Carbon Monoxide Oxidation Promoted by a Highly ... - ACS Publications

reached quickly the same values as under pure oxygen. They stayed roughly stable during the whole reaction over 3 hours, even when the CO partial pres...
0 downloads 0 Views 1MB Size
Subscriber access provided by - Access paid by the | UCSB Libraries

Article

Carbon Monoxide Oxidation Promoted by a Highly Active Strained PdO Layer at the Surface of Au30Pd70(110) Marie-Claire Saint-Lager, Marie-Angélique Languille, Francisco José Cadete Santos Aires, Aude Bailly, Stephanie Garaudee, Eric Ehret, and Odile Robach ACS Catal., Just Accepted Manuscript • DOI: 10.1021/acscatal.8b04190 • Publication Date (Web): 09 Apr 2019 Downloaded from http://pubs.acs.org on April 9, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Carbon Monoxide Oxidation Promoted by a Highly Active Strained PdO Layer at the Surface of Au30Pd70 (110) Marie-Claire Saint-Lager,*1 Marie-Angélique Languille,2,3 Francisco J. Cadete Santos Aires,2,4 Aude Bailly,1 Stéphanie Garaudée,1 Eric Ehret2 and Odile Robach.5

1 CNRS

Institut Néel and Université Grenoble Alpes, 38000 Grenoble, France.

Univ. Lyon, Université Claude Bernard Lyon 1, CNRS, IRCELYON – UMR 5256, 69626 Villeurbanne, France. 2

Present address : Centre de Recherche sur la Conservation (CRC), Muséum National d’Histoire Naturelle, CNRS, Ministère de la Culture, 75005 Paris, France.

3

Laboratory for Catalytic Research, National Research Tomsk State University, 634050 Tomsk, Russia. 4

5 Univ.

Grenoble Alpes, CEA, INAC-MEM, 38000 Grenoble, France. *e-mail : [email protected]

phone : (+33) 4 76 88 74 15

1 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Abstract The evolution of the Au30Pd70(110) surface was studied by coupling Grazing Incidence X-ray diffraction and mass spectrometry under oxygen-rich conditions at moderate temperatures (300 to 470 K). This allows to correlate the depth profile of its structure to its catalytic properties for carbon monoxide (CO) oxidation. Under increasing pressure from ultra-high vacuum up to 100 mbar, both oxygen and CO induce Pd segregation, even at room temperature. However, in pure oxygen the surface is reorganized with a (1x2) missing row reconstruction, whereas in pure CO it is strongly roughened. When oxygen pressure is increased a phase corresponding to the initial step of the oxidation with oxygen dissolution in the subsurface region appears at first. Then, from about 400 K onwards, an oxidized thin Pd layer ( 1 nm) is formed growing in the [100]PdO direction. This PdO phase is strained and does not coincide with the P42/mmc structure usually observed for this oxide under ambient conditions. It is more probably consistent with the high pressure I4/mmm PdO structure strained by epitaxy on the underneath alloy. For higher oxidizing conditions and layer thickness, the oxide will then relax to the usual PdO structure. This strained oxide is easily reduced by CO and exhibits a very high activity for CO oxidation. Its catalytic performance at 470 K is comparable to the one found on surfaces of pure palladium at higher temperatures. Furthermore, on the clean Au30Pd70(110) surface, surface oxidation is hindered up to 470 K if CO is introduced prior to oxygen. This indicates that when Pd is alloyed with gold, its binding with CO is stronger than with oxygen. The weakening of the Pd-O binding by surrounding gold atoms is the key of the formation of a well-ordered and very active thin PdO film on Au30Pd70(110).

Keywords :

CO oxidation, PdO, gold palladium alloy, in situ studies, grazing incidence X-

ray diffraction, catalytic activity

2 ACS Paragon Plus Environment

Page 2 of 37

Page 3 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Introduction Studies on gold-based alloys were stimulated by the pioneering work of Haruta and coworkers, revealing that Au nanoparticles (NPs) supported on oxides may catalyze carbon monoxide (CO) oxidation even at room temperature (RT).1 The key step for CO oxidation is the dissociation of O2. Indeed, large Au NPs, as well as bulk Au, adsorb CO under reaction conditions, but do not activate molecular oxygen.2 However, the role of the synergetic interaction between the NPs and the substrate was very early established.3 More recently, it has been shown that the active sites for CO oxidation on Au/TiO2 catalysts are located at the Au-substrate interface, and the O2 dissociation occurs at a dual Ti-Au site.4 The geometry of this site could play a major role. The maximum in catalytic activity observed for particle size around 2 nm5,6 was indeed correlated to particular sites at the perimeter of the NP/TiO2 interface.7 In any case the conversion rate is as low as a few CO molecules per second and per Au atom. Alloying gold with another metal such as palladium could be another promising route for low temperature CO oxidation as it adds to the gold surface an element capable of dissociating O2. Indeed, Pd is known to dissociate O2 at temperatures as low as ∼150 K. But, due to the CO inhibition, pure Pd catalyzes CO oxidation at much higher temperatures,8 at least 450 K under steady-state conditions.9 It was shown that AuPd(100) is much more active than pure palladium surfaces for CO oxidation around 400 K.10-13 In ultra-high vacuum (UHV), the topmost layer of Au-Pd alloys is strongly enriched in gold, even for high palladium concentration.14-19 Under CO oxidation conditions, palladium segregates towards the surface, because it forms stronger bonds with oxygen and CO than gold does. This yields a mixed composition of the AuPd(100) surface with contiguous Pd ensembles (at least Pd dimers) which are required to achieve its high activity for CO oxidation.11 But the key factor of the improvement of the catalytic activity was attributed to the reduction of the CO binding energy with AuPd(100) compared to pure Pd. This facilitates CO desorption, even near RT, avoiding CO poisoning.10

3 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Au-Pd synergy resulting in the improvement of catalytic properties was not always observed. For instance, in the case of Au-Pd bimetallic NPs supported on TiO2, Pd segregation was also evidenced under O2 and O2/CO with the formation of Aucore-Pdshell structure.20 Nevertheless, no significant improvement of the gold catalytic activity was observed. This was attributed to the possible replacement of Au by Pd atoms in low coordination sites, which were supposed to be at the origin of the high activity in CO oxidation of the supported gold NPs. These conflicting results illustrate how the Au-Pd synergetic effect strongly depends on the system. In previous works, we studied the (110) surface of a Au30Pd70 alloy and showed that Pd segregation occurred under pure oxygen and also pure CO pressure, even at room temperature.18,19 Under near ambient pressure of oxygen, for temperatures above 420 K, an oxidized pure Pd layer grew in the [100]PdO direction. This PdO layer was strained and its structure not completely solved. Such unusual palladium phase can exhibit interesting catalytic properties. Indeed, even if it has been extensively studied as a prototypical catalytic reaction, the mechanism of the CO oxidation proceeding on Pt-group metal surfaces, such as Pd, is still debated. In the past, it was concluded that the CO oxidation reaction proceeds via the Langmuir−Hinshelwood (L-H) mechanism, whereby co-adsorbed CO and O atoms react to form CO2.8 The improvement of surface science techniques has enabled to overcome the “pressure gap” to conduct studies in near-ambient pressure conditions on model catalysts. It led to new insights, raising questions on the reaction mechanism and on the active sites. Some of these studies found a close relationship between the existence of the oxide species and the catalytic activity, and hence suggested a Mars−van Krevelen reaction mechanism.21 In this case, oxide species are formed on the metal substrate and control the reaction.22-24 In contrast, Goodman and coworkers claimed that the most reactive surface is covered by chemisorbed O (not an oxide phase) with undetectable amounts of CO.25-28 The purpose of this article is to study the strained PdO phase that grows on Au30Pd70(110) under high oxidizing conditions. Thus, we used X-ray scattering coupled to mass spectrometry and complemented by Auger spectroscopy, to follow the surface structure and composition under increasing oxygen and CO pressures. The explored temperature range was limited to moderate temperatures, between 300 and 470 K, where most of the crucial effects regarding the catalytic properties of Pd and Pd-Au systems were observed. 4 ACS Paragon Plus Environment

Page 4 of 37

Page 5 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Experiment The experimental setup consists of an UHV chamber connected to a batch reactor to allow sample transfer in UHV conditions. The sample surface was prepared in the UHV chamber where Auger Electron Spectroscopy (AES) measurements were also conducted (see details in Section S1 in the Supporting Information). The reactor is described in detail elsewhere,19,29 only the relevant features are mentioned here. It can be operated from UHV up to reactive conditions at ambient pressure and enables both Grazing Incidence X-ray Diffraction (GIXRD) and activity measurements in static conditions. The sample heating was performed with a high-power fiber-coupled laser diode. The temperature was measured with an external infrared pyrometer (Impac IPE 140 MB 10) that controlled the laser source. The surface heating induced by the exothermic CO oxidation reaction was thus immediately compensated by a decrease of the power supplied. The reactor chamber was connected to a gas manifold. N55 oxygen and N37 CO gas purity were used. Moreover, a Cu-pellet heating trap was installed on the CO gas line to avoid any metal carbonyl and moisture. The reactor gas composition was monitored by a quadrupole mass spectrometer mounted behind a CF40 gate valve. For pressures higher than 10−6 mbar, the reactants and products were sampled through a leak valve. The CO and CO2 partial pressures in the reactor chamber were deduced from the 28 and 44 m/e ionization currents. Experiments were performed at ESRF on beamline BM32, with X-ray photon energy of 18 keV. The incidence angle with respect to the sample surface was set to 0.3°. The crystal basis (a1, a2, a3) used to describe the (HKL) directions was defined with a3 in the direction perpendicular to the sample surface. It is expressed in terms of the cubic Au30Pd70 lattice (marked by the c subscript) with a3=1/2[110]C. a1 and a2 lie in the surface plane, a1 being along the close packed rows of the (110) face of the face-centered cubic structure and a2 perpendicular (see insert in figure 1b), with a1=1/2[-110]c and a2=[001]C. In this basis == =90°, a1 = a3= a0/2 and a2 = a0 with a0 = 3.942Å (Au30Pd70 lattice constant).19 Unless otherwise mentioned, the Miller 5 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

indices H, K and L, in reciprocal space lattice unit (r. l. u.) of the Au30Pd70(110) surface frame, will be used. This unit cell is tetragonal centered and the Bragg peaks appear for even values of H+K+L. The surface was monitored in various environments by recording the diffracted intensity along chosen scans perpendicular and parallel to the sample surface (see details in Section S1 in the Supporting Information). Perpendicularly to the surface, we analyzed the structure factor by line-scans along (HK) crystal truncation rods (CTR), namely the (00) CTR -corresponding to the reflectivity- and the (01) CTR. Along the (00) CTR, the diffracted intensity is not sensitive to the lateral in-plane position of the atoms. But it depends on the composition of each atomic plane and the inter-planar distances which can deviate from their values in the bulk alloy. Regarding the (01) CTR, only atoms in registry with the underneath bulk alloy contribute to its signal. It is the case of atoms in surface planes of which composition or spacing perpendicularly to the surface differ from those in the underneath alloy, as long as they keep the bulk in-plane unit cell (a1, a2). If a surface film is formed with a structure different from the bulk, such as palladium oxide, its atoms will contribute to the (00) CTR but not to the (01) CTR. Scans parallel to the sample surface were also recorded, at L  0, along the closed packed rows of the (110) surface and perpendicularly, H-scan and K-scan respectively. If H and K are integer and H+K is odd, the scan cuts a CTR giving a surface peak such as at (0 1 0.05). It can also meet peaks at fractional values of H and K, which can be generated by an epitaxial film on the alloy, with a different structure. This typically occurs when a Pd oxide film is formed on the surface. In this case, the peaks observed are the Bragg peaks of this film; the crystallography of which can be determined by looking for the other peaks corresponding to this new structure. This X-ray analysis, complemented by Auger spectroscopy, allowed to provide a good evaluation of the depth profile at each step of the alloy surface evolution under reactive gases. It was described by the number of atomic layers modified by the gas exposure, their composition, as well as their displacement from the mean inter-planar spacing. These parameters were deduced from a quantitative analysis by constructing a film that included all the planes differing from the bulk alloy and that covered the entire alloy surface (surface fraction equal to 1). The roughness parameter 30 was used as an approximation of all kinds of disorders in the surface layer (see Section S1 in the Supporting Information). In the following, the atoms are represented by their atomic number Z, which is 8, 46 and 79 for 6 ACS Paragon Plus Environment

Page 6 of 37

Page 7 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

oxygen, palladium and gold, respectively. The composition per plane is thus defined by the mean atomic number normalized to that of palladium. It corresponds to 1 for a pure Pd plane, 1.7 for a pure Au plane and 1.2 for Au30Pd70, assuming in all cases the same lattice parameters as the alloy.

Experimental results Evolution under increasing pure oxygen pressure (UHV up to 500 mbar) The clean Au30Pd70(110) surface was first exposed to pure oxygen pressure at RT following a procedure detailed in figure 1 (negative side of the horizontal time axis). The surface evolved through different phases as identified in our previous work.19 In UHV, the clean surface is (1x1) (see the left insert of figure 1b) and exhibits an intense surface peak at (0 1 0.05). When 2 mbar of oxygen were introduced, this peak rapidly decreased while a broad peak grew at K=1.50  0.01. It was the signature of the (1x2) missing-row reconstruction induced by oxygen adsorption, schematically represented in the second insert of figure 1b and studied in detail elsewhere. 19 After a long exposure to the same O2 pressure, the diffracted peaks began to vanish. Then the O2 pressure was increased up to 500 mbar which provoked the reappearance of a small surface peak at K=1 pointed by the orange arrow in figure 1b (see Section S2 in the Supporting Information). This peak disappeared when the temperature was increased to 420K and a new sharp diffraction signal rose at K= 1.47 corresponding to PdO. At the same time, on AES spectra, the gold signal strongly decreased (red curve of figure 2) yielding a RAu/Pd divided by ten compared to the clean (1x1) alloy surface, indicating a strong Pd segregation.

7 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 1: (a) Variation of partial pressures of oxygen (turquoise and left vertical axis), of CO and produced CO2 (right vertical axis, pink and grey, respectively) over time. 2 mbar of oxygen were introduced at time -3.75h (small green vertical arrow) and the pressure was then increased to 500 mbar; the pink arrow indicates the first CO introduction (at time 0h), the following jumps in the CO pressure correspond to new CO additions. The black dashed line indicates the temperature level. (b) Variation induced on the integrated intensity of the (0 1 0.05) surface peak (in blue) and of the (0 1.47 0.05) oxide peak (in dark green) [ (002)PdO Bragg peak], the dashed lines are guides for the eyes. The two inserts are schematic representations of the (1x1) and (1x2) Au30Pd70(110) surfaces with the close packed rows along [-110]C (see more details in ref. 19). The orange arrow points to the reappearance of the surface peak at K=1 and the two red arrows when the data were recorded on oxidized (O) and reduced (R) surfaces for X-ray analysis reported in figure 3.

8 ACS Paragon Plus Environment

Page 8 of 37

Page 9 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 2: Post mortem Auger spectra of the oxidized surface in red (after 1 mbar CO + 500 mbar O2 at 470 K) and of the reduced surface in blue (after 100 mbar oxygen + 2 mbar CO at 420 K). In these two cases RAu/Pd  0.5. In black, AES spectrum of the clean (1x1) Au30Pd70(110) surface in UHV (RAu/Pd = 4.3  0.2). For each curve, the intensity is normalized to the peak-to-peak height of the Pd(MNN) transition.

Characterization of the thin strained oxide layer under oxygen pressure As shown in our previous work, 19 in presence of the PdO phase, the detected diffraction peaks were at fractional values of H, K, and L and can be related to the tetragonal lattice of bulk PdO. The oxide grows in the [100]PdO direction, with the long axis perpendicular to the closed packed [-110]C rows of the Au30Pd70(110) surface (see figure S1b in Supporting Information). The parameters for the PdO layer unit cell, with orthogonal axes are aoxi = 2.93  0.03 Å, boxi = 3.00  0.05 Å and coxi = 5.39  0.01 Å defining the oxide Miller indices (hkl)PdO. In this frame the (0 1.47 0.05) peak corresponds to the (002)PdO oxide Bragg peak. Remarkably we also observed small peaks at (0 0.73 0.05)  (001)PdO forbidden in the usual phase of PdO. We will discuss further the structure of this PdO strained phase. One of the particularities of this oxide layer is its narrow thickness. In order to get a depth profile of the surface layer, the (00) and (01) CTR were recorded (at point (O) in figure 1b) and analyzed. The (00) CTR, plotted in yellow in figure 3a, presents oscillations characteristic of a fairly well-defined interface between the bulk and the surface layer. Figure 3c shows the mean deduced from fitting the (0 0) CTR for each atomic plane. In the topmost planes, = 0.8 corresponds to PdO. The displacement d relatively to the bulk alloy inter-planar distance is plotted in the same figure; it is about +6.5 % in the topmost planes, close to the PdO value. The mean thickness of this PdO layer is about 7 planes (purple area). For the deepest planes,  1.2 and d  0 corresponding to the bulk alloy parameters (yellow area); in between there is an interface layer made of 3 gold-enriched planes (orange area). As seen by comparing the red and the yellow curves, in figure 3b and 3a, respectively, oscillations along the (01) CTR are much larger than on (00), involving a smaller number of planes. Figure 3c shows a good match between the parameters deduced from the two CTRs for the 3 planes of the interface layer (purple area). In this interface layer, atoms are thus in registry with the bulk alloy, the deepest plane is gold-enriched and the gold concentration 9 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

rapidly decreases when approaching the oxide film. The (01) CTR signal comes thus from the buried interface film. The best fit was obtained for a large roughness parameter ( = 0.7). This indicates that this layer was not well organized with only a fraction of atoms in registry with the bulk underneath. Above this interface film, the occupancy deduced from the (01) CTR in the purple area is zero, meaning that no atoms were found in registry with the alloy. In other words, the whole layer is oxidized with no significant trace of metallic palladium, except if it was in a highly dispersed polycrystalline form. The parameters of the surface layer were obtained from the CTR fitting by using a surface fraction equal to one. It means that these parameters are representative of the entire surface.

Figure 3: Structure factor along (a) the (00) CTR and (b) the (01) CTRs. Points are data and full lines correspond to the calculated values with the parameters plotted with the same color in (c) and (d) for the oxidized and reduced surfaces, respectively. The very intense peak in (b) at L = 1 is due to the (011)C alloy Bragg peak. In (c) and (d) filled circles represent the mean atomic number (left vertical axis) and empty ones the relative variation d compared to the bulk alloy value (right vertical axis)). The horizontal axis is the number of the plane; zero being the surface plane. The colored areas indicate the type of the layer: Au30Pd70 alloy (yellow), gold-rich alloy (orange), PdO (pink) and reduced layer (blue).

Moreover, Figure 4a allows to compare the evolution of the (002)PdO PdO Bragg peak under a reactive mixture of (CO + O2) (1.05h < time < 1.5h) and during the oxide film formation under pure oxygen pressure (time < 0) at the same temperature (420 K). It shows a quite similar behavior, but the oxide peak has a lower intensity during the reaction. This is explained by a 10 ACS Paragon Plus Environment

Page 10 of 37

Page 11 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

thinner film, as evidenced by the reflectivity that exhibits slightly larger oscillations (figure 4b). At 470 K, during the CO oxidation reaction, the intensity and the width of the oxide peak reached quickly the same values as under pure oxygen. They stayed roughly stable during the whole reaction over 3 hours, even when the CO partial pressure was increased, boosting the CO2 formation rate. In the two environments, the width of the (002)PdO peak tends to the value of the surface peak width (in blue in Figure 4a). This indicates that the oxide domains have about the same size as the clean surface alloy in the K direction.

Figure 4: (a) Variation of the width along K of the (010) surface alloy peak (in blue) [q=2K/a2], of the intensity (in dark green) and of q (in orange) of the (002)PdO oxide peak [at (0 1.47 0.05)], of the sample temperature (in black) as a function of time in the same conditions as in figure 1. (b) Reflectivity recorded at 420 K with the oxide film under pure oxygen pressure (in red) and under CO oxidation reaction (in grey), recorded at times pointed with arrows of the same color in (a).

Behavior of the PdO strained phase in reactive mixture CO + oxygen Reducing of the PdO layer by CO To check the reactivity of the present PdO layer, CO was added to 500 mbar of oxygen with progressively increasing pressure at 420 K. As shown in figure 1, introduction of 0.7 mbar of CO at 420 K immediately makes the oxide peak vanish. This peak reappears one hour later but disappears again at the moment of a new introduction of 0.8 mbar CO. The oxide peak rises again when heating the sample at 470 K and stays stable up to a CO partial pressure of about 10 mbar which induces the vanishing of the oxide peak. Auger spectra performed on the reduced surface show that the ratio of Au over Pd signal, RAu/Pd, is close to 0, as for the oxidized 11 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

surface (blue curve in figure 2). The (00) and (01) CTRs were also recorded and analyzed in the stage when the oxide peak disappeared (Point (R) in figure 1b). The results are reported in figure 3 showing a quite different behavior. The oscillations on the (00) CTR of the reduced surface are larger compared to the oxidized surface and, interestingly, they have roughly the same frequency as on the corresponding (01) CTR. This means that the structure has strongly changed. The oxide layer was replaced by a layer with atoms mainly in registry with the bulk alloy. The best fit of the (01) CTR was obtained for a roughness parameter a bit smaller than for the oxidized surface ( = 0.5 instead of 0.7). The values as deduced from the (00) and (01) CTRs, plotted in figure 3d, follow the same trends: they have about the bulk alloy value for the deepest planes and drop below the palladium value when approaching the surface. Since this lowering is similar for the (00) and the (01) CTRs, it cannot be attributed to the disorder. It is rather due to a decrease of the occupancy when approaching the surface. This implies that the height of the layer is irregular. Most importantly, it shows that the surface layer structure is uniform. In particular, this precludes the presence of another phase such as powder-like PdO which would contribute to the reflectivity but not to the (01) CTR. Regarding the inter-planar spacing, there is a tendency to a small expansion compared to the bulk alloy with an average value of 1.3 %; it corresponds to + 2% compared to bulk palladium. Reactivity of the oxide phase Figure 1 also evidences the remarkable correlation between CO and Pd oxidation. The presence of the oxide peak at (0 1.47 0.05) [ (002)PdO Bragg peak] is systematically accompanied with an acceleration of the CO conversion into CO2. When CO is added to 500 mbar of O2 at 420 K, no reaction is observed and the oxide peak disappears. When it reappears, after a long delay, a small rate of CO conversion into CO2 is detected. The pressure ratio O2/CO is then about 800. The conversion rate sharply slows down upon a new CO introduction correlated with the vanishing of the oxide peak. When the sample is heated at 470 K, CO oxidation accelerates again and is accompanied by the comeback of the oxide peak. Increasing the pressure provokes an increase of the CO conversion rate. But a total CO partial pressure of about 10 mbar, decreasing the pressure ratio O2/CO down to 50, makes the oxide peak vanish again and heavily slows down the rate of the CO oxidation reaction. 12 ACS Paragon Plus Environment

Page 12 of 37

Page 13 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

At each temperature, two regimes can be distinguished: one with a low reaction rate and one with a high reaction rate. The low reaction rate regime corresponds to diffraction spectra with no oxide peak and was characterized by a reduced surface. To determine the CO2 formation rate catalyzed by the sample surface itself, we estimated the environment contribution by performing an experiment with no sample (see section S3 in the Supporting Information). The sample environment contribution was negligible at 300 K but significantly increased with temperature. We evaluated its rate as if it was produced by a surface equivalent to the sample one; one site being one surface atom. We found 15 CO2 molecules/s/site at 420 K, and 100 CO2 molecules/s/site at 470 K. In both cases, the rate was comparable to that measured in the low activity regime, which is thus below the accuracy of our measurements. The highly reactive regime occurs when the surface alloy is oxidized. The O2/CO pressure ratio ranges in a small area around 800 at 420 K and from 500 to 50 at 470 K. The CO conversion rates into CO2 obtained at 470 K in several conditions are reported together in figure 5 as a function of the CO partial pressure added to a constant oxygen pressure of 500 mbar. This rate is proportional to the CO pressure up to 10 mbar. It is equal to 500 molecules per second and per surface atom for 5 mbar of CO (O2/CO=100). Figure 5 also shows that the measured value for Pd(110) is on the same curve. However, we were not able to verify if this pure palladium surface is oxidized or not since this measurement was performed in laboratory, with the same set-up, but without X-ray characterization. Anyway, it indicates that the strained PdO formed on the alloy surface behaves as the one formed on a Pd crystal surface.

13 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Figure 5: CO conversion rate into CO2 measured at 470 K as a function of the CO pressure added to 500 mbar O2 on Au30Pd70(110) oxidized surface (in red) and reduced surface (in blue), on Pd(110) surface (in green) and on the sample holder (in black). The purple points were obtained for CO introduced before oxygen (see figure 8a). The CO2 formation rate was calculated assuming that only the free surface of the sample is active. The black horizontal line indicates the background level due to the sample environment.

PdO surface phase formation with CO The formation of this oxide phase was then studied under a mixed pressure of CO and oxygen. For that purpose, the clean surface was directly exposed to CO at room temperature. The surface structure and the gas composition in the reactor were followed when adding oxygen and increasing the temperature (figure 6). The reflectivity and (01) CTR were recorded at several steps of this evolution marked by colored stars. The mean per plane deduced from fitting of the (01) CTR is plotted in figure 7c.

Figure 6: Variation induced by increasing pressure (CO and then oxygen) and temperature over time with the same code as in figure 1: integrated intensity of the (0 1 0.05) surface peak (in blue) and of the (002)PdO oxide peak (in dark green), CO partial pressure (in pink) and produced CO2 (in grey); blue and green dashed lines are guide for eyes. The pink arrow indicates the introduction of 1.3 mbar of CO at RT on the clean surface and the green one the adding of 500 mbar of oxygen (at RT). The stars on the black dashed line of the temperature indicate the time when the CTR, plotted with the same color in figure 7, were recorded.

14 ACS Paragon Plus Environment

Page 14 of 37

Page 15 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 7 : (a) Reflectivity and (b) (01) CTRs with the same color as the stars in figure 6 (and 8a for the light purple curve in (a)), indicating when they were recorded; points being experimental data and lines corresponding to fits. The corresponding mean atomic numbers for each plane are drawn in (c).

CO pressure immediately destroys the (0 1 0.05) surface peak of the UHV clean (1x1) surface as seen in figure 6. Figures 7a and 7b show that the reflectivity and the (01) CTR intensity (black curves) are significantly lowered compared to those measured in UHV (dark blue curves). The CTR analysis indicates that the CO pressure at room temperature favors Pd enrichment of the surface plane. The roughness parameter  also increases (from 0.2 to 0.6), which is consistent with STM results already published.18 The surface plane composition was evaluated to be about half gold and half Pd (black curve in figure 7c). When a pressure of 500 mbar of oxygen is added at room temperature, there is no noticeable change of the surface structure (see light blue curves in figures 7b and 7c). Increasing the temperature to 420 K does not induce CO oxidation reaction but the roughness decreases ( = 0.4). It also accelerates the trend towards Pd segregation observed at RT. This leads to a surface plane mainly composed of palladium atoms and a second plane very rich in gold (see yellow curves in figure 7b and 7c). When heating at 470 K, the CO oxidation starts but it is not correlated to a change in the CTR (red curve of figure 7c). However, the CO2 formation rate, calculated in this case and reported in figure 5, is comparable to the sample environment contribution. This is highlighted in figure 8a where the CO and the CO2 signals are plotted with scales corresponding to that of figure 1a. This means that, if the surface contributes to the CO oxidation, its rate stays too low to be 15 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

extracted from the background. Anyway, this CO conversion lowers its partial pressure in the reactor and thus increases the ratio O2/CO yielding the conditions for the oxide formation. When, at least three quarters of CO are converted into CO2, the PdO peak begins to gradually rise as seen in figure 8a. It should be noted that this corresponds to an O2/CO pressure ratio of about 1500. The more detailed results reported in Figure 8a allow to follow in more detail the oxide formation occurring 1 hour after the temperature reached 470 K. In this zoom, the surface peak intensity, which reappeared after being worn off by the initial CO introduction, has been multiplied by 40. This short reappearance acts as the precursor signal of the structural transition to palladium oxide. Before, the surface evolution is weak and the main changes induced by the reactive gases are localized in the 2 topmost planes. This is still true for the step marked by the red star at the beginning of the temperature plateau at 470 K. The corresponding mean is plotted on the bottom curve of figure 8c, together with the one recorded later, just at the beginning of the surface peak reappearance, but when the oxide peak is still at zero (green curve). During this step, gold segregates towards the alloy: its concentration decreases in the second plane and increases in the third one. The palladium segregation evolves in the opposite direction. In the same time, we found a tendency to an increase of the inter-planar spacing, d-1,0 and d-2,-1, even if the gold concentration has decreased. This can be attributed to insertion of oxygen in the second plane (plane -1). Even before the disappearance of the small surface peak, the (002)PdO oxide peak begins to slowly grow. The reflectivity was recorded just when the surface peak vanishes again and the oxide peak is at  40 % of its maximum value (light purple star in figure 8a). The oxide thickness is about 3 planes. The (01) CTR was recorded 15 minutes later (dark purple star in figure 8a) and its behavior is similar to what was observed for oxidation under pure oxygen (figure 3). However, the oscillations are less pronounced, corresponding to the vanishing of the gold-rich interface layer between the alloy and the oxide.

16 ACS Paragon Plus Environment

Page 16 of 37

Page 17 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

Figure 8 : (a) Zoom on the beginning of the oxide peak formation under 1 mbar CO + 500 mbar O2, time 0 being taken at the beginning of the heating at 470 K. The labels are the same as in figure 6 but with surface peak intensity (in blue) multiplied by 40 and the scale of CO and CO2 signals is adjusted such as their evolution can be compared to that of figure 1a. As in figure 6, the stars indicate the time where the (01) CTRs, plotted in (b) were recorded, the corresponding being drawn in (c). Red data (also shown in figure 7) represent the sample surface just after heating at 470 K and the green ones later at the beginning of the reappearance of the surface peak. The two corresponding mean are plotted in the bottom of (c). Purple data were recorded when the surface peak was vanished during the oxide growth, the deduced being drawn on top of (c) with in dark purple, the values deduced from the (01) CTR and in light purple those from the reflectivity in figure 7a.

Discussion : Palladium segregation As previously evidenced, while the clean surface of Au30Pd70(110) is almost pure gold, an exposure to CO or to oxygen pressure causes palladium segregation towards the surface, even at room temperature.18,19 Under oxygen, the topmost plane becomes pure Pd and the two following ones are gold enriched. Oxygen adsorption induces a (1x2) missing row reconstruction with regular depression every 3 Pd atoms along the remaining row. Under CO pressure, the surface composition is Au-Pd mixed, and the gold surface atoms segregate towards the second plane. The CO effect stays localized in the very near surface region. No reconstruction was observed but a strong roughening occurs, as seen on STM 17 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

images.18 As already pointed out, for a metal surface in contact with CO, the effective step free energy is reduced and this can even lead to the spontaneous formation of steps.31,32 Increasing the temperature, after adding oxygen, reduces the roughness. It also increases the Pd content in the surface plane with no trace of reconstruction induced by oxygen as the (1x2) missing row. Oxidation occurs for a high O2/CO pressure ratio of about 1500 at 470 K. In our batch reactor this was achieved by the CO conversion due to the sample environment. The formation of palladium oxide is preceded by the reappearance of a small surface peak at K = 1. Under pure oxygen it can be observed even at room temperature, but for higher pressures than those necessary for the (1x2) missing row reconstruction. X-ray data analysis, as well as the Auger spectra reported in ref. 19, shows that it corresponds to a Pd enrichment of the surface. The palladium oxide formation requires heating to at least 420 K. X-ray data analysis indicates that the oxide covers the whole alloy surface. The AES spectra suggest that the gold concentration in the PdO film is negligible. Indeed, in this case the very weak gold signal comes from photoelectrons emitted by the gold atoms within the underneath alloy that pass through the 1nm-thick PdO film. The AuNVV Auger electron energy is 69 eV; this energy corresponds to the minimum of the electron inelastic mean free path, namely 0.4-0.5 nm.33 In these conditions, the attenuation of the gold signal is about 90%. It is comparable to the drop of the Auger ratios RAu/Pd between the one measured on the gold surface of the clean sample and the one after oxidation (see figure 2). It is the same when the oxidized surface is reduced, showing that the Pd segregation is irreversible.

Initial oxygen incorporation We first focused on the stage preceding the oxide formation distinguished by the reappearance of a small surface peak at K = 1 (see Section S2 in the Supporting Information). It is observed under pure oxygen at room temperature or under a O2 + CO mixture (for a O2/CO pressure ratio greater than 1500 at 470 K). As shown by the (01) CTR in figures 7b and 8b, its small intensity can be attributed to its closeness to a minimum of the structure factor as expected for a layer with a thickness of 2 atomic planes. This originates from the destructive X-ray interference at this point34 which is thus very sensitive to any change in the surface 18 ACS Paragon Plus Environment

Page 18 of 37

Page 19 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

structure. The (01) CTRs analysis indicates that the number of atomic planes concerned by the changes induced by the gas pressure, have increased from 2 to 3. Palladium segregates towards the surface as shown by the curves in the bottom of figure 8c which is consistent with previous AES measurements.19 In the same time, we deduce from the CTR fit that the inter-planar spacing on each side of the second plane is expanded. This can be explained by oxygen insertion in this plane. Structural evidence of oxygen insertion during the initial step of Pd oxidation is very scarce.35 During this phase, a great amount of dissolved oxygen is localized in the near surface region. LEED and AES reveal that palladium is still in a face-centered cubic (fcc) metallic state.36 The oxide formation on Pd(110) was studied in detail by coupling STM, XPS and DFT.37 Under increasing oxygen pressure, the surface presents first a c(2x4)-O reconstruction and then, the so-called “complex” one.38 The latter is formed by (7x3) and (9x3) domains in which the [110]C Pd rows are decorated by O atoms in a zig-zag pattern as for the c(2x4)-O but with Pd filling the missing row.37 Moreover in a further study by SXRD, the authors attributed to this “complex” structure the reappearance of the (010) surface peak.39 It then disappears with the very beginning of the oxide growth in a quite similar way to what we observed in figure 8a (to be compared to the figure 10a of ref. 39). In other words, this “complex” structure could correspond to the structure of the surface during the step of oxygen insertion in the subsurface region. The initial incorporation of oxygen was studied by DFT for Pd(111).40 This study showed that the most stable state for a 0.75 -1 ML oxygen coverage is a coupling of oxygen adsorbed on a fcc site at the surface and oxygen inserted in tetrahedral subsurface sites. However, the expected metastable state must be quite different on the low density (110) face as compared to that of the densest (111) face with an unlike geometry. Anyway, it is clear that the stable oxygen adsorption site on the (110) surface is of tetrahedral type as seen for the two c(2x4)O and “complex” reconstructions. But, in the bulk, the most probable interstitial sites are the octahedral ones, as already observed.41 The two kinds of sites are represented in figures 9a and b. These octahedral sites are substantially the largest ones and, as already mentioned, no impurity has been found to occupy tetrahedral sites in bulk Pd for which even hydrogen is inserted in octahedral sites.42 The subsurface oxygen is particularly unstable since we observe that the (010) peak quickly vanishes, when the one of the oxide grows (figure 8a). Indeed, in 19 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 37

PdO, due to its 4d8 electronic configuration, the Pd2+ ions are tetrahedral coordinated with a square planar geometry. The role of this subsurface oxygen is thus that of a metastable precursor of palladium oxide.40 The presence of this (010) diffraction signal indicates that oxygen dissolution can occur even at room temperature, but for pure oxygen pressures close to the atmosphere (see Section S2 in Supporting Information). However, the oxygen insertion could be favored in the Au30Pd70 alloy since the lattice is slightly expanded relatively to a Pd single crystal (+1.3 %). Anyway this behavior is consistent with what is observed on Pd(110) since no surface oxide is formed before the formation of the PdO bulk oxide unlike Pd(111) and Pd(100) surfaces.39 Prior to the bulk oxide formation, Pd(111) first pass through a two-dimensional oxide Pd5O4.43 On Pd(100) the structure of the surface oxide is (√5 × √5)R27°-PdO(101),44 which

can be readily

represented by a monolayer of the PdO(101) surface of the usual oxide phase.45 No scan, passing through the (010) position, was recorded during CO oxidation reaction conditions. So we cannot conclude if this dissolution phase also occurs during the switch between the reduced to the oxidized surface. However, we noticed that the (002)PdO peak behaves similarly than during the oxide film formation (figure 4a).

Figure 9: schematic of atomic arrangements in the surface layer formed with Pd under oxidizing conditions. Metallic Pd atoms are represented in blue and the Pd2+ ions in grey, oxygen atoms are in red. (a) Pd surface with adsorbed oxygen atoms in zigzag pattern in tetrahedral sites as in the c(2x4)O and in the “complex” structure, an octahedral site is indicated by the white ellipse on the surface plane and by a dashed one for the second plane. (b) Representation of the Pd unit cell with oxygen

20 ACS Paragon Plus Environment

Page 21 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

(yellow balls) in all the available octahedral sites, those that could be filled during the dissolution in subsurface sites, in the second plane, are in red (dashed ellipse of (a)); (a1, a2, a3) coordinates defined the surface frame of Au30Pd70(110). (c) Unit cell of PdO with the usual P42/mmc structure and (d) unit cell of the HP I4/mmm structure; the PdO(101) plane is represented in blue. The axes are the same for (c) and (d) and are orientated as found for the PdO growth in this study (for the axes see figure S1b in Supporting Information)

Structure of the strained PdO phase The oxide phase observed under 100-500 mbar of oxygen and around 420-470 K, is pure Pd. It has a slightly deformed tetragonal structure grown in the [100]PdO direction with cell parameters: aoxi=2.93  0.03 Å, boxi=3.00  0.05 Å and coxi=5.39 0.01 Å; the unit cell volume is 47  1 Å3. The most common phase of PdO in ambient conditions is of PtS type and has the P4/2mmc structure with a = b = 3.046Å and c = 5.339 Å,46 giving a unit cell volume of 50.1 Å3. It will be referred as the usual or low pressure (LP) phase in the following. The PbO type phase is also mentioned,47 but not documented.48 Under high pressure (HP), PdO crystallizes in a I4/mmm structure which is slightly more elongated along the c direction, with a = b = 2.982 Å and c = 5.383 Å,49 yielding a volume of 47.9 Å3. These two bulk phases are represented in figures 9c and 9d. The measured parameters are slightly closer to the HP tetragonal phase with a contracted volume compared to the usual PdO phase. We can notice that this HP structure can be easily deduced from the Pd with oxygen in octahedral sites by an expansion of the basic square (from 2.75 x 2.75 Å2 for metallic Pd to 2.98 x 2.98 Å2 for HP PdO) and a large lengthening of the long axis (from 3.89 Å to 5.383 Å) (see figure 9b and 9d). In this transformation, the oxygen atoms lose 2 of the 6 first neighbors. In the PtS phase, each metallic ion is in the center of a rectangle with oxygen atoms at the corners forming O-Pd-O angles of 82° and 98° instead of 90° in a square.

21 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Bragg intensities Regarding the Bragg peak intensities, in the two bulk oxide phases (figure 9c and 9d), the Pd atoms are localized on the centered tetragonal lattice. They give rise to intense Bragg peaks with (h+k+l) even. In the HP phase, oxygen atoms have the same symmetry and contribute to the same peaks as Pd. It is not the same in the LP structure where oxygen generates additional peaks of low intensity, following the rule h or k even; h+k+l being odd. The most intense diffracted peaks measured for the strained palladium oxide layer correspond to the intense Bragg peaks common to the two bulk oxide phases. Otherwise the presence of small peaks found at (010)PdO is only consistent with the LP P42/mmc structure where it is induced by oxygen atoms. However, there are remarkable discordances, namely the appearance of a (001)PdO peak (cf. figure 10) forbidden in the P42/mmc symmetry and an overestimated value for the (010)PdO and (012) PdO allowed peaks. Indeed the expected ratio between the structure factor at (002)PdO and at (010)PdO is about 5 when the experimental values of F002/F010 are at least twice as big (see table 1). Table 1 : Experimental characteristic of the (001)PdO, (010)PdO and (002)PdO oxide Bragg peaks (given with their corresponding scattering vector, at L=0, q=2((H/a1)2 + (K/a2)2)½) measured for 3 samples in several experimental conditions: (1) sample oxidized under 100 mbar of pure oxygen at 420 K; (2) sample of figure 1 at the step indicated by the red arrow (O) ; (3) sample monitored in figures 6 and 8a at different steps under 1.3 mbar of CO plus 500 mbar of oxygen, first at minute 80 after heating to 470K, next line at minute 135 and last cell line 150 min after pumping down UHV and cooling to RT. The ML column gives the mean PdO thickness in number of atomic planes as deduced from the (00) CTR fitting, 1ML  1.39Å. The width q (in nm-1 and with the error bar  0.05 in any case) and the mosaicity

 (°) are reported for each peak; F002/F001 and F002/F010 are the ratio of the structure factors.

22 ACS Paragon Plus Environment

Page 22 of 37

Page 23 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

The behavior of the two peaks (010)PdO and (100)PdO is very intriguing. They share common features with the (002)PdO reflection as illustrated by values in table 1 for 3 samples in different conditions. This indicates that they are generated by the same PdO layer. For instance, all the diffraction peaks have the same width  for a given sample. It corresponds thus to the oxide mosaicity. It depends significantly on the way by which the oxide was formed (oxygen pressure and temperature) varying from 2° to 6.5°, but staying stable afterwards. Regarding the width of the momentum transfer q, it increases from (001)PdO to (002)PdO. Except during growth, this gives a lattice gradient of 10-2 and a domain size of 20 nm in the coxi direction, perpendicular to the closed packed rows of Au30Pd70(110). At minute 80 (dark purple star in figure 8a) during the oxide growth on sample (3), the peaks were wider in q and correspond to a smaller coherence domain of 15 nm. The evolution of the (001)PdO and (002)PdO peaks during the growth of the PdO layer can be seen in figure 10 (red and black curves). It shows that (001)PdO intensity was more or less constant whereas the bulk (002)PdO significantly increased. This would indicate that the (001)PdO diffraction signal comes from the PdO layer but, more precisely, from its surface. We have less information about the (010)PdO diffraction signal, but when it was measured, its structure factor had a value comparable to that of the (001)PdO one (see table 1). This indicates that they have the same origin. This means that there is no Bragg reflection added by oxygen atoms and that the PdO structure looks like that of HP PdO I4/mmm structure. These results can be explained since this PdO thin layer is in epitaxy on the alloy. Indeed, it is well known that it is a way to synthesize unusual structures that would only appear under high pressure conditions.

Figure 10: Comparison of the diffracted intensity of the (001) and (002) Bragg peaks of PdO, (blue) sample (1) under 100 mbar of pure oxygen at 420K, (red) sample (3) under a mixture of CO + O2 at 470K and during the growth of the oxide layer corresponding to the dark purple star of figure 8a and (black) the same sample but after cooling to RT and pumping down to UHV.

23 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Comparison with PdO growth on pure Pd surfaces The previous study of the growth of oxide on Pd(110) showed that PdO has relaxed and the [100]PdO axis is tilted by roughly 7° from the direction perpendicular to the Pd(110) surface.39 The (001)PdO peak can be seen in the diffraction pattern and not the (010)PdO one, as expected for the usual LP bulk-like PdO. As previously noticed, this PdO film was significantly different from the strained layer of the present work. It was actually produced in more oxidizing conditions giving a much thicker PdO layer.19A strained epitaxial PdO thin layer was already evidenced by GIXRD for Pd(100) in similar conditions of temperature.23 In this case, the oxide grows in the (101)PdO direction and upon further increase of the O2 partial pressure, a polycrystalline Pd oxide forms on the surface. Besides it was shown that on Pd(111) the oxide also grows in the [101]PdO direction.50

PdO surface As seen in figure 9c and 9d, when PdO grows in the [100]PdO direction, there is a stacking of alternatively pure Pd and PdO2 planes in the usual P4/2mmc structure. These two surfaces are non-stoichiometric and thus belong to the class of polar surfaces that are electrostatically unstable in UHV.51,52 Nevertheless, under oxygen pressure, the PdO2 plane becomes the most stable termination.53 In the HP structure, all planes are stoichiometric. Actually, the termination of the PdO layer formed on Au30Pd70(110) cannot be deduced from the X-ray analysis. Anyway, figure 10 shows that the intensity of the Bragg peaks of the oxide formed under oxygen pressure stay stable even after pumping down to UHV. Since the layers are thin (few ML), surface changes are expected to be reflected in these intensities. This also supports a structure of HP type. In this case, for each surface, a palladium atom is linked to 2 oxygen atoms instead of 4 in the bulk or in the surface plane of the usual PdO structure. This is consistent with the XPS measurement of the Pd 3d5/2 core level during the Pd oxide formation. This signal consists of three components: the bulk one at a binding energy of 334.9 eV and two oxygen-induced components shifted at higher binding energies by 1.3 eV and 0.65 eV with either four or two oxygen neighbors, respectively.43 The component corresponding to a Pd atom with 2 oxygen neighbors remains, even when the oxide is well established.54 The HP PdO surface is stoichiometric for a growth in the [001]PdO direction, but not in the [101]PdO one (cf. figure 9d). The [101]PdO growth is observed for oxidation on Pd(100) and Pd(111).43,44 The 24 ACS Paragon Plus Environment

Page 24 of 37

Page 25 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

stacking along [101]PdO growth axis consists of an alternation of Pd and oxygen planes. It has also been shown that for the usual PdO structure and in oxygen atmosphere the polar Oterminated surface could be stable.53

Activity of the PdO layer for CO oxidation The reactivity of the surface of this strained PdO is quite different from the usual bulk-like one. For instance, we showed that the PdO layer is instantly reduced upon exposure to 0.5 mbar of CO at 420 K whereas reduction starts at higher temperature, above 500 K, for PdO bulk-like NPs55 or deeply oxidized PdO layers.56 Moreover we found that this strained PdO phase is systematically associated with a high CO conversion rate into CO2. It can reach 500 CO2 molecules/s/surface atom under 500 mbar O2 + 5 mbar CO at 470K. Conversely, it was shown that the pre-oxidized Pd(110) surface was almost unreactive for CO oxidation.9 However, such a surface was obtained under strong oxidizing conditions, namely more than 20 minutes above 800 K and under 1 Torr of O2. The O2/CO critical ratio The high CO conversion rate required a O2/CO ratio similar to what was already observed for Pd(110).25 It increases with decreasing temperature: it is around 65 at 470 K and can be extrapolated to 320 at 420 K. 25 Here we found, at 420 K, that the active phase exists in a small range around 1000 with a reaction rate of about 30 CO2 molecule/s/surface atom (after the subtraction of the sample environment contribution). At 470 K, the highly active phase was observed for O2/CO between 500 and 50 correlated to the presence of a thin palladium oxide film. This film was previously formed under pure oxygen pressure (figure 1) and can be considered as pure palladium since gold segregated towards the alloy. But a O2/CO ratio of 1500 is required to form an oxide film when gold is still present close to the surface (figure 8a); this issue will be discussed further. We point out that in the case of reaction in a batch reactor, the conversion by the sample environment may play a role. Indeed, we observe that some CO is consumed. This effect can thus contribute to diminish the CO partial pressure in the reactor and so to reach the oxide formation limit. This can explain the long delay observed before the growth of the oxide peak in figure 6 at 470 K, and also the one observed at 420 K in figure 1.

25 ACS Paragon Plus Environment

ACS Catalysis 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The hyperactive regime? Even though the O2/CO limits for the highly active phase are similar to those found by the Goodman’s group,25 we did not evidence the “hyperactive” surface. Indeed, in this case, the CO2 formation rate is more than ten times higher than the maximum of our values. This would appear as a transitory surface state between (i) a CO-inhibited metallic regime for low temperature displaying a low CO2 formation rate (ii) a high-temperature regime where the CO2 formation rate would be mass transfer-limited.26,27 The mass transfer limit occurs at high pressure when the gas phase diffusion coefficient decreases and the reaction rate exceeds the rate at which the limiting reactant (CO for high O2/CO ratio) can diffuse to the catalyst surface. In this limit, a strong CO pressure gradient is established with a local CO concentration at the sample surface close to zero. During a switch from a low to a high regime, all the CO in the gas surrounding the sample surface quickly transforms into CO2. This generates a spike in the reactivity measurement and can be an alternative explanation of the hyperactive regime.57 In this scenario, the hyperactive phase is not connected to a particular surface state, but rather to the gas phase composition surrounding the catalyst. Imaging of the gas phase provided by planar laser-induced fluorescence during a switch on Pd(110) evidenced such a variation of the CO concentration profile around the sample surface.58 The measured spike intensity, as well as the transition duration, strongly depends on the experimental conditions.25 The hyperactive regime was especially observed in a batch reactor of 0.6 L and more clearly in a 62 L one; with 6 L, the volume of our batch reactor is in between. A transition between a low and a high CO2 conversion rate happened twice during the experimental procedure illustrated in figure 1: at time = 1, when the oxide peak grew “spontaneously” at 420 K correlated to an increase of the CO2 formation rate and at time = 3.2, when the temperature was raised from 420 K to 470 K. The transitory regime, which would manifest itself by a jump in the CO2 pressure, was observed in none of these two cases (see figure S4 of Supporting Information). However, the CO pressures involved were much lower:  1 mbar instead of 8 mbar in the 0.6 L reactor or 12 mbar in the 62 L one.27 CO was thus more rapidly consumed in our reactor. Besides, due to the strong exothermicity of the CO oxidation reaction, the sample temperature can rapidly increase, so that the reaction can become self-sustained.59 In this case the reaction heat induces reactant circulation, thus partially breaking down the mass transport constraints.27 But, in our setup, the temperature 26 ACS Paragon Plus Environment

Page 26 of 37

Page 27 of 37 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Catalysis

is directly measured on the surface with an IR pyrometer, allowing to instantaneously adjust the power supplied to the laser that has, moreover, a fast response (see section 3 of Supporting Information). Therefore, in the present experimental conditions, the transitory regime must not be very pronounced and so short that it cannot be detected by the mass spectrometer. The highly active surface In the range of the O2/CO pressure ratio where the strained oxide exists on Au30Pd70(110), the rate of CO2 formation is directly correlated to a highly reactive regime. The CO2 formation rate was found to vary linearly with CO pressure. It reached a little less than 500 CO2 molecules.site1.s-1

at 470 K for a O2/CO pressure ratio of 100. This value is in the same range as the high

temperature regime plateau in the Arrhenius plot of the CO2 production over Pd(100), in the side of the high temperature regime. 23,28 It was proven that on this plateau the mass-limited transport becomes important.25-28

The absolute value of the measured CO2 formation rate differs from one experimental setup to the other due to the different experimental conditions used. But relative variations are significant. In ref. 23, the CO/O2 ratio varies but the total pressure was kept constant and the high temperature CO2 formation rate did not depend on this ratio. In the measurements performed by Goodman’s group, the CO pressure was fixed and the total pressure increased with the O2/CO ratio. In both cases the plateau observed can be explained by the mass transfer limit. In our case, the total pressure can be considered to be roughly constant at 500 mbar (PCO