Interplay between carbon dioxide enrichment and zinc oxide promotion of copper catalysts in methanol synthesis Journal of Catalysis

Methanol synthesis over Cu/ZnO/Al 2 O 3 is a key industrial reaction. Typically, a ZnO x promoter and CO 2 enrichment of the feed are applied to maximize the syngas conversion. However, understanding the effects of these additives on the performance of the Cu catalysts is obscured by the strong interaction between the ZnO x promoter and oxidic supports. Here, we use Cu nanoparticles on graphitic support to study the interplay between CO 2 concentration and ZnO x promotion. CO 2 enrichment enhanced the activity with an optimum at 3 vol% in the feed, but only if the ZnO x promoter was present, demonstrating the intricate interaction between the two. Interestingly, not only the activity but also the methanol selec- tivity was enhanced by CO 2 enrichment, and even further by the addition of ZnO x . Understanding not only the role of the individual components, but also the interaction between them, is important to design catalysts for processes with more ﬂexible feed compositions.

Small amounts of CO 2 (2-6%) are generally added to the syngas mixture to increase the conversion of Cu-based methanol synthesis catalysts [20][21][22][23]. Although CO can be directly hydrogenated to methanol [21,[23][24][25][26][27], it has been clearly documented that in CO 2 -enriched feed streams the vast majority of the methanol molecules originates from the hydrogenation of CO 2 molecules [11,22,27,28]. The water formed in this process drives the watergas shift (WGS) reaction (CO + H 2 O ? CO 2 + H 2 ) [28,29] by reacting with CO in the feed over a Cu surface. In this way constant but low concentrations of water and CO 2 are sustained in the reaction atmosphere. Due to energy storage considerations it could be desirable to operate the process with high concentrations of CO 2 in the feed, or even with a pure H 2 /CO 2 feed. However, hydrogenation of pure CO 2 results in relatively high water concentrations, which give rise to a poor catalyst stability [30]. Even though the redistribution of the ZnO x promoter can also play a role, deactivation is mostly due to Cu particle growth, as investigated in detail in our lab [31][32][33] and by others [34][35][36][37]. Although the fact that slight CO 2 enrichment increases the catalyst activity and decreases its stability, it is not yet known how and if the promoter plays a role in this.
Likewise, the influence of ZnO x on methanol synthesis is well known with a standard concentration of CO 2 in the feed [6,11,[20][21][22]38,39], but little is understood about its influence in different CO 2 concentrations or while hydrogenating pure CO. Martin et al. [29] showed that for a Cu/Al 2 O 3 catalyst the methanol formation rate increased by one order of magnitude upon ZnO x promotion in the absence of CO 2 in the feed, while Nielsen et al. [25] reported no significant change in the activity between 50 wt % Cu/Al 2 O 3 and Cu/ZnO/Al 2 O 3 catalysts, whereas Studt et al. [27] and Zander et al. [40] even reported a poisoning effect by ZnO x on the activity of a 86 wt% Cu/MgO catalyst. It hence seems that the effect of ZnO x also depends on the type of oxidic support used. ZnO x is known to have a relatively strong interaction with oxidic supports, which also depends on the reaction conditions [41][42][43]. As a result, a large fraction of the ZnO x might not be in the form of zinc oxide, but rather present as for instance zinc aluminates or silicates. Hence, oxidic supports might hinder the understanding of the intrinsic influence of ZnO x on the performance of Cu catalysts. Graphite-like materials as a non-oxidic support may limit these effects and has for example been used as a support for Cu (Zn)O x particles in methanol synthesis [44], Ru particles in Fischer-Tropsch synthesis [45] and bimetallic CuNi in dimethyl carbonate synthesis [46].
A decisive factor in catalysis is the selectivity. Methanol synthesis from CO 2 -containing syngas delivers a high selectivity, typically more than 98% [14,21,34,44,[47][48][49]. This remarkably high selectivity is typically attributed to the limited effectiveness of Cu to dissociate CO, thereby limiting the production of the thermodynamically more stable hydrocarbons. Even though the methanol selectivity is generally high, it is still very relevant to investigate the nature of the side products, and how the product distribution is influenced by the ZnO x promoter and the CO 2 concentration.
Here, we study the combined effect of different amounts of CO 2 (including its absence) to the syngas feed and ZnO x promotion on the methanol selectivity and formation rate on carbon-supported Cu(Zn)O x model catalysts, to obtain insights into how these factors affect the performance of methanol synthesis catalysts. We use graphitic carbon as a model support for Cu(Zn)O x nanoparticles as this avoids strong metal-support interactions [42] and hence is expected to limit spectator species such as zinc mixed metal oxides.  3 and rinsed with water until pH 7 was reached, and finally dried at 393 K overnight before use.

Catalyst synthesis
Typically, ca. 2.3 g of carbonaceous support (high surface area graphite, Brunauer-Emmett-Teller (BET) surface area 509 m 2 g À1 ) was dried at approximately 443 K under dynamic vacuum for 1.5 h to remove water from the pores. After cooling down to room temperature the fine carbon powder was (co-)impregnated to incipient wetness [50], defined as 95% of the total pore volume by N 2 physisorption, under static vacuum with a 0.1 M HNO 3 solution containing copper (and zinc) nitrates. The concentration of copper (and zinc) nitrates were chosen such to obtain the desired weight loading. The impregnated support was dried at room temperature under dynamic vacuum overnight, and reduced at 503 K (ramp 2 K min À1 ) in a 100 mL min À1 flow of 20% H 2 /N 2 for 2.5 h. To be able to store and handle the catalyst in the oxidized state, it was exposed to a flow of 100 mL min À1 flow of 5% O 2 /N 2 for 1 h, heated to 473 K with a ramp of 1 K min À1 and oxidized at 473 K in 15% O 2 /N 2 for 1 h. The obtained X-Cu/C and X-CuZnO x /C catalysts, in which X represents the surface-averaged CuO size in the fresh catalyst, had a theoretical Cu weight loading of ca. 8.1 ± 0.3 wt% and a ZnO loading of 0.1 or 5.3 wt%.

N 2 physisorption
Physisorption was measured on a Micromeritics TriStar II Plus apparatus at 77 K. Prior to analysis the sample was dried at 443 K under a N 2 flow overnight. The BET surface area was determined according to the IUPAC procedure [51]. Pore size distributions were established via a Barrett-Joyner-Halenda (BJH) analysis, using a carbon black reference thickness curve with a Faas correction. The total pore volume V tot was derived from the amount of N 2 adsorbed at p/p 0 = 0.995. The micropore volume V micro was obtained via the t-plot method using a carbon black reference thickness curve (t = 2.98 + 6.45Á(p/p 0 ) + 0.88Á(p/p 0 ) 2 ) fitted in the linear N 2 monolayer adsorption regime. Finally, the mesopore volume V meso was derived from integration of the adsorption-based BJH curve between 2 and 50 nm. Finally, the macropore volume V macro was defined as the difference between the total pore volume and the sum of the t-plot micropore volume and BJH-derived mesopore volume.

EM imaging and EDX mapping
Catalysts were imaged by transmission electron microscopy (TEM) on a Thermo Scientific Talos F200X apparatus, operating at 200 kV and equipped with a high-brightness field emission gun (X-FEG) and Super-X TM energy-dispersive X-ray (EDX) detectors. To this end, holey carbon film-coated Cu or Au grids (Agar, 300 mesh) were dry-loaded with finely ground sample (<25 mm). Number-averaged CuO particle sizes (d N ) were determined by measuring at least 350 individual particles at different locations within the sample. These sizes were translated into surfaceaveraged (d S ) and volume-averaged (d V ) particle sizes, including the standard deviations in the width of the particle size distribu- ter of the i-th particle and N represents the total number of measured particles. Qualitative, chemical compositions were mapped by EDX spectroscopy, while imaging in high-angle, annular, dark-field, scanning transmission electron microscopy (HAADF-STEM) mode. The elemental maps were acquired using Velox TM analytical imaging software. The acquisition time per EDX map was at least 15 min, and the probe current was around 700 pA.

XRD analysis
Powder X-ray diffractograms were recorded at a Bruker AXS D2 Phaser (second generation) diffractometer at room temperature. Samples were irradiated by Co Ka radiation (k = 1.790 Å) at 30 kV and 10 mA. Used catalysts were slowly passivated in static air at room temperature prior to analysis.
Rietveld refinement was performed using Bruker DIFFRAC. SUITE TOPAS software by fitting CuO (monoclinic, C2/c) and ZnO (hexagonal, P6 3 mc) crystallographic data as Lorentzian functions between ca. 36°and 48°2h. No crystal strain or preferred orientation in the lattice planes were incorporated. The background was fitted as a first order Chebyshev polynomial, thereby taking the complete diffractogram into account. The y-coordinate of the O atom in CuO and the z-coordinate of the O atom in ZnO were refined. Crystallite sizes were obtained after the fitting converged to a minimum goodness-of-fit (GOF) value.

H 2 -TPR profiling
Temperature-programmed reduction (TPR) by H 2 was performed on a Micromeritics AutoChem II 2920 apparatus. Prior to the reduction the sample (50 mg, <75 mm granulites) was dried at 393 K under an Ar flow of 50 mL min À1 for 30 min and cooled down to room temperature. Reduction profiles were recorded with a thermal conductivity detector (TCD) when the samples were exposed to a 5% H 2 /Ar flow of 50 mL min À1 up to 873 K with a ramp of 2 K min À1 . H 2 O was captured with a dry ice/isopropanol cold trap.

Catalytic performance measurements
Catalysts were evaluated in a 16-reactor setup (Flowrence, Avantium) for 160 h, operating at 40 bar(g) and 533 K using a H 2 / (CO + CO 2 ) molar ratio of 2. The catalysts powders were pelletized under a maximum pressure of 1 ton and sieved in granulites between 75 and 150 mm. The stainless steel reactors (inner diameter of 2.6 mm) were loaded with 178 mg 10-Cu/C, 51 mg 6-CuZnO x / C or 6 mg commercial Cu/ZnO/Al 2 O 3 /MgO catalysts and diluted with inert SiC (sieve fraction of 212-425 mm). This resulted in an inert fraction by SiC between 25 and 92 vol%, thereby minimizing the error by dilution to a maximum of ca. 3% [52]. The different sizes of the sieve fractions between the catalysts and diluent facilitated post-analysis by EM and X-ray diffraction (XRD).
First, an in situ reduction was performed in 2.8 mL min À1 of 5% H 2 /N 2 at 523 K for 3 h. The temperature was lowered to 393 K before the reactors were flushed with syngas (H 2 /CO/He = 60/30/ 10 vol%) at a GHSV of 400, 1,400 or 24600 h À1 for 1 h for the 10-Cu/C, 6-CuZnO x /C and commercial catalysts, respectively. A tri-phase carbonyl trap (active carbon, c-Al 2 O 3 , ZnO) was located upstream the CO feed to remove metal carbonyls and sulfur species. The flows and catalyst loadings were adjusted to compensate for differences in Cu loading (8 vs 58 wt%) and packing densities of the sieved catalysts (0.52 vs 1.09 g mL À1 ) to achieve conversions in the same range.
Next, the reactors were pressurized to 40 bar(g) and heated to the reaction temperature of 533 K with a ramp of 2 K min À1 . After obtaining catalytic data for 40 h in a H 2 /CO/He gas atmosphere a small amount of CO was replaced by CO 2 and the gas composition was equilibrated for 1 h. After 30 h on stream in a CO 2 -enriched feed, the CO 2 content was stepwise further increased in a similar way up to H 2 /CO/CO 2 /He = 60/23/7/10 vol%. Products were periodically analyzed by online gas chromatography every 15 min. After catalysis the samples were slowly exposed to air at 393 K. The calculations of activity, selectivity and stability are detailed in Section S5.

Structural properties of the catalysts
First we discuss the particle size and elemental distribution in the catalysts as studied by electron microscopy and energydispersive X-ray (EDX) spectroscopy (Table 1). Fig. 1 shows electron micrographs of two selected carbon-supported CuO catalysts, either promoted with ZnO x or not [53]. In bright-field transmission electron microscopy (BF-TEM) (frames C and A) particles of respectively ca. 5.5 nm and 9.3 nm (dark dots) are uniformly distributed over the support surface (light grey). Similarly sized and welldistributed nanoparticles were consistently found throughout these catalysts (more examples in Fig. S1). Elemental mapping (frame E) shows that the ZnO x in the as-synthesized 6-CuZnO x /C catalyst was evenly dispersed over the support surface and that the observed particles in BF-TEM (frames A and C) were hence mostly consisting of Cu(O). However, the elemental distribution after high-pressure methanol synthesis (frame F) is probably more representative of the catalyst during action, and demonstrates a close intimacy between the Cu and Zn species. The high surface area carbon support consisted of stacked graphitic sheets and exhibited a surface area of 509 m 2 g À1 and a mesopore volume of 0.36 mL g À1 , while after Cu and Zn deposition 297 m 2 g À1 and 0.23 mL g À1 were obtained (Section S2). The high surface area of the carbon support clearly allowed a uniform distribution of nanoparticles over the support [46,[54][55][56].
Frames B and D show that the CuO particle sizes were distributed in a lognormal fashion for the fresh 10-Cu/C and 6-CuZnO x /C catalysts. Based on the histograms the particle size distribution was also translated into a surface-averaged size, which is relevant for catalysis, and a volume-averaged size, which can be directly compared to X-ray diffraction (XRD) [57,58] (Table 1). A few agglomerates were present in the used catalysts (Fig. S7) but not taken into account in the corresponding particle size averages obtained from BF-TEM. The preparation and characterization of 10-CuZnO x /C catalysts containing 0.1 wt% ZnO was reproduced three times (Fig. S2). For all three 10-CuZnO x /C catalysts similar number-averaged particle sizes (7.6 ± 1.1 nm) and distributions were obtained, proofing that the catalyst synthesis was reproducible. All three carbon-supported catalysts shown in Table 1 had a similar Cu loading (8.1 ± 0.3 wt%), which was confirmed by temperature-programmed reduction with H 2 (7.7 ± 1.4 wt%, Fig. S3). These results show that the presence of sufficient ZnO x , as in the 6-CuZnO x /C catalyst, limited the growth of CuO x nanoparticles during catalyst synthesis.
Powder X-ray diffraction patterns of the as-synthesized 10-Cu/C and 6-CuZnO x /C catalysts (Fig. 2, frame A) showed, next to the graphite diffraction lines at 30.7°and 50.4°2h, only diffractions attributed to CuO crystallites at 37.9°, 41.4°, 45.2°and 57.1°2h. The crystallite sizes were respectively ca. 16 and 6 nm and corresponded to the volume-averaged CuO particle sizes, derived from the BF-TEM particle size histograms, of 10.4 ± 3.5 nm and 6.2 ± 2. 0 nm, respectively. No diffraction peaks due to ZnO were observed for the fresh 6-CuZnO x /C catalyst, which is in line with the high ZnO x dispersion observed in the fresh catalyst with elemental mapping (Fig. 1, frame E). After catalysis the ZnO (partly) crystallized (Fig. 2, frame B), showing crystallites of 12 nm. This observation was in agreement with the fact that in the used 6-CuZnO x /C catalyst also some larger particles were found in the sample (Fig. S7). In both used 10-Cu/C and 6-CuZnO x /C catalysts (frame B) the diffractions at 50.6°and 59.1°2h are attributed to 13 and 12 nm Cu 0 crystallites, respectively. Please note that these Cu 0 sizes are slightly smaller than the CuO sizes in the fresh catalysts (Table 1), but this difference may be explained by the fact that not all Cu is in the metallic phase. Due to passivation of the samples after catalysis different phases of copper oxide were present.

(frame A) shows diffractions at 35°, 38°and 42°2
h attributed to ZnO and with a width corresponding to an approximate crystallite size of 4 nm. The catalyst typically contained ca. 10 nm-sized Cu(O) particles [10] and 58 wt% Cu with a Zn/ (Cu + Zn) molar fraction of 0.28, which was similar to our 10-CuZnO x /C catalyst (0.35). The commercial catalyst is known to have a close intimacy of Cu and Zn species (Fig. S4), whereas Al 2 O 3 mainly acts as a spacer [7][8][9].

Influence of ZnO x on activity and selectivity during CO hydrogenation
First we tested methanol synthesis with only CO and H 2 in the feed (hence, without CO 2 ). Fig. 3 (frame A) shows the CO conver-  . Surface-averaged CuO particle sizes were 9.9 ± 3.3 nm and 5.8 ± 1.9 nm, respectively. Elemental distribution of Cu and Zn in (E) the fresh 6-CuZnO x /C catalyst in the same area of frame C and (F) the used 6-CuZnO x /C catalyst. sion as a function of time for two selected catalysts, 10-Cu/C and 6-CuZnO x /C, as well as for the commercial Cu/ZnO/Al 2 O 3 /MgO catalyst to put our catalysts into perspective. The commercial catalyst was stable from the start, probably because it had already been exposed to temperatures >573 K [10,44], although the stabilizing effect by Al and/or Mg species could also play a role [5]. For both carbon-supported catalysts an activation period of ca. 20 h was observed, which can be explained by the fact that this was the first time that they were subjected to 533 K and high pressure. The hollow symbols represent a duplicate test from the same catalyst batch, proving that the catalytic testing was reproducible. All these conversion levels are well below the thermodynamic equilibrium of 35.7% (see Table S3 for further details) [60,61]. Table 2 shows the CO conversion, also normalized to the Cu mass in the reactor, as well as the turnover frequencies based on the particle sizes in the fresh and used catalysts. Note that these different activities are not directly reflected in Fig. 3 (frame A) as for these tests the flows and catalyst loadings were adjusted to compensate for differences in Cu loading and packing densities, in order to achieve conversion levels in the same range. By comparing the activities of the 10-Cu/C and 6-CuZnO x /C catalysts it is clear that the ZnO x addition led to an activity increase of about one order of magnitude. After 40 h on stream our 6-CuZnO x /C catalyst had a methanol turnover frequency of 2.7Á10 À3 s À1 , which was similar to the commercial Cu/ZnO/Al 2 O 3 /MgO catalyst with an activity of ca. 2.4Á10 À3 s À1 . The turnover frequency of the 6-CuZnO x /C catalyst may be even higher as a few agglomerates were found in this catalyst, but not taken into account in the particle size average (Fig. S7). As mentioned earlier in the introduction, only a few and somewhat contradictory studies are available regarding the effect of ZnO x while hydrogenating pure CO, which might be induced by the use of (different) oxidic supports. Therefore, we studied the effect of ZnO x addition during CO hydrogenation while using an inert graphitic support [42]. Fig. 3 (frame B) presents the evolution of the methanol formation rate over time. After the activation period all catalysts showed a stable methanol production. However, the methanol formation rate (and selectivity from Table 2) was surprisingly low (<8 mmol g Cu -1 s À1 ), even for the commercial catalyst, compared to what is typically reported for CO 2 -enriched syngas conversion [44,47]. This might be induced by a slight deficiency in hydrogen [61] and the slightly high reaction temperature. Although there are few old studies available on Cu/ZnO/Al 2 O 3 catalysts [21,23,24], as far as we are aware this is the first time that the methanol formation rates are reported for carbon-supported Cu particles in the presence and absence of ZnO x during pure CO hydrogenation. Fig. 4 displays the evolution of the other products formed over the 10-Cu/C and 6-CuZnO x /C catalysts, averaged over at least four separate measurement runs. In both cases CO 2 was the main side product, but also hydrocarbons, ethanol and, in the case of the 6-CuZnO x /C catalyst, dimethyl ether (DME) were detected. The hydrocarbon formation rates were ca. 4 orders of magnitude lower than typically found in iron-based Fischer-Tropsch synthesis [62]. DME was probably formed by dehydration of methanol (2CH 3 -OH ? (CH 3 ) 2 O + H 2 O) [63] on the acidic oxygen vacancy sites on the ZnO x surface [64,65]. Interestingly, CO 2 was produced from CO, while no water or oxygen was present in the feed. Other oxygen sources such as ZnO x or contaminations in the feed, SiC diluent and/or carbonyl trap were excluded (Section S4). This CO 2 production could be explained by the co-formation of hydrocarbons, ethanol and DME. The total formation rate of these compounds for the 10-Cu/C catalyst (5.1 ± 1.1Á10 -8 mol g Cu -1 s À1 ) was similar to the rate  Table 2 Overview of catalytic results during CO hydrogenation in the absence of CO 2 . The initial turnover frequency (TOF) was calculated using the CuO particle size in the fresh catalyst and the activity at t = 0. The final TOF was estimated from the CuO particle size after 160 h of catalysis (see also Fig. S6) and the activity at t = 40 h. The methanol selectivity included CO 2 in the product distribution. The errors indicate the standard deviation between two measurements unless stated otherwise. Conditions were equal to Fig. 3 [A] The error in the TOF reflects the width in the CuO particle size distribution from BF-TEM.
[B] The CuO crystallite size (from XRD) in the fresh catalyst was used in the calculation.
An analogous observation was made for the 6-CuZnO x /C catalyst (6.3 ± 1.7Á10 -7 vs 7.2 ± 1.8Á10 -7 mol g Cu -1 s À1 ). Hence, the following reaction plausibly occurred (next to direct CO hydrogenation to methanol [10,21,[23][24][25][26][27]63]): 2xCO + (x + 1)H 2 ? xCO 2 + C x H 2x+2 . An intriguing question is why the produced CO 2 was not (entirely) consumed by hydrogenation to methanol. From literature it is known that CO 2 is preferred over CO as the carbon source for methanol production at high temperatures [11,28]. The CO 2 concentration in these systems is very low with a corresponding partial pressure of 0.04 and 0.11 bar for, respectively, the 10-Cu/ C and 6-CuZnO x /C catalysts after 40 h on stream, as CO 2 was only generated during the reaction. To our knowledge this is the first time that these relatively low methanol selectivities in a pure H 2 / CO feed have been reported and tentatively explained.

Combined influence of CO 2 and ZnO x
In industrial methanol production a CO 2 -enriched syngas feed is typically used [20]. Fig. 5 (frame A) shows the effect of CO 2 concentration in the syngas feed on the CO + CO 2 conversion for carbonsupported Cu catalysts with and without ZnO x as well as for the commercial catalyst. Please note that this is the net conversion, as we cannot determine the extent of (reverse) WGS taking place in the reactor. The conversion is taken after at least 30 h on stream in a specific gas composition (for the time evolution see Fig. S6). Even 1 vol% CO 2 in the feed (obtained by replacing 3% of the CO with CO 2 ) increased the CO + CO 2 conversion for the 6-CuZnO x /C catalyst from 5.7% to 11.9% and with 3 vol% CO 2 in the feed the maximum CO + CO 2 conversion was reached, while with even higher CO 2 content the conversion level again decreased, down to 6.8% with 7 vol% CO 2 in the total feed. An effect of CO 2 in syngas on the conversion has been observed earlier [6,11,[20][21][22][23]38,39]. A similar conversion increase was observed for the commercial catalyst for low and increasing CO 2 concentrations, but the decline at higher concentrations was much less pronounced. The decrease in the conversion level for the 6-CuZnO x /C catalyst at higher CO 2 concentrations in the feed (frame A) could be explained by a combination of Cu particle growth, ZnO x promotion loss and/or Cu surface oxidation. TEM analysis (Table 1) showed only a minor CuO x particle growth to 8.8 ± 3.1 nm in the used 6-CuZnO x /C catalyst, although probably a significant amount of the Cu atoms were located in several larger particles, but contributing little to the active surface area (Fig. S7). XRD analysis of the used catalyst (Fig. 2, frame B) indicated a significant ZnO crystallization to approximately 12 nm, even though a close interaction between ZnO x and Cu was preserved (Fig. 1, frame F, Fig. S7). Another pos-sible consequence of the higher CO 2 concentration (and the additional H 2 O formation), and hence a more oxidizing atmosphere, could be a higher oxygen coverage of the Cu surface, which is reported to render the catalyst less active [6,22,27].
The influence of CO 2 in the syngas feed on the activity of unpromoted supported Cu particles was so far not clear, as reported results varied depending on the type of oxidic support used [25,27]. Furthermore, as we also found (Fig. S8), Cu-based catalysts can show a relatively long activation period and performance might be influenced on the activation protocol, which is a topic of further study. However, based on carbon-supported model catalysts we clearly show that for unpromoted Cu nanoparticles, in the absence of any oxidic phase, the syngas conversion level (Fig. 5, frame A) is not significantly influenced by the CO 2 content. Hence, synergy between Cu nanoparticles and ZnO x (or another appropriate oxidic promoter) is clearly needed for effective CO 2 hydrogenation, which plays such an important role in the industrial methanol synthesis process [27,66]. Fig. 5 (frame B) presents the methanol selectivity of the catalysts in different syngas compositions, thereby not taking CO 2 in account as a possible product in the product distribution, as a function of the CO 2 content in the feed (for the time evolution see Fig. S6). Interestingly, a modest addition of CO 2 boosted the methanol selectivity by ca. 8% C for the 6-CuZnO x /C and 10-Cu/C catalysts. In CO 2 -enriched syngas feeds the commercial catalyst reached a methanol selectivity of >99% C , which was also reported by our group at similar CO + CO 2 conversion levels [47,67]. The lower selectivity for the 10-Cu/C catalyst (89% C ) was also observed for a Cu/SiO 2 catalyst (92% C ) and might be caused by its low conversion level [68]. It has to be noted that the hydrocarbon formation rate was similar in all syngas atmospheres (2.7Á10 -7 mol g Cu -1 s À1 or 1.5Á10 -4 s À1 for the 6-CuZnO x /C catalyst), showing that the methanol selectivity increase was governed by an increased methanol formation rate (see also Table S2). This remarkable improvement with the addition of a small amount of CO 2 can be understood for the 6-CuZnO x /C catalyst by the DME suppression, as sufficient H 2 O was generated via the CO 2 hydrogenation to methanol. Additionally, the higher H 2 O content seems to play a role in lowering of the CAO bond dissociation activity with respect to the methanol formation rate, as discussed in detail in Section 3.2. Fig. 5 (frame C) illustrates the methanol turnover frequency (TOF MeOH ), serving as a combination of the CO + CO 2 conversion (frame A) and the CO 2 -free methanol selectivity (frame B). For the ZnO x -promoted catalysts, the trends in the TOF MeOH reflect those of the CO + CO 2 conversion. Also, our 6-CuZnO x /C catalyst performed in the same order of magnitude as the commercial catalyst, hence illustrating the relevance of our model catalyst study. The slightly lower conversion at higher CO 2 contents in the feed might be explained by an increased loss of ZnO x promoter effect by crystallization, dealloying and/or overcoverage over active Cu surface area [30,39] compared to the stabilized commercial catalyst by metal oxides. For the 10-Cu/C catalyst the TOF MeOH (frame C) did not change at all (2.1Á10 -4 s À1 ) upon CO 2 enrichment of the feed, again strongly suggesting that the Cu catalyst alone is quite inactive in CO 2 hydrogenation.

Conclusions
We prepared a series of carbon-supported CuZnO x catalysts with a uniform distribution of Cu nanoparticles and, when present, ZnO x in close contact with Cu during catalysis. While hydrogenating pure CO ZnO x promoted the activity by about one order of magnitude and led next to methanol to the production of dimethyl ether. A maximum activity of the promoted catalyst was obtained with 3 vol% CO 2 enrichment of the syngas, whereas the methanol formation rate over unpromoted Cu was not influenced by CO 2 addition. Interestingly, alternative products (>10% C ) such as CO 2 and hydrocarbons were formed for all catalysts in the absence of CO 2 in the feed in concomitance with methanol. The methanol selectivity of all catalysts was very high when CO 2 was present in the feed.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. thereby not taking CO 2 in account as a possible product in the product distribution (see also Equation (S11)), and (C) methanol turnover frequency (TOF MeOH ) at different syngas compositions. The catalysts were exposed to increasing CO 2 concentrations in the feed at a constant H 2 /(CO + CO 2 ) ratio. Values were determined after 30 h on stream in a specific gas composition (40 h on stream for no CO 2 in feed). The TOF MeOH was estimated using the CuO particle size after 160 h of catalysis. Error bars represent the standard deviations between two measurements (frames A and B) or reflect the width in the CuO size distribution in the used catalysts (frame C). Conditions: H 2 /CO + CO 2 /He = 60/30/10 vol%, further equal to Fig. 3.