Magic clusters on group IV surfaces 5

These magic clusters appear to facilitate the propagation of 7x7 domains, where by excess Si adatoms on the “1x1” phase are removed and form magic clusters before being transported to th

Chapter 5: Si magic clusters on Si(111) (7x7)

This chapter discusses the data obtained from STM which was used to probe the formation and structure of Si magic clusters on the Si(111)-(7x7) surface in UHV The first part

of this chapter focuses on the formation of Si magic clusters during the phase transformation from “1x1” to (7x7) reconstruction When the Si(111) surface is quenched rapidly from 1200oC

to room temperature, non-DAS cluster-like Si particles possessing uniform shapes and sizes are observed to be spontaneously formed during this process These magic clusters appear to facilitate the propagation of (7x7) domains, where by excess Si adatoms on the “1x1” phase are removed and form magic clusters before being transported to the step edges

In the second part of this chapter, we will attempt to selectively grow Si magic clusters

by depositing Si adatoms from Si solid source evaporator on the Si(111)-(7x7) and annealing the surface structure instead By carefully controlling the substrate temperature, the structural evolution leading to formation of Si magic clusters is studied We also probe the formation of various pre-cursor structures and a mechanism is proposed to account for their formation We will show that the assembly of 3 tetra-clusters leads to the formation of a magic cluster By analyzing the shape and size of the cluster at different biasing conditions, the structures of both clusters are elucidated

5.1 Si magic clusters on Si(111): via heat and quench

5.1.1 Global Morphology

As discussed in the literature review in Chapter 2, it has been reported that the (7x7) reconstruction disorders at about 870oC into the “1x1” phase and reverts back into the (7x7) structure as the surface is cooled below this transition temperature Localized metastable DAS structures have been observed during this phase transition However STM observation of this process at such high temperatures is difficult due to a high thermal drift effect which renders inconsistency in scan frame capture The imaging of the surface is thus best suited to room temperature scanning However due to the high surface mobility associated with Si adatoms, the adatoms are likely to diffuse very quickly to the step edges which they will attach and integrate into, thus making characterization of surface adatoms difficult In order to overcome this problem, the surface to be studied would have to be cooled very quickly in order to trap the adatoms on the surface This trapping process is only possible when the time required to quench

Si(111)-to low temperatures is shorter than the time for the Si adaSi(111)-toms Si(111)-to diffuse across the terraces Si(111)-to the steps

We first show the starting surface template in Fig 5.1a, which is a 1000nmx1000nm large scan STM image of a clean Si(111) surface This surface was obtained by flashing a H-terminated Si(111) substrate to 1200°C and cooled down to room temperature before STM scanning The image shows large terraces (~ 300nm to 400nm wide) with step edges running along the <110> direction Zoom-in images of this surface shows well-ordered (7x7) surface reconstruction scanned under varying tunneling voltages, V=-2.0V (Fig 5.1b) and V=+2.0V (Fig.5.1c) No surface features such as islands or clusters are observed on the terraces or at the

step edges Hence we use this (7x7) reconstruction as the starting surface template, to generate the “1x1” phase by heating it to 1200oC before quenching rapidly to room temperature at various estimated cooling rates, in order to ascertain the most effective cooling rate as “1x1” transits into (7x7) phase

Figure 5.1: (a) 1000nmx1000nm STM image of Si(111) flashed to 1200oC and scanned at room temperature and 15nmx15nm zoom-in images of clean Si(111)-(7x7) scanned at (b) -2.0V and (c) +2.0V

The clean Si(111) surface was flashed to high temperatures of 1200oC and subsequently cooled at different rates of ~ 1oC/min, ~ 50oC/min and 100oC/min to room temperature before being scanned using STM, as shown in Fig 5.2(a), (b) and (c) respectively

Figure 5.2 shows STM images of Si(111) after flashing to 1200oC and cooled at (a) ~ 1oC/min, (b) ~ 50oC/min and (c) ~ 100oC/min in (i) 1000nmx1000nm (ii)100nmx100nm and (iii) 30nmx30nm scans

The 1000nmx1000nm scans are shown in Fig 5.2a-c(i) Fig 5.2a(i) and Fig 5.2b(i) share similar global morphology which is dominated by flat terraces which are about ~300-400nm wide, with step edges running in the <110> direction In contrast, Fig 5.2c(i) shows large

Step edges Step edges

(9x9)

(5x5) (2x2)

(7x7)

triangular domains with trails leading from the domain apexes to the step edges on the terrace surface Zoom-in 100nmx100nm observations of these respective surfaces are shown in Fig 5.2a(ii), Fig 5.2b(ii) and Fig 5.2c(ii) No significant difference is seen between the two surface morphologies in Fig 5.2a(ii) and Fig 5.2b(ii), which show large terraces and step edges However Fig 5.2c(ii) shows large triangular (7x7) domains co-existing with bright regions of disordered “1x1” phase on wide terraces A closer examination of the surface reveals well resolved cluster-like particles which appear bright and round in shape residing in the “1x1” regions while the (7x7) domains are clearly characterized by large and well ordered triangular domains showing well defined (7x7) unit cells

Zoom-in 30nmx30nm images are shown in Fig 5.2a(iii), Fig 5.2b(iii) and Fig 5.2c(iii) Well ordered (7x7) reconstruction are observed to dominate the terrace surfaces with no other features present as shown in Fig 5.2a(iii) This could be attributed to the slow cooling rate of ~

1oC/min, which have allowed sufficient time for Si adatom diffusion While Fig 5.2b(iii) also shows long range (7x7) ordering, Si cluster-like particles are now observed on the surface, albeit accumulated at the step edges away from the terraces This shows that while the cooling rate of ~

50oC/min allows for the observation of these features, it is still sufficient for Si diffusion to the step edges resulting in the particles decorating the steps Fig 5.2c(iii) shows a considerable number of these same cluster-like particles existing on the disordered “1x1” phase, indicating that the cooling rate of ~ 100oC/min is quick enough to capture sufficient particle populations on the terrace surface Further inspection of the surface shows initial ordering of Si adatoms among the disordered “1x1” phase and the presence of metastable DAS phases such as (5x5), (7x7) and (9x9) as well as non-DAS phases such as (2x2) It is interesting to note from the STM data, that

in addition to the meta-stable structures and single Si adatoms, the feature with the largest occurrence existing on top of the “1x1” phase appears to be cluster-like particles

We obtain the dimensions of each cluster-like particle by taking the average of STM line profile measurements to represent the estimated size and height of each particle An example of the cluster size measurement is shown in Fig 5.3(a) The separation across the area occupied by the bright maxima of the cluster is measured by line profile in 3 directions along the <110> direction The average of these 3 readings is the estimated diameter of each cluster Similarly, the height difference measured from the peak of each cluster maxima to the neighboring (7x7) Si adatom peak is used to obtain the average height of each cluster, as shown in Fig 5.3(b)

The average cluster sizes are counted as shown in Fig 5.3(c) and tabulated into a histogram showing the cluster size distribution for each scan (about ~ 100±5 clusters per scan) The statistical data collected from 5 scans shows a narrow cluster size distribution with the largest occurrence of the estimated average cluster size to be 14.0±0.5Å This information coupled with the STM observation of the same clusters consistently possessing a uniformly round shape suggests that each of these particles are magic clusters This identification of the Si magic clusters is significant, as these clusters are consistently present during the surface evolution during (7x7) reconstruction domain growth from “1x1” As the size of features observed under STM are sensitive to changes in the tunneling bias used during scanning, we will examine the shape and size of the Si magic cluster under varying tunneling biases in the following section

Fig 5.3(a)-(b) is a 8nmx8nm scan showing (a) the measurement of the separation between corresponding bright maxima to obtain the average diameter of each cluster to be ~ 14.0±0.5Å (b) the average height of cluster obtained from the peak of the maxima to the trough of the neighboring “1x1” phase to be 2.1±0.1Å Fig 5.3(c) shows the tabulation of all measured average cluster diameters from a 100nmx100nm STM scan to show the cluster size distribution where the average size of 14.0±0.5Å has the highest occurrence

0 0.5 1 1.5 2 2.5 3 0

0.2 0.4 0.6 0.8 1 1.2 1.4

X[nm]

0 0.5 1 1.5 2 2.5 3 0

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

X[nm]

0 0.5 1 1.5 2 2.5 0

0.5 1 1.5 2

-0 0.5 1 1.5 2 2.5 3 0

0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

D) 2.1Å D

(A)

(B)

20nm

0 11

58

19 6 0 0

10 20 30 40 50 60 70

-(C)

(a)

(b)

(c)

5.1.2 Dual Biasing STM Analysis: Si magic cluster from heat and quench

Using the STM, we scan the Si(111) surface at room temperature which has been

quenched from 1200oC , where the morphology is observed to comprise of ordered (7x7) and dis-ordered “1x1” domains We focus the STM scan frame on the same area which shows a domain boundary with both (7x7) and “1x1” structures (Fig 5.4) The scan frame is

determined to be of the same area by the identification of consistent features such as the kink

in the (7x7) domain boundary as indicated by the arrow in each scan frame We vary the tunneling voltage of the scan from V=-1.8V to +1.8V, in steps of 0.2V, to capture a series of 30nmx30nm STM images of this area as shown in Fig 5.4 The presence of the (7x7)

reconstruction serves as a reference feature (See Chapter 5 – Appendix 1), as the physical appearance of the unit cells change with respect to different tunneling voltages

The STM scans from V=-1.8V to -1.0V show the (7x7) reconstruction appearing as unit cells consisting of 2 triangles with contrasting brightness The “1x1” domain is clearly

resolved to show a region of disordered small round dots with the same brightness as the (7x7) structures The high resolution scans also pick up the other features residing in the

‘1x1” domain such as metastable DAS (9x9) structures and Si magic clusters which appear brighter than the background surface and (7x7) surface It is interesting to note at this point that the Si magic clusters appear as bright blobs which are round in shape and found to be sitting on top of the “1x1” surface

(a)

(b)

At voltage (V) = -0.8V, the “1x1” region appear to be slightly darker than the (7x7) domain, while no changes to the intensity of the other features were noted From V=-0.6V onwards, the “1x1” region becomes gradually darker until it can no longer be resolved at V=-0.05V and appears as dark patches on the STM scan However the Si adatoms of the

metastable DAS structures can still be observed albeit now at the same intensity as the (7x7) domains, which suggests that the DAS structures could be sitting on top of the “1x1” phase

It is also at this stage where the top layer Si adatoms of the (7x7) reconstruction begin to appear brighter which eventually leads to the appearence of (7x7) unit cells as hexagonal rather than triangular arrangements of Si adatoms However the contrast difference between the Faulted and Unfaulted Halves of the unit cell is still discernable albeit less obvious than previously It is observed that the Si magic clusters still remain as the brightest features in the scan frame and still retain the same round appearance as previously seen in higher negative voltage scans

When the tunneling voltage is switched over to low positive values of V=+0.05V, the surface appearance is observed to be similar to that when scanned at V=-0.05V However, as the positive tunneling voltage is being increased from V=+0.2V to +1.0V, the previously dark patches of disordered “1x1” areas are being gradually resolved and begin to appear as faint small dots amidst the (7x7) and DAS structures which remain at the same brightness and appearance The Si magic clusters do not change in appearance and are still observed to

be brighter than the other features, even though the brightness of the clusters is now reduced

When the tunneling voltage is increased from V = +1.0V to +1.8V, the previously faint

“1x1” regions now gradually become brighter and eventually regain the same intensity as it was first observed at V=-1.8V In fact, the Si adatoms in the disordered “1x1” areas can now

be resolved and appear to possess the same brightness as the (7x7) domains and DAS

structures While the DAS structures remain the same in appearance, the Si adatoms in the (7x7) reconstruction which are still seen to be arranged in the hexagonal configuration, now appear uniform in intensity without distinction between adatoms existing in either Faulted or Unfaulted unit cell halves The Si magic clusters are observed to remain similar in shape and size, although the cluster intensity is lower than when scanned at V=-1.8V

In spite of the change in tunneling voltages, the Si magic clusters remain as the brightest features, while the other structures have been observed at similar intensities/brightness at some stage of scanning This suggests that the clusters are likely to be the highest structures existing amongst the (7x7) and “1x1” domains In order to study the clusters more closely,

we use STM to zoom in onto one specific Si magic cluster which is circled in Fig 5.4 and capture high resolution images as shown in Fig 5.5 (cluster is identified by the arrow)

In the series of images in Fig 5.5, we scan the same Si magic cluster over the voltage range of V=-1.8V to +1.8V in steps of 0.2V From voltage scans of V=-1.8V to -0.8V, the magic cluster is observed to be a bright feature possessing a round shape When the tunneling voltages enter the V=-0.6V to +0.05V range, the Si magic cluster retain its shape but appear to be less bright compared to the previous voltage scans (V=-1.8V to -0.8V) When the STM continues the

scans from tunneling voltages V=+0.2V to +1.0V, the Si magic cluster, while still retaining the same shape and size, appear to be gradually resolved into 3 smaller features within the original cluster structure This is particularly evident in the scan at V=+0.8V, where the arrows indicate the 3 separate features which appear to be similar in size and shape to the neighboring electron clouds associated with Si adatoms of the DAS structures or (7x7) reconstruction This observation is interesting as it suggests that the cluster could comprise of 3 Si atoms with single dangling bonds akin to Si adatoms in the DAS structure Further scanning of the same area at voltages V=+1.2V to +1.8V shows that the magic cluster gradually revert back to the original appearance as a single bright and round particle

6.0nm 6.0nm

-1.6V -1.8V

6.0nm

-1.4V

6.0nm-1.2V

We will now study these feature changes of the cluster more closely, in order to better understand the cluster structure Fig 5.6(a) shows 40nmx40nm STM image of the same surface area scanned in the different biases of (i)-1.8V and (ii)+1.8V Both scans show both the (7x7) domain and the “1x1” region In particular, we zoom in onto 4 specific clusters as highlighted by the dark outlined box in both Fig 5.6(a)(i) and (ii) to examine with greater resolution as shown

in the 8nmx8nm scan displayed in Fig 5.6(b)(i)-(vi) The 4 clusters are identified as clusters 1, 2,

3 and 4 to demonstrate that the same clusters are being scanned These clusters are scanned with varied tunneling biases in different scan frames with corresponding biases (i) -1.8V, (ii) -1.5V, (iii) +0.8V, (iv) +1.0V, (v) +1.4V and (vi) +1.8V

In scan (i) and (ii), all 4 magic clusters appear as bright particles sitting on top of the darker disordered “1x1” underlying surface, consistent with the earlier STM descriptions However in scan (iii) and (iv), each individual cluster can now be resolved into 3 smaller sub-unit features of similar shape, size and brightness The 3 arrows in Fig 5.6(b)(iii) clearly indicate the 3 separate sub-units resolved within cluster 3 when scanned at +0.8V It is also interesting to note that each sub-unit feature appears to have the same shape and size as the neighboring Si adatoms of the DAS structures and “1x1” regions We will use line profile measurements to analyze these sub-unit features and compare them to the neighboring Si adatoms in the following section When the tunneling voltage is increased to (v) +1.4v and (vi) +1.8V, the clusters resume the appearance of one singular bright maxima similar to (i) and (ii)

We estimate the average size of each sub-unit feature resolved within a Si magic cluster

to be ~ ~4.5 ±0.5Å each, as shown by the line profile measurements of sub-units A, B and C in

Fig 5.7(a) In the same image, the Si adatoms in the neighboring DAS structures are also estimated to possess a similar average size of ~4.5 ±0.5Å from line profile measurements D and

E It should be noted that the magic cluster sub-unit features and Si adatoms are both round in shape and appear to have the same brightness Fig 5.7(b) shows the collective line profile measurements (A-I) of the separation between each sub-unit feature resolved within the various

Si magic clusters From these measurements, the 3 sub-unit features appear to be arranged in an isosceles triangle with estimated separations of ~5.7 ±0.5Å, ~5.7 ±0.5Å and ~7.5 ±0.5Å

Therefore the data suggests that a Si magic cluster may comprise of 3 smaller sub-unit clusters arranged in an isosceles triangle with an estimated separations of ~5.7 ±0.5Å, ~5.7±0.5Å and ~7.5 ±0.5Å as illustrated in Fig 5.7(c) We also establish that the size of the Si magic cluster

is estimated to be ~14.0 ±0.5Å and exits as a single round entity when imaged in negative biases

We will use this information to elucidate the structure of the Si magic cluster sitting on a (7x7) unit cell in section 5.2.3

Figure 5.6 shows STM scans of (a) 40nmx40nm of same area of ordered (7x7) and disordered

“1x1” domains scanned at (i) -1.8V and (ii) +1.8V (b) 8nmx8nm of the same 4 magic clusters scanned at (i) -1.8V (ii)-1.5V (iii)+0.8V (iv)+1.0V (v)+1.4V and (vi)+1.8V

Figure 5.7(a) shows line profile analysis of the smaller sub-units resolved within the Si magic cluster (A), (B) and (C) to have an average size of ~4.5 ±0.5Å each, and the average size of the neighboring Si adatoms of the “1x1” region to be also ~4.5 ±0.5Å

Figure 5.7(b) shows line profile analysis of the average separation between each smaller sub-unit resolved within the Si magic cluster to be ~5.7 ±0.5Å, 5.7 ±0.5Å and 7.5 ±0.5Å

Figure 5.7(c) shows a schematic of the Si magic cluster

1.6nm

F

1.6nm

ACB

0.2 0.4 0.6 0.8 1

X[nm]

0 0.5 1 1.5 2 0

0.2 0.4 0.6 0.8

0.2 0.4 0.6 0.8 1 1.2 1.4

4.5Å

(C)

5.1.3 Real Time STM: Formation of Si magic clusters

In order to probe the surface structural changes as “1x1” phase transforms to (7x7) phase, fast scanning STM (5 sec/ frame) was used to image the surface in real time immediately after the surface was heated and quenched from 1200oC While the initial images were fuzzy and thermal drifting still occurred due to the high temperatures, the fast scanning technique nevertheless managed to focus the scanner onto almost the same scan area consistently

Figure 5.8 shows 25nmx25nm STM images of the surface quenched from 1200oC after (a) t=5sec (b) t=40 sec (c) t =120sec (d) t=240sec (initial ordering of Si adatoms into meta-stable (7x7), (9x9) and (11x11) phases) (e) t=400sec (formation of (7x7) domains)

(7x7)

(7x7) domain

5nm

Fig 5.8(a)-(e) show the 25nmx25nm STM images of Si(111) surface evolution in real time, as the surface is being cooled from 1200oC Fig 5.8(a) shows the surface immediately after quenching down from 1200ºC at t=5sec (t=time) The image shows a co-existence of small bright spots and larger equally bright disordered blobs, which do not exhibit a defined shape This could be due to thermal drift effects of the still hot surface affecting the STM scanning The smaller bright spots are likely to be the “1x1” phase observed at high temperatures due to the immediate STM monitoring of the surface after quenching, while the larger disordered blobs possess an average size of ~14.0±0.5Ǻ and are therefore Si clusters evolving from this phase undergoing re-structuring In spite of the difficulty in resolving the surface structures, the STM still able to show larger Si clusters which appear brighter than the smaller and darker “1x1” background surface This suggests that the clusters are likely to exist on top of the disordered

“1x1” phase which was undergoing structural change

The surface is imaged at t=40sec, as shown in Fig 5.8(b), which reveals three distinct features showing regions of darker spots, larger and brighter protrusions as well as localized ordered structures The darker regions which are now resolved as disordered dark spots are likely

to be the “1x1” phase Si adatoms, while the brighter and larger features appear to be Si magic clusters which can now be clearly seen to be round in shape and sitting higher than the “1x1” phase, similar to the previous observation The localized features which are observed for the first time appear to resemble a DAS structure and is likely to be the initial nucleation of a (7x7) unit cell on top of the “1x1” phase

Fig 5.8(c) shows the surface at t=120sec, where a larger region of ordered (7x7) adatoms

or DAS structures co-existing with Si magic clusters can now be detected along with the underlying “1x1” phase atoms At time=240sec, meta-stable DAS structures such as (9x9) and (11x11) can be observed alongside the (7x7) reconstructions, as shown in Fig 5.8(d) These observed DAS structures are only made up of outer adatoms configured in a triangular arrangement, which suggests that the formation of edge adatoms in DAS unit cells precedes that

of the central atom formation This observation is consistent with reports by M Chida et al and

W Shimada et al [1-2] who showed that nucleation of DAS structures started at corner holes at

the apexes of the DAS structures and began with Si adatoms arranging themselves into rows along the outside edge of the (7x7) triangular domain Larger regions of (7x7) domains are observed in Fig 5.8(e) at time=400sec The DAS structures now appear more complete with the existence of stacking faults and as well as corner holes bordering each (7x7) unit cell This suggests that the underlying dimer chains have been developed to form the stacking faults and corner holes which are essential in stabilizing the DAS structures and also propagating the (7x7) domains

During the cooling process, it would appear that the growth and propagation of the (7x7) domains on the terrace led to the accumulation and “pushing” of excess adatoms onto the “1x1” region There seems to be sufficient kinetic energy to allow adatoms within the “1x1” domain to move and rearrange into high atomic density structures as such as (9x9) and (11x11) in order to

reduce surface energy It was Williams et al [3] first reported that “1x1” has a greater atomic

density of 6% than the (7x7) phase This estimate was based on the observation of 6% surface coverage of “1x1” islands they observed on the (7x7) reconstructed surface when they heated the

Si(111) to high temperatures before quenching down and scanning with the STM The metastable reconstructions with atomic densities larger than (7x7) such as (9x9), (11x11) and (13x13) were consequently assumed to be intermediate structure regimes which formed to accommodate the expected excess atoms However they did not consider the occurrence or role

of white cluster features with regards to the phase transformation Furthermore, the 6% atomic density difference reported, is based on the estimation of surface coverage rather than based on the counting of difference in number of atoms

Instead of relying on surface coverage, we account for the atomic difference between

“1x1” and (7x7) by counting the total number of Si atoms in 3 layers of a truncated bulk Si(111) based on a (7x7) unit cell as shown in Fig 5.9(a) and compare it to the total number of Si atoms existing in the adatom, restatom and dimer layers of the Takayanagi (7x7) model [4], as shown

in Fig 5.9(b) We counted a total of 147 Si atoms in the truncated bulk Si(111) versus 102 atoms

in the (7x7) structure This would translate into a difference of 45 Si atoms per (7x7) unit cell This represented an estimated difference of ~ 45 Si atoms per unit cell which would have to be removed in order for the “1x1” to transit into the “7x7” phase This observation indicates that the nucleation of one (7x7) unit cell could hence generate up to ~ 11 Si magic clusters onto the

“1x1” surface It is interesting to note from the STM data, the feature with the largest occurrence existing on top of the “1x1” phase appears to be the Si magic clusters

Fig 5.9: (A) Schematic of (7x7) unit cell super-imposed on an ideally terminated (1x1) Si(111) substrate and (B) Schematic of (7x7) unit cell

1stlayer: Adatom

2ndlayer: Rest Atom

Faulted Half Unfaulted Half

3rdlayer: Dimer

1stlayer Si adatoms 2nd layer Si adatoms(A)

(B)

With the “1x1’ phase having a larger atomic density of ~4% per unit cell more than (7x7), the expulsion of this excess Si from “1x1” is therefore necessary in order to effect the transition to (7x7) The STM data which shows the clusters as bright features existing on top of the “1x1” phase, suggests that Si atoms are ejected from the “1x1” phase onto the surface and form into Si magic clusters, while the remaining Si adatoms reconstruct to form DAS (7x7) unit cells As the dual biasing data shows that each Si magic cluster consists of 3 smaller clusters of size ~4.5Ǻ, hence Si adatoms could pop up from the “1x1” surface as Si tetra-clusters and consequently agglomerate into the larger Si magic cluster species (size~14.5Ǻ) comprising of several Si tetra-clusters in a formation mechanism similar to the nucleation of Si clusters on SiC [5-6] However as we are unable to observe this formation mechanism in real time, hence we cannot conclude whether Si pops up as single adatoms and form tetra-clusters/magic clusters or pop up as Si tetra-clusters and form magic clusters Therefore we will address these questions regarding the formation mechanism by studying the progressive formation of Si magic clusters from deposited Si adatoms on Si(111)-(7x7) instead of heating/quenching the Si(111)-(7x7) surface in Section 5.2

The observed clustering of Si adatoms into larger Si magic clusters clearly results in a more stable entity This is perhaps not surprising given the unique stability associated with the magic clusters, which may be attributed to minimization of dangling bonds due to the “magic’ number and configuration of Si adatoms in making up the cluster As the formation of (7x7) unit cells would inevitably lead to propagation of (7x7) domains, one would also expect that the excess Si atoms formed on the “1x1” would have to diffuse away in order to accommodate the growth of (7x7) domains and consequent shrinking of “1x1” domains Since the excess Si on

“1x1” forms into magic clusters, hence we would study in the following section, if Si diffuses as adatoms or magic clusters during the phase transformation

5.1.4 Annealing of quenched surface

This section uses STM to examine the domain evolution of the (7x7) and “1x1” surface

as a function of annealing time and temperature The surface is generated by heating to 1200oC, before we commence the annealing cycles

As a function of annealing time

Fig 5.10(a)(i)-(vi) (100nmx100nm STM scans) shows the surface annealed at increasing time periods of 1min, 2min, 5 min, 15min, 60min and 120min Fig 5.10(b) shows the estimated % surface coverage of (7x7) and “1x1” domains as a function of annealing time at 200oC The data

is expressed as a % obtained from the ratio of number of (7x7) unit cells per 100nmx100nm after each annealing divided by total number of (7x7) unit cells per 100nmx100nm (i.e there are 1634 unit cells)

Fig 5.10(a)(i) shows the surface quenched from 1200oC to room temperature and annealed to 200oC for 1 min The STM scan reveals disordered “1x1” co-existing with triangular domains of well resolved (7x7) structure The (7x7) domains are estimated to cover about ~ 40%

of the scan area (Fig 5.10(b)) Si magic clusters are seen to reside only within the “1x1” areas while no clusters have been observed on the well ordered (7x7) terraces Fig 5.10(a)(ii) shows roughly the same scan area after annealing at the same temperature for an additional 2 min The STM image shows that the (7x7) domains appear to have grown in size and joined up to form larger (7x7) regions while the “1x1” regions now seem to be smaller This trend is similarly

observed when the same surface is annealed for an additional 5 min as shown in Fig 5.10(a)(iii) (7x7) surface coverage is estimated to be ~ 50% at this point (Fig 5.10(b)) When the surface is annealed for another 15min, the (7x7) covers about ~ 60% of the surface area compared to

“1x1”, which are largely confined to regions decorating the step edges and in isolated domains (Fig 5.10(a)(iv)) In fact we also identify a trail of Si magic clusters connecting from the apex of

a “1x1” domain to another as identified in the same image It is interesting to note that these trails are more obvious when “1x1” domain sizes are small enough Upon annealing to 60min and 120min as seen in Fig 5.10(a)(v) and Fig 5.10(a)(vi) respectively, we observe that the area occupied by “1x1” has become significantly smaller This is evidenced by the estimated “1x1” surface coverage of ~ 30% and 10% for Fig 5.10(a)(v) and Fig 5.10(a)(vi) respectively Fig 5.10(b) shows that the overall trend for coverage associated with the “1x1” decreases relative to the (7x7) coverage and is non-linear over the entire period of anneal

As a function of annealing temperature

Fig 5.10(c)(i)-(v) shows 500nmx500nm STM scans of the surface at room temperature after it is annealed for 30min at 100oC, 150oC, 200oC, 250oC and 300oC respectively Fig 5.10(c)(i) shows dark triangular areas of size ~ 80nm, which we identify as (7x7) domains, co-existing with the brighter regions on the surface which we identify as “1x1” As the surface is gradually annealed at higher temperatures of 150oC (Fig 5.10(c)(ii)) and 200oC(Fig 5.10(c)(iii)),

we observe that the average size of the dark triangular areas is correspondingly larger (~100nm and 150nm respectively) With further annealing to 250oC (Fig 5.10(c)(iv)) and 300oC(Fig 5.10(c)(v)), the dark triangular areas are now > 200nm in size and are much larger than the

brighter areas, which are now also confined to triangular domains Similar to data from the previous annealing experiment, we also observe bright trails leading from the apex of the bright domains to the step edges We again observe in the 100nmx100nm image (Fig 5.10(c)(vi)), trails comprising of Si magic clusters running across well ordered (7x7) reconstruction and leading towards the step edges From the STM observation, these cluster trails are only discernable when the “1x1” bright areas are sufficiently small enough Once again, Si magic clusters are not found

on the (7x7) terraces but exist only along the trails observed By estimating the area occupied by (7x7) domains after each annealing, we are able to generate a plot (Fig 5.10(d)) describing the respective % surface coverage The trend is similar to Fig 5.10(b), where coverage associated with the “1x1” decreases while (7x7) increases as a function of temperature

From both sets of STM data, the initial disordered surface consisting of both “1x1” and (7x7) structures undergoes progressive “1x1” to (7x7) phase transformation as the surface is being annealed at higher temperatures (same annealing time) or for longer periods of time (same annealing temperature) We observe that the surface area occupied by (7x7) domains grows while “1x1” domains shrink in size relatively as the surface is being annealed In all the annealing processes, we do not observe Si magic clusters on (7x7) terraces Instead the clusters are found to exist along trails which connect from the apexes of a “1x1” domain to another domain or to the step edges As quenching of the surface freezes the growth front and allows us

to capture snap shots of the surface, the trails of clusters observed could be attributed to Si magic clusters diffusing along domain boundaries While it is generally assumed that surface diffusion

of Si proceeds via adatom diffusion, our STM data, on the other hand, suggests that Si atoms exist as clusters during the “1x1” to (7x7) phase transition and are hence likely to participate in

surface diffusion as clusters Thus the STM data suggests that diffusion of Si could have occurred via Si magic clusters during the growth of (7x7) domains

Although we were unable to see the real time diffusion of Si magic clusters, T.T Tsong

et al [7] observed the diffusion of a similarly sized Si cluster across ordered (7x7) surface and

found its diffusion barrier to be ~ 2eV They also reported that as the magic cluster approached a step edge, it was observed to dissociate into Si adatoms which were incorporated into the step edge Hence our STM data suggests that the phase transition is likely to proceed via;

1) 45 atoms per (7x7) unit cell would have to be removed in order for the “1x1” to transit into the “7x7” phase

2) Formation of one (7x7) unit cell would generate up to ~ 11 Si magic clusters

3) These excess Si adatoms pops onto the “1x1” surface as “1x1” → (7x7) and forms Si magic clusters on the surface

4) Si magic clusters diffuse along trails towards step edges as (7x7) domains propagate 5) Magic clusters dissociate and incorporate into step edges thereby perpetuating step edge growth

6) Hence formation and diffusion of Si magic cluster is a critical step in the nucleation and growth of (7x7) surface

Therefore these observations suggest that Si magic clusters play an important role in facilitating the “1x1” to (7x7) phase transformation Since excess Si adatoms can pop up onto the surface and form Si magic clusters, hence it would be interesting to see if we can selectively grow magic clusters from Si adatoms deposited on (7x7) surface, in the following section Chapter 5.2.1

Figure 5.10(a) shows 100nmx100nm STM scan of the surface after heating to 200oC for (i) 1 min, (ii) 2min, (iii) 5min, (iv) 15min, (v) 60min and (vi) 120min for each anneal Figure 5.10(b) shows the estimated % surface coverage of (7x7) and “1x1” as a function of annealing time at

o

S te p e d e

S te p e d e

S te p e d e

S te p e d e

Series1 Series2

Figure 5.10(c) shows 500nmx500nm STM scan of the surface after heating to (i)100oC, (ii)150oC, (iii) 200oC, (iv) 250oC and (v) 300oC (vi) shows a 100nmx100nm zoom in of the surface at 250oC

Figure 5.10(d) shows the estimated % surface coverage of (7x7) and “1x1” as a function of annealing temperature

Cluster trail

Cluster trail

d (7 x7)

Disordere

d “1 x1”

Cluster trail

Cluster trail

d (7 x7)

Disordere

d “1 x1”

5.2 Selective growth of Si magic clusters on (7x7)-Si(111)

5.2.1 Formation of Si Magic Clusters

From the previous section, we demonstrated that we are able to grow Si magic clusters spontaneously from Si adatoms popping up from Si(111) surface via heating and quenching treatment Hence in this section, we will now explore if we can selectively grow Si magic clusters by first depositing Si adatoms onto Si(111)-(7x7) at room temperature followed by annealing of the surface to nucleate magic clusters

In order to determine the suitable amount of Si to be deposited, we deposited coverages

of 1.5ML, 1.0Ml and 0.6ML from the Si solid source evaporator at room temperature onto the Si(111)-(7x7) surface and observed the respective surface morphologies with STM as the surfaces were being annealed The STM data for these coverages are shown in Chapter 5 Appendix-2 The STM scans shows that Si deposited forms disordered blob-like features (average size ~ 2-3nm) on the Si(111)-(7x7) surface at room temperature These features agglomerate as the surface is being annealed initially, and are later observed to form islands which become larger (average size ~ 6-8nm) at higher temperatures Annealing of the surfaces to

200oC eventually results in the formation of ordered (7x7) islands as the blobs are gradually no longer observed Continued annealing of the surface resulted in all the larger particles eventually nucleating well ordered (7x7) islands However the disappearance of the islands upon further annealing to 400oC suggests that Si detachment from the (7x7) islands occurs, leading to diffusion and incorporation into step edges and consequently resulting in wide terraces of Si(111)-(7x7) As these surface coverages of Si are too high to observe the evolution of the

cluster species, hence we focus on a lower coverage of 0.1ML coverage of Si deposited and study the surface evolution of these clusters as a function of annealing temperature

We use a clean Si(111)-(7x7) surface as seen in the 50nmx50nm STM scan in Fig 5.11(a) as the starting template Using the Si solid source evaporator, we deposit 0.1ML of Si onto this surface at room temperature as shown in Fig 5.11(b) As the Si coverage is not extensive, large areas of well-ordered (7x7) unit cells can still be clearly seen In contrast to the earlier scans of the surface with higher Si coverage at this same temperature where disordered blobs are observed, the STM scans now show single bright round features which are similar to the size of Si adatoms on the (7x7) structure These are likely to be Si adatoms deposited from the Si solid source evaporator, which indicates that single Si atoms are being produced from the evaporator as opposed to Si blobs In addition, we are also able to observe a considerable number

of small and faint features which are well distributed throughout the surface morphology

When the surface is annealed to 70oC, numerous small and bright features which are no larger than a half unit cell area of the (7x7) structure can now be seen in Fig 5.11(c) These features appear to be distributed in a disordered fashion with the underlying (7x7) still clearly exposed When this surface is annealed to 100oC, the STM scan in Fig 5.11(d) does not reveal significant changes to the global morphology However at a higher temperature of 130oC, the image in Fig 5.11(e) shows the appearance of larger and brighter particles which are round in shape co-existing with the earlier smaller features

Further annealing to 150oC leads to a surface which is eventually dominated by these larger and brighter cluster-like particles, as shown in Fig 5.11(f) The smaller features are noted

to be conspicuously absent at this temperature and the bright cluster-like particles appear to possess uniform round shapes and similar sizes In fact, the STM also reveals that the particles prefer to adsorb onto the brighter half of the (7x7) unit cell halves, which has been identified to

be the Faulted Half Unit Cell of the reconstruction

Upon annealing to 200oC, the clusters retain the same shape, size and density; however they become increasingly disordered in spatial distribution as observed in Fig 5.11(g) It is noted that these clusters did not agglomerate to form larger particles at higher temperatures unlike previous observations of Si blob agglomeration At 400oC, as shown in Fig 5.11(h), the particles are no longer observed Instead, islands which appear brighter and thus higher than the underlying surface are observed to display (7x7) periodicity Further annealing of the surface to

500oC leads to the disappearance of the islands, leaving behind only large terraces of (7x7) reconstructed surface areas as shown in Fig 5.11(i)

The STM scans suggests that Si adatoms are deposited onto the (7x7) surface and first form disordered and smaller precursor structures which precede the formation of larger particles which possess uniform shape and size Annealing of the surface to higher temperatures leads to the surface being dominated by these particles which are also ordered in localized domains While this ordering is absent at higher temperatures, the particles retain their uniformity in shape and size and in fact do not agglomerate or increase in size As this suggests that there is stability

in the unique shape and size of the particles observed, we identify them as Si magic clusters

When the surface is heated to higher temperatures, small islands exhibiting (7x7) periodicity are seen before they are no longer observed at even higher temperatures In order to characterize the size and shape of the magic clusters in greater detail, we use higher resolution STM to examine the surface features This will also allow us to study the precursor structures and understand the formation mechanism leading to the nucleation of a Si magic cluster