Regular Paper Adsorption mechanisms of In atoms onto the Si ( 111 )-7 × 7 ; Clustering and substitution for Si atoms

Indium was deposited onto the Si(111)-7×7 reconstruction in order to form an ordered array structure of In nanoclusters. Using scanning tunneling microscopy, the annealing effects on the 0.12-monolayer (ML)-In nanoclusters and 0.24-ML-In nanoclusters have been investigated. In the case of 0.12-ML-In, four types of adsorption mechanism including the clustering of In atoms are observed after annealing at 300◦C, while, in the case of 0.24-ML-In, almost all In nanoclusters are maintained after annealing at 300◦C. This discrepancy is discussed with the relation between the substitution of In for Si adatoms and the clustering of In atoms. [DOI: 10.1380/ejssnt.2005.244]


I. INTRODUCTION
Numerous investigations of the In deposition onto Si surface have been devoted to understand the behavior of metal atoms on semiconductor surfaces.At moderately high substrate temperature, it is well known that In deposition induces several ordered surface reconstructions on the Si(111) substrate surface around the coverage of 1 monolayer (ML) [1][2][3][4].√ 3 × √ 3 and 4 × 1 are major reconstructions induced by In on the Si(111) surface.And these reconstructions have been intensively studied using various techniques, such as scanning tunneling microscopy (STM) [1,2,5], low-energy electron diffraction [6], and surface x-ray diffraction [7,8].Beyond the coverage where ordered reconstructions are formed, the growth of epitaxial In islands has been observed [4,9].For room-temperature (RT) deposition, three In-adsorption mechanisms have been confirmed [2,10,11]; (i) the substitution of In for Si atoms in the 7 × 7 adatom position, (ii) In atoms adsorption on top of Si adatoms, and (iii) covalent bonding between In and Si rest atoms in the outermost atomic layer.The In covered surface is non-metallic due to covalent bonding removing Si surface states near the Fermi Level [10,11].Recently, an ordered two-dimensional array structure of identical In nanoclusters on the Si(111)-7 × 7 was reported by Li, et al [12].It was achieved by regulating both of substrate temperature and In adsorption rate delicately.Such ordered arrays of metal nanoclusters are promising materials for ultra-highdensity recording and nanocatalysis etc.
In this report, we have investigated the early stage of In adsorption at the middle range (200 ∼ 300 • C) of substrate temperature.Furthermore, we pay attention to the formation of array structure consisting of identical In nanoclusters.Using STM, we have investigated the stability of the In nanoclusters as a function of substrate temperature.The coverage of In on the Si surface are 0.12 ML and 0.24 ML.Some differences in the thermal stability of the clusters between the case of 0.12 ML-In and the case of 0.24 ML-In are observed.These differences are discussed from the viewpoint of the conditions to form the identical * Corresponding author: d033016@ems.toyama-u.ac.jp clusters.

II. EXPERIMENTAL
Experiments have been carried out in an ultra-high vacuum (UHV) chamber (Omicron GmbH) with a base pressure of about 2 × 10 −10 Torr.The chamber is equipped with reflection high energy electron diffraction (RHEED), scanning tunneling microscopy (STM) and Auger electron spectroscopy (AES).
The substrates used in the present experiments were cut from a Sb-doped mirror-polished Si(111) wafer with a resistivity of 0.02 Ωcm and were a size of 13mm×3 mm×0.2 mm.The Si(111) substrates were cleaned by a flash anneal process up to 1250 • C after outgassing at 600 • C for 12 hours in UHV chamber.Then the (7 × 7) reconstruction of clean Si(111) surface was confirmed by RHEED and STM.
In was evaporated from a Boron nitride Knudsen-cell.A typical deposition rate of In onto the Si(111)-7 × 7 was estimated to be about 0.012 ML/min at the K-cell temperature of 530 • C. The error margin of this deposition rate is 5%.One-ML coverage of In is referenced to the number of surface Si-atoms of an unreconstructed Si(111) surface (1-ML = 7.8 × 10 14 atoms/cm 2 ).
For STM observations, tungsten (W) tips were employed.These W tips were electrochemically etched by potassium hydroxide (KOH) solution of 5 mol/l, and then oxidized tungsten layer was removed from the etched tip by being exposed to hydrogen fluoride acid (HF).After introducing the etched tips into UHV chamber, they were cleaned by in situ heating.All STM images were acquired in a constant-current mode after cooling the samples to RT.

A. 0.12-ML-In deposition
In order to form an ordered array structure of identical In nanoclusters, 0.12-ML-In was deposited onto the Si(111)-7 × 7 surface at the substrate temperature of 200 • C with very low deposition rate (0.012 ML/min). Figure 1  Figure 2 shows STM images of the sample after annealing at 300 • C for 10 minutes following 0.12 ML-In deposition.On this surface, several types of In adsorption can be found as follows.Type (i): The spots indicated by the arrows (i) appear as bright spots in the empty state image (Fig. 2(a)), whereas they appear dark in the filled state image (Fig. 2(b)).These spots are interpreted as the substitution of In for Si atoms in the 7 × 7 adatom position, since the dangling bond (p z -orbital) on the In adatom is completely empty.
Type (ii): The spots indicated by the arrows (ii) appear as bright spots in both the empty and filled state images.These spots are assigned to In atoms adsorbed on top of the Si adatoms.Some of such spots may be assigned to substituted Si atoms adsorption on top of the Si adatoms, but we cannot distinguish such substituted Si atoms from In atoms.
Type (iii): The spots indicated by the arrows (iii) appear dark in the empty state image, whereas they appear as bright spots in the filled state image.It is thought that these spots arise from charge redistribution caused by In adsorption on the Si rest atoms.The details of these mechanisms were described in the reference [11] by M. Yoon, et al.These present results are very similar to previous reports where In was deposited onto the Si(111)-7 ×7 reconstructed surface at room temperature [10,11].
Type (iv): The HUC indicated by an arrow (c) still contains an original In nanocluster.The typical image of the original In nanocluster can be seen clearly in the filled state image.However, The number of such HUCs containing an In nanocluster is less than 2% of all HUCs.
After annealing at the substrate temperature of 300 • C, almost all original In nanoclusters are thermally resolved, and then, In behaves as individual In atoms on the surface.This result consists with the report of Li [12] where they thought that In nanoclusters can exist up to 200 • C. It is also found that the In-√ 3 × √ 3 reconstruction begin to be formed on the edge of terraces.The ratio of the In atoms substituting for Si adatoms (described as the type (i)) is about 48% of all deposited In atoms.The other In atoms are on the Si atoms (as the type (ii) or (iii)), or are included in the In-√ 3 × √ 3 reconstruction where the local coverage of In atoms is about 0.33 ML.Therefore, after annealing at 300 • C, the type (i) of In adsorption is the predominant process.Within each HUC, In shows a preference to substitute for the edge Si adatoms rather than the corner Si adatoms in the Si(111)-7 × 7 reconstruction; the ratio of In atoms in edge sites to that in corner sites is about 2.2 : 1 under these conditions.This preference is well known, and there are a lot of reports concerning this phenomenon for several kinds of atoms [2,11,16,17].The substituted Si atoms may exist on the atoms in the Si(111)-7 × 7 reconstruction, but, it is difficult to distinguish the substituted Si atoms from In atoms on the Si(111)-7 × 7 reconstruction.Some of such Si atoms may migrate to the edge of terraces, and contribute to form the In-√ 3 × √ 3 reconstruction at the edge of terraces.From these results, the following conclusions are led; At the substrate temperature of 200 • C or less, almost all deposited In atoms migrate on the Si(111)-7 × 7 reconstructed surface, and then, six In atoms form an In nanocluster at the center of each HUC (as shown in the Fig. 1).In substitution for Si adatom is not observed.By annealing at 300 • C, In nanoclusters are thermally dissolved, and In atoms behave as individual atoms on the Si(111)-7 × 7 reconstructed surface.In substitution for Si adatoms becomes predominant process (described as the type (i) of In adsorption).As the annealing temperature increases, the number of In atoms substituting for Si atoms increases.Thus, the Si(111)-7 × 7 reconstruction is disturbed, and then, the In-√ 3 × √ 3 reconstruction begin to be formed.This interpretation seems to be natural.At sufficiently high substrate temperature, as the In coverage increases up to 0.33 ML which is the ideal In coverage of the In-√ 3 × √ 3 reconstruction, a similar disruption process of the Si(111)-7 × 7 reconstruction is expected.
From the above results, it is found that if In atoms have enough energy, In substitutes for Si adatoms rather than forms the nanoclusters.And, in Ref. [11] where 0.03 ML-In was deposited onto the Si(111)-7 × 7 reconstruction at RT, 30% of all deposited In atoms substituted for Si atoms without annealing.On the other hand, in the present experiments, 0.12 ML-In deposition onto the Si surface at 200 • C results in formation of nanoclusters.And In substitution for Si atoms is minor process.Of course, the difference of the In coverage may become an factor of the discrepancy between our results and Ref. [11].However, there is no In substitution for Si atom even on the bare HUC and on the corner Si adatom which does not contribute to form the nanocluster.Since the details of experimental setup are not described in Ref. [11], we cannot discuss why such discrepancy is observed.So, additionally, we investigated 0.03 ML-In deposition onto the Si(111)-7 × 7 reconstruction at RT with high deposition rate (1 ML-In/min) and low deposition rate (0.012 ML-In/min) (not shown here).In the both cases (the case with high deposition rate and the case with low deposition rate) of In deposition, almost all deposited In atoms existed on Si atoms in the Si(111)-7 × 7 reconstruction, only 5% of deposited In atoms substituted for Si atoms, and 15∼20% of deposited In atoms formed into regular In nanoclusters.The present results are different from that of Ref. [11], but are consistent with that of Refs.[10] and [12] very well.In Ref. [10], 0.06 ML-In was deposited onto the Si(111)-7 ×7 reconstructed surface at RT with a deposition rate of 0.05 ML/min, and it is thought that almost all deposited In atoms exist on Si atoms in the Si(111)-7 × 7 reconstruction.In Ref. [12], In was deposited onto the Si(111)-7 × 7 reconstructed surface at the substrate temperature of 100∼200 • C with a deposition rate of 0.01 ML/min, and even at 0.05 ML coverage, almost all In atoms formed regular nanoclusters on the FHUCs of the Si(111)-7 × 7 reconstruction.In either case of Ref. [10] and Ref. [12], In substitution for Si atoms was minor process.Some difference of experimental condition (such as the difference of the effect of radiant heat from In source to the sample and the difference of the amount of defects at the Si(111)-7 × 7 reconstruction etc.) might cause such discrepancy between our results and Ref. [11].However we cannot conclude what cause the discrepancy until now.If we dare to mention the discrepancy, it is thought that there is a chance that In atoms forming regular nanoclusters were counted as the In atoms substituting for Si atoms in Ref. [11], because, at low sample bias (less than ±0.6 V as in Ref. [11]), the HUC containing http://www.sssj.org/ejssnt(J-Stage: http://ejssnt.jstage.jst.go.jp) a regular In nanocluster shows bright spots on each In atom in the empty state, while the center spot of In nanocluster (the center spot can be seen in Fig. 1(b)) does not appear and the HUC shows dark spots on three edge adatom sites in the filled state [12].In other words, at low sample bias, the appearance of a HUC containing a regular In nanocluster is similar to that of a HUC containing three Si atoms in corner adatom sites and three In atoms substituting for Si atoms in edge adatom sites.

B. 0.24-ML-In deposition
The 0.24-ML-In was deposited onto the Si(111)-7 × 7 reconstructed surface with the same condition as the case of 0.12 ML-In deposition except the coverage of In. Figure 3 shows the STM images of the surface after In deposition.On this surface, almost all HUCs of the Si(111)-7 × 7 reconstruction are occupied by In nanoclusters.After all FHUCs have been occupied by In nanoclusters at the In coverage of 0.12 ML, increasing the coverage of In, Un-Faulted Half Unit Cells (UFHUCs) begin to be occupied by the nanoclusters.Further In deposition results in the formation of disordered clusters on the HUCs.It is thought that the irregular blobs observed in the filled state image (Fig. 3(b)) are caused by extra In atoms.In substitution for Si adatoms is not observed here.
Figure 4 shows STM images of the sample after annealing at 300 • C for 10 minutes following 0.24 ML-In deposition.In the filled state image (Fig. 4(b)), many center spots which are assigned to In-Si bonding states in the In nanocluster can still be seen on the HUCs.This is the evidence for existence of the original In nanoclusters.This result is quite different from the case of 0.12 ML-In deposition.The difference between the case of 0.12 ML-In deposition and the case of 0.24 ML-In deposition seems to be only whether UFHUCs are occupied by the In nanoclusters or not.Then, why is such a difference of In nanocluster in the thermal stability observed?Now we have to recall the four types of In adsorption mechanisms observed after annealing in the case of 0.12 ML-In deposition.We especially pay attention to the substitution of In atoms for Si atoms in the 7 × 7 adatom position (type (i) of In adsorption), since it is thought that this mechanism is irreversible at this substrate temperature.In the case of 0.12 ML-In deposition, In atoms activated by the annealing at 300 • C are sometimes released from the regular clusters.And, such activated In atoms prefer to substitute for edge Si adatoms.Therefore, the In nanoclusters are broken by annealing.On the other hand, in the case of 0.24 ML-In deposition, there are few edge Si adatoms since almost all edge Si adatoms are included in the In nanoclusters.Thus, the In atoms activated by the annealing stay in the regular clusters as semi-stable structure.On this surface, corner Si adatoms do not contribute to form the regular nanocluster.Thus, In substitution for the corner Si adatoms (indicated by arrows (i) in Fig. 4) and In adsorption onto the corner Si adatoms (indicataed by arrows (ii) in Fig. 4) can be seen.However, the number of the In nanoclusters after annealing has not changed compared with before annealing.Thus, it is thought that such In atoms substituting for corner Si adatoms come from irregular blobs which include extra In atoms rather than come from the regular clusters.The substituted Si adatoms may exist on the other Si adatoms or on the regular nanoclusters, but we can not conclude where the substituted Si adatoms are.From these results, it is found that In atoms prefer to substitute for the edge Si adatoms rather than form the regular nanoclusters, and prefer to form the regular nanoclusters rather than substitute for the corner Si adatoms.In other words, to avoid In substitution for edge Si adatoms is requisite for formation of the regular In nanocluster.In another experiment where the sample was maintained at 200 • C for 10 minutes after 0.12 ML-In deposition at 300 • C, In substitution for Si adatoms is predominant process while the formation of regular In nanoclusters is minor process (not shown here).This result coincides with the above interpretation.
The tendency of In atoms to form nanoclusters is similar to the case of Al deposition (Ref.[17]), but is not the same.In the case of Al, the formation of the regular http://www.sssj.org/ejssnt(J-Stage: http://ejssnt.jstage.jst.go.jp) nanocluster is preferred under the conditions where the Al substitution for edge Si adatoms can be seen.
On the sample after annealing at 350 • C for 10 minutes following the formation of 0.24 ML-In nanocluster array structure, there is no regular In nanocluster, and the progress of formation of the In-√ 3 × √ 3 reconstruction is observed.

IV. CONCLUSION
In conclusion, we observed four types of In adsorption mechanism after annealing at 300 • C following the forma-tion of 0.12 ML-In nanocluster array structure, and it is found that the substitution of In for Si adatoms is predominant adsorption process at this temperature.On the other hand, after annealing at 300 • C following the formation of 0.24 ML-In nanocluster array structure, almost all In nanoclusters can be maintained.On this surface, almost all edge Si adatoms are included in the nanoclusters, therefore In substitution for the edge Si adatoms is suppressed.It is the reason why the In nanclusters can be maintained after annealing at 300 • C. It is also found that to avoid substitution of In for the edge Si adatoms is one of the prerequisites to form ordered In nanoclusters.

FIG. 1 :
FIG. 1: A 30×30 nm 2 STM images of the sample after deposition of 0.12 ML-In.The sample biases are (a) +2.0 V (empty state image) and (b) −2.0 V (filled state image).The white line shows the unit cell of the Si(111)-7 × 7 reconstruction.An In nanocluster is indicated by an arrow.(c) The Atomic structural model of the unit cell containing an In nanocluster [12].
FIG. 2: A 28 × 28 nm 2 STM image of the sample after annealing at 300 • C following 0.12 ML-In deposition.The sample biases are (a) +2.0 V and (b) −2.0 V.The arrows (i) show an In atom substituting for a Si adatom.The arrows (ii) show an In atom adsorbed on a Si adatom.The arrows (iii) show an In atom adsorbed on a Si rest atom.And, an arrow (c) shows a regular In nanocluster.The In-√ 3 × √ 3 reconstruction can be observed on the step edge.
FIG. 3: A 30 × 30 nm 2 STM images of the sample after deposition of 0.24 ML-In.The sample biases are (a) +2.0 V and (b) −2.0 V.In addition to regular In nanoclusters, several irregular blobs are observed.

Volume 3 (
FIG. 4: A 30 × 30 nm 2 STM image of the sample after annealing at 300 • C following 0.24 ML-In deposition.The sample biases are (a) +2.0 V and (b) −2.0 V.Many regular In nanoclusters can be seen.The arrows (i) show an In atom substituting for a corner Si adatom.The arrows (ii) show an adsorbed atom on a corner Si adatom.Fig. (c) shows the atomic structural models of the Half Unit Cells containing the regular In nanocluster with enlarged STM images.The Half Unit Cell indicated by an arrow A has an In atom substituting for a corner Si adatom.The Half Unit Cell indicated by an arrow B has an adsorbed atom on a corner Si adatom and an In atom substituting for another corner Si adatom.