Electrochemical growth of gold on well-defined vicinal H-Si(1 1 1) surfaces studied by AFM and XRD

Article abstract of DOI:10.1016/S0039-6028(03)00563-6

The possibility of preparing structurally well-defined and periodically arranged metal islands on a semiconductor is very attractive for optical and magnetic applications. In this work we used electrochemical deposition to grow nm-sized gold islands that decorate steps on defect-free vicinal H terminated Si(1 1 1) surfaces prepared by chemical etching. The gold deposits were studied by atomic force microscopy and X-ray diffraction. Results show that gold nucleates exclusively along the steps and that the density of nuclei is controlled by the electrode potential. Nearly prefect replication of the periodic array of straight monatomic steps is achieved at sufficiently negative potential. XRD indicates that the structure of gold films evolves from powder-like, close to the onset potential of nucleation (-1.6 V), to strongly epitaxial with the (1 1 1) orientation at more negative potentials. A reaction model and a growth mechanism are proposed to account for the origin of the selective nucleation and the excellent epitaxy obtained. In particular they discuss whether or not the H-monolayer remains intact under the deposit.

Full text of DOI:10.1016/S0039-6028(03)00563-6

Surface Science 537 (2003) 95–112
www.elsevier.com/locate/susc
Electrochemical growth of gold on well-defined vicinal
H–Si(1 1 1) surfaces studied by AFMand XRD
b
M.L. Munford a,b, F. Maroun , R. Cortes a,1, P. Allongue a, ,1, A.A. Pasa
a,1
*
ꢀ
a
Laboratoire de Physique des Liquides et Electrochimie (CNRS UPR-15), Universiteꢁ P & M Curie, 4 Place Jussieu,
F-75005 Paris, France
b
Laboratorio de Filmes Finos e Superficies, Departemento de Fisica, UFSC, 88040-900 Florianopolis, SC, Brazil
Received 8 January 2003; accepted for publication 8 April 2003
Abstract
The possibility of preparing structurally well-defined and periodically arranged metal islands on a semiconductor is
very attractive for optical and magnetic applications. In this work we used electrochemical deposition to grow nm-sized
gold islands that decorate steps on defect-free vicinal H terminated Si(1 1 1) surfaces prepared by chemical etching. The
gold deposits were studied by atomic force microscopy and X-ray diffraction. Results show that gold nucleates ex-
clusively along the steps and that the density of nuclei is controlled by the electrode potential. Nearly prefect replication
of the periodic array of straight monatomic steps is achieved at sufficiently negative potential. XRD indicates that the
structure of gold films evolves from powder-like, close to the onset potential of nucleation ()1.6 V), to strongly epitaxial
with the (1 1 1) orientation at more negative potentials. A reaction model and a growth mechanism are proposed to
account for the origin of the selective nucleation and the excellent epitaxy obtained. In particular they discuss whether
or not the H-monolayer remains intact under the deposit.
ꢀ 2003 Elsevier Science B.V. All rights reserved.
Keywords: Silicon; Gold; Electrochemical methods; Epitaxy; Growth; Atomic force microscopy; X-ray scattering, diffraction, and
reflection
1. Introduction
on pre-structured 7 · 7-Si(1 1 1) clean surfaces [1]
and on vicinal 7 · 7-Si(1 1 1) surfaces that were
functionalized by pre-depositing gold and CaF2
[2,3]. Metal deposition on silicon is such an impor-
tant technological step that it prompted numerous
studies on clean silicon surfaces and on H-termi-
nated silicon surfaces [4], especially after the work
of Chabal and co-workers [5] who established that
H–Si surfaces are technologically relevant and
stable in air. The comparison of metal evaporation
on the two types of silicon surfaces shows that the
presence of the H-monolayer is sufficient to pro-
mote drastic changes in the growth modes, from a
The preparation of well-defined and periodically
arranged metal islands on a semiconductor is very
attractive for optical and magnetic applications.
Most recent works considered metal evaporation
* Corresponding author. Tel.: +1-6933-4431; fax: +1-6933-
3004.
E-mail address: pa@pmc.polytechnique.fr (P. Allongue).
1
Permanent address: CNRS––UMR 7643, Ecole Polytech-
nique Lab., Physique de la Matiere Condensee, F-91128
Palaiseau, France.
0039-6028/03/$ - see front matter ꢀ 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0039-6028(03)00563-6

96
M.L. Munford et al. / Surface Science 537 (2003) 95–112
layer-by-layer process onto clean silicon to a Vol-
mer–Weber mechanism on the H-terminated sur-
face (for a review see [4]). Epitaxy of the metal film
on H–Si(1 1 1) was reported in few cases such as Ag
and Au for instance. A key issue, often debated,
concerns the eventual segregation (or removal)
of the hydrogen layer under the deposit [4].
of the stepped surface structure is achieved, under
optimum conditions of polarization, by highly
selective nucleation of a large density of gold
clusters at the silicon step edges. Subsequent
growth is 3D. A reaction model and a growth
mechanism are proposed to account for observa-
tions. They address, in particular, whether or not
the H-monolayer remains intact under the de-
posit.
Metal deposition on H-terminated silicon was
also performed using electrochemical growth. The
method constitutes an alternative method, which is
widely used in microelectronics because it is cost
effective and allows fast growth rates. The growth
modes are also simply modified by changing the
applied potential and the solution composition.
Recent studies of this kind concerned the deposi-
tion of copper [6–9], lead [7,10–12], gold [13–15],
magnetic layers Co–Ni–Cu/Cu [16,17], cobalt
[18,19] and nickel [20] on Si(1 1 1) or (1 0 0). Iron
was also deposited on porous silicon [21]. Metal
plating was also performed without applying an
external bias (electroless deposition) from fluoride-
based solutions. The electroless deposition of Cu/
Si(1 1 1) [22], Au/Si(1 1 1) [7,23], Pt and Ni [24,25]
has in particular been investigated. These electro-
chemical studies often focused on one specific
point of the film formation such as the growth
modes using current transient measurements or
STM/AFM observations. In some studies the
structure of the deposits was studied by X-ray
techniques. Very few works combined different
techniques and, to the best of our knowledge,
none succeeded in controlling the spatial arrange-
ment of the metallic nanostructures on the silicon
surface.
This works shows for the first time that it is
possible to prepare well defined and periodically
arranged arrays of gold nm-islands by template
electrodeposition on a defect free vicinal H-ter-
minated Si(1 1 1) surface. The substrates, with a
stepped structure close to perfection, were ob-
tained by controlled chemical etching [26] and
the ultrathin gold layers (1–20 ML) were elec-
trodeposited from an alkaline KAu(CN)2 solu-
tion. The gold films were characterized in details
by ex-situ AFM, X-ray diffraction (XRD) and
electrochemical transient techniques. Results reveal
a strong potential-dependence of the film mor-
phology and structure. The long-range replication
2. Experimental
2.1. Substrate preparation
Silicon samples were cut from 1 to 10 X cm
(1 1 1) wafers (n-type, P doped) with a miscut angle
a ¼ 0:2°, 0.4° or 2° precisely oriented towards
Æ1 1 )2æ. They will be referred to as Si-0.2°, Si-0.4°
and Si-2° in the following. The careful adjustment
of the miscut orientation is a prerequisite to obtain
straight and parallel monatomic steps (Fig. 1a)
terminated by monohydride sites (Fig. 1b) [26].
The step-to-step distance of W ꢀ 140 nm mea-
sured in Fig. 1a corresponds to a misorienta-
tion 0.13°, which is very close to the miscut angle
derived from XRD on this substrate (tanðaÞ ¼
W ðnmÞ=h, where h ¼ 0:314 nm is the height of a
step) [26]. Prior to etching, samples were cleaned in
(98% H2SO4)/(30% H2O2) 2:1 mixture and rinsed
with bidistilled water. The H-termination was ob-
tained by chemical etching in oxygen-free 40%
NH4F [27–29] to avoid the formation of triangular
etch pits on the (1 1 1) terraces. We used 50 mM
(NH4)2SO3 as oxygen scavenger [26,28]. A final
rinse in bidistilled water was performed after
etching. Prior to use the ohmic contact was formed
by applying an InGa eutectic on the rear face. The
sample was then mounted with its lateral edges
protected by an electrolytic scotch tape to expose
only the well-defined (1 1 1) face to solution.
2.2. Gold deposition on Si(1 1 1)
Gold was deposited from 5 mMKAu(CN) 2 + 10
mMNaCN in 2 MNaOH, using a three-electrode
electrochemical cell connected to a potentiostat

M.L. Munford et al. / Surface Science 537 (2003) 95–112
97
2.3. Atomic force microscopy
The morphology of the gold deposits was
characterized by contact mode atomic force mi-
croscopy, AFM(Molecular Imaging microscope,
Phoenix, USA) in a nitrogen atmosphere. Si3N4
cantilevers (Nanoprobes, spring constant 0.12
NmÀ1) were used. In general, stable imaging was
obtained, except when the deposition potential
was too close to the onset of nucleation (U ꢀ À1:6
V). In that case, the Au clusters were wiped by
the AFMtip.
2.4. X-ray diffraction
A home-built five-circle diffractometer was used
ꢂ
with a Ka1 Cu source (k ¼ 1:5405 A). The sample
configuration is presented in Fig. 2. The X-ray
beam is horizontal and parallel to XX0. Before
measurements, the sample optical plane was care-
fully adjusted to bring the normal n parallel to the
rotation axis ZZ0 by minimizing the deflection of a
laser beam upon in-plane rotation of the sample.
To diffract on Au(1 1 1) planes parallel to the
surface, h-scans were performed by varying the
sample angle h around the diffraction angles
hSið1 1 1Þ or hAuð1 1 1Þ with the detector positioned for
Bragg conditions (h ¼ 2hSið1 1 1Þ or 2hAuð1 1 1Þ). For
grazing incidence measurements or U-scans, the
sample angle was h ¼ 0:6° and the detector posi-
tioned in Bragg conditions for {1 1 )1} planes of
gold or silicon. The settings (h, AZ) for the two
types of experiments are given in Table 1. For a
quantitative comparison of different deposits the
XRD spectra were normalized with respect to the
intensity of the Si(1 1 )1) peak.
Fig. 1. (a) AFMimage (5 lm · 5 lm) of a H–Si(1 1 1) surface
with 0.13° miscut angle. (b) Ball model of the Si(1 1 1) surface
with two terraces separated by one step containing two kink
sites. Other sites are monohydride sites. Dark and white spheres
are Si and H atoms. The zigzagging line indicates the step sites.
(Ecochemie, The Netherlands). The reference
electrode was a mercury sulfate electrode (MSE)
and all potentials are quoted against this reference
of potential. The counter electrode was a Pt grid.
The procedure for deposition was as follows: im-
mediately after etching the sample was quickly
mounted (see above) and manually immersed into
the solution at the desired pre-set deposition po-
tential to avoid any chemical etching. After de-
position, the sample was quickly removed from the
solution, rinsed with bidistilled water, and blown
dried with nitrogen.
2.5. Rutherford backscattering (RBS) measure-
ments
The film thickness was determined from RBS
measurements, using the 2 MeV van de Graaf
accelerator at the Groupe de Physique des Solides
ꢁ
(Universite Paris 7). The surface density of Au
atoms was converted into an average thickness
expressed in ML, with 1 ML ¼ 1:38 Â 1015 atoms/
cm2. The film thickness was also derived from

98
M.L. Munford et al. / Surface Science 537 (2003) 95–112
(a) θ-scan : diffracting planes (111)
0.0
-0.2
-0.4
-0.6
without KAu(CN)2
1.0
with KAu(CN)2
h = 2θ
0.8
0.6
sample
RX
θ
X’
X
-2.0
-1.8
-1.6
U (V)
(b) -scan : diffracting planes {-111}
-2.0
-1.9
-1.8
-1.7
-1.6
Φ
U (V)
Z
Fig. 3. Voltammograms of
a n-type H–Si(1 1 1) in 2 M
n
NaOH + 10 mMNaCN, in the presence (––) and in the absence
(- - -) of KAu(CN)2. The sweep rate is 15 mV/s. Inset: variations
of the deposition yield as a function of the deposition potential.
Data correspond to 7 s deposits at each potential.
h
RX
AZ
X’
X
θ = 0.6˚
Φ
Z’
very small for U > À1:5 V (not shown) due to the
rectifying behavior of the interface n-Si/solution.
Direct polarization regime or accumulation of
electrons corresponds to U < À1:5 V, which is
slightly negative of the free potential of the surface
Fig. 2. Definition of angles in XRD experiments. (a) h-scans:
the detector is positioned at h ¼ 2hAuð1 1 1Þ (AZ ¼ 0) to diffract on
Au(1 1 1) planes and the sample angle varied around hAuð1 1 1Þ
;
(b) U-scans: the sample plane is carefully adjusted to bring the
surface normal n parallel to the rotation axis ZZ0. The angle U is
varied with the sample normal n tilted by 0.6° and the detector
position set at h and AZ to diffract on Au{1 1 )1} planes.
2
in this solution. In the blank solution (no
KAu(CN)2 added, dotted line), the small cathodic
current indicates that the kinetics of water de-
composition is very slow on H–Si(1 1 1). In the
gold solution (solid line), the sharp rise of the
current at )1.57 V corresponds to the nucleation
onset of gold deposition. The presence of a peak of
current at )1.65 V indicates that deposition be-
comes limited by mass transport of Au(CN)2À
species in solution. For U < À1:9 V, the decom-
position of water becomes significant on the gold
deposit. The return scan, which corresponds to
gold on gold deposition, indicates that the growth
of gold on gold is easier than the nucleation of
gold on H–Si(1 1 1) since the current is not zero for
U > À1:57 V.
Table 1
Experimental conditions used in X-ray characterizations
Experiment
Planes
h
AZ
h
h-scan
Si(1 1 1) 14.221
Au(1 1 1) 19.087
0
0
28.443
38.173
U-scan
Si(1 1 1)
Au(1 1 1)
0.6
0.6
26.9
36.38
9.35
12.46
See Fig. 2 for the definition of parameters.
integration of the deposition current (using Fara-
dayÕs law).
The deposition was performed by immersing the
sample under potential control. The conditions
will be defined in the following, either by the ap-
3. Results
3.1. Deposition procedure
The voltammograms of an n-type H–Si(1 1 1)
electrode in the presence and absence of
KAu(CN)2 are shown in Fig. 3. The current was
2
The region of potentials U > À1:5 V was not explored to
avoid the passivation of the surface.

M.L. Munford et al. / Surface Science 537 (2003) 95–112
99
plied potential U quoted against the reference
electrode or by the overpotential g ¼ U À U0 > 0,
with U0 ¼ À0:82 V is the Nernst potential of the
reaction Au(CN)À2 + eÀ fi Au (U0 was measured as
the rest potential of a gold wire in an oxygen-free
gold solution). The yield of deposition was deter-
mined by dividing the gold thickness derived from
RBS by the gold thickness derived from FaradayÕs
law using the conversion factor 1 ML ¼ 220 lC/
cm2 (one electrons is exchanged per deposited gold
atom). The curve in inset of Fig. 3 shows that the
deposition yield decreases continuously from uni-
ty, close to the nucleation onset ()1.57 V), to 0.6 at
)1.93 V. The results of Fig. 3 were not significantly
depending on the miscut angle of the substrate.
The transient electrochemical current iðtÞ mon-
itored during deposition is shown in Fig. 4a and b
for two potentials. Solid and dashed lines corre-
spond to Si-2° and 0.2°. Two peaks are observed.
The narrow one (t < 0:1 s) corresponds to the
immersion of the sample under potential control.
A broader peak occurs after some delay. This peak
is related to the nucleation and growth modes and
was characterized by its maximum (absolute
value) i ¼ imax at t ¼ tmax. For a given miscut, imax
increases and tmax decreases with increasing g. At a
given potential, imax increases and tmax decreases
with increasing miscut. Using imax and tmax, the
2
dimensionless transients ði=imax
max
Þ
vs. ðt À t0Þ=
ðt À t0Þ
were plotted (see the review of Oskam
and Searson [14] for more details) to identify the
nature of the nucleation and growth process. The
time t0 represents an induction time corresponding
to the delay of current rise. Fig. 4c shows that t0
decreases with increasing g and that there is no
significant influence of the step density. In Fig. 4d
all the dimensionless transients measured at differ-
ent potentials on Si-0.2° merge into one single plot
(symbols), which fits very well with the hypothesis
of a progressive (solid curve) and not with an in-
stantaneous (dotted line) nucleation process (the
expression of model curves is given in [13]). The
same result is obtained with the Si-0.4° and Si-2°
substrates.
0.0
0.8
U = -1.67 V
(a)
(c)
Si-0.2°
-0.5
-1.0
0.6
Si-2°
0.4
Si-0.2°
Si-2°
0.2
-1.5
-2.0
0.0
-1.90 -1.80 -1.70 -1.60
U (V)
0
1
2
3
4
5
0.0
-1.0
-2.0
-3.0
-4.0
1.0
U = -1.92 V
(b)
(d)
0.8
instantaneous
0.6
0.4
0.2
0.0
progressive
Si-0.2°
Si-2°
Si-0.2°
0.0 0.2 0.4 0.6 0.8 1.0
0
1
2
3
4
t (s)
(t — t0) / (t — t0)max
Fig. 4. (a and b) Current transients for U ¼ À1:67 and )1.9 V recorded in the same solution as in Fig. 3. Solid and dotted lines
correspond to Si-2° and Si-0.2° substrates. (c) Potential dependence of the induction time t0. (d) Dimensionless transients for the
substrate Si-0.2°. See text for explanation. Solid and dashed lines are model curves corresponding to progressive and instantaneous
nucleation [13].

100
M.L. Munford et al. / Surface Science 537 (2003) 95–112
3.2. AFM observations
confirm that the steps are indeed running from top
to bottom in the images and demonstrate that the
gold clusters are precisely located on them. In Fig.
5a–c the island density is nearly independent of
time (ꢀ1.3 · 1010 cmÀ2) and their average height
increases from 19 nm (approximately 80 ML) to 27
nm (114 ML) while their width increases from 47
to 90 nm at FWHM. The islands present therefore
an aspect ratio height/width ꢀ 0.4. At the end of
the growth some of the islands take a well-defined
geometric shape (e.g. see flat top triangles in Fig.
5c). This point will be further discussed below.
Two 9–10-ML thick gold deposits grown at
)1.92 V on Si-0.2° and Si-0.4° are shown in Fig. 6.
The gold islands are again distributed along the
steps. Comparing Fig. 6a with Fig. 5c reveals that
changing the deposition potential from )1.73 V to
)1.92 V increases the island density by a factor 2
In this section we will show that gold deposition
follows a Volmer–Weber mechanism with the
formation of 3D clusters that exclusively decorate
the silicon steps for U À 1:93 V < U < À1:60 V.
The location, density and morphology of the gold
clusters were studied as a function of the potential,
deposition time and step density.
Fig. 5 shows AFMimages of 5.2, 7.2 and 18.4
ML-thick gold films deposited on Si-0.2° at
U ¼ À1:73 V. The films consist in nm-sized clus-
ters which are aligned along parallel lines sepa-
rated, on average, by 90, 115 and 100 nm,
respectively in images (a–c). These distances are
close to the nominal terrace width on this substrate
(W ꢀ 90 nm for 0.2°). The image contrast was
enhanced in Fig. 5d to resolve the silicon steps and
Fig. 5. AFMimages of gold deposits on Si-0.2 ° (U ¼ À1:73 V). The thickness is respectively 5.2 ML (a), 7.2 ML (b) and 18.4 ML (c).
Images are (2 lm · 2 lm). In (d) the image contrast is enhanced to show that the gold islands are located at steps (1.5 lm · 1.5 lm
image). In all images the steps are running vertically from top to bottom in images and are descending from right to left. Their position
is identified by arrows in (d). The bar is 500 nm for all images.

M.L. Munford et al. / Surface Science 537 (2003) 95–112
101
Fig. 6. AFMimages of 9–10 ML-thick gold deposits on Si-0.2 ° (a) and Si-0.4° (b). The deposition potential is U ¼ À1:92 V in both
cases. The silicon steps are running vertically in (a) and oblique in (b). Note that only few Au islands have nucleated on (1 1 1) terraces
(see arrows in a) and that the density of islands along the steps is nearly the same in both images.
and that this increase merely corresponds to an
increase of the linear island density along the steps,
from 105 to 2 Â 105 clusters/cm, because only very
few islands have nucleated on the (1 1 1) terraces at
)1.92 V (arrows in Fig. 6a). Increasing the over-
potential reduces both the size and the size dis-
persion of clusters. At U ¼ À1:92 V the average
height is only 12 nm and the lateral size 55 nm. On
Si-0.4° the island density is 5:2 Â 1010 cmÀ2 at the
same potential of )1.92 V (Fig. 6b), which is twice
as large as on Si-0.2° (Fig. 6a). The factor 2 scales
exactly with the doubled step density on Si-0.4°
and means that there are also 2.5 Â 105 islands/cm
along the steps.
Fig. 7 presents the height derivative of the to-
pography of a deposit performed at U ¼ À1:73 V
during 5 s. The islands appear as well defined flat
top polyhedrons with adjacent edges separated by
an angle of 60° or 120°. Triangles (marked by
triangles) and trapezes (marked with trapezes) are
the most frequently encountered shapes. Hexago-
nal islands are more scarcely observed. A careful
analysis of the islands shows that their edges are
parallel to the preferential directions shown in in-
set of image 7a. These directions are 60° apart
(sixfold symmetry) and one of the direction is
parallel to Æ1 )1 0æ or, equivalently, to the step
direction. A close examination of the triangular
islands in Fig. 7a shows that there are two types of
islands. Triangles of type (A) and (B) which are
respectively pointing to the right and the left of the
image. Zooming inside the rectangular boxes (Fig.
7b–c) evidences that triangles of type (A) and (B)
are overlapping preferentially with the upper ter-
race and the bottom terrace (see also scheme).
3.3. X-ray diffraction measurements
A h-scan close to the Bragg conditions for
Si(1 1 1) planes parallel to the surface (not shown)
gives a peak at the expected angle hSið1 1 1Þ ¼ 14:22°
(Table 1). A U-scan (not shown) with the detector
in Bragg position for Si{1 1 )1} planes exhibits
three peaks at 104.89°, 224.89° and 344.89° cor-
responding to (1 1 )1), (1 )1 1) and ()1 1 1) dif-
fracting planes. After gold deposition a diffraction
peak is found in h-scans at the expected value
hAuð1 1 1Þ ¼ 19:09° for Au(1 1 1) planes parallel to
the surface (Table 1), meaning that the deposit
presents a (1 1 1) texture. In U-scans with the de-
tector in Bragg position for Au{1 1 )1} planes
(Table 1) 6 peaks are observed: three of them
(110.07°, 230.07° and 350.07°) are close to the
peak position of silicon and will be referred to as
Ôeven peaksÕ in the following. The other three
peaks, shifted by 60°, will be referred to as Ôodd

102
M.L. Munford et al. / Surface Science 537 (2003) 95–112
Fig. 7. (a) Height derivative AFMimage (2 lm · 2 lm) of a gold deposit on H–Si(1 1 1)-0.2° (U ¼ À1:73 V, 5 s). Triangles and trapezes
outline triangular and trapezoidal islands. The sixfold orientations taken by the edges of polyhedron islands are given in inset. (b) and
(c) Zoom corresponding to rectangular boxes in (a) to evidence the two sorts of triangular islands. The dotted lines mark the step
position. The islands of type (A) and (B) overlap preferentially with the upper and the bottom terrace respectively (see also schematic
cross section).
peaksÕ. The sixfold symmetry of U-scans indicates
that the deposit is in epitaxy with the surface. The
even peaks correspond to the relationship of epit-
axy Au(1 1 1)[1 )1 0]kSi(1 1 1)[1 )1 0] (with respect
to bulk axis of silicon) and the odd ones to gold
clusters that are rotated by 180°.
rection on h-scans is investigated in a later para-
graph.
3.4. Thickness dependence of h- and U-scans
Fig. 8 shows the h-scans of the gold deposits
imaged in Fig. 5 (U ¼ À1:73 V, Si-0.2°). All scans
were recorded with U ¼ 90°, i.e. with the X-ray
beam parallel to steps. The gold thickness tAu
ranges from 5.2 to 18.4 ML as indicated in the
figure. The strong increase of the peak intensity
with growth (Fig. 8a), scales linearly with tAu in
Fig. 8b, with the intercept of the plot passing
through the origin. More remarkable is the cor-
responding narrowing of the peak linewidth from
2° to 1°. The U-scans of the same deposits are
given in Fig. 9. They exhibit 6 peaks, as explained
above, which intensity increases linearly with the
gold thickness (not shown). A quantitative analy-
sis of the spectra (Fig. 9b) evidences a significant
improvement of the film mosaicity with tAu be-
cause the peak linewidth decreases from 6.6° to
4.5°. Fig. 9b plots also the ratio R of the integrated
intensity of the 6 peaks (after background correc-
tion) to the total integrated signal. R gives the
proportion of the deposit which is in epitaxy with
As a preliminary remark it should be noted that
the linewidth of XRD peaks is not limited by the
grain size. According to the Scherer formula, the
broadening (in degrees) of the linewidth is
ꢀðp=180Þk=ðL Ã cosðhÞÞ where L is the dimension
ꢂ
of the grain in the diffraction plane and k ¼ 1:54 A
the X-ray wavelength. In the case of h-scans, L is
equal to the height of clusters. In the case of U-
scans, the {1 1 )1} diffracting planes are tilted by
57° and L ꢀ 1:2 · height because the aspect ratio of
grains is smaller than unity. Hence, for most de-
posits studied here the peak linewidth of XRD
spectra reflects the film structure since the Scherer
broadening remains <0.4° for islands thicker than
20 nm.
In the following two paragraphs we study the
evolution of h- and U-scans as a function of the
metal thickness (at constant deposition potential)
and as a function of the deposition potential (at
constant thickness). The influence of the step di-

M.L. Munford et al. / Surface Science 537 (2003) 95–112
103
even peaks
odd peaks
Si 0.2°
Φ = 90°
U = -1.73V
400
300
200
100
0
3000
18.4 ML
2000
14.4 ML
18.4 ML
7.2 ML
1000
14.4 ML
5.2 ML
7.2 ML
5.2 ML
16
18
20
22
0
(a)
(°)
θ
0
100
200
300
Φ (°)
(a)
2.0
12
8
6.5
0.90
0.85
0.80
0.75
1.6
1.2
6.0
5.5
5.0
4.5
4
5
10
15
20
4
8
12
16
20
tAu (ML)
(b)
(b)
tAu (ML)
Fig. 8. (a) h-scans around hAuð1 1 1Þ for gold deposits on Si-0.2°
(U ¼ À1:73 V). tAu is indicated in ML in the figure. Solid lines
are Lorentzian fits. (b) Variations of the peak intensity and
FWHMas a function of gold thickness.
Fig. 9. (a) U-scans corresponding to the h-scans of Fig. 8. The
film thickness is indicated in the figure. (b) Quantitative analysis
of spectra. Variations of the peak linewidth and R with the
metal thickness. See text for the definition of R. Note the sixfold
symmetry of spectra.
the Si(1 1 1) surface: R ¼ 0 corresponds to a pow-
der-like structure (no preferential orientation of
the grains) and R ¼ 1 means perfect epitaxy. Ex-
perimentally, R increases approximately linearly
with tAu (Fig. 9b).
data points shown in each graph correspond to
deposits of similar thickness on substrates with
different miscut angles. The same symbols are used
in the four graphs. For a given step density, the
different plots show clear trends for )1:93 V <
U < À1:6 V. Fig. 11a shows that the normalized
intensity increases by a factor 4. The peak line-
width decreases from 6–7° to 2° (Fig. 11b) and R
increases rather abruptly from 0.6, close to the
onset of nucleation (U ꢀ À1:6 V), to 0.9 for
U < À1:68 V (Fig. 11c). The last parameter de-
rived from the analysis of U-scan is R0 the ratio of
the integrated intensity of the odd peaks divided
by the intensity of the even peaks (after back-
ground correction). R0 gives the proportion of the
(1 1 1) gold islands which are rotated by 180° with
respect to the axis of bulk silicon. Fig. 11d shows
3.5. Influence of the deposition potential
The influence of the deposition potential on h-
and U-scans is shown in Fig. 10 for films of
thickness tAu ¼ 10 Æ 1 ML. There is a remarkable
increase, in both types of scans, of the peak in-
tensity with increasing g. There is also a systematic
evolution of the peak linewidth. In U-scans the
FWHMdecreases with increasing g while it in-
creases in h-scans measured with U ¼ 0° (i.e. with
the beam perpendicular to the step direction). This
last point is commented below. Fig. 11 details the
potential dependence of the U-scans. The sets of

104
M.L. Munford et al. / Surface Science 537 (2003) 95–112
that R0 is increasing from 0.6 to 0.7 close to the
nucleation onset and reaches almost 1 at )1.93 V.
The potential dependence of h-scans showed sim-
ilar trends (not shown).
Si 0.2°
(a)
Φ = 0°
300
200
100
0
9-10 Au ML
-1.93V
-1.63V
-1.58V
3.6. Influence of the steps on h-scans
Fig. 12 plots for different U-angles the h-scans of
a 9–10 ML gold deposit on Si-0.2° (U ¼ À1:68 V).
Clearly hAuð1 1 1Þ oscillates with a periodicity of 360°
(Fig. 12a). The amplitude of oscillations (0.3°)
corresponds to the tilt angle between the Au(1 1 1)
planes and the optical plane of the sample surface.
The same experiment around hSið1 1 1Þ on Si-0.2°
(not shown) gives an amplitude of oscillation 0.2°
equal to the miscut angle of this substrate [30,26].
The surprise, however, is that the oscillations of
hAuð1 1 1Þ and hSið1 1 1Þ are out of phase, which means
that the tilt angle hAu–Si, between the Au(1 1 1) and
the Si(1 1 1) planes, is equal to the sum of the two
amplitudes. Hence hAu–Si ¼ 0:2 þ 0:3 ¼ 0:5° for the
considered deposit. Fig. 12b plots hAu–Si as a
function of the deposition potential. On Si-0.2°,
hAu–Si is essentially constant and equal to 0.4°. On
Si-2°, hAu–Si decreases continuously with increasing
g from 2° to close to 0°. Last, Fig. 12c is showing
16
17
18
19
θ (°)
20
21
22
4000
3000
2000
1000
0
(b)
-1.93V
-1.63V
0
100
200
Φ (°)
300
Fig. 10. Series of h- and U-scans showing the influence of the
deposition potential (indicated in the figure) on the film struc-
ture. Note the increase of the peak intensity, in both kinds of
scans, with decreasing deposition potentials.
6
(a)
(b)
80
60
4
2
9-10 ML ; Si-0.2°
15-17 ML; Si-0.4°
14-16 ML; Si-2°
40
20
1.0
0.8
0.7
0.9
0.8
0.7
0.6
(c)
(d)
0.6
-1.9
-1.8
-1.7
-1.6
-1.9
-1.8
-1.7
-1.6
U (V)
U (V)
Fig. 11. Quantitative analysis of U-scans as a function of the deposition potential. For each substrate only deposits of similar thickness
are compared. Triangles, circles and squares respectively correspond to 14–16 ML films on Si-2°, 15–17 ML films on Si-0.4° and 9–10
ML films on Si-0.2° in all graphs. (a) Total normalized intensity integrated under the six peaks. (b) Average peak linewidth. (c and d)
variations of R and R0. See text for the definition of R and R0.

M.L. Munford et al. / Surface Science 537 (2003) 95–112
105
1.6
200
150
100
50
1.5
1.0
0.5
0.0
Si-2°
270°
180°
Φ = 0°
1.2
0.8
90°
Φ = 90°
Si-0.2°
0°
Si(111) 0.2°
0
18
19
θ (°)
20
-1.9 -1.8 -1.7
U (V)
-1.9 -1.8 -1.7
U (V)
(a)
(c)
(b)
Fig. 12. (a) h-scans around hAuð1 1 1Þ recorded for different U angles. (b) Variation of hAu–Si as a function of the deposition potential for
two miscut angles. (c) Variations of the peak linewidth as a function of the deposition potential. Open symbols refer to U ¼ 90° and
bold symbols to U ¼ 0°. Note that the FWHMis smaller for U ¼ 90°.
that, at any deposition potential, the peak line-
width of h-scans is significantly smaller with the
beam parallel to steps (open symbols) than with
the beam perpendicular to them (closed circles).
This is suggesting that the steps have some influ-
ence on the orientation of the gold islands.
Part of the clusters are in epitaxy with bulk silicon,
according to the relationship Au(1 1 1)[1 )1 0]kSi-
(1 1 1)[1 )1 0], while the remaining ones are rotated
by 180°. The epitaxy may be explained by the
atomic model displayed in Fig. 13 where a relaxed
(1 1 1) gold monolayer placed onto a 1 · 1-Si(1 1 1)
surface builds the commensurate 3 · 3 structure
ꢂ
ꢂ
[3 · (Si–Si ¼ 3.84 A) ¼ 4 · (Au–Au ¼ 2.88 A) ¼
3
11.52 A]. From this model we conclude that the
ꢂ
4. Discussion
edges of polyhedron islands are gold densely
packed directions.
Gold electrodeposition leads to a Volmer–
Weber growth on H–Si(1 1 1) likewise gold evap-
oration on the same surface [31]. In agreement
with Oskam and Searson [15], we find that the
nucleation is progressive as indicated by the di-
mensionless transients in Fig. 4c and the wide size
distribution of the gold islands imaged by AFM
(Figs. 5–7). The observation of a constant density
of islands by AFM(Fig. 5a–c) is not contradictory
because the images corresponds to deposition
times greater than tmax (the nucleation stops for
t ꢀ tmax). In the studied range of potential, AFM
shows that the 3D nm-sized islands exclusively
decorate the terrace edges (Figs. 5–7). The im-
provement of the (1 1 1) texture (Fig. 8) and epit-
axy (Fig. 9) upon growth is attributed to the
slower growth rate of the Au(1 1 1) planes with
respect to the other orientations. It is a general rule
that the slower growth planes are dense planes and
that they impose the film texture upon growth. The
sixfold symmetry of the U-scans (Figs. 9 and 10)
indicates epitaxy with the silicon (1 1 1) surface.
4.1. Specificity of the silicon–electrolyte electro-
chemical interface
Unlike metal evaporation, metal electrodeposi-
tion on silicon involves the donation of conduction
band electrons to reduce the metal complexes to
metallic atom. In the case of gold electrodeposition
we may consider the following reactions:
AuðCNÞÀ ! AuðCNÞ þ CNÀ
ð1aÞ
2
ads
AuðCNÞ þ eÀCB ! Au0 þ CNÀ
ð1bÞ
ð2Þ
ads
Si–H þ 2H2O þ 2eÀ ! Si-H þ H2 þ 2OHÀ
CB
3
The (7 · 7)R19°1 is another commensurate structure. It
accounts for the satellites peaks shifted by 19° with respect to
main peaks (Fig. 10b). These were not discussed because of
their low intensity. Clusters rotated by ꢀ20° with respect the
main axis are also occasionally found by AFM(see Fig. 7).

106
M.L. Munford et al. / Surface Science 537 (2003) 95–112
much faster at steps than on terraces because the
Si
Au
splitting of Si–H bonds is more favorable at step
sites for energetic and conformation reasons (Fig.
1b). As the rate of reaction (2) increases with g, the
density of step dangling bonds also increases with
g. In the same way reaction 1b is faster at step sites
because the gold atoms increase their coordination
number with the silicon lattice. The adsorption
stage (reaction (1a)) is, at reverse, thought to be
nearly site independent.
4.2. Origin of the exclusive gold nucleation at steps
[1-10]
As evidenced by AFMobservations (Figs. 5d
and 7b–c) the nucleation of the gold film is not
homogeneous and occurs exclusively at the terrace
edges for )1:92 V < U < À1:60 V. At small g, a
preferential nucleation at steps is not a surprise
because steps are usually favorable sites. It is the
total absence of nucleation on the (1 1 1) terraces,
at larger g which constitutes a strong surprise be-
cause the flux seems sufficiently large to allow the
formation of critical nuclei on atomically smooth
H–Si(1 1 1) terraces (the average flux is ꢀ2 ML/s
for 10–15 ML deposited in 7 s). In the case of in-
dium evaporation, STMobservations show that
islands nucleate on H–Si(1 1 1) terraces for much
smaller fluxes (<0.1 ML/s) [34].
According to the mean field nucleation theory,
the density nx of islands varies as ðD=F ÞÀv on de-
fect free surfaces (i.e. on flat terraces). F denotes
the deposition flux and D is the diffusion coefficient
of adatoms on the surface. The scaling exponent v
is depending on the size of the critical nuclei [33].
This law is general. The conditions of flux and
temperature to reach a given nx are nevertheless
depending on the system though D and v. Using
BruneÕs [33] plots of nx as a function of (D=F ), we
estimate, in the case of indium evaporation, that
ðD=F Þ ꢀ 108 to account for the observed nx ꢀ 1011
islands/cm2 [34]. This corresponds to a quite large
diffusion coefficient (F ¼ 0:01 to 0.1 ML/s) in
agreement with calculation showing that metal
adatoms diffuse much faster on H–Si(1 1 1) than on
clean 7 · 7 surfaces [35]. This is because the satu-
ration of dangling bonds with H atoms removes all
electronic states [36] and promotes a very shallow
Fig. 13. Atomic model of the Au–Si interface accounting for
the observed relation of epitaxy. Only the silicon atoms of the
topmost surface plane are displayed.
Reaction (1) is the reduction of the gold cyanide
complexes in a two-step process as proposed by
Oskam and Searson [14]. In the first step (reaction
(1a)), the gold complex is partially dissociated and
adsorbed on the surface as Au(CN)ads. It is re-
duced to metallic gold in the second step by cap-
ture of one conduction band electron (reaction
(1b)). Reaction Eq. (2) is the decomposition of
water into molecular hydrogen. On the molecular
level [32] this reaction involves the desorption of
one hydrogen atom from the silicon surface in
order to build the molecule of H2. Hence, reaction
Eq. (2) transiently generates dangling bonds on the
surface. The rate of reactions (1) and (2), averaged
over the entire surface, increases with g. From the
curve in inset of Fig. 3, the contribution of reac-
tion (2) to the deposition current becomes signifi-
cant for U < À1:8 V while it may be neglected
closer to the nucleation onset.
One specificity of the electrochemical electron
transfer is that its rate may be highly site depen-
dent on the atomic scale. A striking example of the
influence of the surface structure on electrochem-
ical transfer is the reaction of chemical etching of
silicon, which is totally anisotropic and exclusively
occurs at the silicon step edges because the hy-
drolysis of Si–H bonds into silanol groups is much
faster on such sites [32]. In analogy with the
etching reaction, we infer that reaction (2) is also
ꢂ
electronic corrugation (<0.2 A in STMimages

M.L. Munford et al. / Surface Science 537 (2003) 95–112
107
[37]). In our case, the absence of islands on terraces
as wide as 100 nm implies that ðD=F Þ ꢁ 108 [33].
Accounting for the difference of fluxes between the
two experiments (F ꢀ 2 ML/s in this work), we
conclude that the diffusion coefficient of Au is
several orders of magnitude larger at the solid–
electrolyte interface than that of indium in vac-
uum. Such an enhanced diffusion is consistent with
reaction (1a) because halides and CNÀ anions in-
crease the surface mobility of gold adatoms on Au
surfaces [38]. This effect is amplified here because
the H-monolayer reduces the interactions between
the adsorbates Au(CN)ads and the silicon.
An enhanced surface diffusion is nevertheless
not sufficient to explain the absence of gold clus-
ters on H–Si(1 1 1) terraces. The presence of the
cyanide complexing agents in solution requires
the generation of Si–Au bonds to stabilize a criti-
cal nucleus on the surface. This is the second
prerequisite. Because the step sites are the most
favorable desorption sites of hydrogen, the first
gold adatoms will preferentially form a stable nu-
cleus at steps where the formation of Si–Au bonds
ensure that the Au atoms are not re-dissolved.
Such a mechanism, where the metal atoms is de-
posited at the silicon dangling bonds, was assumed
for nickel electroless deposition from a NiSO4–
NH4F solution because no Ni deposition occurs
if the silicon is not etched at a sufficient rate
[25]. In this case the silicon dangling bonds are
generated by the chemical etching reaction [32].
The same mechanism explains that electroless
deposition from a diluted CuSO4–NH4F [22]
leads to small Cu clusters which decorate the
steps (the steps are the stage of preferential etch-
ing).
the same metal thickness, R reaches 90% at )1.92
V against only 60% at )1.6 V (Fig. 11c). Accord-
ingly the gold islands present a much better-
defined polyhedron shape in AFMimages (Fig. 7).
Such islands are also strongly attached to the
surface (they are not wiped by the AFMtip).
After the formation of a stable nucleus at steps
(Fig. 14a), the growth may proceed on terraces
from such preferential nucleation centers, because
they become soon the stage of the electron transfer
and also because they act as sink for diffusing Au
atoms. A priori, a good epitaxy with the (1 1 1)
terraces implies sufficiently strong interactions
with the substrate. Clean Si(1 1 1) favors metal
epitaxy [4] whereas H-terminated Si(1 1 1) does not
because the saturation of Si dangling bonds with H
atoms prevents the formation of a stable chemical
bond between the metal and the silicon. Therefore
the observed epitaxy of Au electrodeposits suggests
that an intimate Au–Si contact can also be formed
on terraces. The following growth mechanism is
tentatively proposed after formation of the pre-
cursor nucleus (Fig. 14a). At large overpotential,
the island is in direct contact with the silicon (Fig.
14b). The H-layer is progressively unzipped from
the boundary between the H-layer and the gold
step because reaction (2) is catalyzed there and the
rate of H-desorption greatly amplified. We stress
that the rate of spontaneous H-desorption on
naked H–Si(1 1 1) terraces is extremely small. The
growth remains 2D as long as the removal rate of
H atoms is sufficiently large to allow the incorpo-
ration of all incoming gold adatoms. Very soon 3D
growth occurs however due to gold on gold de-
position (electrons are transferred from the de-
posited clusters). The crossover between the 2D
and 3D growth regimes is determined by the size
and the surface density of the metal nuclei [42].
Close to the onset of nucleation the rate of reac-
tion (2) is negligible on the gold islands. As a result
the Au adatoms stick to the nucleus on the step
and the islands grow atop the H-layer (Fig. 14d).
This scenario correlates very well with the obser-
vation that films grown close to )1.6 V are weakly
interacting with the substrate (islands are wiped by
the AFMtip). At intermediate potentials, a mixed
structure such as the one sketched in Fig. 14c
might be obtained.
4.3. Growth mechanism
The main experimental fact is the improved
epitaxy with increasing g. h and U-scans (Figs. 10
and 11) indicate indeed that the film structure is
rather powder-like, close to the nucleation onset
()1.6 V), with 40% grains having no preferential
orientations (R ’ 0:6). Such films are imaged with
difficulty since the AFMtip easily wipes the is-
lands. As soon as U < À1:7 V a well defined
epitaxy with a (1 1 1) orientation is established. For

108
M.L. Munford et al. / Surface Science 537 (2003) 95–112
H-layer ~ 1.6 Å
Au nuclei
(a)
Si(111)
intimate
contact
structural
defect
‘floating’
island
(b)
(c)
(d)
U vs reference
Fig. 14. Qualitative model explaining the correlation between the quality of the epitaxy and the deposition potential. (a) Nucleation
stage: a stable nucleus is formed at the step where preferential desorption of the hydrogen takes place. This nucleus acts as precursor
for the lateral growth on the terrace. (b)–(d) Growth model as a function of the deposition potential: at large overpotential (b) the H-
termination is removed by local H-desorption from the boundary between the H-layer and the gold monolayer. An intimate Si–Au
contact is built. Close to the nucleation onset (d) the gold island is growing atop the H layer. Such islands are wiped by the AFMtips.
At Intermediate potentials (c), the Au island is only partly in contact with the silicon surface.
4.4. Origin of the rotation of gold nanocrystals
Gold evaporation, which constitutes an inter-
mediate case in the scale of reactivity with silicon
[39], promotes a partial removal of the H-layer
upon evaporation on H–Si(1 1 1) [31]. Fig. 14 evi-
dences therefore a major difference between the
two methods of deposition. At the electrochemical
interface, the removal of the H-monolayer is
monitored by the catalytic properties of the de-
posit towards reaction (2). Depending on the po-
tential, the deposit may be in intimate contact with
silicon or may float above the H-monolayer. The
result seems not dependent on the reactivity of
gold with respect to silicon, whereas, in vacuum,
the removal of H atoms is depending, to a large
extent, on the metal reactivity. In the case of in-
dium, a weakly reacting metal, the H-layer is
thought to remain intact under the deposit because
the Si–In bond energy (2.5–2.8 eV) is much smaller
than the Si–H bond energy (ꢀ3.5 eV) [34]. In the
case of very reactive metals (e.g. Ni, Fe, Cu) the
first 2 ML are embedded under the surface plane
and the H-monolayer remains intact (case of Fe
and Ni [40,41]). Above 2 ML the H layer is de-
stroyed and silicide formation is anticipated. Metal
electrodeposition enables the formation of a direct
contact without silicide formation.
The U-scans (Figs. 9 and 10) exhibit a sixfold
symmetry reflecting the symmetry of the silicon
surface and not the threefold symmetry of the
bulk. One could attribute this observation to
stacking faults but it is unrealistic to attain 50% of
stacking faults at )1.93 V. The above sixfold
symmetry rather indicates that the (1 1 1) gold
nanocrystallites take two possible orientations,
which are rotated by 180° from each other. The
same observation has been reported upon silver
evaporation [43]. We believe that our observations
stem from a different reason, principally because
the nucleation occurs exclusively at steps, which
is not the case in the UHV.
The starting point of the discussion is the co-
existence of triangles of type (A) and (B) in Fig. 7
which preferentially overlap with the upper and
the bottom terrace. Their edges correspond to
densely packed directions of fcc gold and XRD
suggests that the facets are (1 1 1) like facets. These
are also energetically favorable. Therefore, we
propose an atomistic model in which the lateral
growth of islands on terraces starts with the for-
mation of the smallest possible pyramidal island

M.L. Munford et al. / Surface Science 537 (2003) 95–112
109
with a triangular (1 1 1) base and (1 1 1) facets.
Such an island is represented in Fig. 15. Its trian-
gular base corresponds to one half of the 3 · 3 unit
cell shown in Fig. 13 with one edge parallel to the
step to conform to Fig. 7. The tip of the triangle is
therefore either pointing towards the upper (Fig.
15a) or on the bottom (Fig. 15b) terrace. Adding
subsequent atomic planes according to the fcc
stacking ABC creates two Au(1 1 1) nanocrystals,
which bulk axis are rotated by 180° with respect to
each other. On the upper terrace the axis of the
cluster are parallel to those of the silicon crystal
(Fig. 15a). Within the frame of this growth model,
two effects might be responsible for the experi-
mental increase of R0 with g. The first one is a re-
duction of the Schwoebel effect [44] at large g,
because the accumulation of electrons at the sili-
con surface likely smoothes out the diffusion bar-
rier across the steps. The second one is the
potential dependence of the structure of nucleation
sites. The desorption of H atoms occurs indeed at
different potentials on kink sites and step mono-
hydrides.
above model that they are in epitaxy with the
silicon, according to the relationship Au(1 1 1)-
[1 )1 0]kSi(1 1 1)[1 )1 0], and that they account for
the even peaks in U-scans. As a consequence, the
triangles of type (B) account for the odd peaks in
U-scans because they are rotated by 180°. Within
this model, the ratio R0 should be equal to the ratio
[volume of triangular islands of type (B)/volume of
triangular islands of type (A)]. We attempted to
verify this from the AFMimage in Fig. 7 by
counting the population of the different islands.
The ratio of triangles of type (A) to triangles of
type (B) is 5:1. Examination of the non-triangular
islands shows that 50% of the islands are over-
lapping preferentially with the upper terrace at
U ¼ À1:73 V, whereas only 10% overlap with the
lower terrace. The center of gravity of the re-
maining islands (40% of the total) overlap equally
with both the upper and bottom terrace. From this
analysis one finds R0 ꢀ 0:33, which is smaller than
the value 0.65 derived from the corresponding U-
scan. The principal cause for this discrepancy
comes form the fact that we could not consider the
volume of islands in our estimate. It was also im-
possible to make such a detailed analysis at more
As the triangles of type (A) are overlapping with
the upper terrace (Fig. 7), we conclude from the
(a)
(b)
[1-10]
Fig. 15. Atomic model accounting for the coexistence of epitaxial gold crystals rotated by 180° on the Si(1 1 1) surface. After nu-
cleation of the precursor nucleus, a half 3 · 3 unit cell is first formed on the Si(1 1 1) plane. Adding the second and third atomic planes
builds a pyramidal cluster with (1 1 1) facets. The cluster on upper terrace (a) is in epitaxy with Si(1 1 1) according to the relationship
Au(1 1 1)[1 )1 0]//Si(1 1 1)[1 )1 0]. It corresponds to triangles of type (A) in AFMimages. The cluster on the bottom terrace (b) is
rotated by 180° and corresponds to triangles of type (B). The precursor nucleus defined in Fig. 14a is not represented.

110
M.L. Munford et al. / Surface Science 537 (2003) 95–112
negative potentials of deposition because the lar-
ger island density hampers the precise localization
of the silicon steps.
16b). The Au(1 1 1) planes are parallel to the op-
tical plane of the sample surface and hAu–Si is equal
to the miscut angle of the substrate, in good
agreement with Fig. 12b. At )1.92 V, the island
size become smaller than the width of two terraces
(ꢀ20 nm). 4 They sit as shown in Fig. 16a, with the
Au(1 1 1) planes practically parallel to Si(1 1 1)
4.5. Influence of the step density on hAu–Si
The influence of steps on the orientation of the
gold islands is clearly evidenced by Fig. 12. On Si-
0.2°, hAu–Si is essentially potential independent. In
this case, simple geometrical considerations may
explain the tilt angle of 0.4° because the islands
size is always smaller than twice the terrace width
(Fig. 16a). The mismatch between the height of a
planes, which reduces hAu–Si
.
The second effect of the step is a broadening of
the linewidth of h-scans when the X-ray beam is
perpendicular to the step direction. This difference
can be simply rationalized by considering that the
Si step induces structural defects in the direction
perpendicular to it if the Au island is positioned
across it. The step is adding a degree of freedom to
accommodate the gold lattice on the silicon lattice.
In Fig. 7, 40% of islands adopt such a position.
The fluctuations of orientations are less significant
parallel to the steps because of the absence of such
surface defects. For the remaining grains, which
are preferentially overlapping with the upper ter-
races, this effect should be less important.
ꢂ
gold monolayer (2.35 A) and that of a silicon step
ꢂ
(3.14 A) makes that the gold plane overlapping
with both terraces does not sit properly on either
of the two terraces. For a 50 nm wide island, the
tilt angle is hAu–Si ꢀ arctg [(0:314 À 0:235)/(25)] ¼
0.18°, which is fairly close to the experimental
value in Fig. 12b. On Si-2°, the stronger effect of
the potential on hAu–Si arises from the fact that the
island size varies in greater proportion. And may
be close or bigger than twice the terrace width
(W ¼ 10 nm). At U ꢀ À1:75 V the islands are 50
nm wide and overlap with several terraces (Fig.
5. Conclusion
In conclusion we have shown that the electro-
chemical growth of Au on structurally well-defined
vicinal H-terminated Si(1 1 1) surfaces occurs by
selective nucleation of 3D nm-sized gold islands at
the silicon steps. The potential of deposition con-
trols both the density of islands along the steps
and the film structure. Under optimum conditions,
i.e. at sufficiently negative potential, a nearly per-
fect replication of the stepped surface structure is
obtained with epitaxial nm-sized islands decorat-
ing the silicon steps. All observations are consis-
tent with the proposed reaction scheme and
growth model. In initial stages, a stable critical
nucleus is first formed on the dangling bonds
generated at steps by the decomposition of water.
0.18°
(a)
Optical
plane
(b)
Fig. 16. Schematic explanation for the tilt angle hAu–Si between
the Au(1 1 1) and Si(1 1 1) planes. (a) When the island is smaller
than the terrace width, the mismatch between the height of a
gold monolayer and the silicon step induces a small tilt of the
second gold atomic plane deposited across the step. (b) When
the islands are wider than two terraces, they are deposited
across several steps. The gold Au(1 1 1) planes are parallel to the
optical plane and hAu–Si is equal to the miscut angle of the
substrate.
4
A precise measure of the island size is difficult on Si-2°
because the tip is not sufficiently sharp. The surface density of
islands being 10 times greater on Si-2°, we may estimate, for an
equivalent metal thickness, that the islands are smaller by a
factor ꢀ101=3 ¼ 2.15 on Si-2° than on Si-0.2°.

M.L. Munford et al. / Surface Science 537 (2003) 95–112
[14] G. Oskam, P. Searson, Surf. Sci. 446 (2000) 103.
111
This nucleus acts soon as preferential nucleation
center for the lateral growth on terraces. During
this phase, and at sufficiently negative potential,
the H-layer is progressively removed from the
surface within a catalytic process occurring at the
Au/H–Si interface, while the H-termination re-
mains intact between the metal islands. Close to
the nucleation onset, the H-layer remains intact
under the deposit and the film has a rather pow-
der-like structure.
[15] G. Oskam, P. Searson, J. Electrochem. Soc. 147 (2000)
2199.
[16] A.P. OÕKeeffe, O.I. Kasyutich, W. Schwarzacher, L.S. de
Olivera, A.A. Pasa, Appl. Phys. Lett. 73 (1998) 1002.
[17] K. Attenborough, H. Boeve, J. De Boeck, G. Borghs, J.P.
Celis, Appl. Phys. Lett. 74 (1999) 2206.
[18] L. Seligman, M.L. Sartorelli, A.A. Pasa, W. Scharzacher,
O.I. Kasyutich, J. Magn. Magn. Mater. 226–230 (2001)
752.
[19] M.L. Munford, L. Seligman, M.L. Sartorelli, E. Voltolini,
L.F.O. Martins, W. Schwarzacher, A.A. Pasa, J. Magn.
Magn. Mater. 226–230 (2001) 1613.
[20] F. Maroun, F. Ozanam, J.N. Chazalviel, W. Teiss,
Vibrational Spec. 19 (1999) 193.
Acknowledgements
[21] F. Ronkel, J.W. Schultze, R. Arens-Fischer, Thin Solid
Films 276 (1996) 40.
M.L. Munford gratefully acknowledges CAPES
(Brazil) for a Ph.D. fellowship during his stay in
France and the ‘‘Direction des Affaires Interna-
[22] T. Homma, C.P. Wade, C.E.D. Chidsey, J. Phys. Chem. 41
(102) 7919.
[23] S. Warren, A. Reitzle, A. Kazimirov, J.C. Ziegler, O.
Bunk, L.X. Cao, F.U. Renner, D.M. Kolb, M.J. Bedzyk,
J. Zegenhagen, Surf. Sci. 496 (2002) 287.
[24] P. Gorostiza, R. diaz, J. Servat, F. Sanz, J.R. Morante,
J. Electrochem. Soc. 144 (1997) 909.
ꢁ
tionales’’ (Universite P & MCurie) for a comple-
mentary financial support. We also thank M.-C.
Bernard for the RBS measurements. We also
acknowledge Siltronix (France) for preparing
the substrates according to our specifications.
[25] P. Gorostiza, M. Kulandainathan, R. Diaz, F. Sanz, P.
Allongue, J.R. Morantes, J. Electrochem. Soc. 147 (2000)
1026.
[26] M.L. Munford, R. Cortes, P. Allongue, Sensors Mater. 13
(2001) 259–269.
References
[27] C.P. Wade, C.E.D. Chidsey, Appl. Phys. Lett. 71 (1997)
1679.
[28] H. Fukidome, M. Matsumura, Appl. Surf. Sci. 130 (1998)
146.
[1] T. Ogino, H. Hibino, Y. Homma, Y. Kkobayashi, K.
Prabhakaran, K. Sumitomo, H. Omi, Acc. Chem. Res. 32
(1999) 447.
[29] P. Allongue, C. Henry de Villeneuve, S. Morin, R. Bouk-
herroub, D.D.M. Wayner, Electrochim. Acta 45 (2000)
4591.
[2] A. Kirakosian, J.L. Lin, D.Y. Petrovykh, J.N. Crain, F.J.
Himpsel, J. Appl. Phys. 90 (2001) 3286.
ꢁ
[30] M.L. Munford, Ph.D. Thesis, Universite Paris 6 in Paris
[3] F.J. Himpsel, A. Kirakosian, J.N. Crain, J.L. Lin, D.Y.
Petrovykh, Solid State Commun. 117 (2001) 149.
[4] K. Oura, V.G. Lifshits, A.A. Saranin, A.V. Zotov, M.
Katayama, Surf. Sci. Report 35 (1999) 1.
(France) and UFSC in Florianopolis (Brazil), 2002.
[31] C. Grupp, A. Taleb-Ibrahimi, Phys. Rev. B 57 (1998) 6258.
[32] P. Allongue, V. Costa-Kieling, H. Gerischer, J. Electro-
chem. Soc. 143 (1993) 1018;
[5] P. Jakob, Y.J. Chabal, J. Chem. Phys. 95 (1991) 2897.
[6] P.M. Hoffmann, A. Radisic, P.C. Searson, J. Electrochem.
Soc. 147 (2000) 2576.
P. Allongue, V. Costa-Kieling, H. Gerischer, Electrochim.
Acta 40 (1995) 1353.
[33] H. Brune, Surf. Sci. Report 31 (1998) 125;
H. Brune, G.S. Bales, J. Jaconsen, C. Boragno, K. Kern,
Phys. Rev. B 60 (1999) 5991.
[7] J.C. Ziegler, A. Reitzle, O. Bunk, J. Zegenhagen, D.M.
Kolb, Electrochim. Acta 45 (2000) 4599.
[8] T. Zambelli, M.L. Munford, F. Pillier, M.C. Bernard, P.
Allongue, J. Electrochem. Soc. 148 (2001) C614.
[9] C. Ji, G. Oskam, P.C. Searson, Surf. Sci. 492 (2001) 115.
[34] F.P. Leisenberger, H. Ofner, M.G. Ramsey, F.P. Netzer,
Surf. Sci. 383 (1997) 25.
€
[35] N. Enomoto, T. Hoshino, M. Hata, M. Tsuda, Appl. Surf.
Sci. 130–132 (1998) 237.
[10] B. Rashkova, B. Guel, R.T. Potzschke, G. Staikov, W.
Lorenz, Electrochim. Acta 43 (2000) 3021.
[11] J.C. Ziegler, R.I. Wielgosz, D.M. Kolb, Electrochim. Acta
45 (1999) 827.
[36] D.M. Kolb, Prog. Surf. Sci. 51 (1996) 109;
Electrochim. Acta, 45 (2000) 2387.
[37] P. Allongue, Phys. Rev. Lett. 77 (1996) 1986–1989.
[12] J.C. Ziegler, G. Scherb, O. Bunk, A. Kazimirov, L.X. Cao,
D.M. Kolb, R.L. Johnson, J. Zegenhagen, Surf. Sci. 452
(2000) 150.
ꢁ
[38] S. Gallego, J. Avila, M. Martin, X. Blase, A. Taleb, P.
Dumas, M.C. Asencio, Phys. Rev. B 61 (2000) 12628.
[39] J.A. Kerr, in: D.R. Lide (Ed.), CRC Handbook of Chem-
istry and Physics 1999–2000: A Ready-Reference Book of
[13] G. Oskam, J.G. Long, A. Natarajan, P. Searson, J. Phys.
D: Appl. Phys. 31 (1998) 1927.

112
M.L. Munford et al. / Surface Science 537 (2003) 95–112
Chemical and Physical Data (CRC Handbook of Chem-
istry and Physics), 81st ed., CRC Press, Boca Raton,
Florida, USA, 2000.
[42] P. Allongue, Modern Aspects of Electrochemistry, B.E.
Conway (Ed.), Plenum Press, New York, Chapter 4, vol.
23, 1992, p. 239.
[40] M.G. Martin, J. Avila, M. Gruyters, C. Teodorescu, P.
Dumas, Y.J. Chabal, M.C. Asencio, Appl. Surf. Sci. 123–
124 (1998) 156.
[43] K. Sumitomo, T. Kobayashi, F. Shoji, K. Oura, Phys. Rev.
Lett. 66 (1991) 1193.
[44] R.L. Schwoebel, E.J. Chipsey, J. Appl. Phys. 37 (1966)
3682;
R.L. Schwoebel, J. Appl. Phys. 40 (1969) 614.
[41] K. Hirose, A. Hanta, M. Uda, Appl. Surf. Sci. 162–163
(2000) 25.