CO oxidation by the atomic oxygen on silver clusters: structurally dependent mechanisms generating free or chemically bonded CO₂

Abstract

Atomic oxygen on silver is the crucial active species in many catalytic oxidation processes, while it is a big challenge to explore the relationship between its activity and molecular-level structures in condensed phases. We carried out kinetic measurements of the gas phase reactions between Ag_nO⁻ (n = 1–8) and CO, in which the oxygen atoms were predicted to be terminal ones in AgO⁻ and Ag₂O⁻, in quasi-Ag–O–Ag chains for Ag₃O⁻ and Ag₄O⁻, and on the two-fold or three-fold bridging positions in Ag_nO⁻ (n = 5–8). All these oxygen species are highly reactive even at a low temperature of 150 K. Ag_nO⁻ (n = 1, 2, 5–8) with terminal or bridging oxygen generate free CO₂, while the quasi-chains of Ag_nO⁻ (n = 3, 4) generate chemically bonded CO₂ with a structural formula of Ag_n–CO₂–Ag₂⁻ (n = 1, 2). Density functional theory calculations well interpreted all experimental observations, showing that no extra excitation energies are needed to initiate all these reactions. The structurally dependent mechanisms and the formation of chemically bonded CO₂ revealed in this work help us to catch a glimpse of some important processes and intermediates on real silver catalysts.

---

1. Introduction

Silver catalysts are used in many oxidation processes with great industrial importance, for example in the partial oxidation of ethylene or propylene to their oxides^1–3 and CO selective oxidation in H₂-rich gas.^4–8 To elucidate the underlying catalytic mechanisms, the states and reactivity of oxygen species on silver have attracted great attention in surface science. The atomic oxygen on bulk silver was firstly noticed about one century ago,⁹ and people realized that these adsorbed oxygen atoms are crucial active species in most of the catalytic oxidation processes as early as the 1940s.^{1,2,6,7,10–15} With the help of modern analysis techniques in surface science, atomic oxygen on silver is now divided into three types: the weakly adsorbed oxygen atoms on surfaces (O_α), the ones inside bulks (O_β), and those strongly adsorbed in crevices or embedded in surfaces (O_γ).^13,16–20 Some experimental hints and theoretical calculations showed that each of these three types include multiform structures. For examples the surface oxygen atoms (O_α and O_γ) could be top oxygen on a single silver atom,^18,21 the bridging oxygen on normal Ag–Ag bonds,²¹ the ones on three or high fold hollow positions,^{3,18,19,21–23} and those bridging on long Ag–Ag bonds or in –Ag–O–Ag– chains.^17,24–27 However, it is a big challenge to prepare one kind of oxygen with a certain defined structure and characterize its activity on bulk surfaces. Oxidation of CO is a clean-off process used to characterize the reactivity of oxygen species on metals. On model silver surfaces like Ag(110), CO reacts with O_α at 100–200 K and with O_γ at above 250 K.^8,28–31 The optimum temperatures for CO oxidation on various silver catalysts extend from room temperature to higher than 400 K.^11,32–36 Determining the operation temperatures of these real catalysts has taken into account other elementary steps, such as generation, transformation and migration processes of the oxygen atoms. At present, most analyses on these complicated processes are based on the aforementioned oversimplified classification rather than the molecular level structures of the surface oxygen atoms.

Metal clusters in the gas phase are ideal models for active sites of real metal catalysts,^37–42 and studies on their reactions with small molecules provide important information to understand the heterogeneous catalytic mechanisms.^43–51 Adsorption and activation of O₂ on gas phase silver clusters were shown to be dominated by clusters' electronic properties.^52–59 The CO oxidation by the O₂ molecule was observed on Ag_n⁻ (n = 7, 9, 11), forming two free CO₂ molecules.⁵⁴ The mechanism for similar catalytic circles was theoretically investigated on Ag₂⁻, Ag₁₃, Ag₅₅, etc.^60–64 The calculations proposed the formation of one OOCO intermediate, which subsequently generated an adsorbed oxygen atom and a free CO₂. The adsorbed oxygen atom on silver is expected to form another free CO₂ with a second CO. Experimentally, the atomic oxygen on silver clusters was generated in the reactions between Ag_n^± and O₂ or NO.^52,53,65,66 Nevertheless, there have been no experimental efforts showing the structure–activity relationships of these adsorbed oxygen atoms, even though they are involved in many proposed reaction mechanisms and are directly related to the active oxygen species on real silver catalysts.

In this work, we generated a series of model clusters Ag_nO⁻ (n = 1–8) in the gas phase. The adsorption structures of their oxygen atoms are quite similar to the predicted ones of those on bulk silver surfaces. The reactions between these model clusters and CO were investigated at a low temperature, which generated free and surprising chemically bonded CO₂ units. The theoretical calculations provided good interpretation of all experimental observations, revealing the structurally dependent reaction mechanisms of various atomic oxygen atoms on silver.

2. Experimental and computational methods

2.1 Kinetic measurements on the gas phase reactions

The kinetic measurements on cluster reactions were carried out on an instrument composed of a magnetron sputter cluster source, a flow reactor, and a time-of-flight (TOF) mass spectrometer. The details of this instrument were described elsewhere.^59,66 Briefly, the cluster series Ag_nO⁻ (n = 1–8) were generated by mixing a trace amount of oxygen (at a ppm level) inside the helium buffer gas in the cluster source. The O₂ molecules dissociated and combined with silver species in the sputtering process. Together with the buffer gas (110 sccm helium) and the sputtering gas (14 sccm Ar), all clusters entered a continuous flow reactor running at 150 K. They firstly went through a thermalization region and then entered the reaction area, where the CO reactant was introduced at defined flow rates. Since the oxygen mixed inside the buffer gases was at the ppm level, its effect on the subsequent reactions inside the reactor was neglected. The parent and product clusters went through a skimmer at the end of the reactor and were directed to the TOF mass spectrometer to be analyzed.

During one set of measurements in the experiment, the flow conditions inside the reactor were kept constant, which implies that the total pressure, the concentration of CO, and the residence time were identical for all cluster sizes. Since the concentrations of metal clusters were much lower than that of CO, their reactions were analyzed based on the pseudofirst order mechanism. Therefore, the relative rate of each cluster size should be proportional to the slope value in linear fitting of −ln(I/I₀) vs. CO flows. Here, I and I₀ stand for the intensities of parent ions with and without the CO reactant, respectively. The slopes of Ag_nO⁻(n = 2–8) were further scaled to that of AgO⁻. The absolute values of the bimolecular rate constants were estimated by comparing the present measurements with that of the reaction between Au₂⁻ and O₂, whose fitting slope under the present conditions and bimolecular rate were previously obtained.^59,67

2.2 Density functional theory (DFT) calculations

The clusters' structures and their reaction paths were investigated using DFT calculations. The B3LYP hybrid functional was used, which makes use of the Hartree–Fock exact exchange and Becke's exchange functional and the Lee–Yang–Parr correlation functional.^68–70 Nearly all possible structures were considered for cluster complexes containing not more than four silver atoms. For Ag_nO⁻ (n ≥ 5), their structural candidates were manually built by putting the oxygen on various positions of many previously reported structures of naked silver clusters. All the structural candidates of Ag_nO⁻ (n = 1–8) were initially optimized with the Lanl2dz basis sets for all elements. The lower lying ones from these preliminary calculations were further optimized using a more expensive method, in which we used the Aug-cc-pvtz-pp basis set for Ag^71,72 and the 6-311g* basis sets for C and O.⁷³ Scalar and spin-orbital relativistic effects of Ag were taken into account via the energy consistent relativistic pseudopotentials.⁷² The obtained structures for the minimum points and the transition states were confirmed by analyzing their vibrational modes, which have no and only one imaginary frequency, respectively. The charge distributions on these structures were obtained using natural bond orbital (NBO) analysis.⁷⁴ Using the intrinsic reaction coordinate (IRC) algorithm, we followed the reaction paths from each transition state to the corresponding reactants and products to reconfirm the proposed reaction mechanism. All calculations were accomplished using Gaussian 09 program.⁷⁵

3. Results and discussion

3.1 Kinetic measurements of the reactions between Ag_nO⁻ (n = 1–8) and CO

Fig. 1(a) shows the mass spectrum of Ag_n⁻ (n = 1–11) and Ag_nO⁻ (n = 1–8) generated by the cluster source with a trace amount of O₂ mixed in the buffer gas. The formation processes of Ag_nO⁻ were very complicated, which involved multistep reactions in the energy rich plasma area. We did not have full control of these processes and can only generate Ag_nO⁻ with n = 1–8 even using optimized conditions. Fig. 1(b) and (c) show the reaction products of these Ag_nO⁻ and CO at flows of 0.02 sccm and 0.20 sccm, respectively. The adsorption of CO on Ag_n⁻ was observed on the sizes with n ≥ 11, as indicated by the small peak of Ag₁₁CO⁻ in Fig. 1(c). A tentative interpretation is that CO is overall an electron donating ligand and the dilute negative charge on large silver clusters makes them easier to adsorb CO than the small sizes. The reactions between Ag_nO⁻ (n = 1, 2, 5–8) and CO were accompanied by increasing the corresponding Ag_n⁻, indicating the formation of these naked silver clusters and free CO₂ molecules. For Ag_nO⁻ (n = 3, 4), the intensities of their Ag_n⁻ were roughly constant when CO was introduced, and the reactions between these two Ag_nO⁻ and CO generated new complexes with the stoichiometric formula of Ag₃CO₂⁻ and Ag₄CO₂⁻. It is worth noting that all Ag_n⁻ (n = 1–8) did not adsorb CO₂, even though we introduced high flow CO₂ (>0.20 sccm) into the reactor. That is to say, the observed Ag_nCO₂⁻ (n = 3, 4) should have some novel structure with a strongly bonded CO₂ unit other than those with a CO₂ molecule weakly attached on Ag₃⁻ or Ag₄⁻. We noticed that the reaction of Ag₅⁻ with CO and O₂ generated one Ag₃CO₂⁻ fragment.⁵⁴ Survival of Ag₃CO₂⁻ in the complicated fragmentation processes as well indicated the strong interaction between the CO₂ unit and the silver moiety in this complex.

Fig. 1 Mass spectra showing the products of the reactions between Ag_nO⁻ (n = 1–8) and CO with flow rates of (a) 0.00 sccm, (b) 0.02 sccm and (c) 0.20 sccm at 150 K. The peak positions of Ag_n⁻ (n = 1–8) are indicated with vertical dashed lines; the Ag_nO⁻ series and other clusters are indicated using the extra atom or group added to the corresponding Ag_n⁻. The small peaks of AgOH⁻ in (c) are also distinguished in the spectra of (a) and (b), which could come from moisture impurities inside the cluster source.

Fitting the decay processes of Ag_nO⁻ (n = 1–8) under the assumption of the pseudofirst order mechanism is shown in the ESI† (Fig. S1). Fig. 2 displays the results of two example sizes Ag₂O⁻ and Ag₃O⁻, including the linear fitting of −ln(I/I₀) vs. CO flows as well as the corresponding mass spectra for all datum points. The obtained slope values and their uncertainties were indicated. After taking into account all uncertainties in fitting and calibration processes, the bimolecular rate constant for AgO⁻ was estimated to be around 3.0 ± 1.5 × 10⁻¹¹ cm³ s⁻¹ at 150 K, which is several percentages of its total collision rate with CO specified by the Langevin theory (∼7 × 10⁻¹⁰ cm³ s⁻¹).^76,77 The slopes of Ag_nO⁻ (n = 2–8) were scaled relative to that of AgO⁻, and are summarized in Fig. 3. The relative rates of Ag_nO⁻ (n = 2–8) are lower than that of AgO⁻, and have a slight even–odd oscillation trend. The sizes with even n (or odd number of total valence electrons) generally have higher rates than their neighbors. However, the kinetic rates of all cluster sizes, including those of Ag₃O⁻ and Ag₄O⁻ which generated strongly bonded CO₂, are roughly in the same order of magnitude. In another word, the size dependence of these clusters' reactivity with CO seems insignificant.

Fig. 2 Linear fitting of −ln(I/I₀) vs. CO flows in the kinetic measurements of the reactions between Ag_nO⁻ (n = 2, 3) and CO at 150 K. I and I₀ stand for the intensities of Ag_nO⁻ with and without CO, respectively. The corresponding mass spectra for all datum points are shown on the right column. The obtained slope values and their uncertainties are indicated.

Fig. 3 Relative kinetic rates of Ag_nO⁻ (n = 1–8) scaled to that of AgO⁻. The estimated bimolecular rate of AgO⁻ was estimated to be 3.0 ± 1.5 × 10⁻¹¹ cm³ s⁻¹. The dashed circles indicated the data of Ag₃O⁻ and Ag₄O⁻, which generated chemically bonded CO₂.

3.2 Exploring the reaction paths forming free or chemisorbed CO₂

The lower lying structural candidates of Ag_nO⁻ (n = 1–8) from DFT calculations are summarized in the ESI,† Fig. S2. The lowest lying ones as well as their symmetries and electronic states are shown in Fig. 4. The oxygen atoms in these lowest lying structures are on the terminal positions for the sizes of n = 1, 2, on the two-fold bridging positions for the sizes with n = 5 and 7, and on the triangle hollow positions for those with n = 6 and 8. The lowest lying structures of Ag₃O⁻ and Ag₄O⁻ are special, in which the distances between the two silver atoms connecting the oxygen atom are apparently longer than a usual Ag–Ag bond. Therefore they can be considered as a class of quasi-Ag–O–Ag chains. The electronic states of Ag_nO⁻ (n = 1–8) oscillate between singlet and triplet with increasing cluster sizes. We also calculated the spins and negative charges localized on the oxygen atoms of these structures in Fig. 4 and summarized these results in the ESI,† Fig. S3. The total spins of Ag_nO⁻ and the localized spins and charges on their oxygen atoms vary widely for different sizes. However, we did not see evident correlations between clusters' kinetic rates and these electronic properties. Comparing the predicted structures of Ag_nO⁻ in Fig. 4 and the reaction channels revealed by the mass spectra in Fig. 1, we found that the quasi-Ag–O–Ag chains generated bonded CO₂ on silver, while the other structures generated free CO₂ molecules. In the ESI,† Fig. S4, we listed the theoretical binding energies of CO₂ on Ag_n⁻ (n = 1–8), among which Ag_n⁻ (n = 1, 2, 5–8) have the structures similar to the silver frames of Ag_nO⁻ and Ag_n⁻ (n = 3, 4) are their lowest-lying structures predicted by the present and previous theoretical calculations.^78–80 On all these naked Ag_n⁻, the adsorption strengths of CO₂ are fairly weak, so they cannot tolerate the large amount of energy released during the formation process, and cannot compensate the entropy losses relative to the desorption states either. These weak interactions well interpret the experimental facts that Ag_nO⁻ (n = 1, 2, 5–8) formed free CO₂ in their reactions with CO, and naked Ag_n⁻ (1–8) did not adsorb CO₂ even at high CO₂ flows. These theoretical results also confirmed the assumption that Ag₃CO₂⁻ and Ag₄CO₂⁻ present in the experiments have some special structures other than the molecularly adsorbed ones in Fig. S4 (ESI†). We extensively searched other structural candidates with the stoichiometric formula of Ag_nCO₂⁻ (n = 3, 4) and the lower lying ones are listed in the ESI,† Fig. S5. The lowest one of each size is displayed in Fig. 5, which has the structural formula of Ag–CO₂–Ag₂⁻ or Ag₂–CO₂–Ag₂⁻. Their charge distributions and dissociation energies relative to various fragments were also listed. According to the NBO charge analyses, the total charges on the CO₂ units of these two structures amount to −0.61 a.u and −0.59 a.u, respectively. These apparent negative charges on the CO₂ units were evidenced by their bending structures. The dissociation energies in Fig. 5 indicated that breaking the Ag–C or O–Ag bond in Ag–CO₂–Ag₂⁻ or Ag₂–CO₂–Ag₂⁻ needs around 0.9–2.2 eV, which are in the range of typical chemical bonds. In a sense, these characteristics of the CO₂ units in the products of Ag_nO⁻ (n = 3, 4) make them more or less like the ester groups in organic complexes.

Fig. 4 The lowest lying structures of Ag_nO⁻ (n = 1–8) from DFT calculations. Their geometries and electronic states are indicated.

Fig. 5 The lowest lying structures of Ag_nCO₂⁻ (n = 3, 4) with chemically bonded CO₂. Their symmetries, electronic states, and the NBO charges (in a.u.) on each atom are indicated. The dissociation energies (in eV), i.e. the energy differences between Ag_nCO₂⁻ (n = 3, 4) and the corresponding fragments, are also listed.

Using the DFT calculations, we also explored the reaction paths between the lowest lying structures of Ag_nO⁻ (n = 1–8) and CO. Fig. 6 displays the structures and the relative energies of the involved reactants, intermediates, transition states, and final products. In one kind of the considered reaction path, CO is pre-adsorbed on silver sites around the oxygen of Ag_nO⁻ (n = 1–8), and then the reaction between the co-adsorbed CO and O occurs (the Langmuir–Hinshelwood mechanism). The calculations showed that CO can strongly adsorb on one silver site of Ag_nO⁻ (n = 1, 3), forming a quasi-linear OC–Ag–O structure. Adsorption of CO on the silver sites of Ag₅O⁻ directly leads to the formation of an adsorbed CO₂. On Ag_nO⁻ (n = 2, 4, 6, 7 and 8), CO interacts very weakly with all silver sites and even tends to move to the oxygen atom in Ag₄O⁻. Except the Ag₅O⁻ which directly forms an adsorbed CO₂, the transition states of all other Ag_nO⁻ have roughly the same energies as those of the initial reactants (Ag_nO⁻ and CO). In the other kind of reaction paths shown in all the panels of Fig. 6, CO is initially put around the oxygen atoms of Ag_nO⁻ (n = 1, 2, and 5–8) or the Ag–O bonds in Ag_nO⁻ (n = 3, 4) (the Eley–Rideal mechanism). Then, the products with physically adsorbed or chemically bonded CO₂ are directly formed, and no apparent barriers can be distinguished in the reaction processes. In brief, the two kinds of reaction paths in each panel of Fig. 6 indicated that no extra excitation energies are needed for the reactions between Ag_nO⁻ (n = 1–8) and CO. This theoretical prediction is consistent with the experimental observation that all reactions were noticeable even at 150 K. The slight even–odd oscillation of the reaction rates in Fig. 3 possibly come from some subtle effects of cluster's spins on the reaction dynamics, which were not included in the reaction figures shown in Fig. 6.

Fig. 6 Theoretical paths for the reactions between Ag_nO⁻ (n = 1–8) and CO. The structures and the relative energies of the reactants, the intermediates, the transition states, and the products are included.

It is noteworthy that, except AgO⁻, the DFT calculations predicted many local minima of Ag_nO⁻ (n = 2–8) other than their global minima. According to the results shown in Fig. S2 (ESI†), the second lowest lying structures of Ag_nO⁻ (n = 2, 3, 5, and 7) are more than 0.15 eV higher than their global minima. Generally, this energy difference is large enough making the lowest structures of small coinage metal clusters dominate in the experiments.^81,82 For Ag_nO⁻ (n = 4, 6, and 8), there are isomers very close in energy to their global minima at the present theoretical level. For Ag₄O⁻, the second lowest lying structure is a rhombic structure with the atomic oxygen bridging on one side (in Fig. S2, ESI†). Theoretical exploration on its reaction with CO indicated that it tends to form Ag₄⁻ and a free CO₂. The dominant product of Ag₂CO₂Ag₂⁻ in the experiments implies that this rhombic structure of Ag₄O⁻ did not appear or was insignificant compared with its lowest lying quasi-chain structure. The lower lying isomers of Ag_nO⁻ (n = 6 and 8) likewise contain two fold or three fold bridging oxygen atoms. Theoretical calculations indicated that the reactions between these isomers and CO form Ag_n⁻ (n = 6 and 8) and CO₂, which is the same as what happens for their lowest lying structures. The present results cannot distinguish if the lower lying isomers of Ag_nO⁻ (n = 6 and 8) appear in the experiments.

3.3 Comparing the gas phase observations with surface reactions on bulk silver

In the present gas phase study, both the CO reactant and the CO₂ product did not adsorb on Ag_n⁻ (n = 1–8), which is parallel to previously observed weak adsorptions of CO and CO₂ on clean bulk silver.^12,83 Generally, the d–s band gaps in silver systems are larger than those in the corresponding species of their congener elements and the early transition metals. The fully occupied and relatively low lying d bands of silver act against their interactions with the frontier orbitals of CO ligands, and therefore do not favor the formation of strong chemical bonds. In this study, the reactions between all Ag_nO⁻ (n = 1–8) and CO were noticeable even at 150 K. We know that the effective temperature for an isolated cluster-ion system is generally different from that of the initial reactants, because the system is a micro-canonical ensemble and the extra energies released during combination of the reactants will dramatically increase the effective temperature value. This scenario is true for AgO⁻ and Ag₃O⁻, which formed linear OC–Ag–O structures with CO binding energies of more than 1.0 eV (shown in Fig. 6). However, CO directly formed adsorbed CO₂ on Ag₅O⁻ and very weakly adsorbed on Ag_nO⁻ (n = 2, 4, 6, 7 and 8). The CO binding energies on all positions of Ag_nO⁻ (n = 2, 4, 6 and 8) were even close to zero. Therefore, the effective temperatures for these reaction systems should be close to the preset one of the initial reactants (∼150 K). On bulk silver surfaces, the reactions between most atomic oxygen species and CO also occur below 200 K.^4,8,29,30 Exceptionally, the terminal oxygen atoms in the bulgy O–Ag–O–Ag– chains react with CO at around 200 K, and the embedded middle ones in these chains react at around 300 K.^26,28,31 The relatively low reactivity of the middle oxygen should relate to some extra barriers for chain reconstruction processes accompanying the reaction processes.^26,27,84 For all gas phase Ag_nO⁻ (n = 1–8), their reactions with CO were predicted to have negligible barriers. The similar CO oxidation processes have low barriers of 0.1–0.3 eV on Ag(110), Ag(111), and Ag(221) surfaces.^29,85,86 The noticeable reaction rates at low temperatures and the relatively low reaction barriers for both the gas phase reactions and the corresponding surface processes indicate their correlations and similarities.

The reactions between CO and atomic oxygen on bulky silver generally form free or weakly adsorbed CO₂. The generated CO₂ has additional chances to react with other surface oxygen species, and forms the widely observed surface carbonate groups.⁸³ In the gas phase experiments, the concentrations of the cluster reactants and the formed free CO₂ were extremely low, and it was impossible to see their subsequent reactions. It is interesting to notice that many surface studies also observed strongly bonded CO₂ on bulk silver, which were desorbed at temperatures higher than 450 K.^27,84,87,88 The STM experiments have shown the structures of –Ag–CO₂–Ag–⁸⁸ and –Ag–Ag–CO₂–Ag–Ag–.^27,84 In –O–Ag–CO₂–Ag–O–, the CO₂ unit was assumed to be on a positive silver position and its bonding interaction with the surface also involved the two neighboring negative oxygen atoms. In the structure of –Ag–Ag–CO₂–Ag–Ag–, the CO₂ unit was on the added silver chain of clean bulk surfaces, and appeared as a protrusion light spot between silver atoms in the STM images. There has been little information for its bonding details. The bonding pattern in the stable structures of Ag₃CO₂⁻ and Ag₄CO₂⁻ provided one possibility for that of the CO₂ in the latter structural case. What is more, the generation process for the observed chemically bonded CO₂ on bulk silver is not clear either. One proposed mechanism is that a surface carbonate group reacts with a CO molecule, forming one free and one chemisorbed CO₂.²⁷ The gas phase experiment in this study showed that CO can react with the oxygen atoms in quasi-Ag–O–Ag chains, and directly formed chemically bonded CO₂. Since CO is among the most common reactants or impurities appearing in the environments of surface studies, and the –Ag–O–Ag– chain structures were widely observed on bulk silver, the reaction path revealed in this study could exist in surface experiments.

4. Conclusions

We carried out kinetic measurements on the reactions between Ag_nO⁻ (n = 1–8) and CO at 150 K, and extensive DFT calculations on the observed reaction paths. These combined experimental and theoretical investigations showed that Ag_nO⁻ (n = 1, 2, 5–8) with terminal or bridging oxygen atoms generated free CO₂, and the quasi-linear structures of Ag_nO⁻ (n = 3, 4) generated chemically bonded CO₂ on silver. The oxygen atoms in all Ag_nO⁻ (n = 1–8) are intrinsically highly reactive even at the low temperature, and their reactivity is only slightly dependent on the sizes, spins and other electronic properties. The chemically bonded CO₂ molecules formed in the reactions of Ag_nO⁻ (n = 3, 4) were predicted to have the structural formula of Ag–CO₂–Ag₂⁻ and Ag₂–CO₂–Ag₂⁻. Similar adsorbed CO₂ units were previously observed on bulk silver, and their formation processes were proposed to involve the reactions between carbonate and CO. This work presented one possible bonding pattern and a new formation mechanism of the chemisorbed CO₂ on bulk silver. Further exploration on this species could provide molecular level insights into some important surface reactions, such as oxidation of CO as well as activation of CO₂ on silver catalysts.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (Grant No. 21273278 and 21673158), the Ministry of Science and Technology of China (Grant No. 2012YQ22011307), and Science & Technology Commission of Shanghai Municipality (14DZ2261100). The authors are greatly thankful to Dr Joel H. Parks in Rowland Institute at Harvard for giving us most of the experimental facilities, and acknowledge helpful discussions with Prof. Lai-Sheng Wang in Brown University and Prof. Xue-Feng Wang in Tongji University.

Notes and references

1. J. G. Serafin, A. C. Liu and S. R. Seyedmonir, J. Mol. Catal. A: Chem., 1998, 131, 157–168.
2. R. M. Lambert, F. J. Williams, R. L. Cropley and A. Palermo, J. Mol. Catal. A: Chem., 2005, 228, 27–33.
3. Y. Lei, F. Mehmood, S. Lee, J. Greeley, B. Lee, S. Seifert, R. E. Winans, J. W. Elam, R. J. Meyer, P. C. Redfern, D. Teschner, R. Schlogl, M. J. Pellin, L. A. Curtiss and S. Vajda, Science, 2010, 328, 224–228.
4. U. Burghaus and H. Conrad, Surf. Sci., 1996, 364, 109–121.
5. G. G. Xia, Y. G. Yin, W. S. Willis, J. Y. Wang and S. L. Suib, J. Catal., 1999, 185, 91–105.
6. Z. Qu, M. Cheng, X. Dong and X. Bao, Catal. Today, 2004, 93–95, 247–255.
7. X. D. Zhang, Z. P. Qu, F. L. Yu and Y. Wang, Chin. J. Catal., 2013, 34, 1277–1290.
8. J. V. Barth and T. Zambelli, Surf. Sci., 2002, 513, 359–366.
9. A. F. Benton and J. C. Elgin, J. Am. Chem. Soc., 1926, 48, 3027–3046.
10. G. H. Twigg, Trans. Faraday Soc., 1946, 42, 657–663.
11. Z. P. Qu, M. J. Cheng, W. X. Huang and X. H. Bao, J. Catal., 2005, 229, 446–458.
12. M. Bowker, M. A. Barteau and R. J. Madix, Surf. Sci., 1980, 92, 528–548.
13. C. T. Campbell and M. T. Paffett, Surf. Sci., 1984, 143, 517–535.
14. A. J. Nagy, G. Mestl and R. Schlogl, J. Catal., 1999, 188, 58–68.
15. S. Linic and M. A. Barteau, J. Am. Chem. Soc., 2003, 125, 4034–4035.
16. C. T. Campbell, Surf. Sci., 1985, 157, 43–60.
17. F. Besenbacher and J. K. Nørskov, Prog. Surf. Sci., 1993, 44, 5–66.
18. G. I. N. Waterhouse, G. A. Bowmaker and J. B. Metson, Appl. Surf. Sci., 2003, 214, 36–51.
19. D. S. Su, T. Jacob, T. W. Hansen, D. Wang, R. Schloegl, B. Freitag and S. Kujawa, Angew. Chem., Int. Ed., 2008, 47, 5005–5008.
20. T. C. R. Rocha, A. Oestereich, D. V. Demidov, M. Haevecker, S. Zafeiratos, G. Weinberg, V. I. Bukhtiyarov, A. Knop-Gericke and R. Schloegl, Phys. Chem. Chem. Phys., 2012, 14, 4554–4564.
21. T. E. Jones, T. C. R. Rocha, A. Knop-Gericke, C. Stampfl, R. Schlogl and S. Piccinin, Phys. Chem. Chem. Phys., 2015, 17, 9288–9312.
22. G. Cipriani, D. Loffreda, A. Dal Corso, S. de Gironcoli and S. Baroni, Surf. Sci., 2002, 501, 182–190.
23. Y. Xu, J. Greeley and M. Mavrikakis, J. Am. Chem. Soc., 2005, 127, 12823–12827.
24. S. R. Bare, K. Griffiths, W. N. Lennard and H. T. Tang, Surf. Sci., 1995, 342, 185–198.
25. M. Taniguchi, K. i. Tanaka, T. Hashizume and T. Sakurai, Surf. Sci., 1992, 262, L123–L128.
26. O. Nakagoe, K. Watanabe, N. Takagi and Y. Matsumoto, J. Phys. Chem. B, 2005, 109, 14536–14543.
27. X. C. Guo and R. J. Madix, J. Phys. Chem. B, 2001, 105, 3878–3885.
28. U. Burghaus and H. Conrad, Surf. Sci., 1997, 370, 17–31.
29. U. Burghaus and H. Conrad, Surf. Sci., 1995, 338, L869–L874.
30. U. Burghaus and H. Conrad, Surf. Sci., 1996, 352–354, 201–205.
31. T. Zambelli, J. V. Barth and J. Wintterlin, J. Phys.: Condens. Matter, 2002, 14, 4241–4250.
32. T. Kou, D. Li, C. Zhang, Z. Zhang and H. Yang, J. Mol. Catal. A: Chem., 2014, 382, 55–63.
33. X. Zhang, Z. Qu, X. Li, M. Wen, X. Quan, D. Ma and J. Wu, Sep. Purif. Technol., 2010, 72, 395–400.
34. K. Shimizu, K. Sawabe and A. Satsuma, ChemCatChem, 2011, 3, 1290–1293.
35. X. Zhang, Z. Qu, X. Li, Q. Zhao, Y. Wang and X. Quan, Catal. Commun., 2011, 16, 11–14.
36. L. B. Yu, Y. Y. Shi, Z. Zhao, H. B. Yin, Y. C. Wei, J. A. Liu, W. B. Kang, T. S. Jiang and A. L. Wang, Catal. Commun., 2011, 12, 616–620.
37. Y. X. Zhao, X. L. Ding, Y. P. Ma, Z. C. Wang and S. G. He, Theor. Chem. Acc., 2010, 127, 449–465.
38. W. Huang, H. J. Zhai and L. S. Wang, J. Am. Chem. Soc., 2010, 132, 4344–4351.
39. H. J. Zhai, S. Li, D. A. Dixon and L. S. Wang, J. Am. Chem. Soc., 2008, 130, 5167–5177.
40. M. Zhou, C. Wang, Z. H. Li, J. Zhuang, Y. Zhao, X. Zheng and K. Fan, Angew. Chem., Int. Ed., 2010, 49, 7757–7761.
41. M. Zhou, X. Jin, Y. Gong and J. Li, Angew. Chem., Int. Ed., 2007, 46, 2911–2914.
42. Y. Gong, M. Zhou, M. Kaupp and S. Riedel, Angew. Chem., Int. Ed., 2009, 48, 7879–7883.
43. G. E. Johnson, R. Mitric, V. Bonacic-Koutecky and A. W. Castleman, Chem. Phys. Lett., 2009, 475, 1–9.
44. A. W. Castleman, Catal. Lett., 2011, 141, 1243–1253.
45. Z. X. Luo, G. U. Gamboa, M. Y. Jia, A. C. Reber, S. N. Khanna and A. W. Castleman, J. Phys. Chem. A, 2014, 118, 8345–8350.
46. S. M. Lang and T. M. Bernhardt, Phys. Chem. Chem. Phys., 2012, 14, 9255–9269.
47. S. M. Lang, I. Fleischer, T. M. Bernhardt, R. N. Barnett and U. Landman, ACS Catal., 2015, 5, 2275–2289.
48. X.-L. Ding, X.-N. Wu, Y.-X. Zhao and S.-G. He, Acc. Chem. Res., 2012, 45, 382–390.
49. L.-N. Wang, Z.-Y. Li, Q.-Y. Liu, J.-H. Meng, S.-G. He and T.-M. Ma, Angew. Chem., Int. Ed., 2015, 54, 11720–11724.
50. M. Schlangen and H. Schwarz, J. Catal., 2011, 284, 126–137.
51. H. Schwarz, Angew. Chem., Int. Ed., 2011, 50, 10096–10115.
52. L. D. Socaciu, J. Hagen, U. Heiz, T. M. Bernhardt, T. Leisner and L. Woste, Chem. Phys. Lett., 2001, 340, 282–288.
53. M. Schmidt, A. Masson and C. Bréchignac, Phys. Rev. Lett., 2003, 91, 243401.
54. L. D. Socaciu, J. Hagen, J. Le Roux, D. Popolan, T. M. Bernhardt, L. Woste and S. Vajda, J. Chem. Phys., 2004, 120, 2078–2081.
55. J. Hagen, L. D. Socaciu, J. Le Roux, D. Popolan, T. M. Bernhardt, L. Woste, R. Mitric, H. Noack and V. Bonacic-Koutecky, J. Am. Chem. Soc., 2004, 126, 3442–3443.
56. Z. Luo, G. U. Gamboa, J. C. Smith, A. C. Reber, J. U. Reveles, S. N. Khanna and A. W. Castleman, Jr., J. Am. Chem. Soc., 2012, 134, 18973–18978.
57. Y.-N. Wu, M. Schmidt, J. Leygnier, H.-P. Cheng, A. Masson and C. Brechignac, J. Chem. Phys., 2012, 136, 024314.
58. M. Schmidt, A. Masson, H.-P. Cheng and C. Brechignac, ChemPhysChem, 2015, 16, 855–865.
59. J. Ma, X. Cao, X. Xing, X. Wang and J. H. Parks, Phys. Chem. Chem. Phys., 2016, 18, 743–748.
60. D. Y. Tang, J. P. Hu, Y. Q. Zhang and C. W. Hu, Acta Chim. Sin., 2009, 67, 1859–1864.
61. P. P. Dholabhai, X. Wu and A. K. Ray, THEOCHEM, 2005, 723, 139–145.
62. D. H. Kim, K. Shin and H. M. Lee, J. Phys. Chem. C, 2011, 115, 24771–24777.
63. C. M. Chang, C. Cheng and C. M. Wei, J. Chem. Phys., 2008, 128, 124710.
64. D. Y. Tang, Z. Z. Chen, J. P. Hu, G. F. Sun, S. Z. Lu and C. W. Hu, Phys. Chem. Chem. Phys., 2012, 14, 12829–12837.
65. M. Schmidt, P. Cahuzac, C. Brechignac and H.-P. Cheng, J. Chem. Phys., 2003, 118, 10956–10962.
66. J. Ma, X. Cao, L. Hao, Y. Baoqi and X. Xing, Phys. Chem. Chem. Phys., 2016, 18, 12819–12827.
67. T. M. Bernhardt, J. Hagen, S. M. Lang, D. M. Popolan, L. D. Socaciu-Siebert and L. Woste, J. Phys. Chem. A, 2009, 113, 2724–2733.
68. A. D. Becke, J. Chem. Phys., 1993, 98, 1372–1377.
69. C. Lee, W. Yang and R. G. Parr, Phys. Rev. B: Condens. Matter Mater. Phys., 1988, 37, 785–789.
70. P. J. Stephens, F. J. Devlin, C. F. Chabalowski and M. J. Frisch, J. Phys. Chem., 1994, 98, 11623–11627.
71. K. A. Peterson and C. Puzzarini, Theor. Chem. Acc., 2005, 114, 283–296.
72. D. Figgen, G. Rauhut, M. Dolg and H. Stoll, Chem. Phys., 2005, 311, 227–244.
73. R. Krishnan, J. S. Binkley, R. Seeger and J. A. Pople, J. Chem. Phys., 1980, 72, 650–654.
74. A. E. Reed, L. A. Curtiss and F. Weinhold, Chem. Rev., 1988, 88, 899–926.
75. M. J. Frisch, G. W. Trucks and H. B. Schlegel, et al., Gaussian 09, Revision A02, Gaussian, Inc., Wallingford, CT, 2009.
76. A. W. Castleman, K. G. Weil, S. W. Sigsworth, R. E. Leuchtner and R. G. Keesee, J. Chem. Phys., 1987, 86, 3829–3835.
77. G. Gioumousis and D. P. Stevenson, J. Chem. Phys., 1958, 29, 294–299.
78. Z. Luo, G. U. Gamboa, J. C. Smith, A. C. Reber, J. U. Reveles, S. N. Khanna and A. W. Castleman, Jr., J. Am. Chem. Soc., 2012, 134, 18973–18978.
79. M.-S. Liao, J. D. Watts and M.-J. Huang, J. Phys. Chem. C, 2014, 118, 21911–21927.
80. G. U. Gamboa, A. C. Reber and S. N. Khanna, New J. Chem., 2013, 37, 3928–3935.
81. H. Hakkinen, B. Yoon, U. Landman, X. Li, H. J. Zhai and L. S. Wang, J. Phys. Chem. A, 2003, 107, 6168–6175.
82. P. Weis, T. Bierweiler, S. Gilb and M. M. Kappes, Chem. Phys. Lett., 2002, 355, 355–364.
83. F. Solymosi, J. Mol. Catal., 1991, 65, 337–358.
84. X. C. Guo and R. J. Madix, Surf. Sci., 2001, 489, 37–44.
85. H.-Y. Su, M.-M. Yang, X.-H. Bao and W.-X. Li, J. Phys. Chem. C, 2008, 112, 17303–17310.
86. H.-Y. Su, Z. Zeng, X.-H. Bao and W.-X. Li, J. Phys. Chem. C, 2009, 113, 8266–8272.
87. U. Burghaus, Prog. Surf. Sci., 2014, 89, 161–217.
88. M. F. Hsieh, H. D. Li, D. S. Lin and K. Morgenstern, J. Phys. Chem. C, 2010, 114, 14173–14179.

---

Footnote

† Electronic supplementary information (ESI) available: Linear fitting of −ln(I/I₀) vs. CO flows in the kinetic measurements of the reactions between Ag_nO⁻ (n = 1–8) and CO at 150 K (Fig. S1), the theoretical structural candidates of Ag_nO⁻ (n = 1–8) (Fig. S2), the spins and negative charges on the oxygen atoms in the lowest lying structures of Ag_nO⁻ (n = 1–8) (Fig. S3), theoretical structures of Ag_nCO₂⁻ (n = 1–8) with CO₂ molecularly adsorbed on Ag_n⁻ (Fig. S4), and structural candidates of Ag_nCO₂⁻ (n = 3, 4) except those with weakly bonded CO₂ (Fig. S5). See DOI: 10.1039/c6cp06741g

---