CO2 Hydrogenation to CH3OH over PdZn Catalysts, with Reduced CH4 Production

Metallic Pd, under CO2 hydrogenation conditions (>175 °C, 20 bar in this work), promotes CO formation via the reverse water gas shift (RWGS) reaction. Pd‐based catalysts can show high selectivity to methanol when alloyed with Zn, and PdZn alloy catalysts are commonly reported as a stable alternative to Cu‐based catalysts for the CO2 hydrogenation to methanol. The production of CH4 is sometimes reported as a minor by‐product, but nevertheless this can be a major detriment for an industrial process, because methane builds up in the recycle loop, and hence would have to be purged periodically. Thus, it is extremely important to reduce methane production for future green methanol synthesis processes. In this work we have investigated TiO2 as a support for such catalysts, with Pd, or PdZn deposited by chemical vapour impregnation (CVI). Although titania‐supported PdZn materials show excellent performance, with high selectivity to CH3OH+CO, they suffer from methane formation (>0.01 %). However, when ZnTiO3 is used instead as a support medium for the PdZn alloy, methane production is greatly suppressed. The site for methane production appears to be the TiO2, which reduces methanol to methane at anion vacancy sites.


Introduction
Currently, over 85 % of global energy is obtained from finite resources (coal, oil or natural gas), giving increased atmospheric CO 2 levels [1] and inevitable climate change consequences. [2] Economic growth has in the past been associated with the availability of finite resources for the production of energy. [3] The specific location of natural resources and their fluctuating price has resulted in intergovernmental frictions, [4] which are expected to intensify as the cheapest natural deposits deplete over time. Hence, the production of energy from renewables is one of the biggest challenges to secure a steady energy supply and to tackle global CO 2 emissions. Current technology (e. g., wind and solar farms, hydroelectric power stations) allows the production of electricity free of carbon emissions. The main drawback for renewable electricity is its intermittent nature, which therefore requires some method of storage of this energy when production is high, to be used when production is low. Currently the most efficient and industrially scalable route to store renewably-produced electricity is through water splitting to produce H 2 , so-called green hydrogen. [5] Hydrogen can be used as an energy vector, however, it has a low energy density per volume, and most current technology has evolved around much more energy-dense molecules. Hence, to readily incorporate green hydrogen into conventional technology, it may be necessary to convert it to a liquid fuel. One possible route is to transform it into methanol by its reaction with captured CO 2 , which in turn will alleviate carbon dioxide emissions.
The selective CO 2 hydrogenation to CH 3 OH (Equation 1) is challenging because of the possibility of simultaneous reactions occurring, such as the reverse water gas shift (RWGS) (Equation 2) and methanation (Equation 3), which produce CO and CH 4 respectively. Thermodynamically, the production of CH 3 OH and CH 4 are favoured at low temperature and high pressure, while high temperature promotes the RWGS.
In a methanol plant, liquid products (CH 3 OH and H 2 O) are separated from gaseous products through some form of condenser, unreacted reagents (CO 2 and H 2 ) and gaseous products (CO and CH 4 ) are recycled into the catalyst bed. [6] The formation of CO is not detrimental for the overall process, since carbon monoxide can also be transformed to CO 2 or methanol, moreover, the presence of CO in the gas feed can result in enhanced methanol yield. [7] However, CH 4 accumulates during gas recycling cycles, and eventually needs to be purged, increasing production costs.
Copper based catalysts are commonly employed for the CO 2 hydrogenation to CH 3 OH, however strong Cu-sintering [8][9][10][11] and coke deposition [12] are observed when CO 2 is used as the feed. Noble-metal based catalysts can be used to overcome the catalyst deactivation observed with Cu-based catalysts, with the PdZn alloy system being one of the materials receiving research attention. Iwasa et al. [13] showed that on Pd-based catalysts at ambient pressure, the support controls the product selectivity. Over palladium black, CO 2 reacts with H 2 at atmospheric pressure to form primarily CO and CH 4 , with no CH 3 OH formation. Supports with little interaction with palladium (SiO 2 or MgO) formed no methanol, while over supports that can form alloys or intermetallics with palladium upon reduction (e. g. ZnO and Ga 2 O 3 ) methanol selectivity significantly increased. For PdZn, the change in product selectivity can be associated to an electron density distribution from the electron rich Pd(4d) to external Pd(5 s), Pd(5p) and Zn(4p), Zn(4 s) orbitals. [14] Díez-Ramírez et al. [15,16] and Bahruji et al. [17] in separate studies over Pd/ZnO catalysts confirmed that the PdZn alloy phase, formed upon pre-reduction in hydrogen at high temperature (> 300°C), acts as the active phase for methanol synthesis. On Pd/ZnO catalysts, the PdZn alloy phase is also active for the formation of CO, [18] whilst metallic Pd sites are responsible for CO [16] and CH 4 formation. [19,20] CH 4 is commonly reported as a minor side product on PdZn alloy catalysts. [13,15,21,22] However, not much attention is paid to CH 4 because of its low selectivity (less than 1 %), which can lead to the misinterpretation that it is not important and that CH 4 formation is inherent to PdZn catalysts. This could limit applications of PdZn based catalysts in a CH 3 OH synthesis plant operating with captured CO 2 and green hydrogen. For comparison purposes, Table S1 shows CH 4 productivity and selectivity reported for Pd-based catalysts employed as CH 3 OH synthesis catalysts.
During the CO 2 hydrogenation on a PdZn/TiO 2 catalyst, methanol is formed over PdZn surfaces, [15,16,23] CO can be produced over metallic Pd [13,15,24] or PdZn, [25] however, little is known about the active sites for CH 4 formation on PdZn alloys. In this work, it is proposed that over PdZn/TiO 2 catalysts CH 4 is not formed through the methanation of CO 2 , but instead as a by-product of CH 3 OH decomposition at TiO 2 sites.

Catalyst characterisation
Powder X-ray diffraction (XRD) patterns were recorded on a (θ-θ) PANalyticalX'pert Pro powder diffractometer fitted with a position sensitive detector using Cu Kα radiation source (40 keV, 40 mA). In situ XRD was recorded on a (θ-θ) PANalyticalX'pert Pro powder diffractometer fitted with a position sensitive detector using a Cu Kα radiation source (40 keV, 40 mA) and an Anton Parr XRK reaction cell connected to a 5 % H 2 /Ar mixture, gas flow was controlled through the use of a Bronkhorst mass flow controller.
X-ray photoelectron spectroscopy (XPS) was carried out on a Kratos Axis Ultra-DLD fitted with a monochromatic Al Kα (75-150 W) source and an analyser using a pass energy of 40 eV. XPS data were analysed using Casa XPS software.
Transmission electron microscopy images were obtained on a JEOL 2100 (LaB6) instrument fitted with a Gatan digital camera (2k 2k) and a dark held HAADF/Z-contrast detector. Specimens were dryprepared on copper TEM-grids prior to analysis, to obtain representative particle size distributions at least 200 particles were analysed.
BET surface areas were measured using a Quantachrome Nova 2200e instrument. Prior to BET analysis samples were degassed in situ (120°C, 4 h).

CO 2 hydrogenation catalyst testing
The catalyst activity for CO 2 hydrogenation was measured in a stainless steel fixed-bed (50 cm length, 0.5 cm internal diameter) continuous flow reactor. 0.5 g of pelleted then crushed (425 -600 μm) catalyst was placed in the reactor tube without diluent, quartz wool was used to secure the catalyst bed in place. Reaction temperature was controlled through a chromel-alumel thermocouple placed in the catalyst bed. Prior to reaction, catalysts were prereduced in pure hydrogen (400°C, 5°C min À 1 , 1 h, 30 ml min À 1 ). Subsequently, the reactor was cooled to 50°C, the gas flow was switched from hydrogen to the reaction mixture (20 % CO 2 , 20 % N 2 , 60 % H 2 , 30 ml min À 1 ), pressurised to 20 bar and heated to the desired reaction temperature (175-250°C, 5°C min À 1 ). Post reactor lines and valves were heated at 130°C to avoid product condensation. Products were analysed via online gas chromatography (Agilent 7890, fitted with a FID and TCD detectors). Details of how to determine reaction metrics (CO 2 conversion, product selectivity and productivities) can be found in the supporting information (equation S1-S9).

Catalytic activity for the thermal CO 2 hydrogenation
Previously we have shown that the synthesis of PdZn alloys prepared by chemical vapour impregnation (CVI) using TiO 2 as support gave improved PdZn dispersion compared to the use of ZnO and Al 2 O 3 , and resulted in improved methanol production rates. [29] However, during the CO 2 hydrogenation (250°C, 20 bar) over 5 wt. % PdZn(1 : 5)/TiO 2 , CH 4 formation was observed at 0.1 % selectivity. [30] Although this may seem like a small amount of this by-product, it nevertheless would result in increased production costs due to the need to purge it after build-up in the recycle system. Hence it is important to minimise CH 4 formation, and to try to identify the exact source for this product. The concentration of active sites for CH 4 formation decreased after increasing the pre-reduction temperature from 400°C to 650°C, which was associated with Zn incorporation into the TiO 2 lattice forming ZnTiO 3 . To determine the active sites responsible for CH 4 formation Pd/TiO 2 , Pd/ZnO, Pd/ZnTiO 3 , PdZn/TiO 2 and PdZn/ZnTiO 3 catalysts were prepared by CVI as described above.
On catalysts containing Pd, CO 2 in the presence of H 2 is converted into CO via the RWGS, [13,24] but as we show below, CH 4 can be a by-product of reaction. On Pd-based catalysts employed for the methanation of CO 2 , debate remains about whether CO is reduced at the metal-support interface [19] or on Pd nanoparticles. [31] As shown in Table 1, the lowest methanol productivity is found for Pd/TiO 2 , whilst it showed the highest CH 4 productivity, in agreement with previous reports on Pdbased catalysts. [20,31,32] Prior to reaction, catalysts were prereduced (400°C, H 2 , 1 h) to form the PdZn active phase for methanol synthesis. [17,30,33,34] As observed for Pd/ZnO, Pd/ZnTiO 3 , PdZn/TiO 2 and PdZn/ZnTiO 3 the formation of βÀ PdZn resulted in enhanced methanol selectivity compared to Pd/TiO 2 throughout the temperature range studied (Table S2). CO 2 is a relatively stable molecule, and high reaction temperature is required for its activation. Increasing reaction temperature resulted in higher CO 2 conversion, however, methanol synthesis is favoured at low temperature and high pressure, [7] hence the decrease in methanol selectivity, seen for instance in figure 1 for Pd/ZnO, in favour to CO production via the RWGS reaction with increasing temperature.
At 250°C no CH 4 was produced within the detection limits of the GC-FID (1 ppm or~0.0005 % effective yield) for Pd/ZnO, where Pd is present as PdZn. Alongside Pd/ZnO, no significant CH 4 formation was observed for PdZn/ZnTiO 3 below 250°C (X CO 2~1 3 %) ( Figure 2). Indicating that CH 4 formation does not occur on ZnO, ZnTiO 3 , PdZn alloy facets or the combination of these. On PdZn/TiO 2 , CH 4 formation was observed at 200°C, CH 4 productivity increased with increasing reaction temperature, reaching a productivity of 0.5 mmol Kg cat À 1 h À 1 at 250°C (X CO 2 12 %). Indicating that TiO 2 is most probably involved in the production of CH 4 . Others suggest that CH 4 can be formed through a CO 2 methanation mechanism, at either Pd surfaces, [31] or via carbonate intermediates formed on TiO 2 [35] and further hydrogenation at the metal-support interface. [19] Another plausible mechanism for CH 4 production on PdZn catalysts is the migration of adsorbed methanol molecules, originating at the PdZn phase, to TiO 2 , where it decomposes via methoxide deoxygenation and methyl-hydrogenation. [36,37] On Pd/ZnTiO 3 , CH 4 production followed the same CO 2 conversion/CH 4 productivity trend as PdZn/TiO 2 ( Figure 2). Blank test for CO 2 hydrogenation, under the same reaction conditions used for PdZn catalysts, using TiO 2 , ZnO and ZnTiO 3 supports, showed very little conversion, as might be expected (Table S3). To discern whether on PdZn catalysts CH 4 is formed through CO 2 methanation or via CH 3 OH decomposition, a physical mixture of Pd/ZnO and TiO 2 was used for the CO 2 hydrogenation (Table S4). At 250°C, a CH 4 productivity of 1.8 mmol Kg cat h À 1 was observed. Firstly, this indicates that CH 4 is formedover the Pd/ZnO physical mixture via CH 3 OH decomposition, and not through CO 2 methanation as reported for Pd catalysts; [38] and secondly, that CH 3 OH can adsorb and decompose at TiO 2 surfaces, and not exclusively at the PdZn-support interface.

Chemical, structural and morphological catalyst characterisation
ZnO, TiO 2 and ZnTiO 3 , with respective BET surface areas of 15 m 2 g À 1 , 50 m 2 g À 1 and 17 m 2 g À 1 , were used as supports for Pd or PdZn. In view of the high loading of 5 wt. % of Pd and 15 wt. % of Zn, the BET surface area of prepared catalysts after annealing (500°C, 16 h) and pre-reduction (400°C, 1 h) was measured. As shown in Table 2, the BET surface areas of the catalysts were comparable to the BET surface areas of the supports, indicating that most of the organic part of the organometallic precursor had decomposed without being deposited at the surface. Highly dispersed PdZn nanoparticles are desired for CO 2 hydrogenation to methanol, because of the improved metal surface area, and the surface area of the support employed is expected to affect the catalyst dispersion and its particle size distribution. [29] However, no significant differences in the average particle size were observed by TEM between PdZn catalysts supported on TiO 2 , ZnO or ZnTiO 3 (Table 2). Nevertheless, particle size histograms for PdZn/TiO 2 and PdZn/ZnTiO 3 , where Pd(acac) 2 and Zn(acac) 2 were impregnated on the support, showed a narrower particle size distribution with a lower frequency of larger nanoparticles (> 7 nm) compared to Pd/ZnO and Pd/ZnTiO 3 , where only Pd (acac) 2 was impregnated on the support ( Figure S2). TEM images for the synthesised catalysts can be found in Figure S3. CVI was employed for the synthesis of Pd/TiO 2 , Pd/ZnO, Pd/ ZnTiO 3 , PdZn/TiO 2 and PdZn/ZnTiO 3 . After annealing in air (500°C, 16 h), the decomposition of the organometallic precursors, Pd(acac) 2 and Zn(acac) 2 , led to the formation of PdO and ZnO respectively (see XRD Figure S4a). A peak at 33.9°for PdO (JCPDS-041-1107) was observed for all catalysts, whilst ZnO, originating from Zn(acac) 2 decomposition, was detected at 31.8°, 34.4°and 36.2°for PdZn/TiO 2 and PdZn/ZnTiO 3 . Reduc-tion treatment prior to reaction is required to form the βÀ PdZn alloy. Previous reports, [33,39,40] suggested that under reducing conditions, PdO is first reduced to Pd metal, followed by hydrogen spill over from Pd to adjacent ZnO, leading to oxide reduction and the formation of the βÀ PdZn alloy. [39,41] The alloy formation mechanism under reducing conditions (5 % H 2 /Ar) was confirmed by in situ XRD for Pd/ZnO (Figure 3). A decrease in the intensity of the peak at 33.9°, which is assigned to PdO, is observed with increasing reduction temperature from 50°C to 210°C, and appears to be complete by the latter temperature. At this temperature, the formation of Pd 0 is detected at 39.9°( JCPDS-046-1043). The peak assigned to metallic palladium remains stable under reducing conditions until 325°C, when a shoulder at 41.4°appears, indicating the incorporation of zinc into the palladium lattice to form the βÀ PdZn alloy. [40] Increasing the reduction temperature to 400°C leads to the further alloying of Pd to PdZn, as indicated by the increase in intensity of the PdZn peak at 41.4°. PdZn formation on the synthesised catalysts is expected to follow the mechanism reported by Penner et al., [39] with PdZn formation starting at the surface of Pd, and growing from the surface inwards.
To study the extent of palladium alloying after prereduction, catalysts were characterised by X-ray photoemission spectroscopy (XPS), figure 5. Even though catalysts were annealed in static air at 500°C for 16 h to remove the acetylacetonate organic moiety of the organometallic precursors, carbon was detected after pre-reduction (400°C, 1 h, 5 % H 2 /Ar), hence, the rest of the elements analysed were calibrated against the adventitious C(1 s) signal at 284.8 eV binding energy (b.e.). [42] The Pd(3d) peak for Pd/TiO 2 was centred at 334.9 eV, [30,42] indicating the presence of metallic palladium. Thorough interpretation of Pd(3d) core-electrons is challenging due to the need to use symmetric and asymmetric peaks for Pd and PdO respectively, the presence of satellites and plasmon contributions. [43] Finite-Lorentzian line shapes with Shirley background were used to fit Pd and PdZn main peaks and satellite contributions, whilst symmetric gaussian peaks were used to fit PdO main peaks and its satellites. Peak fitting on Pd/TiO 2 indicated the presence of PdO at 336.5 eV [30,44,45] (Figure 5). PdO originated from the spontaneous Pd surface passivation in contact with air when transferred into the XPS instrument. [46] For the reduced Pd/ZnO, a 1.2 eV shift towards higher binding energy is observed when compared to Pd/TiO 2 , which indicates the incorporation of Zn within the lattice to form the PdZn alloy, [30,41,47,48] in excellent agreement with the XRD characterisation. A part of PdZn at 336.1 eV, peak fitting suggested the presence of non-alloyed Pd at 335.1 eV, without a clear presence of PdO ( Figure 5). The slight shift towards higher b.e. observed on the non-alloyed Pd peak for PdZn catalysts compared to Pd/TiO 2 can be attributed to stronger metalsupport interactions [33] and to atomic Zn-doping onto metallic Pd. [49] As discussed for Pd/TiO 2 , PdO is formed through oxidation of surface Pd in contact with air, however, alloyed palladium seems to be passivated against oxidation at room temperature when exposed to air. Metallic Pd is believed to be underneath a PdZn alloy layer, according to the PdZn alloy formation mechanism described in the literature, [39] where PdO first reduces to Pd, and further reduction resulted in Zn incorporation into the Pd lattice leading to PdZn alloy, as observed by in situ XRD. The Pd(3d) peaks for Pd/ZnTiO 3 , PdZn/ZnTiO 3 and PdZn/TiO 2 were shifted towards higher binding energy compared to Pd/TiO 2 , in agreement with the formation of PdZn alloy as observed by XRD, peak fitting indicated the coexistence of PdZn and non-alloyed Pd in all catalysts ( Figure 5), the latter presumably at the core of PdZn nanoparticles. [39] Changes in zinc speciation are challenging to detect by standard XPS because of the small change in the Zn(2p) binding energy upon oxidation. For instance, Zn metal and ZnO are reported at 1021.7 eV and 1022.0 eV respectively. [42] No significant changes in the Zn(2p) b.e. were observed between synthesised catalysts ( Figure S5a). The Zn(LM 2 ) Auger peak for zinc oxide is at 988 eV with a minor satellite contribution at 991 eV, [50] whilst for metallic zinc these peaks are much more shifted than the core levels, to 992 eV and 996 eV respectively. [51] Therefore, the Zn(LM 2 ) Auger electron spectra were also analysed, since it has higher discrimination for chemical changes ( Figure S5b). For Pd/ZnO, the two main peaks of ZnO were observed at 988 eV and 991 eV, as expected, but a shoulder, indicating the presence of zinc in a more reduced stated than ZnO, was also observed at 995 eV, [52] which can be assigned to Zn 0 in the PdZn alloy. [29] The presence of PdZn in the Zn(LM) 2 Auger electron region was also observed for Pd/ ZnTiO 3 , PdZn/ZnTiO 3 and PdZn/TiO 2 . Although there is very little metal in the spectra, it is likely that, at least some of the metallic zinc is oxidised upon exposure to air before the XPS. It is interesting to see that the Pd in the alloys does not oxidise significantly (figure 4), indicating that either it is covered by ZnO, possibly due to air exposure, or that it is passivated by it.
In summary our study shows that by careful control of the preparation method and selection of a suitable support, PdZn alloy catalysts can be synthesised that can hydrogenate CO 2 to methanol without the formation of CH 4, which is an undesired product from a production viewpoint.