Dynamics of epitaxial graphene growth and adsorptions of cobalt 1

i revisit the precursor phase prior to formation of graphene on silicon carbide, 6H-SiC0001 substrate i.e.. Recent reported works have also showed the possibility to tune opening the ban

1

Introduction & Research Objectives

There are four main objectives in this thesis i.e (i) revisit the precursor

phase prior to formation of graphene on silicon carbide, 6H-SiC(0001)

substrate i.e (63x63)R30o; (ii) to probe the dynamics and kinetics

occurring during the evolution from 63x63)R30o to graphene, (iii) to

investigate the growth dynamics of Co at room temperature and

comparing the similarities and differences between Co on graphene and

Co on graphite; and (iv) to investigate the change in the growth dynamic

when Co is deposited on graphene at elevated temperatures

This chapter is organised as follows: (a) background and motivation

leading to this thesis; (b) literature reviews; (c) research objectives; and

(d) the organisation of this thesis

1.1 Introduction and Motivation

The discovery of graphene has opened up a new paradigm in nanoelectronics that could surpass the performance of conventional semiconductor devices owing to its superior mobility that near the speed of light [1,2], anomalous half-integer quantum Hall effect [1,3] and quasi-relativistic Klein tunneling [4] Research on graphene, ranging from fundamental interests to exploration of its potential as a new platform for nanoelectronics, have been vigorously pursued since the first discovery in 2004 [5]

Graphene is essentially an isolated single sheet of graphite i.e a two dimensional

sp2–bonded honeycomb lattice of carbon atoms (Fig 1.1a) The peculiarity of its transport properties arises from its unusual valence and conduction band Having four valence

orbitals (2s2 2p2) for each carbon atom, three of the orbitals (one s and two p orbitals) hybridised to form three sp2 bonding orbitals The free lone p orbitals which oriented

perpendicular to the graphene plane hybridise to form the  (valence) and * (conduction)

bands of graphene At the K and K’ point of the graphene Brillouin zone (Fig 1.1b),

these  and * bands meet at the Fermi level, forming a zero gap Instead of showing the typical parabolic dependence, these bands change linearly with the momentum and therefore the electron transport of graphene is essentially obeying the Dirac’s (relativistic) equation (Fig 1.1c) Owing to this, the electrons behave like massless Dirac fermion* and gives rise to many interesting Dirac-properties mentioned above With the emergence of graphene, there is thus a tremendous interest to integrate graphene, carbon nanotubes and fullerenes together as a platform for fully carbon-based nanoelectronics

* Because  and * band touching each other, the Dirac point also coincides with the Fermi level, hence the

Fabrication of graphene-based devices requires reliable and convenient process for mass production Mechanical exfoliation, which was the first method employed to produce ‘free-standing’ graphene in the laboratories [1,3,5], involves lifting off several layers of graphene sheets from a graphite crystal using adhesive tape and transfer onto the SiO2 as supporting substrate This method produces high quality flakes of graphene but it

is time consuming and impractical for large-scale production as it lacks consistency in reproducing flakes of the same sizes Apart from this method of preparation, there are also several avenues to produce graphene on a larger scale These includes epitaxial growth via solid-state decomposition of silicon carbide (SiC) at above 1200oC [6,7,8,9] and cracking of a carbon carrying gas on a metallic surface [10,11,12] Of all these methods, the solid-state decomposition of SiC is perhaps the most direct route to prepare

Fig 1.1 (a) A graphene sheet with atomic arrangement of carbon atoms in a 2-dimensional

honeycomb lattice The sp2 bonding gives shorter (1.42 Å) C-C bonding The unit cell of graphene with lattice parameter of 2.46 Å is also drawn Grey and red spheres represent the two

different sublattice of graphene The basic vectors a1 and a2 of the lattice are shown; (b) the first Brillouin zone of the honeycomb lattice and (c) the linear dependence of conduction () and valence (*) bands around K and K’ points

K’ *





Ek

graphene Besides being able to produce graphene epitaxially, it is also compatible and easier to integrate with current Si-based CMOS technology It has been reported that the thicknesses ranging from single to multiple layers graphene can now be prepared depending on the growth parameters Recent reported works have also showed the possibility to tune (opening) the band gap of graphene via substrate effects from SiC, thus bring this material a step closer to realise carbon-based nanoelectronics [13,14,15,16].*

Some of the unique properties of exfoliated graphene, however, fail to replicate themselves in SiC supported graphene (hereafter epitaxial graphene) Despite having identical atomic structure with the exfoliated graphene and having demonstrated that the

 and * bands obey Dirac’s equation [14,16,17], epitaxial graphene exhibits some dissimilarities to those exfoliated graphene [18] These includes for example, (i) the absence of the quantum Hall effect and this has been attributed to the weak localisation imparted from coupling with underlying substrate [18], (ii) a reduced carrier mobility by two orders of magnitude even after formation of wide terraces (which should aid in reducing electrons scattering) [19] and (iii) instead of forming zero band gap, a small gap

is observed between  and * [13,14,15] At present, it has also been suggested that the first graphene layer interacts weakly with underlying substrate where the unshifted * states and a pinned Fermi level at ~3 eV observed for graphene are similar to graphite characteristics [7,17] This observation does not explain the disparity of properties between exfoliated and epitaxial grown graphene supported on SiC substrates

One of the main challenges in resolving these discrepancies is because epitaxial

* The biggest hurdle to utilise graphene as an electronic material is its zero gap property at Dirac point,

graphene is supported on a rather complex interfacial structure whose electronic and structural details are still not well understood The growth of epitaxial graphene from the Si-terminated SiC(0001) is known to occur when the surface becomes increasingly carbon (C) rich due to gradual depletion of Si during thermal annealing This SiC(0001) surface undergoes a series of surface reconstruction which is increasingly deficient in Si and the formation of graphene takes place when a C-rich phase known as (63x63)R30o

(hereafter 63 for short) has formed and annealed to temperatures above 1200oC [7] Despite being first discovered in 1975 [20], the structural details of this 63 phase are still

in debate There are several different and contrasting structural models reported in the literature A more thorough review of these various models will be given in Section 1.3 (Literature Reviews) The problem becomes more complicated when it was found later that the decomposition of this initial surface to form graphene also resulted in the formation of a new 63-like phase at the interface between the graphene surface and the SiC substrate [15,21,22]

Clearly, the structure and the nature of interactions (covalent vs van der Waals) at the graphene/interface and together with defects/impurities/lattice strains are anticipated

to cause scattering of charge carriers and change the transport properties of graphene For these reasons, it is of interest to revisit this structure and elucidate the salient features leading to formation of epitaxial graphene on the SiC substrates Providing an insight into the structure of 63 will allow better understanding of the interface-graphene interaction (covalent vs van der Waals), growth mode and properties of epitaxial films produced Knowledge of the mechanism leading to the formation of graphene monolayer will also allow the fabrication process to be tailored to control the thickness of graphene and also improve the quality and properties of graphene

Being a platform for nanoelectronic, it is unavoidable that graphene will come in contact with metal to form electrical contact As the technology node always continue to miniaturise, understanding and probing the interaction of graphene with metal at atomic level becomes indispensable Another added-value for using graphene as host for metal studies is its “inertness” which may prove to be an excellent supporting surface for formation of magnetic dots/ clusters of a few hundred atoms Technology exploitation of cluster with properties superior than the bulk requires them to be supported on surfaces or embedded in matrices However, the superior cluster size-dependant magnetism which comes from additional orbital moments [23], often altered or disappeared when these orbitals interacts with or perturbed by the support or the medium For example, non-magnetic Ru has shown the importance of this aspect when its supported clusters display

4d ferromagnetism on inert graphite [24] but appear non-magnetic on noble metals such

as Ag and Au [25]

For the film materials, Co appears to be an ideal candidate due to its ferromagnetism Free Co clusters have been reported to be more magnetic than bulk for sizes up to 400 atoms [23] In addition, recent computation by Xiao et al shows that Co dimer adsorbed on a C(0001) surface displays the highest perpendicular magnetic anisotropy energy (MAE) (~100 meV per atom) [26] This reported value is about an order higher than the next giant MAE reported from Co adatoms on Pt(111) [27,28] For this reason, Co/graphene system will be vital for future molecular magnetic storage [29] Given that properties at the nanoscale are sensitive to the interaction at interface and also the morphology (island sizes, densities and distributions), there is therefore a need to

understand and control the growth kinetics and dynamics involved in producing these atomic clusters The importance of this is further spurred by recent controversies involving Co/C(0001) system At ambient temperature, we have found that Co nucleate as metallic 3-dimensional clusters that physisorbed on graphite surface [30 ] However, results from computational work has been contradictory with some suggested strong binding between Co and graphene sheet [31,32] while other proposed weak binding i.e physisorption [ 33 ] Up to now, there is no experimental study reported for Co on graphene

It is also fundamentally interesting to probe the dynamics of a system involving two elements of very different surface energy Similar to graphite, graphene is expected

to possess very low surface energy This has been recently confirmed via the liquid droplet test [34] In this instance, the surface energy of Co (2709 mJ/m2) [35,36] is almost fifty times higher than graphene monolayer (46.7 mJ/m2) or graphite flakes (54.8 mJ/m2) [34] Hence the dynamics of Co on graphene at elevated temperature may totally draw a very different picture When the system is energetically allowed, such a huge difference

in surface energy will vigorously drives the system to minimise the total surface energy This is normally achieved by either (i) Ostwald ripening to form bigger islands on the surface, or (ii) adsorbates are embedded inside the substrate The second scenario is particular interesting where this mechanism can be use to form embedded magnetic dots

or to form layered structure Although the second scenario is not often reported in comparison to surface coalescence, surface burrowing of gold (Au) by (Bi) surface [37] and Co cluster by Au surface [38] have been observed For the case of metal deposited on graphite or graphene surface, embedding of adsorbate will result in formation of graphite (or graphene)-intercalated structure In this case, Co may insert into graphene as

individual atoms or clusters A 2D magnetism of Co is also possible when Co is sandwiched in between two layers of graphene epitaxially grown from SiC Thus far surface intercalation between graphene and underlying SiC substrate are reported for Ni [39] and more recently Au [40] The dynamic and energetics for surface intercalation has yet to be fully understood as the study of metal intercalation underneath graphene is still relatively new More rigorous studies are required since this structure may give rise to properties similar to the graphite-intercalated-compounds (GICs)

In summary, there are four main goals in this work They are (i) revisit the (63x63)R30o precursor phase structure prior to formation of graphene on 6H-SiC(0001) substrate; (ii) to probe the dynamics and kinetics occurring during the evolution from 63x63)R30o to graphene, (iii) to probe and compare the similarity and difference between the growth dynamics of Co on graphene and Co on highly oriented pyrolytic graphite (HOPG) at room temperature and (iv) the dynamics of Co on graphene at elevated temperatures where surface burrowing of Co is observed in this work

In the next section, the controversies involving the precursor phase of graphene i.e 63 phase will be reviewed Various structural models have been proposed for the 63 phase The significance and short falls of these models will be compared The current status of the graphene growth dynamic on SiC is also discussed in this section The section ends with review on metal adsorption on graphene and graphite, which will provide the background to the third and fourth objective of this thesis

1.2 Literature Reviews

1.2.1 (6 3x63)R30 o

: the precursor phase prior to graphene/6H-SiC(0001)

The structure of hexagonal SiC(0001) and various phases leading to epitaxial graphene will be discussed prior to review of the 63 phase Emphasis is given to 6H-SiC(0001) since our preliminary studies found this polytype produces better quality of epitaxial graphene than 4H-SiC(0001)

(i) Introduction to crystal structure of 6H-SiC(0001)

Figure 1.2a shows a single unit of Si-C structure that consists of a Si atom bonded

to four nearest C atoms (or vice versa) via sp 3 bonding network with tetrahedral angle of 109.5o This single unit of Si-C structure is the building block for all the polytypes of SiC For hexagonal polytypes, repeating this unit in two directions as shown in Fig 1.2b gives rise to a (0001) plane of the hexagonal SiC A SiC bilayer (BL) is defined as a layer consists of Si alternating with the nearest C layer along a direction, in this case the [0001] direction In a hexagonal symmetry, there are three positions (A, B or C) which the layers can be stacked along the [0001] with each position rotated 60o from one another (Fig 1.2c) A change in the stacking sequence resulted in change of the unit cell or a new polytype

Fig 1.2 (a) A single unit of Si-C structure; (b) translation of this unit structure along two directions perpendicular to [0001] direction creates the (0001) plane of hexagonal SiC; (c) the orientation of next layer can be the same or rotated by 60o This rotation creates many polytypes

of SiC; (d) cross section view of a 6H polytype as projected on the (112 0) plane The switch of stacking sequence occurs after the third SiC bilayer (from top) and (e) half of a unit cell of 6H-SiC(0001) The top views i.e (0001) plane for both full and half unit cell in (c) and (d) are also provided at the bottom The top views show that due to switch in stacking sequence by 60o, the unit cell (solid line diamond) is rotated 60o when a 6H unit cell is terminated at half a unit cell

Top view

The most common polytypes for hexagonal (H) SiC is 2H (ABAB), 4H (ABAC) and 6H (ABCACB) The crystal structure of 6H polytype is shown in Fig 1.2d with 6 BL stacked in a sequence of ABCACB (with reference to position of Si atoms) The average Si-C bond length is 1.89 Å, the vertical distance for a Si-C off-bond (bonded sidewise) is 0.63 Å and hence the interlayer distance between Si-Si layers is 2.52 Å (see Fig 1.2d)

The cell parameters are a = 3.08 Å (the nearest distance between two Si or C atoms) and c

= 15.12 Å (height of one unit cell of 6H i.e 6*2.52 Å) The top view of the (0001) plane

is also illustrated in Fig 1.2d with the unit cell drawn as solid line The switch of stacking sequence occurs after the third SiC bilayer with the unit cell rotated 60o Furthermore, there are two possible polarities of the crystal, where in the case of bulk truncation with Si-C bilayer intact, the (0001) surface is Si-terminated, while the (0001) surface is C-terminated [41]

(ii) Surface reconstruction and phases of 6H-SiC(0001)

The Si-terminated surface of 6H-SiC(0001) undergoes various phases when heated under ultra-high vacuum (UHV) conditions to different temperature regimes These phases are in local equilibrium with the annealing temperature and evolve from Si-rich to the more thermodynamically stable C-rich phases As-received SiC wafers are normally covered with layers of native oxides To remove the native oxide and passivate the surface prior to vacuum preparation, the substrates are normally chemically treated using methods similar to hydrofluoric acid (HF) etching of oxides on silicon [42,43,44]

However this ex-situ preparation is not sufficient to prepare atomically clean surface In

principal, the clean surface can be prepared by heating the surface under UHV conditions

at temperatures above 800oC The oxides are normally removed as volatile SiOx species Due to uncontrollable removal of Si from the Si-terminated 6H-SiC(0001), defects are normally generated To minimise these defects, an external supply of Si flux is normally used during heating or annealing such that the arriving Si will “etch” away the oxides on the surface as SiOx and hence minimise the desorption of the intrinsic Si from bulk SiC When the clean surface is prepared without external Si source, the SiC(0001) surface will transform from 1x1 to (3x3)R30o reconstruction [44] However if Si flux is used, a Si-rich phase i.e 3x3 is observed due to higher supersaturation of Si on this surface [45] As the surface is heated to higher temperature regimes, various other phases are created due

to continuous sublimation of Si with temperature Important phases (ascending with temperatures) that consistently observed by others are (3x3)R30o (between 900 to

1000oC), (63x63)R30o (between 1050oC to 1190oC) and finally the graphitised phase (above 1200oC) [46,47,48,49,50] Auger electron spectroscopy (AES) shows that surface prepared using Si flux are Si rich for the 3x3 and (3x3)R30o and C-rich for 63 before the surface becomes completely graphitised [49] Equation (1.1) shows the surface transition:

SiC-1x1+Si, ~850 3x3 (3x3)R30o (63x63)R30o Graphene

(1.1)

- Si

Fig 1.3 Left panel shows filled-state STM images and the respective LEED pattern as the SiC(0001) undergoes from 3x3 [51], (3x3)R30o [52], (63x63)R30o [52] and finally the graphitised states [53,54] as the surface is annealed with increasing temperatures The surface transforms from Si-rich to C-rich as Si is gradually desorbed from the surface The 1x1 LEED diffraction pattern is highlighted using six white circles Except 63 and graphitised phase, the right panel shows the well-received structural model proposed for the 3x3 [57] and 3 phases [58]

Si

SiC bulk

Si SiC bulk

Si SiC bulk

Left panel of Figure 1.3 shows the scanning tunneling microscopy (STM) images and the corresponding low-energy electron diffraction (LEED) pattern of the 3x3 [51], (3x3)R30o [52], (63x63)R30o [52] and graphitised 6H-SiC(0001) [53,54] as the 6H-SiC(0001) surface is progressively annealed from 850oC to above 1200oC Based on the quantitative analyses of Si and C intensities obtained from AES [49] and X-ray photoelectron spectroscopy (XPS) [ 55 ], the 3x3 and (3x3)R30o surface are still relatively Si richer than the SiC-1x1 bulk In addition, XPS results also reveals elemental

Si signals which suggest the protrusions imaged under STM for these surfaces are resulted from Si clusters Since then, several models have been put forward to elucidate the 3x3 [45,56,57,58] and (3x3)R30o phase [44,58,59,60,61] The well-received structural model of these two phases is presented in the right panel of Fig 1.3

The 3x3 structure that proposed by J Schardt et al [57] consists of Si tetramers sitting on top of a full Si adlayer which itself sits on top of the bulk Si-terminated SiC(0001) surface The (3x3)R30o unit cell which is 3 times smaller than 3x3 surface

is easier to elucidate Several groups thus far have agreed on a model where a Si adatom sitting at the T4 site of the bulk Si-terminating layer of SiC(0001) [44,58,59,60,61] This configuration attracts the Si atoms from bulk towards the Si adatoms and induced buckling of first C sub-layer (Fig 1.3) By comparison, the Si density from the 3x3 model

is higher than the 3, which is in agreement with the AES and XPS observations [49,55]

(iii) Controversies related to (6 3x63)R30 o

reconstruction of 6H-SiC(0001)

Since the first discovery back in 1975 [20], the structure of (63x63)R30o

(hereafter 63 for short), which is the precursor phase for graphene, is still in controversies The first controversy comes from disparity between the LEED diffraction pattern and the STM of this surface While the LEED strongly suggests this surface has a long-range order of 63 periodicity, STM suggests otherwise Filled-state STM images in Fig 1.4a-c shows that the surface is predominantly a 6x6 reconstructed surface The imaging of this surface under STM is also staggering where it can manifest in a few different features while maintaining a strong 6x6 periodicity as shown in Fig 1.4a-d In

Fig 1.4 Various filled-state STM images of the (63x63)R30oobserved by different groups Image (a), (b), (c) and (d) are extracted from Refs 52, 62, 63 and 53 respectively The 6x6 unit cell is drawn in each STM image except d(iii)

Fig 1.4a, Mårtensson et al observed the 63 surface seems to be decorated with 6x6 trimer-like clusters with individual clusters scattering around the trimers [52] Chen et al

on the other hand observed this phase as a surface with 6x6 protrusions (Fig 1.4b [62]) or

a 6x6 honeycomb (Fig 1.4c [63]) More recently Riedl et al also observe this surface consists of trimers similar to Mårtensson et al (Fig 1.4d(i)) [53] In addition, they also claimed they observed rings of two different sizes and shapes where the bigger ring is hexagonal while the slightly smaller ring is pentagonal (in Fig 1.4d(ii) and (iii)) They suggest that the alternate arrangement of these two rings can give rise to the 63 periodicity and hence bridge the observation of STM to the LEED pattern observed

The efforts to resolve the structure of 63 surface are also impaired by the complexity of this surface As the AES and XPS in general revealed the Si/C ratio of this surface ranges between 1.6 to 0.6 [49,55], debates have been ongoing whether this surface is fully C-terminated or mixture of both Si and C elements In general, most of the research groups assume this surface is fully C-terminated [7,63] However recent published reports suggest that this 63 surface still consists remnant of Si that are responsible for the bright 6x6 protrusions observed under STM [21,64]

Beginning with models proposed based on a fully C-terminated 63 surface, the first model has been put forward by van Bommel et al where based on their LEED result, they suggested the 63 surface consists of a graphite monolayer resting above a bulk-truncated SiC surface [20] (see Table 1.1, Model 1) This is because in real space, a 30orotated mesh of 13x13 unit cells of graphene matches exactly a (63x63)R30o unit cells

of SiC(0001) This model initially seems accurate when Chang et al., via STM and STS,

graphene overlayer rest above a 3x3 surface (Table 1.1, Model 2) instead of 1x1 bulk [7,47,66] However the presence of a graphene overlayer is later disputed by Mårtesson et

al when they do not find any signature for graphene from their photoemission studies [67], which later confirmed by their STM studies [68] Instead, they attributed the 63 diffraction is a collective diffraction from 5x5, 6x6 and 3x3 phases co-exist on their surface [68] The disparity between these results is later attributed to the 63 phase formed at the interface beneath the graphene layer This structure (graphene/63/SiC) gives rise to 63 diffraction pattern while at the same time graphene lattice is imaged under STM [52] (this can be seen in Fig 1.3 where STM of the graphitised 6H-SiC(0001) showing 6x6 honeycomb beneath the graphene monolayer) As the graphene layers grow thicker, both of the diffraction from SiC-1x1 and 63 patterns become diffuse and disappear with a new 1x1 diffraction originated from graphene layers emerged [7] This new 1x1 pattern rotated 30o from SiC(1x1) and consistent with STM imaging where the graphene layer rotated 30o from SiC More recently, Chen et al proposed a different model for the 63 phase (see Model 3) [63] Based on the STM imaging and the intensity

ratios and binding energies of various C 1s components recorded from their

photoemission studies, they proposed a fully C-terminated 63 surface However, instead

of covering the surface with graphene, they suggested that the 6x6 honeycomb observed under their STM imaging (Fig 1.4c) arises from carbon clusters of two different sizes

bonded via sp3 bonding Surrounding these clusters are small graphene islands Their structure however is at odd with spectroscopy studies reported by others where large amount of non-graphitic carbons and presence of Si are observed [67]

Several recent work continues to report detection of Si-related structures on this