Innovative technologies are urgently demanded for capturing and converting CO2 into value-added species, at low input energy, to mitigate the negative effects of this greenhouse gas and support a sustainable carbon cycle1, 2. Activating CO2 into CO2•‒ radicals or other intermediates is a crucial step for CO2 conversion, while the inertness of CO2 molecules imposes a significant challenge3. To this end, external energy is required, and catalytic systems are commonly engaged to lower the energy barrier for CO2 converting. The choices of both the applied energy and functional materials, and how they are pooled together for CO2 conversion govern the efficiency balances. To date, light and electricity have been extensively used as the triggering energies together with many well-known catalytic systems4-8. However, in general, the existing electro- or photo- empowered CO2 reduction systems suffer from sluggish reaction rates and high energy consumption. Additionally, there has also been no magical definitive catalyst, that could synergistically operate with these energy types to carry out the conversion at low costs and high efficiencies.
Mechanical energy is easy to generate and often a waste source of energy. Except for organic synthesis and polymerisation9, 10, mechanical energy has received little attention in catalytic systems and has not yet been utilised for CO2 reduction. In conductive particles suspensions, the introduction of mechanical stimuli can both increase the temperature11 and generate triboelectrification, as a result of the frictional contact and modulation of gaseous content solubility, respectively12-14.
Solid metallic catalysts have exhibited excellent performance for electrocatalytic or photocatalytic CO2 reduction15-17. However, the active sites of solid metallic catalysts can deteriorate under intense mechanical stimuli and/or can be deactivated when carbonaceous materials adhere onto the catalytic sites during CO2 reduction. In principle, the use of liquid metals, and specific solid structures, can solve these problems. Gallium (Ga)-based liquid metals have shown intriguing properties for catalysis, including tunability by the incorporation of other elements, and remarkable resistance to coking18-21 and also excellent mechanical tolerance. Additionally, it is known that Ga (0) can be oxidized to Ga (I) in the presence of organic materials22. In this work, we explore whether the same effect can also apply to CO2 and whether we can find a process to reduce Ga(I) back to Ga (0) in a closed cycle stimulated by mechanical agitation. We envision that by employing liquid metal mixes of Ga and a compound of Ag, a closed cyclic catalytic system can be developed. This system can then covert CO2 into value-added species in the reversible Ga-Ga+ cycle, which will be fully explored.
Assessment of CO2 conversion for various scenarios
We exploited a suspension of Ga and Ag (I) salt mixes as the precursors of the co-catalysts and ultrasound was initially employed to stimulate the CO2 reduction. Dimethylformamide (DMF), with good stability during mechanical agitation and high CO2 solubility of 0.14 M at 40 °C (to assure the liquid state of Ga), was chosen as the solvent23. We observed that during the reaction, the CO2 molecules near the interface of the suspended particles were reduced to form carbonaceous sheets in a process that will be detailed later.
The best outcomes were obtained when Ga and AgF were mixed in a DMF solution which also contained 0.10 M HCl to remove the native oxide on the surface of Ga. Ga and AgF were sonicated together (Fig. 1a - using a probe sonicator for 30 min) to generate sub-micron Ga droplets of 230 nm median diameters and Ag0.72Ga0.28 rods of micron/sub-micron lengths and median diameters of 160 nm (Fig. 1b and Supplementary Fig. 1). The characterisation outcomes revealing the elemental composition will be presented in a later section.
In the reactor, CO2 was bubbled into and dissolved in DMF through a diffuser (Fig. 1c). The dissolved CO2 is reduced to solid carbonaceous materials at the interface of the Ga droplets. The mechanically enforced CO2 conversion can be scaled up using a variety of mechanical sources that produce frictional contact. To examine this prospect, CO2 conversion by an overhead mixer was also performed and validated (Fig. 1d).
Due to the ultra-smooth nature of the liquid metal droplets, the produced carbonaceous materials on the surface are in the form of sheets18, 24. These low dimensional sheets, on the non-polarized liquid metal surface, are exfoliated during mechanical stimulation (Fig. 1e)18, 24. Most importantly, the carbon sheets migrate to the top of the reactor and can be isolated due to the density difference with reference to that of metallic components (Fig. 1c,d).
The qualitative and quantitative analyses of the production of carbon, when the Ga/AgF (7.0 to 1.0 mass ratio) suspension in DMF is utilised in a 20 mL reactor, are presented in Fig. 1f and Supplementary Fig. 2. The 7.0 to 1.0 mass ratio and the reaction temperature were chosen according to previously optimised data for C-C bond formation reactions22. Additionally, the performance of the system formed by direct alloying of Ga with silver (50 to 1.0 or 20 to 1.0 mass ratio of Ga/Ag, Supplementary Fig. 3e,f) and different silver salts (Fig. 1g-k), including AgCl, AgBr, AgI, AgOTf, AgNO3 (also 7.0 to 1.0 mass ratio) were compared. The homogeneous mixture (20 μL) was drop-casted onto a glass substrate and dried for Raman analysis, with the whole drop-cast region included during the Raman spectroscopy measurement (Supplementary Fig. 3a-d). The changes in the intensity of the carbon D and G bands at 1350 and 1600 cm-1, respectively, were employed to obtain the overall picture. Thermal gravimetric analyses (TGA) and gas chromatography (GC) were also conducted for comparative quantitative assessment of the solid carbon and gaseous products (Supplementary Table 1).
For the Ga/AgF system that exhibited the best performance, the production of carbon was observed in < 1 hour of reaction (Fig. 1f) and increased continuously over time according to TGA. The TGA showed that 4.95 mg of carbonaceous materials were produced per hour in a 20 mL reactor at a CO2 flow rate of ~10 sccm (Supplementary Fig. 2). In comparison, the AgCl, AgBr, AgI, and AgOTf mixes also presented CO2 conversion capability, but they were not as efficient as the AgF system (Fig. 1g-j, for brevity only Raman spectra are shown and not TGA). With no emerging D and G bands after 5 hours of reaction (Fig. 1k and Supplementary Fig. 3e,f), Ga/AgNO3 and Ga-Ag alloys were found to be ineffective for CO2 reduction. The reason for these different conversion activities will be discussed later.
To verify that the CO2 reduction relies on the synergism of Ga and AgF, experiments were conducted by employing Ga and AgF separately (Supplementary Fig. 3g,h), both of which resulted in undetectable carbon production. We note that our attempts with other types of salts (e.g. KCl and NaCl) (Supplementary Fig. 3i,j) and magnetic stirring (less powerful in comparison to ultrasonication and overhead stirring) (Supplementary Fig. 3k), showed no carbon formation, indicating the crucial role of the chosen silver salt (AgF) and a threshold for mechanical energy input into the reaction process. Controlled N2 bubbling also did not show any formation of carbonaceous products (Supplementary Fig. 3l).
We explored the minimum co-catalyst mass required in the system to maintain the efficiency of CO2 conversion. Diluting the catalyst by 10 times offered nearly the same conversion efficiency, still achieving an equivalent production of 4.75 mg of carbonaceous materials per hour at ~10 sccm CO2 bubbling rate (Fig. 1l and Supplementary Fig. 2), whereas the output was dramatically reduced for dilutions of 50 or 100 times (Supplementary Fig. 3m,n. TGA profiles are not shown for brevity).
The amount of CO2 dissolved in solution also significantly influenced the efficiency of CO2 conversion. Ethanolamine (ETA) is a suitable choice for increasing this amount since CO2 solubility is 5.6 M in pure ETA25 in comparison to 0.14 M in DMF at 40 °C. With the addition of 10 vol% ETA in DMF (referred to as DMF+ETA hereafter), CO2 was continuously reduced with a higher efficiency, producing 7.95 mg of carbonaceous materials per hour in the same reactor at ~10 sccm CO2 bubbling rate (Fig. 1m and Supplementary Fig. 2). Interestingly, 22.2 cm3 CO was also produced in one hour (Supplementary Fig. 4). In contrast, when DMSO or H2O were used, the efficiency was very low and carbon products could not be quantified by TGA, owing to their limited CO2 solubility (Supplementary Fig. 3o,p).
Carbonaceous materials characterisation
Carbonaceous materials produced from CO2 were isolated for further characterisation. Scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) analysis (Fig. 2a and Supplementary Fig. 5a) of the isolated carbonaceous materials reveal that the solid product consists of carbon and oxygen before any secondary washing, with trace quantities of the metallic species that can be easily removed. Fourier transform infrared (FTIR) spectroscopy (Supplementary Fig. 5b,c) further confirms that the carbonaceous materials are primarily comprised of C=C and C-O bonds18. Based on X-ray photoelectron spectroscopy (XPS) analysis, the C1s region of the carbonaceous materials shows characteristic peaks of sp2 carbon and C-O bonding, at 284.2 and 286.1 eV, respectively (Supplementary Fig. 5d)26. The presence of C-O bonds is validated from the O1s XPS region of the sample (Supplementary Fig. 5e). Transmission electron microscopy (TEM) and selected area electron diffraction (SAED) images demonstrate that part of the carbonaceous material is akin to that of slightly crystalline graphene oxide (Fig. 2b), and a certain proportion of the product exists in the amorphous state (Supplementary Fig. 5f).
Efficiency and energy requirement for CO2 conversion
The CO2 conversion efficiencies under different configurations were determined using TGA and GC measurements as summarised in Supplementary Table 1 (see Supplementary Information). The conversion efficiency is defined as (captured and reduced CO2/total input CO2)´100%, which are all obtained from the optimum mix of Ga/AgF (to be explained further in the next section).
In the small-volume reactor of 20 mL and height of 4.5 cm, which was used as the characterisation unit in the previous sections, the conversion efficiencies are 1.5% and 6.2%, for DMF and DMF+ETA cases, respectively (Fig. 2c). To demonstrate the scalability, we increased the dimensions of the reactor (Fig. 2d). When the height of the reactor was increased to 40 cm for only DMF solvent (volume of 500 mL), 27% of the input CO2 at the flow rate of ~8.6 sccm could be continuously captured and converted (Fig. 2c). Thus, when the total height of the reactor would be 148 cm, the CO2 conversion could reach the full capacity.
The height of the reactor for near-full conversion could be significantly decreased when DMF+ETA was used as the solvent as this combination could significantly increase the CO2 solubility. The conversion efficiency reached the unprecedented value of 92% (at the flow of ~8.0 sccm CO2) in this case for a reactor as small as 27 cm in height and 330 mL in volume (Fig. 2c and see Supplementary Fig. 6 for the photo of the set-up). The amount of produced O2 and carbonaceous materials (highly oxidised carbon), under different conditions, are presented in Supplementary Fig. 7 and Supplementary Table 1.
According to these measurements, the total energies required for converting 1.0 tonne of CO2, for DMF and DMF+ETA cases, are obtained as 699.5 kW∙h and 228.5 kW∙h, respectively (Supplementary Information). An overall estimation based on the current price of electricity suggests that the operational cost of CO2 capture and conversion using DMF+ETA is lower than any other state-of-the-art technologies27-31.
Analysis of the suspensions
To elucidate how the functional materials are formed during the probe sonication stage, the reactions between Ga and silver salts were investigated by characterising the sonication products. Sonicating Ga with AgF (as an exemplar) forms an intermetallic phase Ag0.72Ga0.28 (Fig. 3a) and GaF3. The existence of metallic Ag (as Ag0.72Ga0.28) is confirmed by the Ag3d XPS peaks at 367.8 and 373.8 eV (Fig. 3b)32. The metallic fluorides can be verified by the F1s XPS peak at 684.3 eV (Fig. 3c)33.
The compositions and morphologies of the materials were investigated and correlated with the CO2 reduction performance. As illustrated in Fig. 3a, XRD patterns of Ga mixed with silver salts, which lead to CO2 conversion (i.e., AgF, AgCl, AgBr, AgI and AgOTf), show the presence of Ag0.72Ga0.28 crystalline peaks. The control experiments show that Ag2Ga particles (generated from the sonication of Ga-Ag alloy, Supplementary Fig. 8) and Ag particle inclusions (using Ga/AgNO3 as the precursors, Fig. 3a,d) are inactive materials for CO2 conversion. These results suggest that the formation of Ag0.72Ga0.28 is a prerequisite for CO2 reduction. Interestingly, the Ag0.72Ga0.28 crystals, generated from different silver salts, show distinct morphologies (Fig. 3e-i and Supplementary Fig. 9) of particles (Fig. 3e-g) or rods (Fig. 3i for AgF) or a combination of both (Fig. 3h for AgCl) together with the Ga spheres. The presence of only rod-shaped morphology (Fig. 3i for AgF) is found to associate with an enhanced CO2 catalytic capability. The Ga/AgF system, which generated the highest efficiency for CO2 conversion, showed the rod-shaped Ag0.72Ga0.28, while Ag0.72Ga0.28 with non-rod morphology from other silver salts (or limited rod morphology for AgCl) exhibited limited catalytic abilities. The high-resolution TEM images, SAED pattern (Fig. 3j,k) and the TEM-based EDS mapping (Fig. 3l-n) further confirm the existence of the Ag0.72Ga0.28 rods and their growth direction along the  plane (Fig. 3k). As shown in Fig. 3o-q, the native oxide layer on the surface of the Ga droplets can be observed when dried for analysis. Furthermore, there were no obvious changes to the Ag0.72Ga0.28 structures after 5 hours of reaction according to both XRD (Supplementary Fig. 9x) and SEM (Supplementary Fig. 10), indicating that the Ag0.72Ga0.28 rods were resilient towards mechanical agitation.
The concentration of gallium and silver ions in solution during the reaction was measured by inductively coupled plasma mass spectrometry (ICP-MS) (Supplementary Fig. 11). The ion concentrations fluctuated without showing any increasing or decreasing trend, indicating that the catalysts are not consumed, and that the system is stoichiometrically stable.
Based on the configuration of the catalytic process, we can consider a scenario in which the contact of the Ga/DMF interface is altered by the interfacial formation of CO2 bubbles. CO2 bubbles are formed as the Ga/DMF interface becomes warmer due to localised friction11. As such, the interfacial solubility of CO2 in DMF decreases. The formation of bubbles induces a significant increase in the transient, capacitive, open circuit voltage through triboelectrification between the separated Ga conductive liquid metal and the DMF dielectric as previously described12, 13. The formation of a closed loop, by the Ag0.72Ga0.28 rods, thereafter, initiates the CO2 conversion. The presence of the triboelectric effect can be validated by a proof-of-concept macro sized experiment, as depicted in Supplementary Fig. 12. The details of triboelectric voltage calculations are presented in Supplementary Information.
The CO2 reduction in our system is completed through a reversible Ga-Ga+ cycle. Cyclic voltammetry was conducted to provide an insight into the catalytic mechanism of the Ag0.72Ga0.28 rods via the suggested route. The results showed that, for the working electrode containing Ga droplets and Ag0.72Ga0.28 rods as the co-contributors, Ga is oxidized to Ga+ at 0.18 V and then reduced to elemental gallium at -0.31 V (Fig. 4a)34. Considering that the triboelectric process generates time-dependent voltages of several volts, the carbonaceous sheets are rapidly produced on the surface of liquid metals as we have previously demonstrated35. We also note that Ga+ reduction was not observed when either Ga droplets (Inset in Fig. 4a) or Ga droplets with non-rod morphology Ag0.72Ga0.28 were used as the working electrode (Supplementary Fig. 13), showing the importance of the long rods for the cyclic reaction.
To investigate the reaction mechanism, we performed a detailed assessment of the reaction intermediates and the by-products generated during CO2 reduction. The overall reaction process in DMF is described by chemical reaction equations (1-6). The equations are separated into ‘liquid metal components’ reactions (equations (1-4)) and ‘solid components’ reactions (equations (5,6)). For the description of the liquid metal component reactions, a series of characterisations were conducted. Nuclear magnetic resonance (NMR) analysis showed that the solvent DMF was not involved in the reaction (Supplementary Fig. 14a). The CO2 reduction is realised via the voltage provided by the nano triboelectrochemical process on the surface of Ga liquid droplets that turn Ga into Ga+, while CO2 is activated into the CO2•‒ radical (equation (1)). The existence of the CO2•‒ radical during the reaction is demonstrated by electron paramagnetic resonance (EPR), which uses 5,5-dimethyl-1-pyrroline N-oxide (DMPO) as a radical trapping agent to form DMPO-CO2•‒ adduct for spectroscopic analysis (Fig. 4b, see Methods)36. The CO2 to CO2•‒ process is followed by the generation of the intermediates CO and O2- radicals (equation (2))37. The former is further converted to carbonaceous materials on the liquid metal surface (equation (3) which is the optimum case and it can be altered according to the C to O ratio in the obtained carbonaceous solid). The equations are as follows:
The description of the ‘solid components’ reactions is as follows. According to the cyclic voltammetry results, the oxidized Ga+ can be reduced to elemental Ga by receiving an electron from the Ag0.72Ga0.28 and the Ag0.72Ga0.28 turns into Ag0.72Ga0.28+ (equation (5)). The catalytic cycle is closed by the electrons provided from the O2- to O2 (equation (4)) to reduce Ag0.72Ga0.28+ back to Ag0.72Ga0.28 (equation (6)), where the existence of O2 can be confirmed through GC (Supplementary Fig. 4 and Supplementary Table 1).
The catalytic cycle of CO2 reduction at the interface of Ga droplets is illustrated in Fig. 4c. This catalytic mechanism also aligns with the DMF+ETA case, and the extra by-products are produced due to the presence of ETA that promotes the process towards CO production according to equation (5) (Supplementary Fig. 4 and Supplementary Fig. 14b)38.
Since the reaction is activated by the triboelectric potential, other forms of mechanical stimuli can also be applied, and the system can be readily scaled up. As a demonstration, we further validate our strategy by coupling an overhead stirrer to a 50 mL reactor. We found that CO2 conversion continuously takes place in a stable manner when the stirring speed exceeds a threshold of 200 rpm (at room temperature, Supplementary Fig. 15a) and the conversion efficiency increases along with the increase of the stirring speed (equivalently, the mechanical energy input) (Supplementary Table 1 and Supplementary Fig. 15b,c).