Identi ﬁ cation of C 2 – C 5 products from CO 2 hydrogenation over PdZn/TiO 2 – ZSM-5 hybrid catalysts †

The combination of a methanol synthesis catalyst and a solid acid catalyst opens the possibility to obtain ole ﬁ ns or para ﬃ ns directly from CO 2 and H 2 in one step. In this work several PdZn/TiO 2 – ZSM-5 hybrid catalysts were employed under CO 2 hydrogenation conditions (240 – 360 (cid:1) C, 20 bar, CO 2 /N 2 /H 2 ¼ 1 : 1 : 3) for the synthesis of CH 3 OH, consecutive dehydration to dimethyl ether and further oxygenate conversion to hydrocarbons. No signi ﬁ cant changes after 36 h reaction on the methanol synthesis catalyst (PdZn/TiO 2 ) were observed by XRD, XAS or XPS. No ole ﬁ ns were observed, indicating that light ole ﬁ ns undergo further hydrogenation under the reaction conditions, yielding the corresponding alkanes. Increasing the aluminium sites in the zeolites (Si : Al ratio 80 : 1, 50 : 1 and 23 : 1) led to a higher concentration of mild Brønsted acid sites, promoting hydrocarbon chain growth.


Introduction
The sustainable production of energy is one of the major challenges of modern society. 1,2 Several technologies have been developed to harvest renewable energy (e.g., solar panels, wind farms) in the form of electricity. 2 However, due to the intermittent nature of renewables, surplus produced electricity must be stored as chemical bonds to ensure a steady production of energy when electricity a bar the selectivity towards CO, oxygenates and hydrocarbons observed was 93.4, 1.8 and 4.8% respectively. C 5+ products were detected, although C 1 and C 2 products accounted for 95.4% of the hydrocarbon product distribution. Over a ZnZrO 2 /SAPO catalyst at 380 C and 20 bar ($18% CO 2 conversion), the CO selectivity was reduced to 47%, whilst C 2 -C 4 olens accounted for 80% of the total hydrocarbon distribution, and the remaining hydrocarbons were assigned to C 2 -C 4 alkanes (14%), C 1 (3%) and C 5+ (3%). 40 Light olens synthesised on hybrid catalysts from CO 2 can also undergo further reduction to their corresponding alkanes on the methanol synthesis catalyst, which effectively acts as a hydrogenation catalyst. Park et al. 41 reported ethane (76.4%) as the main hydrocarbon product over a CuZnOZrO 2 -ZSM-5 catalyst (28 bar, 400 C), with little formation of C 3 (4.5%), C 4 (0.8%) and C 5+ (0.2%) products. However, as reported by Giordano, Frusteri and co-workers 42,43 for CO 2 conversion to DME over CuZnZr/ferrierite hybrid catalysts, Cu remains prone to severe sintering under the reaction conditions (260 C, 30 bar). This makes catalyst stability the bottle neck of this process.
To the best of our knowledge, reports on PdZn hybrid catalysts for CO 2 hydrogenation focus on DME, 27,38 but no detailed attention has been paid to the produced hydrocarbons. We therefore assessed the activity and stability of diverse PdZn/TiO 2 -ZSM-5 hybrid catalysts under CO 2 hydrogenation conditions (20 bar, <360 C) and identied produced hydrocarbons derived from MTH and DMTH.

PdZn/TiO 2 -ZSM-5 hybrid catalyst preparation
The as received (NH 4 )-ZSM-5 zeolites were annealed in static air (550 C, 10 C min À1 , 6 h) to obtain the H-ZSM-5 form prior to reaction. 0.5 g of PdZn/TiO 2 and 0.5 g of treated ZSM-5 were physically mixed in a vial until homogeneous. The mixture was pelleted (10 ton) and crushed (425-600 mm) to obtain the hybrid catalyst. Zeolites are named in the text according to their Si/Al ratio, for instance ZSM-5 with a Si/Al ratio of 23 : 1 will be referred to as ZSM-5(23).

CO 2 hydrogenation and consecutive MTH/DMTH catalyst testing
Catalytic activity for CO 2 hydrogenation to CH 3 OH, DME, olens and hydrocarbons was achieved in a stainless steel xed-bed (50 cm length, 0.5 cm internal diameter) continuous ow reactor. 0.5 g of hybrid catalyst without diluent (or 0.25 g of pelleted PdZn/TiO 2 with 0.25 g of SiC as diluent) were secured in the reactor tube using quartz wool. Prior to reaction, the hybrid catalysts were prereduced in 5% H 2 /He (400 C, 5 C min À1 , 1 h). Subsequently, the reactor was cooled down to 50 C, the 5% H 2 /He ow was switched to the reaction mixture (CO 2 /N 2 /H 2 ¼ 1 : 1 : 3, 30 ml min À1 ), and the reactor was pressurised to 20 bar and heated to the desired reaction temperature (240, 270, 300, 320, 340 and 360 C, 5 C min À1 , 6 h dwell). To avoid product condensation, post reactor lines and valves were heated to 130 C. Products were analysed via online gas chromatography (Agilent 7890, tted with FID and TCD detectors). Details of how CO 2 conversion, product selectivity and productivities were calculated can be found in the ESI (eqn (S1-S11 †)).
Catalyst characterisation X-ray absorption spectroscopy (XAS) was carried out in transmission mode at the Pd K-edge, at the B18 beamline of the Diamond Light Source, Harwell, UK, and a Pd foil was examined simultaneously with the sample and used as a reference. Three spectra were averaged to minimise the noise signal. The X-ray absorption ne structure (EXAFS) was analysed with the Demeter soware package (Athena and Artemis). 44 X-ray photoelectron spectroscopy (XPS) was carried out on a Kratos Axis Ultra-DLD tted with a monochromatic Al Ka (75-150 W) source and analyser, using a pass energy of 40 eV. The XPS data were analysed using Casa XPS soware. Powder X-ray diffraction (XRD) patterns were obtained on a (q-q) PANalytical X'pert Pro powder diffractometer tted with a hemispherical analyser using a Cu Ka radiation source (40 keV, 40 mA). The pore sizes and BET surface areas of the ZSM-5 zeolites were obtained through N 2 adsorption isotherms using a 3-ex Micromeritics instrument. Samples were degassed in situ at 250 C for 10 h prior to analysis. Coke deposition during CO 2 hydrogenation was measured through thermogravimetric analysis on a PerkinElmer TL9000 with a TG-IR-MS interface.

Results and discussion
PdZn/TiO 2 -ZSM-5 hybrid catalysts for direct CO 2 conversion to hydrocarbons In addition to PdZn/TiO 2 employed as a CH 3 OH synthesis catalyst, 20-23 commercial ZSM-5 zeolites with various Si/Al ratios (23, 50 and 80) were used as solid acid catalysts to promote consecutive CH 3 OH dehydration to DME 27-30 and further MTH/DMTH. 32-34 XRD patterns, pore sizes and BET surface areas for the commercial ZSM-5 zeolites aer annealing (static air, 550 C, 10 C min À1 , 6 h) can be found in the ESI ( Fig. S1 and Table S1, † respectively). Thorough characterisation of PdZn/TiO 2 synthesised by chemical vapour impregnation (CVI) with 5 wt% Pd and a Pd/Zn molar ratio of 1 : 5 was previously reported. [25][26][27] Firstly, the catalytic activity of PdZn/TiO 2 (0.25 g catalyst diluted with 0.25 g of SiC) for CO 2 hydrogenation was assessed (20 bar, CO 2 /N 2 /H 2 ¼ 1 : 1 : 3, 240-360 C). As observed in Table S2, † raising the reaction temperature from 240 to 360 C resulted in an increase in the CO 2 conversion from 6.8 to 31.2%, accompanied by an increase in the CO selectivity (from 74.6 to 97.9%) at the expense of CH 3 OH selectivity (from 24.3 to 1.3%), in accordance with the reaction thermodynamics. 45 Below 300 C, CH 3 OH synthesis proceeded in the kinetic regime ( Fig. 1a), as shown by the increase in the CH 3 OH productivity with temperature up to 518 mmol kg cat À1 h À1 . At 300 C, the CH 3 OH yield reached equilibrium (1.45% CH 3 OH yield), 45 and hence, above 300 C CH 3 OH synthesis is controlled by the thermodynamic equilibrium ( Fig. 1b), as shown by the sharp decrease in CH 3 OH productivity. Low selectivity toward CH 4 (<0.8%) and DME (<1.0%), produced by CH 3 OH decomposition on TiO 2 (ref. 46 and 47) and CH 3 OH dehydration, 27 respectively, was observed. Nevertheless, in the absence of ZSM-5 solid acid catalysts, no other hydrocarbons were detected. Comparable CO 2 conversion and CO productivity to PdZn/TiO 2 were observed for all PdZn/TiO 2 -ZSM-5 hybrid catalysts (Fig. 2), indicating that the activity of the methanol synthesis catalyst is not altered by the proximity of solid acid zeolites, and that ZSM-5 zeolites are not active towards the RWGS. To ensure that ZSM-5 zeolites do not act as RWGS or as CH 3 OH synthesis catalysts, blank ZSM-5 (23) was employed for the CO 2 hydrogenation reaction (Table S3 †), and negligible CO 2 conversion was observed at 270 C. The efficient dehydration of CH 3 OH to DME occurred over all PdZn/TiO 2 -ZSM-5 hybrid catalysts. The highest oxygenate productivity (CH 3 OH + DME) was obtained at 270 C. At this temperature almost no methanol to olens (MTH) or dimethyl ether to olens (DMTH) conversion takes place, with only small traces of ethane detected over hybrid catalysts with ZSM-5(50 and 80), and small amounts of higher hydrocarbons observed for PdZn/ TiO 2 -ZSM-5 (23). DME is the major oxygenate product with a selectivity close to 20%. Moreover, the total methanol productivity (CH 3 OH (tot) ), considering that all hydrocarbons originate from CH 3 OH by either dehydration to DME or through the MTH/DMTH process, was higher for all hybrid catalysts compared to PdZn/ TiO 2 (Fig. 3).
The hydrocarbon productivities and CH 3 OH (tot) over PdZn/TiO 2 -ZSM-5 hybrid catalysts can be found in Table S4. † At 300 C, the hybrid catalysts showed higher  CH 3 OH (tot) compared to PdZn/TiO 2 , thus overcoming the theoretical CH 3 OH yield dictated by the equilibrium. Increasing the alumina ratio in the zeolite promoted the formation of longer hydrocarbons. The presence of aluminium sites is related to Brønsted acid sites with mild acidity, and hence, a higher concentration of Brønsted acid sites promotes chain growth via the hydrocarbon pool mechanism. 33,48 Light olens produced as intermediates in MTH 49 undergo further hydrogenation over PdZn/TiO 2 , yielding the corresponding alkanes (ethane and propane); a mixture of n-butane and 2-butene was observed, whilst only olens were detected in the C 5 fraction (1-pentene and 2-cis/trans-pentene). Thus, when PdZn alloys are used for the synthesis of hydrocarbons from CO 2 via a methanol mediated route over hybrid catalysts, they behave as methanol synthesis catalysts but also as olen hydrogenation catalysts, limiting hydrocarbon chain growth as also reported for Cu-based catalysts. 49-51 CH 3 OH (tot) over hybrid catalysts is higher compared to PdZn/TiO 2 at any temperature, with total CH 3 OH productivity surpassing the equilibrium yield above 300 C. The highest total CH 3 OH productivity for hybrid catalysts was observed in the 270-300 C range; hybrid catalysts with ZSM-5(80 and 50) gave the highest oxygenate productivity (CH 3 OH and DME), while ZSM-5(23) led to higher hydrocarbon productivity (Fig. 4) via faster MTH/DMTH. Although the total CH 3 OH productivity was higher than the equilibrium yield, above 300 C, the hydrocarbon productivity is limited by CH 3 OH availability which leads to a decrease in the total CH 3 OH productivity at higher temperatures. The hydrocarbon selectivity based on MTH/DMTH showed that increasing the aluminium concentration in ZSM-5 in the hybrid catalyst resulted in improved selectivity towards higher hydrocarbons ( selectivity via the MTH and DMTH process is lower than the reported values, since CH 4 is also produced as a by-product in CH 3 OH decomposition on PdZn/TiO 2 . 47

Catalytic stability of PdZn/TiO 2 under the reaction conditions
The higher stability of PdZn alloy catalysts compared to their Cu-based counterparts was proven under methanol reforming conditions, 15,52-54 which is the opposite reaction to the intended CO 2 hydrogenation to CH 3 OH. Copper sinters in the presence of water at elevated temperatures, and hence, due to the higher water content when the feed is CO 2 instead of CO, PdZn alloys are employed as stable alternatives to Cu-based catalysts for the synthesis of CH 3 OH from CO 2 . [15][16][17][18][19] According to the Pd-Zn phase diagram developed by Massalski 55 and Vizdal et al., 56 the b-PdZn alloy is thermally stable up to 1200 C. However, in the presence of oxygen at 300 C, the surface of the PdZn alloy segregates into ZnO and metallic Pd, 53 whilst under H 2 , the b-PdZn alloy was experimentally proven to be stable up to 600 C. 57 Chen et al., 58 based on DFT calculations, suggested that Zn segregates from the PdZn alloy when the alloy is supported on ZnO. Nevertheless, Ahoba-Sam et al. 38 reported no changes in the PdZn phase and no formation of extra Pd-phases through operando XAS during CO 2 hydrogenation to CH 3 OH (8 bar, 350 C). The stability of the PdZn phase in the PdZn/TiO 2 methanol synthesis catalyst was assessed through XAS, XRD and XPS characterisation pre-and post-reaction (20 bar, CO 2 /H 2 /N 2 ¼ 1 : 3 : 1, 30 ml min À1 , 240-360 C, 6 h dwell, 36 h reaction). Extended X-ray absorption ne-structure spectroscopy (EXAFS) at the Pd K-edge was employed to detect structural and electronic changes in the PdZn phase aer reaction. No noticeable differences were observed between the EXAFS spectra up to 3.5Å of PdZn/TiO 2 aer pre-reduction (5% H 2 , 400 C, 1 h) and aer CO 2 hydrogenation at up to 360 C (Fig. 5a). Bulk Pd has a cubic structure with a Pd-Pd bond distance of 1.75Å, whilst the b-PdZn alloy has a tetragonal structure. As expected for the intercalation of a Zn atom between two Pd atoms, on average shorter bond distances for the rst coordination shell compared to bulk Pd were observed (shorter Pd-Zn bond distance compared to the Pd-Pd distance in Pd foil). 59 Moreover, comparison with a PdO standard could suggest the presence of a Pd-O bond at 2.02Å, with the concomitant but lower intensity Pd-O-Pd bond at 3.41Å in the second coordination shell. 60 This could be attributed to oxidation of the rst atomic layer of the PdZn alloy in contact with air. 53 Peak tting using Artemis 44 allowed us to obtain the Pd-Pd and Pd-Zn bond distances and the Pd coordination environment ( Table 1). Details of the tting can be obtained from Table S6. † Based on previous reports, PdZn alloy formation begins at the surface of Pd nanoparticles via hydrogen spillover to adjacent ZnO, and the alloy grows inwards, generating a PdZn layer over a Pd core. 53,57 Therefore, the incorporation of bulk Pd-Pd bond distances was necessary to obtain a good t. No differences in the Pd-Zn or Pd-Pd bond distances or coordination number were observed, suggesting high bulk structural stability of the PdZn alloy under reaction conditions, as also reported by Olsbye and co-workers. 38 Despite the apparent bond distance at 2.02Å, no good t was obtained aer the introduction of the Pd-O scattering path, indicating that this contribution can be attributed to noise or to marginal PdO content.
Phase changes in the b-PdZn alloy during reaction were investigated by recording the XRD pattern of PdZn/TiO 2 aer reduction (400 C, 1 h) and aer reaction (240-360 C, 20 bar, 36 h). In agreement with the EXAFS analysis, no changes were observed in the (111) and (200) b-PdZn reections at 41.2 and 44.1 respectively, 15,19,27 showing the high thermal stability of bulk PdZn under the reaction conditions (Fig. 5b). Moreover, no signicant changes were observed in the XRD peak at 40.1 assigned to metallic Pd (PDF 00-046-1043), which presumably is protected underneath a PdZn layer. 53,57 Unincorporated Zn in the PdZn alloy is observed as ZnO at 31.7 , 34.4 and 36.3 (PDF 00-036-1451). In line with PdZn and Pd, no changes in the ZnO reections were observed aer reaction. Nevertheless, the TiO 2 -related reections become more intense aer reaction, suggesting an increase in the particle size of the support (Fig. S2 †). EXAFS and XRD are averaging techniques sensitive to bulk changes. Catalysis, however, is a surface process, and hence small changes at the surface greatly affect the catalytic activity. To follow surface changes in the PdZn alloy composition, the Pd(3d) and Zn(LM 2 ) orbitals of the PdZn/TiO 2 catalyst before and aer reaction were analysed by X-ray photoelectron spectroscopy (XPS). The Zn(2p) orbital is not sensitive towards chemical changes (e.g. the binding energies for Zn 0 and Zn 2+ are reported at 1021.7 and 1022 eV, respectively), 61 and as observed by XRD, the unalloyed zinc remained as ZnO, hence the Pd(3d) and Zn(LM 2 ) orbitals were calibrated against the Zn(2p) orbital at 1022 eV. The Pd(3d) peak for PdZn/TiO 2 aer reduction and aer reaction was centred at 335.6 eV (Fig. 6), between the binding energy values reported for metallic Pd (334.8-335.4 eV) 62-64 and the PdZn alloy (335.6-336.7 eV). [64][65][66] Peak tting using nite Lorentzian line shapes for the Pd and PdZn peaks (including satellites) and Gaussian line shapes for the PdO peaks and satellites with a Shirley background as described previously 47 indicated the presence of Pd, PdZn and PdO at 335.0, 335.9 and 337.2 eV, respectively (Fig. S3a †). 54,64 The presence of Pd and PdZn was conrmed by the XRD and EXAFS bulk characterisation techniques, however bulk PdO was not observed. The broadening of the Pd(3d) peak aer reaction indicated an increase in the proportion of surface PdO, suggesting that surface PdZn phase separation into ZnO and Pd occurred, with concomitant palladium passivation when in contact with air. 53 No signicant changes were observed in the Zn(LM 2 ) Auger electron spectra before and aer reaction (Fig. S3b †); the main peak at 998 eV with a minor satellite contribution at 991 eV was attributed to the presence of ZnO. 67  No signicant changes in the PdZn phase were detected by EXAFS or XRD, con-rming the bulk thermal stability of PdZn under reaction conditions (20 bar, 240-340 C, CO 2 /H 2 /N 2 ¼ 1 : 3 : 1, 36 h). However some surface PdZn phase separation into Pd and ZnO was suggested by the XPS characterisation. Coke deposition is oen reported as the main deactivation mechanism for MTH over acid zeolite catalysts. Coke inhibits CH 3 OH diffusion to the acid active sites by either lling zeolite cavities or blocking pores. 34 Zeolites can be regenerated at high temperature (500-600 C) by oxidising deposited coke to CO 2 with oxygen. 68 To assess the extent of coke deposition during CO 2 hydrogenation (20 bar, 240-340 C, CO 2 /H 2 /N 2 ¼ 1 : 3 : 1, 36 h) over PdZn/ TiO 2 -ZSM-5 hybrid catalysts, TGA-MS was performed on fresh and spent catalysts (Fig. S4 †). The mass loss observed below 200 C can be assigned to physi/chemisorbed water. Just above 200 C, an extra 1 wt% mass loss compared to the fresh samples was observed for the PdZn/TiO 2 -ZSM-5(80 and 50) hybrid catalysts with a corresponding release of CO 2 , which could be assigned to coke deposits with high oxygen content as reported for CH 3 OH conversion to olens over ZSM-5. 69 No additional mass loss at higher temperature, which would be assigned to coke deposits with low oxygen and hydrogen content (e.g. aromatics), was detected. 69 No coke deposition was detected for PdZn/TiO 2 -ZSM-5(23). This should not be interpreted as ZSM-5(23) being less sensitive to coke deposition, since a higher concentration of acid sites generally leads to faster deactivation, 70 but instead the initial rate of coke formation might be slower in this system. This might be attributed to the presence of H 2 O and H 2 in the feed [35][36][37] as well as the low concentration of CH 3 OH throughout the catalyst bed. 69

Conclusions
The combination of a PdZn/TiO 2 methanol synthesis catalyst with solid acid ZSM-5 zeolites in the form of a hybrid catalyst allowed for consecutive CO 2 hydrogenation to CH 3 OH, CH 3 OH dehydration to DME, and MTH/DMTH in a one-pass single bed reactor. Thus, the total CH 3 OH productivity from CO 2 hydrogenation over PdZn/TiO 2 -ZSM-5 hybrid catalysts was higher compared to PdZn/TiO 2 . The synthesised light olens undergo further hydrogenation to the corresponding alkanes (ethane, propane and butane) as also reported for Cu-based catalysts, which limit hydrocarbon chain growth. Hence, future research using PdZn alloys for the synthesis of hydrocarbons via the methanol route should focus on limiting the activity towards olen hydrogenation whilst maintaining good selectivity for CH 3 OH. Increasing the concentration of aluminium sites in ZSM-5, correlated with mild Brønsted acid sites, resulted in the production of higher hydrocarbons. The bulk stability of the PdZn/TiO 2 catalyst up to 360 C under the reaction conditions (20 bar, CO 2 /H 2 /N 2 ¼ 1 : 3 : 1, 30 ml min À1 , 36 h) was conrmed by XAS and XRD. However XPS suggests that some surface PdZn separation into Pd and ZnO occurred during the reaction.

Conflicts of interest
There are no conicts to declare.