Cao Z. (Ed.) Thin Film Growth: Physics, materials science and applications

Подождите немного. Документ загружается.

238 Thin ﬁlm growth

eased by slow cooling of Ni from ~ 1000°C down to room temperature (after

Ni has been enriched with carbon at high temperatures if required) (Kim,

2009). On the contrary, Ir can store much less carbon in its bulk (Arnoult,

1972), which implies that graphene growth proceeds only by impinging

carbon adspecies supplied to the surface.

a variety of carbon sources can be employed for the growth, and it is

probable that others will be used in the future. The most convenient source

is a gas of carbon-containing molecules (Fig. 10.6a). Such molecules are

usually hydrocarbons (Oshima, 1997). Until 2009, the gases were injected

at low-pressure in ultra-high vacuum systems (UHV). In 2009, gas mixtures

(e.g. Ch

/ar/h

) were also employed at ambient pressure (Kim; 2009; Reina,

2009a), opening valuable perspectives for graphene production. In both cases

(UHV and ambient pressure) the carbon-containing molecules are cracked

at the metal surface; for this the metal surface is made catalytically active

by heating it. This process can be considered an heterogeneous chemical

reaction between a gas phase and a solid one, which is why the technique is

often referred to as chemical vapour deposition (CVD). Another option for

graphene growth is the use of an atomic carbon source (Loginova, 2008).

Finally, in the case of metals having a non-negligible carbon solubility,

carbon segregation from the bulk towards the surface might be an alternative

route (Sutter, 2008).

In the following we focus on the growth processes disregarding the way

the carbon ad-species have been formed at the metal surface (CVD, atomic

10.6 Schematics for the different elementary processed during

CVD growth of graphene on a metal, with a methane precursor as

an illustration. (a) A methane molecule is adsorbed at the surface,

cracked into a carbon adatom, leaving two hydrogen molecules

released in vacuum. (b) The carbon adatom diffuses at the surface

(or sub-surface). (c) The carbon adatom incorporates in a carbon

chain, forming a pentamer (on metal surfaces like Ir(111) or Ru(0001),

Loginova, 2008). (d) The carbon chain diffuses at the surfaces and (e)

is incorporated in the graphene ﬂake.

(b)

(a)

(c)

(d)

(e)

ThinFilm-Zexian-10.indd 238 7/1/11 9:42:53 AM



239Epitaxial growth of graphene thin ﬁlms on single crystal metal surfaces

carbon source, segregation). This is legitimate if the nature of the carbon

source only affects graphene growth through the rate of adatom formation.

This is at least the case in prototypical systems that are suited to address

elementary growth processes, for instance growth by CVD with ethene or

by an atomic carbon source at single-crystalline metal surfaces (loginova,

2009b).

10.3.1 Growing graphene nanoﬂakes

A derivative of the CVD growth for graphene consists in (i) adsorbing

carbon-containing molecules at the metallic surface, at such a temperature

that the molecules do not desorb to vacuum, and (ii) increasing the sample

temperature for promoting hydrocarbon decomposition and eventually

graphene growth (Coraux, 2009).

The surface is rst saturated with molecules, and the resulting adlayer

prevents the strong bonding of further molecules at the surface. In practice,

ethene molecules are transformed into ethyledine ones at Pt(111) and Ir(111)

(Nieuwenhuys, 1976) surfaces at room temperature. The amount of carbon

in the resulting adlayer is less than that in a full graphene layer, due to steric

effects.

In the next step the substrate is heated up. In the course of increasing the

sample temperature, dehydrogenation of the molecules proceeds, leaving

pure-carbon adspecies being mobile at the surface. Nanometer-scale carbidic

islands develop from this carbon sea. The electronic interaction between the

carbidic islands and the metal surface is strong, all the more as the islands

are small, as has been shown for graphene on Ir(111). This is because in

small islands the proportion of edge atoms is large: such low coordinated

atoms have a tendency to form bonds with the metal atoms underneath. This

strong bonding was detected through energy shifts in the carbon and metal

atoms core levels. as a result of the enhanced bonding of graphene edges

to the substrate, the small graphene islands were shown to be dome-shaped

and characterized by a small distance to the metal surface, noticeably below

that for a plain graphene sheet on Ir(111), for island radii smaller than 1–2

nm, and all the more as the radii is small (Fig. 10.7) (Lacovig, 2009). Such

small islands, as imaged by STM (Fig. 10.8a), are found at the terraces of

the metallic surface, and occasionally bond to substrate step edges. They are

characterized by a poorly dened height and rounded edges. Further increase

of the temperature modies their morphology (Fig. 10.8b–e). Their height

becomes well dened as seen by STM for islands radii larger than a few

nanometres.

Such islands with well-dened height are bound by straight edges that

are precisely oriented along the <11-20> directions of graphene, consistent

with the prominent proportion of zigzag portions composing these edges. at

ThinFilm-Zexian-10.indd 239 7/1/11 9:42:53 AM



240 Thin ﬁlm growth

sufciently high growth temperature, it was shown for graphene on Ir(111)

that the island edges align to the <1-10> direction of Ir(111) (N’Diaye,

2008b). Graphene islands were also reported on Pt(111) (Land, 1992), Ir(111)

(N’Diaye, 2006; Coraux, 2009), and Co(0001) (Eom, 2009). The occurrence

of straight edges for graphene is the evidence that carbon mobility is active

at edges. In the explored growth temperature range (700–1300°C), these

effects are rather slow as the islands are most often not in their equilibrium,

hexagonal shape. Carbon diffusion at edges could proceed through that of

carbon pentagons continuously formed at graphene edges during growth

(Frenklach, 2004; Whitesides, 2010). Pentagon collisions result in hexagons,

1.62 Å

2.53 Å

(a)

(c)

(b)

(d)

[101

]

[01

2.63 Å

3.13 Å

10.7 Calculated structural models for graphene nanoislands on Ir(111)

as a function of their (small) size, on a side view and on a top view:

for (a) one, (b) three, (c) seven, and (d) 19 carbon rings. Note that the

average height of the islands increases with their size (from Lacovig,

2009, © The American Physical Society, http://prl.aps.org/abstract/

PRL/v103/i16/e166101).

ThinFilm-Zexian-10.indd 240 7/1/11 9:42:54 AM



241Epitaxial growth of graphene thin ﬁlms on single crystal metal surfaces

i.e. graphene building blocks (Whitesides, 2007). On the contrary, edge

smoothing through a 2D carbon gas surrounding the edges, that would be

formed via carbon atom detachment at edges (more readily at regions where

the 2D pressure is higher, i.e. at convex edges), is not a prominent process

at the temperature of interest.

The temperature to which the substrate is heated up determines the size

of the graphene islands: their average diameter can be adjusted between a

few and several tens of nanometres for 600°C and 1200°C respectively on

Ir(111). Through this temperature range, the island density drops by almost

three orders of magnitude, consistent with the fact that the total amount of

carbon at the surface is not modied. As the growth temperature increases,

the proportion of graphene islands bound to substrate step edges increases.

The density and size of the graphene islands vary because of a thermally

induced ripening process. Considering the strength of the C–C bond in

graphene, ostwald ripening, i.e. the preferential growth of large islands at

the expense of the dissolution of smaller ones through their higher pressure

of carbon adatoms, cannot account for the observed efcient ripening below

1300°C. The presence of large islands with irregularly shaped (though zigzag)

edges, exhibiting large vacancies, for growth temperature of 850°C during

(a)

(d)

(b)

(e)

(c)

(f)

10.8 (a–e) STM topographs of typical carbidic (below 600°C) and

graphene (higher temperatures) islands on Ir(111) as a function of

the growth temperature: (a) 600°C (3.2 ¥ 3.2 nm

), (b) 700°C (6.2 ¥

6.2 nm

), (c) 850°C (18 ¥ 18 nm

), (d) 1050°C (32 ¥ 32 nm

), and (e)

1200°C (64 × 64 nm

). The size of the white box in (b–e) corresponds

to the size of (a) (from Coraux, 2009, © Institute of Physics). (f) STM

topograph (73 ¥ 73 nm

) of a graphene island grown on Ir(111) and

presumably formed upon the coalescence of smaller islands.

ThinFilm-Zexian-10.indd 241 7/1/11 9:42:54 AM



242 Thin ﬁlm growth

a few seconds (Fig. 10.8f), suggests that large islands are formed upon the

coalescence of smaller ones that are mobile at the surface (Coraux, 2009).

This process is known as smoluchowski ripening. It is probably effective

due to the incommensuracy of the graphene and metal lattices. For a strictly

incommensurate island the activation energy for lateral motion would vanish,

as for every carbon moving away from its optimum binding site, another

one nds it. The existence of a migration barrier is probably linked to the

stronger binding of the graphene island edge atoms to the substrate. as larger

graphene islands display sizes consisting of an integer number of moiré units

it is likely that the graphene edges bind to specic energetically preferred

substrate sites. Prolonged heating results in islands becoming more compact,

presumably due to efcient carbon mobility at graphene edges.

10.3.2 Growing plain graphene sheets

We now consider the situation where carbon atoms, whatever their source,

are provided at the hot metallic surface.

From graphene islands to plain sheets

Graphene islands preferentially bind to the substrate step edges. The size

(density) of the islands is the larger (lower) the higher the growth temperature.

This is evidence that graphene nucleation along the steps is homogeneous

and not at a specic defect site. The nucleation at step edges was observed

by STM (Fig. 10.9a) and low-energy electron microscopy (LEEM) on Ir(111)

(Coraux, 2009), Ru(0001) (Sutter, 2008), and Pt(111) (Sutter, 2009a), in a

wide range of growth temperatures, and for different types of carbon sources.

The graphene islands nucleate preferentially at the ascending side of the step

edge. as discussed in the next subsection, this is presumably promoted by the

fact that carbon atoms attach via s bonds to the ascending substrate step.

Providing more carbon adatoms at the surface results in the growth of

the islands. This growth is always anisotropic, faster at the lower terrace

than at the upper one. For graphene on Ru(0001), there is almost no growth

at the upper terrace (Sutter, 2008). On Ir(111) and Pt(111) (Coraux, 2009;

Loginova, 2009b; Sutter, 2009a) the onset of growth is delayed (Fig. 10.9b).

Probably it is delayed by an energetically costly transient state, for which the

carbon–metal bond is broken before a carbon–carbon bond is formed across

the substrate step edge. The growth of graphene over steps does not interrupt

the graphene lattice. The carbon rows are continuous and bent vertically

to adapt to the the terrace ow (Pan, 2007; Coraux, 2008); in other words,

graphene rests like a blanket on the substrate.

Growing islands will eventually meet. sTM showed that the moiré

displays continuous rows through the location where the islands coalesced,

ThinFilm-Zexian-10.indd 242 7/1/11 9:42:54 AM



243Epitaxial growth of graphene thin ﬁlms on single crystal metal surfaces

which is evidence that the carbon rows are also continuous (Coraux, 2009).

This observation is puzzling at rst glance, as there is no reason why the

graphene lattices from the two lattices would be in registry. rather we would

expect a shift by a fraction of a lattice constant. lattice distortions, which

could be adapted either locally or at a larger scale, are therefore expected.

Dislocations (heptagon-pentagon pairs) are also found where the islands met

to accommodate for small tilts between the islands.

Prolonged growth will result in a lateral expansion of the graphene akes

which will meet and eventually cover the whole metallic surface. This is

discussed in more detail on page 245.

Graphene interaction with the substrate step edges

The nucleation of graphene at substrate step edges and the partial or total

hindering of uphill growth over step edges are evidence for the interaction

of graphene edges with substrate steps. This interaction is probably lowering

(a)

(b)

= down

20 µm

Down

0 100 200 300

Time (s)

Rate (arb. units)

1.0

0.8

0.6

0.4

0.2

0.0

10.9 (a) Differentiated STM topograph (1 ¥ 1 mm

) of graphene

islands nucleated at Ir(111) atomic step edges (the islands are

highlighted in dark and light gray if nucleated at the upper or lower

terrace respectively) (from Coraux, 2009, © Institute of Physics). (b)

Contours of a growing graphene island on Pt(111) at the vicinity of

an atomic substrate step edge, as a function of the growth time, as

images with LEEM (top panel) and growth rate derived from such a

sequence of images at the upper and lower terrace around the step

edge (from Sutter, 2009a, © The American Physical Society, http://

prb.aps.org/abstract/PRB/v80/i24/e245411).

ThinFilm-Zexian-10.indd 243 7/1/11 9:42:55 AM



244 Thin ﬁlm growth

the edge energies of both graphene and the metal by increasing the low

coordination of edge atoms.

Graphene growth was also observed to induce reshaping of the substrate

step edges. The magnitude of this reshaping varies. On Ir(111) or Ru(0001)

(Coraux, 2009; Loginova, 2010), it is in the form of a reorientation of the

metal step edge that is in contact with one of the graphene island edges, so

that the metal edge transforms into segments aligning the zigzag direction

of graphene (Fig. 10.10a). On Pt(111) and Pd(111), the effects are much

stronger: a whole graphene island can penetrate the metal terraces so that all

graphene edges are bound to metal edges, with the latter aligning graphene’s

zigzag edges (Nakagoe, 2002; Fujita, 2005; Land, 1992; Kwon, 2009) (Fig.

10.10b). Such processes can only be effective if ow of metal atoms develops

at the step edges. The preferential orientation of the reshaped metal edges is

probably driven by maximization of the number of carbon–metal s-bonds.

So far only in the cases of Ni(111) (Helveg, 2004) and Ru(0001) (Loginova,

2010) supports has it been shown that a growth front of graphene is possible

also at the location where it is attached to a metallic step edge. accordingly

graphene grows at the expense of a retraction of the substrate step edge

(Fig. 10.11). It was argued (Loginova, 2010) that the metal atoms are then

displaced away from the step edge and can incorporate in the topmost metal

layer, causing the formation of a dislocation network pattern in this layer.

Metal step edge retraction may also result from the equilibrium between the

step edge and the 2D metal gas surrounding it at the growth temperature,

allowing for an efcient exchange between step atoms and adatoms.

(a)

(b)

10.10 STM topographs evidencing the interaction of metal step

edges upon graphene growth, on Ir(111) (a) and on Pd(111) (b) (from

Kwon, 2009, © The American Chemical Society). The superstructure

is the moiré of graphene/Ir(111) and graphene/Pd(111). The inset in

(a) highlights the structure of the substrate step edge reshaped by

the graphene island. The interaction of graphene with Pd(111) steps

is such that the graphene island is incorporated into a Pd terrace.

Image sizes are 125 ¥ 125 nm

for (a) and 380 ¥ 250 nm

for (b).

ThinFilm-Zexian-10.indd 244 7/1/11 9:42:58 AM



245Epitaxial growth of graphene thin ﬁlms on single crystal metal surfaces

Elementary processes during the growth

as we already discussed, graphene growth starts at the step edges. It was

shown that the amount of graphene does not depend on the size of the terraces

surrounding the step edges (Coraux, 2009). In addition, at steps graphene

islands nucleate with random separations giving rise to earlier coalescence.

These observations indicate the absence of any signicant gradients induced

in the concentration of the adspecies from which the islands grow. Graphene

island growth is not diffusion limited.

on the contrary, it was proposed that the growth is interface limited

because carbon adatoms experience a large energy barrier for attaching to

graphene step edges (Loginova, 2008). A large carbon supersaturation is

needed at the surface so that several atom carbon clusters, being mobile at

the surface, can be formed. such clusters have been calculated to possess a

much lower energy barrier for attachment at graphene edges, and are thus

considered as building blocks for graphene growth. They are as large as

heptamers on Ru(0001) (Loginova, 2008) and Ir(111) (Loginova, 2009b).

These results are qualitatively very similar for two different carbon sources,

CVD with ethylene or atomic carbon ux.

The analysis of the surface coverage with graphene as a function of the

carbon dose provides information about the stability of carbon adspecies

at the bare metal surfaces and regions covered with graphene. For growth

on Ir(111), it was shown that the experimental data are well reproduced by

a mere exponential law (Fig. 10.12), with a rapid increase at initial stages

followed by a slowing down of the growth rate towards zero as full coverage is

approached (Coraux, 2009). Quantitative analysis showed that the exponential

behaviour is well reproduced by assuming (i) the absence of desorption of

carbon adspecies from the Ir surface areas, (ii) the absence of C incorporation

to the bulk and (iii) no sticking of molecules or C adspecies to the graphene

areas. Therefore, under these conditions graphene growth is self-limiting

(a) (b) (c)

10.11 In situ observation of carbon nanotube growth at the surface of

a catalytic (Ni) particle with transmission electron microscopy, where

methane is provided with increasing doses from (a) to (c). Scale

bar is 5 nm. The schematics highlight the graphene growth at the

expense of the retraction of the substrate step edges (from Helveg,

ThinFilm-Zexian-10.indd 245 7/1/11 9:42:58 AM



246 Thin ﬁlm growth

and terminates upon monolayer completion. In the opposite situation where

the bulk of the support has the capability to store carbon and supply mobile

species, like it is the case from 1500°C on Ir(111) or 800–900°C on metals

like Ni or ru, the graphene growth is not self-limiting. a second and even

more layers may form underneath the monolayer graphene.

On the formation and stability of rotational variants

Graphene rotational variants are found at Ir(111) and Pt(111) surfaces (see

Section 10.2.3), and will presumably also be found at the surface of other

metals weakly interacting with graphene, typically those for which the average

graphene–metal distance is close to that between planes in graphite. On Ir(111)

the variant with graphene zigzag edges aligned to Ir dense-packed rows

grows is energetically preferred and dominates the growth; at high growth

temperatures and uxes, other variants may nucleate at its edges (Loginova,

2009b). For graphene on Pt(111) the different variants nucleate independently

and at the same time (Sutter, 2009a). On Ir(111), it was suggested that the

nucleation of the other variants is heterogeneous, taking place at defects

at the rst variant edges. Once the other variants are nucleated, they grow

faster than the rst variant. It was argued that this is due to different carbon

attachment kinetics at graphene edges for the different variants, as carbon

attachment is the limiting step in graphene growth (Lovinova, 2009b).

The growth of other variants can be suppressed if the development of

graphene edges not aligned to the dense-packed rows of the substrate can

be avoided, or at least largely postponed. For this, a fraction of the surface

Total amount

Graphene coverage (%)

100

0 1¥10

–7

2¥10

–7

3¥10

–7

Ethene dose (mbar.s)

10.12 Graphene coverage as a function of the ethene dose during

CVD growth of graphene on Ir(111), as estimated from STM

Institute of Physics).

ThinFilm-Zexian-10.indd 246 7/1/11 9:42:58 AM



247Epitaxial growth of graphene thin ﬁlms on single crystal metal surfaces

might rst be covered by graphene islands having well-dened zigzag edges

following the recipe described in Section 10.3.2 (pre-adsorption of ethene at

room temperature followed by high temperature annealing). accordingly a

high density of islands is achieved, and the growth can be continued by CVD

at high temperature, yielding exclusively the more stable and energetically

preferred variant (van Gastel, 2009). In practice, the rst step may be

performed at 1200°C and the second at 800°C, which results in a single

crystallographic orientation of graphene across the whole sample surface. The

as-grown graphene has a low density of wrinkles (see Section 10.2.4) due

to the low growth temperature of 800°C during graphene layer completion.

The lower growth temperature induces less thermal lattice mismatch during

cooling and consequently less strain relieving defects.

The different variants have distinct stabilities, as suggested by the

preferential high temperature oxygen etching of the graphene variants of

which the zigzag edges do not align to dense-packed rows in the metal (van

Gastel, 2009; Starodub, 2010). This difference in reactivity was employed as

an alternative route to achieve macroscopic graphene samples with a single

orientation on Ir(111), via cycles of CVD growth with ethylene and oxygen

etching of the undesired variants (van Gastel, 2009) (Fig. 10.13).

10.3.3 Graphene multilayers on metals

Graphene multilayers are commonly obtained on metals which can store a

non-negligible amount of carbon in their bulk. This is the case for instance

for Ru, Ni or Co. Relatively slow cooling rates (typically 10°C/s, but this

gure depends on the amount of C stored in the metal) are usually employed

to promote the diffusion of carbon towards the bulk of the metal, leaving

only a limited amount of carbon close to the metal surface, which favours the

growth of few-layer and even single layer graphene (Yu, 2008; Reina, 2009b).

Too slow cooling rates leave too much time for carbon to diffuse towards

the metal bulk, resulting in negligible amounts of carbon near the surface

and accordingly no graphene growth by segregation. on the contrary, fast

cooling rates inhibit carbon diffusion towards the bulk, so that large amounts

of carbon are available close to the surface of the metal: this favours the

growth of multilayer and defective graphene. leeM showed that an additional

graphene layer starts to grow after each layer is completed (Sutter, 2009b,

2009c). So far the question of how the second layer of graphene grows after

the rst is completed remains open: on Pt(111) it was shown that the second

layer does not grow between the rst layer and the topmost Pt layer (Sutter,

2009a), pushing the rst layer upwards, which suggests that carbon atoms

are escaping the bulk towards the graphene surface via defects (e.g. holes)

present in the rst graphene layer. LEEM studies showed that the graphene

growth by segregation is mostly a bulk-diffusion limited process (McCarty,

ThinFilm-Zexian-10.indd 247 7/1/11 9:42:58 AM

