) RM-based electrolyte is injected to a sealed vessel, which is connected to a mass spectrometry with pure CO2 as the purge gas. i Gas evolution rate of the Li–CO2 battery containing Cu(I) RM on charge. The battery is first discharged to 1000 mAh g−1 and subsequently charged back under the measurement of DEMS.

Full size image

To examine the practical electrochemical performance of Cu(I) RM in Li–CO2 batteries, galvanostatic discharge-charge tests were performed by employing home-made Swagelok-type batteries with the addition of Cu(I) RM. No solid catalyst other than Super P carbon was used on the electrode to investigate the efficacy of Cu(I) RM liquid catalysts28. And pre-charged LixFePO4 instead of Li metal was used as the anode to avoid the oxidation of anode by positively charged RM species35 and the crossover of CO2 to Li anode41,42,43. The LixFePO4 potential versus Li/Li+, 3.45 V, was used to express all potentials in this work at the Li scale44. Galvanostatic discharge-charge measurements of batteries were performed at the current density of 100 mA g−1 with a discharge cutoff potential of 2 V. Besides MeCN-based electrolyte, a conventional lithium bis(trifluoromethanesulphonyl)imide/tetraethylene glycol dimethyl ether (LiTFSI/G4) electrolyte was also evaluated in the Li–CO2 battery system for comparison26. All batteries rested for a minimum of 8 h before discharge-charge tests. The OCV of the Li–CO2 battery containing Cu(I) RM was recorded in Supplementary Fig. 4. Upon CO2 pumping, the OCV experiences a rapid rising from 3.12 to 3.38 V. The final OCV greatly higher than that of the conventional Li–CO2 battery, ~2.80 V, declares the change in battery reaction after the addition of Cu(I) RM. In the absence of Cu(I) RM, the Li–CO2 battery deteriorates quickly, exhibiting a small capacity of 19 mAh g−1 (Supplementary Fig. 5), which is consistent with the ignorable current response measured by CV test (Fig. 2a). As presented in Fig. 2b, on the one hand, the Li–CO2 battery containing Cu(I) RM catalyst delivers a larger discharge capacity up to 5846 mAh g−1 and a higher output voltage of about 3.04 V, indicating the great effect of Cu(I) RM on the promotion of discharge performance. On the other hand, the battery shows a reduced charge voltage plateau of 4.27 V compared with those without Cu(I) RM. As is known to all, the conventional Li–CO2 battery usually exhibits poor rechargeability and low energy efficiency45. The decrease in charge potential could be related to Cu(I) RM-mediated discharge reaction path. Notably, a short slope at the beginning of charging in the battery containing Cu(I) RM can be connected with the oxidation of Cu(0)/Cu(I) to Cu(II) after deep discharge. As shown in Supplementary Fig. 6, the deconvoluted Cu 2p spectrum of discharged Super P carbon cathode shows peaks assigned to Cu0/Cu+ and Cu2+46. By contrast, the battery containing Cu(I) RM catalyst has almost no capacity when the CO2 was replaced by Ar, indicating that the discharge capacity is associated with participation of CO2 in the reaction. In the absence of CO2, only the reduction of Cu(I) RM to Cu(0) RM occurs. The small addition of Cu(I) RM in battery accounts for its very small capacity. Furthermore, under a fixed specific capacity of 1000 mAh g−1, the Li–CO2 battery with Cu(I) RM exhibits steady cyclability over 80 cycles (Fig. 2c).

Cu(I) RM-mediated reaction mechanism

In order to unveil the specific discharge product of Li–CO2 batteries in the presence of Cu(I) RM, the morphology and composition of Super P carbon cathodes at different reaction stages, including pristine, after discharge, and after recharge, were characterized. Supplementary Fig. 7a shows the typical morphology of Super P carbon particles. Figure 2d presents Raman peaks corresponding to the D and G bands of Super P carbon at 1338 and 1600 cm−1, respectively. After discharge, randomly arranged products are deposited densely on the surface of Super P carbon cathode (Fig. 2e). Transmission electron microscopy (TEM) image clearly manifests that Super P carbon particles are well-coated with discharge products (Fig. 2f). The corresponding selected area electron diffraction (SAED) pattern indicates that these products are Li2C2O4. Besides, a new peak at 1473 cm−1 assigned to Li2C2O4 is observed for the discharged Super P carbon electrode in the Raman spectrum47 (Fig. 2d). In addition, Fourier transform infrared (FTIR) spectrum displays peaks at 508, 771, 1321, 1422, and 1641 cm−1, which are indexed to ρω(O−C=O), δa(O−C=O), νa(O−C=O), δs(C−O) + ρω(C−O), and νa(C=O) modes of C2O42−, respectively48,49,50,51 (Fig. 2g). And the characteristic peak of O−C=O for Li2C2O4 is detected at 288.95 eV in the C 1 s spectrum through X-ray photoelectron spectroscopy (XPS)43 (Supplementary Fig. 8). It is clearly evidenced that the discharge product is Li2C2O4 rather than Li2CO3. Thus, the formation of Li2C2O4 products accounts for the reduced charge voltage shown in Fig. 2b. When the charge process is finished, these discharged particles disappear, and the cathode surface is recovered (Supplementary Fig. 7b). Meanwhile, all the peaks assigned to Li2C2O4 vanish, confirming the fully reversible decomposition of Li2C2O4 products after recharging (Fig. 2d, g). The reversible formation and decomposition of Li2C2O4 products can be detected even after multiple cycles (Supplementary Fig. 9).

Subsequently, mass spectrometry was conducted to analyze the CO2 gas consumed by Cu(I) RM, and in situ DEMS was performed to reveal the ratio of transferred electrons and generated CO2 gas during charge. As shown in Fig. 2h, a sealed vessel was connected to the quadrupole mass spectrometry with pure CO2 gas stream as the purge gas. The Cu(I) RM-based electrolyte was injected into the vessel when the flux of CO2 reached a stable background line. The flux of CO2 declines sharply in the first 5 min and increases afterwards, demonstrating the rapid reaction kinetics between Cu(I) RM and CO2. After discharged to a limited specific capacity of 1000 mAh g−1, the Cu(I) RM-contained Li–CO2 battery was charged back and monitored by DEMS. No gases were detected other than CO2 (Fig. 2i). The evolution of CO2 increases gradually and stabilizes with a charge-to-mass ratio close to 1 e–/CO2, verifying Li2C2O4 decomposition as the dominant charge reaction (Eq. 5).

$${{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{C}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{4}\to 2{{{{{{\rm{Li}}}}}}}^{+}+2{{{{{{\rm{CO}}}}}}}_{2}+2{e}^{-}$$
(5)

To further elucidate changes in the chemical structures for Cu(I) RM molecules at different electrochemical states, Cu(I) RM-based electrolytes corresponding to different reaction stages in Fig. 3a were explored by ultraviolet–visible (UV–Vis) absorption spectroscopy, Raman spectroscopy, and ESI-MS. No particular features but signals associated with Cu(I) RM and MeCN solvent can be observed for the pristine electrolyte (Fig. 3b–e). A broad band centered at 606.70 nm arises in the UV–Vis spectrum while resting in CO2 for a period (Fig. 3b). The absorbance band matches well with the d–d transition of Cu2+ center in the bridged CO2-derived oxalate group52. And a new peak at 2332 cm−1 referring to the vibration mode of νs(O−C=O) + δs(O−C=O) in oxalate group is recorded in the Raman spectrum49 (Fig. 3c, d). Besides, a prominent signal at the m/z of 379.84 in Fig. 3f is consistent with that of the bridged Cu(II)-oxalate adduct ([Cu4C64H72N12O8S4]4+)39. These results indicate that Cu(I) RM molecules capture CO2 in the electrolyte to form the Cu(II)-oxalate adduct primarily. As calculated by the integration of CO2 consumption rate and time in Fig. 2h, the total amount of CO2 consumed is equal to the summation of reacting CO2 with Cu(I) RM molecules and dissolved CO2 in MeCN53, confirming that the Cu(II)-oxalate adduct is generated by the reaction of two equivalents of CO2 to one Cu(I) RM, as described in Eq. 6. All the characteristic peaks ascribed to Cu(II)-oxalate adduct decrease obviously, and the signal of Cu(I) RM in Fig. 3g appears again after discharge, indicating the regeneration of Cu(I) RM with the formation of Li2C2O4 (Eq. 7). Thus, redox mediation by the Cu(I) RM involves a complexation mechanism and subsequent electron transport (Fig. 3h). The formation of Cu(II)-oxalate adduct can be caused by the rapid reaction between Cu(I) RM and dissolved CO2 in the electrolyte, as evidenced by mass spectrometry (Fig. 2h). After recharge, apart from features of Cu(II)-oxalate adduct (Fig. 3i), a signal consistent with that calculated for the electrochemically oxidized Cu(I) RM ([Cu2C30H36N6S2]4+) is observed (Supplementary Fig. 10). As presented in Supplementary Fig. 11, Cu(I) RM-based electrolyte can remain stable after multiple cycles.

$$2{{{{{\rm{Cu}}}}}}({\rm I})\,{{{{{\rm{RM}}}}}}+4{{{{{{\rm{CO}}}}}}}_{2}\to {{{{{\rm{Cu}}}}}}({\rm I}{\rm I})\mbox{-}{({{{{{{\rm{C}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{4})}_{2}$$
(6)
$${{{{{\rm{Cu}}}}}}({\rm I}{\rm I})\mbox{-}{({{{{{{\rm{C}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{4})}_{2}+4{{{{{{\rm{Li}}}}}}}^{+}+4{e}^{-}\to 2{{{{{\rm{Cu}}}}}}({\rm I})\,{{{{{\rm{RM}}}}}}+2{{{{{{\rm{Li}}}}}}}_{2}{{{{{{\rm{C}}}}}}}_{2}{{{{{{\rm{O}}}}}}}_{4}$$
(7)
Fig. 3: Electrolyte characterization of Li–CO2 batteries with Cu(I) RM-based electrolyte.
figure 3

a Discharge-charge curve of the Li–CO2 battery with the addition of Cu(I) RM, showing three stages, namely rest, discharge, and charge. b UV–Vis and c Raman spectra of Cu(I) RM-based electrolytes in the Li–CO2 battery at different reaction stages. d shows the enlarged view of c. ESI-MS spectra of Cu(I) RM-based electrolytes at different reaction stages, showing e pristine, f before discharge, and g after discharge. h Proposed mechanism of Cu(I) RM-mediated CO2 reduction process. i ESI-MS spectrum of Cu(I) RM-based electrolyte after recharge.

Full size image

On the basis of the above results, Li–CO2 electrochemistry mediated by Cu(I) RM is clear. That is, Cu(I) RM reacts with CO2 to form Cu(II)-oxalate adduct chemically first, then the newly formed adduct gets reduced electrochemically to produce Li2C2O4 product and original Cu(I) RM. Other two possible pathways are also taken into consideration54. One path is that Cu(I) RM gets reduced electrochemically to form Cu(0) RM first, then Cu(0) RM reacts with CO2 chemically. As shown in Fig. 2a, the reduction peak of Cu(I) RM to Cu(0) RM is about 2.68 V (Ec,1), much lower than the practical discharge voltage plateau of 3.04 V in this RM-involved Li–CO2 battery. Thus, this path can be ruled out. Another path is that CO2 gets reduced electrochemically to form CO2− first, then CO2− reacts with Cu(I) RM chemically. Considering the low thermodynamic potential of CO2/CO2− conversion (−1.9 V versus NHE)55, the discharge voltage of a Li–CO2 battery following this path should be no more than 2.0 V versus Li/Li+. In consequence, this path can also be ruled out.

Obviously, the Cu(I) RM-mediated reaction process of Li–CO2 battery possesses some unique advantages. Firstly, the Cu(I) RM can greatly promote the output voltage of battery to above 3.0 V by converting the direct electrochemical reduction of CO2 into the reduction of bridged Cu(II)-oxalate adduct. Then, the Cu(I) RM can increase the discharge capacity of battery significantly until the cathode is covered by enough solid products. Besides, Li2C2O4 products are formed rather than troublesome Li2CO3, leading to a relatively lower charge platform. In addition, aggressive intermediates derived from Li2CO3 decomposition are circumvented, thereby endowing a prolonged cycle life of more than 80 cycles. However, the charge potential plateau beyond 4.0 V is still unsatisfactory. As we know, Super P carbon generally exhibits poor electrocatalytic activity. Thus, more effective catalysts are required to further improve the charge performance of the Cu(I) RM-containing Li–CO2 battery.

Synergistic effect of Cu(I) RM and Ru catalysts

Considering the excellent catalytic properties of Ru in reducing charge overpotential27,28, Ru nanoparticles deposited on Super P carbon (Ru@Super P) were incorporated as the cathode catalyst in this Cu(I) RM-involved Li–CO2 battery. The preparation procedure of Ru@Super P is depicted in the Methods section. All peaks in the X-ray diffraction (XRD) pattern of Ru@Super P can be assigned to metallic Ru (Supplementary Fig. 12a). The TEM image clearly exhibits that Ru nanoparticles are well-dispersed on Super P carbon (Supplementary Fig. 12b). Additionally, the Ru content in Ru@Super P was estimated to be around 17% according to thermogravimetric (TG) analysis (Supplementary Fig. 12c). To demonstrate the stability of Cu(I) RM in the presence of Ru catalyst, the Ru@Super P powder was immersed in the Cu(I) RM-based electrolyte for at least 10 days. It can be seen from Supplementary Fig. 13 that the prominent signal at the m/z ratio of 335.05 consistent with that of Cu(I) RM can also be detected in the ESI-MS spectrum, indicating that the application of Ru catalyst has no effect on the stability of Cu(I) RM. Figure 4a displays the galvanostatic discharge-charge curve of the Li–CO2 battery with Cu(I) RM-based electrolyte and Ru@Super P cathode at the current density of 100 mA g−1. The discharge plateau of around 3.01 V is close to that using the Super P carbon cathode, and an enlarged specific discharge capacity of about 8058 mAh g−1 is obtained. More importantly, the charge platform is remarkably decreased to 3.99 V, and the energy efficiency is enhanced to 75.4%, demonstrating the excellent catalytic activity of Ru nanoparticles toward the charging process. Notably, the Ru catalyst has no effect on the Cu(I) RM-mediated reaction path as Li2C2O4 was detected as the sole discharge product (Supplementary Fig. 14). Moreover, the cycling performance of the battery was assessed with a limited specific capacity of 1000 mAh g−1 at a current density of 200 mA g−1. As shown in Fig. 4b, the battery exhibits robust cycle stability over 400 cycles. The discharge voltage (3.01 V) and energy efficiency (75.4%) of this work are the best results among the reported solid or liquid catalysts during full discharge-recharge process at the current density of 100 mA g−1 25,28,56,57,58,59,60,61 (Fig. 4c). The highly stable and catalytic Ru nanoparticles play an important role in the reversible decomposition of Li2C2O4 products during multiple cycles (Supplementary Fig. 15). Figure 4d summarizes the synergistic effect of soluble Cu(I) RM and solid Ru catalysts in the Li–CO2 battery system. Specifically, Cu(I) RM acting as a molecular shuttle of CO2 manipulates the discharge path and gains Li2C2O4 products through a liquid-liquid catalysis route; Ru nanoparticles provide highly catalytic active sites for Li2C2O4 decomposition on charge. Thus, a high discharge voltage above 3.0 V and a small discharge-charge voltage gap within 1.0 V are obtained, enabling the significantly increased energy efficiency.

Fig. 4: Optimization of charge performance by Ru catalysts.
figure 4

a Galvanostatic discharge-charge curves of the Li–CO2 battery with a Cu(I) RM-based electrolyte and a Ru@Super P cathode during deep discharging at a current density of 100 mA g−1. b Cyclic performance of the Li–CO2 battery with a Cu(I) RM-based electrolyte and a Ru@Super P cathode under a fixed specific capacity of 1000 mAh g−1 at a current density of 200 mA g−1. c Comparison of discharge voltage and energy efficiency with other reported catalysts for Li–CO2 batteries in recent reports 25,28,56,57,58,59,60,61. d Schematic of the synergetic effect of Cu(I) RM and Ru catalysts for the Li–CO2 battery system.

Full size image

In conclusion, we introduced Cu(I) RM as the liquid catalyst to achieve Li2C2O4 products in Li–CO2 batteries. Specifically, the Li–CO2 battery with Cu(I) RM exhibits a remarkably promoted electromotive voltage up to 3.38 V that far exceeds those of other reported Li–CO2 batteries. Electrochemical and spectroscopic investigations clearly reveal the reaction route of Li–CO2 batteries involving Cu(I) RM: Firstly, the Cu(I) RM captures CO2 to form a bridged Cu(II)-oxalate adduct; Secondly, the formed adduct gets reduced with the formation of original Cu(I) RM and Li2C2O4 products during discharge; Finally, Li2C2O4 products are decomposed in the charge process. With the addition of Cu(I) RM, the Li–CO2 battery using the Super P carbon cathode delivers an enhanced discharge plateau of about 3.04 V, an enlarged discharge capacity of 5846 mAh g−1, and a reduced charge voltage plateau of around 4.27 V. Besides, under the fixed capacity of 1000 mAh g−1, the Li–CO2 battery exhibits steady cyclability over 80 cycles. Then, the Ru@super P catalysts are employed to further reduce the charging polarization. Benefiting from the synergistic effect of soluble Cu(I) RM and solid Ru catalysts, a similar discharge potential of 3.01 V, a promoted discharge capacity of 8058 mAh g−1, and a remarkably decreased charge potential of about 3.99 V are achieved. Under the fixed specific capacity of 1000 mAh g−1, the battery shows robust cycle stability over 400 cycles. This study increases the output voltage of Li–CO2 batteries to higher than 3.0 V and provides convictive evidence of the Li2CO3-free discharge route utilizing an effective soluble metal complex, thereby offering an approach to improving the electrochemical performance of Li–CO2 batteries as well as propelling their practical application.

Methods

Materials

All chemicals used in the synthesis were purchased from Sigma–Aldrich or Tokyo Chemical Industry Co., Ltd. without further purification. For the components in the electrolytes, LiClO4 was obtained from Alfa Aesar. LiTFSI, G4, and MeCN were obtained from Sigma–Aldrich. The salts were dried under vacuum at 120 °C for 12 h, and solvents were dried by using activated 4 Å molecular sieves prior to use.

Preparation of Cu(I) RM

The binuclear Cu(I) RM was synthesized as described in previous literature39. Firstly, dipicolylamine (1.196 g, 6 mmol) and propylene sulfide (0.445 g, 6 mmol) were added to ultra-dry MeCN of 10 mL and experienced reflux under Ar atmosphere for 12 h. Then, the solution was stirred in air for 12 h at room temperature. After that, the solvent was evaporated under vacuum to obtain the light yellow oily ligand. Finally, [Cu(CH3CN)4]BF4 (0.3 g, 1 mmol) was added to the MeCN solution containing ligand (0.27 g, 0.5 mmol) in an Ar glove box to acquire the Cu(I) RM.

Preparation of Ru@Super P

Ru@Super P was prepared in accordance with our previous study28: RuCl3•xH2O (50 mg) was dissolved in ethylene glycol of 100 mL. Then Super P carbon (80 mg) was added in the solution, where the suspension was stirred for 3 h at 170 °C via an oil bath. The mixtures were filtered by deionized water and ethanol for several times after cooling to room temperature. The final products were dried at 120 °C under vacuum for 12 h.

Characterization of Cu(I) RM

All samples were transferred to different characterization equipment by using an air-tight sample module. 1H NMR (Bruker DRX500) was applied to analyze the molecular structures of the ligand. ESI-MS (Agilent 6460) was performed to collect information of the ligand and Cu(I) RM.

Battery assembly and electrochemical performance tests

A three-electrode glass cell was first used to conduct CV tests. The working electrode was commercial glassy carbon (GC, Φ 3 mm) that was polished elaborately prior to use. The counter electrode was obtained by rolling the mixture of LiFePO4, Super P carbon, and polytetrafluoroethylene (PTFE) binder (W: W: W = 80: 10: 10) into a film (1.0 × 1.2 cm) and pressing on a stainless steel (SS) current collector. The pre-charged (10% of total capacity) LixFePO4 (x = 0.9) electrode was applied as the reference electrode, which had a stable potential of 3.45 V versus Li/Li+. The galvanostatic discharge-charge measurements were conducted in Swagelok batteries, including a Super P carbon (or Ru@Super P) cathode (Φ 12 mm, 1.13 cm2), a pre-charged LixFePO4 anode (Φ 12 mm), a glassy fiber separator (Φ 12 mm, Whatman), and a gas chamber. The cathode was prepared by rolling the mixture of Super P carbon (or Ru@Super P) and PTFE binder (W: W = 85: 15) into a film and pressing on a SS mesh. The mass loading of the electrode was 0.5 ± 0.2 mg cm−2. The thickness of a glassy fiber separator was 675 μm. All the electrodes were dried at 120 °C under vacuum for at least 12 h before assemblage. LiClO4 (0.1 M) in MeCN with or without Cu(I) RM (0.5 mM) was employed as electrolytes and the amount of electrolyte in each battery was about 300 μL.

CV measurements were carried out on an electrochemical workstation (CHI760E, Chenhua Co., Shanghai, P. R. China) at 25 °C inside an Ar or CO2-filled glove box with a pressure of 1 atm, as well as negligible H2O and O2 levels (<0.1 ppm). Galvanostatic tests were performed on LAND 2001A Battery Testing Systems (Wuhan LAND electronics Co., Ltd, P. R. China) at 25 °C under Ar or CO2 atmosphere. The batteries were discharged and charged at a specific current of 100 mA g−1 and potential cut-offs of 2 V and 4.8 V. Galvanostatic discharge/charge cycling tests were conducted at a constant current density of 100 mA g−1 or 200 mA g−1 and a fixed capacity of 1000 mAh g−1. All current densities and capacities were normalized by the mass of active materials on the cathode. The specific energy based on active substance on the cathode was the product of specific capacity and output voltage.

Electrode and electrolyte characterization

The discharged and recharged electrodes were washed with MeCN and dried sufficiently before characterization. The components of electrodes and electrolytes during different reaction stages were recorded by Raman spectroscopy (Renishaw inVia confocal Raman microscope) with the excitation light of an air-cooled He–Ne laser at 633 nm through a 50× long working distance lens (Leica Microsystems Inc.). To obtain apparent signals on the spectra and avoid the degradation of carbon cathode, the acquisition time was set as 120 s with 10% laser power. The resolution of Raman spectroscopy was around 1.0 cm−1. FTIR measurement was conducted on a FTIR spectroscope (PerkinElmer, Spectrum Two LiTa) with a wavenumber range of 4000–450 cm−1 and a resolution of 1.0 cm−1. The states of surface elements on the cathodes were characterized through XPS (Thermo Fisher Scientific Model K-Alpha spectrometer) equipped with Al Ka radiation (1486.6 eV) at a working voltage of 12 kV and a current of 10 mA. The morphology of cathodes was observed by SEM (Hitachi SU8010). The microstructure was further characterized by TEM (FEI TF20), and the SAED pattern was collected from a Gatan charge-coupled device camera. To evaluate the interaction between Cu(I) RM and CO2 in the electrolyte, ESI-MS and UV–vis absorption spectra data were collected. The UV–vis spectra were evaluated on a UV–vis spectrophotometer (Beijing Purkinje General Instrument Co. Ltd., P. R. China). XRD analysis was performed to analyze the crystalline structure of the catalyst by employing a Bruker D8 advanced diffractometer with Cu–Kα radiation (λ = 1.5406 Å) at a scan rate of 0.064° s−1. TG was carried out on an SDT Q600 TA instrument with a temperature range of 25–800 °C in O2 gas and the heating rate was 5 °C min−1.

Differential electrochemical mass spectrometry characterization

In situ DEMS measurements were performed by chemical/electrochemical reactions. With regard to the chemical reaction, a mixture of CO2/Ar (V: V = 9: 1) was purged continuously to eliminate residual air first, after which 0.5 mM Cu(I) RM in MeCN (1 mL) was injected to the sealed vessel, and the remaining gas was purged to the mass spectrometer chamber (PrismaPro QMG 250 M2). The sealed vessel was connected by using two PEEK valves to the purge gas system. The electrochemical reaction was conducted by a home-made Li–CO2 battery mold with two PEEK valves connected to a quadrupole mass spectrometer with a turbomolecular pump (Pfeiffer Vacuum). During the charge process, ultrapure Ar was employed as carrier with a flux of 0.5 mL min−1. The DEMS battery was also performed on LAND 2001A Battery Testing Systems.