Atomic overlayer of permeable microporous cuprous oxide on palladium promotes hydrogenation catalysis

The interfacial sites of metal-support interface have been considered to be limited to the atomic region of metal/support perimeter, despite their high importance in catalysis. By using single-crystal surface and nanocrystal as model catalysts, we now demonstrate that the overgrowth of atomic-thick Cu2O on metal readily creates a two-dimensional (2D) microporous interface with Pd to enhance the hydrogenation catalysis. With the hydrogenation confined within the 2D Cu2O/Pd interface, the catalyst exhibits outstanding activity and selectivity in the semi-hydrogenation of alkynes. Alloying Cu(0) with Pd under the overlayer is the major contributor to the enhanced activity due to the electronic modulation to weaken the H adsorption. Moreover, the boundary or defective sites on the Cu2O overlayer can be passivated by terminal alkynes, reinforcing the chemical stability of Cu2O and thus the catalytic stability toward hydrogenation. The deep understanding allows us to extend the interfacial sites far beyond the metal/support perimeter and provide new vectors for catalyst optimization through 2D interface interaction.

It is unclear what was the thickness of the Pd nanosheets (2 nm as in ref. 21?), and also of the Cu-oxide film on Pd. For the latter ref. 23 says it is about 3-6 layers. If so, I wonder how such a "thick" film is permeable for alkynes to reach the metal surface. For STM studies, the Cu2O surface was prepared at UHV compatible pressures, and the treatments were also done at low pressures. I wonder whether a well-defined Cu-oxide layer used for calculations remains under H2-rich reaction conditions. The formation of Cu+ in the post-reacted samples could be explained by oxidation of Cu-Pd alloy in air. Figure 2c highlights the pretreatment effect (either H2 or alkyne prior to the reaction) but does not show the results for "normal" reaction when both reactants are exposed simultaneously. I anticipate the result to be similar to that with alkyne pre-adsorption. If so, I would turn this other way around: the pretreatment with pure H2 reduces and eventually destroys the Cu-oxide layer and hence the active sites rather than alkyne "serves as an unprecedented modifier" (p. 10).
As to other alkynes tested, should the reactivity also depend on the alkyne size and triple bond position to adsorb on the metal surface in the "pore"?
What is the "electronic modulation" of the Pd-Cu surface the author talk about in the abstract?
Overall, I think the hypothesis sounds interesting, but at present it is solely based on DFT calculations rather than experiment.
Reviewer #2 (Remarks to the Author): This paper investigates a well-defined system of Pd nanocrystals decorated with Cu overlayer for the semihydrogenation of aryl alkynes such as phenylacetylene. By altering the ratio of Cu/Pd, the authors identify an important change in the speciation of Cu, as quantities of Cu below a monolayer (Cu/Pd < 0.5) prefer to migrate subsurface and above a monolayer (Cu/Pd > 0.5), excess Cu is left on the Pd surface to form Cu2O phase after exposure to air. The formation of some Cu2O phase is identified as being crucial to the highly selective, active, and stable conversion of phenylacetylene to styrene over PdCu@Cu2O. Furthermore, the authors identify that a secondary catalyst effect responsible for the increased stability of Cu2O even in reducing atmospheres is the stabilization of Cu2O phase by dissociatively adsorbed phenylacetylene, which locks in the Cu2O phase by complexing with Cu(I) and increasing the barrier for reduction, thus enabling the active interface between Cu2O and Pd to be stable under reaction conditions. The work follows a very detailed and logical progression and highlights a unique and powerful example of substrate-modified catalyst behavior, which is not often understood. This paper should be published in Nature Communications after a few issues are clarified.
1. The proposed mechanism involves facile adsorption of H2 onto Pd below the Cu2O overlayer due to the large diameter of the Cu2O pore size (5.5 A). However, in similar cases of SMSI using other oxides, small molecule adsorption (CO or H2) is often cited to be highly suppressed. It would be interesting to see what the H2 uptake of PdCu@Cu2O catalysts were for comparison against similar SMSI catalysts to understand the role of Cu2O porosity. Furthermore, this study would be even more illuminating after phenylacetylene pre-treatment, as H2 uptake after phenylacetylene exposure would give insight into whether or not these two species can be co-adsorbed in significant quantities, or if phenylacetylene effectively blocks H2 adsorption. Such site blockage by phenylacetylene might have important impacts on the observed selectivity as well. 2. The reduction of Cu2O at 30˚C is attributed solely to defective sites on the Cu2O lattice, but the possibility of spillover from Pd domains which are not decorated (as evidence by EDS, etc.) which should near-barrierlessly activate H2 and thus facilitate Cu2O reduction even at low temperatures is not explored or discussed. The presence of these defective Cu2O sites is also only assumed and used frequently in theoretical models, but their presence is not necessarily established by any of the characterization. Considering spillover for facilitated reduction is well-documented on multiple systems involving well-mixed noble metals/metal oxides, this may play a larger role than what is addressed here by the authors.
3. Citations 41-43 are works which also investigated dissociative phenylacetylene adsorption, sometimes on Cu2O catalysts, using FT-IR. However, in none of those papers is an IR active mode for Cu:CCPh identified. Citations 43 reports that there is a loss of both the alkyne and alkynal C-H vibration mode upon terminal adsorption. These authors identify the same loss in alkynal C-H, but make an identification of a Cu:CCPh mode which has not been previously been identified, at least in the citations given. This may suggest it is either not associated with the Cu:CCPh complex, or that there is a different geometric bonding mode which makes this vibration IR active. 4. The Cu XPS spectra in Sup. Fig. 36a are not labelled clearly enough to distinguish that there is any change to the materials. The issue is that the energy axis scale is so large compared to the magnitude of the shift that it's hard to visually see the shift which is important to follow the argument being verbally made by the authors. Perhaps including numbers (like in Sup. Fig. 6a) above the Cu XPS signals to highlight that they are shifted or make use a separate zoom where the shifts or changes to the signals are more noticeable.

General Comment:
The manuscript by Liu et al. is not presented in a wellunderstandable way and it was difficult for me to follow. The language has to be improved. A few examples: on p.3: "scarce metals" (?), "…model catalysts … were applied as model catalysts". The sentence on p. 5 begins with "And…". Response: Thank the reviewer for pointing out the language problem.
Following the suggestion, we have tried our best to revise the manuscript accordingly so that it is presented more clearly.
2. Comment: Below this figure (on p.7) the authors state: "It has been widely accepted (?!) that the oxide overlayer should not fully cover the metal surface due lattice mismatch…." I do not understand how the film coverage correlates with the lattice mismatch.

Comment:
As far as I understood, the main results can be summarized as follows. Pd nano-sheets (flakes) alloyed with Cu and then oxidized in air become more active in selective hydrogenation (i.e. alkyne to alkene and not to alkane). The catalyst pretreatment with alkyne (phenylacetylene) was beneficial as compared to that with pure H2. The enhanced reactivity was assigned to the formation of a two-dimensional Cu2O layer which is permeable for hydrogen and alkyne reacts on the Pd1-xCux(111) alloy surface underneath.
Accordingly, the pretreatment effect was explained by that Cu-oxide reduction is suppressed by strongly bound alkynes on defect sites. To support the conclusions, the authors provide low-temperature STM images of the Pd (111) single crystal surface with an ultrathin Cu2O layer grown on top, after several treatments in the ~ 10 -6 mbar pressure range. Although the topic of this study is interesting as it demonstrates the promotional effect of thin oxide layers on reactivity of metal surfaces, the conclusions are primarily based on theoretical calculations rather than solid experimental data. 2) As shown in Fig. 1e, 2c and Supp. Figs. 8 to 10, 22, 36, the difference in catalytic performance and Cu LMM XAES spectra confirmed the defective Cu2O on PdCu@Cu2O was easily reduced to Cu in H2 at 30˚C. Once Cu2O was reduced, the catalyst lost its semi-hydrogenation selectivity. Thus, the semi-hydrogenation reactions should take place over 2D Cu2O/Pd interface stabilized by alkynes. Fig. 3a and Supp. Figs. 26 to 31, 36, the reduction of Cu2O overlayer by H2 was dramatically suppressed after the defective sites were "locked" by Cu(I)-C≡CPh groups, indicating that PhC≡CH served as an unprecedented modifier for stabilizing 2D Cu2O/Pd catalytic interface for the semi-hydrogenation of alkynes. Fig. 3a and Supp. Fig. 27, the PhC≡CH that participate the semi-hydrogenation was in molecular form. 5) Nuclear magnetic resonance (NMR) and in situ FT-IR analysis confirmed that the Cu(I)-O-H would not participate the semi-hydrogenation (Supp.
With all these results, we consider that the proposed mechanism was well supported by both experiments and calculations.  Fig. 36a, b).

Comment
However, when there were Cu(I)-C≡CPh motifs at the defective sites of surface Cu2O, the reduction of Cu2O overlayer by H2 was dramatically suppressed because the removal of interfacial oxygen by H atom has to overcome a much high barrier of 1.85 eV (Supp. Figs. 35-36). All the above results clearly illustrated that Cu(I)-C≡CPh served as an unprecedented modifier for stabilizing surface Cu2O and 2D Cu2O/Pd interface for the semi-hydrogenation of alkynes.
The electronic structure and coordination environment of PdCu@Cu2O remained unchanged during the reaction. Thus, the formation of Cu + in the postreacted samples should not be caused by the oxidation of Cu-Pd alloy in air. 6. Comment: Figure 2c highlights the pretreatment effect (either H2 or alkyne prior to the reaction) but does not show the results for "normal" reaction when both reactants are exposed simultaneously. I anticipate the result to be similar to that with alkyne pre-adsorption. If so, I would turn this other way around: the pre-treatment with pure H2 reduces and eventually destroys the Cu-oxide layer and hence the active sites rather than alkyne "serves as an unprecedented modifier" (p. 10).

Response:
Following the reviewer's comment, we further considered the catalytic performance of PdCu@Cu2O for "normal" reaction when H2 and PhC≡CH are exposed simultaneously. As shown in Fig. R1, when the conversion of PhC≡CH reached 100%, the selectivity toward styrene was only 91.5% for "normal" reaction, which was much lower than that of PhC≡CH pretreatment (96.9%). Moreover, by extending the reaction time, the selectivity towards styrene would gradually decay for "normal" reaction, different from that of PhC≡CH-pretreated PdCu@Cu2O catalyst. These results suggested the formation of Cu-C≡CPh structure should be suppressed in the presence of H2 at "normal" reaction. To further test the influence of steric effect, 2-ethynyltoluene and its positional isomers were chosen as the model reagents. As shown in Fig. R2b, the reactivity was increased from ortho to meta, to para arrangements, and the highest activity was achieved for para-substituted toluene, indicating the "pore" at the PdCu@Cu2O interface would exhibit certain shape selectivity. Response: Part of copper is diffused into palladium lattice to form near surface Pd-Cu alloy whose electronic structure would influence the binding of activated H atoms on metal. The alloyed Pd-Cu surface exhibited weak adsorption of dissociated H atoms so that the reactivity of the hydrogenation activity was enhanced. To illustrate the "electronic modulation" of the Pd-Cu alloy, additional calculations on Pd@Cu2O were conducted, and the energy barrier of TS1 was predicted to be 1.18 eV (Fig. R3), much higher than that of PdCu@Cu2O (0.67 eV). 1. Comment: The proposed mechanism involves facile adsorption of H2 onto Pd below the Cu2O overlayer due to the large diameter of the Cu2O pore size (5.5 A). However, in similar cases of SMSI using other oxides, small molecule adsorption (CO or H2) is often cited to be highly suppressed. It would be interesting to see what the H2 uptake of PdCu@Cu2O catalysts were for comparison against similar SMSI catalysts to understand the role of Cu2O porosity.

Response:
Following the reviewer's suggestion, we compared the adsorption energies of CO and H2 on different surfaces/interfaces, see Figure R4 and Table R1. For CO adsorption, the predicted adsorption energies increased from Pd(111) (-2.07 eV), to PdCu (111) (-1.83 eV), to PdCu@Cu2O (-0.95 eV). And, the same tendency can be found for dissociated adsorption of H2, in which the adsorption energies for the two H atoms were predicted to be -1.03 eV, -0.87 eV and -0.21eV for Pd(111), PdCu (111) and PdCu@Cu2O, respectively. All these findings indicated that the Cu2O overlayer does suppressed the adsorption of both CO and H2, in line with the expectation from SMSI effect.  Response: Thank the reviewer for the comment. As described in the manuscript, the PhC≡CH pre-treatment plays a vital role in stabilizing the 2D Cu2O/Pd interface, and does not block the adsorption and activation of H2 on PdCu@Cu2O. H2-D2 exchange was employed to further characterize the H2 dissociation activity over the PhC≡CH pre-treated PdCu@Cu2O catalyst at ambient pressure. As shown in Fig. R5, HD was produced right after H2 and D2 were co-introduced to the PhC≡CH pre-treated PdCu@Cu2O, indicating the active sites were still accessible for H2 even with protection of PhC≡CH.   (Fig. 2c, 3b, c and Sup. Fig.   37), clearly indicating that the spilled H from Pd domains to the defective Cu2O does not contribute to the high selectivity. Furthermore, the reduction of Cu2O overlayer by H2 would be dramatically suppressed after the defective sites were "locked" by Cu(I)-C≡CPh groups (Sup. Figs. 35 and 36). Therefore, the 2D Cu2O/Pd catalytic interface for the semi-hydrogenation of alkynes should be protected by Cu(I)-C≡CPh groups before exposed to H2.
However, in none of those papers is an IR active mode for Cu:CCPh identified.
Citations 43 reports that there is a loss of both the alkyne and alkynal C-H vibration mode upon terminal adsorption. These authors identify the same loss in alkynal C-H, but make an identification of a Cu:CCPh mode which has not been previously been identified, at least in the citations given. This may suggest it is either not associated with the Cu:CCPh complex, or that there is a different geometric bonding mode which makes this vibration IR active.
Response: Thank the reviewer for the comment. In situ TPD-MS characterization confirmed that the new species on PdCu@Cu2O-used was the dissociated PhC≡C-, and the dissociated PhC≡C-was related to Cu (Sup. Figs. 28-31). The configuration of Cu(I)-C≡CPh structure can be identified by comparing in situ FT-IR spectra. The alkynyl C-H stretching peaks of the free ligands at ∼3300 cm -1 disappear upon adsorption, indicating that the alkynyl ligands (PhC≡C-) should be bound to PdCu@Cu2O (type 1 or 2) ( Fig. 3a and Sup. Fig. 27). In principle, the non-polar group -C≡C-stretching vibration is IR inactive while Raman active. Thus, we also used surface-enhanced Raman scattering (SERS) spectra to detect the vibration of -C≡C-stretching. It was observed that -C≡C-stretching is red-shifted from 2110 cm -1 to 1971 cm -1 (Sup.

Comment:
The Cu XPS spectra in Sup. Fig. 36a are not labelled clearly enough to distinguish that there is any change to the materials. The issue is that the energy axis scale is so large compared to the magnitude of the shift that it's hard to visually see the shift which is important to follow the argument being verbally made by the authors. Perhaps including numbers (like in Sup. Fig. 6a) above the Cu XPS signals to highlight that they are shifted or make use a separate zoom where the shifts or changes to the signals are more noticeable.
Response: Thanks for the suggestion. Following the suggestion, the Cu XPS spectra in Sup. Fig. 36a has been revised to highlight the shift.