Atomic-scale Reconstructions on Metal and Semiconductor Surfaces Andrew Wee Surface Science Laboratory Department of Physics, NUS IMS Workshop, 27 Nov.

Atomic-scale Reconstructions on Metal

and Semiconductor Surfaces

Andrew WeeSurface Science Laboratory

Department of Physics, NUS

IMS Workshop, 27 Nov 04

Surface Science Lab NUS VT-STM/XPS/LEED system +

growth chamber with molecular beam & reactive atom sources [+ cryogenic STM]

Soft X-ray synchrotron end station on SINS beamline [+ growth chamber + STM/AFM]

Cameca IMS 6f Magnetic sector SIMS

VG ESCA MkII/SIMSLAB [EXAFS endstation]

Grand Challenge: Self assembly of single molecule devices

W Ho et al., Science, published online Sept. 4, 2003.

H Park et al., Nature, 417 (2002) 722

Scope of Presentation

1. Structure of Surfaces A rule for structures of open (high index) metal surfaces A high index surface: Cu(210) SiC(0001)-63x63 “honeycomb” reconstruction

2. Adsorbate-induced Reconstructions SiC(0001)-O Cu(210)-O; Cu(210)-Br

3. Surface as Template Monodispersed Co nanoparticles on SiC(0001) honeycomb

template Co ring clusters on Si(111)-(7×7)

1. Structure of Surfaces A rule for structures of open (high index) metal

surfaces A high index surface: Cu(210) Adsorbate-induced reconstructions: Cu(210)-O;

Cu(210)-Br SiC(0001)-63x63 “honeycomb” reconstruction

A rule for structures of open metal surfacesRef: YY Sun, YP Feng, CHA Huan, ATS Wee, Phys. Rev. Lett. 93 (2004) 136102.

Open metal surfaces: The coordination of the atoms in at least two layers is reduced when creating the surface; hence, more than one atomic layer is “exposed” to the vacuum.

Rule: “At bulk-truncated configuration, define a surface slab in which the nearest neighbors of all atoms are fewer than those in the bulk; in the process of relaxation, the interlayer spacing between each pair of atomic layers within this slab contracts, while the spacing between this slab and the substrate expands.”

Surface Slab

Bulk

Contracts

Expands

Density Functional Theory (DFT)

Kohn-Sham equation:

where the last term (the exchange-correlation) is not known exactly.

Various approximations are available. Among others, the LDA and GGA are most widely used.

LDA GGA

Methodology

Plane Wave Expansion:

Advantages: Simple mathematical formulism

Independency of basis set on ion positions

Availability of fast Fourier transform (FFT) between direct and reciprocal spaces

Pseudopotentials: Keep the eigenvalues and scattering properties unchanged compared with those of the real potential.

Softer in the core regions, hence fewer PW’s are needed for the expansion above.

Vienna Ab-initio Simulation Package (VASP) is a very efficient implementation of the pseudopotential plane-wave package.

A rule for structures of open metal surfaces

First-principles calculations:Based on density functional theory with either LDA or GGA approximation for the exchange-correlation functional

Ref: Sun YY, Phys. Rev. Lett. 93 (2004) 136102.

Physical picture: For more open surfaces, electrons from the deeper layers contribute to the smoothing, hence more layers relax.

A rule for structures of open metal surfaces

Further evaluation of the rule

Ni Cu Rh Pd Ag Ir Pt

(311) (- + …) (- + …) (- + …) (- + …) (- + …) (- + …) (- + …)

(331) (- - + …) (- - + …) (- - + …) (- - + …) (- - + …) (- - + …) (- - + …)

(210) (- - + …) (- - + …) (- - + …) (- - + …) (- - + …) (- - + …) (- - + …)

All fcc(311) surfaces have relaxation sequence (- + …)

All fcc(331) and fcc(210) surfaces have relaxation sequence (- - + …)

All these surfaces obey the rule.

Reference: Sun YY, Xu H, Feng YP, Huan ACH, Wee ATS, Surf. Sci. 548, 309 (2004).

Low Energy Electron Diffraction (LEED):Quantitative Determination of Surface Structure

LEED diffraction patternI-V data

collection

Guess a structure

Multiple scattering calculations

Reliability factor

Stop

Adjust

structure

Good

Bad

Q-LEED analysis

A high index surface: Cu(210)

Sun YY, Xu H, Zheng JC, Zhou JY, Feng YP, Huan ACH, Wee ATS, Phys. Rev. B 68 (2003) 115420

Clean Cu(210):

I-V LEED• Studied by layer-doubling LEED analysis and pseudopotential DFT calculations. • Excellent agreement between the calculated and measured I-V curves as judged by small Pendry R factor of 0.12.

LEED DFT

Δd12 (%) -11.1 -16.7

Δd23 (%) -5.0 -4.3

Δd34 (%) +3.7 +6.8

Δr12 (%) -1.9 -1.0

Δr23 (%) -1.9 -0.6

Δr34 (%) +0.6 +1.9

Sun YY, Xu H, Zheng JC, Zhou JY, Feng YP, Huan ACH, Wee ATS, Phys. Rev. B 68 (2003) 115420

Multilayer relaxation of Cu(210) surface: IV-LEED vs DFT

A high index surface: Cu(210)

cf. A rule for structures of open metal surfaces

Structure of 6H-SiC

Wide band gap semiconductor, very hard, good thermal conductor, chemical inert.

Structure: Si-C sp3 configuration, different Si-C bilayer stacking sequence and orientation, ≥200 polytypes, determine the physical property.

A

B

C

A

B

CA

B

CA

7.55Å

A

B

A

C

A

B

A

C

A

B

10.05Å

A

B

C

A

C

B

A

B

C

A

15.11Å

3C-SiC(111) 4H-SiC(0001) 6H-SiC(0001)

C atomSi atom

3C 4H 6H

Eg (eV) 2.3 3.2 3.0

Monodispersed Co nanoparticles on SiC(0001) honeycomb template 6H-SiC(0001) surface reconstruction6H-SiC(0001) surface reconstruction

(1x1) (3x3) (√3x√3) √3x√3) R30R30 (6√3 x 6√36√3 x 6√3 )R301170K 1230K 1250K

30 nm x 20 nm

Photoelectron spectroscopy data of SiC(0001) surface reconstructions

The C1s binding energy of carbon nanomesh is at 285.1 ev; graphite (HOPG) is at 284.4 eV.The well-developed carbon nanomesh surface is formed before the graphitization of the SiC surface. Therefore, the carbon nanomesh surface is not due one monolayer graphite.

Co

un

ts (

a. u

.)

288286284282280Binding energy (eV)

(a) Root 3 (950oC)

(b) Root 3 + nanomesh (1050oC)

(c) nanomesh (1100oC)

(d) nanomesh + graphite (1200oC)

(e) graphite (1300oC)

C 1s

h eV o

284.4 eV

282.9 eV

285.1 eV

Co

un

ts (

a. u

.)

290288286284282280278Binding energy (eV)

S1 (285.1 eV) S2 (283.8 eV) B (282.9 eV)

C 1s for the carbon nanomesh h eV

(a) o

(b) o

S1

S2

B

S1

S2B

C 1s of the carbon nanomesh surface

The carbon nanomesh is a honeycomb superstructure formed by the self-assembly of carbon atoms at high temperature.Two surface-related components for the carbon nanomesh surface have been identified with a binding energy of 283.8 eV and 285.1 eV, respectively.

Building the SiC(0001) honeycomb model

one-layer thick nanomesh; identical honeycomb cells

topmost Si atoms desorb

all the outermost surface atoms are C atoms

C atoms collapse, can substitute Si atoms below

Building model

Class III

III-12

Building model

Class III

III-13b

Unit cell parameters: a=b= 18.450Å, c= 20.0Å.

fixed

C atom

Si atom

H atom

Structure Optimization: force on ion < 10 meV/Å

~ 300 atoms; CPU time ~ 3 weeks

DFT-LDA Calculation results

STM images calculated to compare with experimental images

Partial charge density calculated

Smoothing techniques

STM images

DFT-LDA Calculation results

Model III-12

DFT-LDA Calculation results

Model III-13b

Relaxed structure of model III-12, 2x2x1 cell

DFT-LDA Calculation results

Relaxed structure of model III-13b, 2x2x1 cell

Simulated STM images

Model III-12, V=1.6 eV

(a) VT = 1.5V (b) VT = 1.8V

Model III-13b, V=1.6 eV

Explanation of PES Peaks

Relaxed nanomesh structure consists of graphene-like superstructure bonded to Si atoms below.

Cou

nts

(a. u

.)

290288286284282280278Binding energy (eV)

S1 (285.1 eV) S2 (283.8 eV) B (282.9 eV)

C 1s for the carbon nanomesh h eV

(a) o

(b) o

S1

S2

B

S1

S2B

C1s spectrum can be understood accordingly by:• a graphite-like C-C peak (S1)• an asymmetric low energy tail due to the boundary C atoms which have both C-C bonds and C-Si bonds (S2)• bulk SiC substrate with Si-C bonds (B)

2. Adsorbate-induced Reconstructions

SiC(0001)-OCu(210)-O; Cu(210)-Br

60 × 60 nm2 and detailed 9 ×7 nm2 (insert). (I = 0.30 nA VT = 2.2 V)

LEED, E=70eV

6H-SiC(0001)-3×3

43BA12 1243 6 5

T3T2T1

T0

T3T2T1

T0

T3T2T1

T0

Tetra-cluster

Bulk Si

3

62

1

4

B

5

A

43

2

1

T

T0 T

TT

T

T

T

T

TT

TT

T

T

T

T0T0

T0

T1

T2

T3

T1

T2

T1

T2

T1

T2

T3

T3

T3

1

2

3

2

3

A

4

B

4

5

6

60º 60º30º

Si adlayer

1st-layer2nd-layer

Bulk layer

Top view

Side view

6H-SiC (0001) 3×3 twisted reconstructed modelU. Starke et.al, PRL, 80, 758 (1998); PRB, 62, 10335 (2000).

Initial oxidation mechanism

O2

F. Amy, et. al., Phys. Rev. Lett. 86, 4342 (2001)

O2

O2 reacts with the third Si-layers.

Dangling bond

Si-adatom is much more active.

Si-adatom sites or the third Si-layers?

In-situ oxidation with low tunneling current to minimize the inelastic tunneling electron scattering induced reactions.

Bright sites appear after O2 exposure, and keep increasing.

I = 0.10 nA, VT = 2.2 V 2 nm

Clean Surface 0.2 L O2 1.0 L O2 2.0 L O2

**

**

* *

*

Dark sites appear initially, saturated after 1.0L O2 exposure.

Initial oxidation mechanism

Si adatom+trimer

Si Si SiOr

O2, initial

O2 attach on the dangling bond of Si adatom. Dark sites, O2 depletes the DOS of Si atom

Explanation

Si

O2 inserts into the back bonds of Si adatom. Bright sites, Si atom is lifted by 0.5 Å. Thermal stable sites.

More O2

DFT simulations (Using CASTEP codes)

A1

3

62

1

4

5

A

O

O

A2

3

62

1

4

5

AOO

A12 43 6 5A12 43 6 5

2-O-1=121.2o

2-O-A=128.4o

Top view

Side view

T1

T2

T3

T1

T2

T3

T0

T1 T3T2

T0

T1 T3T2

T0

T1 T3T2

2-O-A=119.5o

A-O-6=120.8o

Models where O2 reacts with the third Si-layer

43B 126 5

B-O-3 =120.1o 3-O-4

=118.5o

6

B

5

43

2

1

O

A3 A4

43B 126 5

B-O-3=117.7o B-O-1

=119.4o

6B

5

43

2

1O

O

Top view

Side view

T0

T1 T3T2

T0

T1 T3T2

T0

T1 T3T2

T0

T1

T3T0

T1

T3

T0

T1

T3

T0

T1

T3

T0

T2

T3T0

T2

T3

T0

T1 T3T2

O

Models where O2 reacts with the third Si-layer

6

321 6

54

321 6

54

C3

21

54

C4

6

C1

3254321 6

54

+

321 6

54

+

O2

T0

T1

T2 T3

T0

T1

T2

T0

T1

T2 T3

T0

T1

T2 T3

T0

T1

T2 T3

T0

T1

T2 T3

Tetra-cluster

1

C2

T3

Models where O2 reacts with Si-adatoms

Oxygen coverage: C

(ML)

Surface models

Chemisorption energy:

∆E (eV/unit cell)

x=1, C = 2/93×3:2O surface

C1C2C3C4A1A2A3A4

-4.10-4.32-5.61-6.93-3.52-3.48-3.50-3.51

The model where O2 insets into the back bonds of the Si-adatoms is thermally most stable!

Chen W, Xie XN, Xu H, Wee ATS, Loh KPAtomic scale oxidation of silicon nanoclusters on silicon carbide surfacesJ PHYS CHEM B 107 (42): 11597-11603 OCT 23 2003

Superstructure formation in the Cu(210)-O system

1000 x 1000 Å2 image of (2x1) reconstruction

Wee ATS, Foord JS, Egdell RG, Pethica JB, Phys. Rev. B 58 (1998) R7548.

(2x1)

(a)

FHS-MR LBS-BRFHS-BR

LBS-MR

L01

[001]

[120]L02

L03

d02

Definition of parameters for LBS-MR

d01 d12

[120]

d03

LBS-MR (oxygen at long bridge site with missing row), LBS-BR (long bridge site with inward buckled row), FHS-MR (four-fold hollow site with missing row) and FHS-BR (four-fold hollow site with inward buckled row)

Tan K. C., Guo Y. P., Wee A. T. S. and Huan C. H. A., Surf. Rev. Lett. 6 (1999) pp. 859-863

O-Cu(210) adsorbate induced reconstructions

Adsorbate-induced Surface Reconstructions

LEED study of oxygen-induced reconstructions on Cu(210)

Buckled (3x1) reconstruction – 2/3 ML

Guo YP, Tan KC, Wang HQ, Huan CHA, Wee ATS, Phys. Rev. B 66 (2002) 165410.

1st Cu-O row (side view) 2nd Cu-O row (side view)

d01’=-0.17Å

D22=0.17Å

D33

D44

d12’

1.03Å (+27.6%)

d23’

0.78Å (-3.4%)

d34’

d01=0.12Å

D22=0.17Å

D33

D44

d12

0.99Å (+22.6%)

d23

0.70Å (-13.3%)d34

L01’=0.54Å

L01=0.25Å

L00=4.84Å(+19.9%)

[001]

[120]

Top view

Adsorbate-induced Surface Reconstructions

(a) (a) 2000 x 2000 Å2 (VB = -1.0 V, IT = 2.5 nA),(b) (b) 300 x 300 Å2 (VB = -1.0 V, IT = 0.30 nA) images after 500 L RT oxygen exposure and

subsequent annealing to 620 K for a few minutes. Analysis of corrugation profiles shows that A and C are at the same height, whereas B is one unit cell below and D one above.

[ ]121

[ ]12 1 [001]

(a)

[001]

(b)

A

B

C

D

Cu(210)-O superstructures

Adsorbate-induced Surface Reconstructions

Cu(210)-Br system 200 x 200 Å2 images of the triangular checkerboard

recorded at VB = -1.0 V, IT = 0.1 nA, showing an inversion of the triangles during different scans but using the same tunnel current and sample bias.

Wee ATS, Fishlock TW, Dixon RA, Foord JS, Egdell RG, Pethica JB, Chem. Phys. Lett. 298, 146 (1998)

[001]

[ ]121

Cu(100)-Br system T.W. Fishlock, J.B. Pethica and R.G. Egdell,

Surf. Sci. 445, L47 (2000)

Adsorbate-induced Surface Reconstructions

Adsorbate-induced Surface Reconstructions

Cu(100)-N system

Adsorbate induced nanostructures also observed in Cu(110), Cu(111)-N systems

F. M. Leibsle, Surf. Sci. 514, 33 (2002)

3. Surface as TemplateMonodispersed Co nanoparticles

on SiC(0001) honeycomb templateCo ring clusters on Si(111)-(7×7)

Self-assembly in a Honeycomb template?

16×16nm2 STM filled state images for the carbon nanomesh with:

(a) 0.1Å Co coverage

(b) 0.2Å Co coverage

(c) line profile 1 for (a) and line 2 for clean surface. VT=2.5V

(a) (b)

4nm

1.7Å 0.1ÅCo

nanomesh

(c)

(1)

(2)

Monodispersed Co nanoparticles on SiC(0001) honeycomb template

At the lower coverage (0.1Å Co), the clusters will adsorb on these active sites, with a diameter of 1.4±0.2nm and a height of 1.7±0.1Å.

At the higher coverage (2.0Å Co), neighbouring Co clusters will coalesce to form big clusters, 3.4±0.2 nm in diameter and 3.3±0.1Å in height.

• Monodisperse Co nanoclusters can be fabricated on SiC honeycomb template under submonolayer condition.

• Boundaries of honeycomb structures serve as active sites for Co cluster growth.8nm×8nm STM image: blue circles

highlight the Co cluster adsorption sites.

Monodispersed Co nanoparticles on SiC(0001) honeycomb template

References: W Chen, KP Loh, H Xu, ATS Wee, Appl. Phys. Lett. 84 (2004) 281 W Chen, KP Loh, H Xu, ATS Wee, to appear in Langmuir.

cf. Boron Nitride NanomeshM. Corso et al., Science, 303 217 (2004)

Hole formation is likely driven by the lattice mismatch of the film and the rhodium substrate.

This regular nanostructure is thermally very stable and can serve as a template to organize molecules, e.g. C60 molecules.

The BN nanomesh was formed by deposition of B3N3H6 on Rh(111).

Co ring clusters on Si(111)-(7×7)

FU

AB

C

C′

Co ring clusters on Si(111)-(7×7)

STM simulation

M.A.K. Zilani, Y.Y. Sun et al., in preparation

Empty state: 1.9 V , 0.1 nA

Published work on other nanotemplates

In nanocluster array formed on Si(111)-7×7 surface. J. L. Li, PRL, 88, 066101 (2202)

3nm

The hexagonal networks were formed by co-deposition of PTCDI and melamine molecules on Ag/Si(111). J. A. Theobald, Nature, 424, 1029 (2203)

Acknowledgements

Current students: Md. Abdul Kader Zilani Qi Dongchen

Past students: Ong Wei Jie Tan Kian Chuan Wang Huiqiong Dr Zheng Jincheng Dr Sun Yiyang* Dr Chen Wei*

* Currently Research Fellow

Research Fellows: Dr Xu Hai Dr Liu Lei Dr Guo Yong Ping Dr Xie Xianning Dr Gao Xingyu

Collaborators: Dr Loh Kian Ping Dr Tok Eng Soon Dr Wang Xuesen A/P Alfred Huan A/P Feng Yuan Ping