Impact of Oxygen Coadsorption on Intercalation of Cobalt under the h

Box 118, 22100 Lund, Sweden, Department of Physics, Uppsala University, Box 530, 75121 Uppsala, Sweden, and V. A. Fock Institute of Physics, St.-P...
4 downloads 21 Views 3MB Size
Download PDF
NANO LETTERS

Impact of Oxygen Coadsorption on Intercalation of Cobalt under the h-BN Nanomesh

2009 Vol. 9, No. 7 2780-2787

A. B. Preobrajenski,*,† M. L. Ng,‡,† N. A. Vinogradov,‡,¶ A. S. Vinogradov,¶ E. Lundgren,§ A. Mikkelsen,§ and N. Mårtensson‡,† MAX-lab, Department of Synchrotron Radiation, Lund UniVersity, Box 118, 22100 Lund, Sweden, Department of Physics, Uppsala UniVersity, Box 530, 75121 Uppsala, Sweden, and V. A. Fock Institute of Physics, St.-Petersburg State UniVersity, 198504 St.-Petersburg, Russia Received April 24, 2009; Revised Manuscript Received May 21, 2009

ABSTRACT The process of penetration of cobalt atoms through the h-BN nanomesh on Rh(111) is investigated with both spectroscopic and microscopic techniques. It is discovered that oxygen coadsorption can drastically modify the physical properties and behavior of the deposited Co clusters upon postannealing. In the absence of oxygen, Co forms small nanoparticles in the pores (bonding parts) of the h-BN nanomesh, which start to agglomerate at elevated temperatures without any considerable intercalation. However, even a tiny amount of coadsorbed oxygen reduces cobalt agglomeration and greatly promotes its intercalation and trapping under h-BN. The oxygen exposure necessary for a complete intercalation of 1-2 monolayers of Co is very low, and the formation of oxidic species can be easily avoided. The nanomesh structure remains intact upon intercalating submonolayer amounts of Co, while further intercalation gradually distorts and finally destroys the periodic corrugation. Fortunately, this process is not accompanied by damaging the h-BN sheet itself, and the original structure can be restored by removing Co upon annealing at higher temperatures.

Introduction. The structure and properties of atomically sharp interfaces between sp2-bonded materials and metallic surfaces are intensively studied nowadays in an effort to utilize their peculiar characteristics in electronics and nanotechnology. Among the most prominent examples of such systems are monolayer graphite (graphene) and hexagonal boron nitride (h-BN) grown on surfaces of transition metals (TMs) by chemical vapor deposition.1 The strength of chemical interaction, adsorbate-substrate separation and the sp2-layer morphology can vary considerably on different substrates for the monolayers of graphite1-4 and h-BN.1,5-12 The properties of the contacts between layered materials and TMs can be further modified in a controllable way by inserting other metal species under the monolayers of sp2hybrids. This fact has stimulated extensive studies of thermally activated metal intercalation under graphene3,13-16 and h-BN monolayer17,18 on different TM surfaces. For practical applications, it is absolutely crucial to control the abruptness of these interfaces. Therefore, it is very important * To whom correspondence should be addressed. E-mail: alexeip@ maxlab.lu.se. † MAX-lab, Lund University. ‡ Uppsala University. ¶ St.-Petersburg State University. § Department of Synchrotron Radiation, Lund University. 10.1021/nl901316p CCC: $40.75 Published on Web 06/03/2009

 2009 American Chemical Society

to identify and understand the factors that can considerably affect the rate of intercalation. The interfacial modification by intercalates can be performed not only for relatively flat graphenelike sheets on TMs, but also for strongly corrugated ones. It is well established by now that for certain combinations of significant interfacial chemisorption and lattice mismatch, the growth of sp2-hybrids on TM surfaces may result in a heavily and regularly buckled monolayer, often referred to as a “nanomesh”, due to their periodicity of a few nanometers. Such nanomeshes were observed first by scanning tunneling microscopy (STM) and ultraviolet photoelectron spectroscopy for h-BN on Rh(111)19 and Ru(0001),20 then their structure was calculated theoretically21,22 and confirmed experimentally by STM23 and core-level spectroscopies.10,24 Very recently, similar nanomeshes were also observed (or predicted theoretically) for graphene on Rh(111)4 and Ru(0001).4,25-27 For graphene on Ir(111), a somewhat lower degree of corrugation has been suggested,4 which is nevertheless sufficient to provide a template function to graphene.28 Evidently, the intercalation of atomic or molecular species can significantly influence the structure and the template functionality of such h-BN and graphene nanomeshes. In this paper, we report on the controllable intercalation of cobalt under h-BN nanomesh on Rh(111) studied by a

combination of core-level photoelectron spectroscopy (PES), near-edge X-ray absorption fine structure (NEXAFS) spectroscopy and STM. Co is a promising metal for growing ensembles of magnetic nanostructure. Recently it has been shown that well-ordered arrays of Co particles with a narrow size distribution can be grown on h-BN/Rh(111) by utilizing soft landing and low growth temperatures.29 Any intercalation is clearly a negative factor for the narrowness of the size and shape distribution of the Co particles. Our primary goal in the present work is to investigate the prospects for controlling the rate of thermally activated intercalation of Co by using small amounts of coadsorbed oxygen. We demonstrate that even tiny amounts of coadsorbed O2 can drastically stimulate penetration of cobalt atoms through h-BN, and discuss the mechanisms of this process. Furthermore, we show how gradual Co penetration affects the periodicity of the h-BN nanomesh and study the prospects for restoring its original structure by removing the intercalated Co. Experiment. All spectroscopic measurements were performed at beamline D1011, MAX-lab, Lund University. PE spectra were measured with a Scienta SES-200 electron energy analyzer. The kinetic energy resolution was set to 75 meV for the N 1s and B 1s spectra. The photon energy resolution was set to 50 meV at the B K-edge (∼190 eV) and to 150 meV at the N K-edge (∼400 eV). The base pressure during the measurements did not exceed 2 × 10-10 mbar. The NEXAFS spectra were recorded in the partial electron yield mode (U ) -100 V) by a multichannel plate detector and normalized to the background curves recorded from clean substrates. Core-level PES data were analyzed and fitted with the FitXPS program.30 The Rh(111) single crystal was cleaned with several cycles of Ar+ sputtering, annealing in oxygen and high-temperature flashing. h-BN nanomesh was prepared by slow thermal decomposition of vaporized borazine (ca. 30 min at T) 780 °C and p[B3N3H6] ) 2 × 10-8 mbar). All PES and NEXAFS spectra, as well as the low-energy electron diffraction (LEED) patterns from the nanomesh were identical to those reported earlier.24 Cobalt was evaporated by electron bombardment of a cylindrical Co rod kept at the potential of ca. +1 kV at the deposition rate of 0.1 ML/min in two regimes: either in ultrahigh vacuum (UHV) or in the low-pressure oxygen ambience (p[O2] ) 8 × 10-9 mbar). Hereafter we will refer to this oxygen pressure as “mild O2 atmosphere”. This preparation results roughly in one adsorbed oxygen atom per three or four Co atoms, as can be judged from photoemission spectra. One monolayer (ML) Co is determined as a single atomic layer in the hcp structure. In different experiments, the samples were postannealed to different temperatures in the range between 300 and 700 °C. All critical preparations and measurements were repeated several times. In order to obtain reference spectra, a monolayer of h-BN was grown also on the Co(0001) single crystal. The temperatures necessary for the borazine cracking and h-BN growth (T ) 750 °C) are much higher than the hcp-to-fcc phase transition temperature in cobalt (T)420 °C), resulting in the macroscopic roughening of the Co(0001) Nano Lett., Vol. 9, No. 7, 2009

surface due to a dislocation network. However, the LEED pattern from this surface remains hexagonal, and we believe that the microscopic details of bonding between h-BN and hexagonal Co face are not affected strongly by this phase transformation, because the structure of the topmost layer is the same for Co(0001) and Co(111). Identical samples were grown for STM measurements. The STM images were recorded using a commercial Omicron TS2 Scanning Tunneling Microscope, operated at room temperature. The STM was positioned inside an UHV system with a pressure better than 1 × 10-10 mbar. The STM tips used in these experiments were chemically etched tungsten tips cleaned by Ar+ sputtering. The images were processed with the WSxM software.31 Results and Discussion. First we will describe the process of thermally activated and oxygen-assisted intercalation of Co under the h-BN nanomesh, and then we will elucidate the specific role of oxygen in this process. Prior to Co deposition, the nanomesh is visible in STM as a highly periodic structure with a period of 3.2 nm, as described in detail previously19,23 and as illustrated in Figure 1a. The inset shows a LEED pattern from one of the six superstructure spots; the entire diffraction pattern can be found elsewhere.19,24 The highly corrugated monolayer of h-BN can easily cover the step edges like an elastic blanket, without breaking the sp2 bonds, but the phase of corrugation is evidently changed at the step edge. The step shape in the system is dictated by the Rh(111) substrate, resulting in many relatively long and straight edges, as the one shown in Figure 1a. The appearance of the surface changes drastically after depositing 0.5 ML Co in mild O2 atmosphere and postannealing it at 450 °C for 30 min (see Figures 1b and 2a), although the nanomesh structure and periodicity are preserved. The shape of the step edges is not straight any more, but rather random, and new flat islands form on the terraces. In addition, many defective sites looking like cavities are visible, which were absent on the pristine nanomesh. The diffraction superstructure spots (Figure 1b, inset) look now somewhat more diffuse, indicating slight disorder, and also their relative intensities have changed. The change in their relative intensities is indicative of a considerable variation in the scattering properties of the surface providing evidence for structural changes after the Co/O treatment. On the basis of the PES overview spectra in Figure 1c, we conclude that Co is embedded between the metal surface and the h-BN nanomesh after annealing. Indeed, since both spectra are normalized to the same intensity of the Rh 3d substrate peak, the increase of N 1s and B 1s intensities upon annealing implies that Co is no longer on top of h-BN, but rather below it. Note that in the case of Co desorption both substrateand h-BN-related peaks would decrease proportionally. The topography profiles across the cobalt-induced islands presented in Figure 2b show that their height (0.24 ( 0.02 nm) roughly corresponds to the 1 ML-thick Co islands (the step-height of Co is about 0.22 nm). The corrugation amplitude of the nanomesh on the surface of these islands and around them is roughly the same. The images in Figures 2781

Figure 2. (a) 135 × 135 nm2 STM image of the h-BN nanomesh on Rh(111) with 0.5 ML Co intercalated underneath (Vs ) 1 V, I ) 0.1 nA); note formation of 2D islands under the h-BN monolayer. (b) Topography profiles along the lines denoted in panel a.

Figure 1. 140 × 140 nm2 STM images of (a) the clean h-BN nanomesh on Rh(111) (Vs ) -1 V, I ) 0.2 nA) and (b) the same nanomesh upon depositing 0.5 ML Co in the presence of O2 (p ) 8 × 10-9 mbar) followed by annealing at 450 °C in UHV (Vs ) 1 V, I ) 0.1 nA); (insets) LEED patterns of a single superstructure spot, E ) 47 eV. (c) Overview XPS spectra (hν ) 730 eV) showing an increase of the N and B signals as compared to Rh signals as a result of Co intercalation upon annealing.

1 and 2 can be interpreted in the following way. All cobalt atoms (0.5 ML) are intercalated between Rh and h-BN under these conditions, probably through structural defects. The intercalated Co atoms seem to spread easier along the step edges from where the fronts of intercalation propagate 2782

laterally along the terraces. Some intercalation may also be initiated directly on the flat terraces. This picture can be further supported by the analysis of the cavities left on the terraces presumably due to the local lack of cobalt. These cavities seem to be distributed randomly at the first glance, but in fact they appear only at sites with a certain symmetry. This is visible in the STM image of Figure 3, showing a terrace with Co intercalated everywhere except an area in the middle and a number of small holes, which are marked with crosses in Figure 3b. The positions of the crosses within every rhombic moire´ cell are almost always the same (in the middle of the left part of the cell in the case of Figure 3). Moreover, these positions coincide with the position of the nanomesh pores formed on the lower lying atomic layer, which are denoted by circles. Evidently, the imperfections appear at the positions of the original nanomesh pores, implying that intercalation happens easier along the wires of the nanomesh. This is logical, because pores are stronger bound to the substrate than the wires.10,21 Probably the most energetically favorite position for intercalated Co is the least bound nanomesh site (i.e., (B,N) ) (hcp,fcc) according to the calculations by Laskowski et al.21,22), while the least favorable is the most bound (pore) site (i.e., (B,N) ) (fcc,top)21,22), as illustrated in Figure 3c. The above interpretation of the STM images does not imply, however, that intercalation happens only or mainly under the wires. Instead, the entire nanomesh becomes lifted Nano Lett., Vol. 9, No. 7, 2009

Figure 4. B 1s PE spectra from 0.5 ML Co deposited on h-BN/ Rh(111) at RT in the presence of O2 (p ) 8 × 10-9 mbar) before (a) and after (b) postannealing at 450 °C in UHV; B 1s PE spectrum from a single h-BN monolayer adsorbed on Co(0001) is shown as a reference (c). The dashed component in panel c accounts for a weak and broad high-energy satellite at ∼193.5 eV.

Figure 3. (a) 25 × 25 nm2 STM images of the h-BN nanomesh on Rh(111) with 0.5 ML Co intercalated underneath (Vs ) 1 V, I ) 0.1 nA); (b) same image with the superimposed mesh of moire´ supercells; regular imperfections are denoted with crosses, nanomesh pores in the lower atomic layer are denoted with circles, and the imperfection, which does not fit in the regular pattern, is denoted with a question mark. (c) Schematics of Co intercalation; darker sites are more preferable.

if a sufficient amount of Co is supplied. A very puzzling issue here is why the same corrugated structure of the h-BN monolayer is preserved after embedding 2D 1 ML-thick Co islands at the h-BN/Rh(111) interface. Since hcp cobalt surface is nearly lattice matched to h-BN, it would be natural to expect that h-BN will abandon its corrugated structure and become more or less flat on a Co monolayer. It will be shown later that this indeed happens, but only if the amount of Co exceeds locally 1 ML. A possible explanation for the stability of the corrugated structure upon embedding submonolayer amounts of Co is that intercalated Co is rigidly attached to the h-BN monolayer in a lattice-matched fashion. Nano Lett., Vol. 9, No. 7, 2009

On the other hand, the interaction between Co and Rh(111) is considerable and should result in the existence of preferential adsorption sites. It is known that even on the less reactive Pt(111) surface the first Co monolayer grows quasi-epitaxially with Co atoms located in the Pt fcc lattice sites.32 As soon as Co is not free, but bound to the h-BN monolayer, it cannot occupy only positions favorable for adsorption. The variation in the strength of attractive forces between favorable and unfavorable adsorption sites leads to a corrugation of the Co monolayer attached to h-BN, just like the corrugation of h-BN itself on Rh(111). In this way the first intercalated Co monolayer rigidly follows the structure of the h-BN nanomesh. On the other hand, it cannot be excluded that the postannealing results in the formation of some Co-Rh surface alloys. In the density functional calculations of the structural and magnetic properties of Co on Rh(111), Dennler and co-workers demonstrated that Co tends to segregate into the subsurface layer at all studied Co concentrations.33 Thus, the annealing required for Co intercalation may result not in the 2D Co islands squeezed between Rh and h-BN, but rather in the Co enrichment of the subsurface layer. In order to check our conclusions regarding Co intercalation, we performed high-resolution B 1s (Figure 4) and N 1s (Figure 5) PES measurements for the 0.5 ML of oxygenactivated Co on h-BN/Rh(111) before and after the postannealing. As long as Co is deposited but not intercalated, both core-level PES spectra of h-BN are nearly identical to those from the pristine nanomesh.10,24 Each of the B 1s (Figure 4a) and N 1s (Figure 5a) signals are made up of two components corresponding to the pores (higher binding energies, BEs) and wires (lower BEs) of the h-BN nanomesh. 2783

Figure 5. N 1s PE spectra from 0.5 ML Co deposited on h-BN/ Rh(111) at RT in the presence of O2 (p ) 8 × 10-9 mbar) before (a) and after (b) postannealing at 450 °C in UHV; N 1s PE spectrum from a single h-BN monolayer adsorbed on Co(0001) is shown as a reference (c).

Figure 6. B 1s NEXAFS spectra from 0.5 ML Co deposited on h-BN/Rh(111) at RT in the presence of O2 (p ) 8 × 10-9 mbar) before (a) and after postannealing in UHV at 450 °C (b) and at 700 °C (c).

Upon annealing at 450 °C, both spectral profiles are modified drastically, but in a similar manner. Note that in our peak fit analysis we assumed that some fractions of the nanomesh may remain untouched due to the incomplete intercalation. Therefore, we postulated that the original spectral shapes shown in Figures 4a and 5a can still be present after the postannealing and kept these spectral shapes and their energies nearly constant throughout the fitting. Furthermore, we assumed that the variations in the spectra can be explained by adding a new component corresponding to the chemical bonds between Co and h-BN. These new components clearly dominate the spectra in Figures 4b and 5b and appear at BEs of 190.4 eV (B 1s) and 398.4 eV (N 1s). The shape and BEs of the new components are very similar to those in the single-peak B 1s and N 1s spectra measured from the reference h-BN/Co(0001) interface (Figure 4c and Figure 5c, respectively). This similarity supports our conclusion that the new components in both spectra result from the h-BN-Co chemical bond, thus confirming the validity of our peak-fit analysis. The fact that only one component remains in both B 1s and N 1s PE spectra upon Co intercalation conforms to the model of Co atoms rigidly attached to the h-BN monolayer from below and following the original corrugation. For future discussion on the role of oxygen, it should be noted here that the B 1s and N 1s spectra remain almost unchanged upon annealing if Co is deposited without any oxygen coadsorption (not shown). The new chemical bonds formed upon oxygen-assisted intercalation of Co under the h-BN nanomesh are reflected also in the NEXAFS spectra of h-BN. In Figure 6a, we show the B 1s NEXAFS spectrum recorded after Co/O deposition but before the postannealing. The spectrum is the same as that from the pristine h-BN nanomesh on Rh(111),24 hence

no new bonds with h-BN are formed at this stage. However, new spectral features (denoted a1, a2, and a3) appear in the B 1s NEXAFS spectrum upon postannealing (Figure 6b). Feature a3 at 193.9 eV results mainly from the B-O bonds, because its intensity strongly depends on the O2 pressure during the Co deposition (see Supporting Information, Figure S3). These bonds can be formed because molecular oxygen chemisorbs dissociatively on Co forming very active atomic oxygen, which can participate in reactions at elevated temperatures. Note that molecular oxygen (without Co) cannot attack the h-BN monolayer; even at p(O2) ) 1 × 10-6 mbar, T ) 700 °C and exposure of 1000 L we could not see any changes in neither PES nor NEXAFS spectra of the nanomesh (not shown). The other two new details in the B 1s NEXAFS, a1 and a2, can be associated with the formation of h-BN-Co bonds accompanying cobalt intercalation. In fact, similar features might appear in this spectrum due to the defects (nitrogen vacancies) induced in h-BN.34 However, in our situation features a1 and a2 cannot be due to the N-vacancies, because the spectrum returns to its original shape at higher annealing temperatures (Figure 6c), when both cobalt and oxygen signals are vanished due to desorption and/or dissolving in the substrate. Since this annealing does not supply new nitrogen, it cannot heal any N-vacancies in the h-BN monolayer. Therefore, features a1 and a2 are not due to these vacancies, but due to the h-BN-Co bonds. The reversibility of the spectral shape in Figure 6 is also a clear indication that neither Co nor O imposes any visible damage on the h-BN monolayer upon Co intercalation. As mentioned earlier, the deposition of Co without O2 coadsorption (in UHV) followed by postannealing would not cause obvious changes to the B 1s and N 1s PE spectra.

2784

Nano Lett., Vol. 9, No. 7, 2009

Figure 7. 140 × 140 nm2 STM images of 1.5 ML Co deposited on h-BN/Rh(111) (a) in UHV (Vs ) -1 V, I ) 0.1 nA) and (b) in the presence of O2 (p ) 8 × 10-9 mbar) (Vs ) -1.6 V, I ) 0.2 nA); both samples are postannealed at 450 °C in UHV. (c) Co 2p NEXAFS spectra from cobalt intercalated under the h-BN nanomesh; “mild O2” and “high O2” correspond to oxygen pressure (during Co deposition) of 8 × 10-9 mbar and 2 × 10-7 mbar, respectively.

The same is true for the NEXAFS spectra of h-BN. The overview PE spectra (same range as for those presented in Figure 1c) also remained unchanged before and after annealing in this case, suggesting a lack of Co intercalation in the absence of oxygen. The influence of O2 coadsorption on the rate of Co intercalation has been studied further with STM. Figure 7 shows STM images of the h-BN/Rh(111) interface covered with 1.5 ML Co and postannealed at 450 Nano Lett., Vol. 9, No. 7, 2009

°C for ca. 30 min, whereby in Figure 7a Co is deposited in UHV, while in Figure 7b it is deposited in a mild O2 atmosphere. Evidently, the difference in morphology of the surface is drastic. Without oxygen coadsorption, Co forms a continuous layer of clusters on the nanomesh, which are partly agglomerated at this coverage. The clusters have preferentially hexagonal shape and their lateral size is between 1 and 5 nm (a). Similar clusters can be observed also just after depositing at RT without annealing, as demonstrated for Co grown on h-BN/Ni(111).17 It is also known that Co clusters can be arranged in the pores of the h-BN nanomesh if they are grown with a buffer layer at low T.29 However, in contrast to the UHV deposition, no Co clusters remain on top of h-BN upon annealing if Co is coadsorbed with oxygen (Figure 7b). It is evident here that Co is intercalated under the nanomesh, since the periodicity of the mesh is somewhat distorted at this coverage (1.5 ML Co). It is also clear that the intercalation does not happen very evenly, because sites with local Co concentration higher than 1 ML are visible. Thus, our STM results confirm that the impact of oxygen on the Co intercalation process is profound. As demonstrated above, the amount of oxygen necessary to promote the intercalation process is very low, that is, one atom of oxygen for several Co atoms is sufficient. At this concentration no oxidic Co species are formed, as can be seen in the rather “metallic” shape of the Co 2p NEXAFS spectrum obtained from the samples with Co intercalated under h-BN nanomesh with oxygen assistance (upper spectrum in Figure 7c). For comparison, we tried higher O2 pressures during Co adsorption. Already at the pressure of 2 × 10-7 mbar Co becomes partly oxidic with the corresponding change of the Co 2p NEXAFS spectral shape, which evolves toward the shape characteristic for CoO (lower spectrum in Figure 7c). Hence, it is sufficient to use O2 at low pressures to stimulate Co intercalation in order to keep Co metallic. The effect described above raises the question about the specific role of oxygen in facilitating the intercalation. Oxygen is well known to act as a surfactant, that is, wetting agent to promote layer-by-layer growth of Co,35 because it can effectively reduce the surface free energy of a Co film, thus preventing clustering and interdiffusion. On an inert substrate like h-BN, Co always tends to form islands to reduce the surface free energy, especially at elevated temperatures. This process can be effectively suppressed by a small amount of oxygen, which may help to transform 3D islands into 2D islands, as has been shown, for example, for Ni and Cu grown on TiO2.36 Furthermore, it can be safely assumed that oxygen not only reduces the size of 3D Co clusters, but also causes a decrease in the strength of the metal-metal bonds in individual particles. Indeed, it is commonly known that atoms and molecules chemisorbed on metal surfaces may often lead to weakening of the metal-metal bonds in the substrate.37 Moreover, this weakening can be much more pronounced in a metal cluster than within a metal surface due to the reduced coordination numbers and hence stronger adsorbate-metal bonds in small 2785

clusters.38,39 In our case the O2 molecules readily dissociate on the Co nanoparticles forming strong O-Co bonds on their surfaces, which can weaken the Co-Co bonds. The smaller the original particles, the more effective is the action of oxygen. This is the reason why it is more efficient to coadsorb O2 with Co, rather than postadsorb it. Upon annealing, the bond-weakened Co clusters gain thermal energy and may dissociate into still smaller particles or individual atoms, thus facilitating the intercalation at step edges and point defects. This scenario is in agreement with the fact that Co can directly (without oxygen assistance) intercalate below h-BN monolayer if deposition happens at elevated temperatures, like in the case of Co/h-BN/Ni(111).17 In this case, intercalation and formation of 3D islands can happen simultaneously. The main effect of intercalation (upon annealing) on the O 1s PE spectra is a strong (2.4 eV) shift of the main line to higher binding energies (see Supporting Information, Figure S1). This can be interpreted as a result of different chemical bonding for oxygen: the O-Co bonds before annealing, and O-B bonds after annealing. Indeed, the binding energy of the O 1s electrons must be lower if oxygen is surrounded by metal atoms (electron suppliers), than in the case of O-B bonds. Therefore, we conclude that O atoms migrate from Co to h-BN upon annealing and Co intercalation. If much higher oxygen exposures are used, the formation of the O-B bonds can be directly seen in the B 1s PE spectra (see Supporting Information, Figure S2). In this case oxygen is present in both states: as O-B and as O-Co (in the CoO form). It is evident that the temperature of annealing is also a critical factor in the competition between Co 3D-agglomeration and intercalation under h-BN. Even if the metal clusters are affected by oxygen, intercalation is suppressed if the annealing temperature is too low. This is illustrated in Figure 8. At RT no intercalated Co can be detected, while at T ) 300 °C both on-top 3D and intercalated 2D islands can coexist. Still higher temperatures (above 400 °C) are necessary to complete the intercalation process. Finally, we demonstrate the effect of “overinflating” the h-BN nanomesh by excessive cobalt. If more than 1 ML of Co is embedded under h-BN, the nanomesh structure starts to lose its periodicity, as shown in the STM image in Figure 9. Structural distortions are more pronounced at places where local Co concentration below the nanomesh is higher. The reason for this behavior is that the h-BN monolayer gradually loses the coherence with the Rh(111) substrate with increasing amount of intercalated Co. As mentioned above, 1 ML thick intercalated Co (or Co-Rh alloy) islands can be formed without disrupting the periodicity of the Rh(111) substrate, thus maintaining the h-BN corrugation (Figure 2). In contrast, thicker Co films try to accommodate the closely packed structure of cobalt, and the nanomesh corrugation becomes relaxed. In the limit of a really thick film with the structure of the Co(0001) or Co(111) surface, the h-BN monolayer should becomes nearly flat due to a very small lattice mismatch. The transition from the original corrugation to a flat monolayer seems to be not gradual. Instead, at a certain 2786

Figure 8. 80 × 80 nm2 STM images of 0.2 ML Co deposited on h-BN/Rh(111) at RT in the presence of O2 (p ) 8 × 10-9 mbar) (a) before (Vs ) 2 V, I ) 0.04 nA) and (b) after postannealing at 300 °C in UHV (Vs ) 2 V, I ) 0.04 nA).

Figure 9. Effect of overinflating the nanomesh: 1.5 ML Co intercalated under h-BN nanomesh on Rh(111) (140 × 140 nm2 STM image, Vs ) -1 V, I ) 0.3 nA). The numbers denote some peculiar features: (1) 2D island due to higher local Co concentration; (2) breakdown of the nanomesh wires and formation of ripples with higher periodicity; (3) not intercalated Co droplet.

local concentration of Co the nanomesh wires break in ripples with higher periodicity, as visible in Figure 9. A formation of ordered Co-Rh surface alloys can also be a reason for this new periodicity. Nano Lett., Vol. 9, No. 7, 2009

Conclusions. In summary, we have studied the process of thermally activated Co intercalation under h-BN nanomesh grown on Rh(111). It has been demonstrated that even tiny amounts of coadsorbed oxygen can dramatically increase the rate of Co intercalation. We attribute this effect to the suppressed agglomeration of Co and weakening of the Co-Co bonds in individual nanoparticles in the presence of oxygen. Oxygen may act as a surfactant even in very small amounts, so that the formation of Co oxidic species can be avoided. Co penetrates somewhat easier under the wires of the nanomesh than under the pores. The embedded Co forms chemical bonds with the h-BN monolayer. The original corrugation of the h-BN sheet is preserved on up to the 1 ML-thick Co islands buried under h-BN, but for thicker Co films inserted between Rh and h-BN the nanomesh structure becomes distorted. Acknowledgment. We are grateful for the financial support by the Swedish Foundation for Strategic Research, the Crafoord foundation, the Edwin and Anna Berger foundation, the Knut and Alice Wallenberg foundation, the Swedish Research Council and the Russian Foundation for Basic Research (Grant N09-02-01278). Supporting Information Available: O 1s photoelectron spectra of oxygen coadsorbed with Co on h-BN/Rh(111) before and after postannealing (Figure S1), B 1s photoelectron spectra (Figure S2) and B 1s NEXAFS spectra (Figure S3) from heavily oxidized and annealed Co/h-BN/Rh samples. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Oshima, C.; Nagashima, A. J. Phys.: Condens. Matter 1997, 9, 1. (2) Aizawa, T.; Souda, R.; Ishizawa, Y.; Hirano, H.; Yamada, T.; Tanaka, K.; Oshima, C. Surf. Sci. 1990, 237, 194. (3) Gall, N. R.; Rut’kov, E. V.; Tontegode, A. Ya. Int. J. Mod. Phys. B 1997, 11, 1865. (4) Preobrajenski, A. B.; Ng, M. L.; Vinogradov, A. S.; Mårtensson, N. Phys. ReV. B 2008, 78, 073401. (5) Paffet, M. T.; Simonson, R. J.; Papin, P.; Paine, R. T. Surf. Sci. 1990, 232, 286. (6) Nagashima, A.; Tejima, N.; Gamou, Y.; Kawai, T.; Oshima, C. Phys. ReV. Lett. 1995, 75, 3918. (7) Rokuta, E.; Hasegawa, Y.; Suzuki, K.; Gamou, Y.; Oshima, C.; Nagashima, A. Phys. ReV. Lett. 1997, 79, 4609. (8) Auwa¨rter, W.; Kreutz, T. J.; Greber, T.; Osterwalder, J. Surf. Sci. 1999, 429, 229.

Nano Lett., Vol. 9, No. 7, 2009

(9) Preobrajenski, A. B.; Vinogradov, A. S.; Mårtensson, N. Surf. Sci. 2005, 582, 21. (10) Preobrajenski, A. B.; Nesterov, M.; Ng, M. L.; Vinogradov, A. S.; Mårtensson, N. Chem. Phys. Lett. 2007, 446, 119. ´ avar, E.; Westerstro¨m, R.; Mikkelsen, A.; Lundgren, E.; Vinogradov, (11) C A. S.; Ng, M. L.; Preobrajenski, A. B.; Zakharov, A. A.; Ma˚rtensson, N. Surf. Sci. 2008, 602, 1722. (12) Laskowski, R.; Blaha, P.; Schwarz, K. Phys. ReV. B 2008, 78, 045409. (13) Nagashima, A.; Tejima, N.; Oshima, C. Phys. ReV. B 1994, 50, 17487. (14) Shikin, A. M.; Prudnikova, G. V.; Adamchuk, V. K.; Moresco, F.; Rieder, K.-H. Phys. ReV. B 2000, 62, 13202. (15) Dedkov, Yu. S.; Shikin, A. M.; Adamchuk, V. K.; Molodtsov, S. L.; Laubschat, C.; Bauer, A.; Kaindl, G. Phys. ReV. B 2001, 64, 035405. (16) Gru¨neis, A.; Vyalikh, D. V. Phys. ReV. B 2008, 77, 193401. (17) Auwa¨rter, W.; Muntwiler, M.; Greber, T.; Osterwalder, J. Surf. Sci. 2002, 511, 379. (18) Goriachko, A.; He, Y. B.; Over, H. J. Phys. Chem. C 2008, 112, 8147. (19) Corso, M.; Auwa¨rter, W.; Muntwiler, M.; Tamai, A.; Greber, T.; Osterwalder, J. Science 2004, 303, 217. (20) Goriachko, A.; He, Y.; Knapp, M.; Over, H. Langmuir 2007, 23, 2928. (21) Laskowski, R.; Blaha, P.; Gallauner, T.; Schwarz, K. Phys. ReV. Lett. 2007, 98, 106802. (22) Laskowski, R.; Blaha, P. J. Phys.: Condens. Matter 2008, 20, 064207. (23) Berner, S.; Corso, M.; Widmer, R.; Groening, O.; Laskowski, R.; Blaha, P.; Schwarz, K.; Goriachko, A.; Over, H.; Gsell, S.; Schreck, M.; Sachdev, H.; Greber, T.; Osterwalder, J. Angew. Chem., Int. Ed. 2007, 46, 5115. (24) Preobrajenski, A. B.; Vinogradov, A. S.; Ng, M. L.; C´avar, E.; Westerstro¨m, R.; Mikkelsen, A.; Lundgren, E.; Mårtensson, N. Phys. ReV. B 2007, 75, 245412. (25) Marchini, S.; Gu¨nther, S.; Wintterlin, J. Phys. ReV. B 2007, 76, 075429. (26) Wang, B.; Bocquet, M.-L.; Marchini, S.; Gu¨nther, S.; Wintterlin, J. Phys. Chem. Chem. Phys. 2008, 10, 3530. (27) Brugger, T.; Gu¨nther, S.; Wang, B.; Dil, J. H.; Bocquet, M.-L.; Osterwalder, J.; Wintterlin, J. Phys. ReV. B 2009, 79, 045407. (28) N’Diaye, A. T.; Bleikamp, S.; Feibelman, P. J.; Michely, T. Phys. ReV. Lett. 2006, 97, 215501. (29) Brihuega, I.; Michaelis, C. H.; Zhang, J.; Bose, S.; Sessi, V.; Honolka, J.; Schneider, M. A.; Enders, A.; Kern, K. Surf. Sci. 2008, 602, L95. (30) Adams D. L. FitXPS (version 2.12). http://www.sljus.lu.se/ download.html (accessed May 28, 2009). (31) Horcas, I.; Ferna´ndez, R.; Go´mez-Rodrı´guez, J. M.; Colchero, J.; Go´mez-Herrero, J.; Baro, A. M. ReV. Sci. Instrum. 2007, 78, 013705. (32) Lundgren, E.; Stanka, B.; Schmid, M.; Varga, P. Phys. ReV. B 2000, 62, 2843. (33) Dennler, S.; Hafner, J.; Marsman, M.; Morillo, J. Phys. ReV. B 2005, 71, 094433. (34) Jime´nez, I.; Jankowski, A. F.; Terminello, L. J.; Sutherland, D. G. J.; Carlisle, J. A.; Doll, G. L.; Tong, W. M.; Shuh, D. K.; Himpsel, F. J. Phys. ReV. B 1997, 55, 12025. (35) To¨lkes, C.; Struck, R.; David, R. Phys. ReV. Lett. 1998, 80, 2877. (36) Zhou, J.; Kang, Y. C.; Ma, S.; Chen, D. A. Surf. Sci. 2004, 562, 113. (37) Somorjai, G. A.; van Hove, M. A. Prog. Surf. Sci. 1989, 30, 201. (38) van Santen, R. A.; Neurock, M. Catal. ReV. Sci. Eng. 1995, 37, 557. (39) van Santen, R. A.; Neurock M. Molecular Heterogeneous Catalysis; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, 2006; Chapter 3.

NL901316P

2787