First principles study of graphene related materials 2

4, we use density-functional theory calculations to explain the experimentalobservations during the synthesis of graphene quantum dots from the decomposition of C60 on Ru0001 surface, an

In this part, we discuss two methods for the synthesis of graphene-related materials.

In Chap 4, we use density-functional theory calculations to explain the experimentalobservations during the synthesis of graphene quantum dots from the decomposition

of C60 on Ru(0001) surface, and also during the chemical vapor deposition growth of

hexagonal boron nitride on Ru(0001)

64

Understanding the synthesis process of

graphene-related materials

Abstract: To utilize graphene-related materials in mass-produced devices, we need ways

to synthesize these materials in large-scale, in good quality, and in the desired size and shape Computational studies can shed light on the observations made by experimentalists during the synthesis process of these materials, thus assisting the experimentalists to find better ways to synthesize these materials In this chapter, we use density-functional theory (DFT) calculations to investigate the process of synthesizing geometrically well-defined graphene quantum dots through the catalytic decomposition of C60 on the Ru(0001) sur- face, and also the nucleation and growth of a hexagonal boron nitride monolayer on the stepped Ru(0001) surface using chemical vapor deposition.

* Only the DFT simulations were conducted by the Ph.D candidate The experimental results were obtained by others and are included here for a coherent discussion of the topic.

4.1 Graphene quantum dots from catalytic

decomposi-tion of C60

4.1.1 Introduction

In Sec 1.2, we reviewed the current methods for synthesizing graphene and noted thepromising method of chemical vapor deposition (CVD) on metal surfaces for producingmacrosize graphene If we wish to use the CVD method to grow graphene nanoislands

of smaller sizes, it is clear that we must limit the aggregation of carbon fragments onthe metal surface, and there are two ways that this can be accomplished Firstly, thediffusion path length of a carbon fragment before meeting another fragment must be

65

sufficiently long, which means that the amount of precursors deposited on the surfaceshould be as little as possible Secondly, the diffusion velocity of the carbon fragmentsshould be as low as possible, so it is advantageous to use metals that form strong bondswith carbon.

Furthermore, if we want the graphene nanoislands grown to have highly regular shapes,some thought has to be given to the choice of carbon precursors, since we want theresulting carbon fragments to be regular in shape and bond to each other in predictableways For instance, Treier et al created regular-shape graphene nanoribbons and quan-tum dots through the cyclodehydrogenation of cyclic polyphenylene on the Cu(111) sur-face.100Inspired by the metal-catalyzed cyclodehydrogenation of polyaromatic molecules

to form C60 and the direct transformation of graphene to C60,171,172 we believe that

the reverse process – which is the synthesis of regular-shape graphene quantum dots(GQDs) from the metal-catalyzed fragmentation of C60 – should be possible Thus far,

there are only studies reporting the appearance of large graphitic domains generatedfrom the decomposition of C60 on metal surfaces.173,174We hypothesized that the frag-

mentation of C60 should produce uniform-size fragments, and by controlling the

diffu-sion path length and velocity of the fragments as described above, GQDs with uniformgeometries may be produced Herein, we report our experiments confirming the syn-thesis of well-defined GQDs from the cage-opening of C60 catalyzed by the Ru(0001)

surface.175

4.1.2 Results and discussion

4.1.2.1 Experimental results: Graphene quantum dots from C60 decomposition

Fig.4.1(a) shows the scanning tunneling microscope (STM) image of the infinite graphiticlayer that is formed on Ru(0001) when 0.7 monolayer (ML) of C60 is deposited on the

Ru surface and annealed at 1200 K for 5 minutes Note that a monolayer of C60 film

would produce five graphene layers if all the C atoms had remained on the surface.Since a single layer of graphene was observed on the Ru(0001) surface after the anneal-ing, substantial gasification of the decomposed carbon fragments must have occurred

Due to the lattice mismatch between the graphene layer and the Ru substrate, a graphenemoiré structure with a lattice constant of ∼ 30 Å was formed Portions of the graphenestructure sitting close to the Ru substrate (the ‘valleys’) show up as dim spots on theSTM image, whereas areas of graphene that are buckled upwards from the metal surface(the ‘humps’) show up as bright spots.

For free-standing graphene, all 6 C atoms in each hexagonal ring of graphene appear

as individual bright spots under STM imaging.176 In cases where some of the atoms inthe hexagonal ring are coupled to a substrate: for example in bernal-stacked graphene

of a few layers, only 3 atoms out of 6 atoms that are not directly coupled to anothercarbon atom in the adjacent graphene layer are imaged under STM.177 For the infinitegraphene layer on Ru(0001), we observed that all 6 atoms of the hexagonal ring wereimaged for the humps, whereas only 3 out of 6 atoms per hexagonal ring were imagedfor the valleys

When the C60 coverage is reduced to 0.03 ML and annealed to 500 − 600 K, the C60

molecules were observed to diffuse across the Ru surface and become embedded intothe Ru surface [Fig.4.1(b–c)] The decrease in apparent height for the embedded C60

was ∼ 0.5 Å The embedding of C60 molecules adsorbed on metal surfaces have been

observed before on Pd(110),178 Pt(111),179 Au(110),180 and Ag(111).181 The nomenon has been attributed to the formation of metal-atom vacancies underneath the

phe-C60 molecules, which favorably enhances the C60-metal adsorption enthalpy since C60

is able to form more bonds with the metal surface The increased adsorption enthalpymore than compensates for the unfavorable enthalpy in forming the metal-atom va-cancy We hypothesize that such vacancies form under the C60 molecules on Ru(0001)

too, and confirm this hypothesis through density-functional theory (DFT) calculationsthat will be shown later

Upon flash annealing of the embedded C60molecules to 725 K for 2 minutes, numerous

carbon clusters appear on the Ru surface, whose 2.7 Å apparent height is 60% lowerthan the 6.5 Å apparent height for C60 under a similar bias voltage About 23% of the

decompo-= 0.3 V, I decompo-= 0.25 nA; (e) V decompo-= 0.3 V, I decompo-= 0.2 nA; (f) V decompo-= 0.3 V, I decompo-= 0.16 nA.

decomposition fragments are flower-shaped clusters with a three-fold symmetry andlateral diameter of ∼ 0.7 nm [Fig 4.1(d)]; the other 67% are hexagonal mushroom-shaped clusters with lateral diameter of ∼ 0.9 nm [Fig 4.1(e)] When the annealingtemperature is increased to 825 K, larger clusters with lateral diameter of ∼ 1.2 nm areseen [Fig.4.1(f)].

If the surface coverage of C60 on the surface is increased to 0.08 ML, large graphene

quantum dots (GQDs) are seen in addition to the clusters described above, as shown inFig 4.2(a) A 1 minute annealing at 725 K results in a (15 ± 1)% yield of triangularGQDs with apparent lateral size of 2.7 nm [Fig.4.2(b)] Further annealing of the sample

at 825 K for 1 minute produces perfect hexagonal GQDs with lateral diameter of 5 nm in

a (30 ± 2)% yield [Fig.4.2(e)], while carbon clusters with diameters < 5 nm completelydisappear Other shapes with lower yields like parallelograms [Fig.4.2(c)], trapezoids[Fig.4.2(d)], and large hexagons [Fig.4.2(f)] are also seen These GQDs exhibit size-dependent band gaps; the differential conductance (dI/dV) characteristics of the GQDsobtained using scanning tunneling spectroscopy (STS) are shown in Fig.4.2(g), and theband gaps calculated from the analysis of the dI/dV curves are 0.8 eV, 0.6 eV, 0.4 eVand 0.25 eV for the triangular, parallelogram, hexagonal, and large hexagonal GQDs,respectively

The triangular GQDs show the same moiré pattern and dim/bright contrast as infinitegraphene on Ru(0001) Curiously, in contrast to the infinite graphene on Ru(0001), only

3 out of 6 atoms per hexagonal ring were imaged even for the humps (bright regions

in the STM image) that are not strongly coupled to the Ru substrate This is likely due

to the ‘topological frustration’ of the π-electron conjugation in finite-size graphene, asexplained in Sec 1.1.1 We simulate the STM image using DFT calculations and theTersoff-Hamann approximation for a triangular GQD that is similar in size and shape toFig.4.1(e), to show that edge effects alone are responsible for the ‘3-out-of-6’ image

f (IV), and for infinite monolayer graphene on Ru(0001) (V) STM parameters: (a,e) V

= 0.5 V, I = 0.1 nA; (a,e inset) V = 0.3 V, I = 0.2 nA; (b–c) V = 0.3 V, I = 0.2 nA; (d,f)

V = 0.3 V, I = 0.1 nA

4.1.2.2 DFT results: C60adsorption and fragmentation mechanism

To investigate if C60 adsorbs on the Ru surface with the formation of a metal atom

vacancy, we calculate the adsorption energies of C60 on Ru(0001) at different

adsorp-tion sites using density-funcadsorp-tional theory (DFT) Our simulaadsorp-tion model consists of a5-layer Ru(0001) periodic surface slab measuring 16.08 × 13.93 × 26.44 Å3with a C60

molecule on the surface [Fig.4.3(a)] The third to fifth layer of Ru atoms are held fixed

in position during any geometry optimizations The C60 molecules showed a

‘clover-leaf’ STM image with 3-fold rotational symmetry [Fig 4.3(b)], which indicates thatthe C60 molecule is oriented such that one of its hexagon rings is lying flat on the sur-

face.182,183 Thus, we only consider configurations where the C60 molecule is oriented

with a hexagon ring lying flat on the Ru surface for our simulations

We conduct spin-polarized calculations based on DFT using the SIESTA code,151 anduse the local density approximation for the exchange-correlation functional.184 We ap-ply Troullier-Martins pseudopotentials;185in the case of Ru, relativistic and non-linear

-9.077 a 1.361 b

Table 4.1: The lowest-energy adsorption configurations of C60 on different locations

on the Ru(0001) surface and their respective adsorption energies The top hemisphere

of the C60 is not shown for clarity aAdsorption energy was calculated with respect tothe energy of an Ru atom in the bulk metal, and bwith respect to an isolated Ru atom(see text)

core corrections are added Double-ζ plus polarization localized basis orbitals are used,with the C (2s, 2p) and Ru (4d, 5s) electrons treated as valence A meshcutoff of 300 Ryand a 3 × 3 × 1 Monkhorst-Pack186 sampling scheme are used Geometry optimiza-tions are conducted using the conjugate gradient algorithm until the Hellmann-Feynmanforce on each atom is less than 0.05 eV/Å

There are four different high-symmetry locations that the hexagon ring of C60 can sit

on the Ru(0001) surface: the fcc, hcp, bridge and on-top locations (see Table4.1) Thecenter of the hexagon ring of C60is sitting on the face-centered cubic (hexagonal close-

packed) interstitial site of the Ru(0001) surface for the fcc (hcp) location, while thecenter of the hexagon ring is located between two Ru atoms in the bridge location Theon-top location has the center of the hexagon ring on top of an Ru atom There is alsoanother degree of freedom associated with the rotation of the C60 molecule along the

axis that is perpendicular to the Ru surface for each location, and we investigate thedifferent high-symmetry configurations of C60 on the Ru surface based on the rotation

of the C60as well The adsorption energy of C60on the Ru slab is calculated by

where ERu+C60 is the energy of the system with C60 adsorbed on the Ru slab, and ERu

(EC60) is the energy of the isolated Ru slab (C60) The adsorption energies of the most

stable configurations are shown in Table 4.1, and the on-top configuration is the moststable Next, we investigate to see if the formation of a metal-atom vacancy underneaththe C60 molecule, as was seen for some transition metal surfaces,74,178–180 is energet-

ically favorable for the Ru(0001) surface If we remove the central Ru atom in theon-top configuration to form a Ru vacancy site (on-top_vac configuration in Table4.1),the adsorption energy of C60 is now calculated as follows:

Eads= ERu+C60− ERu− EC60− µRu, (4.2)

where µRuis the chemical potential of the Ru atom that has been removed The value

of µRuis likely to be between that of the Ru atom in the bulk metal and atomic Ru

Tak-ing these two scenarios into account, the possible adsorption energy of the on-top_vacconfiguration ranges from −9.077 eV (bulk Ru) to 1.361 eV (atomic Ru) If we fur-ther consider the fact that vacancy formation is an entropically favorable process, theon-top_vac system is certainly the most energetically favorable configuration

With the creation of the Ru vacancy, the interaction between the C60 molecule and the

Ru substrate is enhanced The bond lengths between the C atoms of the bottom hexagonring and the closest Ru atoms [grey-pink bonds in Fig 4.4(a)] are on average, 2.0%shorter in the on-top_vac than for the on-top configuration; while the bond lengths be-tween the second-tier carbon atoms of C60 and the closest Ru atoms [black–pink bonds

in Fig 4.4(a)] are on average 2.5% shorter in the on-top_vac configuration spondingly, certain C–C bonds in the C60 molecule are weakened due to the interaction

Corre-(a) (b)

Figure 4.4: (a) on-top_vac configuration of the C60 molecule The green arrow cates the top-down point of view, from which (b) is derived (b) C–C bond lengths (Å)

indi-of the bottom hemisphere indi-of the C60in (a) The top hemisphere is not shown for clarity

between C60 and Ru Fig 4.4(b) shows the bond lengths of the bottom hemisphere of

the optimized C60 of the on-top_vac configuration The C–C long bonds, which are

between a hexagon and pentagon ring in C60 (numbers labeled in red) in Fig 4.4(b),

have lengthened by 2.6% on average compared to the isolated C60 molecule, and the

C–C short bonds between two hexagon rings in C60 (numbers labeled in blue) have

been lengthened by 3.0% We surmise that these lengthened bonds constitute a faultline that causes the C60 to rupture upon heating into two asymmetrical hemispheres

The surface-retained fragment [red structure in Fig 4.4(b)] derived from the bottomhemisphere of the ruptured C60 evolves eventually into the observed surface-stabilized

clusters on the Ru surface [Fig 4.1(c–d)], while the top hemisphere of the C60 cage

desorb into the gas phase

4.1.2.3 DFT results: ‘3-for-6’ STM image of triangular GQDs

To address why only 3 out of 6 atoms per hexagonal ring of the triangular GQDs areimaged under STM, the local density of states of a 2.7 nm triangular GQD, similar

in size to the the one in Fig 4.1(e), with and without H atoms attached to the edges,was calculated We simulate two cases because the dangling bonds at the edges of theGQD should be partially quenched by interactions with the Ru substrate This bondingsituation is likely to be between the case of a GQD fully edge-terminated by H, and the

of the molecular model represent C (H) atoms.

case where the GQD is not edge-terminated by H at all STM image simulations wereconducted under the simplifying assumptions of the Tersoff-Hamann theorem,187whichstates that the STM images may be modeled by the summation of the local density ofstates (LDOS) from the Fermi level of the system to the tunneling voltage of interest

In Fig.4.5(b), the STM simulation of a triangular GQD without any edge termination

is shown, and as compared to Fig.4.5(a) where all the edge atoms are terminated by Hatoms, the STM signal at some atoms are diminished, but the ‘3-for-6’ STM pattern isstill clearly visible Both cases show the presence of electronic states near the Fermilevel that are localized on the C atoms of one sublattice only, which is consistent withthe prediction by benzenoid graph theory31 and Lieb’s theorem32,35 (see Sec 1.1.1),

even though the theory is strictly applicable only to the case of H-terminated GQDs Toconclude: the ‘3-for-6’ STM pattern seen experimentally for the GQDs are the purelydue to edge effects, and is not due to interactions with the Ru(0001) substrate

In conclusion, we have demonstrated the formation of regular-size graphene quantumdots (GQDs) on Ru(0001) substrate using C60 as a precursor We found through DFT

calculations the most stable configuration of C60 on Ru(0001), and confirmed that a

metal-atom vacancy is formed underneath C60during the adsorption process, similar to

what happens on the Pt(111) surface The interactions between C60and Ru

correspond-ingly weaken certain C–C bonds in C60, leading to a fault line in the C60 that ruptures

when the temperature is high enough That only 3 out of the 6 C atoms in each hexagonring of the GQDs shows up under STM imaging is wholly due to edge effects and is notdue to interactions with the Ru substrate

4.2 Growth of h-BN on Ru(0001)

4.2.1 Introduction

For technological applications, it is crucial to minimize structural defects and synthesizehigh-quality h-BN films on a large scale Many methods have been devised to synthe-size monolayer h-BN, as summarized in Sec.1.3.1 The use of metal surfaces as thecatalyst and deposition substrate in thermal catalytic chemical vapor deposition (CVD)

is the most promising approach for large-scale growth of h-BN.188 A wide polytype ofh-BN superstructures had been previously grown on metal surface, depending on thelattice mismatch and the interaction strength between h-BN and the underlying metal.The strength of interfacial bonding is determined by the amount of charge transfer be-tween the h-BN π states and the metal d band, which is small if the d-shells of the metalatoms are full Hence, Cu (d10) or Pt (d9) with greater filled d-shells have weaker bond-

ing to h-BN as compared to Ru (d7) For the Ru surface with a large lattice mismatch

and strong bonding with h-BN, the resulting h-BN topology becomes highly corrugatedwith periodic ‘nanomesh’ structures consisting of moiré ‘pores’ (h-BN domains thatsit close to the metal substrate) and ‘wires’ (h-BN domains that are further away fromthe metal surface).123 The relationship between the periodicity of the moiré structureand defect formation during the growth process has thus far been unexplored There isalso a lack of atomistic insight on the formation of vacancies and grain boundaries inthe CVD growth of h-BN To this end, we carried out a systematic scanning tunnelingmicroscopy (STM) study of the nucleation of h-BN on the Ru(0001) surface with theaim of understanding the formation of defects and how they can be suppressed.189

4.2.2 Results and discussion

4.2.2.1 Experimental observations during initial h-BN growth process

To observe what happens in the initial stages of h-BN nucleation during CVD, we heat

a clean Ru(0001) surface to 700 K and expose it to a low dosage [9.6 Langmuir (L)] ofborazine (B3N3H6) STM images show that the Ru surface becomes uniformly deco-

rated with triangular h-BN clusters, as shown in Fig.4.6(a–b) This is in contrast to theinitial stages of the CVD growth of graphene, where high-symmetry hexagonal clustersare seen in addition to the triangular ones.190–193Fig.4.6(b) shows the magnified STM

image of homogeneous magic clusters with lateral size of 1.5 nm The 3-D STM heightprofile analysis of a magic cluster is shown in Fig 4.6(c), where the center region ofthe magic cluster appears to be darker as compared to the corner regions due to thestronger interaction with the substrate, which gives rise to a ‘bowl-like’ adsorption con-figuration It is likely that the edge atoms of the clusters are passivated with hydrogen.The presence of hydrogen atoms in the initial stage of growth of h-BN has also beenproposed in a previous report.194

Larger h-BN islands with lateral sizes ranging from 10 − 50 nm appear after exposure

of Ru to 5 L borazine gas at 950 K Most of the h-BN islands are triangular, while therest show a truncated triangular shape with three long and three short edges Intrigu-ingly, the orientations of these triangular islands are rotated 60◦between adjacent steps

[Fig 4.6(d–e)], giving rise to antiparallel domains which alternate across the steps ofthe Ru(0001) surface

4.2.2.2 Simulations to explain experimental observations

To explain the alternating geometry of the h-BN islands between the adjacent steps ofthe Ru(0001) surface, we investigate the atomistic model of the adsorption of h-BNislands on the hexagonal close-packed (hcp) Ru(0001) surface For infinite h-BN ad-sorbed on transition metal surfaces, the N atoms were found to preferentially sit ontop of a Ru atom due to the strong hybridization between the N lone pair and Ru d

BN clusters with edges terminated by N atoms and B atoms, respectively (h) Atomisticmodel of a hexagonal h-BN cluster.

states.195 The B atoms can either sit in the high-symmetry hcp interstitial sites, or onthe face-centered cubic (fcc) interstitial sites; we denote these two configurations by

NtopBhcpand NtopBfcc, respectively Fig.4.7(b–c) shows a 1-ring h-BN cluster adsorbed

on Ru(0001) in these two different configurations We expect that an energy preferencefor either the NtopBhcpor NtopBfccconfiguration is what dictates the antiparallel orienta-

tion of the triangular islands between adjacent steps To compare the energy differencebetween h-BN clusters in the NtopBhcp or NtopBfcc configurations on the Ru(0001) sur-

face, we investigate the adsorption energies of 1-ring, 3-ring, and 6-ring h-BN clusters

on the Ru(0001) surface using density-functional theory (DFT) calculations

We perform the DFT calculations using the SIESTA package.151 The electronic ation is terminated when the energy difference is less than 0.1 meV The local-densityapproximation is used for the exchange-correlation functional Troullier-Martins pseu-dopotentials are used to describe the interactions due to ions and core electrons.185Themesh cutoff is 250 Ry We use double-z with polarization (DZP) atomic basis functions

YZ

= Ru

, ,

YX

YX(b)

(c)

Figure 4.7: (a) Sideview of the 8-layer thick Ru(0001) surface slab used to model theRu(0001) terrace (b) & (c) The top views of the NtopBfcc and NtopBhcp configurations

of a 1-ring h-BN cluster on the Ru(0001) surface, respectively

for B, N, and bulk Ru atoms, while Ru atoms at the surface have additional f tion orbitals We add 5s DZP ‘floating orbitals’ above the Ru surface to better describethe decay of the wave functions into the vacuum at the surface, which results in a bettermatch between the experimental and calculated surface energies.196 The atomic posi-tions are relaxed using the conjugate gradient algorithm until the Hellmann-Feynmanforce on each atom is less than 0.05 eV/Å We model the Ru(0001) surface using ahexagonal supercell that has 8 Ru atomic layers and 15 Å of vacuum in the directionperpendicular to the surface, as shown in Fig.4.7(a) The top 2 layers of atoms at bothexposed surfaces are allowed to relax The surface energy is 3.14 J/m2(cf 3.11 J/m2in

polariza-Ref 197) To compare the energy difference between h-BN clusters in the NtopBhcp or

NtopBfcc configurations on Ru(0001), we investigate the adsorption energies of 1-ring,

3-ring, and 6-ring h-BN clusters on Ru(0001) hexagonal supercells with lateral size of10.66, 13.33, and 15.99 Å, respectively The k-points are sampled using the Monkhorst-Pack method and the maximum spacing between the k-points is less than 0.07 Å-1 We

use a symmetric slab model where BN clusters are adsorbed on both exposed surfaces

of the slab The BN clusters and the top 2 layers of Ru atoms are allowed to relax The

energy difference per B atom is then calculated according to

Ediff= (Ehcp− Efcc)/2NB,

where Ehcp(Efcc) is the energy of the BN cluster adsorbed on the Ru(0001) slab with the

NtopBhcp (NtopBfcc) configuration and NB is the number of B atoms in the BN cluster

Table 4.2(a) shows Ediff per B atom for all the clusters studied, showing that NtopBfcc

configuration is more stable than NtopBhcp by 0.04 eV and the value is independent of

the cluster size

Thus, in order to adopt the energetically favorable NtopBfcc absorption configuration

exclusively, all triangular clusters become aligned in the same direction on one race However, it is necessary for the h-BN triangles to rotate by 60◦between adjacent

ter-terraces to accommodate the alternating ABAB packing of the hcp Ru(0001) steppedsurface, thus giving rise to the antiparallel arrangement of the h-BN islands observed inFig.4.6(d–e) On Cu or Ni substrate, the most stable binding configuration adopted bythe h-BN islands is also the NtopBfcc type However, all the h-BN clusters are aligned

parallely whether on the same or adjacent terraces, due to the fcc crystal packing ofthese metals In addition, the energy difference between the NtopBhcp and NtopBfcc

configurations is significantly reduced due to the weak interfacial bonding (0.009(2)eV/B atom for Ni198), resulting in a higher percentage of NtopBhcp oriented clusters on

Ni(111) or Cu(111) surfaces, as compared to Ru(0001)

Since the initial growth clusters are predominantly triangular in shape, we surmise thatthere must be a significant energy difference between the clusters terminated at the edgesolely by B atoms, and those that are terminated by N atoms As shown in Fig.4.6(f–h), triangular clusters are terminated at the edges solely by B or N atoms, whereashexagonal clusters must exhibit both edge types We investigate the formation energies

of h-BN clusters with B-terminated and N-terminated edges on the Ru(0001) surface

in the NtopBfcc orientation The formation energy of a 3-ring BN cluster is calculated

BN cluster Ediff(eV)

B-terminated

BN cluster Ediff (eV)

BN cluster Eform (eV)

Table 4.2: (a) Calculated Ediff for BN clusters of different sizes (b) Eform of the terminated and B-terminated 3-ring BN cluster adsorbed on Ru(0001)

N-according to

Eform= EBN+Ru− ERu− NB(EB+Ru− ERu) − NN(EN+Ru− ERu),

where EBN+Ru, EB+Ru, and EN+Ru are the total energies of the BN cluster, B, and

N atom adsorbed on the Ru(0001) supercell, respectively ERu is the energy of the

clean Ru(0001) supercell and NB(NN) is the number of B (N) atoms in the BN cluster

Eform is a measure of how readily individual B and N atoms adsorbed on the Ru(0001)

surface come together to form BN clusters Table 4.2(b) shows that Eform of the

N-terminated 3-ring BN cluster is lower in energy than the B-N-terminated one by 1.87 eV.The consequence of this is a symmetry reduction of the h-BN islands such that theymanifest almost entirely as triangular clusters, in contrast to the appearance of high-symmetry hexagonal islands in the initial stages of graphene CVD

4.2.2.3 Origin of void defects and fault lines in h-BN film growth

Fig 4.8(a) displays the growth of larger-sized islands after exposing Ru to 9.6 L razine gas at 800 K The majority of the islands observed here are ∼ 4 nm in size[circled in Fig 4.8(a)] These clusters are formed by the coalescence of three 1.5 nmmagic clusters which inadvertently trap a triangular void in the center region, as shown

bo-(a) (b) (c)

Figure 4.8: (a) STM images of the Ru(0001) surface after exposure to 9.6 L borazinegas at 800 K (b) Magnified view of the 5 nm size triangular BN islands circled in (a).(b inset) Height profile recorded along the green dotted line in (b) shows that the edge

of triangular islands is 0.4 Å higher than the center, which indicates that the edges ofthe h-BN islands are terminated with hydrogen (c) Sample (a) was heated at 900 K

to remove edge H atoms and the height profiles show that the center of island is nowhigher than edges

in Fig.4.8(b) The height profile in the inset of Fig.4.8(b) clearly reveals that the edge

of bowl is 0.4 Å higher than the center region of the bowl This topographic structure isreminiscent of the moiré ‘pore’ and ‘wire’ structure imaged in infinite h-BN nanomesh

on Ru(0001) Upon heating to 900 K, ∼ 95% of the triangular voids in h-BN are paired due to a move towards thermal equilibrium The topographic contrast of theh-BN clusters is also reversed, with the center region appearing brighter than the edge[Fig 4.8(c)] The center region is now ∼ 0.5 Å higher than the edge from the heightprofile analysis, as shown in the inset of Fig 4.8(c) This can be rationalized by thedesorption of H atoms from the edges of the h-BN clusters at high temperature, whichallows the edge atoms to couple to the Ru substrate Fig.4.9(a–e) shows the Ru(0001)surface after exposure to different amounts of borazine at 950 K, demonstrating that

re-∼ 5% of the voids cannot be annealed regardless of the amount of borazine introduced.This is presumably due to the large lattice distortion in the corners of the triangularvoids where different h-BN islands merge Incorporation of N and B adatoms in thecorners and the edges of voids are likely discouraged due to the high edge energy thatwill be generated consequently

Other than void defects, fault lines are also oft-seen during the CVD growth of h-BN.Fault lines, a type of grain boundary defect, occur wherever misregistry between twogrowth nuclei prevents their coherent merging at the growth front [Fig.4.9(f)] We shall

describe the three main candidates responsible for the development of these fault linesbelow.

The periodic unit cell (called the moiré unit cell) of a compound system consisting ofmonolayer h-BN adsorbed on a metal surface, is usually different from the periodicunit cell of the h-BN or the metal surface alone The size of the moiré unit cell isdetermined by the lattice matching condition: nh-BN× ah-BN= nM× aM, where nh-BN(nM) is the number of unit cell repetitions of h-BN (metal), while ah-BN (aM) is the

corresponding lattice constant.195 nM= 12 for the Ru(0001) surface, whereas nM =

9 for Pt(111) Fig 4.10(a) shows one precise nucleation site out of 144 possibilities(12 × 12) for a h-BN cluster to merge coherently with a second pre-existing cluster.The dark blue regions in the inset of Fig.4.10(a) shows the locations of the moiré porewhere clusters must nucleate in order that they merge coherently For Pt with a smallermoiré unit cell (9 × 9), the chance for two NtopBfcc domains to merge coherently will

be increased to 811 Although Cu is considered as a good substrate due to the smaller