Dynamics of epitaxial graphene growth and adsorptions of cobalt 5

5 Adsorption of Co on graphite, Crich 63 and graphene surfaces at room temperature The interaction, nucleation and growth of submonolayer cobalt (Co) on graphite, C-rich (63x63)R30o (hereafter 63 for short) and epitaxial graphene/6HSiC(0001) at room temperature are studied here. Co is found physisorbed on all three surfaces. Co/63 displayed more complex interaction where elemental Si (Si tetramer) is released from the 63 surface upon Co adsorption. On all the three surfaces, Co adopts 3-dimensional (3D) or V-W growth mode and they form dome-shaped clusters. Co clusters of similar dimension and density are found on the Co/HOPG and Co/graphene. Smaller Co but higher density are found forming on 63 due to impedance of Co diffusion by the corrugation of 63 surface. Despite similarities of atomic structure between graphite and graphene, scaling analysis of the cluster volume distributions reveals very different critical Co nucleus size, i* on these two surfaces. i* equals to zero is found for Co/HOPG while i* equals to atoms is found for Co/graphene. The difference is attributed to the influence of underlying substrate where charge transfer to the graphene from its interface causes the i* of Co to increase. For Co/63, i* of atom is deduced with the nucleation on top of Si-rich 6x6 maxima as the preferred adsorption site for Co. Adsorption of Co at room temperature 5.1 Adsorption of Co on graphite (HOPG) 5.1.1 Interaction of Co with HOPG As presented in Section 4.1, only HOPG prepared in-situ will be used for adsorption studies of Co. Binding energy of elemental sp2-type C 1s (HOPG) and Co 2p3/2 (sputtered-clean Co foil) are measured to serve as reference. They are located at 284.4 and 778.1 eV respectively. The binding energy of C 1s is in agreement with theoretical estimation by Davis and Shirley [1]. Figure 5.1 shows the coverage-dependant spectra of C 1s and Co 2p3/2 for Co grown on graphite at room temperature. These spectra have been arbitrary normalized for ease to distinguish any energy shift or features variation. As the Co increases from submonolayer (0.3 monolayer (ML)) to high coverage (11 ML), C 1s spectra (Fig. 5.1a) are unaffected i.e. remained at 284.4 eV while Co 2p3/2 spectra (Fig. 5.1b) are shown slightly shifted by +0.2 eV (778.3 eV) for coverage of 0.3 and 0.6 ML. As the Co 2p3/2 remains asymmetric, a common property for metallic element, Co is believed to be remained unreacted on the surface. More significantly, we not observe any changes to the C 1s throughout the growth. This shift is believed to be caused by sizeeffect of small-cluster [2]*. In this respect, the core-level spectra are also cross-checked with XPS valence band (VB) spectra since they are more sensitive towards electronic states of graphite and Co. * Screening of positive hole created by photoelectron emission reduces with lower coordination numbers and hence the core levels shift to higher binding energies. 151 Chapter (a) (b) (b) Fig. 5.1 (a) C 1s spectra from low to high coverage of Co grown on graphite shows neither new features nor change in binding energy; (b) Co 2p3/2 spectra shows +0.2eV shift to higher binding energy due to incomplete screening from small cluster. All spectra in (a) and (b) have been arbitrary normalised. The VB spectra of clean graphite and polycrystalline cobalt are first acquired to serve as reference for pure electronic states for these elements. Figure 5.2a depicts the VB spectra of clean graphite, clean polycrystalline Co and a series of different coverage of Co grown on graphite. The VB of graphite consists of the 2p and 2s atomic states. The VB acquired using XPS is resolved into three most prominent states of graphite, similar to those acquired by others using Al K [3] and those using He source (hυ= 21.22 eV) [4]. These three prominent peaks arise from the three valence electrons of C that form coplanar bonding with neighbouring C atom. They are located at (i) between and 11 eV (peak I) which are very broad and weak, derived from px,y-σ band [5,6], (ii) between 11 and 14 eV (peak II) which are less intense and narrower, arising from hybridisation of C 2s and 2p orbitals; and (iii) between 14 and 22 eV (labelled as peak III) that are generally intense and fairly broad arising from two nearly degenerate s-like σ bands [3]. The forth valence electron, C 2pz which does not take part in the co-planar bonding and responsible for the conductivity of graphite form the p band and can be located at ~EF + eV. Upon deposition of Co, weak emission near EF can be observed at submonolayer coverage and 152 Adsorption of Co at room temperature this emission gets more pronounce as the coverage increases, as depicted in Fig. 5.2a. This near EF feature is emission from Co 4s and 3d photoelectrons. At higher coverage i.e. ML, the EF is filled with much stronger Co 4s and 3d emission and the substrate signals are almost suppressed with peak I observed as residue. This spectrum is now almost identical with the spectrum of clean polycrystalline Co. (b) (a) III II I Fig. 5.2 (a) XPS valence band of Co grown on graphite. Substrate signals are seen attenuated and Co signals increase tremendously with coverage, (b) difference counts spectra for Co 4s and 3d band with emission from graphite subtracted. The sharp edge near Fermi edge at coverage as low as 0.18 ML suggested that no electron exchange between graphite and Co overlayers. Stacking of these spectra (inset) show that they overlap well with bulk Co spectrum. To elucidate these Co filled states from graphite, emission from clean graphite is subtracted from each of the spectrum in Fig. 5.2a and plotted as difference curves in Fig. 5.2b. These difference curves show no differences between Co in the submonolayer regime and bulk polycrystalline Co. For better comparison, these spectra are normalized to same intensity and stacked as shown in the inset of Fig. 5.2b where they clearly overlap with the bulk Co valence band and indicates there is no bonding between Co and graphite, 153 Chapter implies that the energy shift observed from the core level are indeed due to effects from cluster size. Hybridisation is perhaps not possible given the large energy separation between graphite  bands (EF + eV) and Co 4s and 3d bands. As the Co adatoms not chemically react with graphite, these Co adatoms are physisorbed on graphite surface via weak physical attractive forces such as van der Waals. This result shows that graphite is an inert substrate. The absence of chemical interaction between metal and graphite at ambient temperature was also reported for Ti [7], Ni [8], Al [9,10] and Cu [11]. 5.1.2 Growth mode of Co on HOPG as probed by XPS Integration of the core-level peaks of C 1s and Co 2p3/2 in Fig. 5.1 gives their coverage-dependant intensity where their variations with film thickness tell us about the growth mode of the Co on HOPG. The background to this concept has been previously presented in Chapter 3. Figure 5.3 shows the intensity, I (percentage, %) of C 1s and Co 2p3/2 as a function of deposition time, t and coverage, θ. I (%) is also known as atomic percentage. In this case, %Co is 100*X(Co 2p3/2)/(X(Co 2p3/2)+ Y(C 1s)) where X and Y are intensity of Co 2p3/2 and C 1s respectively after normalised with their respective sensitivity factor (see Eqs. (3.4) – (3.7) for more details). The I-t plot in Fig. 5.3 increases linearly and gradually with Co coverage where upon 15 ML, the %Co is just about 10%. The slow and linear increment of the I-t plot without any change of slope strongly implies Co/HOPG system adopted 3-dimensional (3D) or Volmer-Weber (V-W) growth mode. Slow variation of Co percentages is caused by weak electrons emission of adsorbates due to self- attenuation within their 3D structure. The gradual increment of adsorbate signal is 154 Adsorption of Co at room temperature often complimented by slow decrease of substrate signal due to relatively large uncovered surface of substrate in 3D system [12]. The result implies that graphite is unable to support layer-by-layer growth of Co primarily due to weak interaction of Co clusters on graphite (physisorbed) and also lower surface energy of graphite. Fig. 5.3 Intensity (percentage) of Co 2p3/2 and C 1s as a function of deposition time and coverage. Slow increment of adsorbate signals indicates 3-dimensional growth mode. Solid lines are linear fit to the experimental data. 5.1.3 Nucleation and growth of Co on HOPG Figure 5.4a shows a series of STM images depicting morphologies of Co sequentially grown on HOPG at room temperature using a Co flux of 2.4 x 10-3 MLs-1. ML (0.2 nm thickness) is defined according to density of atoms in Co hcp (0001) plane i.e. 3.03 x 1013 atoms cm-2. As seen, the Co clusters appear round from top. Cross-section of one the Co clusters at 0.14 ML (inset) shows this Co cluster is dome-shaped with height, h of 0.9 nm and basal width, w of 6.4 nm. The distributions of h and w for a few selected 155 Chapter coverages are shown in Fig. 5.4b. All the width has been corrected for tip shape effect as described in Section 3.4.3. In agreement with the prediction by XPS (Fig. 5.3), Fig. 5.4b clearly shows that Co on graphite adopts 3D or V-W growth mode where cluster with multilayer thickness up to nm already formed at 0.14 ML. One of the main reasons for Co/HOPG to adopt 3D growth is Co are physisorbed on graphite and it has a much higher surface energy than graphite. Forming 3D clusters will keep the total surface energy low by maximise the substrate (graphite) surface energy. In addition, interfacial energy created is expected to be very small for this physisorbed system. The average height (), basal width () and volume, () for a few selected coverages are given in Fig. 5.4c. The is seen to increase from 25.1 ± 2.3 nm3 (at 0.14 ML) to 86.7 ± 4.1 nm3 (at 0.58 ML). The distributions of cluster aspect ratio (h/w) for these coverages are also given in Fig. 5.4d. Interestingly, irrespective of the coverage, the aspect ratio in general increases with the basal width of cluster, w. This implies that for any given cluster that grows with arrival of Co adatoms, the vertical growth of the Co clusters is faster than the lateral growth. It reflects the continual preference of this physisorbed system to keep the graphite surface exposed as much as possible. The STM images show that the density of the Co clusters increases with Co coverage, θ. Measurement of cluster density, N as a function θ and deposition time, t is plotted in Fig. 5.4e. Based on the change of slope, three regimes can be identified as previously described in Chapter 2. The first regime is nucleation dominated. At this very early stage of growth, nucleation actively takes place at a constant nucleation rate, J due to vast supply of nucleation site on graphite and unrestricted diffusion of Co adatoms. Regime II takes place when nucleation slows down and dominated by cluster growth. This happens when diffusion length of new arriving adatoms are beginning to be 156 Adsorption of Co at room temperature restricted by the decrease of nearest-neighbour distance as N increases. Hence majority of them are captured by existing Co clusters before they are able to meet another adatom to nucleate a new cluster. Nucleation of fresh nuclei is still possible although at much lower rate than regime I and hence the slope of regime II is smaller than regime I. The decrease in N begins when the growth region of a cluster overlaps with another neighbouring cluster causing them to coalesce. A saturation of cluster density, Nsat is created when nucleation and coalescence rate are comparable. As coalescence rate takes over J, the N starts to decrease gradually as seen in regime III. The coverage-dependant N can be described empirically according to Eq. (5.1) below where the change of N with time, dN(t)/dt, will depend on J i.e. the initial slope near t= and also the probability of nucleation which can be assumed as (Nsat – N(t))/Nsat [13]:  N  N (t )  dN (t )  J  sat  dt Nsat   N (t ) 0 dN (t ) J dt  0t Nsat  N (t ) Nsat   J  N (t )  Nsat 1  exp   t     Nsat    (5.1) where N(t)= at t= 0. From the best fit of pre-coalescence regime in Fig. 5.4e (i.e. regime I and II), we obtained J as 9.5 x 108 clusters cm-2sec-1 and Nsat as 1.65 x 1011 clusters cm-2 for Co flux of 2.4 x 10-3 MLs-1. Empirically, in one second, there are approximately 7.3 x 1010 atoms cm-2† competing for 3.82 x 1015 adsorption sites cm-2 on graphite. Substituting J and Nsat into Eq. (5.1) gives us N(t)= 9.47 x 108 clusters cm-2 in one second. Hence on average each cluster contains 77 atoms assuming complete condensation with diffusion length of Co at room temperature as 3.25 x 10-3 m‡. † ‡ Assuming atomic density of Co hcp at 3.03 x 1013 atoms cm-2. Diffusion length, dl ~ 1/N, where N is the cluster density. 157 Chapter (a) 0.14 ML 0.29 ML 0.43 ML 0.86 ML 2.16 ML z (Å) 10 50nm 12 16 20 24 x (nm) 0.58 ML 4.0 (c) 0.14 ML 0.29 ML 3.0 0.43 ML 3.5 0.58 ML 2.5 100 Average cluster size (nm; nm ) Height, h (nm) (b) 2.0 1.5 1.0 0.5 80 60 Width, Height, Volume, 40 20 0.0 10 12 14 16 0.0 0.2 0.4 0.6 Co coverage (ML) Cluster width, w (nm) (e) 0.30 I 0.43 ML II III 1.6 11 0.58 ML 0.20 2.32 2.0 0.29 ML 0.25 Aspect ratio (h / w) Co coverage (ML) 1.16 0.58 1.74 0.14 ML Cluster density ( x 10 /cm ) (d) 0.8 0.15 0.10 0.05 0.00 1.2 0.8 0.4 0.0 10 12 14 Cluster width, w (nm) 16 10 12 14 16 Deposition time (min) Fig. 5.4 (a) STM images (300nm x 300nm) of sequential Co growth on HOPG at room-temperature with Co flux of 2.4 x 10-3 ML/s; cross-section of one of the Co clusters at 0.14 ML is shown as inset, (b) distribution of Co cluster heights and widths; (c) average cluster size (width, height and volume) of few selected coverage, (d) aspect ratio of Co clusters and (e) density of Co clusters, N as a function of deposition time and coverage. The density profile can be divided into regimes i.e. regime I (nucleation-dominated), II (growth-dominated) and III (coalescence regime). Solid line is the best fit for cluster density in pre-coalescence (regime I) based on Eq. (5.1). Nsat= 1.65 x 1011 clusters cm-2, J= 9.5 x 108 cluster cm-2s-1. 158 Adsorption of Co at room temperature 5.1.4 Influence of surface contamination Next we shall probe the effect of having a minute contaminants present on graphite surface on interaction and morphology of Co. As discussed in Section 4.1, the ex-situ cleaning will leave behind minute amount of oxygen and hydrocarbon on the graphite surface. Similar to Co growth on clean graphite (in-situ prepared), Co was deposited at 2.9 x 10-3 ML/s on the contaminated graphite (ex-situ prepared) at room temperature. The coverage-dependant C 1s spectra are shown in Fig. 5.5a. Unlike Co/clean graphite (Fig. 5.1a), a new feature located on the low binding energy side (283.3 eV) for Co/contaminated graphite is seen and it gets more pronounce with Co coverage (inset). To resolve the sub-components inside these C 1s spectra, peak fitting using software was carried out and as shown in Fig. 5.5b, they can be resolved into subcomponents which are labelled as A1, A2 and A3 for convenience of discussion. Peak fitting of C 1s of Co/ clean graphite are also carried out to provide comparison. Together they are tabulated in Table 5.1 below. (b) (a) (b) (a) A2 (C-C) A1 (C-H) A3 (C-Co) Fig. 5.5 (a) C 1s spectra for Co grown on contaminated graphite. Shoulder (also shown in inset) on the lower energy scale increases as the Co coverage increases. They indicate carbide (C-Co) formation; and (b) peak fitting of C 1s from 35 ML of Co grown on contaminated graphite. Details of peak fitting and peak assignment are tabulated in Table 5.1. 159 Chapter For the nucleation on T site, we found that the Co clusters most of the time not sit directly on top of the 6x6 maxima but next to these maxima. As earlier photoemission studies shows that elemental Si content on the surface increases with Co coverage, the adsorption site studies by STM suggest that this elemental Si would come from the Sirich 6x6 maxima where the Co clusters nucleates on. We suggest this is due to partial charge transfer arising from Co to Si that result in weakening of Si tetramer bonding to it’s underneath layer and hence increment of elemental Si content on the surface. The STM images that capture the morphologies of Co after a sequential deposition on 6√3 at room temperature are shown in Fig. 5.18a. It is evident that the density of Co clusters and their sizes increase with Co coverage. As shown in the cross-section of one of the selected clusters (Fig. 5.18b) and the coverage, θ dependant size distributions presented in Fig. 5.18c, Co adopts a 3D growth mode on 63 surface where clusters with several monolayer height are in fact are already formed even at θ of 0.06 ML. § The cluster size however is smaller than their counterpart on HOPG. As discussed in Section 4.4.1, this small cluster size is attributed to surface roughness of 63 that impeded the diffusion of Co. Interestingly the Co/63 clusters have the same cross-section as those observed for Co/HOPG i.e. 3D dome-shaped and share the same range of aspect ratio as Co/HOPG i.e. between 0.1 and 0.2 as shown in Fig. 5.18d. However, the shift of the size distribution (Fig. 5.18c) and aspect ratio (Fig. 5.18d) with θ not follow the trend of Co/HOPG (Figs. 5.4b and 5.4d). By comparison, the width distributions for Co/63 seem to be confined within certain window i.e. between 0.5 and 3.5 nm regardless of the θ of Co deposited, while the height increases. § One monolayer (1ML) is defined according to Co hcp crystal structure where ML is equivalent to Å. As shown by the cross-section in Fig. 5.17b, the thickness of the cluster is approximately ML. 190 Adsorption of Co at room temperature (a) 0.06 ML 0.10ML 20nm 0.44ML 0.18ML (c) 1.0 (b) 0.06 ML 0.10 ML Height, h (nm) 0.8 3.4 nm z (Å) 6.7 Å 0.6 0.4 0.2 10 x (nm) 15 0.0 20 (d) Cluster width, w (nm) Co coverage (ML) (e) 0.10 0.19 0.29 0.30 I 11 Cluster density ( x10 / cm ) Aspect ratio (h / w) 0.38 0.48 40 0.06 ML 0.10 ML 0.25 0.20 0.15 0.10 0.05 II III 30 20 10 0.00 Cluster width, w (nm) 0.0 0.4 0.8 1.2 1.6 2.0 Deposition time, t (min) Fig. 5.18 (a) STM images (100nm x 100nm) of sequential Co growth on 6√3/6H-SiC(0001) at room-temperature with Co flux of 4.0 x 10-3 ML/s; (b) cross-section given by a line profile of a selected Co cluster from 0.1 ML; (c) cluster size (height and width) distributions of Co on 63; (d) aspect ratio of Co clusters and (e) density of Co clusters as a function of deposition time and coverage. Solid line is best fit for cluster density in the pre-coalescence regime according to Eq. (5.1). Nsat= 2.97 x 1012 clusters cm-2, J= 9.0 x 109 cluster cm-2s-1. 191 Chapter The density of Co/63 cluster vs. deposition time (or coverage) is provided in Fig. 5.18e. Three regimes i.e. nucleation (I), growth (II) and coalescence (III) are identified based on the slope change. Fitting the pre-coalescence regime (regime I & II) according to Eq. 5.1 gives Nmax and J of 2.97 x 1012 clusters cm-2 and 9.0 x 109 cluster cm-2s-1 respectively. Both values are about an order higher than those found on Co/HOPG which is reasonable since the size of the Co/63 clusters is much smaller than Co/HOPG. 5.3.2 Nucleation and growth of Co on graphene Figure 5.19a shows the sequential growth of Co on graphene/6H-SiC(0001) at room temperature after the clean substrate is exposed to a constant Co flux of 4.3 x 10-3 ML/s. Similar to Co deposited on HOPG, the Co grows 3D on this surface showing the characteristic of a physisorption system. The density of Co and the size of each cluster are also observed to increase with θ. The growth however displays one dissimilarity from Co/HOPG. Unlike Co/HOPG (Fig. 5.8a(i)), there is no preferential of nucleation of Co at the steps. One of the possible reasons for this is lack of dangling bonds at the steps since a continuous layer of graphene across the steps of 6H-SiC(0001) would formed as described in Section 4.5.3. Unlike Co/63, there is also no preferential adsorption (i.e. T, H and B site) seen on the graphene surface. One interesting observation however is the size variation with coverage between these three carbon surfaces. Similar to Co/63 (Figs. 5.18c-d), the size and aspect ratio of Co/graphene seems to confine within a certain cluster width window regardless of the θ (Figs. 5.19b-c). Such confinement is clearly not observed for Co/HOPG. 192 Adsorption of Co at room temperature (a) 0.13 ML 0.26 ML 0.51 ML 1.28 ML 2.04 ML 50nm 0.77 ML (b) Co coverage (ML) 0.52 1.03 1.55 2.06 2.58 0.13 ML 0.26 ML 1.0 0.8 0.6 0.4 11 Aspect ratio (h / w) 0.25 0.20 0.15 0.10 0.05 0.2 0.0 0.00 10 Cluster width, w (nm) 2.0 I 0.30 0.13 ML 0.26 ML 1.2 Cluster density ( x 10 / cm ) 1.4 Height, h (nm) (d) (c) II III 1.6 1.2 0.8 0.4 0.0 10 Cluster width, w (nm) 10 Deposition time, t (min) Fig. 5.19 (a) STM images (300nm x 300nm) of sequential Co growth on graphene/6H-SiC(0001) at room-temperature with Co flux of 4.3 x 10-3 ML/s; (b) cluster size (height and width) distribution of Co on graphene; (d) aspect ratio of Co clusters and (d) density of Co clusters as a function of deposition time and coverage. Solid line is best fit for cluster density in the precoalescence regime according to Eq. (5.1). Nsat= 1.85 x 1011 clusters cm-2, J=1.45 x 109 cluster cm-2s-1. The density vs. θ plot is also fitted with Eq. (5.1) with Nsat and J as 1.85 x 1011 clusters cm-2 and 1.45 x 109 cluster cm-2s-1 respectively. These values are an order lower than those obtained for Co/63 using the same incident Co flux. For a given Co incident 193 Chapter flux, Nsat and J are very similar between Co/graphene and Co/HOPG. As previously discussed in Section 4.4.1, one of the reason is perhaps impedance of Co diffusion by the surface roughness of 63, which is almost twice the roughness of graphene/6H-SiC(0001). 5.4 Nucleation and critical nucleus size, i* 3.0 12 Cluster density (x 10 / cm ) 3.5 2.5 Co/ 3 6Ö3 Co/ HOPG Co/ graphene 0.2 2.0 0.1 1.5 0.0 1.0 0.0 1.0 2.0 Coverage (ML) 0.5 0.0 0.0 0.4 0.8 1.2 1.6 2.0 2.4 Coverage (ML) Fig. 5.20 Plot of cluster density vs. coverage, θ for Co clusters on three C surfaces i.e. 63/6HSiC(0001), HOPG and graphene/6H-SiC(0001). Co fluxes used are 4.0 x 10-3 ML/s, 2.4 x 10-3 ML/s and 4.3 x 10-3 ML/s respectively. The cluster nucleation dynamic on a surface will depend on the interplay between surface temperature, concentration of adatoms on the surface (arrival flux) and the type of surface (host) [37]. The competition between them will in turn define nucleation rate J, and cluster densities such as N(t) and Nsat. In this work, when same growth temperature (room temperature) and similar flux (2.0 to 4.0 x 10-3 ML/s) are used, the nucleation dynamics are mainly affected by the host. As shown in Fig. 5.20, the cluster density of Co 194 Adsorption of Co at room temperature on these three carbon surfaces are very different. Co/63 has the highest density since the 63 surface is more corrugated than the atomically smooth graphite and graphene. For Co/HOPG and Co/graphene, they appear to have same nucleation kinetic as shown by the similarity of their density-vs-θ plot (Fig. 5.20 and enlarged in the inset) and also the similar cluster sizes (Figs. 5.4 and 5.19). However, as discussed previously their cluster height vs. cluster diameter plot (Fig. 5.4b and Fig. 5.19c respectively) show a rather two distinctive behaviours, implying their nucleation kinetic may be different on atomistic level and will be probed as follow. The influence of underlying substrate (host) onto nucleation can be probed in more detail by investigating the critical nucleus size, i* (atom) needed to form the smallest stable cluster, (i*+1) on the surface. By definition, a cluster of size (i*+1) or larger has higher probability to grow than decay (cluster dissociation) and hence all the clusters observed under STM are the stable clusters that formed from i*. i* strongly depends on the substrate temperature where bigger i* is needed to form a stable cluster when temperature increases. As previously introduced in Chapter 2, a scaling model developed by Bartelt et al. has been widely used to deduce i* and it has the following form [38]: N s  s 2   f  s  s  (5.2) where Ns denotes the density of cluster size s, is mean size of coverage  and defined as =  /NT. Scaling function, f(x) defines the shape of the cluster size distribution as a function of x-variable i.e. s/, the scaled cluster size. This function is normalized such that  x0 f ( x)dx   x0 xf ( x)dx 1 . This model also works best for low coverage regime notably the pre-coalescence regime where diffusion of adatoms is not limited by θ. This 195 Chapter model is applicable to isotropic system i.e. no preferential diffusion along certain direction and islands formed are strain free where scaling the size distributions from various coverage collapse onto a common curve with a single peak (monomodal) around s/= [39,40]. Scaling according to Eq. (5.2) is performed for Co grown on these three surfaces in order to further understand the nucleation kinetics of these physisorption systems. In this case, the cluster volume, v is used as the scaling size, s. (i) Co/HOPG (a) 0.40 (b) 1.2 0.14 ML Co/HOPG 0.29 ML 0.30 1.0 N v / θ 0.43 ML 0.8 0.58 ML N v/ N T fitting i*=1 0.14 ML 0.29 ML 0.43 ML 0.58 ML Co/HOPG 0.20 0.6 0.4 0.10 0.2 0.0 0.00 50 100 150 v (nm ) 200 250 300 0.0 1.0 2.0 v / 3.0 4.0 Fig. 5.21 (a) Distributions of cluster volume (v) for Co on HOPG. These distributions are measurements made from STM images in Fig. 5.4; and (b) scaling of volume distributions in (a) according to Eq. (5.2). The red solid line for 0.26 ML is drawn to guide the eyes. Theoretical fit for i*=1 (solid black line) is also provided. Fig. 5.21a shows the volume distributions of Co/HOPG from 0.1 to 0.6 ML. These coverages are chosen mainly because they are in the pre-coalescence regime (Fig. 5.4e). The average cluster volume and the width (FWHM) of the distributions increase with coverage. The increment of FWHM is mainly caused by cluster volumes from fresh nuclei which are smaller than the existing clusters from previous coverage. However, 196 Adsorption of Co at room temperature when nucleation rate of fresh nuclei decreases i.e. crossing from region I to II in Fig. 5.4e, the cluster volume distributions for coverages near this cross-over i.e. 0.43 ML versus 5.8 ML are less affected by fresh nuclei and becomes comparable as seen in Fig. 5.21a. Upon applying the scaling analysis, the distributions at various coverages in Fig. 5.21a (except 0.14 ML) collapse onto a common curve as shown in Fig. 5.21b. The scaled data is best fitted with a theoretical curve with a critical island size i*= 1, suggesting that i= (from i*+1) i.e. dimer is the smallest stable Co nuclei formed during growth at these coverages. The scaled data however is slightly skewed to the left with the peak of v/ at ~ 0.75 instead of 1. For 0.14 ML Co, scaling analysis reveals a very different nucleation behaviour of Co/HOPG. The scaled distribution is now akin to i*= (see Fig. 2.5) where the maximum of the scaling curve peak at v/ = [41,42]. In this instance, Co monomer (i= 1) is the smallest stable nucleus and not the dimer. Ganz et al. [19] using STM has been able to observe monomers of Ag, Au and Al appearing on β or bridge site of HOPG and stayed stable for more than several seconds. They attribute this to the stability of the monomer being strongly bound to substrate. In addition, recent DFT calculations have also shown that Co monomer can also bound strongly to the hole site of graphene [43,44,45]. The long life time at these sites may promote nucleation of stable cluster upon additional of one adatom to these monomers. Apart from this, the scaling analysis also revealed a noticeable hump near the value v/ ~ 2. The distribution thus appears to be bimodal. The scaling analysis therefore shows nucleation behaviour of Co/HOPG is rather complex where the distributions (Fig. 5.21a) as well as the critical island size, i* (from to 1) change as the Co coverage increases. The effect of coverage on i* has been seen in 197 Chapter the growth of 3D InAs quantum dots on GaAs(001) [46] as well as in the growth of Ni on GaAs(110) [47]. For the growth of InAs quantum dots on GaAs(001) in the regime before saturation island density, Nsat, i*= is observed and this was attributed to the effects of strain due to the presence of lattice mismatch. At higher InAs coverage beyond the Nsat where strain effects are minimal, i*= is seen. For the case of Ni/GaAs(110) growth, nucleation is proposed to occurs via exchange with the surface atoms and this gives rise to i*= 0. Nucleation via exchange mechanism also resulted in a bimodal size distribution with two maxima upon scaling [48,49,50,51]. Comparing these reported work with the present study where Co clusters are physisorbed on HOPG, it is unlikely that strain or exchange with surface atom are the dominating factors leading to the peculiar nucleation behaviour seen. The coverage-dependant distribution and i* for Co/HOPG may be related to change in the mobility of the Co clusters as their size grow with coverage. Following the computation work on nucleation behaviour involving mobile and immobile islands, Kuipers and Palmer [52] showed that as the island mobility increases from the case of immobile islands to more mobile islands, the scaled size distribution changes shape and as a result splits into two parts, giving rise to a bimodal distribution. The first part is a broad maximum obtained without island mobility while the second part arises from the dramatic increase of islands size that bigger than the islands in first part as a result of island mobility. For our experimental observations, the effect of cluster mobility on i* of Co/HOPG can be postulated as follow. DFT calculations have shown that when the coverage is extremely low where only Co monomers are formed on a graphene surface, these monomers are strongly bound to the hole site of graphene [43,44]. However as the size of the nuclei increases gradually with coverage i.e. from monomer (i= 1) to tetramer 198 Adsorption of Co at room temperature (i= 4), the initial adsorbate-substrate bonding strength per Co atom is found to decrease. This is due to a stronger adsorbate-adsorbate bonding. Thus it may be anticipated that as the nuclei changes from i= to i= 2, and 4…, they becomes more increasing weakly bound to the surface. Therefore, during the initial growth, strong Co monomer-substrate bonding leads to immobile cluster mobility with i*= 0. As the nuclei grow, they become increasing weakly bound to graphite (physisorbed) and gain mobility. Clusters with higher mobility are expected to grow faster than those less or immobile clusters. As such the co-existence of freshly nucleated immobile cluster and mobile clusters which are larger in size would thus give rise to the bimodal distribution. The effect of cluster mobility is strongly manifested in scaling of 0.14 ML in Fig. 5.21b. While the first part of the distribution shows i*= 0, a noticeable hump in the second part is associated with cluster mobility. Nucleation rate as seen in Fig. 5.4e decreases with increasing coverage and becomes negligible when Nsat regime is reached. At higher coverages above 0.14 ML, the average volume of Co clusters also increases. The mobility of these larger clusters may become restricted. The effects of cluster mobility on scaling behaviour may become less. As such the distribution of clusters arising from i*= 0, distribution arising from mobile clusters and distribution arising from large immobile clusters may overlap and hence giving rise to a monomodal distribution but skewed as depicted by the scaling of 0.29 to 0.58 ML Co (Fig. 5.21b). Since the scaling skewed towards the regime of v/ less than which indicate population of small clusters are relatively higher, it is anticipated that the cluster mobility decreases more dramatically than nucleation of fresh island with i*= 0. It is clear that without the scaling results from 0.14 ML, it might be misinterpreted 199 Chapter that i*= operates for Co/HOPG. In conclusion, Co monomers are the smallest stable cluster (i*= 0) on HOPG. However, due to stronger Co-Co bonding as the clusters grow, the clusters become increasing weakly bound to graphite surface and gain mobility. This gives rise to bimodal distributions. As the clusters mobility decreases as their size increases, the distributions become skewed towards regime v/ less than 1. The nucleation dynamics of Co/HOPG is thus more complex and it is not sufficient to describe the process by assuming typical homogenous nucleation scaling analysis which ignores the influence of cluster mobility. (ii) Co/63 (a) 0.30 (b) 1.0 0.06 ML Co/ 63 0.10 ML 0.20 Nv/NT 0.06 ML 0.8 N v / θ 0.25 fitting i*=1 Co/ 63 0.15 0.10 0.10 ML 0.6 i*= 1(69%) and (31%) 0.4 0.2 0.05 0.0 0.00 0.0 1.0 2.0 v (nm3) 3.0 4.0 0.0 1.0 2.0 3.0 4.0 v / Fig. 5.22 (a) Distributions of cluster volume, v for Co on 63 surface. Solid line is guide for the eye; and (b) scaling of these volume distributions according to Eq. (5.2). The distributions are measurements made from STM images in Fig. 5.18. The scaling for Co/63 fits a scaling profile simulated for i*= 1(69%) and i*= (31%). Figures 5.22a shows the distributions of cluster volume of Co adsorbed on 63 surface where the average volume and width (FWHM) of the distribution increases with coverage. Upon scaling, these distributions collapse onto a common curve as shown in 200 Adsorption of Co at room temperature Fig. 5.22b. The main peak for the scaled data resembled most closely to theoretical fit of i*= (dotted-line curve), suggesting the smallest stable clusters as dimer. Theoretical fit with i* of higher order (i*= 2, 3…) will have bigger maximum and smaller FWHM, which are obvious will not fit the scaled data as well as i*= 1. The scaling for Co/63 also appears to be skewed slightly towards v/ less than regime i.e. 0.85. In this system, we exclude the effect of cluster mobility as the source for the skewness observed since the scaling of Co/63 does not change with coverage and the degree of the skewness is less than that observed for Co/HOPG. Instead we attribute the skewness observed to multiple adsorption sites on 63. As previously observed there are three possible adsorption sites on 63 surface i.e. 69% of the Co clusters are found next to the Si-rich 6x6 maxima (T site) and 31% on the bridge (B) and hole (H) site (Fig. 5.17). The presence of different adsorption sites in principle can result in different nucleation behaviour. One effect is on the formation and stability of the smallest stable nucleus, i*+1. The preferential of Co at T site would suggest smaller critical nucleus, i* at this site than those nucleate on B or H site. Statistically the distribution of Co clusters generated from smaller i* will dominates over higher i* since their nucleation is preferred by the system. In this case, clusters nucleated on the T site gives rise to the main peak in Fig. 5.22b which is best fitted with i*= 1. To investigate the influence of multiple adsorption sites on scaling, we simulate such environment by assuming adsorptions at B and H site have i* of and in accordance with experimental observation, the combined distribution of clusters has 69% of i*= (T site) and 31% of i*= (B and H site). The simulated scaling curve is plotted in the same chart in Fig. 5.22b as solid red line. By comparison, both scaled data and simulated curve are skewed to regime of v/ less than 1. Both distributions are also broader in the regime of v/ more than where two small 201 Chapter humps can be seen. As indicated by our simulation, the distributions in this regime are due to intensities from i*= 2. Since the peak area for scaling function f(v/) is conserved, the additional intensities from i*= at regime v/ more than causes the scaling curves (both simulated and scaled data) to peak at a maximum lower than scaling curve of i*= (100%). (iii) Co/ epitaxial graphene (a) 0.16 (b) 1.6 0.13 ML 0.26 ML Co/ G 1.2 N v / θ 0.12 Nv/NT fitting i*=2 fitting i*=3 0.13 ML 0.26 ML Co/ G 0.08 0.8 0.04 0.4 0.00 0.0 10 20 30 v (nm3) 40 50 60 0.0 0.5 1.0 1.5 2.0 2.5 3.0 v/ Fig. 5.23 (a) Distributions of cluster volume, v for Co on graphene surface and (b) scaling of these volume distributions according to Eq. (5.2). The distributions are measurements made from STM images in Fig. 5.19. The scaling in (b) fits the profile for i*= 3. Figure 5.23a shows the volume distributions of Co clusters nucleated on graphene. Similar to Co/HOPG and Co/63, the average cluster volume and FWHM of the distribution increases with Co coverage. The scaling analysis performed on these distributions is provided in Fig. 5.23b. As observed, the distributions from two coverages collapse onto a common scaling curve that best fitted with i*= (after comparing with i*=2), indicating i= or Co tetramer as the most stable nuclei on graphene. The scaling of 202 Adsorption of Co at room temperature Co on graphene shows very interesting characteristics when compared to Co/HOPG. Despite HOPG and graphene are both sp2-bonded C atoms in honeycomb lattice and also the similarities of Co morphology (density, size) on both surfaces, the i* for these two system are rather different. In comparison to Co/HOPG and Co/63, the scaled distribution for Co/graphene is also more akin to isotropic systems where the f(v/) is symmetrical at v/ = [39,40,53,54]. The difference of f(v/) between Co/HOPG (i*= 0) and Co/graphene (i*= 3) implies that graphene is more inert towards Co adsorption than graphite. One possible reason for this dynamic is perturbation of the  electrons of graphene. It has been shown by several others that there is a net charge transfer from the 63 interface to the graphene overlayers [21]. Aizawa et al. demonstrated that charge transfer from their TaC(111) substrate to the thin graphite overlayer weaken the  bonding of graphite [55]. Hence we expect the similar nature occurs to the  bonding of graphene/6H-SiC(0001). Since formation of i*= for Co on HOPG is facilitated by strong interaction between the d electrons of Co and  electrons of graphite, weakening of these  bonding in graphene is believed to greatly modify the nucleation dynamic of Co on graphene. As such bigger i* is expected for nucleation of stable cluster on graphene. 203 Chapter 5.5 Conclusions Co is supported on pristine graphite via physisorption where no silicides or carbide is formed at the interface. Co grows 3D in the form of hemispherical domeshaped nano-clusters. The nucleation dynamic is found to be very sensitive to contamination where upon contaminations that less than a monolayer, Co changed from dome-shaped clusters on pristine surface to irregular islands on contaminated surface. These irregular islands have better wetting on contaminated graphite via carbide formation. XPS reveals that the carbide formation between Co-C involves only adventitious carbons and not carbon atoms of graphite. The study above highlights that preparation of clean surface is crucial for interface and morphology studies where presence of minute amount of contaminants can give misleading observations. Co adsorption on 63 is rather more complex where adsorption of Co is accompanied by increase of elemental Si content on the surface and have been found to originate from 6x6 maxima. As the adsorbed Co exhibits metallic states, it is postulated that the interaction between Co and 6x6 maxima occurs via partial charge transfer, resulting in weakening of bonding between Si tetramer and its underlying layer. Co adsorption on graphene is more akin to Co/HOPG i.e. Co physisorbed on graphene. An increase in elemental Si also detected but in lesser amount than those observed for Co/63. They are believed to occur via defects, deep pits and exposed step edges where penetration of Co to the 63 interface takes place. The distribution of cluster size of Co/graphene is similar to those on HOPG. Co 204 Adsorption of Co at room temperature clusters formed on 63 has a smaller size but higher density than Co/HOPG and Co/graphene. The main difference is attributed to the corrugation of 63 surface that impeded the diffusion of Co adatoms. The diffusion length of Co is expected to be much higher on atomically flat graphite and graphene. All three systems obey the typical diffusion and aggregation processes on surface where the coverage-dependant cluster density of Co on HOPG, 63 and graphene can be divided into three regimes i.e. nucleation dominated, growth-dominated and coalescence-dominated. Upon performing scaling analyses on the cluster size distributions, critical nucleus size, i* of these three systems are found to be very different with i*= for Co/HOPG, i*= for Co/63 and i*= for Co/graphene. The small i* for Co/HOPG is attributed to initial strong monomersubstrate bonding. As for Co/graphene, weakening of  bond due to charge transfer from the 63 interface is responsible to observation of the larger i*. The last section of this work demonstrated modification of nucleation dynamic of Co once the electronic state of the host (HOPG vs. graphene/63) is perturbed by the underlying interface. 205 [...]... min Co 10 min Co 1 min Co Graphene 10 min Co 1 min Co 65 63 61 59 57 55 106 2p1/2 104 Binding energy (eV) (c) Si 2s 2p3/2 102 100 98 Binding energy (eV) (d) 50 o h 350 eV 50 o C 1s h 350 eV 50 o o 50 153 Intensity (a.u.) Intensity (a.u.) 2 85 B2 0o C6 C5 0o B1 30 min Co 156 154 152 Binding energy (eV) 10 min Co 1 min Co Graphene 158 30 min Co 10 min Co C4 C3 C2 1 min Co 150 Graphene 148 290 288 286 284... Graphene B1 B4 C6 50 o C5 C3 C2 Intensity (a.u.) Intensity (a.u.) 50 o 0o 63 0o 63 C4 C6 C5 C3 C2 C1 B3 50 o B4 B2 50 o B1 0o 0o 158 156 154 152 150 Binding energy (eV) 148 290 288 286 284 282 Binding energy (eV) 280 Fig 5. 9 (a) VB of 63 and graphene recorded at two different take-off angles i.e 0o and 50 o using a photon energy of 60 eV The energy range from 14 – 26 eV recorded from graphene, which contains... dome-like and more regular, (b) Right panels show 0.45ML of Co on contaminated graphite Clusters are irregular and (c) cross section of Co clusters labeled as A, B and C in (a) and (b) The aspect ratio (height/diameter) for cluster A and C is ~0.08 and for cluster B is ~0.02 1 65 Chapter 5 5.2 Photoemission studies of Co on (63x63)R30o and graphene/ 6H-SiC(0001) 5. 2.1 Comparison between clean 63 and graphene. .. extend of charge transfer on each band [22] 167 Chapter 5 (a) (b) Graphene Valence band h 60 eV o Si 2p h 350 eV 50 Graphene Graphene o 0 A4 50 o 20 16 0o 63 A1 0o Intensity (a.u.) Intensity (a.u.) 24 A3 50 o A4 63 50 o A3 2p3/2 0o 2p1/2 [0001] 50 θ h 0 16 14 12 10 8 6 4 2 0 -2 -4 Binding energy (eV) (c) A1 o 1 05 (d) Si 2s h 350 eV A2 o 103 101 99 Binding energy (eV) 97 C 1s h 350 eV C4 B3 Graphene Graphene... h 350 eV 98 97 50 o C 1s h 350 eV 50 o 286 152 Intensity (a.u.) Intensity (a.u.) 50 o C4 C6 C5 0o B4 8 min Co 8 min Co 3 min Co 1 min Co 63 3 min Co 1 min Co 63 158 156 C3 C2 C1 0o B3 B2 B1 154 152 150 Binding energy (eV) 148 146 290 288 286 284 282 Binding energy (eV) 280 Fig 5. 12 High resolution core-level spectra of (a) Co 3p; (b) Si 2p; (c) Si 2s and (d) C 1s acquired using photon energy of 350 ... states of graphitised surface The graphene valence band recorded at 50 o is again also similar to those observed by Johansson et al [20] The VB spectra can be described as follows: (i) ~3 eV: a very small bump associated to p band of graphene; (ii) 4 .5, 5. 5, 6 .5 and 8 eV: these states are subcomponents of px,y-derived σ band; (iii) 11 eV: a less intense peak and they are associated with hybridisation of. .. acquired at 50 o (more surfacesensitive) for each core-level is also stacked together for easier comparison and they are displayed as inset 179 Chapter 5 (a) (b) 50 o Co 3p 50 o Si 2p h 350 eV h 350 eV 60 101 Intensity (a.u.) Intensity (a.u.) 50 o 50 o 0o 8 min Co 3 min Co 1 min Co 63 3 min Co 1 min Co 65 (c) 63 61 59 57 Binding energy (eV) 50 o Si 2s 55 A4 A2 A3 0o 8 min Co A1 2p3/2 2p1/2 1 05 104 103 102... (Fig 5. 14b), the Si 2p (Fig 5. 14c )and Si 2s (Fig 5. 14d) is located at 101.6 eV and 153 .0 eV respectively for Si-C bond Similar to Co/63, the elemental Si signals are seen to increase upon adsorption of Co and Co remains metallic (more evident in the VB scan) 184 Adsorption of Co at room temperature (a) (b) 50 o Co 3p h 350 eV 50 o Si 2p h 350 eV 60 50 o 102 A1 o Intensity (a.u.) Intensity (a.u.) 50 0o... core-level spectra of Si 2p, Si 2s and C 1s for both 63 and graphene using photon energy of 350 eV 169 Chapter 5 These spectra, which recorded at two different take-off angles (0o and 50 o), are displayed in Figs 5. 9b-d respectively They are distinctively different between these two surfaces and the results arising from peak fitting are also displayed in Fig 5. 9 and tabulated in Table 5. 2 The Si 2p peaks... contains the C 2s states of graphite, is shown as inset; high-resolution core-level of (b) Si 2p, (c) Si 2s and (d) C 1s recorded using photon energy of 350 eV Results from peak fitting of each core-level are tabulated in Table 5. 2 The geometry of measurement is shown in (a) 168 Adsorption of Co at room temperature Table 5. 2 Peak fitting for Si 2p3/2, Si 2s and C 1s of 63 and graphene recorded at 0o . band of graphene; (ii) 4 .5, 5. 5, 6 .5 and 8 eV: these states are subcomponents of p x,y -derived σ band; (iii) 11 eV: a less intense peak and they are associated with hybridisation of C 2s and. (a.u.) 14 8 150 152 154 156 158 Binding energy (eV) Intensity (a.u.) 97991011031 05 Binding energy (eV) Intensity (a.u.) 0 o 50 o Graphene 50 o 0 o h  350 eV 63 Si 2p A3 A1 A2 A3 A1 50 o 0 o Graphene 50 o 0 o h. -4 24 20 16 Graphene 0 o 50 o Intensity (a.u.) Bindin g ener gy ( eV ) 6  3 Graphene 50 o 0 o 50 o 0 o h  60 eV Fig. 5. 9 (a) VB of 63 and graphene recorded at two different take-off angles