CO2-to-Methanol Hydrogenation on Zirconia-Supported Copper
Nanoparticles: Reaction Intermediates and the Role of the Metal–
Support Interface
Kim Larmier,[a] Wei-Chih Liao,[a] Shohei Tada, [a] Erwin Lam,[a] René Vérel,[a] Atul Bansode,[b] Atsushi
Urakawa,[b] Aleix Comas-Vives*[a]and Christophe Copéret*[a]
Abstract: Methanol synthesis via CO2 hydrogenation is a key step in methanol-based economy. This reaction is catalyzed by supported copper
nanoparticles and displays strong support or promoter effects. Zirconia is known to enhance both the methanol production rate and the
selectivity. Nevertheless, the origin of this observation and the reaction mechanisms associated with the conversion of CO 2 to methanol still
remain unknown. Here, we present a mechanistic study of the hydrogenation of CO 2 on Cu/ZrO2. Using kinetics, in situ IR and NMR
spectroscopies and isotopic labeling strategies, we examined the surface intermediates during CO 2 hydrogenation at different pressures.
Combined with DFT calculations, we show that formate species is the reaction intermediate and that the zirconia/copper interface is a key for
its conversion to methanol.

The catalytic hydrogenation of carbon dioxide to methanol is a key process in the sustainable methanol-based economy.[1] While
copper-based catalysts are highly active for this transformation, [2] their activity and selectivity strongly depend on the support and/or
the promoters. Understanding the copper-support interaction – its effect on the activity and product selectivity – has been a very
intensive field of research over the last decade. While the reaction mechanisms and the nature of the active sites on Cu/ZnO systems
have been extensively investigated,[3] copper supported on zirconia and related materials also exhibits high activity and selectivity in
CO2 hydrogenation to methanol (Eq. 1) by minimizing the formation of CO, a byproduct often resulting from the competitive reverse
water-gas shift reaction (Eq. 2).[4]
CO2 + 3H2 = CH3OH + H2O
CO2 + H2 = CO + H2O

∆rH° (500 K) = –62 kJ.mol-1 (1)
∆rH° (500 K) = +40 kJ.mol-1 (2)

Although the copper-zirconia interface was proposed to play a key role in the selective formation of methanol,[4c, 4e-g] the active site and
the reaction mechanism, including the role of the interface on methanol selectivity, are still not understood. In fact, mechanistic
investigations using Diffuse Reflectance IR Fourier Transform spectroscopy (DRIFTS) led to opposite conclusions: formate is an
intermediate in methanol formation[4c, 4d] vs. CO2 is first reduced to CO that is in turn hydrogenated to methanol through a carboxyl
intermediate.[4f]
Herein, by using a combined experimental and computational approach on realistic models, we investigated the reaction mechanism
of CO2 hydrogenation to methanol on a Cu/ZrO2 catalyst. Kinetic investigation, in situ and ex situ spectroscopies – FTIR and NMR –
together with isotopic labeling and computational modelling showed that methanol is a primary product formed by the hydrogenation of
formate as a reaction intermediate.
First, narrowly dispersed copper nanoparticles supported on monoclinic zirconia were prepared by a molecular approach. [5] Grafting of
[Cu(OtBu)]4 on the surface hydroxyl groups of the support (Figure S1-S2, Scheme S1) followed by a treatment under H2 at 500 °C for
5 h[6] yields small and narrowly distributed Cu nanoparticles: 2.2 ± 0.5 nm with 0.8 wt% Cu loading (Figure S3a). We also prepared via
the same method a Cu/SiO2 catalyst as a prototype of pure copper particles in order to probe their specific reactivity. This sample
contains Cu particles with a size distribution of 2.1 ± 0.5 nm for 2.3 % of Cu loading (Figure S3b). Their catalytic activities were
measured in a fixed-bed flow reactor at 230 °C and 25 bars (H2/CO2 molar ratio 3:1) under steady-state conditions (Figure S4). For
each sample, the contact time was increased by decreasing the volumetric flow rate Q of the feed gas. Figure 1a shows the evolution
of the formation rates of methanol and CO as a function of the contact time. For both catalysts, the rates strongly depend on the contact
time. Extrapolation of the initial rates to zero conversion (zero contact time) for the two catalysts (Figure 1b) clearly show a strictly
positive initial rate of formation for both CO and methanol, which indicates that they are both primary products. Thus, the intermediacy
of CO in the formation of methanol is unlikely. Second, while the
rate of CO formation is of the same order of magnitude on both
catalysts, the rate of methanol formation is dramatically increased
[a]
Dr K. Larmier, Mr. W.-C. Liao, Dr. S. Tada, Mr. E. Lam, Dr. R. Vérel,
on Cu/ZrO2. Thus, CO formation likely occurs on the Cu surface,
Dr. A. Comas-Vives, Prof. C. Copéret
Department of Chemistry and Applied Biosciences
while methanol is formed at a much higher rate when both copper
ETH Zürich
and zirconia are present. As a result, Cu/ZrO2 displays much
Vladimir-Prelog Weg 1-5, CH-8093 Zürich, Switzerland
higher initial activity and methanol selectivity (15 μmol.s-1.gCu-1,
E-mail: comas@inorg.chem.ethz.ch
75 %) than Cu/SiO2 (2.6 μmol.s-1.gCu-1, 50 %). On Cu/SiO2, the
E-mail: ccoperet@inorg.chem.ethz.ch
[b]
Dr. A. Bansode, Prof. A. Urakawa
rate of CO formation increases with contact time while the
Institute of Chemical Research of Catalonia (ICIQ),
opposite is observed for methanol, which is likely due to partial
The Barcelona Institute of Science and Technology,
methanol decomposition into CO. On Cu/ZrO2, similar trends are
Av. Països Catalans 16, 43007 Tarragona, Spain
observed for CO formation. However, the decrease in methanol
E-mail: aurakawa@iciq.es
formation rate with contact time is much sharper, which is likely
This work was presented at the CO2 Catalysis Conference , April
due to inhibition of methanol formation by reaction products, such
2016, Albufeira, Portugal
as water or methanol itself. Both molecules are basic and strongly
bind ZrO2, suggesting that the active sites may involve Lewis
Supporting information for this article is given via a link at the end of
the document.
acidic Zr atoms. This inhibition is essentially reversible as going

back to a low contact time mostly restores the activity. A slight deactivation occurs over 40 h on stream (empty data point in Fig. 1a).
Overall, the methanol selectivity strongly decreases with contact time and conversion on both catalysts (Figure S4c-d).

MeOH --Cu/ZrO2
MeOH Cu/ZrO2
MeOH Cu/SiO 2
MeOH --Cu/SiO2

Formation rates
(μmol.s-1.gcu-1)

15

CO - Cu/ZrO2
CO- Cu/ZrO2
CO- Cu/SiO2
CO - Cu/SiO2

10

5

75%

(b) 15
Initial formation rates
(μmol.s-1.g cu-1)

(a)

CO
MeOH

10

5

49%

0

0
0

1

2

Contact time mcat /F (s.g.mL-1)
Contact time mcat/Qin (s.g.mL-1)

3

Cu/SiO2
Cu/SiO2

Cu/ZrO2
Cu/ZrO 2

Figure 1. (a) Evolution of the rate of formation of CO and methanol with contact time on Cu/ZrO 2 and Cu/SiO2 measured in a flow reactor at 230 °C and 25 bars
(H2/CO2 = 3:1). The empty points show the activity when the first data point was repeated after 40 h of reaction, showing slow deactivation. CO2 conversion was
kept between 0.5 and 6 %. (b) Extrapolated rates of formation at zero conversion. The selectivity to methanol is indicated.

In order to obtain information about surface reaction intermediates, CO2 hydrogenation on Cu/ZrO2 was investigated by in situ DRIFTS
at 230 °C under varying pressures (1-20 bars). The IR spectrum of the pristine catalyst shows only OH stretching frequencies at 3774
and 3670 cm-1 (Figure S5a). After contacting pre-reduced Cu/ZrO2 with H2/CO2 mixture (3:1) at 230 °C and 1 bar, features appeared in
the -CH (Figure S6a) and -CO regions (Figure S6b). The main broad -OCO features at 1593 cm-1 can be attributed to carbonate or
bicarbonate (CO3* or (CO3H*),[7] and the bands at 2978, 2878, 2736, 1567 and 1387 cm-1 are characteristic of formate species adsorbed
on zirconia (HCOO_ZrO2, ESI Section 5).[4d, 7-8] The presence of formate on copper cannot be excluded as a characteristic band at
1357 cm-1 is also observed.[4d] Increasing the reaction pressure to 5 bars (Figure S6a-b) or higher pressures resulted in the appearance
of new features at 2942, 2828 cm-1 and the characteristic C-O stretching frequencies at 1154 and 1049 cm-1 that were attributed to
methoxy adsorbed on zirconia.[4f, 7] Interestingly, when pure ZrO2 was used, only carbonate, bicarbonate and formate species were
observed (Figure S6). This suggests that the presence of Cu on ZrO2 is responsible for the formation of methoxy species.
We subsequently investigated the nature of the reaction intermediates by solid-state NMR spectroscopy. Note that NMR will preferably
provide information about species adsorbed on zirconia, as adsorbed species on metallic copper will suffer from signal broadening and
disappearance (ESI, Section 5). Cu/ZrO2 was contacted with a H2/13CO2 mixture (3:1) at 230 °C for 12 h in a high-pressure glass reactor
at 1 and 5 bars, respectively. After cooling down to room temperature, the gas phase was evacuated and the solid was analyzed by
solid-state NMR. The 1H-13C heteronuclear correlation (HETCOR) spectrum of the sample after reaction at 1 bar is shown in Figure 2a.
The correlation of the 13C NMR chemical shift at 168 ppm to the 1H NMR signal at 8.4 ppm is consistent with the presence of formate
species; the line broadening in the 1H dimension suggests the presence of formate in different chemical environments. After reaction
at 5 bars (Figure 2b), an additional correlation is observed at δ(13C) = 51 ppm and δ(1H) = 4.0 ppm consistent with the formation of
surface methoxy species. The NMR data are in agreement with what is observed by in situ DRIFTS. However, the absence of
carbonates in both, ex situ NMR or IR experiments (Figure S5b-c) shows that these species are weakly adsorbed.

Figure 2. Ex situ MAS-NMR 1H-13C HETCOR spectra of Cu/ZrO2 reacted with H2/13CO2 (3:1) at 230 °C for 12 h at (a) 1 bar or (b) 5 bars. (c) Ex situ MAS-NMR 1H13
C HETCOR spectrum of Cu/ZrO2 after two step hydrogenation of 13CO2 : (1) hydrogenation at 1bar, 230 °C, 12h (2) deuteration at 5 bars, 230 °C, 12h . For
HETCOR experiments, ramp cross polarization (1H-13C) was used with contact time of 0.5 ms. The recycle delay was 1 s. External projections of the 1D 13C and 1H
spectra are applied in all spectra.

On the contrary, the formate and methoxy species are strongly adsorbed and remain on the surface even after evacuation under high
vacuum. Finally, when the similar experiments were performed on pure ZrO 2, formate species were observed (Figure S16) but no
methoxy, as previously observed by in situ DRIFTS. On Cu/SiO2, IR shows (Figure S13) the formation of formate identified by bands
at 2937, 2858, and 1544 cm-1 associated to a very broad and weak feature at 168 ppm in CP-MAS, and no methoxy could be observed
(Figure S14-S15). Thus, copper particles by themselves or the pure support are not able to generate significant amounts of methoxy
from formate under these conditions (230 °C; 5 bars).

Scheme 1. Reaction scheme derived from the spectroscopic measurements.

In order to evaluate whether or not formate is a reaction intermediate towards methanol synthesis, additional isotopic labeling
experiments were carried out. First, we selectively prepared 13C labeled formate adsorbed on the surface (H13COO*) by reacting
Cu/ZrO2 with a H2/13CO2 mixture (3:1) at 1 bar and 230 °C (Figure 2a). In a second step, following cooling down to room temperature
and evacuation of the gas phase (10-4 mbar, 2 h), the sample was treated with 5 bars of D 2 at 230 °C for 12 h. MAS-NMR and IR
spectra of the resulting solid were recorded (Figure 2c and S5d). 1H-13C CP-MAS NMR spectrum shows two signals at 168 and 51 ppm,
attributed to formate and methoxy species, respectively. In this experiment, the formate initially present on the surface after the first
step (treatment by 13CO2 and H2) is the only source of 13C and 1H, so that the methoxy surface species must be formed from deuteration
of the initial formate species, supporting its involvement as a reaction intermediate to methanol formation. In order to further characterize
the nature of the intermediates adsorbed on the surface, and the isotopic source of the methoxy species, the sample was extracted
and analyzed by solution NMR; it was first divided in two fractions prior to extraction. The first fraction was extracted with D2O and a 1H
solution NMR spectrum was recorded (Figure S18), and the second fraction was extracted with H 2O and a 2D solution NMR spectrum
was recorded (Figure S19). The 1H spectrum shows a doublet of pentet centered at 3.12 ppm, with coupling constants J(1H-13C) =
141 Hz and J(1H-2D) of 1.7 Hz, consistent with the formation of 13CHD2OD. The 2D spectrum shows a doublet centered at 3.15 ppm
(J(2D-13C) = 21 Hz), but did not show additional 1H-2D coupling, thus showing that the most abundant species correspond to fully
deuterated methanol 13CD3OH. This is also consistent with the absence of formate signal in the 1H-13C HETCOR spectrum (Figure 2e).
Taken together, the results suggest that the formate can exchange its hydrogen with deuterium prior to its subsequent deuteration,
indicating that the formation of methoxy species is slower than the H/D exchange of formate as shown in Scheme 1. IR spectroscopy
also reveals a high degree of H/D exchange upon contact of the sample with D2 (Figure S5d). Nonetheless, quantitative 13C direct
excitation NMR spectrum (Figure S17) showed that a significant proportion of formate (deuterated or not) was converted into methoxy
(about 60 %). Furthermore, the 2D spectrum in solution allowed a rough estimation of the number of methanol adsorbed on the surface
prior desorption to about 0.04 par nm-2, which may be considered as an indication of the number of active sites. This number is close
to the density of particles that can be estimated (0.01 per nm-2) indicating a rather small number of active sites per copper particle
(details in ESI). The NMR spectroscopic investigation shows that methoxy (methanol) is formed from formate species adsorbed on
zirconia, as proposed earlier.[4d] It also highlights the determining effect of pressure in the formation of the intermediates. We also show
that both copper and zirconia are required for the reaction under such conditions. Thus, it can be proposed that the reaction takes place
at the interface between zirconia and copper particles, as suggested in previous reports. [4b, 4e].
In order to get molecular insights into the reaction mechanisms and possible involvement of this interface, we turned to DFT
calculations. A model for the supported copper particles on ZrO2 and the potential interfacial active sites was constructed by depositing
̅
a Cu38 particle (ø = 0.8 nm, truncated octahedron from fcc structure) on a m-ZrO2 (111) slab, which is the main termination exposed by
monoclinic zirconia (Figure S20).[9] No significant deformation of the cluster was observed, consistent with the small change in the
cohesive energy of the Cu nanoparticle upon adsorption (–266 and –269 kJ.mol-1 prior and after adsorption, respectively). The ZrO2
slab model and a pristine (111) surface of fcc copper were also used for comparison. The adsorption modes of H 2 and CO2 were
assessed on these three model materials. In line with previous findings, the dissociative adsorption of H 2 is endoenergetic on the
zirconia surface, either through heterolytic (ΔrE = +57 kJ.mol-1) or homolytic cleavage (ΔrE = +241 kJ.mol-1),[9] while the dissociation of
H2 is exoenergetic by –40 kJ.mol-1 on the Cu (111) facet.[10] On the supported nanoparticle, the dissociation is slightly more favorable
than on the extended surface (ΔrE = –50 kJ.mol-1). CO2 virtually does not bind the Cu (111) surface (–2 kJ.mol-1, Figure 3a). On ZrO2,
CO2 adsorbs as a carbonate, or bicarbonate if a hydroxyl group is available, with adsorption energies of –65 and –70 kJ.mol-1,
respectively (Figure 3b and S21). However, we found a much more favorable adsorption mode of CO 2 at the interface between copper
and zirconia (ΔrE = –179 kJ.mol-1, Figure 3c), where CO2 adopts a bent structure with the carbon atom bound to the copper particle
surface, and the two oxygen atoms interact with Zr 4+ Lewis acidic centers of the zirconia surface. A similar adsorption mode has been
found for the adsorption of CO2 on Ni/Al2O3 interface.[11] Bader charge analysis of this intermediate revealed a transfer of electron
density from the copper particle to the CO2 molecule (-1.1 compared to about 0 for CO2 on Cu(111) and carbonates on ZrO2), which is
thus negatively charged and partially reduced. The positive charge is delocalized on the copper particle. This adsorption mode of CO2
at the Cu/ZrO2 interface is therefore a very good candidate for further reduction by H 2, and was used as a starting point to calculate
hydrogenation pathways.
We first examined the transformation of CO2 into three intermediates commonly proposed:[3a, 12] i) direct decomposition of CO2 into CO
(first step of the reverse water-gas-shift reaction, Figure 3d), ii) carboxyl intermediate (COOH*, Figure 3e) and iii) formate intermediate
(HCOO*, Figure 3f). The activation barriers to form these intermediates from CO 2 chemisorbed at the interface and H2 dissociated on
the particle are given in Figures 3d-f. Formation of the formate intermediate is the lowest energy pathway from all those evaluated, with
an activation barrier of only 74 kJ.mol-1, compared to 119 and 193 kJ.mol-1 for the formation of CO and COOH*, respectively. It is also
the most thermodynamically favorable one (–67 kJ.mol-1) in comparison with the formation of CO and COOH* which are endoenergetic
by 54 and 67 kJ.mol-1, respectively. The whole pathway leading from CO2 to methanol was thus calculated through the most favorable
formate intermediate. The resulting energy and Gibbs free energy diagrams are depicted in Figure 4, along with the main intermediates.
Note that the highest point in the free energy diagram (19 kJ.mol-1) lies lower than the transition states for the initial activation of CO2
to CO* (50 kJ.mol-1) or COOH* (124 kJ.mol-1), confirming the relevance of the formate route as the most favorable pathway. From the
formate, adsorbed on ZrO2 in close vicinity to the Cu particle, transfer of an additional hydrogen atom from the Cu particle can generate
an acetal-like species H2C(O)2*, where both oxygen atoms are bound to two Zr4+ sites, with a low free energy barrier (71 kJ.mol-1). The
further hydrogenation of the acetal is the most energetically demanding step. A hydrogen atom is first transferred to one of the oxygen
atoms to form H2COOH* (ΔrG‡ = 105 kJ.mol-1). Note however that the relative energy barrier for this step is lower than the formation of
CO* and COOH* intermediates. This high-energy intermediate is easily converted by the transfer of an additional hydrogen atom (ΔrG‡

= 52 kJ.mol-1). In a concerted, SN2-like step, the third C-H bond is formed while the C-OH bond is broken, leading to methoxy and
hydroxyl groups adsorbed on the formally Zr 4+ sites. Both can then further undergo protonation by hydrogen-transfer from the copper
particle. Methanol, and water are thus formed (ΔrG‡ = 52 and 102 kJ.mol-1, respectively), and desorb with a desorption energy of –84
and –87 kJ.mol-1, respectively.
0.0

(a)

(c)

(b)
-0.1

CO2 on Cu(111)

-1.1

CO2 on ZrO2(-111)

(d)

CO2 on Cu/ZrO2

(f)

(e)

CO* + O*
ΔrE‡ = 119
∆rE = 54

Caboxylate COOH*
ΔrE‡ = 193
∆rE = 67

Formate HCOO*
ΔrE‡ = 74
∆rE = -67

̅
Figure 3. (a)-(c) Optimized structures of CO2 adsorbed (a) on Cu (111) (b) on m-ZrO2 (111) (c) at the interface between copper and zirconia. Bader charge on the
CO2 molecule are indicated. (d)-(f) Structures of the possible intermediates at the interface between copper and zirconia (d) CO* and O* (e) COOH* (f) formate.
Activation barriers and formation enthalpy from adsorbed CO2 + H2 are given in kJ.mol-1. Cyan: zirconium, red: oxygen, orange: copper, Grey: carbon, white:
hydrogen.

100

G or E (kJ/mol)
‡

‡

0

‡

‡

G

‡

Formate
HCO2* + H*

‡

H3COH* +H2O*

H3COH* +HO* + H*

H3CO* +HO* + 2H*

H3CO* +HO*

H2CO2 H* + H*

H2CO2* + 2H*

H2CO2*

HCO2* + H*

-400

CO2*

-300

CO2* + 2H*

-200

H2O* + H3COH(g)
H2O(g) + H3COH(g)

E

-100

Acetal
H2CO2* + 2H*

Methoxy
H3CO* +HO*

Figure 4. Energy and Standard Gibbs free energy pathway at 473 K calculated for the hydrogenation of CO 2 to methanol at the copper-zirconia interface The
energetics are referenced to the Cu/ZrO2 model and CO2,(g) + 3 H2,(g). The lowest intermediates in Gibbs free energy are shown on the right

The lowest points in the diagram are formate and methoxy, and the calculated 13C and 1H chemical shifts are in good
agreement with the experimental data (Table S1 and Figure S22). Overall, the Gibbs free energy diagram is rather flat,
with an energy span of 151 kJ.mol-1. It is worth mentioning that it is significantly lower than what is found on a flat Cu(111)
surface (267 kJ.mol-1, Figure S23) or even stepped Cu(211) facets (over 200 kJ.mol-1 calculated at the GGA-PBE level of
theory). [3a]
In conclusion, we show that CO2 can be transformed into carbonate or bicarbonate, formate, methoxy, upon adsorption
and hydrogenation on zirconia-supported Cu nanoparticles. Experimentally, carbonate and bicarbonate can only be
observed under in situ conditions, consistent with the low calculated adsorption energies (about –70 kJ.mol-1), in contrast
with the strong adsorption of formate and methoxy species. The acetal, which is the next most stable intermediate, was
not observed under the described reaction conditions. In fact, calculations show that it lies about 20 kJ.mol-1 higher in
energy than the formate, with a low formate-acetal interconversion barrier of about 70 kJ.mol-1. Thus both intermediates
can be in equilibrium (in favour of the formate). This interconversion is consistent with the observed H/D exchange of the
formate intermediate (Scheme 1). Overall, the formate species is a key and observable reaction intermediate in the
hydrogenation of CO2 to methanol on Cu/ZrO2. The proposed mechanism is reminiscent to this recently proposed for Rumolecular catalyst, bridging the gap between molecular catalysis and heterogeneous catalysis. [13] The obtained results
clearly point out to the crucial molecular role of the interface between the copper particles and zirconia, [4b, 4e] paving the
road toward a more rational design of efficient heterogeneous CO 2 hydrogenation catalysts.

Acknowledgements
K.L., E.L. and S.T. were supported by the SCCER – Heat and Energy Storage program, W.C.L by SNF (200020_149704),
and A.C.V. by the SNF funding program (Ambizzione Project PZ00P2_148059). K.L. thanks the ETH Career Seed Grant
SEED-21 16-2. S. T. also thanks the Japan Society for the Promotion of Science for a fellowship (JSPS, No. 15J10157).
A.B. and A.U. acknowledge MINECO, Spain for financial support (CTQ2012-34153) and for support through Severo Ochoa
Excellence Accreditation 2014–2018 (SEV-2013-0319). The authors thank Dr. T.C. Ong (ETHZ) for many useful
discussions and the quantitative 13C SSNMR measurement, T.-H. Lin (ETHZ) for his precious help in using the highpressure glass-reactor, and Juan José Corral (ICIQ) for assistance during high-pressure DRIFTS measurements.
Keywords: CO2 hydrogenation, solid-state NMR, in situ DRIFTS, DFT, copper, zirconia
[1]
[2]
[3]

[4]

[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]

[13]

A. Goeppert, M. Czaun, J. P. Jones, G. K. S. Prakash, G. A. Olah, Chem. Soc. Rev. 2014, 43, 7995-8048.
M. Saito, T. Fujitani, M. Takeuchi, T. Watanabe, Applied Catalysis a-General 1996, 138, 311-318.
a) M. Behrens, F. Studt, I. Kasatkin, S. Kuhl, M. Havecker, F. Abild-Pedersen, S. Zander, F. Girgsdies, P. Kurr, B. L. Kniep, M. Tovar, R. W.
Fischer, J. K. Norskov, R. Schlogl, Science 2012, 336, 893-897; b) S. Kuld, M. Thorhauge, H. Falsig, C. F. Elkjaer, S. Helveg, I.
Chorkendorff, J. Sehested, Science 2016, 352, 969-974; c) T. Lunkenbein, J. Schumann, M. Behrens, R. Schlogl, M. G. Willinger, Angew.
Chem. Int. Ed. 2015, 54, 4544-4548; d) F. Studt, M. Behrens, E. L. Kunkes, N. Thomas, S. Zander, A. Tarasov, J. Schumann, E. Frei, J. B.
Varley, F. Abild-Pedersen, J. K. Norskov, R. Schlogl, ChemCatChem 2015, 7, 1105-1111; e) S. Zander, E. L. Kunkes, M. E. Schuster, J.
Schumann, G. Weinberg, D. Teschner, N. Jacobsen, R. Schlogl, M. Behrens, Angew. Chem. Int. Ed. 2013, 52, 6536-6540; f) S. Kuld, C.
Conradsen, P. G. Moses, I. Chorkendorff, J. Sehested, Angew. Chem. Int. Ed. 2014, 53, 5941-5945.
a) T. Witoo, J. Chalorngtham, P. Dumrongbunditkul, M. Chareonpanich, J. Limtrakul, Chem. Eng. J. 2016, 293, 327-336; b) F. Frusteri, G.
Bonura, C. Cannilla, G. D. Ferrante, A. Aloise, E. Catizzone, M. Migliori, G. Giordano, Applied Catalysis B-Environmental 2015, 176, 522531; c) I. A. Fisher, H. C. Woo, A. T. Bell, Catal. Lett. 1997, 44, 11-17; d) I. A. Fisher, A. T. Bell, J. Catal. 1997, 172, 222-237; e) F. Arena,
G. Italiano, K. Barbera, S. Bordiga, G. Bonura, L. Spadaro, F. Frusteri, Applied Catalysis A-General 2008, 350, 16-23; f) S. Kattel, B. Yan,
Y. Yang, J. G. Chen, P. Liu, The Journal of the American Chemical Society 2016, 138, 12440-12450; g) I. Ro, Y. Liu, M. R. Ball, D. H. K.
Jackson, J. P. Chada, C. Sener, T. F. Kuech, R. J. Madon, G. W. Huber, J. A. Dumesic, ACS Cat. 2016, 6, 7040-7050.
C. Coperet, A. Comas-Vives, M. P. Conley, D. P. Estes, A. Fedorov, V. Mougel, H. Nagae, F. Nunez-Zarur, P. A. Zhizhko, Chem. Rev.
2016, 116, 323-421.
A. Roussey, P. Gentile, D. Lafond, E. Martinez, V. Jousseaume, C. Thieuleux, C. Coperet, J. Mat. Chem. C 2013, 1, 1583-1587.
E. Guglielminotti, Langmuir 1990, 6, 1455-1460.
F. Ouyang, J. N. Kondo, K. Maruya, K. Domen, Catal. Lett. 1998, 50, 179-181.
a) O. Syzgantseva, M. Calatayud, C. Minot, J. Phys. Chem. C. 2010, 114, 11918-11923; b) O. A. Syzgantseva, M. Calatayud, C. Minot, J.
Phys. Chem. C. 2012, 116, 6636-6644.
a) K. Mudiyanselage, Y. X. Yang, F. M. Hoffmann, O. J. Furlong, J. Hrbek, M. G. White, P. Liu, D. J. Stacchiola, J. Chem. Phys. 2013, 139;
b) E. M. Mccash, S. F. Parker, J. Pritchard, M. A. Chesters, Surf. Sci. 1989, 215, 363-377.
M.-C. Silaghi, A. Comas-Vives, C. Copéret, ACS Cat. 2016, 6, 4501-4505.
a) Y. Yang, D. H. Mei, C. H. F. Peden, C. T. Campbell, C. A. Mims, ACS Cat. 2015, 5, 7328-7337; b) Q. L. Tang, Z. P. Liu, J. Phys. Chem.
C. 2010, 114, 8423-8430; c) 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, Science 2014, 345, 546-550; d) M. D. Marcinkowski, C. J. Murphy, M. L. Liriano, N. A. Wasio, F. R.
Lucci, E. C. H. Sykes, ACS Cat. 2015, 5, 7371-7378.
S. Wesselbaum, V. Moha, M. Meuresch, S. Brosinski, K. M. Thenert, J. Kothe, T. V. Stein, U. Englert, M. Holscher, J. Klankermayer, W.
Leitner, Chem. Sci. 2015, 6, 693-704.

S6

Supporting Information

Reaction Intermediates and Role of the Interface in the CO2
hydrogenation to CH3OH on ZrO2-supported Cu
Nanoparticles
Kim Larmier,[a] Wei-Chih Liao,[a] Shohei Tada,[a] Erwin Lam,[a] René Vérel,[a] Atul
Bansode,[b] Atsushi Urakawa,[b] Aleix Comas-Vives,[a]* Christophe Copéret[a]*
[a] ETH Zürich, Department of Chemistry and Applied Biosciences, Vladimir-Prelog Weg 2,
CH-8093 Zürich, Switzerland
[b] Institute of Chemical Research of Catalonia (ICIQ), The Barcelona Institute of Science
and Technology, Av. Països Catalans 16, 43007 Tarragona, Spain
* Corresponding author: comas@inorg.chem.ethz.ch, ccoperet@inorg.chem.ethz.ch

S7

1. Methods
Experimental.
pXRD – Powder XRD patterns were recorded on a STOE STADI P apparatus at a voltage of
40 kV and a current of 35 mA.
N2 physisorption – The specific surface area of the support was measured from a nitrogen
physisorption isotherm recorded at 77 K on a BEL JAPAN BELSORP-min apparatus. The data
were analyzed by the BET method.
TEM – The morphology of the samples was obtained by transmission electron microscopy
(TEM, Philips CM12) and by high resolution TEM (FEI Tecnai F30).
Fourier-Transform InfraRed spectroscopy - FTIR measurements were carried out on a Bruker
Alfa-T spectrometer (inside a glovebox under Ar atmosphere). The powdered samples were
pressed into a thin disk using a 7 mm die set. Typically, 24 scans were collected for each
spectrum at a resolution of 4 cm-1. Background spectra were collected under Ar atmosphere.
In situ DRIFTS (Diffuse Reflectance Infrared Fourier Transform Spectroscopy) – In situ
DRIFTS measurements were performed in a custom-made high-pressure reaction cell having a
cylindrical cavity (3 mm in diameter and 3 mm vertical length) for the sample placement. The
cell, resembling that reported previously,[1] was mounted in a Praying Mantis (Harrick) optical
accessory. Approximately, 10-15 mg of catalyst powder was placed in the cell and reduced at
250 C for 1 h under continuous flow of H2 (15 mL·min-1), prior to the exposure to the reaction
gas mixture (CO2:H2 = 1:3). For the measurements under ambient pressure conditions, a
continuous flow of the gas mixture was passed over the catalyst at 15 mL·min-1. For the
experiments, involving reaction pressure up to 20 bars, a high-pressure needle valve was placed
at the outlet of cell to build the pressure in the cell. The flow of the gas mixture was stopped
after attaining the desired reaction pressure. The spectra were collected with 4 cm -1 resolution
on an FT-IR spectrometer (TENSOR 27 / Vertex 70, Bruker) equipped with a liquid-nitrogencooled MCT detector. The DRIFT spectrum recorded under H2 flow at the reaction temperature
was used as the background.
NMR spectroscopy –
The solution NMR experiments were done on a Bruker 500 MHz AVANCE III HD
spectrometer with a 5 mm probe. The chemical shifts were referenced to the residual proton in
the solvent.
The solid-state NMR experiments were done in a Bruker 400 MHz AVANCE III HD
spectrometer with a 4 mm MAS triple resonance probe operating in double resonance mode.
The MAS frequency was set to 10 kHz. The chemical shift scale was calibrated using
admantane as an external secondary reference. Ramped cross polarization (1H-13C) was used
for most experiments with 1H excitation frequency (γβ1) at 100 kHz. The contact time was 2 ms
for 1D experiments and 0.5 ms for 1H-13C HETCOR experiments. Additionally, for 1H-13C
HETCOR experiment, DUMBO homonuclear (1H-1H) decoupling was used during t1.[2]
Ab initio calculations.
Periodic DFT calculations were carried out with Vienna Ab Initio Simulation Package
(VASP)[3] code using the projector augmented wave (PAW) method[4] with a plane wave energy
cutoff of 400 eV and the PBE exchange-correlation functional. The criterion convergence
chosen for the SCF cycle was 10-5 eV, and optimizations were considered converged once the
forces on all atoms were lower than 0.1 eV.Å-1.
̅
To model the zirconia surface, we considered a slab of the (111) termination of monoclinic
zirconia. The slab was 13.6 x 14.7 Å large, and 9 Å thick (144 atoms, 48 ZrO2 units), with a
vacuum separation of 17 Å. The three lowest atomic layers were frozen during optimizations
and transition state calculations. The Brillouin zone was sampled with a 2 x 2 x 1 k-point grid.
S8

Electron occupancies were determined according to a gaussian scheme with an energy smearing
of 0.1 eV. For the supported copper particle, a Cu38 cluster of truncated octahedron shape from
Cu fcc structure was deposited on the surface. The ensemble was optimized before studying the
adsorption and reactivity of CO2 and H2.
Cu(111) was simulated using a periodic 3 x 3 four-layer slab consisting of 36 Cu atoms with a
̊
vacuum separation of 15 A. During optimization the first Cu(111) layer was kept frozen. The
Brillouin zone was sampled with a 3 x 3 x 1 k-point grid. Electron occupancies were determined
according to a Methfessel-Paxton scheme with an energy smearing of 0.2 eV.
Transition states were determined using the Climbing Image Nudge Elastic Band (Cl-NEB)
method with 8 images.[5] The transition states were confirmed by frequency analysis.
Chemical shifts calculations were performed using the linear response method implemented in
VASP. For these calculations, the convergence criterion on the SCF cycles was set to 10-8 eV.
Thermodynamic calculations including enthalpic and entropic corrections were performed to
calculate the Gibbs free energies by calculating the vibration frequencies for the gas-phase
molecules, the adsorbed species and the transition states. The procedure used in this publication
is fully described in Larmier et al [6] :
For a given species i, G(i,T,P) can be calculated through the formula (S1)
G(i,T,P) = E(i) + Uvib(i,T) + Urot(i,T) + Utrans(i,T) + PVm –T x Svib(i,T) –T x
Srot(i,T) – T x Strans(i,T,P)

(S1)

The vibrational, rotational and translational enthalpic and entropic contributions for a gas phase
molecule considered as an ideal gas can be calculated with equations (S2) – (S8):
ℎ𝜈 𝑛
ℎ𝜈 𝑛 × exp⁡(−
)
1
𝑘𝐵 𝑇
𝑈 𝑣𝑖𝑏 (𝑖, 𝑇) = 𝑁 𝐴 [∑ ℎ𝜈 𝑛 + ∑
]
ℎ𝜈 𝑛
2
1 − exp⁡(−
)
𝑛
𝑛
𝑘𝐵 𝑇

(S2)

𝑈 𝑡𝑟𝑎𝑛𝑠 (𝑖, 𝑇) + 𝑈 𝑟𝑜𝑡 (𝑖, 𝑇) + 𝑃𝑉 𝑚 (𝑖, 𝑇) = 4𝑅𝑇 (non-linear molecules)

(S3)

𝑈 𝑡𝑟𝑎𝑛𝑠 (𝑖, 𝑇) + 𝑈 𝑟𝑜𝑡 (𝑖, 𝑇) + 𝑃𝑉 𝑚 (𝑖, 𝑇) = 7/2𝑅𝑇 (linear molecules)

(S4)

ℎ𝜈 𝑛
ℎ𝜈 𝑛
× exp⁡(−
)
ℎ𝜈 𝑛
𝑘𝐵 𝑇
𝑘𝐵 𝑇
𝑆 𝑣𝑖𝑏 (𝑖, 𝑇) = 𝑁 𝐴 𝑘 𝐵 [∑
− ∑ ln⁡(1 − exp⁡(−
))]
ℎ𝜈 𝑛
𝑘𝐵 𝑇
1 − exp⁡(−
)
𝑛
𝑛
𝑘𝐵 𝑇

(S5)

3

𝑆 𝑟𝑜𝑡 (𝑖, 𝑇) = 𝑁 𝐴 𝑘

√ 𝜋 8𝜋 2 𝑘 𝑇 2
ln⁡[ 𝜎 ( ℎ2 𝐵 ) √ 𝐼 𝑥,𝑖
𝐵

× 𝐼 𝑦,𝑖 × 𝐼 𝑧,𝑖 ] (nonlinear molecules)

(S6)

S9

1 8𝜋 2 𝑘 𝐵 𝑇
) 𝐼𝑖]
ℎ2

𝑆 𝑟𝑜𝑡 (𝑖, 𝑇) = 𝑁 𝐴 𝑘 𝐵 ln⁡[ 𝜎 (

(linear molecules)

5
5
𝑆 𝑡𝑟𝑎𝑛𝑠 (𝑖, 𝑇, 𝑃) = 𝑁 𝐴 𝑘 𝐵 ( ln(𝑇) − ln(𝑃) + ln(𝑀 𝑖 ) − 1.165)
2
2

(S7)

(S8)

where the vibrational contributions can be deduced from the calculated frequencies νn of the
model (note that equation (S3) includes the zero point vibrational energy in its first term). Ix,i,
Iy,i, Iz,i (Ii for a linear molecule) are the moments of inertia of the molecule i, and Mi its molecular
mass. Rotational contributions have been calculated using the Colby Rotational Constant
Calculator (http://www.colby.edu/chemistry/PChem/scripts/ABC.html).[7]
For an adsorbed molecule or for a surface, the rotational and translational contributions are
converted into vibration modes, so that the only remaining terms are the vibrational ones – that
can be calculated using the equations (S2) and (S5) – and the electronic energy E(i). For these
phases, we also consider that the PVm term is very small with regard to the energetic terms, and
is therefore neglected, and thus we consider H = U in this case.

S10

2. Sample preparation method and characterization
Support. Commercial ZrO2 (DK-1 from DAIICHI KIGENSO KAGAKU KOGYO CO., LTD.,
25 m2.g-1 from BET analysis of N2 physisorption measurement) was calcined at 500 °C for 2
hours to remove organic impurities. The support was then partially dehydroxylated at 500°C
for 20 h under high vacuum (10-5 mbar), and then cooled down to room temperature prior
storage under argon atmosphere. This sample is referred to as ZrO2-500. The support shows the
Powder X-Ray Diffraction (pXRD) pattern of monoclinic zirconia (Figure S1-(a)).
Commercial SiO2 (AEROSIL 200, Evonik, SBET = 206 m2.g-1) was partially dehydroxylated at
500, for 20 h under high vacuum (10-4 mbar), and then cooled down to room temperature prior
storage under argon atmosphere. This support is referred to as SiO2-500.

6000
ZrO2

Count (a.u.)

Cu/ZrO2
4000

2000

0
20

25

30

35

40

45

50

55

60

65

2θ (°)

Figure S1. XRD pattern of ZrO2 (a) before (b) after copper particle synthesis. Only monoclinic
zirconia phase is detected (International Center for Diffraction Data, Entry 00-083-0939).
Copper particles synthesis and characterization. All samples were prepared by the grafting
method using copper (I) tert-butoxide,[8] [Cu(OtBu)]4, as a precursor. 100 mg of [Cu(OtBu)]4[9]
were dissolved in dry pentane (20 mL). The solution was contacted with 1.0 g of the zirconia
support under argon atmosphere in a double Schlenk. After 4 hours of stirring at room
temperature, the solid was washed three times with 20 mL of pentane to remove unreacted
copper precursor and dried under high vacuum conditions (~10-5 mbar). The sample was
reduced at 500 °C under H2 atmosphere for 5 h. After reduction, the sample was cooled down
to room temperature under H2, before vacuum treatment and storage under argon.
The operation can be followed by IR spectroscopy at the various stages of the preparation.
Figure S2-(a) shows the IR spectrum of ZrO2-500, that mainly displays ν-OH stretching
frequencies at 3774 and 3670 cm-1 related to Zr-OH hydroxyl groups. After grafting, C-H
stretching bands appear around 2966 cm-1 while the intensity of the Zr-OH bands is strongly
decreased (Figure S2-(b)). This confirms the grafting of the molecular precursor by chemical
reaction with the Zr-OH groups according to Scheme S1. Upon reduction, the ν-CH bands
S11

Absorbance (a.u.)

(a)

2966

3774
3670

disappear while the ν-OH bands reappear as an indication that the ligands are decomposed
(Figure S2-(c)). At the same time, the color of the sample turns from pale yellow to pink, and
particles can be observed by TEM (see below). Scheme S1 shows a schematic representation
of the whole process. Finally, the copper content (0.8 %) was determined by ICP (Inductively
Coupled Plasma) spectroscopy.

(b)

ν-CH
(c)

ν-OH
4000

3500

3000

2500

2000

1500

1000

Wavenumber (cm-1)

Figure S2. FTIR spectra of the sample during preparation. (a) ZrO2-500 (b) grafted
[Cu(OtBu)]4 on ZrO2-500 (c) after reduction under H2 at 500 °C for 5 hours.

Scheme S1. Schematic representation of the grafting-reduction sequence.
The pXRD pattern remains unchanged by this procedure (Figure S1-(b)). No diffraction peak
for copper could be detected. TEM imaging shows indeed that the copper particles are small
(2.2 ± 0.5 nm) and thus below the detection limit of pXRD (Figure S3-(a)).
A Cu/SiO2-500 sample was prepared according to a similar synthesis method. Using 100 mg of
[Cu(OtBu)4] in pentane for 1 g of silica yields a sample with 3.7 % of copper loading after
grafting and reduction (H2, 500 °C, 5 hours). TEM imaging on Figure S3-(b) shows the particles
and the particle size distribution for this catalyst.

S12

(a)

70

2.2 ± 0.5

60

Count

50
40
30
20
10
0
0

2

4

6

8

10

Particle size (nm)

(b)
2.1 ± 0.4

100

80

Count

60

40

20 nm
2 0 nm

20

0
0

1

2

3

4

5

6

7

8

9

10

Particle size (nm)

Figure S3. (a) TEM imaging and particle size distribution of Cu/ZrO2-500 prepared as described
above (b) TEM imaging and particle size distribution of Cu/SiO2-500 prepared as described
above

S13

3. Catalytic testing in flow reactor
The CO2 hydrogenation was conducted in a fixed-bed tubular reactor at 25 bars (PID
Eng&Tech). The reaction temperature was measured at the catalyst bed by a K-type
thermocouple. After loading about 300 mg of a catalyst powder and 5.0 g of SiC in the reactor
under air, the catalyst was treated under a flow of 17%H2/N2 (60 mL min-1) at 300 ºC for 30
min under ambient pressure. After cooling to 270 ºC, a flow of CO2/H2/N2 (1/3/1, 62 mL min1
) was passed through the catalyst bed for 12 h at 25 bar. The conditions were then set to the
measurements conditions (230 °C, 25 bars), and the products were analyzed by an online gas
chromatograph (Agilent 7890A) equipped with FID (for methanol) and TCD (N2, CO2, CO,
CH4). In all experiments, CH4 concentration in the outlet gas was negligible.
The formation rates of CO and MeOH and the selectivity to methanol were calculated using the
following equations.
Fout ⁡[mol⁡h−1 ] = ⁡

Fin ⁡×⁡CN2 ,in
CN2 ,out
Fout ⁡×Cx,out ⁡

rx ⁡[molx ⁡h−1 ⁡g −1 ] = ⁡
Cu
SMeOH =

FmeOH,out

⁡wCu

FCO,out +FMeOH,out

⁡

(S9)
(S10)
(S11)

where Fin is a total gas inlet flow rate (mol h-1), Fout is a total gas outlet flow rate (mol h-1),
Cx,in is the inlet gas fraction of species x, Cx,out is the outlet gas fraction of species x, rx is the
production rate of species x (molx h-1 gcat-1) and wCu the copper content in the catalytic bed (gCu).
The volumetric flow rate Q (in mL.min-1) of the feed was decreased in order to increase the
contact time (defined as mcat/Q). Each flow rate was kept constant for about 4 hours before
switching to the next one. For each flow rate, the formation rates of CO and methanol are
constant save for a rather large scatter of the CO rate due to the integration of from rather small
TCD peaks (Figure S4 (a) and (b)). After about 30 hours on stream, the flow rate is changed
back to the initial value. On Cu/SiO2 the initial value is recovered, indicating negligible
deactivation. On Cu/ZrO2, slightly lower values are obtained (about 10 % lower than the initial
one), indicating a very slow deactivation process for both CO and methanol productions. Thus,
steady-state is a good approximation for both catalysts. The five points were averaged for the
calculation of the formation rates and selectivities for a given contact time (flow rate). The rates
were extrapolated using a second order polynomial function. This function has no physical
meaning and is merely used to allow for estimating the initial rates at zero conversion – and
hence selectivities.

S14

(a) Cu/ZrO2
Formation rates (μmol.s-1.gcu-1)

Decreasing flow rate / increasing contact time
16

62

37

31

19

25

12

62

7.5

10

14
12
10
8
6
4

MeOH
CO

2
0
0

5

10

15

20

25

30

35

Time on stream (h)

(b) Cu/SiO2
Decreasing flow rate / increasing contact time

Formation rates (μmol.s-1.gcu-1)

10

62

37

31

19

25

12

10

62

7.5

8

6

4

2

MeOH
CO
0
0

5

10

15

20

25

30

35

Time on stream (h)

S15

(c)
80
60

4

40
2
20
0

Methanol Selectivity (%)

6

CO2 Conversion (%)

100

Conv. -- Cu/ZrO2
Conv. Cu/ZrO2
Conv. Cu/SiO2
Conv. -- Cu/SiO2
S MeOH --Cu/SiO22
MeOH. Cu/ZrO
S MeOH.--Cu/ZrO2
MeOH Cu/SiO2

0
0

1

2

3

4

Contact time mcat/Q (s.g.mL-1)

(d)

MeOH - Cu/ZrO2
Cu/ZrO2
CO - Cu/ZrO2
CO- Cu/ZrO2
MeOH - Cu/SiO2
Cu/SiO2
CO- Cu/SiO2
CO - Cu/SiO2

100

Selectivity (%)

80
60
40
20
0
0

1

2

3

CO2 Conversion (%)

Figure S4. Catalytic results of the hydrogenation of CO2 in a flow reactor (230 °C, 25 bars of
H2/CO2/N2 (3:1:1), mcat ~ 300 mg) (a) Evolution of the formation rates with time on stream for
Cu/ZrO2 (b) Evolution of the formation rates with time on stream for Cu/SiO2(c) Evolution of
the CO2 conversion and methanol selectivity with contact time. (d) Selectivity versus
conversion plot.

S16

4. Additional IR spectroscopic measurements

1139
1031

2921
2824

1633
1533
1424
1340

(a)

2940
2857

3776
3670
3596

Ex-situ IR measurements. Similarly to the Ex-Situ NMR spectra presented in the main article
(Figure 1), Ex-situ IR measurements were recorded after reaction of Cu/ZrO2 in a closed vessel
with 13CO2 and H2 and/or D2, followed by evacuation of the gas-phase under high vacuum (104
mbar).

Absorbance (a.u.)

(b)

(c)
2700

(d)

2149 2048

(e)

0
4000

3500

3000

2500

2000

1500

1000

Wavenumber (cm-1)

Figure S5. (a) IR spectra of the starting Cu/ZrO2 sample. (b-e) Ex-situ IR spectra taken after
reaction of Cu/ZrO2 with 13CO2 and H2 at 230 °C for 12 hours at (b) 5 bars (c) 1 bar (d) After
reacting sample from spectrum (b) with D2 (5 bars) at 230 °C for 12 hours (e) Ex-situ IR
spectrum taken after reaction of pure ZrO2 with 13CO2 and H2 at 230 °C for 12 hours at 5bars.

In situ DRIFTS measurements.

S17

1800

1600

1400

1200

Absorbance (a.u.)

1049

2550 2000

1154

2750

Wavenumber (cm-1)

1524

2736

2950

20 bars
10 bars
bar
5 bars
1 bar

1357
1387

(b)

2942
2978

3150

1567
1593

2828
2878

(a)

1000

Wavenumber (cm-1)

(c)

2873
1623
2965

1432
1329
1221

Absorbance (u.a.)

20 bar
10 bar
5 bar

1 bar

4000

3000

2000

Wave number

1000

(cm-1)

Figure S6. In situ DRIFT spectra of Cu/ZrO2 at 230 °C under 1 to 20 bars of a H2/CO2 mixture
in the regions of the (a) C-H stretching and (b) C-O stretching frequencies. The DRIFT spectra
recorded under H2 flow at the reaction temperature were used as the background. (c) In situ
DRIFT spectra of pure ZrO2 contacted with a CO2/H2 mixture (1:3) at 230 °C and at various
pressures (1 to 20 bars).

S18

5. Adsorption of Formic acid and methanol on the catalysts (Cu/SiO2 and Cu/ZrO2)
In order to assign the spectroscopic signals observed during reaction, we adsorbed formic acid
on the catalysts and measured IR spectra and 1H solid-state NMR spectra. A small amount of
the reduced sample is loaded in a glass tube in an Ar glovebox. The tube is evacuated at room
temperature under high vacuum (10-4 mbar), and the solid is contacted with either 30 mbar of
formic acid or 80 mbar of methanol for 5 minutes. The gas-phase is then evacuated under high
vacuum at room temperature for 6 hours. The sample is then transferred back in a glovebox,
where the IR spectrum is recorded. In the case of formic acid, the sample is packed into a
2.5 mm rotor and a MAS NMR 1H spectrum is recorded (spinning rate 20 kHz).
Detection of formate adsorbed on copper by NMR
The adsorption of formic acid on silica gives rise to small features at 2946 cm-1 and 1734 cm-1
(Figure S7). The latter band is similar to the C=O stretching frequency of gas-phase formic
acid. Thus these features can be assigned to formic acid bound to silica through hydrogen
bonding. The 1H NMR spectrum shows a weak signal at 8.7 and 8.4 ppm. The region between
0 and 5 ppm shows several peaks that are already present in the silica itself, thus the changes
can be attributed to perturbation of the hydrogen bond network (Figure S8).
Adsorption of formic acid on Cu/SiO2 shows the band at 1735 cm-1 and two additional features
in the C=O region at 1668 and 1549 cm-1, that are consistent with formic acid adsorbed on
copper and formate on copper, respectively (see Table S2). These are associated to bands in the
CH region at 2858, 2876, 2938 and 2954 cm-1. These bands are much stronger in intensity than
the CH band of formic acid on silica, suggesting a much greater surface species concentration
(Figure S7). Thus, in this sample, three species coexist: formic acid on silica, formic acid on
copper and formate on copper. However, the MAS NMR 1H spectrum shows no additional
signal around 5-10 ppm (Figure S8).
On the contrary, formic acid adsorbed on ZrO2 or Cu/ZrO2 shows strong IR frequencies in the
CH region at 2962, 2932, 2875, 2854 cm-1 and 1565 and 1387 cm-1 in the CO region (Figure
S9), that are attributed to formate adsorbed on zirconia. Very little difference is observed
between Cu/ZrO2 and ZrO2, suggesting that most species are on the support. MAS NMR 1H
spectrum (Figure S10) also shows a strong signal at 8.8 ppm.

S19

Adsorption of formic acid on Cu/SiO2 and SiO2

(b)

2938

Absorbance (a.u.)

2858
2954

SiO2
SiO2

SiO2
SiO2

Cu/SiO2
Cu/SiO2

2876
2696

2946

3050

2950
2850
2750
Wavenumber (cm-1)

1357

Cu/SiO2
Cu/SiO2
Absorbance (a.u.)

(a)

1735 1668

2111

2100

2650

1549

1734

1900
1700
1500
Wavenumber (cm-1)

1300

Figure S7. FTIR spectra of formic acid adsorbed on Cu/SiO2 and SiO2, after evacuation at
room temperature overnight. Spectra are normalized with respect to the Si-O-Si bands of silica
between 2100 and 1500 cm-1.

2.7

8.7 8.4

HCOOH_Cu/SiO2

HCOOH_SiO2

SiO2

Figure S8. Solid-state MAS NMR 1H NMR of formic acid adsorbed on SiO2 or Cu/SiO2. The
starting silica is also shown. Spectrometer 700 MHz (16.5 T), MAS frequency 20 kHz, 200
scans, line broadening 50 Hz.

S20

Adsorption of formic acid on Cu/ZrO2 and ZrO2

(a)

2875

ZrO2
ZrO2
Cu/ZrO2
Cu/ZrO2

2932

1565
1387

Absorbance (a.u.)

Absorbance (a.u.)

2962

(b)

ZrO2
Cu/ZrO2
ZrO2
Cu/ZrO2

2854

2950

1309

2729

3050

2950
2850
2750
Wavenumber (cm-1)

2650

1800

1600
1400
1200
Wavenumber (cm-1)

1000

Figure S9. FTIR spectra of formic acid adsorbed on Cu/ZrO2 and ZrO2 after evacuation at room
temperature overnight.

2.7
8.8

HCOOH_Cu/ZrO2

HCOOH_ZrO2

ZrO2

Figure S10. Solid-state MAS NMR 1H NMR of formic acid adsorbed on ZrO2 or Cu/ZrO2. The
starting zirconia is also shown. Spectrometer 700 MHz (16.5 T), MAS frequency 20 kHz, 200
scans, line broadening 50 Hz.

S21

Adsorption of methanol on Cu/SiO2 and SiO2
2959

(a)

Cu/SiO2
Cu/SiO2
SiO2
SiO2

2858

(b)

SiO2
SiO2
Cu/SiO2
Cu/SiO2

2929

2893

3050

2823

2950
2850
Wavenumber (cm-1)

1462
1435

Absorbance (a.u.)

Absorbance (a.u.)

2996

2750

2100

1900
1700
Wavenumber (cm-1)

1500

1300

Figure S11. FTIR spectra of methanol adsorbed on Cu/SiO2 and SiO2. Spectra are normalized
with respect to the Si-O-Si bands between 2100 and 1500 cm-1.
Adsorption of methanol on Cu/ZrO2 and ZrO2
1166

(a)

2931

(b)

2817

3050

2950
2850
Wavenumber (cm-1)

2750

ZrO2
ZrO2
Cu/ZrO2
Cu/ZrO2

1465

Absorbance (a.u.)

Absorbance (a.u.)

ZrO2
Cu/ZrO2
ZrO2
Cu/ZrO2

1061

1445

1800

1600
1400
Wavenumber (cm-1)

1200

1000

Figure S12. FTIR spectra of methanol adsorbed on Cu/ZrO2 and ZrO2.

6. Reaction of CO2 and H2 on Cu/SiO2
The 13CO2 hydrogenation at 5 bars and 230 °C in batch as reported for Cu/ZrO2 was also
performed on the Cu/SiO2 sample according to a similar procedure. Solid-state 1H and 1H-13C
CP-MAS NMR spectra were recorded. Figure S13 compare the spectra obtained with Cu/ZrO2
and Cu/SiO2. On the proton NMR spectrum, Cu/ZrO2 shows quite intense signals at 8.4 and 4.1
ppm for formate and methoxy, respectively. With Cu/SiO2, no signal is observed in the 7-10
ppm region, although more scans were used to record this one (256 vs. 64 on Cu/ZrO2). On the
CP-MAS 13C spectrum, intense signals are obtained on Cu/ZrO2 at 166 and 52 ppm while we
could only observe a very weak and broad signal at 168 ppm (formate) with Cu/SiO2. In the
latter case, longer mixing times (2 ms vs. 0.5 ms) and much more scans (7344 vs. 64) where
S22

used, which should have resulted in stronger signals. This signal can be assigned to formate
adsorbed on copper, as confirmed by IR spectroscopy (see Figure S13, performed with 12CO2
and H2 to allow comparison with spectra in Figures S7 to S12),[10] although its observation by
NMR is made very difficult because of fast relaxation of the polarization. Alternatively, it might
arise from very small amounts of formate or formic acid spilt over to silica.
To confirm the presence of formate only at the surface of Cu/SiO2, the catalyst was washed
with D2O after the reaction (without exposure to air), and liquid 13C NMR of the supernatant
was measured (Figure S15). The signal at 168 ppm of formic acid appears, while no methanol
is observed.
1354
2937

(a)

Cu/SiO2
Cu/SiO2
Cu/ZrO2
Cu/ZrO2

2857

(b)
Cu/ZrO2
Cu/ZrO2
Cu/SiO2
Cu/SiO2

Absorbance (a.u.)

Absorbance (a.u.)

2694
2875
2929
2970

2857
2829

1544
1573
1375
1387 1357

2739

3050

2950
2850
2750
Wavenumber (cm-1)

Figure S13. FTIR spectra of
Cu/ZrO2 (b) Cu/SiO2.

12

1600
1500
Wavenumber (cm-1)

1400

1300

CO2 and H2 reacted at 5 bars and 230 °C for 12 hours on (a)

2.3

(a)

1700

2650

(b)
168

8.4

52

Cu/SiO2

166
4.1

Cu/ZrO2

Cu/SiO2

*

Cu/ZrO2

Figure S14. (a) Ex-situ MAS NMR 1H spectrum of 13CO2 and H2 reacted with Cu/SiO2 (NS =
256) or Cu/ZrO2 (NS = 64) at 5 bars and 230 °C for 12 hours (b) Ex-situ CP-MAS NMR 13C1
H spectrum of 13CO2 and H2 reacted with Cu/SiO2 (NS = 7344, mixing time 2 ms) or Cu/ZrO2
(NS = 64, mixing time 0.5 ms) at 5 bars and 230 °C for 12 hours. No HETCOR spectrum could
be recorded on Cu/SiO2. * : spinning side band.

S23

168

Figure S15. Liquid 13C NMR spectrum of the supernatant of Cu/SiO2 washed with D2O.

S24

7. Additional solid-state NMR spectra

Figure S16. Ex-situ MAS NMR 1H-13C HETCOR spectrum of ZrO2 after reaction with
13
CO2/H2 (1:3) at 5 bar and 230 °C. The signal at δ(13C) = 167 ppm, δ(1H) = 8.0 ppm is
attributed to formate, the signal at δ(13C) = 179 ppm, δ(1H) = 1.6 ppm to bicarbonate.

(a)

(b)

Figure S17. Ex-situ MAS NMR 13C direct excitation spectra of Cu/ZrO2 (a) after contacting
13
CO2/H2 (1:3) at 1 bar at 230 °C (b) after contacting the sample from spectrum (a) with 5 bars
of D2. The recycle delay of spectrum (b) was set as 240 s in order to have a quantitative analysis.
Neither carbonate (expected at 180-210 ppm) nor CO2 (around 120 ppm) can be observed.

S25

8. Liquid state NMR spectroscopy
J (1H-2D) = 1.7 Hz

J (1H-13C) = 141 Hz
3.24 ppm
Figure S18. Liquid-state 1H NMR after washing Cu/ZrO2 with D2O after the two steps
hydrogenation of 13CO2. A doublet of pentet centered at 3.24 ppm can be observed, related to
the 1H-13C and 1H-2D J-couplings. This is consistent with a 13CHD2 group. This signal merges
into a single peak at 3.24 ppm when a 13C decoupling pulse is applied. The other peaks are
impurities introduced during the process. They do not change upon 13C decoupling.
H2O
J (2D-13C) = 22 Hz

Figure S19. Liquid-state 2D NMR after washing Cu/ZrO2 with H2O after the two steps
hydrogenation of 13CO2. A doublet centered at 3.19 ppm can be observed, related to the 2D-13C
J-coupling. No fine structure can be distinguished (no 1H decoupling was applied).
Quantification of the number of desorbed methanol
Based on the deuterium spectrum shown in Figure S19, it is possible to evaluate the number of
methanol released in solution, and hence to give a rough estimate of the number of active sites
on the catalyst.
S26

The deuterium signal associated to water at 4.7 ppm is due to the naturally abundant deuterium,
(0.0115 %). The concentration of proton in water is of 111.1 mol.L-1; hence the concentration
of deuterium is of 1.3 10-2 mol.L-1. The peaks assigned to methoxy integrate for 0.041 of the
water peak at 4.7 ppm.
If we assume that the predominant form of methanol in solution is CD3OD, and if we neglect
the desorption of some OD groups on the surface in the form of water in solution, we can
calculate the concentration of methanol in solution to be 1.8 10-4 mol.L-1. As 2.0 mL of water
were employed to wash the solid, this represents 3.5 10-7 mol of methanol.
The spectrum was obtained after washing 200 mg of sample (SSA = 25 m2.g-1). Thus, the
surface concentration of adsorbed methanol was of 0.07 μmol.m-2 (0.04 methanol.nm-2).
Assuming that one molecule of methanol was adsorbed on each active site, this number should
be the number of active sites.
We may also estimate the number of particles at the surface of the catalyst. Considering the
loading (0.8 % Cu), the size of the particles (~ 2 nm, which is constituted of about 300 Cu
atoms) and the specific surface area, we calculate the average density of particles at the surface
to be of 0.01 particle.nm-2. This number is about the same order of magnitude than the number
of active sites estimated (~ 4 sites/particles), which rather supports the involvement of a small
number of interfacial sites – although these estimates are quite rough.

S27

!!!a)!!!!!!!!!!!!!!!!!!!!!!!!!b)!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!c)!

9. Additional DFT structures

!!!a)!!!!!!!!!!!!!!!!!!!!!!!!!b)!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!c)!

(a)

(c)

(b)

(d)

Figure S20. Models for the supported copper particle (a) Cu38 particle (ø = 0.8 nm) (b)
moniclinic ZrO2 slab (top view). The frame highlights the limits of the simulation box. (c-d)
Optimized Cu38/ZrO2 model (c) Top view (d) Side view.

(a)

∆adsE (H2O) = -102 kJ.mol-1

(b)

∆adsE (CO2) = -65 kJ.mol-1

Figure S21. (a) Water molecule adsorbed on the ZrO2 surface. The molecule is dissociated in
two hydroxyl group. (b) Bicarbonate (CO3H*) generated by adsorption of CO2 on one hydroxyl
group.

S28

10. Various calculations
Table S1. Calculated 1H and 13C chemical shifts and Bader charges on the carbon atom for
̅
various intermediates optimized on the Cu (111) and ZrO2 (111) facets, as well as at interfacial
sites of the Cu/ZrO2 model. When two types of H atoms are present in the intermediate, both
are given in the order of the formula.
Surface
Species
δ (1H)
δ (13C)
Charge on C
COOH*
3.0
219
1.2
HCOO*
2.6
179
1.6
HCOOH*
1.0/-1.3
180
1.5
Cu (111)
H2COO*
1.1
90
0.9
H3CO*
-2.6
54
0.4
H3COH
-4.1/-4.0
61
0.3
CO3*
158
2.1
CO3H*
15.3
169
2.2
̅11)
HCOO*
6.6
180
1.5
ZrO2 (1
H2COO*
8.7
87
1.1
H3CO*
3.4
67
0.3
CO2*
219
1.2
COOH*
10.9
227
1.1
HCOO*
6.5
177
1.5
Cu/ZrO2
H2COO*
9.2
114
1.0
H3CO*
3.6
66
0.5
H3COH*
3.3
58
0.3
CO2(g)
135
2.1
CO2(g) - experimental
125
-4
-2

Chemical Shift (ppm)
150

100

50

0
0
2
4
6
8

Chemical Shift (ppm)

200

1H

13C

10
1

13

Figure S22. Simulated H- C HETCOR spectra of the formate (red) and methoxy
̅
intermediates (blue) on the Cu (111) surface (open circles), on the ZrO2 (111) facet (open
diamond) and on the Cu/ZrO2 interface (open triangles). Experimental points are shown
(squares).
Surface
Cu (111)
̅
ZrO2 (111)
Cu/ZrO2
MeOH(g)

Species
H3CO*
H3COH
H3CO*
H3CO*
H3COH*

νC-H
3028, 3024, 2966
3078,3023,2949
2995,2964,2904
2959,2859,2837
3093, 3037,2965
3061, 2980,2944

δC--H
1446,1441,1417
1455,1440,1421
1452,1447,1424
1436,1422,1407
1468,1446,1422
1461,1449,1433

δO-C--H
1135,1132
1135,1046
1146,1138
1158,1133
1144,1093
1130,1062

νC-O
1012
983
1102
1130
1000
1010

S29

Table S2. Calculated vibration frequencies for the formate/formic acid and methanol/methoxy
intermediates
Surface
Cu (111)
̅
ZrO2 (111)

Species
HCOO*
HCOOH*
HCOO*
HCOOH*
HCOO*

Cu/ZrO2
HCOOH(g)

νC-H
2954
3001
2980
2895
3014
2984

νO-C-O
1529
1663
1532
1694
1525
1763

δC-H
1318, 1313
1360,1289
1348,1365
1340,1202
1369,1330
1352,1261

δC-O
983
1001
1014
1002
1056
1000

11. Methanol synthesis on the Cu(111) surface
There are some mechanistic differences during CO2 hydrogenation to methanol. The
hydrogenation on Cu(111) surface via a formate (HCO2*) intermediate shows that in contrast
to Cu/ZrO2, the reduction of formate (HCO2*) to acetal (H2CO2*) is less favorable than going
through formic acid (HCO2H*) with a difference in overall activation barrier of ΔΔG‡ = 20
kJ/mol. The overall activation barrier for the favored pathway is ΔG‡ = 267 kJ/mol. The
difference in preferred reaction mechanism is due to the Lewis acidic site of ZrO2 which can
bind more strongly the reaction intermediates and hence favors to hydrogenate carbon a second
time to keep the bidentate bonding. The decomposition of hydroxymethoxy (H2CO2H*) to
methoxy (CH3O*) and hydroxyl (OH*) happens in a stepwise fashion by first forming
formaldehyde and hydroxide upon which the formaldehyde gets further reduced to methoxy
(CH3O*).

S30

Figure S23. Enthalpy and Gibbs free energy at 475 K for the hydrogenation of CO2 to methanol on Cu(111) surface via two mechanistic pathway
via formic acid (straight line) and acetal (dashed line).

S31

REFERENCES
[1]
[2]

A. Urakawa, N. Maeda, A. Baiker, Angew. Chem. Int. Ed. 2008, 47, 9256-9259.
A. Lesage, D. Sakellariou, S. Hediger, B. Elena, P. Charmont, S. Steuernagel, L. Emsley, J. Magn. Reson. 2003, 163, 105-113.

[3]
[4]

a) G. Kresse, J. Furthmuller, Phys. Rev. B 1996, 54, 11169-11186 b) G. Kresse, J. Hafner, Phys. Rev. B 1994, 49, 14251-14269.
G. Kresse, D. Joubert, Phys. Rev. B 1999, 59, 1758-1775.

[5]

a) G. Henkelman, H. Jonsson, J. Chem. Phys. 2000, 113, 9978-9985 b) G. Henkelman, B. P. Uberuaga, H. Jonsson, J. Chem. Phys. 2000,
113, 9901-9904.
K. Larmier, C. Chizallet, N. Cadran, S. Maury, J. Abboud, A. F. Lamic-Humblot, E. Marceau, H. Lauron-Pernot, ACS Cat. 2015, 5, 44234437.

[6]

;

;

;

[7]
[8]

a) H. Tetrode, Annalen Der Physik 1912, 38, 434-442 b) O. Sackur, Annalen Der Physik 1912, 40, 87-106.
A. Roussey, P. Gentile, D. Lafond, E. Martinez, V. Jousseaume, C. Thieuleux, C. Coperet, J. Mat. Chem. C 2013, 1, 1583-1587.

[9]

a) K. L. Fujdala, I. J. Drake, A. T. Bell, T. D. Tilley, J. Am. Chem. Soc. 2004, 126, 10864-10866 b) T. Tsuda, T. Saegusa, Hashimot.T, J.
Am. Chem. Soc. 1972, 94, 658-659.

[10]

a) I. A. Fisher, A. T. Bell, J. Catal. 1997, 172, 222-237 b) S. Kattel, B. Yan, Y. Yang, J. G. Chen, P. Liu, The Journal of the American
Chemical Society 2016, 138, 12440-12450.

;

;

S32