Carbon Stabilised Saponite Supported Transition Metal-Alloy Catalysts for Chemical CO2 Utilisation via Reverse Water-Gas Shift Reaction

Chemical CO2 upgrading via reverse water gas shift (RWGS) represents an interesting route for gas phase CO2 conversion. Herein, nature inspired clay-based catalysts are used to design highly effective materials, which could make this route viable for practical applications. Ni and transition metal promoted Ni saponite clays has been developed as highly effective catalysts for the RWGS. Saponite supported NiCu catalyst displayed a remarkable preference for the formation of CO over CH4 across the entire temperature range compared to the saponite supported NiCo and Ni catalysts. The NiCu sample is also highly stable maintaining ~ 55% CO2 conversion and ~ 80 % selectivity for CO for long terms runs. Very importantly, when compared with reference catalysts our materials display significantly higher levels of CO2 conversion and CO selectivity. This confirmed the suitability of these catalysts to upgrade CO2-rich streams under continuous operation conditions.


Introduction
Extensive use of fossil fuels is largely contributing to CO2 emissions and global warming.
The current efforts of the scientific community is to develop sustainable technologies that can reduce CO2 emissions or even reach net zero or negative emission through the capture and conversion of CO2 [1]. CO2 conversion into CO, CH4, cyclic carbenes, polymers, etc can be achieved by a variety of methods [2][3][4][5][6]. CO2 capture and utilisation to produce fine chemicals accounts for a small percentage of the emitted CO2 levels [7]. A possible way to reach net zero emissions of CO2 is to use fuels that are derived from emitted CO2 [8,9]. However, CO2 is a stable molecule that requires a great deal of energy to activate, and therefore, efficient and low-J o u r n a l P r e -p r o o f cost conversion methods and catalysts for CO2 activation and conversion are highly sought.
Current reseach in this area concern the use of Cr, Fe, Ni and Cu doped with a variety of materials (i.e. Ce, Cs , Zr or Y) or using photocatalysts in a photo-assisted revese water gas shift . Although the active phase varies quite considerably among these materials, the support media largely remains the same, usually metallic oxides like Al2O3, CeO2, ZrO2 or doped/mixed combinations.
Production of higher hydrocarbon fuels through processes like the Fischer-Tropsch (F-T) synthesis and hydrogenation of CO2 to form methanol via reverse-water-gas-shift reaction (CAMERE process) are promising routes to utilise emitted CO2 [10,11]. F-T synthesis and CAMERE processes reported better efficiencies (approximately 20%) when CO generated from RWGS reaction was used as raw material [11]. Conversion of CO2 to CO through RWGS reaction is shown in (eq. 1). The reaction is endothermic and as such, is expected to demonstrate increased efficiencies at higher temperatures. There exists a major competing exothermic methanation reaction (Sabatier process), (eq. 2) that occurs at lower temperatures producing methane. A highly unwanted material where F-T or CAMERE processes are concerned.
CO2 + H2↔ CO + H2O; ΔH298 = +41 kJ/mol (1) CO2 + 4H2↔ CH4 + 2H2O; ΔH298= -165 kJ/mol (2) As an additional drawbackthe Sabatier reaction consumes 4 moles of hydrogen per mol of CO2 thus imposing extra process cost. In this regard, if we aim to desing an efficient reverse water gas shift unit, it is of paramount importance to control the competition CO2-Methanation/RWGS to ensure the process is selective towards carbon monoxide. In this sense, a variety of catalytic materials has been investigated for RWGS reaction [12][13][14][15]. Noble metals, such as Au, Pt, Pd, Rh and Ru, exhibit high activity, stability and selectivity for CO2 reduction to CO. However, due to their cost and scarcity, it is desirable to replace these materials by introducing more ecominally appealing catalysts. Transition metals-based catalysts represent an economically interesting alternative [16]. For instance, copper-based catalysts were found to be more selective for CO production favouring RWGS reaction [17][18][19][20][21][22][23][24]. However, they suffered stability problems [25]. Modified Ni-based catalysts were also designed to achieve better selectivity and stability for RWGS reaction [26]. Bimetallic or metal alloy catalysts of Cu, Ni and Co exhibited activities comparable to noble metals and their alloys but, had stability issues [27]. Hence, in addition to catalyst activity and selectivity J o u r n a l P r e -p r o o f for CO, stability and sustainability of the chosen catalyst at reaction conditions are also critical.
The nature of the catalyst support is known to influence the coking characteristics, stability and dispersion of catalysts. Different supports interact differently with the active catalyst [28,29]. For example, CeO2 supported Au was more active than the TiO2 supported Au catalyst due to the higher oxygen mobility of CeO2 [30]. In situ generated carbon support due to decomposition of a metal organic framework precursor showed high stability of the catalyst [31]. Mixed oxide supports of CeO2/ZrO2 or CeO2/Al2O3 supporting Ni altered the activity and selectivity of the catalyst when compared to unsupported Ni [26,32]. Reduced surface acidity of Al2O3 modified with CeO2 caused lesser extent of coking and retained catalyst activity for a longer time [33]. Specialised methods such as magnetron sputtering, atomic layer epitaxy (ALE), atomic layer/chemical vapour deposition, etc have also been employed to increase catalyst activity and stability [34,35]. Avoiding sophisticated, expensive techniques and excessive use of chemicals in catalyst preparation would make the whole process commercially more viable.
Accordingly, it would be advantageous to design a sustainable synthesis of transition metal-based catalysts supported on low-cost, environmentally benign supports such as clays.

Synthesis of Catalyst
Ni(OH)2, Cu(OH)2 and Co(OH)2 were used as the transition metal sources, and 1adamantanecarboxylic acid, used as the carboxylate source (All chemicals were procured from Sigma Aldrich and used as received without further purification). Deionised water (18 MΩ.cm resistivity, Millipore water purification system) was used throughout the experiment. Na + -saponite, NaMg6(Si7Al)O20(OH)4, was synthesized hydrothermally by a procedure reported by Kawi and Yao [38].
In a typical synthesis of saponite supported NiCu-adamantanecarboxylate (NiCu-Ada/Sap), 1 g of Na + -saponite was stirred in 100 ml of water for 2 days at room temperature to produce exfoliated colloidal dispersion of saponite clay.  Raman spectra were recorded using Renishawin Raman Microscope with a 785 nm red laser operating WiRE® version 4.2. The data was obtained using a 10s exposure time with 5 -10% laser power.
X-ray Photoelectron Spectroscpopy (XPS) was undertaken using a K-ALPHA Thermo Scientific device, utilising Al-K radiation (1486.6 eV) and a twin crystal monochromator to produce a focussed x-ray spot at 3 mA x 12 kV (400 µm major axis length of the elliptical shape). Prior to the spectral acquisition, the samples were pre-reduced simulation the activation treatment. The data was then processed using the Avantage software package.
N2 adsorption isotherms and textural analysis was performed on a Micrometrics 3Flex

Catalytic Activity Studies
The RWGS experiments were conducted in a tubular quartz reactor (10 mm ID) at atmospheric pressure. The catalysts were supported on a layer of quartz wool acting as a bed. The reactant gas flow used for temperature screening and stability tests contained CO2 and H2 in a ratio of 1:4 balanced with N2 to maintain a WHSV = 15 L/gcath. The gas products were analysed using an online gas analyser (ABB AO2020, ABB Ltd., Zurich, Switzerland) equipped with both an IR and TCD detectors. All catalysts were reduced pre-reaction in the reactor by flowing 100 mL min −1 , 20% H2/ 80% N2 at 850 o C for 1 hour. Temperature screening reactions were conducted using a temperature range of 300-850 o C at 50 o C intervals. The stability study was conducted at 500 o C for 89 hours.

Catalyst Characterization
Saponite supported metal-adamantanecarboxylates were prepared by hydrothermally treating the respective metal hydroxides with twice the number of moles of 1adamantanecarboxylic acid in the presence of exfoliated saponite clay layers as described in the experimental section. Only the required stioichiometric amounts of metal hydroxides (metal ion source) and 1-adamantane carboxylic acid (carboxylate ion source) were used for the synthesis and no excess amounts of chemicals were used. Exfoliation of smectite clays in water is spontaneous and well known [40]. Exfoliated clay layers have been used as 2D starting materials for synthesis of various composites [41]. Using exfoliated clay layers as PXRD patterns of the as synthesized catalyst precursors and pristine saponite are shown in Fig. 1 (a-d). Saponite (Fig. 1a)         could be due to loss of water molecules. Pristine saponite sample (Fig. S2a) loses mass in two more steps at 557 °C and 749 °C due to loss of water of hydration, dehydroxylation of outer and inner hydroxyl ions as reported in literature [41]. Mass loss of Ni-Ada/Sap (Fig. S2b), NiCo-Ada/Sap (Fig. S2c) and NiCu-Ada/Sap (Fig. S2d) in between 200 -550 °C could be due to the degradation of the metal-adamantanecarboxylates in the catalyst precursors.   Fig. 4a, b and c respectively.
Reflections due to the Ni metal, NiCo and NiCu alloys were observed along with those due to saponite. The reflections of Ni-metal (Fig. 4a)   FTIR as shown in supporting information, Fig. S3. All samples show vibrations due to Si-O at 950 cm -1 due to saponite and C=C vibration (1600 cm -1 ) due to residual carbon. The samples also show a broad vibration at 3400 cm -1 due to adsorbed water.
The metal loadings on saponite were determined by XRF analysis. Table 1 shows the amount of Cu, Co and Ni present in various samples per gram of catalyst. The resultant catalysts were further characterized with Raman spectroscopy (Fig. S4, supporting information). The Raman spectra of the freshly prepared catalysts showed two intense bands which are attributed to vibration modes of sp 2 -bonded carbon atoms. The G-band observed at 1594 cm -1 is due to the sp 2 carbon stretching modes in aromatic rings derived from the incomplete decomposition of 1-adamanatanecarboxylate unit. The peak at approx. 1336 cm -1 is the graphitic D-band that becomes active in the presence of structural disorders [44]. The role played by the solid surface is essential in catalysis. Herein, x-ray photoelectron spectroscopy (XPS) allows us to determine the oxidation states and electronic environment of the elements in the outermost layers of the material. The Ni 2p3/2 spectra of all samples can be found in Fig. S5a, with the associated Cu 2p3/2 and Co 2p3/2 regions for the NiCu-

J o u r n a l P r e -p r o o f
Sap and NiCo-Sap catalysts in Fig. S5b and S5c, respectively. Table 2 contains a summary of the main peaks found in Fig. S5. Prior to this analysis, all samples were reduced under the same conditions used before a reaction (850 o C, 1 hour, 20% H2:80% N2). As seen in both Fig. S5a and Table 2, there are several Ni oxidation states that exist following reduction and are seen in the deconvoluted spectra. The bands ca. 851-852 eV are characteristic of Ni 0 , while the bands at 852-854 eV are attributed to Ni 2+ species interacting with the support [26,45,46]. The band centred at binding energy (BE) 857 eV, in the case of the NiCu-Sap catalyst, is attributed to Ni 2+ as part of surface NiCu alloy species [47,48]. This assignment is corroborated when considering the Cu 2p3/2 region that details a shift towards BEs associated with NiCu alloys [49]. The remaining bands are the shake-up satellite peaks associated with the previous species. However, the BE displayed in the NiCo-Sap catalyst at 855.48 eV is indicative of Ni 3+ cations present [51,52], characteristic of NiCo alloys [53].
The Cu 2p3/2 region in the NiCu saponite, Fig. S5b, shows two significant bands. One band at 932 eV can be assigned to the Cu + /Cu 0 species, with the higher bands at 934, 940 and 943 eV attributed to the Cu 2+ species and two shake up satellites, respectively [45,54].
The main bands for Cu at 932 and 934 eV can be explained to be at higher binding energies than monometallic Cu as found in literature, due to the charge transfer from Cu to the partially empty d-band present in Ni and the oxidation of Cu [55,56]. Such electronic interaction between the two metals has been theorised to contribute to increased catalytic activity since it results in an electronically rich metal-metal interface which is ideal for reactants activation [48,49]. Furthermore, the Ni-Cu interface has been identified to be the active site for the forward and reverse water gas shift reactions by enhancing CO/CO2 adsorption and supressing methane production [50]. Additionally, it has been found that high compositional contents of Cu in a Cu-Ni alloy sufficiently increased the reducibility and mesoporosity of the structure to subsequently increase the catalytic activity [45].
Finally, the Co 2p3/2 spectrum for the NiCo saponite found in Fig. S5c details two main bands at 778 and 780 eV, which are attributed as Co 3+ and Co 2+ , respectively [53]. While the other two bands are the associated shake-up satellites [57,58]. Another key factor found in the data for this region, is that the splitting (spin-orbit coupling) energy between the Co2p1/2 and Co 2p3/2 orbitals (not shown) is approximately 15 eV, further indicating the coexistence of Co 2+ /Co 3+ species [57]. The textural properties of the active catalysts were analysed using N2 adsorption. Fig. 5 shows the adsorption isotherms of all the samples and surface properties are tabulated in  Ultimately, however, beyond the onset relative pressure of the loop, the isotherms confirm the presence of narrow or slit shaped mesopores within the material that are confirmed by the BJH analysis of the material (Table 3) confirming average pore diameters between 2-50 nm. J o u r n a l P r e -p r o o f  Saponite support displays coherent distribution of Mg, Al and Si accounting for homogeneously formed saponite clay layers. In situ generated carbon in the freshly prepared catalysts was found to be distributed homogeneously over the saponite clay layers. The presence of carbon in the catalysts was also indicated in the Raman analysis as discussed earlier. where typically the methanation reaction is the dominant process. This is seen in the selectivity plots, Fig. 8b and c, where there is little to no methane produced, while CO is being produced in abundance. Overall the Ni-Cu catalyst is the best material within the studied series and hence we have compared its performance with reference systems. As shown in Table 4 the NiCuSap shows either markedly improved or comparable performance as a number of transition metal and noble metal catalysts reported recently [22,26,31,59,[62][63][64] using the same temperature window and reaction mixture (CO2:H2 1:4).

Catalytic Testing for RWGS Reaction
Only the bimetallic Fe-Ni catalysts reported in [26] outperforms our CO2 conversion levels but the selectivy of the material is much lower than that exbited by our Ni-Cu catalysts.
Hence the Ni-Cu/saponite catalyst represents an excellent balance activity/CO selectivy when compared with benchmark materials. The homogenous distribution and high dispersion of the Ni-Cu active centres shown in the STEM study, along with the Ni-Cu electronic interaction discussed in the XPS section, can explain the excellent performance of this sample. Cu suppresses the methanation activity of Ni and the Ni-Cu ensemble is an advanced active phase for the RWGS reaction, leading to high levels of CO2 conversion in the whole temperature range and remarkable selectivity levels towards CO. In fact the presence of Cu opens up the possibility to conduct the RWGS reaction via redox and/or formate mechanism as previously reported elsewhere thus favouring the CO route over the CH4 pathway [65] In any case, the superior behaviour of our catalysts compared to reference materials is indeed a very encouraging result and showcases the viability of low cost nature inspired multicomponent catalysts for the reverse water gas shift process. Due to this behaviour, the J o u r n a l P r e -p r o o f NiCuSap catalyst was selected for a stability study at 500 o C as this temperature indicated significant conversion at lower temperature for the RWGS reaction, while not reaching equilibrium.  The suppression of methanation by copper containing materials has been previously reported by our team in the forward WGS reaction [60,61]. However, the potential of this alloy for the reverse water gas shift process in still under explored. The enhanced selectivity to CO at low temperatures is an encouraging result to achieve the successful coupling of RWGS with downstream processes such as Methanol synthesis and Fischer-Tropsch which typically take place at around 250-350 o C. This way we could close the cycle: CO2 conversion to fuels and chemicals in a two step-process with the RWGS as front unit and the F-T or methanol conversion reactors as second unit to produce the upgraded end products.

Characterization of Spent Catalysts
The x-ray diffraction patterns of the spent catalysts from the temperature screening experiments are shown in Fig. 10. All the diffractograms display peaks at around 28.2°, 31.05° and 35.5° 2θ due to saponite support. Additionally, the expected peaks for the loaded a b c  Combusting the spent material under air from room temperature to 900°C at 10 °C/min revealed several zones (Fig. 11). Each sample underwent an initial loss between room temperature (RT) -160 o C that is attributed to free water loss. Each sample then displayed a significant weight gain (+5-14 wt %) in between 200-450 °C for NiCuSap and J o u r n a l P r e -p r o o f NiCoSap that is attributed to the oxidation of the metals. The same weight gain zone for the NiSap occurred at the slightly higher zone of ca. 275-500 °C.
The spent NiCoSap catalyst displayed a two-step weight loss totalling 3.8 wt% which is attributed to the loss of surface carbon and then engrained carbon. In a similar fashion, the spent NiCuSap catalyst details a one-step weight loss (1.5 wt %), which is also attributed to the loss of surface carbon. These conclusions are supported by the presence of an exothermic heat flow curves for the oxidation of the metals and the associated oxidation of the amorphous carbon (not shown). The weight gain of these materials being related to metallic oxidation is further supported by the curve displayed for the NiSap material, which details a much smaller increase owing to its monometallic loading. Interestingly, however, the NiSap material did not display any weight loss at higher temperatures,. In any case the TGA profiles corroborate the absence of crystalline carbon deposits in good agreement with XRD. This observation along with the lack of metallic sintering validate that these "nature inspired" catalysts developed in this work are not only highly active, but also very robust for the RWGS reaction.

Conclusions
This work demostrates the viability of nature inspired transition metal based catalyst for gas phase CO2 upgrading via RWGS. Aqueously exfoliated saponite magnesio- Interestingly, the undesired parallel reactionthe Sabatier processwhich typically is the dominant reaction in the low temperature window can be supressed using Ni-Cu alloys as active phases. Indeed, the bimetallic Ni-Cu system is the best performing material within the studied series with an outstanding balance activity/CO selectivity in addition to be a very stable catalysts for long term runs. The electronic interaction Ni-Cu evidenced by XPS contributes to this exceptional behaviour. Indeed, such a close metal-metal contact results and electronically rich Ni-Cu interface which is ideal for CO2 activation. No signs of carbon deposition due to the reaction, nor metallic sintering were observed, explaining the enhanced stability of this material.
Considering a potential application where the RWGS unit is coupled to a downstream process using syngas such as Fischer-Tropsch or methanol synthesisthe obtained results are very encouraging since our Ni-Cu catalyst is very active and selective towards CO in the low temperature range minimising the temperature gap between RWGS and the Fischer-Tropsch or methanol unit. In other words, the catalysts developed in this study may facilitate the integration of a RWGS reactor with a syngas convertorsuch an integrated dual system would enable the direct conversion of CO2 to added value chemicals.

declaration of interests
The authors declare that they have no competing interests.
through grants EP/N024540/1 and EP/N009924/1, as well as the Buchan Chair in Sustainable Energy Engineering.
J o u r n a l P r e -p r o o f