Carbon films grown on Pt(1 1 1) as supports for model gold catalysts

Carbon films were grown on a Pt(1 1 1) single crystal by ethylene decomposition at elevated temperatures (1000-1300 K). Depending on the preparation conditions, different carbon structures formed on the metal surface such as flat and curved graphitic layers, carbon particles and carbon nanowires. Although these carbon films exhibited a high density of surface defects, gold interacted only weakly with the carbon surface. CO adsorption on the Au/carbon systems was very similar to that observed for various Au/oxide systems previously studied. This finding strongly indicates that CO adsorption on gold is essentially independent of the nature of support.     

High-pressure study on the adsorption and oxidation of CO on gold/titania model catalysts

The results of an IR study on the interaction of CO/O₂ gas mixtures with planar Au/TiO₂ model catalysts at elevated pressures and at room temperature are presented. The model catalysts were prepared by deposition of a flat titania film on a Ru(0 0 0 1) substrate and subsequent evaporation of gold on the titania film. In the presence of the gas mixtures, an IR band in the CO stretching region was formed, pointing to linearly adsorbed CO. The position of this band is nearly independent of the Au coverage employed. Compared to pure CO, the IR band is shifted to higher wave numbers when CO/O₂ gas mixtures are used. Although the production of CO₂ was detected in the CO oxidation reaction on the model catalysts, the formation of other IR bands, revealing the build-up of carbonates or other side-products which is usually observed for Au/TiO₂ real powder catalysts, was very weak.     

Oxygen adsorption and oxidation reactions on Au(2 1 1) surfaces: Exposures using O2 at high pressures and ozone (O3) in UHV

Nanosized gold particles supported on reducible metal oxides have been reported to show high catalytic activity toward CO oxidation at low temperature. This has generated great scientific and technological interest, and there have been many proposals to explain this unusual activity. One intriguing explanation that can be tested is that of Nørskov and coworkers [Catal. Lett. 64 (2000) 101] who suggested that the “unusually large catalytic activity of highly-dispersed Au particles may in part be due to high step densities on the small particles and/or strain effects due to the mismatch at the Au-support interface”. In particular, their calculations indicated that the Au(2 1 1) stepped surface would be much more reactive towards O₂ dissociative adsorption and CO adsorption than the Au(1 1 1) surface. We have now studied the adsorption of O₂ and O₃ (ozone) on an Au(2 1 1) stepped surface. We find that molecular oxygen (O₂) was not activated to dissociate and produce oxygen adatoms on the stepped Au(2 1 1) surface even under high-pressure (700 Torr) conditions with the sample at 300-450 K. Step sites do bind oxygen adatoms more tightly than do terrace sites, and this was probed by using temperature programmed desorption (TPD) of O₂ following ozone (O₃) exposures to produce oxygen adatoms up to a saturation coverage of θ_O = 0.90 ML. In the low-coverage regime (θ_O ? 0.15 ML), the O₂ TPD peak at 540 K, which does not shift with coverage, is attributed to oxygen adatoms that are bound at the steps on the Au(2 1 1) surface. At higher coverages, an additional lower temperature desorption peak that shifts from 515 to 530 K at saturation coverage is attributed to oxygen adsorbed on the (1 1 1) terrace sites of the Au(2 1 1) surface. Although the desorption kinetics are likely to be quite complex, a simple Redhead analysis gives an estimate of the desorption activation energy, E_d, for the step-adsorbed oxygen of 34 kcal/mol and that for oxygen at the terraces near saturation coverage of 33 kcal/mol, values that are similar to others reported on Au surfaces. Low Energy Electron Diffraction (LEED) indicates an oxygen-induced step doubling on the Au(2 1 1) surface at low-coverages (θ_O = 0.08-0.17 ML) and extensive disruption of the 2D ordering at the surface for saturation coverages of oxygen (θ_O ? 0.9 ML). Overall, our results indicate that unstrained step sites on Au(2 1 1) surfaces of dispersed Au nanoparticles do not account for the novel reactivity of supported Au catalysts for CO oxidation.     

Effects of predosed oxygen and hydrogen on CO adsorption on Ni(1 0 0)

We present the results of high-resolution electron energy loss experiments on (CO/O)/Ni(1 0 0) and (CO/H)/Ni(1 0 0) systems. Oxygen and hydrogen interact differently with Ni(1 0 0) surface, nevertheless, both species do not affect to a great extent the vibrational properties of CO. A phase of CO molecules weakly bonded to the surface and unaffected by coadsorbed oxygen and hydrogen, is found. Coverage of 0.5 ML of predosed oxygen chemically passivates the Ni(1 0 0) surface and inhibits any CO adsorption at room temperature. CO sites are unambiguously determined for each predosed Ni(1 0 0) surface.     

Adsorption of carbon monoxide Au/Pd(1 0 0) alloys in ultrahigh vacuum: Identification of adsorption sites

The adsorption of carbon monoxide is studied on Au/Pd(1 0 0) alloys by means of reflection-absorption infrared spectroscopy (RAIRS) and temperature-programmed desorption (TPD). The alloy was formed by adsorbing a four-monolayer thick gold film on a Pd(1 0 0) substrate and by heating to various temperatures to form alloys with a range of palladium coverages. The alloy was characterized using X-ray photoelectron spectroscopy and the composition of the outermost layer measured using low-energy ion scattering spectroscopy. CO adsorbs on palladium bridge sites only for palladium coverages greater than 0.5 monolayers (ML) suggesting that next-nearest neighbor sites are preferentially populated by palladium atoms. CO adsorbs on atop palladium sites and desorbs at ∼350 K corresponding to a desorption activation energy of ∼117 kJ/mol. However, at lower palladium coverages, these sites are not occupied and CO desorption states are detected 170 and 112 K corresponding to desorption activation energies of ∼53 kJ/mol and ∼35 kJ/mol, respectively, for these states. It is suggested that these states are due to a restructuring of the surface to form low-coordination gold sites that obscure the atop palladium site.     

CO adsorption on oxygen-modified molybdenum surfaces

Structures of carbon monoxide layers on the oxygen-modified Mo(1 1 0) and Mo(1 1 2) surfaces have been investigated by means of density-functional (DFT) calculations. It is found that CO molecules adsorb at hollow sites on the O/Mo(1 1 0) surface and nearly atop Mo atoms on the O/Mo(1 1 2) surface. The favorable positions for adsorption are shown to be near protrusions of electron density above the Mo surface atoms. The presence of oxygen on the molybdenum surface significantly reduces the binding energy of the CO molecule with the substrate; on the oxygen-saturated Mo(1 1 0) surface, the adsorption of CO is completely blocked. The calculated local densities of states (LDOS) demonstrate that the O 2s peak for O adsorbed on Mo(1 1 0) surface is at −19 eV (with respect to the Fermi level), while for the oxygen atom of an adsorbed CO molecule the related 3σ molecular orbital gives rise to a peak at −23 eV. This difference stems from the bonding of the O atom either with Mo surface for adsorbed O or with C atom in adsorbed CO, and therefore the position of the O 2s peak in photoemission spectra can serve as a convincing argument in favor of either the presence or absence of the CO dissociation on Mo surfaces.     

Oxidation of CO on Fe2O3 model surfaces

Using first principles calculations based on a gradient corrected density functional formalism we show that Fe₂O₃ nano-particles with (1 0 0) and (0 0 0 1) surface orientations can oxidize CO to form CO₂ with or without the presence of O₂. However, depending on the surface orientation, the oxidation occurs through differing sequences. On the (1 0 0) surface, in the absence of O₂, two CO molecules are required for one CO oxidation in a concerted reaction while on a oxygen terminated (0 0 0 1) surface, a single CO molecule itself, without the aid of a second CO, can react with the lattice oxygen atoms to form CO₂. In the presence of O₂, the O vacancies created by an initial oxidation through lattice oxygen act as the favored sites for O₂ adsorption which can subsequently oxidize the incoming CO. Detailed reaction paths and the corresponding energetics for the proposed mechanisms are also studied.     

CO interaction with Au atoms adsorbed on terrace, edge and corner sites of the MgO(1 0 0) surface. Electronic structure and vibrational analysis from DFT

The interaction of CO with Au atoms adsorbed on terrace and low-coordinates sites (edge and corner) of the MgO(1 0 0) surface was studied using the density functional theory (DFT) in combination with embedded cluster models. Surface anionic (O²⁻) and neutral oxygen vacancy (F_s) sites were considered. In all the cases, the CO stretching frequencies are shifted with respect to free CO with values between −232 and −358 cm⁻¹. In particular, the values for Au on F_s at edge and corner are shifted to higher stretching frequencies by 100 and 59 cm⁻¹, respectively, with respect to the value on a perfect MgO(1 0 0) surface. This result is in agreement with recent scanning tunneling microscopy and infrared spectroscopy experiments where a corresponding shift of 70 cm⁻¹ was observed by comparing the measurements on perfect and O-deficient MgO(1 0 0) surfaces. However, these results are different than expected because Au atoms on F_s centers are negatively charged and, therefore, according to the generally accepted scheme the CO frequency should be red-shifted with respect to the adsorption on anionic five-coordinated site where the Au atom is essentially neutral. The following picture emerges from the present results: the single occupied HOMO(α) of Au atom on F_s at low-coordinated sites consists in two lobes extended sideward the Au atom. For symmetry reasons, this MO overlaps efficiently with the 2π^∗ MO of CO. This bonding contribution to the Au-CO link is counteracted by a Pauli repulsion between the 5σ MO of CO and more internal orbitals (the HOMO-1(α) and the HOMO(β)) centered on Au. In consequence, CO is forced to vibrate against a region with a high electron density. This is the so-called “wall effect” which by itself contributes to higher CO frequency values.     

Phase mixing and phase separation accompanying the catalytic oxidation of CO on Ir{1 0 0}

The co-adsorption of CO and O on the unreconstructed (1 × 1) phase of Ir{1 0 0} was examined by low energy electron diffraction (LEED) and temperature programmed desorption (TPD). When CO is adsorbed at 188 K onto the Ir{1 0 0} surface precovered with 0.5 ML O, a mixed c(4 × 2)-(2O + CO) overlayer is formed. All CO is oxidised upon heating and desorbs as CO₂ in three distinct stages at 230 K, 330 K and 430 K in a 2:1:2 ratio. The excess oxygen left on the surface after all CO has reacted forms an overlayer with a LEED pattern with p(2 × 10) periodicity. This overlayer consists of stripes with a local p(2 × 1)-O arrangement of oxygen atoms separated by stripes of uncovered Ir. When CO is adsorbed at 300 K onto the surface precovered with 0.5 ML O an apparent (2 × 2) LEED pattern is observed. LEED IV analysis reveals that this pattern is a superposition of diffraction patterns from islands of c(2 × 2)-CO and p(2  × 1)-O structures on the surface. Heating this co-adsorbed overlayer leads to the desorption of CO₂ in two stages at 330 K and 430 K; the excess CO (0.1 ML) desorbs at 590 K.LEED IV structural analysis of the mixed c(4 × 2) O and CO overlayer shows that both the CO molecules and the O atoms occupy bridge sites. The O atoms show significant lateral displacements of 0.14 Å away from the CO molecules; the C-O bond is slightly expanded with respect to the gas phase (1.19 Å); the modifications of the Ir substrate with respect to the bulk-terminated surface are very small.     

Adsorption behavior and reaction properties of NO and CO on Rh(1 1 1)

Reactions between NO and CO on Rh(1 1 1) surfaces were investigated using infrared reflection absorption spectroscopy, X-ray photoelectron spectroscopy, and temperature-programmed desorption. NO adsorbed on the fcc, atop, and hcp sites in that order, whereas CO adsorbed initially on the atop sites and then on the hollow (fcc + hcp) sites. The results of experiments with NO exposure on CO-preadsorbed Rh(1 1 1) surfaces indicated that the adsorption of NO on the hcp sites was inhibited by preadsorption of CO on the atop sites, and NO adsorption on the atop and fcc sites was inhibited by CO preadsorbed on each type of site, which indicates that NO and CO competitively adsorbed on Rh(1 1 1). From a Rh(1 1 1) surface with coadsorbed NO and CO, N₂ was produced from the dissociation of fcc-NO, and CO₂ was formed by the reaction of adsorbed CO with atomic oxygen from dissociated fcc-NO. The CO₂ production increased remarkably in the presence of hollow-CO. Coverage of fcc-NO and hollow-CO on Rh(1 1 1) depended on the composition ratio of the NO/CO gas mixture, and a gas mixture with NO/CO ? 1/2 was required for the co-existence of fcc-NO and hollow-CO at 273 K.     

Segregation at the surface of an Au/Pd alloy exposed to CO

We use density functional theory (DFT) with the generalized gradient approximation (GGA) and the revised Perdew-Burke-Ernzerhoff (rPBE) functional, to study the surface composition of the (1 1 1) and (1 0 0) dilute Pd/Au alloy. We find that the energy of Pd atoms is lower when they substitute an Au atom in the bulk than when they substitute an Au atom in the surface layer, or when they are adsorbed on the surface. Whether they are in the surface layer or in the bulk, the Pd atoms interact very weakly with each other. CO adsorbs on the Pd atom in the surface layer and the energy of this complex is lower than that of CO in gas and Pd atom in the bulk. The interaction between the PdCO complexes formed when CO adsorbs on a Pd atom imbedded in the surface layer, is also negligible. We use these energies, equilibrium thermodynamics, and a simple lattice-gas model to examine the equilibrium composition of the surface layer, as a function of temperature, CO pressure and the Pd/Au ratio. We find that the surface Pd concentration for a nanoparticle of an Au/Pd alloy differs from that in a bulk sample. The difference is due mainly to the fact that in a nanoparticle the migration of Pd atoms to the surface depletes the bulk concentration while in a large sample; the bulk provides an infinite source of Pd atoms to populate the surface sites. This system is of interest because Pd/Au alloys are selective catalysts for vinyl acetate synthesis when the Pd concentration on the surface is very low.     

Au deposits on graphite: On the nature of high temperature desorption peaks in CO thermal desorption spectra

Due to the discovery of Au as a catalyst for low temperature CO oxidation, the adsorption of CO on Au surfaces has attracted a lot of attention recently. On stepped and rough single crystal surfaces as well as on deposited particles two characteristic desorption states above 100 K have been observed via TPD. We have studied Au deposits on graphite in order to elucidate the nature of these desorption peaks in more detail. For this purpose, Au was deposited at 100 K and 300 K on HOPG as a weakly interacting support. In analogy to other supports, we obtain two desorption states (∼140 K and ∼170 K) whose relative intensities depend strongly on the deposition temperature with the high temperature peak being much more pronounced for the 100 K deposits. After annealing to 600 K, both states drastically lose intensity. XP spectra, on the other hand, show virtually no decrease of the Au 4f intensity as would be expected for desorption or significant changes of the particle morphologies. We conclude that both desorption peaks are defect-related and connected with under-coordinated Au atoms that are lost for the most part upon annealing. These sites could be located at the perimeter of dendritic islands or on small, defect-rich particles in addition to larger particles not adsorbing CO at 100 K. Preliminary STM results are in favour of the second interpretation.     

The effect of preadsorbed K on the size distribution of Au nanoparticles on TiO2(1 1 0) surface

The effect of K on the morphology of Au nanoparticles deposited on TiO₂(1 1 0) surface is investigated by STM-STS and AES methods. For comparison, the enhanced concentration of oxygen defect sites generated by Ar⁺ bombardment was also studied. It was found that both the K additive and the oxygen defect sites induce a pronounced decrease in the average size of the Au nanoparticles evolved at 320 K. On the clean TiO₂(1 1 0) the average size of Au particles is 4.3 nm at approximately monolayer coverage of gold, while in the presence of K or oxygen vacancies this value decreased to 2.5 nm. In spite of the reduced average diameter detected at room temperature, the mean size of the Au nanoparticles increased significantly from 2.5 nm up to 7 nm on the effect of annealing at 500-700 K for K precoverages of 0.3-1 ML. For the clean and the Ar⁺ pretreated TiO₂(1 1 0) surfaces the mean size of the Au particles changed only slightly on the effect of the same thermal treatments.     

Density-functional study of the CO adsorption on ferromagnetic Co(0 0 0 1) and Co(1 1 1) surfaces

The regular CO overlayers at coverage θ = 1/3 adsorbed on the (0 0 0 1) surface of hcp Co and (1 1 1) surface of fcc Co are studied by first-principles density-functional theory with the exchange-correlation component in the PBE form. Adsorption in atop, bridge, and three-fold hcp or fcc position are considered. The adsorption energies, CO stretching frequencies, geometry, work function, and local magnetic moments are studied, and, when possible, compared with experimental or theoretical data. Particularly, we show that the recently proposed correction to adsorption energy of CO prefers correctly the atop adsorption site, whereas the remaining sites are almost degenerate in energy. The CO molecule lowers magnetization on neighbouring Co atoms, and the effect decreases with the adsorption site coordination. We show, however, that this trend is not the result of the different C-Co separation at different adsorption sites. A very small magnetic moment appears on CO that couples antiferromagnetically to Co. Most results are very similar for the Co(0 0 0 1) and Co(1 1 1) surfaces.     

Carbon monoxide adsorption on Co deposited Pt(1 0 0)-hex: IRRAS and LEED investigations

Infrared reflection absorption spectroscopy (IRRAS) was used to investigate carbon monoxide (CO) adsorption on Pt(1 0 0) surfaces deposited with Co layers with different thicknesses. Pt(1 0 0) surfaces cleaned in ultrahigh vacuum showed surface reconstruction, i.e., Pt(1 0 0)-hex: two absorption bands ascribable to adsorbed CO on the 1 × 1 surface and hex domains emerge at 2086 and 2074 cm⁻¹, respectively, after 1.0 L CO exposure. Deposition of a 0.3-nm-thick-Co layer on Pt(1 0 0)-hex at 333 K changes the low-energy electron diffraction (LEED) pattern from hex to p(1 × 1), indicating that the deposited Co lifts the reconstruction. The IRRAS spectrum for 1.0-L-CO-exposed Co_0.3 nm/Pt(1 0 0)-hex fabricated at 333 K yields a single absorption band at 2059 cm⁻¹. For Co_0.3 nm/Pt(1 0 0)-hex fabricated at 693 K, the LEED pattern shows a less-contrasted hex and the pattern remains nearly unchanged even after CO exposure of 11 L, although only 1.0 L CO exposure to Pt(1 0 0)-hex lifts the surface reconstruction. A Co_0.3 nm/Pt(1 0 0)-hex surface fabricated at 753 K exhibits an absorption band at 2077 cm⁻¹, which is considered to originate from CO adsorbed on the Pt-enriched surface alloy. Co_0.3 nm/Pt(1 0 0)-hex surfaces fabricated above 773 K show a clear hex-reconstructed LEED pattern, and the frequencies of the adsorbed CO bands are comparable to those of Pt(1 0 0)-hex, indicating that the deposited Co atoms are diffused near the surface region. The outermost surface of the 3.0-nm-thick-Co-deposited Pt(1 0 0)-hex is composed of Pt-Co alloy domains even at a deposition temperature of 873 K. Based on the LEED and IRRAS results, the outermost surface structures of Co_x/Pt(1 0 0)-hex are discussed.     

Sulfur poisoning of the CO adsorption on Co(0 0 0 1)

CO adsorption on a sulfur covered cobalt surface at 185 K has been studied using XPS, TDS, LEED, and WF measurements. As in the case of CO adsorption on the clean Co(0 0 0 1) surface, CO adsorbs and desorbs molecularly and no dissociation was observed. The saturation coverage of CO decreases linearly from 0.54 ML to 0.27 ML when the S pre-coverage increases to 0.25 ML. The WF increased during CO adsorption, but did not reach the value obtained for CO adsorption on the clean surface. The smaller work function change is explained by the reduced adsorption of CO on the sulfur-precovered surface. A reduction in the activation energy of desorption for CO from 113 kJ/mol to 88 kJ/mol was observed indicating weaker bonding of the CO molecules to the surface. The behavior of the CO/S/Co(0 0 0 1) system was explained by a combination of steric and electronic effects.     

Reaction mechanism for CO oxidation on Cu(3 1 1): A density functional theory study

The microscopic reaction mechanism for CO oxidation on Cu(3 1 1) surface has been investigated by means of comprehensive density functional theory (DFT) calculations. The elementary steps studied include O₂ adsorption and dissociation, dissociated O atom adsorption and diffusion, as well as CO adsorption and oxidation on the metal. Our results reveal that O₂ is considerably reactive on the Cu(3 1 1) surface and will spontaneously dissociate at several adsorption states, which process are highly dependent on the orientation and site of the adsorbed oxygen molecule. The dissociated O atom may likely diffuse via inner terrace sites or from a terrace site to a step site due to the low barriers. Furthermore, we find that the energetically most favorable site for CO molecule on Cu(3 1 1) is the step edge site. According to our calculations, the reaction barrier of CO + O → CO₂ is about 0.3 eV lower in energy than that of CO + O₂ → CO₂ + O, suggesting the former mechanism play a main role in CO oxidation on the Cu(3 1 1) surface.     

Au/TiO2/Ru(0 0 0 1) model catalysts and their interaction with CO

Au/TiO₂/Ru(0 0 0 1) model catalysts and their interaction with CO were investigated by scanning tunneling microscopy and different surface spectroscopies. Thin titanium oxide films were prepared by Ti deposition on Ru(0 0 0 1) in an O₂ atmosphere and subsequent annealing in O₂. By optimizing the conditions for deposition and post-treatment, smooth films were obtained either as fully oxidized TiO₂ or as partly reduced TiO_x, depending on the preparation conditions. CO adsorbed molecularly on both oxidized and reduced TiO₂, with slightly stronger bonding on the reduced films. Model catalyst surfaces were prepared by depositing submonolayer quantities of Au on the films and characterized by X-ray photoelectron spectroscopy and scanning tunneling microscopy. From X-ray photoelectron spectroscopy, a weak interaction between the Au and the TiO₂ substrate was found. At 100 K CO adsorption occurred on both the TiO₂ film and on the Au nanoparticles. CO desorbed from the Au particles with activation energies between 53 and 65 kJ/mol, depending on the Au coverage. If the Au deposit was annealed to 770 K prior to CO exposure, the CO adsorption energy decreased significantly. STM measurements revealed that the Au particles grow upon annealing, but are not encapsulated by TiO_x suboxides. The higher CO adsorption energy observed for smaller Au coverages and before annealing is attributed to a significantly stronger interaction of CO with mono- and bilayer Au islands, while for higher particles, the adsorption energy becomes more bulk-like. The implications of these effects on the known particle size effects in CO oxidation over supported Au/TiO₂ catalysts are discussed.     

Adsorption CO2 on the perfect and oxygen vacancy defect surfaces of anatase TiO2 and its photocatalytic mechanism of conversion to CO

The adsorption energies for physisorption and the most stable chemisorption of CO₂ on the neutral charge of perfect anatase [TiO₂] (0 0 1) are −9.03 and −24.66 kcal/mol on the spin-unpolarized and −12.98 and −26.19 kcal/mol on the spin-polarized surface. The small activation barriers of 1.67 kcal/mol on the spin-unpolarized surface and of 6.66 kcal/mol on the spin-unpolarized surface were obtained. The adsorption mechanism of CO₂ on the oxygen vacancy defect [TiO₂ + V_O] surface of anatase TiO₂ using density functional theory calculations was investigated. The energetically preferred conversion of CO₂ to CO was found either on the spin-unpolarized or spin-polarized surfaces of oxygen vacancy defect surface [TiO₂ + V_O] as photocatalyst.     

Formation, stability and CO adsorption properties of PdAg/Pd(1 1 1) surface alloys

The formation, stability and CO adsorption properties of PdAg/Pd(1 1 1) surface alloys were investigated by X-ray photoelectron spectroscopy (XPS) and by adsorption of CO probe molecules, which was characterized by temperature-programmed desorption (TPD) and high resolution electron energy loss spectroscopy (HREELS). The PdAg/Pd(1 1 1) surface alloys were prepared by annealing (partly) Ag film covered Pd(1 1 1) surfaces, where the Ag films were deposited at room temperature. Surface alloy formation leads to a modification of the electronic properties, evidenced by core-level shifts (CLSs) of both the Pd(3d) and Ag(3d) signal, with the extent of the CLSs depending on both initial Ag coverage and annealing temperature. The role of Ag pre-coverage and annealing temperature on surface alloy formation is elucidated. For a monolayer Ag covered Pd(1 1 1) surface, surface alloy formation starts at ∼450 K, and the resulting surface alloy is stable upon annealing at temperatures between 600 and 800 K. CO TPD and HREELS measurements demonstrate that at 120 K CO is exclusively adsorbed on Pd surface atoms/Pd sites of the bimetallic surfaces, and that the CO adsorption behavior is dominated by geometric ensemble effects, with adsorption on threefold hollow Pd₃ sites being more stable than on Pd₂ bridge sites and finally Pd₁ a-top sites.