DOI: 10.1002/chem.201403140

Minireview

& Surface Chemistry

Carbon Monoxide-Induced Dynamic Metal-Surface Nanostructuring Sophie Carenco*[a, b, c]

Chem. Eur. J. 2014, 20, 1 – 11

These are not the final page numbers! ÞÞ

1

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

&

Minireview ability of metal surfaces and nanoparticles to undergo restructuring after exposure to CO under fairly mild conditions, generating nanostructures. This Minireview proposes a brief overview of recent examples of such nanostructuring, which leads to a discussion about the driving force in reversible and non-reversible situations.

Abstract: Carbon monoxide is a ubiquitous molecule in surface science, materials chemistry, catalysis and nanotechnology. Its interaction with a number of metal surfaces is at the heart of major processes, such as Fischer–Tropsch synthesis or fuel-cell optimization. Recent works, coupling structural and nanoscale in situ analytic tools have highlighted the

Introduction

Brief Summary about CO Structure and Its Interaction with a Metal Surface

Carbon monoxide is an ubiquitous molecule that takes part into a variety of biochemical[1–3] and chemical processes.[4, 5] Whether it comes into play as a feedstock (e.g. Fischer–Tropsch Synthesis),[6] as a reaction intermediate (e.g. Sabatier process)[7] or as a poison (e.g. Pt-based fuel cells),[8] understanding its interaction with metal clusters and surfaces is of prime importance. While the field of metal carbonyls has been largely explored in the last century, both for monometallic species and clusters,[9–15] the interaction of CO molecules with metal surfaces is still at the heart of mechanistic studies, for example, in the case of Fischer–Tropsch synthesis.[16–22] More generally, the interaction of CO with metal clusters and surfaces is deeply involved in the field of catalysis. For example, metal carbonyls are precursors for supported or heterogeneous catalysts and they can also be active catalytic species.[23–26] During the last decade, development of controlled synthesis of nanomaterials gave access to monodispersed nanoparticles (NPs) as well as bimetallic objects with nanostructures (e.g. core-shell or dimmers). Huge progress has also been made in the field of single-nanoparticle characterization (ex situ and in situ). Consequently, the local consequences of CO interaction with nanoparticle surfaces are being investigated anew. This Minireview discusses recent results on structure evolution at the nanoscale of surfaces that strongly interact with CO. Examples were selected across a range of structures, from the most ideal ones (single-crystal surface) to the most complex (bimetallic nanoparticles). They involve a set of recent experimental characterization tools to illustrate the relevance of dynamic restructuring processes for further development of catalysts.

CO is a well-known ligand in coordination chemistry, and its binding to a metal M mainly involves two of its frontier orbitals (Scheme 1). CO acts as a Lewis base through the HOMO that is

Scheme 1. Diagram of molecular orbitals of carbon monoxide and their possible overlap with d orbitals of a metal M.

strongly polarized on the carbon atom, forming a s bond C M bond (donation). It also acts as a Lewis acid through its LUMO, also polarized on the carbon atom, by accepting electron density from the metal in its p orbital (back-donation). This second interaction d is most often not negligible and contributes to the binding of CO to M, in particular when M is an electronrich Group 10 metal (i.e. with filled dxy, dyz or dxy orbitals). It also results in a weakened carbon–oxygen bond because of the anti-bonding character of the LUMO, and as a consequence the C O stretching mode vibration, measured by IR spectroscopy, shifts to a lower frequency. A CO molecule interacts with a metal center also as a function of the environment (e.g. other ligands, other metal atoms in a cluster, slack of atoms underneath in a crystal), on the geometry of the site and on the number of CO molecules already present. The reactivity of both the metal and the CO molecule can be considerably affected by these parameters, from the enthalpic and from the entropic point of views. As a consequence, there is no simple relation between the nature of the metal or metal alloy, the geometry of the adsorption site and the average CO coverage and the energy of adsorption. Direct calorimetric experiments and DFT modelling are of great help to evaluate this energy, as a function of coverage and nature of the adsorption sites (top, bridge, etc.).[27, 28] As an example, Figure 1 shows the adsorption energy of CO at low coverage of 0.25 monolayer (ML—one ML corresponding to one molecule per surface atom) on the most stable site of a variety of

[a] Dr. S. Carenco Sorbonne Universits-UPMC Univ Paris 06 UMR 7574, Chimie de la Matire Condense de Paris Collge de France, 11 place Marcelin Berthelot 75231 Paris Cedex 05 (France) [b] Dr. S. Carenco CNRS, UMR 7574, Chimie de la Matire Condense de Paris 11 place Marcelin Berthelot 75231 Paris Cedex 05 (France) [c] Dr. S. Carenco Collge de France, Chimie de la Matire Condense de Paris 11 place Marcelin Berthelot 75231 Paris Cedex 05 (France) Fax: (+ 33) 1-44-27-15-04 E-mail: [email protected]

&

&

Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

2

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Minireview techniques in addition to MS. However, it is much more delicate to detect in situ surface restructuring, without using structural analysis, such as LEED for single-crystal or HRTEM for nanoparticles. These structural changes can be easily overlooked since they do not result in a chemical alteration of the surface. In the following, we will mostly discuss cases in which the surface does not get oxidized or carbidized, but for which its structure is nonetheless modified because of its interaction with CO. We aim at giving a few illustrations of some of the most impressive restructuring schemes of metal surfaces or nanoparticles as a consequence of CO exposure. This will emphasize the growing role of in situ characterization techniques for understanding these transformations.

Figure 1. Adsorption energy of CO for a variety of metal surfaces, on the most stable adsorption site (i.e. on top site, except when otherwise indicated in parenthesis): Mo(110), W(110), Re(0001), Fe(110), Ru(0001), Co(0001), Rh(111), Ir(111), Ni(111) (hollow hcp), Pd(111) (hollow hcp), Pt(111), Cu(111), Ag(111), Au(111), Zn(0001), Al(111), plotted from the values calculated in ref. [27].

CO Interaction with Monometallic Systems: The Example of Platinum Surfaces Single-crystal surfaces

metal surfaces, as calculated in ref. [27]. The metals were ordered by group. Binding of CO on the metal surface involves a complex interaction of the molecular orbitals of CO (not limited to the frontier ones) and the d-band of the metal. A detailed description and discussion on the trends can be found elsewhere.[29] For the purpose of our discussion, we will simply highlight the large difference of adsorption energy when going from Group 6–10 to 11–13. In particular, CO forms much weaker bonds with Au than with Pd. Once coordinated to a metal atom or adsorbed on a metal surface, the fate of the CO molecule can be manifold, as illustrated on Scheme 2. Most common high-symmetry adsorption sites are depicted on the top part of the scheme. NPs have a high proportion of low-coordination number sites such as edges and corners. CO may bind in a linear (1) or tilted (2) fashion, and at higher coverage multi-carbonyl species (3) are favoured. Back-donation from the metal surface to the 2p* orbital of CO results in weakening the C O bond as illustrated in (4). The general trend is that this effect is more pronounced when going from the right to the left of the periodic table, although it also depends on the adsorption site (top, bridge, etc.).[27] Once adsorbed, CO can go back to the gas phase by desorption, which is favoured by temperature increase. It can migrate on the surface as a molecular adsorbate, or dissociate into C and O. Surface migration may also involve a metal atom, which can lead to the formation of islands on the surface. Desorption of metal carbonyl is also possible (e.g. Ni(CO)4). Along with CO pressure, the nature of the surface will influence the relative occurrence of each of these processes, but they will in return be able to modify the surface structure. The situation becomes even more complex when the surface is bimetallic, or when species (e.g. hydrogen, oxygen) are co-adsorbed with CO. Luckily, CO is an easy molecule to detect on a surface, whether one uses FTIR, XAS, XPS, or 13C NMR spectroscopy. Resulting oxides, carbides, or gas-phase species, such as CO2 and alkanes, can also quite easily be tracked by using the same Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

These are not the final page numbers! ÞÞ

Pt surfaces interact strongly with CO, which is one of the reasons for CO being a poison in Pt-based fuel cells. At the atomic level, adsorption sites and angles have been very well described by using single-crystal Pt surface cleaned in ultrahigh vacuum (UHV) chambers. On a closed-packed surface, such as Pt(100), extensive description of surface-phase transition induced by the adsorption of CO is available,[30] for example, under oxidative conditions.[31] Because of the repulsive CO–CO interaction, the coverage generally saturates below 1 ML, and complex structures, such as a Moir pattern form, as a function of CO background pressure, as was shown on the Pt(111) surface.[32] The lattice constant of the Moir pattern increases with the CO partial pressure in the 10 7–103 Torr range to accommodate the increasing number of CO molecules (from ca. 0.50 to 0.70 ML). A high-density Moir pattern is depicted on Figure 2, top, and corresponds to a coverage of 0.70 ML, obtained with a pressure of 720 Torr (i.e. 1 bar) at

Sophie Carenco graduated in France from Ecole Polytechnique, Palaiseau, in 2008. She obtained her Ph.D. in 2011 from University Pierre and Marie Curie (UPMC), Paris, for her work on the synthesis and applications of metal phosphide nanoparticles, under the cosupervision of Pr. Clment Sanchez (UPMC) and Dr. Nicolas Mzailles (Ecole Polytechnique). In 2012 and 2013, she was a postdoctoral fellow in the group of Pr. Miquel Salmeron at Lawrence Berkeley National Lab, Berkeley, California, where she used synchrotron based in situ X-ray spectroscopy (XPS, XAS) to monitor the surface of metal nanoparticles during catalytic reactions such as Fischer–Tropsch synthesis. Since January 2014, she is a researcher at Collge de France-UPMC-CNRS, Paris, in the team “Hybrid Materials and Nanomaterials”. She is the author of 21 publications and one book chapter. She was awarded the European Young Chemist Award from EuCheMS in 2010 and the C’Nano National Award in 2012 for her Ph.D. work.

3

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

&

Minireview nal structure (noted hex-Pt(100)) of approximately 20 % higher density than the bulk-terminated structure, which would have a square symmetry. The interaction between the last two layers is not as strong as in Pt(111), which has an underlying hexagonal structure, meaning that hex-Pt(100) is more prone to further reconstruction upon CO adsorption. Structural techniques, such as Rutherford backscattering (RBS) and low-energy electron diffraction (LEED), are particularly suited to identify ensemble surface structures on single crystals. In early work by Norton et al., CO was shown to help the surface Pt atoms move back into registry when it covered more than 0.5 ML.[33] Very interestingly, the strong decrease of the RBS surface peak with coverage indicated that the movement of more than one atomic plane had to be involved. The authors concluded that lateral displacements of atoms of the second and potentially subsequent layers participated to the surface reconstruction from hex-Pt(100) to (1  1) Pt upon CO adsorption. At the time, experimental techniques operating at high pressure gave only ensemble measurements: it was hardly possible to identify non-periodic structures forming at the atomic or nanoscale. Almost three decades later, the developments of high-pressure STM allowed looking anew into this question, confirming the hypothesis of a profound surface restructuring. Exposure of hex-Pt(100) to 10 5 Torr of CO at room temperature resulted in the formation of clusters of metal atoms of diameter in the 0.5–3.5 nm range covering about 45 % of the surface (Figure 2, bottom).[34] CO molecules are still coordinated on these clusters. The examples of Pt(111) and hex-Pt(100) illustrate well two different behaviours upon CO adsorption. In the first case, an incommensurate CO hexagonal structure forms on the surface (also of hexagonal structure), without inducing Pt surface restructuring. On the second case, the surface is reconstructed at first, with a hexagonally-structured top layer sitting on a lattice of square symmetry, and coordinating CO on the surface helps the Pt atoms to fall back into registry. This is accompanied by the formation of surface Pt clusters. One could argue, in the second case, that the high tendency of the surface to undergo restructuring is solely due to the incommensurate structure of the top layer compared with the bulk. However, it was shown that adsorption of other molecules, such as ethylene, does not affect the hex-Pt(100) up to 1 Torr.[35] The strong effect of CO on the formation of surface clusters was further confirmed in the same study by the fact that, in the presence of 5.10 6 Torr

Scheme 2. Most common fates of a CO molecule landing on a surface.

Figure 2. Top) Moir superstructure of CO molecules (720 Torr) on Pt(111) at room temperature: ball model on left (each red dot is a CO molecule), STM images on right. Adapted with permission from ref. [32]. Copyright (2004) American Chemical Society. Bottom) Restructuring of hex-Pt(100) under CO (10 5 Torr): ball model on left, STM image on right. N.B: STM images do not allow determining if the islands grow in registry with the surface: the ball model merely illustrates the presence of the islands on top of the Pt surface. Adapted with permission from ref. [34]. Copyright (2009) American Chemical Society.

room temperature. The Pt(111) surface does not undergo restructuring even under these conditions. This stands in contrast with the Pt(100) surface, which will be discussed next. Pt(100) belongs to a group of bare Pt surfaces with low surface atom density that already reconstruct under vacuum to lower the surface free energy. As a result, the last layer of atoms is not in registry with the crystal structure. On the Pt(100) surface, the last layer of atoms adopts a quasi-hexago&

&

Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

4

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Minireview a Pt(557) surface, lowered this temperature to only 550 K.[41] Any restructuring of a stepped surface is thus expected to have consequences on reaction kinetics. Restructuring of stepped Pd(111) was recently studied by high-pressure STM.[42] Under pressures of CO as low as 5.10 8 Torr, the initial six-atom wide and one-atom high terraces of Pt(557) restructured to terraces of double that height and width (Figure 3, left). Under a higher pressure of 1 Torr, these double terraces broke into nanoclusters of triangular shapes and of 2.1  2.2 nm width, pointing in the direction of the lower terraces (Figure 3, right). Very interestingly, this nanostructuring was reversible: by lowering the pressure back to the 10 8 Torr range, the nanoclusters disappeared and the double terraces were formed anew. Reversible nanostructuring under 1 Torr of CO was also observed for the Pt(332) surface, though it occurred without the intermediate double-terrace formation. On this surface, the nanoclusters were of parallelogram shape, a geometry that is likely related to the original step orientation. Ambient-pressure XPS (APXPS) allowed measuring the CO coverage in each situation (Figure 3, bottom). On Pt(557), CO coverage was as high as 0.97 ML under 0.5 Torr, and dropped to a more typical value around 0.5 ML under 10 8 Torr. This two-fold variation of coverage might seem comparatively small compared with the pressure differential (from 10 8 to 0.5 Torr), but one should remember that CO molecules repel themselves on a surface. The authors proposed that the break-up of the surface under 0.5 Torr was driven by the maximization of Pt CO bonds along with the minimization of the repulsive CO–CO interaction. Nanostructuring created a high number of low-coordinated atoms on which CO could bind in a tilted configuration, which also limited the repulsive intermolecular interaction. APXPS also provided complementary information about restructuring of the Pt(557) surface, in particular regarding the more drastic nanostructuring observed under high CO pres-

of ethylene, few CO molecules (typically those from the background, at a very low partial pressure likely below 10 8 Torr), interact enough with the surface to trigger the formation of 2.3  1.4 nm Pt islands, quite similar to those observed in Figure 2, bottom, and described in ref. [34]. This confirms the specific role of CO in favouring the restructuring of hex-Pt(100) into nanoscaled Pt islands. This first set of examples, dealing with model surfaces, illustrated two main ideas: 1) different surfaces have different abilities to undergo restructuring (e.g. hex-Pt(100) vs. Pt(111)) and 2) on metals for which adsorption is strong, CO can trigger restructuring even at comparatively low partial pressure. Stepped surfaces To gain further information regarding surfaces that are closer to those of real catalysts, stepped surfaces have been thoroughly studied in the last decade. For example, the introduction of steps on a Pt(111) surface, giving Pt(332) or Pt(557) surfaces, was shown to influence their chemical and electrochemical performances in reactions, such as CO oxidation.[36–39] Pt(332) and Pt(557) are stepped (111) surfaces, with six-atomwide terraces of (111) and (100) orientation, respectively. This gives rise to a wider variety of adsorption sites, in particular those with low coordination numbers. Some reaction pathways, such as CO oxidation to CO2, are well-known to depend on the nature of these sites. For example, early work by Ohno et al. showed that CO2 forms from the reaction of adsorbed CO with adsorbed O atoms on flat terraces, but with adsorbed O2 molecules on steps.[40] Energy barriers also depend on the presence of steps, kinks and defects. By using sum-frequency generation operating under 40 Torr, CO dissociation temperature was shown to be lower (500 K) on the less stable Pt(100) surface, which is prone to surface roughening, than on more stable Pt(111) surface (673 K). Introducing steps on this latter surface, that is, using

Figure 3. Restructuring of Pt(557) surface under CO. Adapted with permission from ref. [42]. Copyright (2010), American Association for the Advancement of Science. Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

These are not the final page numbers! ÞÞ

5

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

&

Minireview sure.[42] The Pt4f core level indicated the ratio of atoms in each coordination geometry (bulk, surface and low-coordinated sites) while the O1 s region allowed tracking the bonding sites of CO (top, bridge, low-coordinated sites). Upon a pressure increase to 1 Torr, the ratio of Pt in low-coordination sites increased, as did the ratio of CO bound to low-coordination sites. Lowering the pressure restored the original bonding situation for Pt and CO. The driving force for surface restructuring lies firstly in the strength of the Pt–CO interaction. Molecular dynamics simulation showed that the step-doubling reconstruction of Pt(557) occurred when the coverage was high enough (0.5 ML), through adatom diffusion and step wandering, whereas it did not occur on a Au(557) surface that has a much weaker interaction with CO.[43] The reversible character of the nanostructuring on Pt(557) highlights the paramount role of in situ techniques, here, highpressure STM and APXPS, for tracking the 3D structure of a material exposed to a reactive environment. Very high reactivity was observed in this particular example, thanks to the strong adsorbate–metal interaction and to the fact that the initial surface was clean and already reconstructed.

proximately 2 nm. These comparatively large Pt NPs were less sensitive to restructuring than the small ones: they did not further increase in size even under more strenuous conditions (600 K and ca. 1 Torr of CO), because their interaction with CO was weaker. CO was thus able to induce mass transport only below a critical size of 2 nm on this MgO support. On Pt NPs, co-adsorption of CO and another strong ligand allows synergistic effects. This was investigated with the purpose of controlling the shape of large NPs, by stabilizing selected facets. Co-adsorption of oleylamine with CO during the growth of Pt nuclei in solution was shown to stabilize the Pt(100) surface.[47] Without CO, the synthesis conducted in oleylamine and oleic acid leads to a poorly defined dentritic shape (Figure 4a). Introducing a flow on CO during the reac-

Pt nanoparticles Figure 4. TEM images of Pt nanoparticles synthesized at 180 8C with oleylamine and oleic acid (4:1 ratio). a) Pt dendrites synthesized without CO and b) Pt nanocubes synthesized under a CO flow. Adapted with permission from ref. [47]. Copyright (2010), Royal Society of Chemistry.

Shifting to nanoparticulate systems, such as ligand-covered Pt colloids, or alternatively supported Pt nanoclusters, brings in an additional level of complexity, as one has to consider both the interaction of CO with the exposed Pt sites and the interaction of ligands with the nanoparticles. These latter interactions already play a role in defining the facet orientation of the nanoparticles and are often not negligible. Moreover, restructuring of nanoparticles depends on the strength of their interaction with the support (e.g. metal oxide). Regarding the support effect, the case of Pt nanostructure synthesis from Ptx(CO)y clusters gives a good example of the various interactions at stake. Deposition of Chini-type Pt carbonyl clusters ([Pt3(CO)6]62 species)[44] by radiolysis on a carbon support provided Pt islands. Some CO molecules stayed coordinated to the islands: they were observed by Xray absorption spectroscopy, which showed the conservation of a 1.5  bond of Pt with CO.[45] These islands had a structure reminiscent of the starting Pt–CO polynuclear clusters. In contrast with this, deposition on a-alumina favoured a more profound restructuring: Pt–CO dissociation was favoured, and resulted in Pt fcc clusters with a high proportion of Pt Pt bonds. Once formed on a surface, Pt NPs are still sensitive to further restructuring upon interaction with CO, much alike the singlecrystal surfaces described above. In situ grazing incidence small-angle X-ray scattering (GISAXS) allowed monitoring the NPs size distribution first under increasing CO pressure, and second under annealing in vacuum of CO-covered NPs.[46] At a relatively low pressure of approximately 10 3 Torr, CO was shown to induce the scavenging of 1 nm Pt nanoparticles deposited on MgO(001), forming surface Pt–carbonyl species with high mobility. Increasing the temperature up to 650 K resulted in the coalescence of the NPs in larger structures of ap&

&

Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

tion resulted in Pt nanocrystals with a cubic shape with (100) surfaces (Figure 4b). DFT calculation highlighted the collaborative effect of CO and oleylamine in this nanostructuring process. While bare Pt(100) is less stable than bare Pt(111) by 0.26 eV/atom, adsorption of 0.25 ML amine (alone) reduces this difference to 0.20 eV/atom. Co-adsorption of amine and CO (0.25 ML each) brings down the energy difference to 0.02 eV/ atom, which makes the (111) surface as stable as the (100). This last example illustrated how CO can be used as a tool to direct NPs structure during the growing step. To the best of our knowledge, CO-driven restructuring of preformed nanoparticles in solution was not attempted yet, but in the light of all the previous examples including those about single-crystal surfaces, it could be interesting to open this line of work. One could prepare a colloidal solution of Pt NPs and then expose it to CO, looking for ripening mechanisms and/or changes in nanoparticles shapes. Such phenomena could be followed by routine tools (e.g. TEM, XRD), as well as more advanced ones (e.g. in situ TEM). This may provide a post-synthetic route to adjust nanocrystals structure and could have implications for in-depth understanding of Pt NPs behaviour in major catalytic processes, such as CO (electro)oxidation. All the examples above dealt with Pt, which interacts strongly with CO. More generally, CO can induce surface reconstruction in any system in which the interaction is strong enough and for which the starting surface is reactive enough (e.g. existing constraints on the surface, unsaturated sites, etc.). 6

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Minireview Among other late transition metals, nickel also forms strong bonds with CO, which makes possible the leaching of the Ni atom through the formation of Ni(CO)4 gaseous species, at temperatures as low as 145 K.[48] This phenomenon was shown to occur through the formation of surface subcarbonyls (Ni(CO)3,ads) preferentially on defects, and lead to the corrosion of the nanoparticles supported on SiO2. This destructive kind of surface restructuring is not desirable, because it affects the composition of the catalyst during the reaction, and because it may change the size of the nanoparticles. Since many elemental processes, such as adsorption or dissociation, strongly depend on nanoparticle size[49–51] and on the exposed facets,[52] this highlights the importance of in situ monitoring of the interaction of CO with the nanoparticles surface. These tools are particularly suited for bimetallic structures, which present an additional degree of complexity in the sense that the two metals may have very different interactions with CO. This will be discussed now.

Figure 5. STM images of Au/Ni(111) surface alloy. Top) before exposure to CO (inset : atomic resolution image, the Au atoms are imaged as depression) and bottom) after exposure to 1 bar CO (left inset: line scan; inset: atomic resolution of an area between the Au islands, showing a clean Ni(111) surface). Schematics of the pictures inset are plotted on the right side of the figure. Adapted with permission from ref. [55]. Copyright (2005), American Physical Society.

CO Interaction with Bimetallic Surfaces and Advanced Nanostructures The literature on surface alloys is rich as these surfaces have always raised the question of the co-localisation of each metal. In particular, early works were discussing the relative behaviour of metals as a function of surface composition and bulk composition, for example, for Pt–Au alloys.[53] However, local analysis on few-atom aggregates was hardly possible using techniques such as LEED or photoemission electron microscopy.[54] STM has brought unique insights into these questions, as shown below.

the bulk. At the nanoscale, similar examples are found and exhibit a strong propensity to nanostructuring, as discussed next with the example of copper and cobalt, which are not miscible at the macroscale but can form alloys at the nanoscale. Cobalt–copper nanoparticles: CO can undergo dissociative adsorption on cobalt. As a consequence, exposure to CO of cobalt–copper nanoparticles with a cobalt-rich shell resulted in the surface oxidation of the NPs at 250 8C, as evidenced by the appearance of an oxide peak in the O1 s region of APXPS (under 200 mTorr of CO) and by a pre-edge peak in Co l-edge XAS (under 1 bar of CO).[56] Some CO could still adsorb molecularly on the surface and was detected on the O K-edge spectrum as a specific absorption peak at 534 eV from p*(CO) (Figure 6, route a). No NPs restructuring was observed under these conditions. In contrast, mixing CO with a more reductive gas, H2, allowed keeping the surface in a reduced state and observing the restructuring caused by CO molecular adsorption on the bimetallic surface (Figure 6, route b). At 250 8C, dealloying of about 5 % of the NPs was observed and resulted in few NPs that were almost cobalt-free, as evidenced by the EDS maps performed in scanning-TEM mode (Figure 6, bottom right). The formation of cobalt–carbonyl species on the reduced Co-rich surface was proposed to be at the origin of this profound restructuring. Low solubility of cobalt in copper may also have played a role as a driving force for this segregation process. These phenomena would benefit from being studied by morphological in situ tools such as environmental TEM, in order to delineate the structural features that promoted CO-induced dealloying on some of the nanoparticles.

Compositional change through metal–carbonyl formation NiAu surface alloy: As a model surface for a bimetallic catalyst, NiAu surface alloys have been prepared by evaporating approximately 0.3 ML of Au on a Ni(111) crystal in UHV and annealing to 800 K (Figure 5, left). Such a surface is interesting for the present discussion, as both metals have a contrasted behaviour toward CO. On one side, nickel interacts strongly with CO and can be leached away, as mentioned above. On the other side, Au interaction with CO is very weak (about 1 eV less energetic than Ni). Using a high-pressure STM apparatus, Vestergaard et al. showed that exposure of this surface to 1 bar of CO induced a phase separation of the two metals and a restructuring of the surface into Au nanoparticles of a few nm diameter and one or two monolayer height (Figure 5, right).[55] Mechanistic studies evidenced that this nanostructuring was due to the departure of Ni–carbonyl to the gas phase. More interestingly, STM real-time movies and DFT modelling highlighted that CO and Au played a synergistic effect in driving Ni leaching: CO strong adsorption was favoured not only by the presence of steps and kinks on the surface, but also because it was compressed on Ni because of the presence of Au. This example on a model surface dealt with a couple of metals that can alloy on the last layer of a surface, but not in Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

These are not the final page numbers! ÞÞ

7

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

&

Minireview

Figure 7. DFT modelling of Pd coverage for Pd0.2Au0.8 NPs exposed to CO (1 bar) as a function of size and temperature. Adapted with permission from ref. [59]. Copyright (2007) Elsevier.

Figure 6. CoCu NPs exposed to carbon monoxide (route a) or syngas (route b). Bottom) TEM pictures in high-angle annular dark-field mode (left) and EDS maps of two regions (right) showing two populations of NPs resulting from route b: cobalt-rich (red-circled, left EDS map) and copper-rich (green-circled, right EDS map).[56] The second population represents 5 % of all the nanoparticles. Adapted with permission from ref. [56]. Copyright (2013) American Chemical Society.

Shifting to nanoscale system, the question of CO-induced surface enrichment and segregation has to be reviewed, because there is not anymore an infinite reservoir of Pd from the bulk. Equilibrium concentrations on NP surfaces, calculated from DFT, are thus different than on single-crystal and vary with the NP size. For Au-rich alloys in particular, smaller NPs reach lower CO surface coverage than larger ones because there are not enough Pd atoms to cover the whole surface (Figure 7).[59] For example, at 400 K, Pd0.2Au0.8 NPs of 2 nm diameter will exhibit 35 % of Pd on its surface versus 85 % for a NP of 8 nm diameter, according to DFT modelling. Significant differences of catalytic efficiency are thus expected for reactions such as alkene hydrogenations, for which Pd is the active species.

Both bimetallic systems described so far (Ni–Au and Co–Cu) were characterized by a low miscibility of the two metals and the fact that one of them (Ni or Co) could readily undergo metal–carbonyl formation. In the next section, we describe the case of highly miscible metals that are not prone to metal–carbonyl formation. These two factors will be shown to impact the mechanism of CO-induced restructuring. Pd–Au system: a contrast of adsorption energy that promotes restructuring

Experimental observation of CO-induced Pd segregation in AuPd NPs: Thanks to the high sensitivity of the CO stretching mode to the bonding situation, DRIFTS provides a tool of choice to monitor the number and nature of adsorption sites on NPs, especially during a catalytic reaction. Over time, exposure of supported AuPd NPs (ca. 2–3 nm diameter) to a CO/He flow resulted in the increase of the bands related to CO on Pd (linear and multi-bonded carbonyls) while the bands for CO on Au decreased.[60] The surface was enriched in Pd. A similar phenomenon was also observed in the context of CO oxidation: Pd atoms were found to migrate to the surface at room temperature, probably in low coordination sites (edges, corners), as it induced a decrease of catalytic activity.[61] However, in this case, it is possible that surface Pd underwent oxidation, a reaction that is susceptible to drive the segregation. For small NPs of diameter below 3 nm, one can hardly define a crystalline structure that could provide a starting point for discussing NPs restructuring. However, techniques dedicated to poorly crystalline objects, such as Pair Distribution Function (PDF) analysis, can track these structural changes by providing information on a local environment for each atom. For PdAu NPs exposed to CO for a week at room temperature, PDF evidenced an increase of Au Au bonds and a de-

Au and Pd form a solid solution over the whole compositional range, and can crystallize as Au3Pd or AuPd3 phases.[57] But they are much contrasted in terms of their interaction strength with CO. Besides, these alloys are not prone to oxidation due to their noble character, neither do they form a metal–carbonyl: they constitute an excellent model to look for CO-induced restructuring. DFT modelling of a Au–Pd surface interaction with CO: On a single-crystal PdAu(111) surface, modelling showed that moderate pressure of CO in the 10 2 Torr range induces segregation of Pd to the surface, eventually forming monomer and dimer patches.[58] These patches allow accommodating a higher number of coordinated CO molecules while limiting their repulsive interactions, which is a driving force for extracting Pd from the lattice (as discussed above for the Pt(557) surface). At lower pressure, restructuring already results in an enrichment of Pd in the last layers because of the stronger energy of adsorption of CO on Pd compared with Au. These energies are still sensitive to the alloy composition and to the nature of the adsorption site, but CO prefers Pd to Au by at least 1 eV in every case examined.[58] &

&

Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

8

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Minireview crease of Au Pd and Pd Pd bonds, which was interpreted as a segregation of Pd toward the surface of the NPs.[62] In the same work, a similar behaviour was identified for PtCu NPs initially made of a PtCu core surrounded by a Cu shell: Pt was found to segregate to the surface under CO, in accordance with the relative M CO bond energy, as for the Pd Au system. For nanoalloys in which both metals interact strongly with CO, segregation of one particular metal to the surface as a result of CO adsorption becomes more difficult to anticipate than in the cases illustrated above. In this case, chemically resolved in situ techniques are of great help to monitor surface restructuring. PdPt and RhPd nanoparticles under CO and NO

Figure 8. Reversible surface restructuring of RhxPd1 x NPs as a function of composition (x = 0.2, 0.5, 0.8) and gas atmosphere. Bottom: Surface atomic fractions of Pd and Rh measured by APXPS with electrons of 0.7 nm mean free path. Adapted with permission from ref. [63]. Copyright (2010) American Chemical Society.

Like CO oxidation reaction, NO oxidation by CO produces surface oxide in addition with CO adsorbed species. The discussion about the driving force for the NP restructuring is thus more delicate, but still of interest. In particular, APXPS allows tracking the formation of the oxide. Spectra collection with variable photon energy provides at the same time composition as a function of depth within a 2 nm thick surface layer of the NP. Using this tool, composition-controlled PdxPt1 x NPs (16 nm diameter, x = 0.8, 0.5, 0.2) were studied under exposure of CO, NO and their mixtures at approximately 0.1 Torr pressure.[63] After synthesis, a Pd shell was observed on the nanoparticles with x = 0.8 and 0.5, in agreement with the lower surface energy of Pd versus Pt (1.90 for Pd(100) and 2.48 J. m 2 for Pt(100)), and the shell was thicker for the higher value of x. Ptrich NPs (x = 0.2) also exhibited a relative surface enrichment in Pd closer to the surface, although Pt was still the major species. Exposure of the NPs to NO gas (0.1 Torr at 300 8C) resulted in surface oxidation, which further favoured the segregation of Pd to the surface because it is less noble than Pt. Changing the gas atmosphere to pure CO (0.1 Torr at 300 8C) reduced the species to a metallic state. The Pd/Pt did not evolve for any of the compositions. In contrast with this, RhxPd1 x NPs of the same diameter exhibited a strong and reversible restructuring under the same conditions: the surface energy of metallic Rh (2.9 J. m 2 for Rh(100)) is higher than those of Pd, but Rh oxides are more stable than Pd oxides. As a consequence, switching the atmosphere from mildly reducing (CO + NO) to oxidizing (NO) resulted in oscillation of the relative surface composition of Rh and Pd (Figure 8). For example, APXPS measurement performed with electrons of 0.7 nm mean free path showed that Pd-rich NPs (x = 0.2) exhibited a Pd surface ratio of less than 80 % under oxidizing conditions (NO) versus more than 80 % under reducing conditions (NO + CO or H2). These latter two examples did not illustrate the restructuring effect of CO alone but in a mixture with NO. More classically, they showed the influence of oxidation state on the relative surface energies, instead of highlighting the role of CO as a strong surface ligand. It would be of great interest to further investigate these systems under CO only, while keeping the Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

These are not the final page numbers! ÞÞ

NPs metallic, to identify if the bonding of CO to the surface could promote by itself restructuring of the NPs surface. This would be particularly interesting for bimetallic systems of tuneable composition of size, since both parameters will modulate the strength of the metal–CO interaction.

Conclusion and Outlook This Minireview proposed a journey from single-crystal surfaces to bimetallic nanoparticles upon strong interactions with carbon monoxide. Several modes of CO-induced restructuring were described: atom migration on a surface, sometimes producing nanoclusters, atom leaching from the surface and dealloying, atom migration from the NP core to the NP surface, etc. Some were reversible, other were not. All these processes have in common a modification of the nanoscaled structure of the metal surface exposed to CO (alone or mixed with another gas) that is best characterized with in situ tools, such as highpressure STM, DRIFTS and APXPS, and that strongly benefits from the support of DFT modelling. In fact, this could also help in evaluating whether the restructuring yielded a thermodynamically stable surface or if it was kinetically limited by the choice of experimental conditions. Besides, entropic contributions to these rearrangements could also be calculated and may help in rationalizing the experiments in addition to enthalpic considerations. Examples dealing with bimetallic systems (surface alloys or NPs) particularly highlighted the importance of coupling chemical and morphological information to delineate the driving force of the nanostructuring. The future of the field probably lies in a further understanding of these processes, as a function of NP size and composition, for both reversible and non-reversible restructuring. Systematic and quantitative measurement of CO surface coverage should be pursued and coupled to DFT modelling to evaluate the weight of repulsive CO–CO interactions. Co-adsorption of CO with other gas or ligands should also be at the heart of explorative work, as it is most 9

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

&

Minireview relevant for real catalytic processes and can be responsible for complex spillover mechanism.[64] Besides, CO-induced nanostructuring could be envisioned as a post-treatment for modifying surface textures on purpose, for example, for creating a nanopattern of small Au NPs on a surface for plasmonic applications. Lastly, these studies would benefit to be extended to covalent non-metallic systems, such as metal carbides, nitrides, borides, phosphides, etc.[65, 66] These surfaces could also exhibit strong interactions with CO, hence, would also potentially exhibit nanostructuring phenomena.

[32] S. R. Longwitz, J. Schnadt, E. K. Vestergaard, R. T. Vang, I. Stensgaard, H. Brune, F. Besenbacher, J. Phys. Chem. B 2004, 108, 14497. [33] P. R. Norton, J. A. Da Vies, D. K. Creber, C. W. Sitter, T. E. Jackman, Surf. Sci. 1981, 108, 205. [34] F. Tao, S. Dag, L.-W. Wang, Z. Liu, D. R. Butcher, M. Salmeron, G. Somorjai, Nano Lett. 2009, 9, 2167. [35] Z. Zhu, D. R. Butcher, B. Mao, Z. Liu, M. Salmeron, G. A. Somorjai, J. Phys. Chem. C 2013, 117, 2799. [36] N. P. Lebedeva, A. Rodes, J. M. Feliu, M. T. M. Koper, R. A. van Santen, J. Phys. Chem. B 2002, 106, 9863. [37] G. A. Attard, O. Hazzazi, P. B. Wells, V. Climent, E. Herrero, J. M. Feliu, J. Electroanal. Chem. 2004, 568, 329. [38] W.-X. Li, J. Phys. Condens. Matter 2008, 20, 184022. [39] H.-C. Wang, S. Ernst, H. Baltruschat, Phys. Chem. Chem. Phys. 2010, 12, 2190. [40] Y. Ohno, J. R. Sanchez, A. Lesar, T. Yamanaka, T. Matsushima, Surf. Sci. 1997, 382, 221. [41] K. McCrea, J. S. Parker, P. Chen, G. Somorjai, Surf. Sci. 2001, 494, 238. [42] F. Tao, S. Dag, L.-W. Wang, Z. Liu, D. R. Butcher, H. Bluhm, M. Salmeron, G. A. Somorjai, Science 2010, 327, 850. [43] J. R. Michalka, P. W. McIntyre, J. D. Gezelter, J. Phys. Chem. C 2013, 117, 14579. [44] G. Longoni, P. Chini, J. Am. Chem. Soc. 1976, 98, 7225. [45] K. Torigoe, H. Remita, G. Picq, J. Belloni, D. Bazin, J. Phys. Chem. B 2000, 104, 7050. [46] N. Chabane, R. Lazzari, J. Jupille, G. Renaud, E. Avellar Soares, J. Phys. Chem. C 2012, 116, 23362. [47] B. Wu, N. Zheng, G. Fu, Chem. Commun. 2011, 47, 1039. [48] M. Mihaylov, K. Hadjiivanov, H. Knçzinger, Catal. Lett. 2001, 76, 59. [49] I. V. Yudanov, A. Genest, S. Schauermann, H.-J. Freund, N. Rçsch, Nano Lett. 2012, 12, 2134. [50] C. Strebel, S. Murphy, R. M. Nielsen, J. H. Nielsen, I. Chorkendorff, Phys. Chem. Chem. Phys. 2012, 14, 8005. [51] A. Tuxen, S. Carenco, M. Chintapalli, C.-H. Chuang, C. Escudero, E. Pach, P. Jiang, F. Borondics, B. J. Beberwyck, A. P. Alivisatos, G. Thornton, W.-F. Pong, J. Guo, R. Perez, F. Besenbacher, M. Salmeron, J. Am. Chem. Soc. 2013, 135, 2273. [52] J.-X. Liu, H.-Y. Su, D.-P. Sun, B.-Y. Zhang, W.-X. Li, J. Am. Chem. Soc. 2013, 135, 16284. [53] J. W. A. Sachtler, J. P. Bibrian, G. A. Somorjai, Surf. Sci. 1981, 110, 43. [54] K. Asakura, J. Lauterbach, H. H. Rotermund, G. Ertl, Surf. Sci. 1997, 374, 125. [55] E. Vestergaard, R. Vang, J. Knudsen, T. Pedersen, T. An, E. Lægsgaard, I. Stensgaard, B. Hammer, F. Besenbacher, Phys. Rev. Lett. 2005, 95, 126101. [56] S. Carenco, A. Tuxen, M. Chintapalli, E. Pach, C. Escudero, T. D. Ewers, P. Jiang, F. Borondics, G. Thornton, A. P. Alivisatos, H. Bluhm, J. Guo, M. Salmeron, J. Phys. Chem. C 2013, 117, 6259. [57] H. Okamoto, T. Massalski, Bull. Alloy Phase Diagrams 1985, 6, 229. [58] M. Garca-Mota, N. L pez, Phys. Rev. B 2010, 82, 075411. [59] V. Soto-Verdugo, H. Metiu, Surf. Sci. 2007, 601, 5332. [60] A. Hugon, L. Delannoy, J. Krafft, C. Louis, J. Phys. Chem. C 2010, 114, 10823. [61] L. Delannoy, S. Giorgio, J. G. Mattei, C. R. Henry, N. El Kolli, C. Mthivier, C. Louis, ChemCatChem 2013, 5, 2707. [62] S. M. Oxford, P. L. Lee, P. J. Chupas, K. W. Chapman, M. C. Kung, H. H. Kung, J. Phys. Chem. C 2010, 114, 17085. [63] F. Tao, M. E. Grass, Y. Zhang, D. R. Butcher, F. Aksoy, S. Aloni, V. Altoe, S. Alayoglu, J. R. Renzas, C.-K. Tsung, Z. Zhu, Z. Liu, M. Salmeron, G. Somorjai, J. Am. Chem. Soc. 2010, 132, 8697. [64] E. A. Lewis, D. Le, A. D. Jewell, C. J. Murphy, T. S. Rahman, E. C. H. Sykes, ACS Nano 2013, 7, 4384. [65] C. Giordano, M. Antonietti, Nano Today 2011, 6, 366. [66] S. Carenco, D. Portehault, C. Boissire, N. Mzailles, C. Sanchez, Chem. Rev. 2013, 113, 7981.

Acknowledgements Collge de France, CNRS and UPMC are acknowledged for financial support. S.C. acknowledges Dr. D. Portehault and Pr. C. Sanchez for helpful discussions about this review. Keywords: carbon monoxide · metal surfaces · nanoparticles · nanostructuring · surface chemistry [1] S. H. Heinemann, T. Hoshi, M. Westerhausen, A. Schiller, Chem. Commun. 2014, 50, 3644. [2] P. Amara, J.-M. Mouesca, A. Volbeda, J. C. Fontecilla-Camps, Inorg. Chem. 2011, 50, 1868. [3] T. R. Johnson, B. E. Mann, J. E. Clark, R. Foresti, C. J. Green, R. Motterlini, Angew. Chem. 2003, 115, 3850; Angew. Chem. Int. Ed. 2003, 42, 3722. [4] M. Haruta, T. Kobayashi, H. Sano, N. Yamada, Chem. Lett. 1987, 405. [5] R. Franke, D. Selent, A. Bçrner, Chem. Rev. 2012, 112, 5675. [6] A. Y. Khodakov, W. Chu, P. Fongarland, Chem. Rev. 2007, 107, 1692. [7] W. Wang, S. Wang, X. Ma, J. Gong, Chem. Soc. Rev. 2011, 40, 3703. [8] J. J. Baschuk, X. Li, Int. J. Energy Res. 2001, 25, 695. [9] A. A. Blanchard, Chem. Rev. 1940, 26, 409. [10] J. S. Anderson, Q. Rev. Chem. Soc. 1947, 1, 331. [11] J. W. Cable, R. K. Sheline, Chem. Rev. 1956, 56, 1. [12] E. W. Abel, Q. Rev. Chem. Soc. 1963, 17, 133. [13] P. Chini, J. Organomet. Chem. 1980, 200, 37. [14] U. Radius, F. M. Bickelhaupt, A. W. Ehlers, N. Goldberg, R. Hoffmann, Inorg. Chem. 1998, 37, 1080. [15] A. Hughes, Coord. Chem. Rev. 2000, 197, 191. [16] O. R. Inderwildi, S. J. Jenkins, D. A. King, J. Phys. Chem. C 2008, 112, 1305. [17] M. Ojeda, R. Nabar, A. U. Nilekar, A. Ishikawa, M. Mavrikakis, E. Iglesia, J. Catal. 2010, 272, 287. [18] M. Zhuo, A. Borgna, M. Saeys, J. Catal. 2013, 297, 217. [19] M. P. Andersson, F. Abild-Pedersen, I. N. Remediakis, T. Bligaard, G. Jones, J. Engbæk, O. Lytken, S. Horch, J. H. Nielsen, J. Sehested, J. Catal. 2008, 255, 6. [20] S. Shetty, R. A. van Santen, Catal. Today 2011, 171, 168. [21] S. Shetty, A. P. J. Jansen, R. A. van Santen, J. Am. Chem. Soc. 2009, 131, 12874. [22] Q. Zhang, B. Han, X. Tang, K. Heier, J. X. Li, J. Hoffman, M. Lin, S. L. Britton, A. Derecskei-Kovacs, H. Cheng, J. Phys. Chem. C 2012, 116, 16522. [23] D. C. Bailey, S. H. Langer, Chem. Rev. 1981, 81, 109. [24] R. Ugo, R. Psaro, J. Mol. Catal. 1983, 20, 53. [25] B. C. Gates, Chem. Rev. 1995, 95, 511. [26] A. Choplin, F. Quignard, Coord. Chem. Rev. 1998, 178 – 180, 1679. [27] F. Abild-Pedersen, M. P. Andersson, Surf. Sci. 2007, 601, 1747. [28] K. Liao, V. Fiorin, D. S. D. Gunn, S. J. Jenkins, D. King, Phys. Chem. Chem. Phys. 2013, 15, 4059. [29] M. Gajdo, A. Eichler, J. Hafner, J. Phys. Condens. Matter 2004, 16, 1141. [30] R. J. Behm, P. A. Theil, P. R. Norton, G. Ertl, J. Chem. Phys. 1983, 78, 7437. [31] R. J. Schwankner, M. Eiswirth, P. Mçller, K. Wetzl, G. Ertl, J. Chem. Phys. 1987, 87, 742.

&

&

Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

Received: April 17, 2014 Published online on && &&, 0000

10

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Minireview

MINIREVIEW & Surface Chemistry

Nanoparticles and reactive surfaces: An overview of selected CO-induced nanostructuring (see scheme). Recent examples of metal-surface and nanoparticles restructuring as a consequence of exposure to CO are discussed and show that nanoscaled structures can be obtained under fairly mild conditions. Several cases of mono- and bimetallic compounds are described and show a range of behaviours in relation with the metal–CO interaction strength.

S. Carenco* && – && Carbon Monoxide-Induced Dynamic Metal-Surface Nanostructuring

CO-Induced Nanostructuring Carbon monoxide is an ubiquitous molecule that takes part into a variety of biochemical and chemical processes; therefore, understanding its interaction with metal clusters and surfaces is of prime importance. In her Minireview on page && ff., S. Carenco provides a brief overview of recent examples of CO-induced nanostructuring on metal surfaces and discusses the driving forces involved in reversible and non-reversible situations.

Chem. Eur. J. 2014, 20, 1 – 11

www.chemeurj.org

These are not the final page numbers! ÞÞ

11

 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

&