Magic clusters on group IV surfaces 4

The formation of the type “A” clusters on the surface is thus likely to arise from these individual tetra-clusters “popping” out from the 3x3 reconstruction and as a consequence, creatin

CHAPTER 4: Role of Si magic clusters in the phase

transformation of 6H-SiC(0001):

(3x3) →(6x6) clusters→(6x6) rings

This chapter discusses the data obtained from STM and XPS which was used to probe the local atomic structure of the 6H-SiC(0001) surface in UHV The evolution of surface structures is observed with progressive annealing; starting from (3x3) phase at

850oC → Si rich (6x6) clusters at 1000oC → (6x6) rings at 1100oC This chapter will also include STM analysis of the various surface features such as defects, tetramer formation as well as tetramer agglomeration leading to the formation of the various surface structures The data is co-related with XPS information in order to propose a mechanism for the observed phase transformation

4.1 Global morphology of 6H-SiC(0001): STM and XPS

The SiC(0001) sample is prepared in–situ and ex-situ as described in Chapter 3

before being annealed for 30min at various temperatures and scanned correspondingly using the STM and XPS This section focuses on the global morphological evolution as the surface is progressively annealed

Figure 4.1a shows the STM image of the surface after initial deposition of Si at room temperature The surface is characterized by three-dimensional islands, which appear to be round in shape, and possesses an average island size and height of 50±10.0nm and 20.0±5.0nm respectively On annealing the surface to 800°C for 30min (Figure 4.1b), the island density decreases while the average island size increases Flat terraces as shown in Figure 4.1c are beginning to form when the surface was further annealed to 850°C At this stage, annealing the surface progressive to higher temperatures

of 900°C (Figure 4.1d), 950°C (Figure 4.1e), 1000°C (Figure 4.1f) and 1050°C (Figure 4.1g) results in the formation of increasingly larger terraces respectively STM line profile analysis across adjacent terraces reveals the average step height to be in multiples

of 15Å, which is the height of a 6H-SiC structure in the c-direction This indicates formation of an ordered surface with progressive annealing

Figure 4.1: A series of 1000nmx1000nm STM images of the global surface morphology after Si was deposited at (a) room temperature and after annealing to temperatures of (b)800°C, (c)850°C, (d)900°C, (e)950°C, (f)1000°C and (g) 1050°C respectively Zoom

in 50nmx50nm images of (c), (f) and (g) shows (c)(i) (3x3) reconstruction, (f)(i) clusters and (g)(i) (6x6)-ring-like structures respectively

(6x6)-While the global morphology of the surface annealed from 850°C to 1050°C (Fig 4.1c-g) shows flat terraces and is not visibly different, the microstructure revealed by high resolution STM images (see Fig 4.1(c)(i), (f)(i) and (g)(i)) show the surface undergoing significant atomic re-arrangement from the Fig 4.1(c)(i)(3x3) phase → (f)(i)

clusters (6x6) → (g)(i) ring-like (6x6) structure In the following sections we will discuss the (3x3) reconstruction first observed at 850oC followed by the self assembly of (6x6) clusters at 1000oC leading to the eventual formation of the ring-like (6x6) structure at

1050oC Figure 4.2 shows the XPS trend from the initial deposition of Si up to annealing

at 1050°C corresponding to the surface morphology shown in Figure 4.1

Figure 4.2: The evolution of the Si 2p XPS peak (i.e elemental Si component at 99.3eV and the Si-C component at 103.0eV) after annealing the substrate to different temperatures

Elemental Si at 99.3eV

The XPS spectrum of the Si 2p peak consists of elemental Si and Si-C signals at 99.3eV and 103.0eV respectively The peaks were normalized with respect to the Si-C signal from the as-received wafer It is evident that with progressive annealing to higher temperatures, the Si-C signal increases while the elemental Si signal decreases The Si-C content is thus greater due to the increasing presence of Si-C bonding at the surface This trend could be a consequence of excess Si desorbing from the surface and/or a reaction between the deposited Si and bulk C to seed the growth of 6H-SiC(0001) polytype Both processes would lead to the formation of flat terraces, which have step heights in multiples of 15Å As it is well known that annealing at elevated temperatures (>1200oC) often induces graphitization, the desorption of Si is expected to eventually dominate It should be noted in the present study that no graphitization was detected and the surfaces observed are still in the Si rich regime

4.2 6H-SiC(0001)-(3x3) reconstruction

STM images of the SiC(0001) surface after annealing at 850°C for 30min are shown in Figures 4.3a and 4.3b below The images obtained are of the same surface area

at albeit at different sample bias of +3.0V and -3.0V for Figure 4.3a and 4.3b respectively

Figure 4.3: A 40nmx40nm STM scan of the SiC(0001)-3x3 surface at sample bias of (a) +3.0V and (b) -3.0V Inset figures show the (3x3) surface at higher magnification (12nmx12nm) in both respective biases Line profiles illustrate the 3x periodicity of the surface

The (3x3) reconstruction was imaged as bright protrusions in the empty state and

as small dark depressions or bright rings in the filled state The difference in appearance

is further highlighted in the respective inset figures The separation between each adjacent bright protrusion or between neighboring dark depressions measured along the [2110] and [1210] azimuths is ~ 9.0Å This periodic arrangement is 3x the SiC(0001)-(1x1) unit cell (lattice parameter =3.08Å) These observations may be explained using the

(3x3) structural model proposed by J Schardt et al [1]

Figure 4.4b shows the ball and stick models corresponding to the plane view and side view of the surface reconstruction The (3x3) reconstructed surface is believed to result from arrangement of Si-tetra clusters These Si tetra-clusters are described as having one Si adatom sitting above a Si trimer structure in the second layer The Si tetra cluster is positioned above a full Si adlayer which is arranged on top of the Si terminated 6H-SiC(0001) substrate The (3x3) reconstruction as shown in Figure 4.4a would consist

of one tetra-cluster per unit cell All the Si atoms are four fold coordinated except for the

top Si adatom in the tetra-cluster, which has a single dangling bond

The bright protrusions observed in empty state imaging on the (3x3) surface could therefore be attributed to tunneling into the empty states associated with the single dangling bond on each of the top Si adatom The observation of the rings with depressions in the center under filled state imaging would therefore be attributed to contribution from the back bonding of the 2nd layer Si trimer and third layer Si adlayers

Figure 4.4: (a) A 8nmx8nm STM image of (3x3) surface structure after annealing at 850°C Line profiles analysis shows that the (3x3) tetra clusters are about 5.0±0.5Å in size and 1.5 ± 0.5Å in height (b) A plane and side view of the ball and stick model illustrating the (3x3) surface structure [1]

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

First layer: Si adatom with one dangling bond

Second layer: Si trimers

Third layer: Si adlayer in (1x1) position

Fourth layer: Si atoms in top layer of bulk

Fifth layer: C atoms

I II

III IV

[ ] 2 10

[ 2110 ]

(a)

(b)

Figure 4.4a shows a high-resolution 8nmx8nm STM picture of the (3x3) reconstruction Line profiles (i) and (ii) as indicated in the STM image measures the average full width half maximum size of a protrusion in the (3x3) unit cell to be ~ 5.0 ± 0.5Å while line profile (iii) shows the average height of the protrusions to be ~ 1.5 ± 0.5Å These average scan sizes were obtained from the line profile measurements of the same features scanned under different biases in 3 different directions The plane view of the surface as illustrated in Figure 4.4b shows the structure with the Si-adlayer atoms occupying bulk SiC positions and the Si atoms within the tetra-cluster having a Si-Si bond lengths of ~2.3Å [1] In this configuration, the Si tetra-cluster would have a diameter 5.0 ± 0.5Å and a height of 1.5 ± 0.2Å This size is equivalent to the measured average size of each protrusion occupied by one Si tetra-cluster in the (3x3) unit cell as labeled in Figure 4.4a

4.3 6H-SiC(0001)-(6x6) clusters

4.3.1 Formation of (6x6) clusters from the (3x3) reconstruction

This section discusses STM data of the 6H-SiC(0001) surface evolution as the starting (3x3) surface is progressively annealed from 850ºC to 1050°C A series of ball and stick diagrams are used to illustrate the formation of the various surface features at the corresponding annealing temperatures

Figure 4.5a(i) shows a 40nmx40nm STM image of the (3x3) surface after annealing to 850ºC for 10 minutes The surface shows long range ordering of (3x3) reconstruction on the surface with low defect density At higher resolution (Figure 4.5a(ii)), the STM data shows the appearance of recessed protrusions existing amongst the bright protrusions, previously attributed to the Si tetra-clusters Line profile measurements of these recessed protrusions reveal an average depth of only ~ 0.6 ± 0.2Å The recessed protrusion is likely to be caused by a missing Si adatom from the tetra-cluster, as illustrated in the corresponding ball and stick diagram shown in Figure 4.5a(iii) The remaining underlying 3 Si atoms exposed would therefore contribute to the

“shallow” appearance in the STM image These defects are identified as “shallow” holes

Subsequent annealing of this surface to 900ºC results in the formation of two distinctive defects Figure 4.5b(i) shows that the surface is characterized by regions of depressions (dark features) and clusters (bright features) coexisting with the (3x3) reconstruction The bright clusters, which are identified as “A”, are observed to be round

in shape Analysis of the “A” type clusters in Figure 4.5b(ii) shows the average size and height to be ~ 5.0 ± 0.5Å and 1.5 ± 0.5Å respectively, while the deep holes (labeled as C) possess average sizes of 9.0±0.5Å and depths of ~ 2.0 ± 0.5Å The dimension of the clusters and the “deep” holes are thus comparable These defects appear to be real and are not due to electronic effects, as STM biasing does not reveal any changes to the feature sizes of these defects In addition, these “A” clusters are usually sited adjacent to the dark vacancies It is also interesting to note that the size of the clusters is similar in shape and dimension to a (3x3) tetra-cluster The formation of the type “A” clusters on the surface

is thus likely to arise from these individual tetra-clusters “popping” out from the (3x3) reconstruction and as a consequence, creating “deep” holes as illustrated in the corresponding ball and stick diagram shown in Figure 4.5b(iii)

Further annealing to 950ºC (Figure 4.5c(i)) shows that there are now more depressions and clusters on the surface Closer analysis of these defects (Figure 4.5c(ii)) shows that a second species of clusters, termed as “B”, are now observed along with the

“A” type clusters The size of the depressions is now also larger than before These new

“B” clusters are similarly round in shape and they are also visibly larger than the “A” type Line profile analysis shows that these randomly arranged “B” clusters have an average size of 14.3±0.5Å and a height of 2.3±0.2Å As previous STM evidence pointed towards the creation of the “A” clusters through the ejection of whole (3x3) tetra-clusters, the observation of the larger “B” clusters after annealing at higher temperatures could be attributed to the agglomeration of a few of the “A” clusters together to form “B” clusters

as shown in Figure 4.5c(iii).At this stage, no ordering of clusters was observed even after several hours of cooling in the STM stage

Figure 4.5: Shows the surface after it was annealed to (a) 850°C, (b) 900°C and (c) 950°C respectively at a scan size of (i) 40nmx40nm and (ii) 20nmx20nm The corresponding structural model is given by (iii) Clusters and deep holes labeled as “A” and “C” respectively are clearly observed when the sample was annealed to 900°C Line profiles show the average size of an “A” cluster to ~ 5.0±0.5Å and height of ~ 1.5±0.2Å while the average size of a deep hole “C” is ~9.0±0.5Å at a depth of ~2.0±0.2Å Annealing to

950oC, leads to observation of type “B” clusters Line profile shows the average “B” cluster size to be ~14.3±0.5Å and height to be ~2.3±0.2Å

B Disordered (3x3)

Figure 4.6: Shows the STM images (40nmx40nm) of the surface (a) immediately after annealing to 1000°C, (b) after cooling down for 60mins and (c) further cooling for another 10mins or longer A greater number of disordered “B” clusters are now observed compared to Fig5(c) Type “B” clusters were observed to exhibit a (6x6) ordering on the surface in larger domains existing near step edges Inset shows a 12nmx12nm zoom-in image of the ordered clusters Line profiles (i) and (ii) show that the average separation between clusters is typically ~18.5±0.2Å in both [2110] and [1210] directions Line profiles (iii) and (iv) show that clusters possess an average diameter of ~14.3±0.5Å and

an average height of ~2.3±0.2Å

Figure 4.6a shows the surface imaged immediately after quenching from 1000ºC The previously observed (3x3) long range ordering is no longer evident at this stage Instead, the surface appears to be highly disordered The larger “B” type clusters are now present in greater number density and hence, are noticeably more prevalent in comparison to the almost absent “A” type clusters This observation further supports the agglomeration of “A” clusters together to form the “B” type clusters

Figure 4.6b shows the same surface after cooling down for 60mins The previously disordered “B” type clusters are now observed to organize into small domains consisting of six to eight clusters Within each domain, these clusters are observed to have taken on a hexagonal arrangement with uniform cluster separation.Further study of the image reveals that disordered regions of the (3x3) phase, as indicated in Figure 4.6b, are also observed to co-exist with these ordered cluster domains Cooling of the surface for another 10mins or longer only revealed the formation of larger ordered domains of these type “B” clusters

These cluster domains are also observed to be predominantly located near the step edges as shown in Figure 4.6c at higher resolution Line profile analysis given by (i) and

(ii) shows the average cluster separation to be ~18.5±1.0Å in both the [2110] and [1210] azimuths This arrangement of clusters thus gives rise to a (6x6) ordering on the 6H-SiC(0001) surface We confirm that these are indeed B type clusters by comparing the size and heights obtained from line profile analysis annotated for example by (iii) and (iv) The clusters have an average diameter of 14.3±0.5Å and heights of 2.3±0.2Å These dimensions are similar to the larger clusters first observed in Figure 4.5c In general, the determination of the size of clusters were obtained as an average of the distance between opposite fringes of the bright cluster protrusions in three crystallographic directions ([2110], [1010] and [1210]), while the height was obtained from an average of the difference between the maxima of the cluster and its neighboring minima These topography measurements (i.e sizes and heights) were also obtained by performing analysis of the STM data collected at different biases (ranging from +0.1V to +4.0V) and tunneling currents (0.05nA to 1.00nA) This is to reduce any electronic effects associated with changes with the electron density distribution In our experiments, no changes in the features were observed

It is interesting to observe that these “B” type clusters do not increase in size beyond the dimension of 14.3±0.5Å This size and shape are in fact similar to the Si

“magic” clusters reported by Tsong et al [2-3] The “magic” Si clusters which were

studied on the Si(111)-(7x7) surface, however, did not exhibit any ordering The implication of the above results is the occurrence of a critical or “magic” Si cluster size

on the 6H-SiC(0001) surface

When this surface is further annealed to 1050°C, a ring-like structure is observed

to form (see Figure 4.1(g)(i) inset) At this stage, we failed to detect any presence of

these “B” clusters The rings are observed to exist in large domains of up to ~100nm in

size and appear to be aligned also in the [2110] and [1210] azimuths of the substrate Line profile analysis shows a separation of 18.5±0.2Å between neighboring rings in both directions This translates into a 6x periodicity for the observed surface structure with respect to the SiC(0001)-(1x1) unit cell The formation of these rings as well as the atomic structure will be discussed in the subsequent sections However, it should be noted that this surface, as revealed by XPS analysis (Figure 4.2), is still Si-rich albeit less than the initial (3x3) surface

We now summarize our STM results: heating the surface from 900ºC to 1000ºC results in the disordering of the original (3x3) phase and encourages the formation of round “A” clusters These clusters are similar to the (3x3) tetra-clusters and appear to be ejected from the reconstruction during heating They appear to combine and form the larger type “B” clusters These type “B” clusters organize into (6x6) regular arrays upon cooling from 1000ºC The observation of the different defects such as the shallow and deep holes, “A” and “B” type clusters coexisting initially with the (3x3) reconstruction suggests that the (3x3) surface undergoes structural re-arrangement involving not only

the top layer Si tetra-clusters but also the Si adlayer With the aid of J Schardt et al’s

model for the (3x3) reconstruction [1], we will now proceed to propose a structural model

to account for the type “B” cluster size, the (6x6) periodicity and the observed structural transformation

4.3.2 The (6x6) Si rich magic cluster structure

Referring to the plane view model of the (3x3) surface (Figure 4.4(b)), four types

of Si atom bonding situation within the Si adlayer can be identified [1] These atoms are labeled in the figure as type-“I” or “II”, “III” and “IV” The type-“I” or “II” atoms have four bonds of which one is bonded down directly to the Si atom in the bulk position of the substrate The remaining three bonds are in turn bonded to three other neighboring Si atoms through sp2–like co-planar bonds, leading to a twisted bonding configuration The type-“I” and “II” bonding configurations are similar and they differ only in orientation The type “III” atoms are also bonded directly down to the Si atom in the bulk position of the substrate While one bond is linked to the tetra-cluster structure, the remaining two bonds are bonded co-planarly to neighboring Si atoms

Each tetra-cluster is supported by 6 of these type-“III” atoms The final type-“IV”

Si atom sits below the tetra-cluster, occupying the position directly beneath the top Si adatom Unlike the type-“I”, “II” or “III” atoms, the type-“IV” atom has the closest sp3character, with three of its bonds supporting the Si tetra-cluster above, while the remaining one bonds to a Si atom below Hence the type-“IV” atoms are most similar to bulk Si crystal atoms The Si atoms making up tetra-cluster are described to adopt a tetrahedral sp3 bonding configuration [1]

If we analyze the (3x3) unit cell, there would be two type-“I /II”, six type-“III” and one type-“IV” Si atom in the Si adlayer Consequently, this translates into only ~11%

of the Si adlayer atoms possessing close to ideal type “IV” of bonding configuration This would suggest that the Si adlayer is highly strained due to the immense twisting and stretching of the co-planar Si-Si bonds of type-“I”, “II” and “III” atoms Furthermore, the 20% difference in the Si-Si and Si-C bond lengths also creates a further mis-match, resulting in a stress field inherent in the (3x3) reconstruction, which exists predominantly

at the Si/SiC interface Therefore it is not surprising to expect a high strain energy associated with the SiC(0001)-(3x3) surface Although the final (3x3) structure appears to

be driven by optimizing dangling bond saturation through the reduction of nine dangling bonds per (3x3) unit cell to one [1], the release of this inherent strain may encourage further structural changes to occur especially when the substrate is heated to high temperatures In a tetrahedral bonding geometry, it is reasonable to deduce that atoms with a less than ideal sp3 configuration (i.e type-“I”, “II” and “III” Si atoms) are more susceptible to breaking and rearrangement than the type “IV” or the tetra-cluster Si atoms

In this section, we describe the evolution of the surface structure that may be expected when the substrate is heated to 850oC and above Figure 4.7a shows a ball and stick model of the (3x3) reconstruction Six tetra-clusters A1 to A6 are identified The underlying Si adlayer atoms identified as type-“I” and “II” Si atoms are assigned as atom-

1 to atom-9 respectively, while type-“III” Si atoms supporting the six tetra-clusters are labeled as atom-10 to atom-37 respectively

Figure 4.7(a): The ball and stick model of the initial (3x3) reconstruction

To account for the STM observation of the 6x periodicity observed in the [1210] and [2110] azimuths, we selectively move three clusters A2, A4 and A6 (or A1, A3 and A5) tetra-clusters from the (3x3) reconstruction as shown Figure 4.7a However, this process alone cannot account for the observation of larger type “B” cluster In order to describe a cluster with a diameter of 14.3Å would require the agglomeration of tetra-clusters on the surface Preserving the integrity of a tetra-cluster (i.e size of ~5Å and height of ~1.5Å) and using it as a basic building block, the A2, A4 and A6 tetra-clusters

A1

A3

A5

16 17

18 19 20 21

23 24

25

26 27

22 29

A6

A2

A4

30 31 32 33 28 8 9

12 13 14 15 10 1 11

3

5 35

2 37

36

34 7

will need to occupy positions directly above atoms labeled as 3, 5 and 9 respectively, as shown in Figure 4.7b

Figure 4.7(b): Movement of tetra-clusters A2, A4 and A6 with corresponding bond breaking and formation to form a proposed intermediary structure A dotted circle drawn

to the size of a measured cluster size (~14.3±0.5Å) is superimposed on the on this proposed structure

The detailed bond re-arrangement shown in Figure 4.7a and 4.7b is as follows; the

three tetra-clusters A2, A4 and A6, would be moved to occupy atom positions 3, 5 and 9

If we take A6 as an example, each cluster would have to break six type-“III” bonds

A1

A3

A5

16 17

18 19 20 21

23 24

25

26 27

22 29

A6

A4

30 31 32 33 28

8 12

13 14 15 10 1 11

35

2 37 36

34 7

~14.3±0.5Å

(atoms 10 to 15) and three type-“IV” bonds To accommodate A6 in its new position, three type-“III” bonds (i.e between atoms-23 and 16, 21 and 12 as well as 24 and 11), and three type-“I”/”II” bonds (atom-9) would be broken per cluster Hence a total of 9 type-“I”/”II” bonds, 27 type-“III” bonds and 9-type “IV” bonds would be broken in the process To anchor the tetra-cluster A6 down, three type-“I”/”II” bonds, six type-“III” bonds (atoms-10 to 15) and three type-“IV” bonds (atom-9) would have to be formed Consequently, each of the atoms labeled 3, 5 and 9 would now have sp3 type-“IV” of tetrahedral bonding to the tetra-clusters sitting directly above instead of the previous type-“I”/”II” bonding However, in the wake of the tetra-cluster departure, the dangling bonds created from the now exposed type-“IV” atoms are likely to be saturated as well This could be achieved through bonding with neighboring type-“III” atoms and hence result in the formation of three type “I”/”II” bonds (in the case of A6; atom-38 forms type

“I”/”II” bonds with atoms 11, 13 and 15) and three new bonds (i.e between atoms 11 and

12, 13 and 14 and 10 and 15 respectively) which are less type “III”-like In summing the net bond count, a total count of 9 type “I”/”II”, 27 type “III” and 9 type “IV” bonds are being broken comparing 9 type “I”/”II”, 18 type “III” and 9 type “IV” bonds being formed in this structural transformation Hence, there would be difference of 9 less type

“III” bonds after the bond re-structuring

Superimposed onto Figure 4.7b is a dotted circle representing the actual measured diameter of a “B” cluster (14.3±0.5Å) from STM The size of the proposed model, which

is made up of six agglomerated Si tetra-clusters, appears to fit within the measured diameter Moving the tetra clusters (A2, A4 and A6) to these new positions, however,

would require bonds to be broken and new bonds to be formed in order to anchor the new structure This atomic re-arrangement of atoms and tetra-clusters will involve breaking a total of 9 type-“I”/”II” bonds, 27 type-“III” bonds and 9 type-“IV” bonds while a total of

9 type “I”/”II” bonds, 18 type “III” bonds and 3 type “IV” are re-formed [see Section 4.2.3] The new surface structure would now have 9 less type “III” bonds

However, the resultant ring-like appearance does not fit the STM description of the “B” cluster Hence we propose that a seventh tetra-cluster be added to the centre of the ring, as shown in Figure 4.7c This addition would allow the breaking of 3 more type-

“I”/”II” bonds which further reduces the “sp2 character” within the Si adlayer and thus leading to a structure with a lower strain energy associated However, a total of seven dangling bonds per (6x6) unit cell would be created as a consequence This increase in dangling bond density would make this structure energetically not as favorable compared

to initial (3x3) structure, which would have an equivalent of four dangling bonds per (6x6) unit cell As previous STM analysis also shows that the “B” clusters have an average height of ~ 2.3±0.2Å, the structure in Figure 4.7c would not fit this dimension as it is likely to be only ~1.5±0.2Å in height Therefore, we propose that an additional tetra-cluster is required, which will sit on top of this seven tetra-cluster structure to complete the (6x6) cluster model as shown in Figure 4.7d

Figure 4.7(c): The (6x6) cluster base structure incorporating a centre tetra-cluster.

A1

A3

A5

16 17

18 19 20 21

23 24

25

26 27

22 29

A6

A4

30 31 32 33 28 12

13 14 15 10 1 11

35

2 37 36

34 7

Figure 4.7(d): The formation of the (6x6) cluster base structure including the top inverted

Si tetra-cluster with respect to the SiC(0001)-(1x1) lattice, arranged along [2110] and [1210] azimuths

This top tetra-cluster is inverted in orientation with respect to the surface and is bonded down to the top adatoms of the tetra-clusters A1 to A6 as well as the centre tetra-cluster directly below Although this bonding arrangement results in a “co-planar” tetra-cluster configuration, the number of dangling bonds per (6x6) unit cell is reduced to three, which is associated with the three topmost Si adatoms of the inverted(6x6) cluster From the perspective of dangling bonds minimization as a driving force, the proposed structure

24

25

26 27 22

A6

A4

30 31 32 33 28 12

13 14 15 10 1 11

35

2 37 36

34 7

is now more energetically favorable than before as it results in the formation of three dangling bonds per (6x6) unit cell instead of seven This model will also be able to account for the STM description of the (6x6) cluster as a bright protrusion with a height

of 2.3Å

Based on the size comparison obtained from STM analysis of the diameter and height of one (3x3) tetra-cluster and a (6x6) cluster at the same scanning bias, the (6x6) cluster structure is thus proposed to be made up of eight Si tetra-cluster sub-units Seven

of the tetra-clusters form a hexagonal cluster base possessing six fold symmetry An additional inverted Si tetra-cluster unit sits above this base structure, forming the top layer of the cluster in completing the (6x6) unit cell The clusters are assembled in a hexagonal (6x6) arrangement on top of a SiC(0001)-(1x1) surface Each unit cell of the (6x6) cluster would consist of 32 atoms The assembly and formation of these clusters will thus inevitably be accompanied by significant surface disordering, defect formation and eventual disruption of the (3x3) reconstruction as observed by STM

Thus the surface evolution from a (3x3) surface to a (6x6)-cluster surface is summarized as; (a) formation of (3x3) reconstruction consisting of Si tetra-clusters arranged on top of a full Si adlayer above Si-terminated SiC bulk at 850°C (b) ejection

of tetra-clusters onto surface due to breaking of highly strained co-planar Si-Si bonds at 900°C, resulting in shallow defects (c) agglomeration of 8 tetra-clusters to form larger type “B” clusters at 950°C in order to reduce dangling bond density and (d) self assembly

of type “B” clusters at 1000°C to form the (6x6) cluster phase as shown in the side view schematic illustrated in Fig 4.8

Figure 4.8 is a schematic side view showing (a) the (3x3) reconstruction formed at 850°C (b) ejection of a tetra-cluster (size~8.0Å height~1.5 Å) from the (3x3) reconstructed surface at 900°C leading to the formation of a shallow defect (c) agglomeration of 8 tetra clusters at 950°C to form a type “B” cluster of size ~ 14.3Å and height ~ 2.3 Å (d) the ordering of type “B” clusters at 1000°C to form the (6x6) cluster phase on the 6H-SiC(0001) surface

The ball and stick model proposed for the (6x6) structure represents an attempt to elucidate the structure from consideration of dangling bonds and comparison of sizes of clusters observed from STM measurements We understand that the experimental observations reported are confined within the constraints of the STM measurements, as the existence of artifacts or distortion due to electronic effects could affect the accuracy

of our size measurement Nevertheless, we took caution during our scans by applying a slow scan speed through a large bias range to eliminate drift and electronic effects in ensuring that our images are topographical representations of the surface features Clearly,

a first principle calculation of the (6x6) cluster surface will be useful in confirming the cluster structure Considering the large of number of atoms and different possible types of

bonding configurations, a theoretical approach similar to that taken by Que et al [4-5]

may be necessary in order to solve this structure efficiently

In this section, we investigate with STM, the SiC(0001)-(3x3) phase transition to Si-rich (6x6) reconstruction as a function of temperature We were able to observe the formation of (6x6) clusters from the (3x3) surface, after annealing at 1000°C High temperature annealing causes tetra-clusters to “pop-out” of the (3x3) surface reconstruction Agglomeration of these tetra-clusters on the surface occurs as a precursor

to the (6x6) cluster self-organization We have combined J Schardt et al’s (3x3) atomic

model as a basis for discussion with STM evidence to demonstrate the phase transition mechanism We have proposed a hexagonal cluster structure model consisting of 8 sub-units of tetra-clusters to explain the cluster size and the 6x-periodicity The process involving the ejection of silicon tetra-clusters, agglomeration of the tetra-clusters and subsequent self-assembly is driven by the formation of a structure with a lower strain energy and minimum number of unsatisfied dangling bonds In the following section, the same approach in analysis will be applied to study the formation of (6x6) ring structures from the (6x6) cluster structure

4.4 6H-SiC(0001)-(6x6) rings

4.4.1 Formation of (6x6) rings from (6x6) clusters: global morphology

In the previous section, the (3x3) surface reconstruction was observed to undergo disordering with the formation of type “A” (size~5.0Å) and “B” (size~14.3Å) clusters when the surface was heated from 850oC to 950oC This was superceded by the self-assembly of type “B” clusters into an ordered (6x6) arrangement with no type “A” clusters detected at 1000oC [6-7] In an effort to resolve the surface structures observed at higher temperatures, STM and XPS is used to study the surface structural evolution when the (6x6) cluster surface is heated beyond 1000oC In particular, we will show that; (i) disordering of the (6x6) cluster surface occurs with diminishing type “B” clusters density and the reemergence of type “A” clusters on the surface The surface is eventually dominated by type “C” Si adatoms (size~3.0Å) which decorate the ring structure at

1100oC This structural evolution appears to occur with the loss of Si from the surface; (ii)

By analyzing the unit cell dimensions, crystallographic azimuths and auto-correlation of the real space, we will show that this ring structure has a (6x6) and not (6√3x6√3) R30° unit cell; (iii) By incorporating the existence of Si as tetra clusters, we propose a ball and stick model to describe the structural transformation and the formation of the ring structure from the initial (3x3) reconstructed surface We will show that each structural phase observed can be accounted by moving or removing at least M x 4 Si atoms (where

M = Number of clusters) Figures 4.9(a)-(f) show the STM images (40nmx40nm) of the surface after annealing the 6H-SiC(0001) substrate to temperatures of 850°C, 900°C,

950°C, 1000°C, 1050°C and 1100°C respectively The inset images are higher resolution images of the corresponding surface obtained at a scan size of 16nm x 16nm

[2110]

[1120]

5nm1100°C

[2110]

40Å

5nm[1120]

Figure 4.9(a)-(f) shows a series of 40nmx40nm STM images of the surface morphology after annealing to temperatures of (a)850°C, (b)900°C, (c)950°C, (d)1000°C, (e) 1050°C and (f) 1100°C Corresponding inset images show high resolution 1.6nmx1.6nm STM pictures of (a) (3x3) reconstruction, (b) formation of type “A” from (3x3) and line profile

of type “A” (average size ~5Å, height~1.5 Å), (c) co-existence of type “A” and “B” clusters, (d) self assembly of type “B” clusters into (6x6) super-structure, (e) absence of type “B” leaving only type “A” co-existing with rings and (f) (6x6)-ring-like structures

dominating surface structure

Similar to the previous results, by annealing the substrate to a temperature of

850oC, we obtain the initial (3x3) surface structure (Fig 4.9(a)) At higher temperatures, a (6x6) cluster superstructure is obtained of 1000oC (Fig 4.9(d)), while formation of ring-like structures over wide terraces occurs at 1100oC (Fig 4.9(f)) In the temperature regime going from (3x3) (Fig 4.9(a)) to the (6x6) clusters phase (i.e Fig 4.9(c)), occurrence of two types of clusters with an average size of ~5.0 ± 0.5Å and ~14.3±0.5Å,

as well as an average height of ~1.5 ± 0.5Å and ~2.3±0.5Å are also observed on the surface We have previously labeled these clusters as type A and type B clusters respectively [section 4.2.1] A more interesting observation however is the appearance of similar cluster features in shape and size respective to the type “A” and “B” clusters when the surface transforms from the (6x6) cluster phase at 1000°C to a disordered phase

at 1050°C (Fig 4.9(e)) and later in the formation of ring structures with depression at the center at 1100°C (Fig 4.9(f))

By comparing the density of atoms on the surface of the (3x3) reconstruction prepared at 850°C with the final ring structure obtained at 1100°C, as shown in Fig 4.9, the surface has clearly changed from a compact and high density Si-rich structure to a

surface that is decorated with depressions/vacancies This morphological changes clearly indicate a loss of material and it occurs probably through desorption of Si when the substrate is progressively heated to 1100oC from 850oC Apart from STM data, evidence for this loss of Si material can also be inferred from the XPS measurements (Fig 4.10 and 4.11) corresponding to the surfaces shown in Figure 4.9, which is discussed in the following section

4.4.2 Formation of (6x6) rings from (6x6) clusters: XPS

Fig 4.10 shows the best fit XPS spectra for the Si 2p signal after chemical etching and outgassing in UHV (as-received), after Si deposition at room temperature and after annealing to temperatures ranging from 800°C to 1200°C In the Si 2p core level spectrum, we can distinguish two contributions at 99.3eV and 102.0eV The component

at 102.0eV is detected from the surface after chemical etching and is attributed to the

Si-C bonding in the SiSi-C bulk The component at 99.3eV only appears upon Si deposition at room temperature and is due to Si-Si bonding derived from the presence of elemental Si now present at the surface

In order to extract information about the surface Si stoichiometry corresponding

to the various surface reconstructions arising from progressive annealing, we compare the relative atomic concentration of Si-C and Si-Si as shown in Fig 4.11, obtained from the following expression;

TF(ASF)

ITF

(ASF)I

TF(ASF)

I

2p Si

C - Si 2p

Si

Si - Si

2p Si

Si - Si

ITF

(ASF)I

TF(ASF)

I

2p Si

C - Si 2p

Si

Si - Si

2p Si

C - Si

Figure 4.10 shows the peak fitted Si 2p signal (i.e elemental Si component at 99.3eV and the Si-C component at 103.0eV) due to surface after chemical etch, after Si deposition at room temperature and after annealing the substrate to different temperatures