Reichenbach 2018
Reichenbach 2018
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
a r t i c l e i n f o a b s t r a c t
Article history: Methanol formation from CO2 and molecular hydrogen on ZnO/Cu catalysts is studied by gradient cor-
Received 6 November 2017 rected density functional theory. The catalytically active region is modeled as a minimum size inverse
Revised 19 January 2018 catalyst represented by ZnXOY(H) clusters of different size and a ZnO nano-ribbon on an extended Cu
Accepted 30 January 2018
(1 1 1) surface. These systems are chosen as a representative of thermodynamically stable catalyst struc-
tures under typical reaction conditions. Comparison to a high level wave function method reveals that
density functional theory systematically underestimates reaction barriers, but nevertheless conserves
Keywords:
their energetic ordering. In contrast to other metal-supported oxides like ceria and zirconia, the reaction
Inverse catalyst
methanol
proceeds through the formation of formate on ZnOX/Cu, thus avoiding the CO intermediate. The differ-
CO2 ence between the oxides is attributed to variance in the initial activation of CO2. The energetics of the
Renewable energy formate reaction pathway is insensitive to the exact environment of undercoordinated Zn active sites,
hydrogenation which points to a general mechanism for Cu-Zn based catalysts.
DFT Ó 2018 Elsevier Inc. All rights reserved.
Coupled cluster
1. Introduction fossil resources) from the greenhouse gas CO2 and H2 is a promis-
ing strategy to store renewable energy and produce a chemical
Mankind is facing considerable challenges in their quest to feedstock (CO2 + 3H2 ! CH3OH + H2O) [2,3]. Contrary to H2,
secure sufficient energy supply for the future and to solve the glo- methanol is liquid at room temperature and thus it can easily be
bal warming problem [1]. Future energy sources have to be renew- stored, transported and further processed e.g. to oxymethylenedi-
ables. Most of all wind and solar energy already cover an methylethers (a novel class of fuels that promises rich applications
unexpectedly large part of our energy production and are likely [4]). Furthermore, conversion of CO2 to methanol is of great envi-
to dominate the energy market in near future. However, renewable ronmental relevance as a strategy for reducing the concentration
energies are often produced when they are not needed while their of this anthropogenic green house gas in the atmosphere [5,6].
production is low at times of large demand. Therefore, large scale The catalyst used for synthesizing methanol from synthesis gas
effective energy storage is needed. Electric energy can be used to (syngas) containing CO and CO2 on industrial scale corresponds to
split water resulting in H2 as energy carrier. Since molecular hydro- composites of Cu/ZnO/Al2O3 [7–9] and the role and interplay of the
gen is delicate to handle, it should be processed further. The syn- different components under reaction conditions are under active
thesis of the base chemical methanol (currently produced from debate. Some studies claim that the active phase corresponds to
the CuZn alloy formed under reaction conditions upon the partial
reduction of the ZnO phase [10,9,11]. In contrast, industrial cata-
q
Supplementary Material (SM) available: [Details of the computational method- lysts were reported to contain a ZnOX-overlayer on metallic Cu-
ology, the global minima search and comparison to (DLPNO-)CCSD(T) calculations. nanoparticles [7] and recent experimental results show evidence
Methodology for the inclusion of thermodynamics. Evaluation of the crucial steps of an improved catalytic activity [12,13] of ZnOX particles on Cu
using different models. Structure of the nano-ribbons. Wannier orbitals analysis of (1 1 1) as compared to the conventional metal-on-oxide configura-
CO2. Energetics and structures of the full reaction network.]
⇑ Corresponding author at: Freiburger Materialforschungszentrum, Stefan-Meier- tion [14–16]. These results strongly suggest that the active sites in
Str. 21, 79104 Freiburg, Germany. methanol synthesis catalysts correspond to the ZnO phase or to the
E-mail address: Michael.Walter@fmf.uni-freiburg.de (M. Walter). ZnO-Cu interface [8,12,7,17,18,13]. In addition to their often supe-
https://doi.org/10.1016/j.jcat.2018.01.035
0021-9517/Ó 2018 Elsevier Inc. All rights reserved.
T. Reichenbach et al. / Journal of Catalysis 360 (2018) 168–174 169
rior activity, inverse catalysts have become a valuable tool for Zn3O Zn7O3 Zn9O7 W3
investigating reaction mechanisms and the role of the oxide and
the metal-oxide interface [19–22]. Similarly, recent density func-
tional theory studies of supported oxide-clusters-models brought
new insights to the catalytic activity of PtCo/TiO2 and PtCo/ZrO2-
catalysts regarding CO2-hydrogenation [23]. The CeOX/Cu(1 1 1)-
system was reported to show an even superior performance as
compared to ZnOX/Cu(1 11 ), with the reaction mechanism consist-
Fig. 1. Global minima structures of Zn3O, Zn7O3, Zn9O7 and a zigzag edge ZnO nano-
ing of a reverse water–gas shift (RWGS, CO2+H2 ! CO + H2O) fol- ribbon (W3) supported on Cu(111). Zn: light blue, O: red, Cu: brown. (For
lowed by the CO hydrogenation (CO + 2H2 ! CH3OH) [19]. A interpretation of the references to colour in this figure legend, the reader is
similar reaction path was recently proposed on Cu supported referred to the web version of this article.)
ZrO2 and on TiO2 [24]. In contrast, the majority of theoretical and
experimental investigations indicate that the hydrogenation of
CO2 does not proceed through a CO-intermediate using Cu/ZnO forms. This results in a stabilized Zn3O unit on the extended Cu
catalysts [8,25,18,26,27,13]. Although the ZnOX/Cu(1 1 1) was (111) surface as a minimal model for ZnO/Cu catalysts. A more
already suggested and explored in early experiments [12,28], only detailed description of the full thermodynamic analysis performed
very recently a copper supported Zn6(OH)7 structure was consid- is in preparation and will be published elsewhere.
ered as possible catalyst [13]. Our work further rationalizes the In the next step, reaction pathways and transition barriers for
importance of such structures, explores their global minima struc- methanol formation catalyzed by Zn3O/Cu(111) are determined.
ture, and shows the broad independence of the results from the However, before studying the entire reaction network the applica-
explicit size and composition of the reduced ZnO/Cu inverse bility of DFT was verified.
catalyst. The PBE functional is able to predict the relevant gas-phase
energetics in good agreement to CCSD(T), DLPNO-CCSD(T) [42–
49] (see SM for the details) and experiment as is shown in Table 1,
2. Computational setup
where BEEF-vdW performs much worse [50,51]. The only excep-
tion is RWGS with a deviation of 0:45eV with respect to experi-
We have explored the complex reaction network for methanol
ment. This error can be attributed to a destabilization of gas-
synthesis catalyzed by Cu-supported zinc oxide by means of ab ini-
phase CO by the PBE functional [52]. Accordingly, CO adsorption
tio modeling. Because of the heavy computational burden of the
on Cu(111) is known to be overestimated within the generalised
high level quantum chemistry, we use the smallest possible sup-
gradient approximation (GGA) by 0.32–0.41 eV [53]. The erroneous
ported zinc oxide unit that still describes the relevant reactions.
gas-phase CO energetics in GGAs is therefore not to be expected to
Methods based on density functional theory (DFT) are used for
affect the energetic ordering of supported species discussed below
large scale electronic structure calculations of nanostructures sup-
and a correction [51] is not needed.
ported by extended surfaces. The accuracy of these calculations is
DFT within the generalized gradient approximation is also
verified by comparing the DFT energetics of gas-phase reactions
known to underestimate reaction barriers [54,55] (database tests
and a subset of catalytic processes with the Coupled Cluster
for catalysis showed mean signed errors of PBE up to 0:59 eV
method (CCSD(T)), the gold standard of quantum chemistry, that
[56]), however. In order to approve the accuracy and applicability
is considered to give accurate agreement with experiment [29–
of our DFT approach in the present systems we have performed
32] if numerical settings are converged [33,34].
single point high level DLPNO-CCSD(T) quantum chemistry calcu-
The main part of the calculations use the projector augmented
lations for selected key reactions. DLPNO-CCSD(T) is computation-
wave method as implemented in GPAW [35,36] where the wave
ally very demanding such that the reactions have to be described
functions are represented on real space grids with grid spacing
on a copper cluster model of finite size.
h 0:18 Å. The exchange correlation energy is modeled in the gen-
We have chosen the Cu4Zn3O cluster as structural model (see
eralized gradient approximation as proposed by Perdew, Burke and
lower panels of Fig. 2 a). The four Cu-atoms were fixed using the
Ernzerhof (PBE) [37]. We contrast these results using the BEEF-
experimental lattice constant 3:61 Å of bulk Cu to imitate the Cu
vdW-functional [38], which is better suited for van der Waals con-
(111)-surface [57], while the supported Zn3O and the adsorbates
tributions. A detailed description of the methods applied and
were relaxed using PBE. The values reported for BEEF-vdW and
parameters chosen can be found in Supplemantary Material (SM).
for DLPNO-CCSD(T) were obtained from subsequent single-point
calculations.
3. Discussion Four key reaction steps have been chosen as benchmark. A: ini-
tial transition of CO2 from a linear to a bent configuration, B:
As the first step, we have searched for global minima of the ZnX- hydrogenation of oxygen in a carboxyl intermediate, C: second
OY/Cu(111)-system using a recently implemented genetic algo- hydrogenation of an HCO intermediate and D: hydrogenation of
rithm [39,40] in the size range X ¼ 3 9 and selected Y. The the carbon atom in the carboxyl group. The energetics was deter-
ZnXOY units were modeled on a continuous Cu(111) slab to take mined for the Zn3OCu4 cluster model as well as for the Zn3O/Cu
into account the metallic nature of the copper support. This study (111) extended surface model for all four test reactions. The differ-
revealed that the typical ZnO/Cu-interface can be conveniently ences between the energies of the initial state EIS and the final state
approximated by a Zn3O-unit, i.e. a reduced form of ZnO, c.f. EFS compare reasonably well between all computational
Fig. 1. Interestingly, the Zn3O-unit is part of the bulk ZnO wurtzite approaches on the cluster model, i.e. these energy differences are
structure and this structural motif can also be found in ZnO over- fairly well described by DFT.
layers in industrial Cu/ZnO catalysts under certain conditions [7]. The most important quantities are the barriers determined by
Larger ZnXOY units show similar motifs and additional oxygen their corresponding transition state energy ETS . These were deter-
atoms do not improve stability (see Supplementary Material for mined by the nudged elastic band method [58] combined with a
details). According to our ab initio thermodynamics analysis, climbing image condition [59]. Fig. 2 shows both forward
zinc-oxide is strongly reduced at typical reaction conditions [8], (ETS EIS ) and backward barriers (ETS EFS ) for the chosen reac-
such that Zn3O and Zn7O3 from Fig. 1 are thermodynamically stable tions. The predicted barrier heights from PBE and BEEF-vdW are
170 T. Reichenbach et al. / Journal of Catalysis 360 (2018) 168–174
Fig. 2. Selected reactions and barrier heights computed by different ab initio methods. a) The structures for initial (IS), transition (TS) and final states (FS) for the surface and
corresponding cluster models. (b) Energetic differences between final and initial state, (c) forward barriers (from initial to transition state) and (d) backward barriers (from
final to transition state). Energies are in eV (1 eV = 96.49 kJ/mol). Zn: light blue, O: red, Cu: brown, C:grey, H: white. (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)
Table 1
Electronic energy differences of gas phase reactions compared to experimental data [41] corrected for thermodynamic and zero point energy contributions (uncorrected values
are given in brackets). Ensemble estimate errors [38] are given for BEEF-vdW. See Supplementary Material for the computational procedure.
very similar. The latter gives slightly higher values in particular for (see SM for details). All reaction steps are also explored on the
the backward barriers of steps C and D, where both H2CO and Zn3O/Cu(111) models where only marginal deviations to Zn3OH/
HCOOH are van der Waals bonded to Zn3O. The DFT results system- Cu(111) are found as detailed below. The reactions need many dis-
atically underestimate the barriers predicted by DLPNO-CCSD(T), sociated hydrogen atoms that are considered to be available on the
in particular the PBE values lie approximately 0:5 eV below Cu(111) surface under reaction conditions [8] (despite a small pos-
DLPNO-CCSD(T). Despite its good description of transition state itive Gibbs free energy difference DG ¼ 0:15 V). Gibbs free energies
energies for dissociation on transition metal surfaces [60,61] are calculated using ab initio thermodynamics at T = 500 K, 40 bars
BEEF-vdW generally has the same tendency, but improves in cases of H2, 10 bars CO2 and 1 bar of methanol, CO and H2O, correspond-
where van der Waals contributions are important [62]. The under- ing to the operating conditions for ZnO/Cu catalysts. While changes
estimation is rather independent of the reaction itself such that the in these parameters might strongly influence reaction rates [3],
relative heights of the DFT barriers generally show good qualitative their influence on the calculated energetics is rather small.
agreement to DLPNO-CCSD(T). Most importantly, the ordering of All reaction routes, except the lowest energy pathway, start
the barriers is not changed. We therefore conclude that the DFT with an initial activation of CO2 by transforming it to a bent config-
description of the underlying processes is qualitatively correct uration with the carbon atom attached to the Cu support and one
and can indeed be used to determine the most probable reaction of the oxygen atoms to Zn. This bent configuration is typical for
path. reductive CO2 activation by hetero-bimetallic systems and was
Some larger variations can be observed when the cluster model observed in manifold for molecular systems in organometallic cat-
is compared to the surface model. The deviation between the mod- alysts. This may lead to either activation [65] or even a complete
els is largest for steps C and D, in particular for the ETS EIS term. disruption of the C = O bond in CO2 and subsequent formation of
The energy difference in these configurations is caused by the a M-O-M0 -CO system [66]. In contrast to these coordination com-
hydrogen atom that is attached solely to copper, a situation that plexes and other oxides [19,24] this configuration is energetically
obviously is poorly represented by the Cu4 cluster in comparison unfavorable on reduced zinc-oxide and leads to relatively large
to the extended Cu(111) surface. The Cu4 cluster cannot account barriers. Combined with the loss of entropy upon adsorption,
for the metallic nature of the copper surface due to its molecule 1.07 eV is required to activate CO2 in this way. The corresponding
like orbitals that do not resemble a band structure. Otherwise, backward reaction CO2⁄⁄!CO2⁄ is barrier-less (0:01 eV) such that
the cluster model captures the main features of the reactions on a further forward barrier of 0.58 eV has to be overcome in order to
the surface model, particularly those related to the ZnO-part, bind H⁄ to oxygen, resulting in a total barrier of 1.64 eV (see Fig. 3).
where the reaction steps take place. Splitting off CO in the RWGS reaction needs up to 2:08 eV. The
We proceed to investigate various possible reaction pathways alternative trans-COOH pathway is strongly hindered energeti-
[63,64] on the Zn3O/Cu(111) surface model (depicted schemati- cally, having to overcome an effective barrier of 2.96 eV due to
cally in Fig. 3) employing PBE gradient corrections. Since H2 is pre- insufficient stabilization of reaction intermediates (see SM for the
sent in large quantities during the reaction, the adsorption of energetics of the whole reaction network). These reaction paths
hydrogen atoms on various sites is considered. We find that hydro- are thus rather improbable on the Cu/ZnO system.
gen adsorption is energetically favored only on the O-atom of the We have further investigated the initial activation of CO2 by
Zn3O-unit. Therefore, one hydrogen atom on top of the oxygen is analyzing the corresponding Wannier orbitals, shown in Fig. 4.
included resulting in an additional Zn3OH/Cu(111) surface model Wannier orbitals are formed from the occupied Kohn-Sham orbi-
T. Reichenbach et al. / Journal of Catalysis 360 (2018) 168–174 171
Fig. 4. (a) Definition of bond lengths du ; dM and CO2 angle h for the values listed in
Table 2. (b) Wannier orbitals of [CO2], [HCOO] (formate) and HCOOH (formic
acid), and CO2 bound straight to Zn, bent to Zn, Ce and Zr in the surface model. Only
the binding metal atoms of the surface model are colored. Wannier orbitals
corresponding to triple bonds are depicted in green, double bonds are depicted in
yellow and Wannier orbitals corresponding to single bonds in blue. Representative
formal Lewis structures are depicted below. O:red, C:gray, Cu: brown, Zn: light blue,
Ce: light yellow, Zr: cyan. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
Table 2
Atomic distances du ; dM in Å, CO2 angle h in degrees, CO2 adsorption energy (negative energies are exothermic) Eads in eV and the Bader charge on the CO2-part qCO2 and on the
metal atom qM in unit charges for the molecules in different configurations (see Fig. 4 for the labelling). The values are given for the MZn2OH and the MZn2O model [in square
brackets].
du dM h Eads qCO2 qM
Fig. 5. a) Reaction pathway of CO2 hydrogenation to methanol on the Zn3OH/Cu(111)-system and the Gibbs free energy diagram calculated at T = 500 K and partial pressures
of 40; 10; 1; 1, and 1 bars of H2, CO2, methanol, CO, and H2O, respectively. All energies are relative to CO2 in the gas phase, 6H⁄/Cu(111) far from ZnO and the bare cluster
Zn3OH/Cu(111). Gas-phase species are marked by (g), while molecules marked with ⁄ are adsorbed. In addition, energy levels of transition states are denoted with the
corresponding forward barrier. Only the metal atoms participating in bonds with the reactants are indicated. (b) Electronic energies of the rate limiting barriers enclosed in
the orange rectangle in (a) for different model structures. (c) Structure of H2CO2 intermediates on Zn7O3H and Zn7O3.
hyde proceeds via a surprisingly low barrier (0.14 eV) leading to of 1.30 eV as in the other models. Similarly, none of the models sta-
the strongly bound CH3O radical that reacts to the methanol pro- bilizes activated CO2 significantly (see Table S2 in SM) underlining
duct with an activation energy of 0.70 eV. the generality and the importance of the initial activation for selec-
The rate limiting step for methanol synthesis on the ZnOH/Cu tion of the formate path.
(111) is the hydrogenation of formate towards H2COOH in agree-
ment to the literature [8,63,13]. To validate the universality of this 4. Conclusions
minimal model, we have investigated this main step in the reaction
path as well as the energetics of the CO2 activation also on Zn7O3(- In conclusion, our ZnXOYH/Cu(111) and ZnO-nanoribbon/Cu
H)/Cu(111)) and on a nano-ribbon supported on Cu(111) generated (111) models for the industrial ZnOX/Cu catalyst show that the
by cutting a planar ZnO sheet [73–76] along the zigzag edge. The reaction proceeds via the formate pathway, thus avoiding CO inter-
ribbon is three ZnO units wide and the reactions are considered mediates. This finding is in agreement with studies using isotopi-
on the Zn terminated edge that resembles the Zn3O motif (see cally labeled CO2 species [79], showing that methanol is formed
Fig. 1). The reaction barriers compared in Fig. 5 b) are very similar solely from CO2, and is in contrast to theoretical results for ceria
on all systems with the overall barrier ranging from 1.30 eV (Zn7- based catalysts that favor the reverse water gas shift reaction path.
O3H) to 1.67 eV (Zn3O). The lowest barrier for formate hydrogena- This indicates that the preferred reaction mechanism of methanol
tion is found on Zn7O3H where the reaction product H2CO2 finds a formation can be changed by varying the oxide (e.g. from ZnOX to
low energy binding configuration thus lowering the barrier accord- CeOX) - a result that was already suspected in literature [2]. The
ing to the Bell-Evans-Polanyi principle [77,78] [see Fig. 5 c) for the difference between the oxides can be traced back to the altered
‘‘usual” binding motif on Zn7O3 vs. the special case of Zn7O3H]. This ability of the more positively charged metal to stabilize a delocal-
exceptionally strong bonding in turn increases the barrier for the ized formiate (Zn) or localized Cu-C(=O)-O-M structure (M = Zr,
next hydrogenation step and thus leads to a similar overall barrier Ce). Thus, the higher charged metals in higher oxidation states
T. Reichenbach et al. / Journal of Catalysis 360 (2018) 168–174 173
(III, IV) naturally seek for a localized structure as for example found [4] J. Burger, V. Papaioannou, S. Gopinath, G. Jackson, A. Galindo, C.S. Adjiman, A
hierarchical method to integrated solvent and process design of physical CO2
in Zr(IV)-carboxylates [80], while the less charged Zn(II) seeks a
absorption using the SAFT-c Mie approach, AIChE J. 61 (10) (2015) 3249–3269.
delocalized environment as found frequently in the paddle wheel [5] A. Goeppert, M. Czaun, J.-P. Jones, G.K.S. Prakash, G.A. Olah, Recycling of carbon
structures [M2(O2CR)4] of the 3d-series (M = Cr-Zn) [81]. dioxide to methanol and derived products – closing the loop, Chem. Soc. Rev.
It is relevant to highlight that the reaction mechanisms on other 43 (23) (2014) 7995–8048.
[6] M.D. Porosoff, B. Yan, J.G. Chen, Catalytic reduction of CO2 by H2 for synthesis
systems involving under-coordinated Zn sites such as the Zn- of CO, methanol and hydrocarbons: challenges and opportunities, Energy
terminated ZnO(0001) surface [82] and CuZn alloy stepped sur- Environ. Sci. 9 (1) (2016) 62–73.
faces [8] were also found to proceed through the formate path; [7] T. Lunkenbein, J. Schumann, M. Behrens, R. Schlögl, M.G. Willinger, Formation
of a ZnO overlayer in industrial Cu/ZnO/Al2O3 catalysts induced by strong
we have checked that the initial CO2 binding energy as well as metal–support interactions, Angew. Chem. Int. Ed. 54 (15) (2015) 4544–4548.
the rate determinating barriers in Fig. 5 (HCOO !H2CO2) are very [8] M. Behrens, F. Studt, I. Kasatkin, S. Kühl, M. Hävecker, F. Abild-Pedersen, S.
similar for various reduced ZnO structures supported on Cu(111) Zander, F. Girgsdies, P. Kurr, B.-L. Kniep, M. Tovar, R.W. Fischer, J.K. Nørskov, R.
Schlögl, The active site of methanol synthesis over Cu/ZnO/Al2O3 industrial
(see SM). Our results are robust against changes in the form and catalysts, Science 336 (6083) (2012) 893–897.
size of the reduced zinc-oxide. These similarities suggest that the [9] S. Kuld, M. Thorhauge, H. Falsig, C.F. Elkjær, S. Helveg, I. Chorkendorff, J.
reaction path found is generally the same on any under- Sehested, Quantifying the promotion of Cu catalysts by ZnO for methanol
synthesis, Science 352 (6288) (2016) 969–974.
coordinated Zn sites in ZnO/Cu catalysts, i.e. in the border region [10] S. Kuld, C. Conradsen, P.G. Moses, I. Chorkendorff, J. Sehested, Quantification of
of a real ZnO catalyst in contact with metallic copper. zinc atoms in a surface alloy on copper in an industrial-type methanol
The gain in entropy in the gas-phase causes all stable species synthesis catalyst, Angew. Chem. 126 (23) (2014) 6051–6055.
[11] M.M.-J. Li, Z. Zeng, F. Liao, X. Hong, S.C.E. Tsang, Enhanced CO2 hydrogenation
(CO2, CO, HCOOH, H2CO, H3COH, and H2O) to desorb. While this
to methanol over cuzn nanoalloy in ga modified Cu/ZnO catalysts, J. Catal. 343
effect helps preventing water poisoning, desorption simultane- (2016) 157–167.
ously drops the methanol formation rate as it competes with the [12] S.D. Senanayake, P.J. Ramirez, I. Waluyo, S. Kundu, K. Mudiyanselage, Z. Liu, Z.
reactions HCOOH!H2COOH and H2CO!H3CO. Decreasing the Liu, S. Axnanda, D.J. Stacchiola, J. Evans, J.A. Rodriguez, Hydrogenation of CO2
to methanol on CeOx/Cu(111) and ZnO/Cu(111) catalysts: Role of the metal–
reaction temperature helps in this respect, but also hinders over- oxide interface and importance of Ce3+ sites, J. Phys. Chem. C 120 (3) (2016)
coming the reaction barriers. 1778–1784.
A promising strategy to improve the catalyst could involve [13] S. Kattel, P.J. Ramírez, J.G. Chen, J.A. Rodriguez, P. Liu, Active sites for CO2
hydrogenation to methanol on Cu/ZnO catalysts, Science 355 (6331) (2017)
reducing the highest barrier in the main reaction pathway, i.e. 1296–1299.
the hydrogenation of formate. Therefore, substrate modifications [14] L. Martínez-Suárez, J. Frenzel, D. Marx, Cu/ZnO nanocatalysts in response to
that either bind formate more weakly or H2CO2 more strongly environmental conditions: surface morphology, electronic structure, redox
state and CO2 activation, Phys. Chem. Chem. Phys. 16 (47) (2014) 26119–
could improve catalytic performance. Such modifications consist 26136.
of e.g. alloying Cu with other metals or modifying the oxide via [15] L. Martínez-Suárez, N. Siemer, J. Frenzel, D. Marx, Reaction network of
the introduction of nitrides [83] or other metal oxides. A related methanol synthesis over Cu/ZnO Nanocatalysts, ACS Catal. 5 (7) (2015) 4201–
4218.
suggestion to tune activity by using different oxides was recently [16] K. Larmier, W.-C. Liao, S. Tada, E. Lam, R. Verel, A. Bansode, A. Urakawa, A.
given for oxide-supported PtCo catalysts [23]. However, as we have Comas-Vives, C. Copéret, CO2-to-methanol hydrogenation on zirconia-
just shown for Zn, Ce, and Zr, modifications of the metal oxide supported copper nanoparticles: Reaction intermediates and the role of the
metal–support interface, Angew. Chem. 129 (9) (2017) 2358–2363.
influences the bonding mechanism of CO2, which leads to varia-
[17] R. Naumann d’Alnoncourt, X. Xia, J. Strunk, E. Loffler, O. Hinrichsen, M. Muhler,
tions in the preferred reaction pathways. It is thus clear that the The influence of strongly reducing conditions on strong metal-support
synergies between different oxides in contact with a metal offer interactions in Cu/ZnO catalysts used for methanol synthesis, Phys. Chem.
rich possibilities, and that these can be explored by means of com- Chem. Phys. 8 (2006) 1525–1538.
[18] F. Studt, M. Behrens, E.L. Kunkes, N. Thomas, S. Zander, A. Tarasov, J.
putational models of inverse oxide-on-metal catalysts. Schumann, E. Frei, J.B. Varley, F. Abild-Pedersen, J.K. Nørskov, R. Schlögl, The
mechanism of CO and CO2 hydrogenation to methanol over cu-based catalysts,
Acknowledgements ChemCatChem 7 (7) (2015) 1105–1111.
[19] J. Graciani, K. Mudiyanselage, F. Xu, A.E. Baber, J. Evans, S.D. Senanayake, D.J.
Stacchiola, P. Liu, J. Hrbek, J.F. Sanz, J.A. Rodriguez, Highly active copper-ceria
Funding from sustainability center Freiburg within the HyCO2 and copper-ceria-titania catalysts for methanol synthesis from CO2, Science
project and through the project BioMethanol funded by the Federal 345 (6196) (2014) 546–550, arXiv,URL.
[20] J. Graciani, A.B. Vidal, J.A. Rodriguez, J.F. Sanz, Unraveling the nature of the
Ministry of Education and Research Germany is acknowledged. The oxide–metal interaction in ceria-based noble metal inverse catalysts, J. Phys.
authors are thankful for the computing time granted by the John Chem. C 118 (46) (2014) 26931–26938.
von Neumann Institute for Computing (NIC) and provided on the [21] J.A. Rodriguez, S. Ma, P. Liu, J. Hrbek, J. Evans, M. Pérez, Activity of CeOx and
TiOx nanoparticles grown on au(111) in the water-gas shift reaction, Science
supercomputer JURECA at Jülich Supercomputing Centre (JSC) as 318 (5857) (2007) 1757–1760.
well as the computational resource bwUni-Cluster funded by the [22] J.A. Rodriguez, J. Hrbek, Inverse oxide/metal catalysts: a versatile approach for
Ministry of Science, Research and the Arts Baden-Württemberg activity tests and mechanistic studies, Surf. Sci. 604 (3–4) (2010) 241–244.
[23] S. Kattel, W. Yu, X. Yang, B. Yan, Y. Huang, W. Wan, P. Liu, J.G. Chen, CO2
and the Universities of the State of Baden-Württemberg, Germany,
hydrogenation over oxide-supported PtCo catalysts: The role of the oxide
within the framework program bwHPC. support in determining the product selectivity, Angew. Chem. Int. Ed. 55 (28)
(2016) 7968–7973.
[24] S. Kattel, B. Yan, Y. Yang, J.G. Chen, P. Liu, Optimizing binding energies of key
Appendix A. Supplementary material intermediates for CO2 hydrogenation to methanol over oxide-supported
copper, J. Am. Chem. Soc. 138 (2016) 12440–12450.
Supplementary data associated with this article can be found, in [25] E.L. Kunkes, F. Studt, F. Abild-Pedersen, R. Schlögl, M. Behrens, Hydrogenation
of CO2 to methanol and CO on Cu/ZnO/Al2O3: Is there a common intermediate
the online version, at https://doi.org/10.1016/j.jcat.2018.01.035. or not?, J Catal. 328 (2015) 43–48.
[26] Y. Yang, J. Evans, J.A. Rodriguez, M.G. White, P. Liu, Fundamental studies of
References methanol synthesis from CO2 hydrogenation on Cu(111), cu clusters, and Cu/
ZnO(0001), Phys. Chem. Chem. Phys. 12 (2010) 9909–9917.
[27] O. Martin, C. Mondelli, A. Cervellino, D. Ferri, D. Curulla-Ferré, J. Pérez-Ramírez,
[1] J. Hansen, L. Nazarenko, R. Ruedy, M. Sato, J. Willis, A. Del Genio, D. Koch, A.e.
Operando synchrotron x-ray powder diffraction and modulated-excitation
Lads, K. Lo, S. Menon, T. Novakov, J. Perlwitz, G. Russell, G.A. Schmidt, N.
infrared spectroscopy elucidate the CO2 promotion on a commercial methanol
Tausnev, Earth’s energy imbalance: confirmation and implications, Science 308
synthesis catalyst, Angew. Chem. 128 (37) (2016) 11197–11202.
(5727) (2005) 1431–1435.
[28] T. Fujitani, I. Nakamura, T. Uchijima, J. Nakamura, The kinetics and mechanism
[2] M. Behrens, Heterogeneous catalysis of CO2 conversion to methanol on copper
of methanol synthesis by hydrogenation of CO2 over a zn-deposited cu(111)
surfaces, Angew. Chem. Int. Ed. 53 (2014) 12022–12024.
surface, Surf. Sci. 383 (2–3) (1997) 285–298.
[3] E. Frei, A. Schaadt, T. Ludwig, H. Hillebrecht, I. Krossing, The influence of the
[29] F. Claeyssens, J.N. Harvey, F.R. Manby, R.A. Mata, A.J. Mulholland, K.E.
precipitation/ageing temperature on a Cu/ZnO/ZrO2 catalyst for methanol
Ranaghan, M. Schütz, S. Thiel, W. Thiel, H.-J. Werner, High-accuracy
synthesis from H2 and CO2, ChemCatChem 6 (6) (2014) 1721–1730.
174 T. Reichenbach et al. / Journal of Catalysis 360 (2018) 168–174
computation of reaction barriers in enzymes, Angew. Chem. Int. Ed. 45 (41) [54] J.P. Perdew, S. Kurth, A Primer in Density Functional Theory, Springer Berlin
(2006) 6856–6859. Heidelberg, Berlin, Heidelberg, 2003, Ch. Density Functionals for Non-
[30] P. Jurečka, J. Šponer, J. Černý, P. Hobza, Benchmark database of accurate (MP2 relativistic Coulomb Systems in the New Century, pp. 1–55..
and CCSD(T) complete basis set limit) interaction energies of small model [55] A.J. Cohen, P. Mori-Sanchez, W. Yang, Challenges for density functional theory,
complexes, DNA base pairs, and amino acid pairs, Phys. Chem. Chem. Phys. 8 Chem. Rev. 112 (1) (2012) 289–320.
(17) (2006) 1985–1993. [56] K. Yang, J. Zheng, Y. Zhao, D.G. Truhlar, Tests of the rpbe, revpbe, s-hcthhyb,
[31] M. Pitoňák, P. Neogrády, J. Černý, S. Grimme, P. Hobza, Scaled MP3 non- xb97x-d, and mohlyp density functional approximations and 29 others
covalent interaction energies agree closely with accurate CCSD(T) benchmark against representative databases for diverse bond energies and barrier heights
data, ChemPhysChem 10 (1) (2009) 282–289. in catalysis, J. Chem. Phys. 132 (16) (2010) 164117.
[32] G.H. Booth, A. Grüneis, G. Kresse, A. Alavi, Towards an exact description of [57] N. Ashcroft, N. Mermin, Solid State Physics, Science: Physics, Saunders College,
electronic wavefunctions in real solids, Nature 493 (7432) (2013) 365–370. 1976.
[33] X. Xu, W. Zhang, M. Tang, D.G. Truhlar, Do practical standard coupled cluster [58] H. Jónsson, G. Mills, K.W. Jacobsen, Nudged elastic band method for finding
calculations agree better than Kohn–Sham calculations with currently minimum energy paths of transitions, in: B.J. Berne, G. Ciccotti, D.F. Coker
available functionals when compared to the best available experimental (Eds.), Classical and Quantum Dynamics in Condensed Phase Simulations,
data for dissociation energies of bonds to 3d transition metals?, J Chem. Theor. 1998, pp. 385–404.
Comp. 11 (5) (2015) 2036–2052. [59] G. Henkelman, B.P. Uberuaga, H. Jónsson, A climbing image nudged elastic
[34] R. Würdemann, H.H. Kristoffersen, M. Moseler, M. Walter, Density functional band method for finding saddle points and minimum energy paths, J. Chem.
theory and chromium: Insights from the dimers, J. Chem. Phys. 142 (12) Phys. 113 (22) (2000) 9901–9904.
(2015) 124316. [60] S. Mallikarjun Sharada, T. Bligaard, A.C. Luntz, G.-J. Kroes, J.K. Nørskov, Sbh10:
[35] J.J. Mortensen, L.B. Hansen, K.W. Jacobsen, Real-space grid implementation of A benchmark database of barrier heights on transition metal surfaces, J. Phys.
the projector augmented wave method, Phys. Rev. B 71 (3) (2005) 035109. Chem. C 121 (36) (2017) 19807–19815.
[36] J. Enkovaara, C. Rostgaard, J.J. Mortensen, J. Chen, M. Dulak, L. Ferrighi, J. [61] K. Duanmu, D.G. Truhlar, Validation of density functionals for adsorption
Gavnholt, C. Glinsvad, V. Haikola, H.A. Hansen, H.H. Kristoffersen, M. Kuisma, energies on transition metal surfaces, J. Chem. Theory Comput. 13 (2) (2017)
A.H. Larsen, L. Lehtovaara, M. Ljungberg, O. Lopez-Acevedo, P.G. Moses, J. 835–842.
Ojanen, T. Olsen, V. Petzold, N.A. Romero, J. Stausholm-Møller, M. Strange, G.A. [62] J. Wellendorff, T.L. Silbaugh, D. Garcia-Pintos, J.K. Nørskov, T. Bligaard, F. Studt,
Tritsaris, M. Vanin, M. Walter, B. Hammer, H. Häkkinen, G.K.H. Madsen, R.M. C.T. Campbell, A benchmark database for adsorption bond energies to
Nieminen, J.K. Nørskov, M. Puska, T.T. Rantala, J. Schiøtz, K.S. Thygesen, K.W. transition metal surfaces and comparison to selected dft functionals, Surf.
Jacobsen, Electronic structure calculations with gpaw: a real-space Sci. 640 (Supplement C) (2015) 36–44, reactivity Concepts at Surfaces:
implementation of the projector augmented-wave method, J. Phys.: Cond. Coupling Theory with Experiment.
Mat. 22 (25) (2010) 253202. [63] D. Cheng, F.R. Negreiros, E. Apra, A. Fortunelli, Computational approaches to
[37] J.P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made the chemical conversion of carbon dioxide, ChemSusChem 6 (6) (2013) 944–
simple, Phys. Rev. Lett. 77 (1996) 3865–3868. 965.
[38] J. Wellendorff, K.T. Lundgaard, A. Møgelhøj, V. Petzold, D.D. Landis, J.K. [64] Y. Li, S.H. Chan, Q. Sun, Heterogeneous catalytic conversion of CO2: a
Nørskov, T. Bligaard, K.W. Jacobsen, Density functionals for surface science: comprehensive theoretical review, Nanoscale 7 (2015) 8663–8683.
Exchange-correlation model development with bayesian error estimation, [65] L.H. Gade, Highly polar metal–metal bonds in early-late-heterodimetallic
Phys. Rev. B 85 (2012) 235149. complexes, Angew. Chem. Int. Ed. 39 (15) (2000) 2658–2678.
[39] L.B. Vilhelmsen, B. Hammer, A genetic algorithm for first principles global [66] J.P. Krogman, B.M. Foxman, C.M. Thomas, Activation of CO2 by a
structure optimization of supported nano structures, J. Chem. Phys. 141 (4) Heterobimetallic Zr/Co Complex, J. Am. Chem. Soc. 133 (37) (2011) 14582–
(2014) 044711. 14585.
[40] A. Larsen, J. Mortensen, J. Blomqvist, I. Castelli, R. Christensen, M. Dulak, J. Friis, [67] K.S. Thygesen, L.B. Hansen, K.W. Jacobsen, Partly occupied Wannier functions:
M. Groves, B. Hammer, C. Hargus, E. Hermes, P. Jennings, P. Jensen, J. Kermode, construction and applications, Phys. Rev. B 72 (12) (2005) 125119.
J. Kitchin, E. Kolsbjerg, J. Kubal, K. Kaasbjerg, S. Lysgaard, J. Maronsson, T. [68] N. Marzari, A.A. Mostofi, J.R. Yates, I. Souza, D. Vanderbilt, Maximally localized
Maxson, T. Olsen, L. Pastewka, A. Peterson, C. Rostgaard, J. Schiøtz, O. Schütt, Wannier functions: theory and applications, Rev. Mod. Phys. 84 (4) (2012)
M. Strange, K. Thygesen, T. Vegge, L. Vilhelmsen, M. Walter, Z. Zeng, K.W. 1419–1475.
Jacobsen, The atomic simulation environment – a python library for working [69] R.F.W. Bader, Atoms in Molecules: A Quantum Theory, 1st ed., Clarendon Press,
with atoms, J. Phys.: Cond. Mat. 29 (2017) 273002. Oxford England: New York, 1994.
[41] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 85th ed., CRC Press, [70] W. Tang, E. Sanville, G. Henkelman, A grid-based Bader analysis algorithm
Boca Raton, 2004. without lattice bias, J. Phys.: Cond. Mat. 21 (8) (2009) 084204.
[42] C. Riplinger, F. Neese, An efficient and near linear scaling pair natural orbital [71] L. Kristinsdóttir, E. Skúlason, A systematic DFT study of hydrogen diffusion on
based local coupled cluster method, J. Chem. Phys. 138 (2013) 034106. transition metal surfaces, Surf. Sci. 606 (17) (2012) 1400–1404.
[43] C. Riplinger, B. Sandhoefer, A. Hansen, F. Neese, Natural triple excitations in [72] C. Liu, B. Yang, E. Tyo, S. Seifert, J. DeBartolo, B. von Issendorff, P. Zapol, S.
local coupled cluster calculations with pair natural orbitals, J. Chem. Phys. 139 Vajda, L.A. Curtiss, Carbon dioxide conversion to methanol over size-selected
(2013) 134101. cu4 clusters at low pressures, J. Am. Chem. Soc. 137 (27) (2015) 8676–8679.
[44] P.J. Malinowski, D. Himmel, I. Krossing, Coordination chemistry of diiodine and [73] C. Tusche, H.L. Meyerheim, J. Kirschner, Observation of depolarized ZnO(0001)
implications for the oxidation capacity of the synergistic Ag+/X2 (X=Cl, Br, I) monolayers: Formation of unreconstructed planar sheets, Phys. Rev. Lett. 99
System, Angew. Chem. Int. Ed. 55 (32) (2016) 9262–9266. (2007) 026102.
[45] H. Böhrer, N. Trapp, D. Himmel, M. Schleep, I. Krossing, From unsuccessful H2- [74] M. Topsakal, S. Cahangirov, E. Bekaroglu, S. Ciraci, First-principles study of zinc
activation with FLPs containing B(Ohfip)3 to a systematic evaluation of the oxide honeycomb structures, Phys. Rev. B 80 (2009) 235119.
Lewis acidity of 33 Lewis acids based on fluoride, chloride, hydride and methyl [75] H. Si, L.J. Peng, J.R. Morris, B.C. Pan, Theoretical prediction of hydrogen storage
ion affinities, Dalton Trans. 44 (16) (2015) 7489–7499. on ZnO sheet, J. Chem. Phys. C 115 (18) (2011) 9053–9058.
[46] F. Scholz, D. Himmel, L. Eisele, W. Unkrig, A. Martens, P. Schlüter, I. Krossing, [76] H. Guo, Y. Zhao, N. Lu, E. Kan, X.C. Zeng, X. Wu, J. Yang, Tunable magnetism in a
The acidity of the HBr/AlBr3 system: stabilization of crystalline protonated nonmetal-substituted ZnO monolayer: a first-principles study, J. Chem. Phys. C
arenes and their acidity in bromoaluminate ionic liquids, Chem.–Eur. J. 21 (20) 116 (20) (2012) 11336–11342.
(2015) 7489–7502. [77] R.P. Bell, The Theory of Reactions Involving Proton Transfers, Proc. R. Soc. Lond.
[47] D. Himmel, S.K. Goll, F. Scholz, V. Radtke, I. Leito, I. Krossing, Absolute Brønsted A Math. Phys. Eng. Sci. 154 (882) (1936) 414–429.
Acidities and pH scales in ionic liquids, ChemPhysChem 16 (7) (2015) 1428– [78] M.G. Evans, M. Polanyi, Inertia and driving force of chemical reactions, Trans.
1439. Faraday Soc. 34 (0) (1938) 11–24.
[48] D. Himmel, S.K. Goll, I. Leito, I. Krossing, Bulk Gas-Phase Acidity, Chem.–Eur. J. [79] G.C. Chinchen, P.J. Denny, D.G. Parker, M.S. Spencer, D.A. Whan, Mechanism of
18 (30) (2012) 9333–9340. methanol synthesis from CO2/CO/H2 mixtures over copper/zinc oxide/
[49] D. Himmel, S.K. Goll, I. Leito, I. Krossing, Anchor points for the unified Brønsted alumina catalysts: use of 14C-labelled reactants, Appl. Catal. 30 (2) (1987)
acidity scale: the rCCC Model for the calculation of standard Gibbs Energies of 333–338.
proton solvation in eleven representative liquid media, Chem.–Eur. J. 17 (21) [80] G. Yalcin, A. Kayan, Synthesis and characterization of Zr, Ti, Al-phthalate and
(2011) 5808–5826. pyridine-2-carboxylate compounds and their use in ring opening
[50] F. Studt, F. Abild-Pedersen, J.B. Varley, J.K. Nørskov, CO and CO2 hydrogenation polymerization, Appl. Catal. A: General 433-434 (2012) 223–228.
to methanol calculated using the beef-vdw functional, Catal. Lett. 143 (1) [81] S. Vagin, A.K. Ott, B. Rieger, Paddle-wheel zinc carboxylate clusters as building
(2013) 71–73. units for metal-organic frameworks, Chem. Ing. Techn. 79 (6) (2007) 767–780.
[51] R. Christensen, H.A. Hansen, T. Vegge, Identifying systematic DFT errors in [82] A.J. Medford, J. Sehested, J. Rossmeisl, I. Chorkendorff, F. Studt, J.K. Nørskov, P.
catalytic reactions, Catal. Sci. Techn. 5 (11) (2015) 4946–4949. G. Moses, Thermochemistry and micro-kinetic analysis of methanol synthesis
[52] A.A. Peterson, F. Abild-Pedersen, F. Studt, J. Rossmeisl, J.K. Norskov, How on ZnO (0 0 0 1), J. Catal. 309 (2014) 397–407.
copper catalyzes the electroreduction of carbon dioxide into hydrocarbon [83] R. Shanmugam, A. Thamaraichelvan, B. Viswanathan, Methanol formation by
fuels, Energy Environ. Sci. 3 (2010) 1311–1315. catalytic hydrogenation of CO2 on a nitrogen doped zinc oxide surface: an
[53] M. Gajdoš, J. Hafner, CO adsorption on cu(1 1 1) and cu(0 0 1) surfaces: evaluative study on the mechanistic pathway by density functional theory,
Improving site preference in DFT calculations, Surf. Sci. 590 (2–3) (2005) 117– RSC Adv. 5 (2015) 60524–60533.
126.