Epitaxial Growth of Hexagonal Boron Nitride on Ir(111) | The Journal

The formation of a hexagonal boron nitride (h-BN) layer through dissociation of ... Epitaxial Growth of a Single-Domain Hexagonal Boron Nitride Monola...
0 downloads 0 Views 3MB Size
ARTICLE pubs.acs.org/JPCC

Epitaxial Growth of Hexagonal Boron Nitride on Ir(111) Fabrizio Orlando,†,‡ Rosanna Larciprete,§ Paolo Lacovig,|| Ilan Boscarato,^ Alessandro Baraldi,†,‡ and Silvano Lizzit*,|| †

Physics Department and Center of Excellence for Nanostructured Materials, University of Trieste, Via A. Valerio 2, 34127 Trieste, Italy IOM-CNR, Laboratorio TASC, AREA Science Park, S.S. 14 Km 163.5, 34149 Trieste, Italy § CNR-ISC, Via Fosso del Cavaliere 100, 00133 Roma, Italy Sincrotrone Trieste S.C.p.A., AREA Science Park, S.S. 14 Km 163.5, 34149 Trieste, Italy ^ Department of Chemical Sciences, University of Trieste, Via A. Valerio 2, 34127 Trieste, Italy

)



ABSTRACT: The formation of a hexagonal boron nitride (h-BN) layer through dissociation of borazine (B3N3H6) molecules on Ir(111) has been investigated by a combination of X-ray photoelectron spectroscopy, near-edge X-ray absorption fine structure, temperature-programmed desorption, and lowenergy electron diffraction. At low temperature (T = 170 K), molecular borazine adsorption occurs with the plane of the benzene-like ring parallel to the substrate. Dehydrogenation is observed at temperatures higher than 250 K and extends up to 900 K, with a maximum H2 desorption rate around 300 K. Besides dehydrogenation, room temperature adsorption of borazine leads to the formation of atomic and molecular fragments due to the break-up of part of the BN bonds. The epitaxial growth of h-BN starts at temperature higher than 1000 K where an extended and long-range ordered layer is obtained. The presence of a corrugation in the h-BN layer with moire periodicity of (13  13)/(12  12) BN/Ir unit cell is reflected in the double component structure of the B 1s and N 1s core level spectra.

’ INTRODUCTION Ultrathin films grown on metal surfaces have attracted considerable attention because of their potential application in nanotechnology. One of the main interests is represented by the possibility of using these systems as templates for arranging molecules in a controlled ordered fashion,1,2 therefore opening the door to the engineering of large-scale nanodevices.3 In this context, the single layer of h-BN is considered a promising candidate, especially after the discovery by Corso et al.4 of the formation of a self-assembled nanostructure, the so-called nanomesh, on the Rh(111) surface. The interest in h-BN is motivated by its appealing properties, such as the excellent thermal stability4 and, in particular, the large band gap5 which makes it an insulator isostructural to graphene,6 the new challenging material for future nanoelectronics. Recently, it has been shown that using h-BN as a substrate for graphene electronics leads to an improvement of the device performances in terms of charge carrier mobility.7 Among the several ways to synthesize a single layer of h-BN, the most common method used to obtain high-quality films is by chemical vapor deposition (CVD) of a molecular precursor, e.g., benzene-like B3N3H6, at transition metal (TM) surfaces. Most of these studies concern the growth on substrates with the same C3v symmetry as the h-BN, i.e., (111) and (0001) hexagonal surfaces of face-centered cubic (fcc) and hexagonal closed-packed (hcp) crystals, respectively. Important examples are Ni(111),811 Pt(111),8,12 Pd(111),8,13 Ru(0001),14 and Rh(111).4,15 In addition, r 2011 American Chemical Society

there are several investigations on h-BN monolayers grown on top of more open surfaces, such as Ni(110),16 Cr(110),17 Pd(110),18 and Mo(110).19 Most of the research focuses on the characterization of the morphology and electronic structure of the resulting h-BN films, but little is known on the processes that lead chemisorbed borazine molecules to form an extended and long-range ordered h-BN layer. The analysis of the island boundaries on Ni(111)20,21 and Rh(111)15 suggests that the BN bond breaks during the self-assembly process. However, up to date only a few investigations report on the adsorption and interaction of borazine with metallic surfaces.2225 Depending on the substrate, two adsorption geometries were identified: borazine bound with the molecule ring either perpendicular (Pt(111)22,23) or parallel to the substrate (Au(111)23 and Ru(0001)24). This behavior resembles that of benzene (C6H6) adsorption on TMs, although the C6H6 molecules adsorb prevalently in flat geometry.2629 An exhaustive picture of the chemisorption properties of borazine on metallic substrates is therefore still missing, motivating further studies on this topic. In the present article, we report on the formation of the h-BN layer on Ir(111) focusing on the low-temperature adsorption and the dissociation of borazine molecules up to Received: August 7, 2011 Revised: November 17, 2011 Published: December 05, 2011 157

dx.doi.org/10.1021/jp207571n | J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C

ARTICLE

the completion of the h-BN single layer. Using in situ highenergy resolution photoelectron spectroscopy, in combination with complementary surface science techniques, we find that at low temperature borazine molecules adsorb intact on Ir(111) with the molecule ring parallel to the surface. Molecular dissociation occurs near room temperature, with dehydrogenation accompained also by partial cracking of the benzene-like ring, a process which precedes the formation of the BN layer.

’ EXPERIMENTAL SECTION All measurements were performed at the SuperESCA beamline30,31 of Elettra, the synchrotron radiation facility in Trieste, Italy. The ultrahigh vacuum system, with a background pressure better than 2  1010 mbar, is equipped with a sputter ion gun for sample cleaning, a mass spectrometer, low-energy electron diffraction (LEED) optics, and a Phoibos hemispherical electron energy analyzer (150 mm mean radius) with an in-house developed delay line detector. The Ir(111) substrate was cleaned by repeated cycles of Ar+ sputtering (T = 300 K, p = 2  106 mbar) and annealings in O2 (p = 5  107 mbar) between 500 and 1070 K, followed by a final flash annealing to 1400 K to remove residual oxygen. The sample cleanliness and ordering were verified by monitoring the presence of eventual contaminants, such as C or O, and by checking the quality of the (1  1) LEED pattern, which exhibited sharp spots and a low background. For the preparation of the h-BN layer, we used thermal decomposition of benzene-like borazine, which was synthesized following the procedure described by Wideman et al.32 Borazine was stored below 250 K at all times to avoid degradation. Prior to deposition, borazine was regularly outgassed by freezethaw cycles to pump off the residual vapor in the frozen state. Coverages are reported in monolayers (ML), where 1 ML corresponds to the surface density of the Ir atoms in the Ir(111) surface, i.e., to 1.57  1015 atoms cm2. High-energy resolution fast X-ray photoelectron spectroscopy (XPS) Ir 4f7/2, B 1s, and N 1s core-level spectra were collected using photon energies of 130, 284, and 500 eV, respectively, with an overall energy resolution (electron energy analyzer and X-ray monochromator) ranging from 40 to 100 meV. The surface normal, the incident beam direction, and the electron emission direction are all in the same horizontal plane, with the angle between the photon beam and the electron energy analyzer fixed at 70. All the binding energies (BEs) presented in this work are referenced to the Fermi level. The core level spectra were fitted using DoniachSunjic (DS)33 functions convoluted with a Gaussian. In addition, a linear background was included in the fit. The lineshape parameters are the Lorentzian width (Γ), the Anderson singularity index (α), and the Gaussian width (G). Near-edge X-ray absorption fine structure (NEXAFS) spectra were acquired at the B and N K-edge in Auger yield mode at photon incidence angles of θph = 0 and 70 with respect to the sample surface normal. In this case the spectra were normalized to the incident beam intensity, monitored with a gold mesh intercepting the beam. Temperature-programmed desorption (TPD) spectra were acquired in the range 1701150 K, using a heating rate of ∼3 K s1. During desorption experiments, the sample was placed in front of the mass spectrometer equipped with a Feulner cup34 to enhance the surface signal. Quoted borazine exposures are given in units of Langmuir (1 L = 1  106 Torr s1).

Figure 1. (a) Selection of B 1s fast-XPS core level spectra of the B3N3H6 uptake on Ir(111) at 170 K together with the fitting components (Bmol, molecular borazine; Bm, multilayer structure). (b) Evolution of the intensity of the B 1s components as a function of the borazine exposure. The intensities corresponding to each spectrum are normalized to the saturation value.

’ RESULTS Low-Temperature Borazine Adsorption. Figure 1(a) shows time-lapsed B 1s spectra acquired during exposure of the Ir substrate to B3N3H6 at 170 K. Up to ∼1 L the sequence of spectra displays only one component (Bmol) at 189.7 eV. At higher dose, Bmol diminishes in intensity and shifts (∼0.2 eV) progressively toward lower BEs. In parallel, a new component (Bm) emerges at ∼190.3 eV. This new feature grows continuously with increasing borazine exposure indicating the formation of a condensed multilayer structure, in agreement with the assigment made for analogous systems, e.g., benzene adsorbed on Pd(111).35 The integrated peak intensities of the two populations as obtained by the fitting procedure are shown in Figure 1(b). The lineshape of the two components has been kept fixed for each spectrum of the uptake sequence. This fitting procedure gave low residual modulation, proving the existence of no more than two nonequivalent molecular species. The continuous shift of the Bm component (∼0.6 eV) toward high BEs with increasing borazine exposure most likely arises from final state effects. 36,37 In particular, if compared to the first-layer chemisorbed molecules, the core-hole screening by the metal surface becomes less efficient as the thickness of the multilayer film increases. N and B K-edge absorption spectra for low B3N3H6 exposure (1.1 L) and different θph are reported in Figure 2(a). The N K-edge spectra show four π-related sharp resonances, A1, A2, A3, and A4 at 398.2, 399.3, 400.7, and 402.0 eV, respectively. In the photon energy range 405415 eV, instead, two σ-related broad peaks B and C can be observed. By comparison with borazine electronic structure,38 A1 and A3 resonances are assigned to transitions from the N 1s orbital to π*(e00 ) and π*(a00 2) molecular orbitals. These features show a pronounced angular dependence, being clearly visible at grazing incidence (black curve, θph = 70) and almost completely absent at normal incidence (red curve, θph = 0). Therefore, these peaks are related to molecules that adsorb with the plane of the hexagonal ring parallel to the surface. A2 and A4 are assigned to π*(e00 ) and π*(a00 2) transitions too. However, these π*-resonances can be observed at both incidence angles, indicating that some molecules are tilted away from the surface. These assignments are supported by the 3 eV shift 158

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C

ARTICLE

Figure 3. B 1s spectrum obtained after B3N3H6 saturation at T = 330 K (25 L) together with the fitting components (Bad, atomic boron; Bmol, molecular borazine; B0, molecular fragments). The inset shows the uptake curve for the three components, normalized to the saturation value, as obtained by fitting the B 1s spectra. Note that the saturation is reached already at about 1.8 L.

the tilted population, in accordance with the development of the multilayer structure. The close similarity between the gas-phase borazine EELS38 and our multilayer NEXAFS spectra can be explained considering that the disorder of the multilayer is quite similar to the random molecular orientation in the gas phase. This further supports our assignment of π* transitions to different adsorption geometries. From Molecular Adsorption to Hexagonal BN. Figure 3 shows the B 1s spectrum after 25 L of borazine exposure at T = 330 K. The spectrum can be decomposed into three contributions: B0, Bmol, and Bad, at BEs of 190.5, 189.6, and 188.6 eV, respectively. The inset in Figure 3 displays the intensity evolution of these components as a function of borazine exposure. The lineshapes of the three peaks were fixed during the fitting analysis of the whole sequence. At the beginning of the uptake, only Bmol is observed, while the other two components rise at slightly higher exposures (above 0.3 L). It is interesting to note that B0 and Bad start their growth simultaneously, although with different rates, thus suggesting a common origin. At saturation, above 2 L, these species are 52% (B0) and 32% (Bad) of the concentration of the Bmol peak. A possible attribution of B0 and Bad to nonequivalent B atoms within the borazine molecule appears unlikely, because the ratio B0/Bad is not constant during the uptake. On the other hand, the high intensity of these components and the expected low density of surface defects rule out also the possibility of preferential adsorption at defective sites. Moreover, induced photodissociation has to be excluded since the photoemission spectra do not undergo modification with time when the surface is exposed to the photon beam. Hence, the existence of multiple peaks at 330 K in the B 1s spectra suggests that at this temperature a fraction of borazine molecules decomposes. The BE of the Bmol peak is close to the value reported in the previous section for borazine adsorption at low temperature. This contribution is therefore assigned to borazine molecules which adsorb intact on Ir(111). Conversely, molecular dissociation leads to the appearance of B0 and Bad. It is interesting to note that (i) after deposition at room temperature the intensity of the Bad component is approximately 17% of the total B 1s signal (Figure 3) and that (ii) by heating the system to obtain a h-BN layer (as will be explained later on) the photoemission intensity decreases by

Figure 2. N and B K-edge NEXAFS spectra for the Ir(111) exposed to (a) 1.1 L, thinner curves, and (b) 3.3 L, thicker curves, B3N3H6 at 170 K.

between π*(e00 ) and π*(a00 2) resonances, which is quite close to the value reported for gas-phase B3N3H6 (3.3 eV).38 The interpretation of the B K-edge spectra is in line with this analysis. Also for the B resonances it is possible to single out a polarization-dependent (A1 and A3) and a polarization-independent (A2) contribution of π* symmetry. However, the B K-edge resonances are quite broad, and the exact peak positions are not well-defined, except for the A2 and the A3 components located at 190.5 and 191.5 eV, respectively. At exposures of 1.1 L, the low coverage adsorption state reaches saturation, while the multilayer state only starts to be populated (Figure 1). Hence, we attribute the more intense A1 and A3 resonances of the flat-lying borazine to the Bmol component and associate the tilted geometry to Bm. NEXAFS spectra of the borazine-saturated (3.3 L) Ir surface are reported in Figure 2(b). A quite similar behavior is found both for N and B K-edges. As compared to the low B3N3H6 dose, in this case the A2 resonance develops into a strong component detected at both incidence angles. In addition, A1 appears only at grazing incidence. These findings suggest a preferential growth of 159

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C

ARTICLE

Figure 4. Background subtracted thermal desorption spectra obtained during heating (3 K s1) of a B3N3H6 saturated (11 L at 170 K) surface. H2 desorbs between 250 and 900 K, with a maximum desorption rate at 330 K.

almost the same amount, the drop being of about 15%. Therefore, this reduction can be reasonably explained in terms of a reduced amount of boron atoms on the surface. In addition, the Bad peak location is close to the value of 188.4 eV reported for diluted boron in Rh(111).39 From the above considerations, it is therefore realistic to explain the Bad component as due to single boron atoms and B0 as due to molecular fragments. To confirm that B3N3H6 dissociation occurs on the sample surface, we performed a series of TPD experiments on the borazine-saturated surface at low temperature. Figure 4 shows the resulting H2 thermal desorption spectra (m/e = 2) from the Ir surface exposed to 11 L of B3N3H6 at 170 K. This spectrum shows a broad desorption feature resulting from dissociation of the adsorbed B3N3H6 layer. The dehydrogenation starts already at 250 K and goes on over a broad temperature range up to 900 K, with a maximum desorption rate at about 330 K. These findings are similar to those obtained on B3N3H6/Pt(111) by Simonson et al.,23 reporting a broad H2 desorption peak with a maximum at 285 K. In addition, the temperature of 330 K falls within the temperature range of H2 desorption from the Ir(111) surface exposed to hydrogen, which lies between 290 and 370 K depending on the initial coverage.40 Therefore, our desorption data are compatible with a borazine dehydrogenation process followed by a second-order recombinative H2 desorption. Figure 5 shows the sequence of B 1s and Ir 4f7/2 spectra measured at room temperature after annealing the borazine-saturated surface to increasing temperatures, ranging from 470 to 1370 K. B0 grows with temperature and moves toward lower BEs, while Bmol and Bad intensity gets reduced. Upon heating to 620 K, a new component (B1, from spectrum (3)) at ∼190.8 eV has to be included in the analysis to achieve a low residual modulation. The intensity of this component increases progressively up to the annealing temperature of 1120 K (spectrum (5)). Between 1120 and 1370 K, a complete suppression of Bmol and Bad is observed, while the remaining peaks (B0 and B1) shift toward lower BE by ∼0.3 eV. At this point, a h-BN layer is formed, as confirmed by B0 and B1 peak positions and lineshapes which are quite close to those corresponding to the best h-BN layer obtained in the present work (see next section). The analysis of the integral area of the B 1s core level region indicates a drop of about 15% going from spectrum (1) to (6), as mentioned above. According to this result, for the same sequence of spectra the signal from the N 1s region diminishes by about 20% (not shown here). This reduction can be attributed to a decrease of the B and N species on the surface, although photoelectron diffraction effects cannot be

Figure 5. High-energy resolution B 1s and Ir 4f7/2 core level spectra for the clean surface (bottom) and B3N3H6 saturated (25 L, T = 330 K) Ir surface (1) and after fast annealing to the listed temperatures (26). The different spectral contributions as extracted from the peak fit analysis are represented by the colored peaks. All curves are plotted after linear background removal.

excluded. In the former case, the diffusion of atomic boron into the bulk and/or desorption of NHx or N2 species41 seem to be the most plausible processes. Ir core level spectra for the same heating sequence are shown on the right side of Figure 5. As previously reported,42 the Ir 4f7/2 spectrum of the clean surface exhibits two well-resolved components: a low BE peak centered at 60.28 eV, assigned to photoemission from Ir atoms of the topmost layer (Irs), and a high BE feature at 60.83 eV due to atoms in the deeper layers, i.e., bulk atoms (Irb). The two-component analysis gives best-fit values of 160 (360) meV for the Lorentzian width, 0.16 (0.21) for the asymmetry parameter, and 230 (120) meV for the Gaussian width of the bulk (surface) component. The lineshape parameters and the measured surface core level shift (SCLS), i.e., the BE difference between the surface and the bulk peak, of 540 meV are in good agreement with the previous experimental42 and theoretical findings.43 Upon borazine deposition, a new broad feature (Iri) grows between the surface and the bulk components at 60.61 eV, paralleled by a drop of the Irs peak intensity and a slight reduction of its SCLS to 505 meV (spectrum (1) in Figure 5). For the adsorbate-induced component Iri, the same lineshape parameters of the clean surface component were used. Upon increasing the annealing temperature, Iri continuously decreases, while Irs almost recovers the intensity of the clean surface 160

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C

ARTICLE

Figure 7. B3N3H6 deposition on Ir(111) at 1270 K. (a) Selection of high-energy resolution B 1s spectra measured during the uptake. (b) Evolution of the integrated intensities for the two spectral contributions resulting from fitting of the B 1s spectra. The upper part shows the B1/B0 ratio, the initial fluctuations are due to the poor signal-to-noise ratio at low coverage.

saturated with borazine at room temperature. In fact, the Ir spectrum in Figure 6 displays the same three components: Irb, Irs, and Iri. However, the intensity of the adsorbate-induced component Iri is higher in this case, as can be judged by comparing the analogous peak of spectrum (6) in Figure 5. The N 1s and B 1s photoemission spectra display a quite similar double peak structure: both signals are dominated by a well-resolved main component located at 397.52 eV (N0) and 189.84 eV (B0). In addition, a second weak contribution emerges at 398.62 eV (N1) and 190.61 eV (B1). To shed light on the origin of the spectral features, we monitored in situ the evolution of N 1s and B 1s core level regions during borazine uptake at 1170 K. Figure 7 shows the B 1s XPS spectra in panel (a), while panel (b) displays the intensity behavior of the two components. The spectral contributions grow together with a constant B1/B0 ratio of about 0.10. Similar results have been obtained for the N components (not shown here). The fact that the ratio between the two components is constant during the growth process suggests that B1 and N1 are not related to layer defects. On the basis of these findings and according to ref 44, the spectral contributions are attributed to h-BN regions differently interacting with the Ir substrate: N0 and B0 are associated to the fraction of the BN film weakly bounded to the substrate atoms, represented by Irs, while N1 and B1 are attributed to h-BN regions strongly interacting with the Ir surface, represented by Iri. A quantitative interpretation of the overlayer morphology cannot be directly extracted from XPS spectra presented in Figure 6 and Figure 7 since the intensity of the photoemission peaks can be strongly affected by diffraction effects. To properly evaluate the intensity ratios N1/N0 and B1/B0, we thus followed the procedure proposed in ref 44, measuring the photoemission spectra at different photon energies, corresponding to a kinetic energy range of 40200 eV. The average value of N1/N0 and B1/B0 obtained in this way attests at 0.17, which well compares to the relative weight of the iridium adsorbate-induced feature with respect to the total surface signal, Iri/(Iri + Irs), i.e., 0.16. Similar trends were found also for the corrugated h-BN layer grown on top of Rh(111) and Ru(0001) surfaces.44 This result is in accordance with the overall assignment of the different components: N0,

Figure 6. (a) LEED pattern of the h-BN/Ir(111) taken at E = 77 eV. Ir spots are surrounded by the h-BN spots; the satellite spots reflect the periodicity of the moire. (b) Line profile of the diffraction pattern along the (01) direction. To measure this profile, the sample has been moved a few degrees off normal. (c) Ir 4f7/2, B 1s, and N 1s core level spectra together with the spectral contributions resulting from the peak-fit analysis of h-BN/Ir(111) resulting from dosing B3N3H6 at T = 1170 K. N 1s fitting parameters (Γ, α, G): N0 and N1 (100 meV, 0.08, 800 meV). B 1s fitting parameters (Γ, α, G): B0 and B1 (100 meV, 0.11, 760 meV).

and Irs moves toward lower BE eventually reaching a SCLS of about 535 meV, which is very close to the value of the clean Ir(111). At the same time, as expected, the BE position of the bulk peak remains fixed. Single Hexagonal BN Layer. The growth of a single h-BN layer was achieved by borazine on the Ir(111) surface at a sample temperature of 1170 K. Exposures above 20 L were needed to grow an extended and long-range ordered layer. Panel (a) in Figure 6 shows the LEED pattern from h-BN/Ir(111): first-order substrate spots are surrounded by a six-fold arrangement of diffraction spots, indicating the formation of a coincidence lattice between the h-BN layer and the Ir(111) substrate (for a schematic drawing of the moire superlattice see for instance Figure 3A in ref 4). The line profile analysis (panel (b)) reveals a superstructure with a periodicity of (13  13) BN units (a = 0.250 nm) on top of (12  12) Ir unit cells (a = 0.272 nm). The high-resolution N 1s, B 1s, and Ir 4f7/2 core level spectra of h-BN/Ir(111) are shown in panel (c). The Ir spectrum relative to the h-BN layer obtained by dosing borazine at high temperature is similar to the one measured for the annealed surface previously 161

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C

ARTICLE

indicates that: (i) the 1.1 L exposed Ir(111) is completely saturated with an extended close-packed parallel molecular configuration and (ii) the saturation coverage is determined by borazine lateral interactions. We propose therefore that the most favorable configuration for the low-temperature adsorption is with the molecular plane, defined by the hexagonal ring, parallel oriented to the Ir substrate (Bmol component in XPS spectra) and bound to it via π bonds. The subsequent surface crowding leads to the occupancy of all these sites and, above 0.20 ML, inhibits additional planar adsorption, forcing the molecules to anchor in tilted configurations and, eventually, to form a condensed structure on top of the first chemisorbed layer (Bm). Under similar growth conditions, the borazine molecule was proposed to lie flat on Ru(0001),24 Re(0001),25 and Au(111)23 substrates. On the contrary, in previous studies based on vibrational spectroscopy a vertical bonding geometry on Pt(111) was also reported.2224 The XPS results clearly indicate that B3N3H6 molecules undergo dissociation on Ir(111) at T = 330 K. However, the peaks corresponding to atomic boron and molecular fragments can be detected only for exposures higher than 0.3 L, when the slope of the Bmol intensity curve decreases (inset in Figure 3). These findings indicate that borazine decomposition does not occur in the initial stage of the adsorption; i.e., the first adsorbed molecules keep their benzene-like ring intact, giving rise to the Bmol component. Borazine dissociation takes place only above 0.3 L, leaving dehydrogenated molecules and molecular fragments (B0) and boron adatoms (Bad) on the surface. A temperature as high as 930 K was suggested for the BN bond breaking process taking place on Rh(111).15 Here, instead, the presence of atomic boron indicates that the borazine ring cracks already at 330 K. This value compares to the detection of atomic B and N species by Auger electron spectroscopy of the Re(0001) surface dosed with borazine and annealed to 570 K.25 Increasing the substrate temperature to 470 K results in further molecular dissociation, as depicted by the intensity drop of Bmol and the increase of B0 components (Figure 5). This process results also in a decrease of Bad, indicating a depletion of boron adatoms. Given this information, and the fact that the overall B 1s signal is reduced by 15% at the end of the annealing cycle, we suggest that the boron atoms undergo bulk diffusion. A similar behavior has been previously observed for boron adsorbed on Pd(111)47 and Ni(100).48 Above T = 620 K, this trend in the intensity evolution of B0, Bmol, and Bad continues, and in addition, a new feature (B1) grows. Our findings can be explained by comparing the h-BN growth process on Rh(111).15 By using scanning tunneling microscopy, Dong et al. evidenced that a first reorganization into nanometer-scale clusters takes place at T = 690 K, while the h-BN structure emerges already below T = 900 K. According to this picture, we intepret the observed changes in our XPS spectra as the earliest rearrangement of dehydrogenated molecules and fragments to form islands with h-BN morphology. These islands are already characterized by regions differently interacting with the Ir surface, i.e., boron and nitrogen atoms strongly (B1 and N1) and weakly bonded (B0 and N0) to the substrate. As the temperature is further increased, B0 and B1 gradually shift toward lower BEs, finally reaching a position which resembles the value obtained for our best h-BN layer, while the Bmol and Bad components disappear when an extended layer is formed. Hence, the evolution, in terms of BE position and intensity of B0 component together with the appearance of B1, reflects the coalescence of single dehydrogenated BN rings and molecular fragments which ends up with the completion of the h-BN layer. Accordingly, this process is reflected

Figure 8. N K-edge NEXAFS spectra of the h-BN single layer on Ir(111).

B0, and Irs stem from the atoms in the low-interacting regions while N1, B1, and Iri to the atoms residing in the regions where the h-BN layer strongly interacts with the Ir substrate. The geometry of the BN lattice on the Ir surface was examined by recording the N K-edge spectra at two photon incidence angles. As can be seen in Figure 8, the marked angular dependence allows us to distinguish between π* resonances (A0 , A, and A00 ) and σ* features (B and C). The complete absence of π* components at normal incidence indicates that the h-BN lattice is lying flat on the substrate. According to the work by Preobrajenski et al. for h-BN growth on selected metal surfaces,44 A0 and A00 resonances are attributed to h-BN regions significantly interacting with the Ir atoms. Both these resonances occur due to the orbital mixing between the Ir d orbitals and the h-BN π states. In particular, A0 can be assigned to excitations to adsorptioninduced gap states of h-BN,45 while A00 relates to excitations to interlayer conduction-band states.9 On the other hand, the π resonance A is detected also in the bulk h-BN NEXAFS spectra9,44 and is thus related to the h-BN site weakly bound to the metal substrate.

’ DISCUSSION Two different geometries were identified for low temperature borazine adsorption on clean Ir(111): a flat geometry, assigned to the Bmol component, and a tilted configuration, attributed to the Bm peak (Figure 1). Because of the fundamental role played by the availability of space in the molecular adsorption mechanism, an estimate of the maximum saturation coverage is in order. From the LEED analysis the B and N coverages for the best h-BN layer resulted to be 1.17 ML. The intensity of the B 1s core level measured on this surface can be used to evaluate the boron coverage at low temperature. This method results in a B coverage of 0.6 ML at the saturation limit of the Bmol component, reached at 1.1 L (Figure 1), i.e., to a molecular coverage of about 0.2 ML. Considering the borazine molecule as a regular hexagon, we calculated a van der Waals area of 32.74 Å2 46 corresponding to a surface density of 3.05  1014 molecules cm2. Accordingly, the theoretical borazine saturation coverage is 0.19 ML. This value must be considered only as a crude approximation of the maximum packing since it does not take into account photoelectron diffraction effects, that could modulate the photoemission intensity, or the adsorbatesubstrate registry. Nevertheless, the good agreement between theoretical and experimental values 162

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C in the hydrogen desorption curve in Figure 5. After initial surface decomposition of borazine, as indicated by the main desorption feature around room temperature, dehydrogenation extends over a wide temperature range. This continuous desorption is due to the dehydrogenation of the intact-ring molecules and of the molecular fragments adsorbed on the surface. The variations of the adsorbate chemical composition and structure lead to measurable changes also in the Ir 4f7/2 spectra. The SCLS of the clean Ir(111) surface is primarily due to the reduced coordination of first-layer Ir atoms with respect to the bulk, which determines a narrowing of the surface d-band that results in a shift of the surface component Irs toward lower BEs. Upon borazine adsorption at T = 330 K the Ir 4f7/2 spectrum shows a decrease of Irs together with the appearance of Iri, which stems from the Ir atoms interacting with the chemical species adsorbed on the substrate. It is well-known that atomic and molecular adsorption on TM surfaces typically induces a broadening of the substrate d-band, together with a lowering of the density of states in the Fermi level region. For TMs with a more than half-filled d-band this results in a shift toward higher BEs of the d-band of the surface atoms interacting with the adsorbate with respect to that of the clean atoms, which is reflected in the measured adsorbate-induced SCLSs.49 This effect applies also to the case of Iri. Upon increasing the annealing temperature, Iri decreases, while the clean surface peak Irs gains intensity. This behavior clearly reflects the reorganization of the adsorbate layer from a disordered structure, composed of several chemical species strongly interacting with the Ir surface, into an ordered h-BN layer, which on the average is weakly interacting with the substrate. Similar findings have been obtained for the related hexagonal lattice of graphene grown on Ir(111).43,50 In that case, upon ethylene adsorption at room temperature and subsequent annealings,43 a third feature between surface and bulk components due to Ir first-layer atoms differently interacting with the adsorbed species was found. Moreover, as in our case, increasing the substrate temperature resulted in a continuous reduction of this peak paralleled by the growth of the clean surface component. This behavior has been explained as a result of the evolution of the different species on the surface from ethylene, strongly interacting with the substrate, to the formation of a graphene layer, which is very weakly interacting with Ir(111).43 These considerations can be applied also to the h-BN growth process on Ir(111). High-temperature deposition of borazine leads to the formation of a more extended h-BN single layer compared to that obtained by heating the substrate saturated at room temperature. Indeed, the XPS integral area shows that the h-BN film obtained by the latter procedure covers an area which is about 58% that of our best layer, achieved with the former growth technique. This is further supported by comparison of the corresponding Ir 4f7/2 spectra: for the room temperature deposition, the lower intensity of Iri with respect to Irs is due to a higher fraction of uncovered surface. At room temperature, in fact, part of the hydrogen atoms are attached to borazine molecules, which therefore occupy a greater area than the hexagon in the h-BN configuration.15 The presence of more than one surface component in the Ir 4f7/2 core level, as well as that of two components in the B 1s and N 1s spectra, reflects different strength of interaction of the h-BN layer with the substrate. In fact, while Iri is related to substrate regions strongly interacting with the h-BN lattice, the main surface component Irs has a SCLS of 540 meV, which is practically identical to that of the clean surface, denoting the small interaction between these surface atoms and the h-BN thin film. Interestingly, the

ARTICLE

interaction between Ir(111) and the graphene layer is quite different, and no adsorbate-induced components are detected in the Ir spectra.43 The strength of the chemical bonding gives rise to substantial changes also in the morphology of the overlayer.44 Indeed, according to ref 44, N1 and B1 spectral features are related to the layer atoms that are close to the substrate, while the main N0 and B0 contributions correspond to the elevated part of the h-BN overlayer. Since the chemical bonding between h-BN and Ir(111) is known to be halfway between that of Pt(111), weakly interacting, and Rh(111),44 strongly interacting, we therefore expect a rather poor degree of corrugation of the h-BN film.

’ CONCLUSIONS We have presented a study of the interaction of borazine with the Ir(111) surface, from low-temperature adsorption to dissociation and formation of a h-BN single layer. Experimental data for borazine adsorbed on Ir(111) at T = 170 K indicate that borazine adsorbs molecularly with the plane of the ring parallel to the substrate; the molecular saturation coverage is estimated to be 0.20 ML. Subsequent annealing of the borazine-covered surface results in a continuous dehydrogenation from ∼250 up to ∼1000 K, with a maximum H2 desorption rate at 330 K, and eventually ends up with the formation of the h-BN layer. Above T = 330 K the molecule ring cracks, leading to the formation of molecular fragments and atomic species, as revealed by the detection of atomic boron. Deposition of borazine on the Ir surface kept at high temperature leads to the formation of an extended h-BN layer. LEED images of the single h-BN film show a moire pattern consistent with the formation of a (13  13)/(12  12) BN/ Ir coincidence structure. XPS and NEXAFS measurements indicate the presence of h-BN regions differently interacting with the Ir surface. ’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT We thank J. Kaspar, S. Fornarini, and R. Zanoni for precious support in the production of borazine and for stimulating discussions. A.B. acknowledges financial support from the University of Trieste under the program FRA2009. ’ REFERENCES (1) Dil, H.; Lobo-Checa, J.; Laskowski, R.; Blaha, P.; Berner, S.; Osterwalder, J.; Greber, T. Science 2008, 319, 1824. (2) 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. 2007, 119, 5207. (3) Barth, J. V.; Costantini, G.; Kern, K. Nature 2005, 437, 671. (4) Corso, M.; Auw€arter, W.; Muntwiler, M.; Tamai, A.; Greber, T.; Osterwalder, J. Science 2004, 303, 217. (5) Watanabe, K.; Taniguchi, T.; Kanda, H. Nat. Mater. 2004, 3, 404. (6) Geim, A. K.; Novoselov, K. S. Nat. Mater. 2007, 6, 183. (7) Dean, C. R.; Young, A. F.; Meric, I.; Lee, C.; Wang, L.; Sorgenfrei, S.; Watanabe, K.; Taniguchi, T.; Kim, P.; Shepard, K. L.; Hone, J. Nat. Nanotechnol. 2010, 5, 722. (8) Nagashima, A.; Tejima, N.; Gamou, Y.; Kawai, T.; Oshima, C. Phys. Rev. Lett. 1995, 75, 3918.  (9) Preobrajenski, A. B.; Vinogradov, A. S.; Martensson, N. Phys. Rev. B 2004, 70, 165404. (10) Auw€arter, W.; Kreutz, T.; Greber, T.; Osterwalder, J. Surf. Sci. 1999, 429, 229. 163

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164

The Journal of Physical Chemistry C

ARTICLE

(47) Krawczyk, M. Appl. Surf. Sci. 1998, 135, 209. (48) Desrosiers, R. M.; Greve, D. W.; Gellman, A. J. J. Vac. Sci. Technol. A 1997, 15, 2181. (49) Baraldi, A. J. Phys.: Condens. Matter 2008, 20, 093001. (50) Lizzit, S.; Baraldi, A. Catal. Today 2010, 154, 68. (51) Bondi, A. J. Phys. Chem. 1964, 68, 441.

(11) Grad, G. B.; Blaha, P.; Schwarz, K.; Auw€arter, W.; Greber, T. Phys. Rev. B 2003, 68, 085404. (12) Cavar, E.; Westerstr€om, R.; Mikkelsen, A.; Lundgren, E.;  Vinogradov, A.; Ng, M. L.; Preobrajenski, A.; Zakharov, A.; Martensson, N. Surf. Sci. 2008, 602, 1722. (13) Morscher, M.; Corso, M.; Greber, T.; Osterwalder, J. Surf. Sci. 2006, 600, 3280. (14) Goriachko, A.; He; Knapp, M.; Over, H.; Corso, M.; Brugger, T.; Berner, S.; Osterwalder, J.; Greber, T. Langmuir 2007, 23, 2928. (15) Dong, G.; Fourre, E. B.; Tabak, F. C.; Frenken, J. W. M. Phys. Rev. Lett. 2010, 104, 096102. (16) Greber, T.; Brandenberger, L.; Corso, M.; Tamai, A.; Osterwalder, J. e-J. Surf. Sci. Nanotechnol. 2006, 4, 410. (17) M€uller, F.; H€ufner, S.; Sachdev, H. Surf. Sci. 2008, 602, 3467. (18) Corso, M.; Greber, T.; Osterwalder, J. Surf. Sci. 2005, 577, L78. (19) Allan, M.; Berner, S.; Corso, M.; Greber, T.; Osterwalder, J. Nanoscale Res. Lett. 2007, 2, 94. (20) Auw€arter, W.; Suter, H. U.; Sachdev, H.; Greber, T. Chem. Mater. 2004, 16, 343. (21) Auw€arter, W.; Muntwiler, M.; Osterwalder, J.; Greber, T. Surf. Sci. 2003, 545, L735. (22) Simonson, R.; Trenary, M. J. Electron Spectrosc. Relat. Phenom. 1990, 5455, 717. (23) Simonson, R.; Paffett, M.; Jones, M.; Koel, B. Surf. Sci. 1991, 254, 29. (24) Paffett, M.; Simonson, R.; Papin, P.; Paine, R. Surf. Sci. 1990, 232, 286. (25) He, J.-W.; Goodman, D. Surf. Sci. 1990, 232, 138. (26) Koel, B. E.; Crowell, J. E.; Mate, C. M.; Somorjai, G. A. J. Phys. Chem. 1984, 88, 1988. (27) Lehwald, S.; Ibach, H.; Demuth, J. E. Surf. Sci. 1978, 78, 577. (28) Jakob, P.; Menzel, D. Surf. Sci. 1989, 220, 70. (29) Grassian, V. H.; Muetterties, E. L. J. Phys. Chem. 1987, 91, 389. (30) Baraldi, A.; Comelli, G.; Lizzit, S.; Kiskinova, M.; Paolucci, G. Surf. Sci. Rep. 2003, 49, 169. (31) Baraldi, A.; Barnaba, M.; Brena, B.; Cocco, D.; Comelli, G.; Lizzit, S.; Paolucci, G.; Rosei, R. J. Electron Spectrosc. Relat. Phenom. 1995, 76, 145. (32) Wideman, T.; Sneddon, L. G. Inorg. Chem. 1995, 34, 1002. (33) Doniach, S.; Sunjic, M. J. Phys. C 1970, 3, 285. (34) Feulner, P.; Menzel, D. J. Vac. Sci. Technol. 1980, 17, 662. (35) Lee, A. F.; Lambert, R. M.; Goldoni, A.; Baraldi, A.; Paolucci, G. J. Phys. Chem. B 2000, 104, 11729. (36) Chiang, T.-C.; Kaindl, G.; Mandel, T. Phys. Rev. B 1986, 33, 695.  (37) Tillborg, H.; Nilsson, A.; Hernn€as, B.; Martensson, N.; Palmer, R. E. Surf. Sci. 1993, 295, 1. (38) Doering, J. P.; Gedanken, A.; Hitchock, A. P.; Fischer, P.; Moore, J.; Olthoff, J. K.; Tossell, J.; Raghavachari, K.; Robin, M. B. J. Am. Chem. Soc. 1986, 108, 3602. (39) M€uller, F.; H€ufner, S.; Sachdev, H.; Gsell, S.; Schreck, M. Phys. Rev. B 2010, 82, 075405. (40) Engstrom, J. R.; Tsai, W.; Weinberg, W. H. J. Chem. Phys. 1987, 87, 3104. (41) Weststrate, C.; Bakker, J.; Gluhoi, A.; Ludwig, W.; Nieuwenhuys, B. Catal. Today 2010, 154, 46. (42) Bianchi, M.; Cassese, D.; Cavallin, A.; Comin, R.; Orlando, F.; Postregna, L.; Golfetto, E.; Lizzit, S.; Baraldi, A. New J. Phys. 2009, 11, 063002. (43) Lacovig, P.; Pozzo, M.; Alfe, D.; Vilmercati, P.; Baraldi, A.; Lizzit, S. Phys. Rev. Lett. 2009, 103, 166101. (44) Preobrajenski, A.; Nesterov, M.; Ng, M. L.; Vinogradov, A.;  Martensson, N. Chem. Phys. Lett. 2007, 446, 119. (45) Preobrajenski, A. B.; Krasnikov, S. A.; Vinogradov, A. S.; Ng,  M. L.; K€a€ambre, T.; Cafolla, A. A.; Martensson, N. Phys. Rev. B 2008, 77, 085421. (46) The following internuclear bond distances were used: BN, 1.44 Å; BH, 1.20 Å; NH, 1.02 Å. The H (aromatic) van der Waals radii is 1.00 Å.51 164

dx.doi.org/10.1021/jp207571n |J. Phys. Chem. C 2012, 116, 157–164