In situ electrochemical conversion of CO2 in molten salts to advanced energy materials with reduced carbon emissions

See allHide authors and affiliations

Science Advances  28 Feb 2020:
Vol. 6, no. 9, eaay9278
DOI: 10.1126/sciadv.aay9278


Fixation of CO2 on the occasion of its generation to produce advanced energy materials has been an ideal solution to relieve global warming. We herein report a delicately designed molten salt electrolyzer using molten NaCl-CaCl2-CaO as electrolyte, soluble GeO2 as Ge feedstock, conducting substrates as cathode, and carbon as anode. A cathode-anode synergy is verified for coelectrolysis of soluble GeO2 and in situ–generated CO2 at the carbon anode to cathodic Ge nanoparticles encapsulated in carbon nanotubes (Ge@CNTs), contributing to enhanced oxygen evolution at carbon anode and hence reduced CO2 emissions. When evaluated as anode materials for lithium-ion batteries, the Ge@CNTs hybrid shows high reversible capacity, long cycle life, and excellent high-rate capability. The process contributes to metallurgy with reduced carbon emissions, in operando CO2 fixation to advanced energy materials, and upgraded conversion of carbon bulks to CNTs.


The heavy dependence on fossil fuels and carbon-intensive industrial processes has raised notable concerns with the global energy and environmental crises (1). As sustainable alternatives to these processes, the electrochemical technologies using green, trace-free, and activity-tunable electrons as energy carriers have been actively perused in diverse areas, such as electrochemical CO2 reduction (25), lithium-ion batteries (LIBs) (610), and electrometallurgy (1113).

CO2 reduction is an imperative and demanding task (1, 5). The difficulty in chemical reduction of CO2 associates with the high bond energy of C═O (14). Besides, separation of CO2 from its practical sources (e.g., the atmosphere and industrial flue gas) is technologically challenging and costly but unavoidable for the massive deployment (15, 16). Considering these challenges and the fact that CO2 is an exploitable resource, the in operando capture and utilization of CO2 are more promising from the viewpoint of real applications. On the other hand, the implementation of high-performance LIBs is another route to reduce carbon emissions (610, 1721). As an effective method to enhance the performance of LIBs, encapsulating high-capacity electrode materials (e.g., Ge and Si) in a carbon matrix can buffer volume variation during the alloying/dealloying process and ensure structure integrity of the electrode for cycling stability (1820). Therefore, it is appealing to use CO2 as the carbon source to encapsulate the electrode materials, which can enable CO2 fixation and improve battery performance simultaneously (22, 23). The related works, however, are seldomly achieved.

Molten salt electrolysis of metal oxides is a typical industrial process for extraction of aluminum and rare earth metals (11, 2426). This process uses carbon as the anode and causes huge CO2 emissions (11). For example, hundred million tons of CO2 are annually released on the carbon anode during molten salt electrolysis of alumina (2Al2O3 + 3C = 4Al + 3CO2) for primary aluminum production (27, 28). Using carbon anode–generated CO2 in a molten salt electrolyzer as carbon source to encapsulate high-capacity electrode materials can realize both the performance improvement of energy materials and the reduction in CO2 emissions. Planting O2 evolution on a carbon anode is also beneficial to reduce carbon footprints, which unfortunately is a thermodynamically uphill and kinetically sluggish reaction in O2−-containing systems. Therefore, rationally manipulating the electrode mechanism of the carbon anode to simultaneously promote in operando CO2 fixation and facilitate O2 evolution is highly desirable for electrometallurgical technology to diminish CO2 generation.

In this study, we report an unprecedented cathode-anode synergy for metallurgy in a molten salt electrolyzer with reduced carbon emissions. The electrolyzer consists of molten NaCl-CaCl2-CaO (750°C) as electrolyte, soluble GeO2 as Ge feedstock, conductive substrate (e.g., C, Cu, Ni, or Fe) as cathode, and carbon as anode. Coelectrolysis of soluble GeO2 and in situ–generated CO2 produces cathodic Ge nanoparticles encapsulated in carbon nanotubes (Ge@CNTs), with the evolution of O2 on the carbon anode being enhanced at the same time. As schematically illustrated in Fig. 1A, soluble GeO2 is electroreduced to Ge on the cathode (red arrow in Fig. 1A, reaction 1 in Fig. 1B), and at the same time the released O2− discharges at the carbon anode and generates CO2 (red arrow in Fig. 1A, reaction 2 in Fig. 1B). In the melts, the existed O2− can render a swift capture of in situ–generated anodic CO2 to form CO32− (black arrow in Fig. 1A, reaction 3 in Fig. 1B) (29, 30). As a result, electrometallurgy of Ge from GeO2 and electrodeposition of carbon from CO32− occur concurrently at cathode (black arrow in Fig. 1A, reaction 4 in Fig. 1B), while the evolution of O2 is correspondingly enhanced on the carbon anode (blue arrow in Fig. 1A, reaction 5 in Fig. 1B). The capture of anodic CO2 is greatly accelerated by the purposely added O2− (3032), which provides sufficient supplement of CO32− for continuous electrodeposition of carbon on the cathode. Such a cathode-anode synergy bridges a closed-loop carbon cycle in the molten salt electrolysis (reaction 6 in Fig. 1B), affording a theoretical decrease of CO2 emissions by 44% based on life cycle assessment. In particular, the in operando capture and conversion of CO2 (i.e., CO2 emission, capture, and conversion are fulfilled concurrently in one integrated step) can avoid the separation of CO2 from post-emission sources and heat loss. Liquid CaxGe is revealed as the intermediate for catalyzing the formation of the tubular nanostructure (Fig. 1C). In the initial stage, a small amount of CaxGe is formed as the catalyst for carbon nanotube (CNT) formation (Fig. 1C), with the detailed reactions as follows2xCaO(l)+2GeO2+x+2Canode=2CaxGe+(x+2)CO2(x+2)CO2+(x+2)CaO(l)=(x+2)CaCO3(l)(x+2)CaCO3(l)=(x+2)CaO(l)+(x+2)CCNT+(x+2)O2Overall reaction:2xCaO(l)+2GeO2+x+2Canode=2CaxGe+(x+2)CCNT+(x+2)O2

In the subsequent stage, formation of CaxGe is not needed, and the reaction 6 in Fig. 1B becomes dominant, which means that the reaction mechanism is as followsGeO2+Canode=Ge+CO2CO2+CaO(l)=CaCO3(l)CaCO3(l)=CaO(l)+CCNT+O2Overall reaction: GeO2+Canode=Ge+CCNT+O2

The Ge@CNT composite is harvested readily by rinsing the cathodic product in water to remove Ca from CaxGe. The obtained Ge@CNT hybrid shows excellent performance as an anode material for LIBs. The results imply that the cathode-anode synergy in a molten salt electrolyzer can contribute to the metallurgy process with reduced carbon emissions and value-added conversion of CO2 to high-performance anode materials for LIBs.

Fig. 1 Mechanisms of the cathode-anode synergy and morphology evolution.

(A) Schematic illustration on coelectrolysis of soluble GeO2 and in situ–generated CO2 at carbon anode to cathodic Ge@CNTs and anodic O2 in molten NaCl-CaCl2-CaO. (B) The corresponding reactions. (C) Formation mechanism of Ge@CNTs.


Electrochemical mechanisms

Constant voltage electrolysis at 2.2 V between a graphite anode and a carbon paper cathode is conducted in NaCl-CaCl2 molten salt at 750°C containing 2 weight % (wt %) CaO and 0.3 wt % GeO2. CaO is added to accelerate the capture of anodic CO2 to form soluble CaCO3 (3032). Complete dissolution of GeO2 is realized by keeping the GeO2 pellet and CaO in the molten salt for 24 hours (28, 33). By comparing the two red lines in Fig. 2, an enhanced evolution of O2 on the carbon anode is observed in the case of “2 wt % CaO and soluble GeO2.” In addition, the formation of CO2 on the carbon anode is decreased for “2 wt % CaO and soluble GeO2” by comparing the two gray lines in Fig. 2. The overall anodic current efficiency is high up to 86% in the case of “2 wt % CaO and soluble GeO2” (fig. S1A). Specially, the current efficiency for formation of O2 increases from a negligible value in the case of “0 wt % CaO and solid GeO2” to 35% in the case of “2 wt % CaO and soluble GeO2” (fig. S1A). Besides, the anodic current efficiency for CO2 generation decreases from 78 to 51% after addition of 2 wt % CaO (fig. S1A) (31). The above results indicate that the deliberately added CaO triggers the facilitated evolution of O2 on the carbon anode, which results in the coevolution of O2 and CO2 on the carbon anode and substantial reduction in CO2 emissions. When a SnO2 anode is used for electrolysis, only O2 is detected, with negligible formation of CO2 (fig. S2A). This phenomenon suggests that the evolution of CO2 comes from the carbon anode. In addition, the control experiments under varied electrolysis conditions indicate that the added CaO can facilitate the transportation of anodic CO2 to the cathode for CO32− formation, which finally boosts the electrodeposition of carbon at the cathode (fig. S1, C to H). Using gaseous CO2 as a carbon source, Ge@CNT can be also obtained with an SnO2 anode (fig. S2, B to F). However, the long-term electrolysis of GeO2 in molten salts using an SnO2 anode is not available because of the corrosion and dissolution of SnO2 in molten salts (Fig. 2F) (11, 13).

Fig. 2 Concentration variations (ΔC) of CO2 and O2 during the electrolysis of GeO2 under different conditions.

(A) Two weight percent CaO and soluble GeO2. (B) Zero weight percent CaO and solid GeO2.

Powder x-ray diffraction (XRD) patterns of the cathodic samples are indexed to cubic Ge [Joint Committee on Powder Diffraction Standards (JCPDS) no. 04-0545] (fig. S3A) (33, 34). The weak and broaden peak (the dashed square in fig. S3A) in the XRD pattern of the sample obtained in the presence of CaO is attributed to the (002) plane of carbon (JCPDS no. 01-0646), and this peak becomes more clear after leaching the sample in a dilute HNO3 solution to remove Ge (fig. S3B) (32). In the Raman spectra, the four peaks at 1350 (D band), 1576 (G band), 2700 (2D band), and 2925 cm−1 (D + G band) correspond to CNTs; and the peak at 300 cm−1 relates to crystalline Ge (fig. S3, C and D) (33, 35). X-ray photoelectron spectroscopy further verifies the coexistence of carbon and Ge in the cathodic product (fig. S4), consistent with results of XRD and Raman tests. The detected GeO2 in the product may originate from the minor surface oxidation of Ge metal in air (fig. S5A) (33). The Ge content in the composite is determined to be ~39 wt % by inductively coupled plasma optical emission spectrometry (ICP-OES), which is very close to the value (38.4 wt %) obtained by thermogravimetric analysis (TGA; fig. S5B) (18). The Ge content in the cathodic product can be easily tuned by varying the electrolysis voltage (fig. S5, B to K). The overall cathodic current efficiency is 80%, involving the contribution of 32 and 48% for the generation of Ge and CNT, respectively (fig. S1B). These findings indicate the formation of Ge@CNT composite through the coelectrolysis of GeO2 and in situ–generated anodic CO2.

In NaCl-CaCl2 molten salt, solid GeO2 is electrodeoxidized to cathodic Ge (reaction 1 in Fig. 1B) and anodic CO2 (reaction 2 in Fig. 1B) (11). In this case, O2− diffuses into the melts and discharges at the carbon anode. As shown in Fig. 3A, the standard potential of O2− discharging for CO2 formation (green line) is 1.03 V more negative than that for O2 evolution (black line), agreeing well with the generation of CO2 at the carbon anode (Fig. 2B). With addition of CaO, the anodic CO2 reacts with O2− to form soluble CO32− (reaction 3 in Fig. 1B) (3032). In such a scenario, both CO32− and O2− might discharge at the carbon anode. The magenta and blue lines in Fig. 3A show that the standard potentials for formation of O2 and CO2 from CO32− at a carbon electrode becomes comparable, suggesting a facilitated O2 evolution at the carbon anode upon occurrence of CO32−. Therefore, with the accumulation of CO32− at the interface between electrode and melt (CO2 + O2− = CO32−), the evolution of O2 and CO2 takes place simultaneously on the carbon anode (as shown in Fig. 2A). A large part of CO2 can be captured by abundant O2− in the melt to form CO32− on the occasion of its generation, resulting in decreased CO2 emissions and increased O2 evolution in the electrolysis (Fig. 2A). As depicted by reaction 3, CO32− is generated at the expense of O2− (30), indicating increased activity of CO32− and decreased activity of O2− in the carbon/molten salt interface. Resultantly, CO2 formation is restrained (blue line in Fig. 3A), while O2 evolution is kinetically promoted (magenta line in Fig. 3A). The equilibrium potentials for O2 evolution is thus more negative than that for CO2 formation, rendering enhanced evolution of O2 at the carbon anode (Fig. 2A). Such a cathode-anode synergy contributes to electrodeoxidation of GeO2 (reaction 1 in Fig. 1B) and electrodeposition of carbon on cathode (reaction 4 in Fig. 1B) (30), together with enhanced O2 evolution at the anode (reaction 5 in Fig. 1B). As a result, GeO2 and graphite anode are converted to cathodic Ge@CNTs and anodic O2, with great reduction in CO2 emissions (reaction 6 in Fig. 1B).

Fig. 3 Thermodynamic considerations and carbon emissions.

(A) Thermodynamic data based on HSC Chemistry 7.0 and (B) a comparison of theoretical carbon emissions based on life cycle assessment. kg CO2 eq., equivalent carbon emissions.

With 100 kWh of electricity input for electrolysis, the cathode-anode synergy process of “GeO2 + Canode = Ge + CCNT + O2” consumes 4.26 kg of GeO2 and 4.61 kg of graphite, producing 2.95 kg of Ge, 4.61 kg of CNT, and 1.30 kg of O2. The equivalent carbon emissions (CO2 eq.) based on life cycle assessment (Fig. 3B) are 113.8 kg (CO2 eq.). With the same input electricity, the process of “GeO2 + Canode = Ge + CO2” leads to a much higher carbon emission of 204 kg (CO2 eq.), with an occurrence of 44.4 kg of GeO2, 5.10 kg of graphite, 30.8 kg of Ge, and 18.7 kg of CO2. Compared with the electrolysis of solid GeO2 without CaO in the molten salts, the developed cathode-anode synergy mechanism achieves a theoretical decrease of carbon emissions by 44%.

Microstructure of cathodic product

Morphology of the Ge@CNT composite is revealed by field-emission scanning electron microscopy (FESEM) and transmission electron microscopy (TEM). The FESEM images and TEM images show that homogeneous CNTs with embedded Ge nanoparticles are achieved, and the diameter of CNTs is about 50 to 100 nm (Fig. 4, A to D). Further TEM analyses display that Ge nanoparticles with an average size of 3 to 10 nm are anchored on the inner wall of CNTs (Fig. 4, E and F). In the high-resolution TEM (HRTEM) image (Fig. 4G), the distinct lattice fringes with an interlayer spacing of 0.32 nm correspond to the (111) crystal plane of cubic Ge (18, 34). Besides, the clear interlayer distance of 0.37 nm assigned to the (002) crystal plane of carbon is observed on the wall of CNTs (17). The distribution of Ge nanoparticles inside CNTs is examined by the high-angle annular dark-field scanning TEM (HAADF-STEM) image (Fig. 4H) and the corresponding energy-dispersive x-ray spectroscopy (EDS) elemental mapping images (Fig. 4, I to K). A closer observation from the HAADF-STEM image (Fig. 4L) and the EDS spectrum (Fig. 4M) further suggests the embedded Ge nanoparticles inside CNTs, as schematically illustrated in Fig. 4N. At higher voltages, the tubular space inside the CNT is completely stuffed with Ge particles, suggesting that the Ge content in Ge@CNT can be adjustable by tuning the electrolysis conditions (fig. S5, C to K). In comparison, metallic Ge nanowires or nanoparticles (fig. S6, A to D) are obtained from electrolysis of solid GeO2 in NaCl-CaCl2 molten salts (33).

Fig. 4 Microstructure characterizations of the cathodic product obtained from electrolysis of soluble GeO2 in NaCl-GaCl2-GaO molten salt.

(A and B) FESEM images, (C to F) TEM images, (G) HRTEM image, (H and L) HAADF-STEM image, and (I to K) the corresponding elemental mappings of C, O, and Ge. (M) EDS spectrum of the crossline-marked point in (L). a.u., arbitrary units. (N) Structure illustration of Ge@CNT.

Like the classic chemical vapor deposition process, a liquid metallic catalyst is the prerequisite for CNT formation (17, 36). The CaxGe alloys are herein verified as the catalytic seed in the molten salt electrolysis strategy. According to the Ge-Ca phase diagrams (fig. S7A), the melting point of CaxGe alloys (745°C) is lower than the electrolysis temperature (750°C). CaxGe alloys have been confirmed as a cathodic product of the electrolysis of solid GeO2 in molten NaCl-CaCl2 at high overpotentials (also see fig. S7B) (33). After electrolysis for 1 min, only sparse particles are observed (Fig. 5A, region a), and the particle size matches well with the diameter of CNTs (Fig. 4). Further operating the electrolysis to 10 min leads to the evolution of a core-shell structure (Fig. 5B, region b), which is obviously rooted out from the liquid seeds (17, 36). After reaction for 20 min, CNTs begin to appear (Fig. 5C, region c), and the catalyst seed in the tip of CNT is also observed (Fig. 5, D to F). According to the HRTEM image (Fig. 5F), the interlayer spacing in the tip area is 0.32 nm, corresponding to the (111) crystal plane of cubic Ge (18, 34). The elemental mappings also reveal the distribution of Ge in the tip of CNT (Fig. 5, H to J).

Fig. 5 Morphology evolution.

(A to C) FESEM images of cathodic samples after electrolysis for (A) 1, (B) 10, and (C) 20 min. (D to F) TEM images of cathodic sample after electrolysis for 2 hours. (G) HAADF-STEM image and (H to J) the corresponding elements mapping images of cathodic product after rinsing in water. (K to P) Characterization results of cathodic product after rinse in dimethyl sulfoxide: (K and O) HAADF-STEM images and (L to N) the corresponding elements mapping images, and (P) the EDS spectrum of point 3 in (O).

The Ge@CNT material is harvested by rinsing the cathodic product in water to remove Ca from the CaxGe alloys [CaxGe+2xH2O = Ge + xCa(OH)2 + xH2] (33). When the cathodic product is washed by dimethyl sulfoxide (33), both Ca and Ge are observed in the tip of CNTs (Fig. 5, L to N). Moreover, the EDS spectrum reaffirms the presence of Ca and Ge in the tip (Fig. 5, O and P), which further suggests the formation of CaxGe during the molten salt electrolysis process. It is found that CNTs are only formed when the temperature exceeds the melting point of CaxGe (fig. S7, C to E). Also, CNTs are generated only at a higher voltage (fig. S7, F and G), in line with generation of CaxGe at higher overpotentials (fig. S7B). In addition, the cathode substrates have negligible influence on the evolution of CNTs (fig. S8), as CNT formation is also observed with Cu, Ni, Fe, or carbon paper as the cathode. The above results indicate that CaxGe alloy is the catalyst seed for CNT formation during the molten salt electrolysis (Fig. 1C).

The control experiments demonstrate that when SnO2 is adopted as the anode, Ge particles are the cathodic product without the formation of any CNTs, suggesting that cathodic CNTs originated from the graphite anode (fig. S9, A and B). In addition to graphite, amorphous carbon anodes can also be converted to cathodic CNTs (fig. S9, A and D), which reflects the universality of the molten salt electrolysis for upgraded conversion of low-grade carbon bulks to value-added CNTs.

Performance of Ge@CNTs for lithium storage

The as-prepared Ge@CNT composite is evaluated as a negative electrode material for LIBs. In cyclic voltammograms (CVs) (Fig. 6A), the cathodic peak at around 1.1 V in the initial cycle, which disappears in the following cycles, is attributed to the irreversible formation of solid electrolyte interface (1820, 3739). The cathodic peaks between 0.6 and 0.01 V represent the formation of LixGe alloy (6, 3739). The corresponding oxidation peaks from 0.15 to 0.55 V are attributed to the dealloying reactions of LixGe (1820, 3739). Almost overlapped CV curves are observed from the second cycle, indicating the high reversibility of the electrode (6, 18, 3739). The comparison of charge-discharge profiles of Ge@CNTs and the commercial CNT (C-CNT; fig. S10, A and B) electrodes indicates the enhanced performance of Ge@CNTs for lithium storage (Fig. 6B). The first Li-uptake capacity of Ge@CNTs and C-CNTs is 1742 and 489 mAh g−1, with the initial Coulombic efficiencies (ICE) being 66 and 24%, respectively. Moreover, the first Li-uptake capacity and ICE of Ge@CNTs are also higher than the corresponding values of metallic Ge obtained by electrolysis of solid GeO2 pellets (fig. S6, E and F).

Fig. 6 Lithium storage capability.

(A) CV curves swept at 0.1 mV s−1, (B) galvanostatic charge-discharge curves at 200 mA g−1, and (C) cycling performance of Ge@CNT electrode. (D) Rate capability of Ge@CNT and C-CNT electrodes. (E) Cycling performance of Ge@CNT and C-CNT electrodes at different current densities.

The cycling performance and rate capability of Ge@CNTs, C-CNTs, and metallic Ge electrodes are investigated and compared. The Ge content in Ge@CNT is determined to be ~39 wt %, which means a theoretical capacity of 851 mAh g−1 (1600 × 39% + 372 × 61%) (18, 20, 21). Such a relatively low amount of Ge is advantageous to reduce the total cost of the composite electrode (18, 20). As shown in Fig. 6C, after 110 continuous cycles at 200 and 500 mA g−1, the capacities of the Ge@CNT electrode stabilize at 801 and 634 mAh g−1, corresponding to 94 and 75% of the theoretical value. After 350 cycles at 500 and 1000 mA g−1 (fig. S10C), the reversible capacity is maintained at 401 and 288 mAh g−1, respectively. The Ge@CNT electrode well maintains its original microstructure (fig. S10D), and the film remains closely attached onto the Cu foil substrate (inset in fig. S10D) after 350 cycles at 500 mA g−1, revealing high stability of the Ge@CNT electrode (17, 34). The capacity based on Ge content is high up to 2053 mAh g−1 at 200 mA g−1, suggesting a very high utilization efficiency of Ge atoms in Ge@CNT (18). It should be mentioned that the capacity can be further improved by increasing the Ge content in Ge@CNT, which can be realized by electrolysis at higher voltages (fig. S5, B to K).

The Ge@CNT electrode also exhibits greatly improved rate capability compared to the C-CNTs and metallic Ge electrodes (Fig. 6D and fig. S6F). The Ge@CNT electrode delivers a reversible capacity of 438 mAh g−1 at 5000 mA g−1, which is more superior than the C-CNTs (29 mAh g−1 at 5000 mA g−1) and metallic Ge (200 mAh g−1 at 1000 mA g−1) electrodes. The long-term stability and high-rate capability of Ge@CNTs are further highlighted by a reversible capacity of 276 mAh g−1 after 1000 cycles at a large current density of 2000 mA g−1 (Fig. 6E), corresponding to an average capacity decay rate of only 0.04% per cycle (18). The rate performance in Fig. 6D and long-term cycling performance in Fig. 6E are successively tested using the same battery. Therefore, a part of the capacity is already lost after 90 cycles of test in Fig. 6D, leading to the capacity of Ge@CNT at 2000 mA g−1 in Fig. 6E slightly lower than that in Fig. 6D. The performance of the Ge@CNTs for lithium storage is comparable to that of similar Ge/C composites in literatures (6, 1721, 34, 3739). The enhanced lithium storage capability of Ge@CNTs is mainly attributed to the unique structural and compositional merits (18, 20, 34, 38). To be specific, the encapsulated Ge core can provide high capacities for Li+ storage, while the CNT shell with large free space can buffer volume variation during lithiation/delithiation process to ensure excellent structure stability and mechanical integrity (18, 20, 21). Meanwhile, the CNT sheath is Li+ permeable and can retard oxidation of Ge to avoid formation of thick GeOx that may obstruct permeability of Li+ ions (17, 21, 34). Moreover, the Ge nanoparticles are intimately attached to the inner wall of CNTs, which can effectively promote the charge transfer and increase the utilization efficiency of Ge atoms (20, 21), permitting enhanced rate capability.


In summary, we demonstrate an unprecedented cathode-anode synergy in a rationally designed molten salt electrolyzer for metallurgy with reduced carbon emissions. In molten NaCl-CaCl2-CaO, coelectrolysis of soluble GeO2 and in situ–generated CO2 at a carbon anode results in cathodic Ge@CNTs. Such a cathode-anode synergy triggers enhanced evolution of O2 at the carbon anode. The upgraded conversion of carbon bulks to CNTs is achieved with liquid CaxGe intermediate as the catalyst. When evaluated as an anode material for LIBs, the Ge@CNT electrode shows a high reversible capacity (801 mAh g−1 at 200 mA g−1, 94% of the theoretical value), superior cycling performance (a fading rate of 0.04% per cycle in ~1000 cycles at 2000 mA g−1), and excellent high-rate capability (438 mAh g−1 at 5000 mA g−1). The molten salt electrolyzer has a similar configuration to the Hall-Heroult cell for industrial Al production. The technological maturation promises a large-scale deployment. Joule heat formed by the current that flowed through the resistant and high-heat capacity molten salts can offset heat loss during molten salt electrolysis, promising a self-heating process like the industrial Al production occurring at much higher temperatures. These merits make the process an up-scalable, high-valuable, and sustainable protocol for green metallurgy and in operando CO2 conversion.


Molten salt electrolysis

All used reagents were of analytical purity and used as received. NaCl, CaO, GeO2, and C-CNTs were purchased from Sinopharm Chemical Reagent Co., Ltd. CaCl2 (anhydrous, >99% purity) was obtained from Shanghai Titan Scientific Co., Ltd. Electrolysis was conducted in eutectic NaCl-CaCl2 molten salt (500 g in total mass, with a molar ratio of 48:52), with a graphite rod as the anode (20 mm in diameter). An alumina crucible (75 mm by 125 mm in inside diameter × height) containing the salts was put into an alumina tube reactor, followed by heating at 400°C for 24 hours to remove the residual moisture. After that, the reactor was heated to 750°C. Pre-electrolysis was performed at 2.6 V for 12 hours using a nickel plate as the cathode (30 mm by 20 mm). For preparation of Ge@CNTs, 2 wt % (10 g) of CaO and 0.3 wt % (1.5 g) of GeO2 were added into the molten salt. Before electrolysis, the molten salt was kept at 750°C for 24 hours to ensure complete dissolution of CaO and GeO2. For comparison, electrolysis without addition of CaO was also conducted (only GeO2 is added in the molten salt). A carbon paper (30 mm by 20 mm; WOS1002, purchased from Cetech Co., Ltd.) was used as the cathode for the above-mentioned two cases. Solid GeO2 pellets (diameter, 10 mm; 1.5 g) prepared by die pressing of GeO2 powder at 8 MPa were also electrolyzed in the same molten salt without addition of CaO. All the electrolysis was conducted at 2.2 V unless otherwise mentioned specifically. During electrolysis, the reactor was continuously flushed by a high-purity Ar flow (150 ml min−1). The concentration of CO2 and O2 in the outlet was detected by an online multicomponent gas analyzer (NGP50-M500, Ningbo Gaopin Technology Co., Ltd.). The concentration variation of O2 was obtained by subtracting the lowest value of O2 concentration detected during 2 hours of electrolysis period. Correspondingly, the concentration variation of CO2 was obtained by the similar method. After electrolysis, the cathodic products were obtained by thorough washing in dilute hydrochloric acid and deionized water and then vacuum-dried at 60°C. To avoid surface oxidation, all the obtained samples were stored in an Ar-filled glove box [Super, MIKROUNA; H2O < 0.1 parts per million (ppm), O2 < 0.1 ppm]. To investigate the influence of cathode materials on electrolysis products, Ni (purity >99.9 wt %), Cu (purity >99.9 wt %), and Fe (purity >99.9 wt %) plates (30 mm by 20 mm; provided by Sinopharm Chemical Reagent Co., Ltd.) were also used as the cathodes, respectively. In one case, a SnO2 rod (20 mm in diameter, provided by Dyson Thermal Technologies) was used. Specially, an amorphous carbon rod (20 mm in diameter, provided by Beijing Dragon, Carbon Crystal Technology Co., Ltd.) without undergoing a high-temperature graphitization process was used as an anode to investigate the influence of anode materials. CVs of GeO2 were conducted in the same NaCl-CaCl2 molten salt at 750°C. A three-electrode system controlled by an electrochemical workstation (Solartron Analytical 1470E) was adopted. A homemade GeO2 powder–modified Mo-cavity plate, alumina-sealed Ag/AgCl (NaCl-CaCl2 melt containing 2 wt % AgCl), and graphite rod were used as the working electrode, reference electrode, and counter electrode, respectively.

Materials characterizations

XRD patterns of samples were obtained on Rigaku MiniFlex600 at a scan rate of 4° min−1 with Ni-filtered Cu Kα radiation (λ = 1.5406 Å). A field-emission scanning electron microscope (ZEISS Sigma; extra high tension, 15.00 kV; working distance, 5.7 mm), a transmission electron microscope (JEM2010-HT; 200 kV), and a high-resolution transmission electron microscope (JEM 2010-FEF; 200 kV) were used to analyze the morphology and structure of samples. The composition and elemental mapping images of the samples were analyzed by EDS (GENESIS 7000 and OXFORD IET 200) attached to a TEM apparatus. X-ray photoelectron spectra were collected on an x-ray photoelectron spectrometer (ESCALAB250Xi, Thermo Fisher Scientific) with the results calibrated by C 1 s (284.8 eV). Raman analysis was conducted on a Renishaw inVia Raman microscope with a 325-nm Ar+ laser as the excitation source. The sample was dissolved in a dilute HNO3 solution, and then the Ge contents in the sample were determined by analyzing the Ge concentration in the HNO3 solution using an Agilent ICPOES 730 ICP-OES. TGA of the sample in air was carried out using a Mettler Toledo DSC1 TGA apparatus at a heating rate of 5°C min−1 from room temperature to 900°C.

Electrochemical characterizations

The slurry was prepared by homogenously mixing the active material, acetylene black, and binder in an agate mortar with the weight ratio of 7:2:1, which was then pasted onto a Cu foil (12 μm in thickness). The binder was composed of sodium carboxymethyl cellulose and polyacrylic acid (CMC/PAA, 1:1 in weight). After drying at 120°C in vacuum for 12 hours, small Cu disks (1.2 cm in diameter) containing the active materials (ca. 1.5 mg cm−2) were obtained by cutting the Cu foil. Using the electrode disk as the working electrode, the metal Li foil (1.6 cm in diameter) as the counter electrode, 1 M LiPF6 in ethylene carbonate and diethyl carbonate (EC/DMC, 1:1 in weight) as the electrolyte, and polyethylene (25 μm in thickness) as the separator, CR 2025 coin-type cells were assembled in an Ar-filled glove box (Super, MIKROUNA; H2O < 0.1 ppm, O2 < 0.1 ppm). The electrochemical analysis of charging and discharging was performed on a LAND battery tester in voltage range between 0.01 and 1.5 V. CV tests were conducted on the Solartron Analytical 1470E workstation by using a scanning speed of 0.1 mV s−1 at 0.01 to 1.5 V.

Life cycle assessment

Life cycle assessment of the two processes (i.e., GeO2 + Canode = Ge + CO2 or GeO2 + Canode = Ge + CCNT + O2) was calculated on the basis of CO2 eq. collected from the Life Cycle Inventory (LCI) database. The equivalent carbon emissions of 1 kWh of electricity (based on the global average value), 1 kg of graphite, and 1 kg of GeO2 are 0.928, 3.43, and 1.69 kg (CO2 eq.), respectively. Data for GeO2 are not found in the LCI database. The carbon emission for commodity was mainly determined by the production process. Most germanium resources are found as associated element in lead-zinc deposit, which means that the extraction process of GeO2 is similar to that of ZnO. Therefore, the data of ZnO were prudently used for calculating the carbon emission for GeO2 production.


Supplementary material for this article is available at

Fig. S1. Current efficiency and digital photos of the carbon paper cathodes before and after electrolysis.

Fig. S2. Concentration variations of outlet gas using a SnO2 anode and morphology of cathodic product using gaseous CO2 as carbon source.

Fig. S3. XRD patterns and Raman spectra of different samples.

Fig. S4. X-ray photoelectron spectroscopy spectra of the cathodic product.

Fig. S5. XRD patterns of the obtained cathodic products kept on different conditions, TGA curves, TEM images, and the corresponding elements distribution images of cathodic products obtained at various voltages.

Fig. S6. XRD patterns and FESEM images of metallic Ge obtained from electrolysis of solid GeO2 and its performance as anode materials in LIBs.

Fig. S7. Ge-Ca phase diagram, CV curve of GeO2, and morphology characterizations of cathodic products obtained on various conditions.

Fig. S8. FESEM images of cathodic products obtained using different cathode substrates.

Fig. S9. XRD patterns and the corresponding FESEM images of cathodic product using various anode materials.

Fig. S10. Morphology (FESEM and TEM images) of C-CNTs, the cycling performance of Ge@CNTs and C-CNTs electrodes, and the FESEM images of the electrode after cycling tests.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: W.X. and W.W. acknowledge the funding support from the National Key R&D Program of China (2018YFE0201703), the National Natural Science Foundation of China (51722404, 51674177, 51804221, and 91845113), the China Postdoctoral Science Foundation (2018 M642906 and 2019 T120684), and the Hubei Provincial Natural Science Foundation of China (2019CFA065). Author contributions: W.W. and B.J. performed molten salt electrolysis, materials characterizations, and LIB tests. Z.W. performed the life cycle assessment. W.X. conceived the idea, analyzed all data, and wrote the manuscript. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials 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.

Stay Connected to Science Advances

Navigate This Article