2012 Thesis Koen Lauwaet

!"
#
$ % &
'((' )" *) +
'((% ,-." *) +
'((/ )" *% +
'((
'((# ,0"
'((-1203
'((2 ' 4
(" *+
"
),1# 5(()5% 6 !!5
013!!"-)50 7
9 ) 1 5 (()5 % 6 !!5
013!!"-)50 7
3
9 ) # :.5 -5 % .5 013;<!
.50 7
4
: # 5 ) = # 10 5, %:: ;3591
!" ;# 59
c 2012 Faculteit Wetenschappen, Geel Huis, Kasteelpark Arenberg 11,
3001 Heverlee (Leuven)
Alle rechten voorbehouden. Niets uit deze uitgave mag worden ver-
menigvuldigd en/of openbaar gemaakt worden door middel van druk,
fotokopie, microfilm, elektronische of op welke andere wijze ook zonder
voorafgaande schriftelijke toestemming van de uitgever.
All rights reserved. No part of this publication may be reproduced in
any form by print, photoprint, microfilm, electronic or any other means
without written permission of the publisher.
ISBN 978-90-8649-488-0
Wettelijk depot D/2012/10.705/5
Dankwoord
“Leg ons nu eens uit wat je precies doet? ” Deze vraag, of varianten
ervan, is me zeer veel gesteld de laatste vier en een half jaar. Wat ik
gedaan heb is niet zo in één twee drie uit te leggen, vandaar dat ik er dan
ook een boekje over geschreven heb, hetwelk nu voor u ligt. Natuurlijk
ben ik niet alleen tot dit eindresultaat geraakt. Onderweg hebben vele
mensen me geholpen en ik wil hun graag hiervoor bedanken.
Mijn promotoren prof. dr. Peter Lievens en prof. dr. Chris Van
Haesendonck wil ik bedanken voor me de mogelijkheid te geven om aan
dit doctoraat te beginnen en me te begeleiden naar dit eindpunt. Ik wil
hun ook graag bedanken voor hun vertrouwen in mij en de bijhorende
vrijheid om mijn eigen stempel op dit onderzoek te zettten en zo nieuwe
hoeken van de nanowereld te exploreren. Ook al werden sommige van
mijn ideeën vreemd bekeken door de collega’s (“Gij wilt met zout gaan
werken, zout? ”), dankzij jullie kreeg ik de kans om mijn ding te doen
zowel binnen de Class als binnen de SPM groep. Ook dank voor de
soms broodnodige verse inspiratie en wetenschappelijke ideeën wanneer
ik vastzat met op dat moment voor mij moeilijk te verklaren resultaten.
Al mijn onderzoek heeft zich afgespeeld in het Laboratorium voor
Vaste-Stoffysica en Magnetisme (VSM). Graag wil ik dan ook het vorig
voormalig en het huidig hoofd van VSM, prof. dr. Chris Van Hae-
sendonck en prof. dr. Jean-Pierre Locquet, bedanken voor het terbe-
schikkingstellen van de koffie, apparatuur en middelen die nodig waren
voor mijn onderzoek.
I would like to thank the members of my doctoral committee: prof.
dr. Roger Silverans, prof. dr. Margriet Van Bael, prof. dr. Kristiaan
Temst, dr. Koen Schouteden, prof. dr. Hans-Gerhard Boyen, and prof.
dr. Gianfranco Pacchioni, for the interest that they showed in my work
and time they spend reading and improving this thesis.
Ook het Agentschap voor Innovatie door Wetenschap en Technologie
ben ik dankbaar voor hun financiele steun gedurende vier jaar van mijn
i
doctoraat.
Dr. Mario I. Trioni, dr. Livia Giordano, and prof. dr. Gian-
franco Pacchioni, thank you for your theoretical calculations on the
NaCl/Au(111) system. Your input was very valuable and taught me
a lot about the system that I was experimenting on.
Ook “die andere Koen” verdient een speciaal bedankje. Voor al
de uren achter de STM samen, de discussies over al-dan-niet fysica, de
motivatie, de STM humor,... dankjewel! Dank ook aan Ewald voor de
dagdagelijkse begeleiding en de vele correcties. Christian, for all the
hours we spent together at the CDA, for all the jury rigging or proper
jobs we did, and for the sometimes conflicting (but always fresh!) views
on things, thank you. Aurélie, thank you for guiding me through my
first steps in the experimental world. For all the members of the Class
group present in these last four and a half years, thank you for the
fruitful discussions and the opinions on what works in the cluster world,
and what does not work.
Monique en Liliane, ik heb het veel te weinig tegen jullie gezegd,
maar hartelijk bedankt om de administratieve kant van de experimentele
wereld in goede banen te leiden en ons te beschutten van de administratie
van de K.U.Leuven. Ik ben er zeker van dat ik zonder jullie hulp nooit
half zover was geraakt.
Ook een hartelijk dankjewel aan de technische staf, zonder wie ik
met een hoop roestvrij staal zou zitten in plaats van werkende appa-
ratuur. Dankjewel aan Phillippe, Phillippe, Steven en Stijn voor altijd
klaar te staan met het juiste gereedschap of een babbel. Dankjewel ook
aan de mechanische werkplaats voor jullie ondersteuning en het maken
van de massafilter. Lucien, Philip en de rest van elektronica, bedankt
voor jullie hulp bij het repareren van controllers en het maken van alle
benodigde kabels. Erik en Marc, bedankt voor de vele gassen en liters
N2 en He, nodig om alles koel te houden. Hartelijk dank ook aan Bas
Opperdoes voor me bij te staan met raad, daad, MBE-hulp en én een
reservefiets wanneer nodig.
Ook al staan fysici er niet om bekend, ontspanning was er ook. Er is
dan ook een hele verzameling collega’s die ik om uiteenlopende redenen
wil bedanken. Sportcoördinator Bas, voor het orkestreren van onze vele
overwinningen bij het voetbal (zonder Jokks, Jo, Tom, Pieterjan, Pieter
en Alejandro te vergeten) en voor het uitstippelen van tochtjes voor
de VSM fietsclub, die je dan natuurlijk wel zelf won. Ook dank aan
de leden van de koffie-club en andere activiteiten: Koen, Stijn, Steven,
ii
Maarten, Christian, Violetta, Tom, Pieterjan, Jo, Bart, Bert, Kelly,
David, Katrien, Bas, Denitza, Tobias, Thomas, Nele, Nele en iedereen
die ik vergeten ben in het lijstje. Dank ook aan alle bureaugenootjes,
zowel de oude als de nieuwe, voor de vele toffe gesprekken en af en toe
de gelegenheid om stoom af te blazen: Christian, Jo, Sabina, Pieterjan
en Kelly; dankjewel.
Dank ook aan mijn ouders die me de mogelijkheid hebben gegeven
om altijd mijn studiekeuze te volgen, al was het soms tegen het advies
van de leerkrachten in. Dankje voor je steun gedurende zoveel jaren.
And last but not least: gracias Elena por tu apoyo, ayuda y pacien-
cia conmigo. Thanks for the time and laughs we shared, all the dances
together and the time and dances to come. Thank you for being you in
short, dankjewel.
iii
iv
List of Abbreviations
0D Zero Dimensional
1D One Dimensional
2-DEG Two Dimensional Electron Gas
2D Two Dimensional
3D Three Dimensional
BS Bulk State
CDA Cluster Deposition Apparatus
DFT Density Functional Theory
DOS Density of States
fcc face centered cubic
FT Fourier Transform
hcp hexagonal close packed
IS Interface State
LDOS Local Density of States
SFM Scanning Force Microscopy
SPM Scanning Probe Microscopy
SS Surface State
STM Scanning Tunneling Microscopy
v
STS Scanning Tunneling Spectroscopy
UHV Ultra-High Vacuum
vi
Contents
List of Abbreviations v
The nanoscale: Where size does matter 5

Chapter overview . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1 Reducing dimensions 11
1.1 Confinement of particles . . . . . . . . . . . . . . . . . . . 12
1.1.1 The infinite square well . . . . . . . . . . . . . . . 12
1.1.2 Particles trapped in three dimensions . . . . . . . 14
1.2 The nearly-free electron gas inside metals . . . . . . . . . 15
1.2.1 Metal as a 3D electron box . . . . . . . . . . . . . 15
1.2.2 Nearly-free electron model . . . . . . . . . . . . . . 18
1.3 Electrons at metal surfaces . . . . . . . . . . . . . . . . . 19
1.4 0D systems . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2 Exploring the nanoworld with a scanning tunneling mi-

croscope 27
2.1 Imaging surfaces at the atomic level . . . . . . . . . . . . 28
2.1.1 Scanning tunneling microscopy . . . . . . . . . . . 29
2.2 Revealing the density of states . . . . . . . . . . . . . . . 34
2.2.1 Single point tunneling spectroscopy . . . . . . . . . 34
2.2.2 Mapping of the local density of states . . . . . . . 35
2.3 The low-temperature UHV STM . . . . . . . . . . . . . . 36
2.3.1 Preparation of STM tips . . . . . . . . . . . . . . . 38
3 Realization of systems with reduced dimensionality 39

3.1 The Au(111) surface: An atomically flat base . . . . . . . 39
3.1.1 Growth and cleaning of Au(111) films . . . . . . . 40
3.1.2 The “herringbone” surface reconstruction . . . . . 40
3.1.3 The electronic surface state . . . . . . . . . . . . . 41
1
2 Contents
3.2 Depositing preformed clusters in UHV . . . . . . . . . . . 43

3.2.1 Cluster deposition apparatus . . . . . . . . . . . . 43
3.2.2 Sample transport . . . . . . . . . . . . . . . . . . . 47
3.3 Growth of thin NaCl layers . . . . . . . . . . . . . . . . . 47
3.4 Creation of F-centers and deposition of Co atoms . . . . . 48
4 TiOx nanoclusters on Au(111) 49

4.1 TiOx clusters on Au(111) without annealing . . . . . . . . 51
4.2 Influence of annealing to elevated temperatures . . . . . . 53
4.3 Electronic properties of TiOx nanoparticles on Au(111) . 56
4.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5 Tuning the NaCl/Au(111) interface state 63

5.1 Morphologic Characterization . . . . . . . . . . . . . . . . 65
5.1.1 As deposited . . . . . . . . . . . . . . . . . . . . . 65
5.1.2 After annealing . . . . . . . . . . . . . . . . . . . . 66
5.2 Dispersion relations . . . . . . . . . . . . . . . . . . . . . 67
5.2.1 Bilayer NaCl/Au(111) . . . . . . . . . . . . . . . . 67
5.2.2 Trilayer NaCl/Au(111) . . . . . . . . . . . . . . . . 71
5.2.3 Influence of the Au(111) reconstruction . . . . . . 72
5.3 Interaction between the NaCl and the Au(111) surface . . 74
5.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6 Surface-reconstruction at play 77
6.1 NaCl/Au(111) interface state . . . . . . . . . . . . . . . . 79
6.2 Resolving all NaCl atoms . . . . . . . . . . . . . . . . . . 84
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
7 Co atoms on and F-centers in NaCl/Au(111) 91

7.1 F-Centers in NaCl(2 ML)/Au(111) . . . . . . . . . . . . . 91
7.2 Co atoms deposited on NaCl(3 ML) . . . . . . . . . . . . 94
7.2.1 Topography . . . . . . . . . . . . . . . . . . . . . . 94
7.2.2 Local density of states . . . . . . . . . . . . . . . . 95
7.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8 Imaging Co nanoclusters with atomic resolution 99

8.1 Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . 100
8.2 Electronic properties . . . . . . . . . . . . . . . . . . . . . 103
8.3 Atomic reconstruction . . . . . . . . . . . . . . . . . . . . 105
8.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Contents 3
Conclusions 109
Nederlandse Samenvatting 113
Bibliography 117
Publications 129
Curriculum vitae 133

4 Contents
The nanoscale: Where size
does matter
Leucippus proposed the concept of the atoma (atom) in the fifth cen-
tury BC when thinking about the behavior of a substance as it becomes
smaller and smaller. The reverse process from an atom to its bulk coun-
terpart is now known to involve quantum size effects, making that the
properties of conglomerates of atoms and molecules do not simply scale
with the particle size. It are precisely these quantum size effects that
lead to the expression “small is different”, explaining why the synthesis
of nanoparticles with controlled size and composition is of fundamental
and technological interest. On the experimental side, the rapid develop-
ment of new techniques for producing and for probing nanoparticles has
resulted in a phenomenal increase in our knowledge of these systems.
The effort to understand the physics of ever smaller structures has been
paralleled by attempts to exploit their properties in applications, and
has indeed led to new ideas for advanced materials and devices. These
efforts have ensured that the controlled growth and the detailed char-
acterization of objects of reduced dimensionality is currently one of the
most active research areas in modern science.
Two major classes of effects lead to this interest. One class of effects
is related to the large ratio of number of surface atoms to bulk atoms in
a nanostructure. As an example, for a spherical nanoparticle of radius
R composed of atoms with an average spacing a, the ratio is given by:
Nsurf /N ∼ 3a/R [1]. For R = 6a ∼ 1 nm, half of the atoms are on
the surface. The large surface area of nanoparticles is advantageous
for applications in gas storage, where molecules are adsorbed on the
surfaces, or in catalysis, where reactions occur on the surface of the
catalyst (Fig. 1). The increased surface area and size-specific surface
chemistry of metal nanoparticles are being used for some time now to
5
6 The nanoscale: Where size does matter
optimize the activity and specificity of catalysts [2,3].
Figure 1. Schematic representation of a cubo octahedral metal

cluster on an insulating substrate. Chemically different sites on
the cluster surface are colored differently. Figure taken from [4].
The second class of effects is due to the electronic energy levels in

low dimensional structures, whose energy spacing becomes larger than
kb T due to the reduction in size. The increased energy spacing induces
a breakdown of the bulk band structure as illustrated in Fig 2 A for
both a metallic and a semi-conducting quantum dot. This breakdown
can happen in one, two, or all three dimensions (Fig 2 B), and deter-
mines many important properties of the nanostructured material [5]. In
the case of semiconductor nanocrystals size-tunable optical properties
have been shown to exist, and have been integrated into exploratory
optical and electronic devices [6], while similar size dependent proper-
ties in metal nanoparticles are being explored for use in medicine [7].
It is this dependency of electronic states on their confinement
conditions that is the primary subject of this thesis.
A first class of confined electrons under study is the so-called two
dimensional electron gas (2-DEG), provided by electrons confined to the
first few atomic layers of a metal substrate. We are concerned with both
the interaction between the 2-DEG and electrons confined to nanopar-
ticles, and how the properties of the 2-DEG changes upon changing its
environment. The interaction between the 2-DEG and nanoparticles is
Figure 2. (a) Schematic illustration of the density of states in

metal and semiconductor clusters. (b) Density of states in one
band as a function of dimension. Figure taken from [6].
realized by depositing nanoparticles on the metal surface, whereas the

environment of the 2-DEG is changed by growing a thin insulating film
on top of the metal surface.
The second class of confined electrons are electrons strongly confined
in all three dimensions, experimentally realized in three different ways in
this work. First, when creating a defect in an insulating layer electrons
occupy spatially confined defect states localized at the vacancy site,
called a F-center. These F-centers are created in the thin insulating
film on top of the metal surface. A second way of obtaining strongly
confined electrons is using single atoms deposited onto the insulating
layer. The third and final system of electrons confined in all dimensions
is an individual nanoparticle on the insulating surface.
Investigating these confined electron systems in a reliable way re-
quires appropriate tools for measuring at the nanoscale. Diffraction
techniques such as X-ray diffraction and small angle X-ray scattering
are not suited for this research. Although they yield both detailed
structural and chemical sample information, this sample information
is averaged over a large sample area, making it impossible to study in-
dividual nanosized objects. Real-space information at the atomic scale

can be provided by scanning probe techniques, in particular by scan-
ning tunneling microscopy (STM) and scanning force microscopy. In
contrast to diffraction experiments, both techniques are able to probe
individual nanostructures and are therefore suited to study the spatial
distribution and shape of individual nanoscale systems. For this study
we made use of STM which, when combined with scanning tunneling
spectroscopy (STS), under controlled conditions offers the ideal means
to study in detail down to the atomic scale both the morphology and
electronic properties of objects at the nanoscale.
Chapter overview
This thesis is divided into eight chapters.
In Chapter 1 we will give a brief introduction of the properties of
electrons for different types of confinement regimes. We will quantita-
tively introduce the phenomenon of electrons confined to metal surfaces
and explain the basics of the quantum mechanical principles laying at
the heart of the interest in these nanoscale systems.
Chapter 2 is devoted to the measurement techniques that are used
for this work, i.e., STM and STS. We will first discuss the general config-
uration of a scanning probe microscope as well as the underlying physical
quantities that are needed to understand the working principle of STM.
Afterwards, the physics needed to interpret the measurements correctly
is discussed and it is explained how information on the sample density
of states can be extracted by means of different types of STS. At the
end of the chapter, the low-temperature UHV STM setup used in this
work and its facilities are briefly presented and the preparation details
of our STM tips are given as well.
In Chapter 3 we provide experimental details. The atomically flat
Au(111) surface is introduced and we give the deposition and clean-
ing conditions to obtain this surface. Its characteristic surface recon-
struction is discussed and we present a short overview of its electronic
structure. We then continue to introduce the cluster deposition setup
and give typical conditions for the growth and deposition of clusters on
surfaces.
In Chapter 4 TiOx clusters on atomically flat Au(111) films are
investigated. We present measurements of both the morphology and the
electronic properties of these clusters. First, pure Ti clusters are pro-
duced in the gas phase and are deposited onto a clean Au(111) surface
under controlled ultrahigh vacuum conditions. Next, the clusters are
oxidized. Finally, STM measurements are performed before and after
annealing up to high temperatures (970 K), indicating a high thermal
stability of the deposited clusters. STS reveals an n-type semiconductor-
like gap in the TiOx cluster density of states. Local influence of the clus-
ters on the electronic surface state of the Au(111) substrate is observed,
which is relevant for catalytic cycles where certain steps of the cycles
take place at the metal-oxide interface.
To minimize the interaction between clusters and surface, we investi-
gate in Chapter 5 the growth and the electronic properties of crystalline
NaCl layers on Au(111) surfaces, which can isolate the clusters from the
Au(111). Deposition of NaCl on Au(111) at room temperature leads

to the growth of bilayer NaCl, which can be changed into trilayer NaCl
by post annealing. The Au(111) Shockley surface survives as an inter-
face state band at the NaCl/Au(111) interface. Using Fourier-transform
images of maps of the local density of states, the energy versus wave
vector dispersion of this interface state and of the bulk electrons were
measured. By changing the bilayer NaCl into trilayer NaCl, we were
able to modify the dispersion of both the interface and the bulk state.
Ever since the first scanning probe measurements with atomic reso-
lution on alkali halides, only one of the atomic species could be imaged.
[8,9] Simultaneously resolving both alkali and halogen atoms, which has
always been the ultimate goal of surface microscopy, is essential for quan-
titative scanning probe microscopy studies of point defects, [10] adsor-
bates, [11,12] tip induced manipulation, [13] and other surface processes
on the atomic scale. In Chapter 6 we demonstrate the imaging of
NaCl with unprecedented atomic resolution by exploiting the Au(111)
herringbone reconstructed surface to simultaneously visualize both the
sodium and the chlorine component of the thin alkali halide bilayer.
In Chapter 7 we use the thin NaCl film from the previous two chap-
ters as a substrate for the creation of extremely small zero-dimensional
systems. First, the properties of F-centers in trilayered NaCl are experi-
mentally determined as well as the interaction between F-centers in close
proximity. The second part of this chapter is an experimental report on
the behavior of adsorbed Co atoms on bilayered NaCl.
Finally, in Chapter 8 we use STM to determine the morphology
of preformed Co nanoclusters deposited onto a thin insulating film with
atomic precision. As the exact atomic configuration of a nanoparticle
strongly determines its properties (Fig. 1), this is a very important re-
sult. Nevertheless, such a determination has only been done before in a
few cases [14,15], and never with STM.
Chapter 1
Reducing dimensions
In this era of miniaturization of electronic devices there is an increas-

ing need to understand the properties of structures with dimensions of
the order of nanometers. The engineering of electron wave functions in
reduced dimensions has allowed researchers to explore and visualize fun-
damental aspects of quantum mechanics and has also led to new ideas
for advanced materials and devices.
However, one does not need to go to very special geometries to

encounter confined electrons. Due to the strength of electromagnetism,
in all normal substances electrons are bound to atoms, molecules, or
bulk matter. This makes quantum confinement of electrons very relevant
when developing models for solid state physics. Therefore, even when
discussing electrons in large pieces of matter (section 1.2), one must
start with introducing the general quantum mechanical behavior of a
confined particle (section 1.1).
After this general introduction the remainder of the chapter deals

with the reduction of dimensions in two sections. Section 1.3 deals with
surface states, i.e., electrons confined to the nanoscale in one direction,
but not in the other two. Systems of electrons confined in one dimen-
sion are called two dimensional (2D) systems, because of the two “free”
dimensions. Section 1.4 introduces the reader to the properties of elec-
trons confined to the nanoscale in all three dimensions, so-called zero
dimensional (0D) systems.
11
12 Reducing dimensions
1.1 Confinement of particles

The laws governing atomic-size particles are quite different from the
laws governing matter on the normal human length scale. A new set of
laws has had to be developed to deal with these small particles, since
our intuition and our classical laws fail us when dealing with the small.
This new set of rules is known as quantum mechanics and they tell us
that the behavior of atomic scale particles is guided by so-called wave
functions, commonly denoted with ψ. These wave functions are obtained
by solving the Schrödinger equation of the system under study, which
in the time-independent one-dimensional case is given by
~2 d 2 ψ
− + V ψ = Eψ , (1.1)
2m dx2
where V denotes the potential energy of the system [5]. To gain an
understanding of confinement of charge carriers in quantum mechanics,
one has to start by solving this equation.
1.1.1 The infinite square well

The infinite square well is the simplest theoretical model to study the
phenomena of confinement in quantum mechanics. As stated in [5], de-
spite the simplicity of this system it serves as a wonderfully accessible
test case where most of the important properties of quantum confine-
ment can be studied. In the model a particle of mass m is confined to a
length L by infinitely high barriers. The potential energy of the electron
inside the well is zero, and infinite elsewhere. The time-independent
Schrödinger equation for this system in one dimension for 0 < x < L
reads [5]
~2 d 2 ψ
− = Eψ . (1.2)
2m dx2
As in [1], we use the term orbital to denote a solution of the wave
equation for a system of only one particle. The term allows us to distin-
guish between an exact quantum state of the wave equation of a system
of N interacting electrons and an approximate quantum state which we
construct by assigning the N electrons to N different orbitals, where
each orbital is a solution of a wave equation for a single non-interacting
particle. The orbital model is exact only if the interactions between
electrons can be neglected.
1.1 Confinement of particles 13
Figure 1.1. (a) First three energy levels and wave functions of a
particle of mass m confined to a line of length L. The energy levels
are labeled according to the quantum number n which gives the
number of half-wavelengths in the wave function. (b) Square of
the wave-functions shown in (a). Note that the quantum number
n is one less than the amount of minima of ψ 2 . (c) Energy versus
~k dispersion for both a confined and a free particle.
From the boundary conditions it follows that ψ(0) = ψ(L) = 0 [5].

From these boundary values and Eq. 1.2, it follows that the solutions of
Eq. 1.2 take the following form [5]:
r
2 nπ
ψ(x) = sin(kn x) , kn = , (1.3)
L L
with n = 1, 2, 3, ... (Fig. 1.1 a). The energy En of each solution is given
by [5]
~2 kn2
En = . (1.4)
2m
Note that, in sharp contrast with classical mechanics, a quantum
particle cannot have a continuously varying energy, but it has to have
one of the specific allowed energy values [Fig. 1.1 (c)]. Historically, one of
the first systems where this energy quantization became apparent is the
atom. An electron bound to a proton can only be in certain orbitals,
explaining the line spectrum of hydrogen [16]. For very large L the
energy spectrum becomes continuous, as is expected classically, because
the spacing between the allowed energy levels varies proportional to 1/L2
and thus vanishes for large L.
So far we did not mention what these wave functions are and what
they do for you once you find them. This is provided by Born’s sta-
tistical interpretation of the wave function [5], which learns that ψ 2 (x)
[Fig. 1.1 (b)] gives the probability of finding the particle at point x. As
an example, for the lowest energy state (n = 1) it is very likely to find
the particle in the middle of the potential well, and very unlikely to find
the particle at the edges. For certain systems, it is possible to measure
this ψ 2 , as will be discussed in section 1.4.
1.1.2 Particles trapped in three dimensions

In the real world eventually each system, including the atom, is three
dimensional to a certain extent. Fortunately, extending the model of the
one-dimensional infinite square well to three dimensions can be done
in a straightforward way. The generalization of the time-independent
Schrödinger equation for three dimensions is [5]
~2 2
− ∇ ψ(~r) + U (~r) = Eψ(~r) , (1.5)
2me
where ∇2 = ∂ 2 /∂x2 +∂ 2 /∂y 2 +∂ 2 /∂z 2 and ~r is a vector with components

x, y, and z. Following the one-dimensional infinite square well it is not
difficult to see that the solutions for a particle in a three-dimensional
box of volume L3 with infinitely high walls are given by [5]
2
ψn (x, y, z) = ( )3/2 sin(knx x) sin(kny y) sin(kny y) , (1.6)
L
with knx = nx π/L, nx = 1, 2, 3, ..., and similar expressions for kny and
knz . The allowed energies are given by
~2 (kn2 x + kn2 y + kn2 z )

En = . (1.7)
2m
Equation 1.7 also holds for a rectangular box, instead of a cube, with
the modification that then knx = nx π/Lx where Lx is the size of the box
in the x direction, and similarly kny = ny π/Ly , and knz = nz π/Lz . If
the bottom of the well is nonzero, one has to add an additional offset
energy E0 to Eq. 1.7.
1.2 The nearly-free electron gas inside metals 15
1.2 The nearly-free electron gas inside metals

1.2.1 Metal as a 3D electron box
Many of the properties of metals can be explained in terms of the free
electron model [17]. According to this model, when atoms of a metallic
element are brought together, the valence electrons of the atoms become
detached and can move almost freely through the volume of the metal.
These free electrons then form a nearly-free electron gas within a box,
defined by the size of the metal. In everyday conditions, the interference
between the electrons is averaged out, leading to a remarkable accuracy
for simple models that do not include electron interference [17].
A model for a metal box filled with electrons that follow the Pauli
exclusion principle, explained below, is presented in Fig. 1.2. The model
used here is the Jellium model in which it is assumed that the positive
charges (i.e. atomic nuclei) are uniformly distributed in space [18]. The
energy needed to take an electron from the box and bring it to the
vacuum is given by the work function Φ, while the depth of the potential
well that contains the electrons in the metal is given by V0 .
Figure 1.2. Metal as a 3D box filled with non-interacting elec-

trons up to the Fermi energy EF , following the Pauli exclusion
principle. The total depth of the potential well is V0 , the sum of
EF and the work function Φ. Figure taken from [19].
Fermi energy
A very important aspect of governing the behavior of these electrons
inside this metal box, is the Pauli exclusion principle. As discussed in
section 1.1, in each system only specific wave functions ψ are allowed,
each with its own energy. An electron belonging to a particular wave

function ψ is said “to sit in state ψ”. The Pauli principle now states that
if one electron sits in a well-defined described state, this state should be
regarded as occupied and cannot hold another electron [20]. So far in our
discussion we have ignored the property “spin” of the electrons. Is does
not change the discussion as it is brought here substantially, but due to
spin two electrons can share a certain wave function, as long as their
respective spins are opposite. The Pauli principle implies that when we
start adding electrons to the metal box, the first two electrons will sit
in the state with the lowest energy, and the energy of the filled levels
will subsequently increase. At temperature T → 0, where no additional
energy is available, only the lowest levels will be filled with electrons.
The energy of the highest occupied state is referred to as the Fermi
energy (EF ) and states above EF are empty (Fig. 1.2).
The concept of the Fermi energy is a very important concept in solid
state physics. The main reason for this high importance of the Fermi
energy is its large value in comparison to other physical quantities. For
most metals the Fermi energy is of the order of 5 to 10 eV. On the other
hand, at room temperature the available thermal energy kb T is 0.026 eV.
For an electron to absorb this available energy, it needs to be able to go
to a level with a higher energy than its original energy. Due to the Pauli
principle an electron can only jump to a state with higher energy if this
state is unoccupied. Therefore, electrons far below the Fermi energy
cannot take part in physical processes involving energies of a fraction of
an eV, as there are no unoccupied states available to jump to (Fig. 1.3).
The combination of the large Fermi energies and the Pauli principle
implies that the properties of metals for physical processes which happen
on a small energy scale compared with EF are determined by a small
fraction of electrons with an energy close to the Fermi energy. If we take
Cu as an example (EF = 7 eV [17]), a quick estimation yields that only
electrons in about 0.4 % of the energy range (kb T /EF ) are influenced
by the thermal energy at room temperature. Another everyday physical
process that happens on an energy scale much smaller than EF is the
electrical conductivity of copper. As the mean free path of an electron
is of the order of 10 nm [17], the energy given to an electron in a 1 m
copper wire by applying 100 V is of the order of:
V
W = eEd = e × 100 × 10 nm = 0.000001 eV , (1.8)
m
revealing that electrical conductivity in a normal copper wire only in-
1.2 The nearly-free electron gas inside metals 17
volves small energies when looking at single electrons.
Figure 1.3. Density of single-particle states as a function of en-

ergy for a free electron gas in three dimensions. The shaded area
represents the filled orbitals at absolute zero. The dashed curve
represents the density of filled orbitals at a small but finite tempe-
rature T . The effect of the thermal energy on the electron dis-
tribution is not drawn to scale. The average energy is increased
when the temperature is increased from 0 to T and electrons are
thermally excited from region 1 to region 2. Figure taken from [1]
Density of states
As most of the important properties of materials are determined by the
electrons close to the Fermi energy, it is important to know how much
electron states are available in a small energy interval. The amount
of available states per energy and per unit volume in a small energy
range dE is the density of states (DOS). In a three dimensional box
as described in section 1.1.2, the distance between allowed kn values is
2π/L, as indicated in Eq. 1.6. The number of states between k and
k + dk then equals
3
dN3D L
=2 4πk 2 . (1.9)
dk 2π
Eliminating k from Eq. 1.9 using the E(k) relation of Eq. 1.7 and divid-
ing by the volume (L3 ) gives then the DOS as:
√ 3
dN3D 2π 2 m 2 √
g(E)3D = = E, E ≥ 0 . (1.10)
V dE ~3
For a three dimensional metal the DOS increases with energy, as illus-
trated in Fig. 1.3.
1.2.2 Nearly-free electron model

In our discussion so far we have, except for the positive background
that they provide, neglected the ion cores of which solid state matter
exists. However, the theory used so far explains to a remarkable degree
most of the important properties of metals. It turns out that in metals
the electrons can “sneak by all the ions in a metal” [21], as proved by
Bloch [22]. Because the electrons in a perfect crystal are arranged in a
regular periodic array, the potential energy of electrons in a crystal is of
the form
U (~r + ~a) = U (~r) (1.11)
for all Bravais lattice vectors ~a.

It turns out that the solutions of the Schrödinger equation con-
taining a potential energy of this form differ only from the solutions of
Eq. 1.5 by a periodic modulation, i.e.,
~
ψ~k (~r) = eik·~r u~k (~r) , (1.12)
where
u~k (~r + ~a) = u~k (~r) . (1.13)
Electrons fulfilling these requirements are called Bloch electrons, in con-

trast to free electrons.
The form of the equation for the energy of Bloch electrons is to a
good approximation the same as for the equation for the energy of free
electrons (Eq. 1.7), but one has to substitute the mass of the particles by
an effective electron mass m∗ , which can be experimentally determined
as will be done in chapter 5. For the full implications of the Bloch
theorem we refer the reader to [17] or [1].
1.3 Electrons at metal surfaces 19
1.3 Electrons at metal surfaces

In the previous section we have assumed that all three dimensions of our
metal box are of the same order, and very large. Let us now consider
what happens if one dimension (z) is made very small (∼ d). Electrons
in such a system have the following energy [19]:
h2 n2z ~2 (kn2 x + kn2 y )

En = E0 + + . (1.14)
8md2 2m
If d is much smaller than either Lx or Ly , knx or kny have to become very
large before it is favorable to put an electron in a state with nz = 2. As a
result one usually works with kx and ky , which can be seen as continuous
variables. In this situation the quantum number nz is called the sub-
band index and for very strong confinement all particles are in the first
sub-band. Electrons in these systems can only jump to a higher energy
by changing their momentum in the x and y direction, as changing nz
requires too much energy. The electrons are effectively bound to move
in x − y planes and these systems are called 2D electron gases.
An important class of 2D electron systems is formed by the so-called
“surface states” at bulk crystal surfaces. Unlike in traditional solid state
physics where one deals with quasi-infinite periodic solids, in any real
crystal the solid is terminated by surfaces. At these surfaces the Bloch
waves are scattered and reflected back into the solid. Besides these bulk
states, new states can arise on the surface under certain conditions [23].
The electrons in these new states are being confined to the top atomic
layers (by the vacuum barrier at one side and at the other side by the sp-
band gap in the bulk valence states that is created by the termination
of the crystal at the surface [23]) and therefore act as a (quasi) two-
dimensional free electron gas [24–26].
The presence of a surface state is a typical property of various (111)
oriented surfaces, including Cu(111), Ag(111), and Au(111), as well as
other surfaces such as Be(0001) [27]. The Au(111) surface state, im-
portant in this work (section 3.1.3), was visualized for the first time by
Hasegawa et al. in 1993 by means of STM at room temperature [28].
Surface states play a very important role in several processes, in
particular the behavior of adsorbates on the metal surface. Indeed,
unlike bulk states which rapidly die out near the surface, surface states
extend further into the vacuum region, thereby strongly interacting with
molecules in the vicinity of the metal surface [29]. Therefore, surface
states play a mayor role in determining the chemistry of metal surfaces.
On the other hand, the adsorbates will influence the surface states as
well, turning them into a sensitive probe for surface processes.
A second reason for the interest in surface states comes from fun-
damental physics, where one engineers the surface states to investigate
and exploit in great detail the properties of the quantum world.
Surface states
For a normal piece of solid state matter there is no chance to solve the
Schrödinger equation exactly, for it includes all the constituting ions
and electrons. Therefore, many approximations are developed to deal
with this problem in solid state physics. In bulk metal solutions of
the Schrödinger equation constitute of three dimensional Bloch func-
tions (section 1.2.2), respecting the Born - von Karman boundary con-
ditions [17]. The Bloch functions lead to the formation of the bulk
electron band structure and are at the heart of solid state physics [17].
However, this picture breaks down at the surface where the translation
symmetry is broken in the direction normal to the surface.
The formation of surface states can be explained straightforward in a
quantitative way. As stated in section 1.2.2, the electron wave functions
in crystalline matter are of the following form [17]:
~
Ψ (~r) = eik·~r u(~r), u(~r + ~a) = u(~r) , (1.15)
where ~k denotes the Bloch wave vector and ~r is a spatial coordinate.

Note that ~k does not need to be real in these equations. If the electron
wave vector has an imaginary component ~κ, we can rewrite 1.15 as
~
Ψ (~r) = [eik·~r u(~r)]e−~κ·~r . (1.16)
Here ~k is now only the real part of the Bloch wave vector. The wave
function described by Eq. 1.16 grows exponentially in the direction op-
posite to ~κ. In an infinite crystal, like one obtains when respecting the
Born - von Karman boundary conditions, such levels have no physical
meaning, since the electron density has to be finite everywhere. On the
other hand, if there is a surface perpendicular to ~κ, then one can join
solutions of the form 1.16 within the crystal, which grows exponentially
towards the surface, with one that is exponentially damped outside the
crystal (Fig. 1.4). Solutions obtained this way are referred to as surface
states. For a more complete explanation of this phenomena, we refer
the reader to [30].
1.3 Electrons at metal surfaces 21
Figure 1.4. One dimensional semi-infinite lattice model poten-

tial (solid line). The surface state shows the highest localization
at the surface and decays exponentially into the vacuum as well
as into the crystal (dashed line). The figure is taken from [31].
Parallel to the surface, however, the electrons can move freely, lead-
ing to two dimensional Bloch functions with a real wave vector ~k|| . For
each ~k|| a surface state exists, localized at the surface of the metal.
We thus obtain a two-dimensional surface state band, which can be de-
scribed using Eq. 1.14, leading to the following energy versus wave-vector
dispersion:
E(k|| ) = E0∗ + ~2 k||2 /2m∗ . (1.17)
Here m∗ is the effective electron mass, typically given in units of the

free electron mass me , and E0∗ is the onset energy of the surface state,
obtained by adding the energy from the nz = 1 state to the original onset
energy E0 . Values for the surface states on noble metal (111) surfaces
can be found in Table 1.1
Density of states in two dimensions

Similar to what was done in section 1.2.1, we will calculate the DOS,
but now in the two-dimensional case. The distance between allowed kn
values is still 2π/L for the x and y direction, as indicated by Eq. 1.6.
The number of states between k and k + dk now becomes:
dN2D L
= 2( )2 2π~k . (1.18)
dk 2π
noble metal E0∗ (meV)) m∗ /me

Cu(111) -435 0.412
Ag(111) -63 0.397
Au(111) -487 0.255
Table 1.1. Values of the surface state of Cu(111), Ag(111), and

Au(111). E0∗ is given with respect to EF . All values are taken
from [32].
Eliminating k from Eq. 1.18 using the E(~k) relation of Eq. 1.17 and
dividing by the surface area (L2 ) gives then the DOS:
dN2D 4π4m
g(E)2D = = , E ≥ E0∗ . (1.19)
SdE π~2
Thus in two dimensions the DOS is a constant above the onset energy
E0∗ , as confirmed experimentally in section 3.1.3.
1.4 0D systems
0D systems are formed by strongly confining electrons, not only in one
dimension as in section 1.3, but in all three dimensions. For electrons,
such systems can be obtained by confining surface state electrons in the
directions along the surface, or by reducing the size of a bulk slab of
metal to the nanoscale. Finally, this reduction in size leads to the most
well known 0D systems, i.e., the single atoms, which are the smallest
possible version of 0D electron systems. In literature there is no well
defined upper limit for 0D systems, but in this work we will restrict
ourselves to systems with a size in the order of nanometers and thus a
countable number of atoms.
When we go to the nanoscale L becomes small, implying that the
energy spacing between allowed electron levels (see Eq. 1.7), becomes
noticeable. This implies that in this case the DOS is not a continuous
function of energy, but is only nonzero at specific allowed energies. As
an example, for a spherical Au nanoparticle with a radius of 2 nm the
average level spacing is ∆E ∼ 2 meV [1]. Following the reasoning
of section 1.2.1 electrons in these nanoparticles cannot be influenced
by processes that happen on an energy scale smaller than ∆E, since
there are no available allowed states for electrons to jump to. Since ∆E
1.4 0D systems 23
scales inversely proportional with the size of the nanoparticles, one can
tune this cutoff in interaction energy by selecting nanoparticles with the
appropriate size.
Using STM the individual orbitals of 0D systems and their energy
spacing can be revealed by dI/dV maps (chapter 2.2.2), as is illustrated
in Fig. 1.5. The 0D system presented in Fig. 1.5 is obtained by confining
surface state electrons, naturally trapped in the z direction, in the x and
the y directions. As explained in section 1.1.2, certain quantum numbers
nx , ny , and nz correspond to each quantum state. As the measured
electrons come from a surface state, nz = 1. One can obtain the other
two quantum numbers by counting the minima of the measured orbitals
for each direction, and subtracting one, as explained in section 1.1.1
(Fig 1.5). By changing the boundary conditions of the Schrödinger
equation, one can also explain the energy levels of electrons confined to
more complicated shapes than rectangles [33].
Figure 1.5. Array of 28 Mn atoms forming a rectangle of size

90 × 100 Å2 . Panels (a)-(h) show constant current topographs
recorded at the indicated sample voltages. Panels (i)-(p) are
dI/dV maps recorded simultaneously. The first four dI/dV maps
represent respectively the orbitals (0,0,1), (1,1,1), (1,2,1), and
(2,2,1). Figure taken from [34].
Electron transport through nanoparticles

To allow for electronic transport through a 0D system, one has to bring
the system in contact with electronic leads. If the contact is strong,
the electrons of the 0D system can easily escape to the leads, thereby
destroying the phenomena stemming from quantum confinement. As we
want to investigate these quantum effects precisely, we need to work in
the limit where the contacts between the 0D system and its environment
are weak. This way, the electrons are located either in the leads or on the
nanoparticle. To achieve this the time it takes an electron to tunnel to

and from the 0D system should be larger then any quantum fluctuation
[5] for which
~
∆E∆t ≥ . (1.20)
2
A typical time needed to charge or discharge a 0D system weakly con-
nected to electronic leads is classically given by:
∆T = Rt C , (1.21)
where Rt denotes the resistance between the 0D system and the leads,
and C is the capacitance between the 0D system and its environment.
The relevant energy to be considered here is the charging energy, EC ,
of the 0D system. This energy is given by EC = e2 /2C, where e is the
fundamental electron charge. Therefore, Eq. 1.20 now becomes:
e2 ~
Rt C ≥ . (1.22)
C 2
To retain the 0D nature of a system, the resistance Rt between the
system and its leads should be (much) larger than 2e~2 . Only when this
criterium is fulfilled one can investigate the properties of an isolated 0D
system.
Coulomb blockade
When investigating the electronic transport through 0D systems, there
is an important secondary effect that should be taken into account, i.e.,
the presence of a so-called “Coulomb blockade” [35]. This effect stems
from the fact that the charge on an island can only change by a quantized
amount e. When one wants to add an electron to the 0D system, one
needs to overcome the electrostatic potential (EC = e2 /2C) needed to
charge the island. Therefore, in a region of e/C around zero bias, one
cannot add or remove electrons for the 0D system and therefore no
current can flow in the circuit [Fig. 1.6 (a)].
When the voltage across the 0D system is raised above the charg-
ing energy, electrons can start to go one by one from the lead with the
highest energy through the 0D system to the other lead (Fig. 1.6). How-
ever, if the voltage is only slightly above the charging energy, only one
of the possible quantized states of the 0D particle (section 1.1.2) can be
1.4 0D systems 25
used for this electron transport [Fig. 1.6 (b)]. Upon raising the voltage
further, more and more states can contribute to the transport and the
current will increase with each new available state [Fig. 1.6 (c)]. Thus,
by measuring I(V ) curves on single 0D systems, one can learn about
their intrinsic allowed electron orbitals and thus their specific electronic
properties. However, in order to measure these individual level spacings
one needs to work at low temperatures as quantum fluctuations will
destroy the discreteness of the electronic states when kb T ∼ ∆E [35].
Note that in this discussion it was assumed that the level spacing in
the 0D system was much smaller than the charging energy EC . For an
experimental measurement of a Coulomb blockade, see section 8.2.
Figure 1.6. Energy band diagrams of a 0D system coupled to

two leads at zero temperature. (a) The voltage drop across the
nanoparticle is zero and no current can flow due to the Coulomb
blockade. (b) The voltage is raised above the charging energy and
electrons can flow through one of the orbitals of the 0D system, as
indicated by the gray lines. (c) The voltage is raised even higher
and electrons can now pass through the first two orbitals of the
0D system, increasing the current.
Chapter 2
Exploring the nanoworld

with a scanning tunneling
microscope
For the detailed exploration of the new aspects of the nanoworld ap-
propriate measurement tools are needed. Since a reliable exploration
requires simultaneous registration of both the local morphologic and
electronic properties of individual nanostructures with unprecedented
resolution, a special microscope with very high resolution is needed.
The special microscope used in this work is the scanning tunneling mi-
croscope. Here we will present the general working principles of STM
and briefly introduce the physical principles that make STM an unique
and powerful imaging tool in nanoscience (based on the work of K.
Schouteden [36]). For a deeper treatment of the technique and the the-
ory involved we recommend the book of Wiesendanger to the reader [37].
Together with its somewhat younger companion the atomic force
microscope, the scanning tunneling microscope is the founding mem-
ber of the family of scanning probe microscopes. STM was the first
technique that was able to generate real-space images of surfaces with
atomic resolution [38]. It is therefore not surprising that only five years
after its invention in 1981 by Binnig and Röhrer at the IBM Zurich Re-
search Laboratory, its spiritual fathers were awarded the Nobel Prize in
Physics. The reason for the ability to measure surfaces with atomic res-
olution is explained in section 2.1. Apart from topographic information,
also information on the local density of states (LDOS) of the sample
surface can be obtained by means of the spectroscopic STM variant,
27
28 Exploring the nanoworld with a STM
i.e., STS. How this is done is the topic of section 2.2.

The two complementary measurements make sure that STM is the
ideal tool to study quantum confinement phenomena on surfaces. For
this work we made use of STM in ultra-high vacuum, operating at low
temperatures, which will be introduced in section 2.3.
2.1 Scanning tunneling microscopy:

Imaging surfaces at the atomic level
Figure 2.1. (a) General setup of a scanning probe microscope.

(b) The concept behind scanning probe microscopy, reading
Braille at the atomic scale. If the interaction between tip and
sample decays sufficiently rapidly on the atomic scale, only the
two atoms that are closest to each other are able to “feel” each
other. (b) is taken from [39].
The general operation of any kind of scanning probe microscopy

(SPM) is based on two principles. First, all of the microscopes make use
of a sharp tip/probe that is brought in close enough vicinity of the sample
under investigation to detect the tip - surface interaction [Fig. 2.1 (b)].
Second, this interaction is measured while moving the tip according to a
raster pattern across the sample surface. Although the SPM setup has
been the subject of continuous developments to improve its operation,
the basic concept of the setup has never been altered. The general
configuration of an SPM is schematically presented in Fig. 2.1 (a).
To ensure accurate lateral and vertical displacement of the tip with
respect to the sample, either tip or sample [Fig. 2.1 (a)] is mounted
on a high resolution piezoelectric scanner. The piezoelectric material
2.1 Imaging surfaces at the atomic level 29
contracts or extends with increments of only a fraction of an interatomic

distance, when a voltage is applied to it. During the raster scanning of
the surface with the tip the local interaction between tip and sample is
recorded and regulated by means of a feedback system.
The resolution of SPM on flat surfaces is of the order of the atomic
sized features [Fig. 2.1 (b)]. However, for relatively pronounced surface
features with heights of 1 nm and more the finite bluntness of the tip (tip
radius R is typically a few nanometers) becomes increasingly important
and broadens to a certain extent the recorded size and shape of the
visualized surface features, as illustrated in Fig. 2.2. On the other hand,
the vertical resolution remains unaffected by the tip apex and can be as
good as 0.001 nm.
Figure 2.2. Schematic representation of a SPM scan line (dashed

line) over a single nanoparticle. Due to the finite radius R of the
STM tip, the width of the nanoparticle is overestimated in the
measurement. The measured height h of the individual particle
is, however, unaffected by the size of the SPM tip.
2.1.1 Scanning tunneling microscopy

In STM the tip-sample interaction can be described in terms of the
wave-function overlap of empty and filled states of a tip and sample (or
vice versa), respectively. This overlap makes electron transfer between
tip and sample possible, i.e. electrical current, while not making an ac-
tual mechanical contact between the two. According to classical physics
such a current is not possible, since classical physics allows only a cur-
rent flow between two electrically conducting materials when they are in
mechanical contact with each other. Quantum mechanically, however,
the wave functions associated with the electrons allow for a small cur-
rent of electrons to flow if tip and sample are brought to within only a
few interatomic distances (∼ 0.2 nm) of each other. This results in an
overlap of the wave functions of tip and sample, and therefore to a finite
probability of finding an electron beyond the barrier (being air, vacuum,
...). Under the influence of an externally applied electric field electrons
preferentially “tunnel” through this barrier from one electrode to the
other, resulting in a net current flow, which is consequently referred to
as the tunneling current I. This is schematically represented in Fig. 2.3,
where the electrodes are the STM tip and the sample. In our experi-
ments the externally applied bias voltages V are defined with respect
to the sample, while the STM tip is virtually grounded. Below we will
give a brief overview of the theoretical background regarding STM and
STS.
Tersoff-Hamann model
Just a few years after the invention of STM in 1981 [38], Tersoff and
Hamann developed one of the first successful theories that was able to
accurately describe electron tunneling behavior between a conducting
surface and an STM tip [40]. Starting from quantum mechanics, it
is found that electronic wave functions extend beyond the mechanical
boundaries of metals. Moreover, the surface wave functions ψν decay
exponentially outside of the metal, in the direction normal to the surface
(z ) [5,17]:
|ψν (→
−
z )|2 ∝ |ψν (0) exp [−z/ℓ0 ]|2 , (2.1)
where ℓ0 is defined as the decay length for wave functions into the vac-
uum barrier (ℓ0 is typically of the order of 0.1 nm):
~
ℓ0 = √ . (2.2)
2mΦ
Tersoff and Hamann demonstrated that the tunneling current I also
exhibits an exponential dependence on the tip - sample distance d [40]:
I ∝ exp (−2d/ℓ0 ) . (2.3)
To obtain this result, the tip apex was treated as a (locally) spherical
potential well with radius R (see Fig. 2.3), and only small externally ap-
plied voltages (on the order of meV) and moderate temperatures (room
Figure 2.3. Due to the exponential dependence of the tunneling

current on the tip-sample distance d, current flows mainly via the
outermost atom at the tip apex whenever an external voltage V
is applied between tip and sample. Apart from d, the resulting
tunneling current also depends on the electronic properties of tip
and sample, including the local density of states (ρt and ρs ) and
the work function (Φt and Φs ). Figure is adapted from [40].
temperature or below) were considered. The exponential dependence of

the tunneling current on the tip - sample distance is a very important
result, since it implies that the tunneling current is very sensitive to to-
pographic height variations at the sample surface. By raster scanning of
the sample with the STM tip in the x - and y-direction a topographic im-
age can be achieved. Generally, when one refers to an topographic STM
image, one refers to an image where the movements of the piezoelectric
scanner that are needed to maintain a constant tunneling current at a
certain bias voltage during scanning of the sample (closed feedback loop)
are plotted.
In addition, it can be derived from the Tersoff-Hamann model that
the effective lateral space resolution is related to the tip radius R and
the tip-sample distance d as [(R+d)ℓ0 ]1/2 [40]. Typical values (R = 9 Å,
d = 10 Å, and ℓ0 = 0.4 Å) yield an estimated lateral resolution of about
8 Å [40]. Although applicable to metal surfaces having a relatively large
surface periodicity, the spherical tip model is not able to explain atomic
resolution images on close-packed metal surfaces such as Au(111). An
increased spatial resolution can however be explained by the presence
of localized surface states or dangling bonds at the tip apex [41]. For
a more elaborate discussion on this we refer to [37]. This increased
spatial resolution can be as good as 0.01 nm when imaging atomically
flat surfaces.
Beyond the Tersoff-Hamann model
Although the Tersoff-Hamann model constitutes a valuable contribution

to understanding STM, the low-voltage approximation is often violated
in practice, as many STM experiments are performed at bias voltages of
up to 1 V and more, i.e., comparable to the barrier height Φ. At these
elevated voltages V, the tunneling current arises from a whole range of
states that lie within an energy eV from the Fermi level, as illustrated
in Fig. 2.4. The figure presents energy diagrams for different conditions
of the tunneling of electrons between STM tip and sample and for the
simplified case of a constant tip LDOS. The sample LDOS is assumed to
exhibit some specific broadened energy states related to the finite size of
the nanostructure under investigation. In the absence of an external bias
voltage V the tip-to-sample and sample-to-tip tunneling probability are
the same and there is no net tunneling current [Fig. 2.4 (a)]. Under these
conditions the Fermi levels of tip and sample are aligned. Note that the
different metal work functions Φt and Φs result in a non-rectangular
trapezoidal vacuum barrier. In the presence of an external bias voltage
V the sample LDOS is shifted either upward [Fig. 2.4 (b)] or downward
[Fig. 2.4 (c)] with respect to the tip LDOS, depending on the polarity of
the applied voltage V to the sample (negative or positive, respectively).
As a result, electrons tunnel in an energy window [0, eV] either via the
occupied or unoccupied sample states to or from the tip, respectively.
The expression for the tunneling current at large bias voltages can
be obtained by taking the integral over the involved tip and sample
states [37], yielding
Z eV
I(V ) ∝ ρt (EF − eV + E) ρs (→
−
r 0 , EF + E) dE , (2.4)
0
where ρt (E) is the LDOS of the STM tip. ρs (→−

r0 , E) is the LDOS of the
sample evaluated at the center of curvature of the tip →−r0 . ρs (→
−
r0 , E) can
Figure 2.4. Energy diagrams of an STM tip and a sample in

tunnel contact at different tunneling conditions. A constant tip
LDOS is assumed. EF is the Fermi level of tip and surface, U (V
in the text) is the applied sample bias. Φt and Φs are the work
functions of tip and surface, respectively. (a) Energetic equilib-
rium at zero bias. Net tunneling current (b) from sample to tip at
negative sample bias and (c) from tip to sample at positive sample
bias. Figure taken from [31].
be related to the LDOS of the sample evaluated at the sample surface

according to
ρs (→
−
r 0 , E) ∝ ρs (E) T (E, V, d) , (2.5)
via a generalized version of the transmission coefficient T(E, V, d) that
is based on the WKB formula (Wentzel, Kramers, and Brillouin) for elec-
tron tunneling through a one-dimensional potential barrier of arbitrary
shape [5]:
" 1 #
2m eV 2
T (E, V, d) = exp −2(R + d) Φ + − (E − E k ) , (2.6)
~2 2
in which the previously defined decay length ℓ0 of Eq. 2.2 now also
includes a dependence on the applied bias voltage V and on the energy,
including the total electron energy E and the electron energy component
Ek parallel to the sample surface [37].
The different lengths of the horizontal arrows in Fig. 2.4 reflect
the voltage dependence of the transmission coefficient: Higher energy
electrons experience a smaller effective barrier height and therefore have
a higher tunneling probability, implying that they also have a larger
contribution to the tunneling current I (see Eq. 2.4).
2.2 Scanning tunneling spectroscopy:

Revealing the density of states
2.2.1 Single point tunneling spectroscopy
As already discussed in Section 2.1, the exponential dependence of the
recorded tunneling current I on the applied bias voltage V provides
high-resolution topographic information about the sample surface. As
is clear from Eq. 2.4, however, the tunneling current contains more valu-
able physical information than just this height information, since I also
depends upon the sample LDOS ρs (E), where the range of contributing
energy states is determined by the applied bias voltage V (see Fig. 2.4).
This sample LDOS can be extracted by taking the first derivative of I
with respect to V in Eq. 2.4, yielding the following expression for the
differential conductance:
dI
(V, d) ∝ ρt (EF ) ρs (EF + eV ) T (E = EF + eV, V, d) (2.7)
dV
Z eV
d
+ ρt (EF − eV + E) ρs (EF + E) T (E, V, d) dE
0 dV
Z eV
d
+ ρt (EF − eV + E) ρs (EF + E) T (E, V, d) dE .
0 dV
Since the transmission coefficient T(E,V,d) varies only smoothly and

monotonously with the applied bias voltage (see Eq. 2.6) and under
the assumption of a constant tip LDOS ρt (E), the contribution of the
second and the third term can be neglected, yielding that (dI /dV )(V )
is proportional to the sample LDOS ρs (E) at energy E = EF + eV .
Measurement of (dI /dV )(V ) as a function of the applied bias voltage
V can thus yield the energy resolved LDOS at any location of the sample
surface.
For some cases, normalization if (dI /dV )(V ) by dividing by I/V
is needed to minimize the effect of the voltage dependence of T and the
influence of the tip-sample spacing. However, a significant problem can
arise if the current is non-zero at zero bias. In that case an artificial
band-gap develops around zero bias. Such a non-zero current can arise
from working with a lock-in amplifier (described below) or a small offset
in the voltage amplifier of the scanning tunneling microscope. As in gen-
eral the quantity (dI /dV )(V ) gives a good estimation of the LDOS [42],
we have not performed the (dI /dV )/(I/V ) normalization in this work.
2.2 Revealing the density of states 35
Local dI /dV data can be obtained in two different ways. In a

first method, the tunneling current I is recorded while linearly ramp-
ing the bias voltage, after having disabled the feedback loop with the
tip positioned above the location of interest. One thus obtains a local
I(V) curve which can then be numerically derived to obtain the sam-
ple LDOS ρs (E) at E =EF + eV . An alternative method is to directly
record the (dI /dV )(V ) spectrum by means of harmonic detection with
a lock-in amplifier, an STS technique that was first applied by Binnig
et al. [43]. For this purpose, a small sinusoidal modulation voltage Vmod
(30 to 60 mV in our work) with frequency ωL (700 to 900 Hz in our
work) is superimposed on the linearly ramped DC bias voltage VDC .
Consequently, the recorded tunneling current I[VDC + Vmod cos(ωL t)]
exhibits an oscillating behavior with frequency ωL . This signal is fed
into a lock-in amplifier where it is multiplied by a lock-in reference sig-
nal of frequency ωL , yielding an output signal that is proportional to
dI /dV |V =VDC (which is the first harmonic of I[VDC + Vmod cos(ωL t)]).
Note that the thus obtained sample LDOS ρs (E) is “averaged” over
the energy region [eV − eVmod , eV + eVmod ].
The energy resolution ∆E with which spectral features can be dis-
cerned in (dI /dV )(V ) measurements is limited by thermal broaden-
ing effects (due to the Fermi distribution) that smear out quantized
electronic states by an amount ∆Etherm ≈ 3.5kB T (≈ 1.4 meV at
4.5 K) [44,45]. As mentioned above, an additional broadening ∆Emod ≈
1.7 eVmod must be taken into account when acquiring dI /dV spectra by
means of the mentioned lock-in technique [45,46]. (dI /dV )(V ) curves
presented in our work are measured using the lock-in technique due to
the higher signal to noise ratio, unless stated otherwise.
2.2.2 Mapping of the local density of states

As discussed in the previous section, recording the dI /dV signal as a
function of the applied voltage V allows to determine the LDOS at a
specific location on the sample surface. Alternatively it is possible to
visualize the LDOS in differential conductance images dI /dV(x,y), often
referred to as “LDOS mapping”. These maps can be obtained in two
different ways.
First, LDOS maps can be extracted from local I(V) curves measured
with open feedback loop in a grid of typically 200 × 200 points, which
can be combined with topographic imaging with closed feedback loop.
From these data sets, the LDOS at a certain electron energy EF + eV is
extracted, as explained in previous section, and the spatial variation of

the thus obtained local dI /dV data at the bias voltage V is plotted. In
literature, this method is often referred to as current imaging tunneling
spectroscopy (CITS) [37].
Second, differential conductance images can be acquired using a
closed feedback loop by means of harmonic detection with a lock-in am-
plifier. For this purpose, a modulation voltage Vmod is added to the DC
bias voltage V that is applied to the sample. As discussed in the pre-
vious section, the lock-in amplifier extracts the tunneling conductance
dI /dV from the tunneling current I that is recorded while raster scan-
ning the sample surface in constant current mode, providing a map of
the sample LDOS at energy EF + eV . In the present work modulation
frequencies in the 700 − 900 Hz range and modulation amplitudes in the
30 − 60 mV range were used. All LDOS maps presented in our work are
obtained using this second technique.
2.3 The low-temperature UHV STM

STM and STS measurements were performed with a commercial scan-
ning tunneling microscope (Omicron NanoTechnology). The setup con-
sists of two chambers, the actual STM chamber and a sample prepa-
ration chamber, which are independently pumped by means of an ion
pump and additional titanium sublimation pump, ensuring operation
under UHV conditions at a base pressure below 5 × 10−11 mbar in both
chambers. The setup is presented in Fig. 2.5 (a).
Samples are introduced via a separate load-lock chamber that can
be pumped/vented separately from the STM and preparation chamber.
Samples are mounted on a large rotational transfer arm that further
transports the sample to the preparation chamber and the STM cham-
ber. This transfer arm is equipped with a heater stage to resistively
anneal samples in situ in the preparation chamber (Tmax ≈ 1200 K).
The chamber is further equipped with an argon ion gun (AG21 from
VG Scientific, see Fig. 2.5 (b)) and two evaporation cells [EFM3 from
Omicron Nanotechnology, see Fig. 2.5 (c)].
The microscope is mounted in a thermally shielded compartment
directly attached to a cryostat that consists of an inner and an outer
part, allowing measurements at room temperature (Tsample ≈ 300 K)
and at low temperatures. For working at low temperatures the outer
part is filled with liquid nitrogen, while the inner part can be filled with
2.3 The low-temperature UHV STM 37
Figure 2.5. (a) The low-temperature ultra-high vacuum STM

setup, consisting of the STM measurement chamber and the
preparation chamber. Inset (taken from Omicron Nanotechnol-
ogy): In the STM chamber the scanner and the sample stage are
present. The sample can be brought in contact with the cryo-
stat for cooling to low temperatures. The preparation chamber is
equipped with (b) an argon ion gun and (c) two evaporation cells.
Figure taken from [36].
either liquid nitrogen or liquid helium. This way, stable sample temper-
atures Tsample of 78 K (liquid nitrogen) and 4.5 K (liquid helium) can be
achieved. For optimal topography and energy resolution all measure-
ments were performed at 4.5 K, unless indicated otherwise in the text.
The time available for STM measurements at liquid helium temperatures
without refilling the cryostat is around 24 hours. A spring suspension
system with Eddy current damping ensures optimal vibration isolation.
The z-resolution of the STM is better than 0.1 nm.
The bias voltages V given in the text and figure captions are always
with respect to the sample, while the STM tip is virtually grounded.
Image processing was performed by Nanotec WSxM [47].
2.3.1 Preparation of STM tips

For the STM experiments we have relied on electrochemically etched W
tips (Omicron Nanotechnology) and on mechanically cut PtIr (10% Ir,
MaTeck GmbH) tips. PtIr tips were cleaned in situ by applying high
voltage pulses up to 10 V while being in close proximity with a clean
Au(111) surface, until stable topography and spectroscopy is achieved.
The W tips were cleaned in situ by repeated flashing well above 2000 K
in order to remove the surface oxide layer and any other surface contam-
ination. The known electron surface state of the Au(111) surface is used
as a quality label for our spectroscopic measurements (see Section 3.1.3
for details on the Au(111) surface state).
Chapter 3
Realization of systems with

reduced dimensionality
Reliable investigation of nanoscale systems requires the creation of very

clean and well-defined samples. This chapter gives a description of the
equipment and methods used to produce the systems investigated in
this work. The chapter is ordered so that the reader can follow the
production process step by step.
In section 3.1 we introduce the substrate used for all the experiments
in this work, i.e., the Au(111) surface. This surface provides us with a
good base on which the experiments are conducted. After introducing
the substrate, section 3.2 explains how nanoparticles can be created
and deposited on the substrate. However, as argued in section 1.4, to
investigate the properties of individual nanosystems, they need to be
electrically isolated from their environment. This isolation is achieved by
growing a thin NaCl layer on top of the Au(111) prior to the deposition of
the nanoparticles (section 3.3). Finally, section 3.4 explains the creation
of 0D systems on the NaCl by depositing Co atoms on top of the NaCl,
or by creating F-centers in the NaCl.
3.1 The Au(111) surface:

An atomically flat base
In order to systematically study individual nanoparticles by means of
local probe techniques such as STM, a very flat substrate is required.
This substrate must be conducting, in order to allow the tunnel current
to flow. An ideal candidate for this purpose is the (111) surface of Au.
39
40 Experimental
Au(111) oriented films can be grown on mica substrates and have been
used before in the Laboratory of Solid-State Physics and Magnetism [36,
48]. As the Au(111) substrate is used in all experiments of this work,
we will here give an overview of its most relevant properties based on
previous studies [36,48].
3.1.1 Growth and cleaning of Au(111) films

Epitaxially grown 140 nm thick Au(111) films on freshly cleaved mica
were prepared ex situ by molecular beam epitaxy (MBE) at elevated
temperatures [48]. The freshly cleaved mica is degassed in UHV by
annealing during 1 hour at T = 550◦ C prior to the deposition. Au is
deposited by thermal evaporation from a Knudsen cell containing high
purity gold (99.999 %). Evaporation rates (typically 0.03 - 0.1 nm/s)
are controlled by temperature stabilization and in situ calibrated by
quartz crystal oscillators. Au films with large atomically flat terraces
are obtained when the substrate is kept at a temperature of 490◦ C during
evaporation and at 510◦ C during post-annealing for 1 hour.
Sample transport from the MBE setup to the STM setup described
in section 2.3 is performed under ambient conditions. To remove surface
contamination the Au(111) films are cleaned in the preparation chamber
of the STM setup by repeated cycles of Ar ion sputtering (4 minutes at
4 keV and 1.10−6 mbar) and annealing (12 hours at 720 K). After this
cleaning procedure we end up with gold films consisting of large atom-
ically flat areas that are separated by steps of one or more monolayers
high (1 ML Au(111) = 0.235 nm). For all samples surface quality is
checked by STM prior to the growth/deposition of NaCl or clusters.
3.1.2 The “herringbone” surface reconstruction

In contrast to other close-packed (111) surfaces, gold is the only fcc
metal that exhibits a surface reconstruction. It was only in 1989 that
this reconstruction was visualized for the first time with atomic resolu-
tion by means of STM by Wöll et al. [49]. The mechanism behind this
peculiar deformation originates from a reduction of the surface tension
that is achieved by a local contraction of the top atomic layers of only
a few percent
√ along one of the three <11̄0> directions (see Fig. 3.1).
The 22 × 3 unit cell of the surface therefore contains 23 surface atoms
on top of 22 substrate atoms. A modulated surface structure is hereby
formed, locally varying between bulk fcc stacking (broader region be-
3.1 The Au(111) surface 41
tween the ridges in Fig 3.1) and hcp stacking (narrower regions between
the ridges in Fig 3.1), with a lateral periodicity of 6.3 nm. At the domain
boundaries between the two different stacking regions, atoms are slightly
squeezed out and form two ridges that run with an equal distribution
along the 3 symmetry directions of the Au(111) surface.
Figure 3.1. (a) 177 × 177 nm2 constant current STM image of
the Au(111) “herringbone” reconstruction. (b) Model for the re-
construction of the Au(111) surface. The crosses denote the po-
sitions of atoms in the second layer. The open circles denote the
positions of atoms in the reconstructed top layer. C and A mark
the positions of fcc and hcp stacking, respectively. The straight
line indicates the displacement of atoms along the
√ [110] direction.
The dotted parallelogram indicates the 22 × 3 unit cell. (c)
Height profile is taken along the white dotted lines in (a). Panel
(b) adapted from [49].
This remarkable reconstruction of zigzag alternating line features

is commonly referred to as the Au(111) “herringbone”, reconstruction.
At the elbows of this reconstruction a single atomic point dislocation is
known to be present [50,51].
3.1.3 The electronic surface state

Besides its characteristic topographical features, the Au(111) surface
also exhibits intriguing electronic properties that are dominated by its
sp-derived Shockley-type surface state (section 1.3). Surface electrons
42 Experimental
Figure 3.2. (a) 34.0 × 34.0 nm2 constant current image of a

Au(111) surface after 7 s Ar+ sputtering at 2.4 keV and 6 × 10−7
mbar. (b) The corresponding LDOS map at +500 meV. Standing
wave interference patterns arise from electron scattering of surface
defects and impurities created by the Ar+ sputtering. (c) dI /dV
curves of Au(111) reveal a step like onset around −480 meV, the
signature of the surface state.
are scattered at step edges, surface defects and impurities and the re-
sulting interference of the incident and reflected electron waves give rise
to energy dependent periodic spatial oscillations of the electron density
of states at the surface. The scattering is nicely illustrated by the defect
in the Au(111) surface in Fig. 3.2 (a) (indicated with the black arrow)
and the corresponding LDOS map in Fig. 3.2 (b) taken at +500 meV.
Brighter features correspond to a higher LDOS, while darker features
correspond to a lower LDOS. The Au(111) surface state can also be ob-
served in local (dI /dV )(V ) spectra [Fig. 3.2 (c)]. A sudden increase in
tunneling conductance around −460 meV is the well known signature of
the Au(111) surface state (section 1.3).
3.2 Depositing preformed clusters in UHV 43
3.2 Depositing preformed clusters in UHV

The nanoparticles that are investigated in this work are preformed in
the gas phase and are deposited with low kinetic energy onto single-
crystalline flat surfaces. This preparation route offers a high degree of
freedom to tune the properties of nanostructured materials by select-
ing the appropriate particle size, composition, kinetic energy, and sub-
strate [52]. We refer to nanoparticles created in this manner as clusters.
Preformed clusters are grown and deposited under UHV conditions, al-
lowing for the creation of clean and well-defined samples. Section 3.2.1
gives a brief overview of the used cluster deposition apparatus (CDA),
whereas for a detailed description we refer the reader to [53].
3.2.1 Cluster deposition apparatus
Figure 3.3. Schematic overview of the cluster deposition appa-

ratus.
The CDA is schematically presented in Fig. 3.3. The setup consists

of four differentially pumped vacuum chambers, named hereafter the
source chamber (1 × 10−8 mbar), the extraction chamber (2 × 10−10
mbar), the deposition chamber (2 × 10−10 mbar), and the load/lock
(1 × 10−8 mbar). The load/lock (not included in Fig. 3.3) allows for
the loading of samples into the setup without breaking the vacuum in
the other chambers. The most important components of the setup are
discussed below.
44 Experimental
Cluster source
The cluster source used in this work has been described extensively
elsewhere [54], [53]. Figure 3.4 presents a schematically overview of this
source.
Figure 3.4. Schematical overview of the dual-target dual-laser

vaporization cluster source [53]. The main components are: (1)
pulsed gas valve, (2) ceramic placeholder, (3) formation chamber,
(4) laser channels, (5) glass plate, (6) nozzle, (7) target holder,
(8) laser beams, and (X) and (Z) targets.
We vaporize material from a target using focused radiation of a

pulsed laser. The vaporized material expands into the formation chan-
nel of the source. This formation channel is filled with high purity He
(99.9999 %) gas (9 bar), inserted typically about 700 µs before the light
of the vaporizing lasers hits the target. The atoms of the “hot” evap-
orated material are mixed into the “cold” He, whereupon condensation
occurs and cluster formation starts. The mixture of gas, atoms, and
clusters expands into the vacuum through a conical nozzle after which
cluster growth stops because of the rapidly decreasing density. Char-
acteristic for laser-vaporization cluster sources is that the metal vapor
is quenched, producing a beam of clusters with a temperature below
300 K. The speed of the clusters depends mostly on the gas-dynamics
of the He carrier gas and is about 73 × 10 m/s when exiting the source.
The constant speed implies that the inherent kinetic energy of the clus-
ters depends on their mass. Since we bring He into the setup dur-
3.2 Depositing preformed clusters in UHV 45
ing formation of the clusters, the pressure will rise during operation.
Typical values for the pressure during operation are: source chamber
(4 × 10−5 mbar), extraction chamber (6 × 10−7 mbar), and deposition
chamber (8 × 10−8 mbar).
Time of flight mass spectrometer

In the deposition chamber the cluster size distribution is measured by
a time-of-flight mass spectrometer (TOF-MS) before the substrate is
inserted into the beam and clusters are deposited. The process of opti-
mizing the free cluster production crucially depends on various experi-
mental parameters, such as a proper choice of the time interval between
the gas pulse and the firing of the laser. TOF mass spectrometry is
based on the principle that the mass of charged particles that acquired
the same kinetic energy can be determined from the travel time they
need to reach the detector. The charged particles are accelerated by
the electrodes of the extraction optics in the extraction chamber. After
passing a potential difference V , the velocity v of particles with a mass
m and charge q is the following:
r
2qV
v= . (3.1)
m
As V is the same for all particles, particles with the same charge to mass
ratio hence have the same velocity. If the particles travel a length l, after
being accelerated with voltage V before being detected, their mass, and
hence the amount of constituting atoms can be inferred from the time
of arrival tf at the detector [Fig. 3.5 (a)].
Once the mass-distribution is known, it can be converted to a size-
distribution. For this conversion we assume that the clusters are spher-
ical with a diameter d given by
√
d = 2r 3 n , (3.2)
with n the number of atoms in the cluster and r the Wigner-Seitz-radius

of the atoms of the target material [Fig. 3.5 (b)].
A typical mass spectrum of free cationic Ti clusters, used in chap-
ter 4, as a function of (a) the number of atoms in the cluster and (b)
the corresponding cluster diameter is presented in Fig. 3.5. The diam-
eter was calculated assuming the clusters to be spherical and using the
Wigner-Seitz radius for Ti (0.17 nm, see [17]). The maximum in the
46 Experimental
Figure 3.5. Abundance spectra of free cationic Ti clusters as a

function of (a) the number of atoms in the cluster and (b) the
corresponding cluster diameter. The mean cluster size was tuned
to be around 750 atoms (≈ 3.1 nm cluster diameter). The cluster
diameter was calculated assuming a spherical shape.
cluster size distribution was tuned to around 750 atoms (≈ 3.1 nm clus-
ter diameter). Ti clusters can be made with a very low oxygen content
(a few atoms per cluster), as observed in experiments on free Ti clus-
ters [55,56]. This ensures that the bond between the Au(111) substrate
and the Ti cluster will be formed by the pure metals.
Deposition chamber
Samples are introduced via a separate load-lock chamber that can be
pumped/vented separately from the deposition chamber. Samples are
mounted on the sample deposition stage for deposition. This stage can
be raised when recording mass spectra, and is subsequently lowered for
sample deposition (Fig. 3.3). The sample deposition stage is equipped
with a resistive heater to anneal samples in situ (Tmax ≈ 600 K). The
chamber is further equipped with a Knudsen-Cell, for evaporation of
additional materials. This Knudsen-Cell is used for the evaporation of
3.3 Growth of thin NaCl layers 47
the NaCl, which will be discussed in detail in section 3.3. There is

a quartz microbalance installed to monitor the cluster-film thickness.
Deposition times were chosen to achieve a low coverage of clusters on
the substrate, i.e. well below one monolayer, to allow for individual
characterization.
3.2.2 Sample transport

To avoid sample contamination and oxidation, a home-built UHV trans-
port vessel was developed and made compatible with both the cluster
deposition apparatus and the STM setup [36]. This allows us to deposit
the clusters onto substrates that are cleaned in the preparation chamber
of the STM setup and to transport the samples under UHV conditions.
The vessel consists of a small chamber provided with a long transport
arm and an ion pump that can run autonomously for 7 hours using a
battery. Substrates are transported to the cluster deposition apparatus
and back with the home-built UHV vessel with a base pressure typically
in the 1 × 10−10 mbar range.
3.3 Growth of thin NaCl layers

The study of the intrinsic properties of individual 0D systems with scan-
ning tunneling spectroscopy (STS) is limited by the difficulty to create
well characterized and atomically flat insulating layers, which decouple
the 0D system from the substrate (section 1.4). Alumina, magnesium
oxide, silica, and sodium chloride (NaCl) are some of the proposed ma-
terials to support and electronically isolate clusters. Among these, crys-
talline NaCl is most promising because it can be grown as atomically flat
layers on various substrates. In addition, when it is sublimated it main-
tains its stoichiometry, which makes it extremely suitable as support for
the characterization of individual clusters.
In this work we grow thin NaCl films on Au(111) by thermal evap-
oration (at 800 K) of the NaCl (99.999 % purity, Alfa Aesar) from the
Knudsen-Cell installed in the deposition chamber of the CDA (Fig. 3.3).
The Au(111) substrate is kept at room temperature during evaporation.
The evaporation rate (typically 0.1 nm/min) is controlled by tempera-
ture stabilization and in situ calibrated with a quartz crystal microbal-
ance. The properties of the thin NaCl grown in this manner are discussed
in chapters 5 and 6.
48 Experimental
3.4 Creation of F-centers and deposition of Co

atoms
We did not limit ourselves to 0D systems formed by clusters, but looked
for a broader range of atomic scale features. Naturally, as an extension
of the clusters we investigated the smallest possible cluster, the single
atom (Co in our case). However, as it turns out, one can not only obtain
0D systems by adding extra atoms, but also by removing single atoms.
A single Cl vacancy in NaCl is an atomic-scale defect in an otherwise
defect-free environment, and introduces a single occupied electron state
in the bulk band gap. These defects, intrinsic in bulk NaCl crystals,
are fairly simple, well defined, and stable. They have been studied both
experimentally and theoretically in detail and are commonly referred
to as color or F-centers [10,57]. For examples we refer the reader to
chapter 7, while the procedure for Co deposition and F-center creation
is discussed below.
Co atoms
Single Co atoms where evaporated onto our substrates in the preparation
chamber of the STM, using the evaporators available there (section 2.3).
Evaporations were performed using deposition rates in the range 0.01
- 0.05 nm/min with atoms evaporating from a high purity Co rod (2.0
mm diameter, 99.995 % purity).
F-centers
Cl vacancies were formed with the STM tip using electron-stimulated
desorption experiments, following the procedure stated in [58]. If we
retract the tip by more than 1 nm from the surface and apply a large
(> 7 V) sample voltage we can desorb single Cl ions from the surface.
Chapter 4
Morphology and electronic

properties of thermally
stable TiOx nanoclusters on
Au(111)
The results presented in this chapter are based on:
Morphology and electronic properties of thermally stable

TiOx nanoclusters on Au(111)
Koen Lauwaet, Koen Schouteden, Ewald Janssens,
Chris Van Haesendonck, and Peter Lievens
Phys. Rev. B 83, 155433 (2011).
Transition metal particles, and in particular TiO2 clusters are of

high fundamental and technological interest due to their optical, catalyt-
ical and photocatalytical properties. The particles are being applied in
various fields, including sunscreen agents [59], removal of various types of
aqueous pollutants [60] and antibacterial and detoxification effects [61].
Information about the electronic structure of free Ti clusters has been
obtained by performing core level spectroscopy [55], and by combining
photoelectron spectroscopy with theoretical calculations [62]. Deposi-
tion of preformed TiOx clusters has been performed in order to study
the properties of thick cluster assembled films [56,63,64]. Both the final
morphology and the optical gap of the films were shown to depend on
the initial cluster size [64,65].
49
50 TiOx nanoclusters on Au(111)
In the case of individual particles on a surface the morphology and

electronic properties of a variety of nanoparticles have been investigated
in detail in a controlled environment [66–68]. However, attempts to ex-
plore individual nanosized TiO2 (titania) particles are scarce. Recently,
TiOx nanoparticles with x close to 2 have been grown by atomic evap-
oration of Ti onto thin layers of H2 O and NO2 adsorbed on Au(111).
The buffer layer subsequently was evaporated, resulting in the forma-
tion of various self-assembled TiOx nanoparticles on the Au(111) sur-
face [69,70]. The morphology of these particles strongly depends on the
buffer layer and on the annealing temperature used in the growth pro-
cess, and their crystal structure is typically that of rutile titania [70].
TiOx particles are known for their high catalytic activity [71,72], and ex-
hibit a highly temperature dependent morphology [69]. Although clean
Au(111) is not catalytically active for the water gas shift (WGS), gold
surfaces that are 20 to 30% covered by titania nanoparticles have a
catalytic activity comparable to and even higher than that of a good
WGS catalyst such as Cu(100) [71]. By interaction of H2 O and CO
with TiOx /Au(111), the water dissociates on oxygen vacancies of the
titania and CO adsorbs on gold sites located nearby. Subsequent re-
actions steps, leading to the formation of CO2 and H2 , take place at
the metal-oxide interface. The interaction between TiOx and Au(111)
is supposed to play an important role in the catalytic activity because
of cooperative effects at the oxide-metal interface [71]. For Au parti-
cles grown on TiO2 it is known that there occurs a charge transfer from
the defect sites of TiO2 to the gold nanoparticle [73]. In the opposite
case of TiOx nanoparticles grown on Au(111), the presence of a simi-
lar charge transfer and its relevance for heterogeneous catalysis are still
open questions.
In the present study we report on the investigation of individual

TiOx nanoparticles on clean Au(111) created by deposition of preformed
pure Ti clusters on a Au(111) surface (section 3.2.1), which are subse-
quently oxidized. The Ti clusters are preformed in the gas phase and
are deposited with low kinetic energy onto single-crystalline flat surfaces.
This alternative preparation route offers a high degree of freedom to tune
the properties of nanostructured materials by selecting the appropriate
particle size, composition, kinetic energy, and substrate [52]. Preformed
clusters are grown and deposited under ultra high vacuum (UHV) con-
ditions, allowing for the creation of clean and well-defined samples and
ensuring to obtain particles that are immobilized on the Au(111) sur-
4.1 TiOx clusters on Au(111) without annealing 51
face by strong Ti-Au bonds, since very little oxygen is present during
the deposition of the particles.
The alternative preparation route enabled us to investigate the mor-
phology as well as electronic properties of nanometer sized clusters con-
sisting of a few up to several hundreds of atoms by means of scanning
tunneling microscopy (STM) and scanning tunneling spectroscopy (STS)
under controlled conditions. STM combined with STS offers the ideal
means to study in detail, down to the atomic scale, both the geomet-
rical and electronic properties of individual nanoparticles on metallic
substrates. The influence of annealing on the cluster morphology was
studied up to high temperature (970 K). With STS we probed the lo-
cal electronic properties of the clusters as well as their influence on the
surrounding Au(111). The possible relevance of our findings for the un-
derstanding of the catalytic performance of the TiOx /Au(111) system
is discussed.
4.1 TiOx clusters on Au(111) without annealing

The Ti clusters were oxidized before measuring with STM. Oxidation
of the deposited clusters is achieved by lowering the vacuum conditions
in the transport vessel. For this purpose we switched from ion pumping
to pumping with a turbomolecular pump. This way, the vacuum is
kept in the 10−7 mbar range for a few minutes. When analysing the
gasses present in the vacuum chamber with a mass spectrometer, the
highest partial pressures were found for O2 (8 · 10−8 mbar) and H2 O
(4 · 10−7 mbar), with only a low background pressure of hydrocarbons
(10−9 mbar range). Under these conditions the Ti clusters interact with
both O2 and H2 O to form titanium oxide [69]. We can, however, not
fully exclude that a small amount of hydroxyl groups is adsorbed on the
Ti surface as well [69]. It is known that oxide can penetrate up to 5 nm
into the surface of Ti films, implying that the here investigated clusters
may be (nearly) completely oxidized [74]. From previous experiments
using reactive sputtering of Ti with an oxygen-rich plasma, it is known
that instead of stoichiometric TiO2 an oxygen deficient TiOx phase with
x close to 2 is formed [75].
Figure 4.1 (a) presents a typical STM image of TiOx clusters on
an atomically flat Au(111) surface before annealing. Particles of var-
ious lateral sizes and with different heights are retrieved on the sur-
face. The particles are randomly distributed across the surface. Some
Figure 4.1. (a) 100 × 100 nm2 STM image of deposited TiOx
clusters on a Au(111) surface (V = 1.0 V, I = 1.0 nA). The clus-
ters show negligible mobility after deposition. (b) Height profile
taken along the dashed white line in (a) comprising three TiOx
clusters. (c) Normalized height histogram of deposited TiOx clus-
ters and complementary abundance spectra of free cationic Ti
clusters as a function of cluster diameter in the spherical approx-
imation (gray) and as a function of cluster radius in the hemi-
spherical approximation (black).
contamination was observed on the Au(111) substrate after Ti cluster

deposition and oxidation (less than one monolayer). This contamina-
tion leads to a roughening of the substrate as seen in Fig. 4.1 (a) and
Fig. 4.1 (b). Locally the atomically flat Au(111) surface can still be
discerned (Fig. 4.1 (a) on the left hand side). Reaction of TiOx with
excess oxygen leads to oxygen spill over from titanium nanoparticles to
the gold surface, possibly explaining the observed contamination on the
Au(111) surface [76].
The amount of TiOx nanoparticles retrieved on the surface is around
20 per 100 nm2 and agrees well with the amount of deposited clusters
estimated from measurements with the quartz micro balance (24 clusters
per 100 nm2 ). The TiOx nanoparticles are distributed randomly across
the Au(111) surface and do not show any preferential positioning at, e.g.,
Au(111) step edges, nor show they any clear alignment at elbows of the
Au(111) reconstruction as was the case for self-organized Co islands in
a previous study [68]. These findings imply that the TiOx nanoparticles
4.2 Influence of annealing to elevated temperatures 53
do not exhibit significant diffusion on Au(111) at room temperature.

The height distribution of the particles after deposition is presented
in Fig. 4.1 (c). Height histograms presented in this chapter are based on
measured particles heights of around a hundred particles. When mea-
suring with STM, the particle height should equal the diameter of the
free cluster provided it remains spherical upon deposition. In case the
clusters flatten out and become hemispherical, the height of the particle
should equal the radius of a hemisphere having the same volume of the
free cluster. Figure 4.1 (c) includes an abundance spectrum of the free
cationic TiOx clusters in the gas phase, both as a function of cluster
diameter in the spherical approximation (ds , gray) and as a function
of cluster radius in the hemispherical approximation (rh , black). They
relate to each other as rh = 2−2/3 ds . Comparison of the abundance
spectra with the height distribution of the TiOx clusters clearly reveals
that the deposited clusters can be considered as individual hemispheri-
cal particles after deposition. This result is similar to what is observed
for deposited Co clusters on Au(111) [68]. The observed cluster height
is used to determine the flattening of the particles. The observed broad-
ening cannot be used, since with STM the width of the particles can
not be accurately measured due to the finite size of the tip [68]. Be-
low we demonstrate that there occurs a strong interaction between the
TiOx clusters and the Au(111) surface, which can account for this flat-
tening. In contrast to Co clusters on Au(111) [68], the height histogram
of the TiOx nanoparticles on Au(111) does not reveal preferential clus-
ter heights. This conclusion should be considered with care because of
the uncertainties in the height determination that are caused by the
contaminants. The contaminants can be moved by the tip while scan-
ning, leading to an increased uncertainty on the height determination of
around 0.2 nm. Detailed STS measurements are hampered as well by
this layer.
4.2 Influence of annealing to elevated tempera-

tures
In a next step we have performed a systematic investigation of the influ-
ence of annealing the sample. Figure 4.2 (a) presents a topographic STM
image after annealing of the TiOx nanoparticles up to a temperature of
970 K. Comparison with Fig. 4.1 (a) shows that the substrate is now
atomically flat and the Au(111) herringbone reconstruction is observed,
Figure 4.2. (a) 100 × 100 nm2 STM image of TiOx nanoparticles
on a Au(111) surface after annealing to 970 K (V = -1.0 V, I = 0.1
nA). (b) Line profile taken along the dashed white line indicated
in (a). (c) Normalized height histogram of the TiOx nanoparticles
after annealing to 670 K (dashed) and after annealing to 970 K
(shaded).
indicating a clean surface. The contamination layer is evaporated, which

happened already after annealing to 570 K (data not shown). The re-
moval of this layer also eliminates the uncertainty for the height mea-
surement. Still no preferential cluster heights are observed when looking
at an ensemble of a lot of different clusters.
Analysis of the STM images after annealing to 670 K and 970 K
does not reveal a significant difference in the distribution of particle
heights (Fig. 4.2 (c)) when compared to the substrate before annealing
(Fig. 4.1 (c)). This indicates that TiOx nanoparticles do not “sink”
into the Au(111) surface as is the case for, e.g., Co nanoparticles [77].
Meanwhile also the density of particles on the surface remains constant
(around 20 per 100 nm2 ). After annealing the particles are still randomly
distributed across the surface and they do not coalesce. From these
results it can be concluded that even at high temperatures the TiOx
nanoparticles do not diffuse and remain on the surface as single entities.
The maximum applied annealing temperature (970 K) is higher than the
temperature used for annealing of the gold film after Ar bombardment
(720 K). At 720 K Au atoms are able to diffuse on a clean Au(111)
4.2 Influence of annealing to elevated temperatures 55
surface to a degree that even nanosized islands and vacancies are able
to flatten out [78]. However, after depositing the TiOx nanoparticles on
the Au(111) no additional changes of the Au(111) surface are observed
up to 970 K.
A similar annealing procedure was followed by Osgood et al. in or-
der to investigate the evolution of TiOx nanoparticles created by atom
deposition after H2 O adsorption on Au(111) [69]. The TiOx nanopar-
ticles in the cited study showed a clear evolution, i.e., coagulation of
particles, leading to an increased particle size after annealing to 300 K,
500 K, and 700 K. Our annealing temperature of 970 K is higher than
700 K, but nevertheless we do not observe any significant changes in
substrate morphology. TiOx nanoparticles grown by deposition of Ti
clusters on Au(111), which are subsequently oxidized, therefore have an
enhanced stability when compared to their counterparts grown by atom
deposition on H2 O.
One could question if the thermal stability can be explained by
bonding of the clusters to defects in the surface. Previous experiments
on Ag clusters deposited on graphite demonstrated that mobile clusters
can be captured and stabilized by surface defects [79]. The Au(111)
surface contains intrinsic defects at the elbows of the herringbone recon-
struction [51]. However, in Fig. 4.2 (a) one can spot clusters that are
not located at elbow sites of the Au(111) surface reconstruction after
annealing (e.g., the upper cluster on the dashed white line). Stability of
these clusters can therefore not be attributed to intrinsic surface defects
present under the clusters.
Another possibility is that the clusters create their own defects
in the gold surface upon landing. When cluster deposition is done
with low kinetic energy, we do not expect that defects are created
upon deposition [80]. In the case of deposition of, e.g., Ag clusters
on graphite the required energy for “self pinning” due to defect cre-
ation is 10.4 eV/atom [81]. This value is significantly higher than the
0.13 eV/atom used in our experiments and hence no surface damage is
expected. Therefore, surface defects cannot explain the stability of our
clusters.
The low mobility of the clusters on the Au(111) surface can be
explained by the growth process of the TiOx nanoparticles. The clusters
made in the gas phase consist of metallic Ti, ensuring upon landing
available atoms to form Ti-Au bonds. The bond between Au and Ti is
known to be strong, i.e. 2.5 eV for the diatomic species [82], compared
with 2.3 eV for diatomic gold [82], and 1.5 eV for diatomic Ti [83].
Once oxidized the oxygen will prevent the Ti from diffusing into the Au
substrate during the annealing process, as would happen in the case of
pure Ti nanoparticles [84,85].
Next, we zoomed in on individual clusters. Two typical results are
presented in Fig. 4.3 (a) and Fig. 4.3 (d). For reasons of clarity the
“derivative image” of the recorded topography image is presented [47].
One can discern that there are nanoparticles that exhibit a hexagonal
shape on the gold. The nanoparticles seem to have a flat top and clearly
defined steps (step height: 0.23 nm ± 0.03 nm). Our observations are
consistent with previous experiments on TiO2 nanoparticles formed on
Au(111) by different methods, from which it was concluded that the
hexagonal TiO2 particles have a rutile crystal structure with their (100)
plane oriented parallel to the Au(111) surface [70,76]. Similar hexagonal
formations have been observed when TiO2 nanoparticles are formed on
Au(111) by different methods [70,76]. Figure 4.3 (d) reveals the presence
of “dots” on the cluster surface. Adhesion of molecular oxygen on the
titanium surface is likely to cause such “dots” since similar features
where observed after the oxidation of a titanium film under controlled
conditions [86].
4.3 Electronic properties of TiOx nanoparticles

on Au(111)
After annealing to 570 K the layer of surface contamination is evapo-
rated and stable spectroscopy can be performed. Figure 4.3 (b) is a
LDOS map taken at -600 mV on the cluster presented in Fig. 4.3 (a).
The step edges of the atomic layers of the nanoparticle stand out clearly
in the LDOS map. These edges can also be observed in the LDOS map
taken at +600 mV (Fig 4.3 (c)). On clusters where the “dot”-like fea-
tures are present (see Fig. 4.3 (d)), the LDOS map is strongly affected
by these “dots” (see Fig. 4.3 (e) and Fig. 4.3 (f)), indicating that the
oxygen molecules locally disturb the electron density of the cluster. It
is not surprising that we do not observe two-dimensional standing wave
patterns on the uppermost facet of the cluster. The impurities will
induce additional scattering of the electrons inside the cluster and inter-
ference patterns purely defined by topography will be washed out. Part
of the electron density in the TiOx clusters is, however, still related to
their topography since step edges stand out in the LDOS maps. Previ-
4.3 Electronic properties of TiOx nanoparticles on Au(111) 57
Figure 4.3. (a) and (d) 12 × 12 nm2 and 10 × 10 nm2 derivatives

of the STM images of 2 TiOx clusters on Au(111) after annealing
to 700 K (V = +600 mV, I = 0.3 nA). (b) and (e) corresponding
LDOS maps taken at -600 meV. (c) and (f) corresponding LDOS
maps taken at +600 meV. (c) reveals the formation of standing
wave patterns at the Au(111) surface due to scattering of Au
surface state electrons at the TiOx cluster.
ous observations of the LDOS of deposited Co clusters also revealed a

topological dependence [68].
Upon careful inspection of the LDOS map of Fig. 4.3 (c) it can
be seen that interference patterns are present on the Au(111) surface
surrounding the TiOx cluster, which can be related to scattering of Au
surface state electrons. As already discussed above, the clusters do not
damage the Au(111) surface upon deposition, implying that surface de-
fects cannot explain the observed scattering of Au(111) surface electrons.
We therefore conclude that the clusters act as a strong scatterer for Au
surface state electrons, indicative of a strong interaction between the
cluster and the Au(111) surface.
Typical I (V ) and corresponding dI /dV (V ) spectra taken on a TiOx
nanoparticle and on the surrounding Au(111) are presented in Fig. 4.4
(a) and (b), respectively. The nanoparticle exhibits a rectifying I (V ) be-
Figure 4.4. (a) I (V ) and (b) corresponding dI /dV (V ) spec-

tra taken at the center of a TiOx nanoparticle (dashed curves)
and at the surrounding Au(111) surface (solid curves). Inset:
35 × 35 nm2 derivative of an STM image of the TiOx cluster
after annealing to 670 K (I = 0.4 nA, V = 1.0 V).
havior: the tunneling current is much higher for positive than for nega-
tive bias, where the current is nearly completely suppressed. This asym-
metric I (V ) behavior corresponds to that of a rectifying element in paral-
lel with a leak resistor and can be described by I(V ) = a [exp(αV ) − 1]+
bV , where a is the current of the system for reverse bias, α denotes the
ideality of the diode and b corresponds to the inverse leak resistance of
the system [87]. The latter can be related to an incomplete oxidation of
the TiOx cluster (x < 2). By fitting the equation to the experimental
data (see Fig 4.4 (a)), we find that α is much larger than 1, which is
typical for nanoscale diodes [88,89].
It is known that non-stoichiometric TiOx behaves as a n-doped semi-
conductor (the Fermi level is located near the conduction band). [90]
Charge transfer hence takes place between the semiconducting TiOx
nanoparticle and the metallic Au(111), leading to the formation of a
nanosized Schottky diode. Such charge transfer was previously reported
for the opposite case of Au nanoparticles on TiO2 substrates [91]. It
must be noted that the amount of charge transfer strongly depends on
the oxidation state of the nanoparticle [92].
The formation of a Schottky diode indicates that oxygen also is
present inside the nanoparticle and that the presence of oxygen is not
limited to the dot-like features that are observed on the surface of the
clusters (see Fig. 4.3 (d)). As already indicated above, it was shown that
oxide can penetrate up to 5 nm into the surface of Ti films, implying that
the here investigated clusters are (nearly) completely oxidized (x close
to 2) [74]. In addition, it is known from previous experiments that an
oxygen content x of at least 1.35 is needed for TiOx to become n-type
semiconducting [75]. Nevertheless, formation of oxygen vacancies at the
surface of the nanoparticles during annealing to elevated temperatures
can be expected. This can be interesting in view of technological appli-
cations, since oxygen vacancies are essential for the catalytic behavior
of the TiOx /Au(111) system, e.g., for the WGS [71]. However, it was
shown by Biener et al. that, for the case of TiOx nanoparticles grown
on Au(111) by Ti atom deposition, non-oxidized Ti diffuses into the
Au(111) surface at elevated temperatures, while TiO2 remains at the
Au(111) surface [76]. Similar results have been reported more recently
by Potapenko et al. [85]. Within the resolution of our measurements we
do not observe any significant change of the cluster dimensions in our
STM images as well as in our cluster height histograms after annealing.
We therefore conclude that the oxidation state of the TiOx clusters is
Figure 4.5. (a) dI /dV spectra at the center of a TiOx cluster

(dashed) and at the surrounding Au(111) surface (solid). Inset:
21 × 21 nm2 derivative of a STM image of this TiOx cluster that
has been annealed at 670 K. (b) 2D visualization of the dI /dV
spectra along the white line in (a) reveals the spatial dependence
of the LDOS inside the cluster and the bending of the Au(111)
surface state near the edge of the cluster (I = 1.0 nA, V = 1.0 V).
close to 2 and at least higher than 1.35.

Figure 4.5 (b) presents a 2D visualisation of dI /dV (V ) spectra
recorded along the white line crossing the cluster presented in the inset
of Fig. 4.5 (a). A so-called “tip state” occurs at 0.1 V in the dI /dV
spectrum of the Au(111) surface presented in Fig. 4.5 (a). Since this
maximum does not occur for other STM tips, it can be attributed to the
presence of an electronic state localized at the apex of the STM tip. The
cluster exhibits a clear gap in the DOS (dark region in the lower part
of Fig. 4.5 (b)) that varies across the cluster surface and is asymmetric
around zero bias voltage.
When crossing from the cluster to the gold, atomically sharp tran-
sitions in the DOS are observed (at distances of 8 nm and 17 nm in
Fig. 4.5 (b)). These abrupt transitions can be attributed to intrinsic
tip switching. Whereas the atomically flat Au(111) surface as well as
the top of the TiOx clusters are probed by the outermost tip asperity,
the cluster edges are probed by different asperities at both sides of the
tip [68]. The “tip state” around 0.1 V is related to the outermost tip
asperity and is therefore observed on the whole Au(111) surface in the
dI /dV spectra in Fig. 4.5 (b). On the Au(111), near the edges of the
cluster, one can observe an influence of the nanoparticle on the Au(111)
surface state: the onset of the surface state shifts toward higher energies
(by about 80 meV) in the vicinity of the nanoparticle. The tip related
maximum at 0.1 eV remains constant as expected. The shift of the sur-
face state at -480 meV can be attributed to a local electron transfer from
the TiOx nanoparticle to the Au(111) surface, resulting in a shift of the
onset energy of the surface electrons. This result further supports the
idea of the formation of a nanosized Schottky diode.
The modification of the electronic structure of the substrate close to

deposited particles is a known phenomenon [91], and is of high impor-
tance for the catalytical activity of a system. On a metal that exhibits
a surface state, stable molecules such as CO experience a strong Pauli
repulsion due to the electrons stemming from the surface state [29]. The
observed repression of the surface state around the nanoparticle leads
to a smaller Pauli repulsion and can significantly lower the activation
energy for dissociation of adsorbates [29]. The observed shift of the
Au(111) surface state near the edges of the cluster is thus relevant for
the high catalytic activity of the TiOx /Au(111) system. In particular
the high performance of the WGS reaction on TiOx /Au(111) catalysts
has been shown to heavily rely on the direct participation of the metal-
oxide interface [71]. The adsorption and dissociation of water takes place
on the oxide, whereas the CO adsorbs on sites of the gold substrate lo-
cated in the vicinity of the cluster. All subsequent steps to complete
the catalytic cycle occur at the oxide-metal interface. In this context,
the observed bending of the surface state near the cluster edges induces
a preferential binding of the CO on the Au(111) surface in the clus-
ter vicinity [29,93]. Next, the local charge transferred to the Au(111)
surface from the TiOx nanoparticle can be partially transferred to the
CO molecules. Such charge transfer weakens the C-O bond [94], thereby
leading to reduced potential barriers for the reaction of the CO molecules
with atomic oxygen [94] stemming from the aforementioned dissociation
of H2 O on the TiOx surface. The cooperative effect between the TiOx
and the Au(111) surface can thus enhance the catalytic performance of
the TiOx /Au(111) system.
4.4 Summary
The morphology and electronic properties of preformed Ti clusters de-
posited and subsequently oxidized on clean Au(111) under controlled
UHV conditions were investigated before and after annealing up to high
temperatures by means of STM and STS. By systematic analysis of the
STM data we demonstrated that the TiOx clusters experience only a lim-
ited flattening resulting from the deposition and that their shapes can
be approximated by hemispheres. A negligible mobility of the clusters
on the Au(111) surface was observed upon annealing to temperatures as
high as 970 K. After annealing some of the nanoparticles are observed
to exhibit hexagonal facets. STS measurements reveal that the clusters
exhibit a rectifying behavior, which is attributed to the formation of
a Schottky junction between the n-type semiconducting TiOx clusters
and the Au(111) surface. A charge transfer to the Au(111) substrate in
the vicinity of the cluster was observed. This is relevant for the high
catalytic activity of this system, since the charge transfer locally creates
a preferential binding site for reagents on the Au(111) surface in the
vicinity of the cluster and facilitates the activation of adsorbates. Both
issues are essential for catalytic cycles where reaction steps take place
at the metal-oxide interface.
Chapter 5
Tuning the NaCl/Au(111)

interface state by the NaCl
layer thickness
Except for section 5.3, the results presented in this chapter are based
on:
Tuning the NaCl/Au(111) interface state by the NaCl layer
thickness
To be submitted
Ever since the first observations of standing waves formed by the
two-dimensional electron gas of the Cu(111) surface by means of STM
[95], the 2-DEG has proven to be an ideal playground for the investiga-
tion of quantum-mechanical phenomena. Engineering of electron wave
functions in reduced dimensionality allowed researchers to explore and
visualize fundamental aspects of quantum mechanics, thereby giving rise
to new ideas for advanced materials and devices [96]. The local density
of electronic states at the surface strongly influences chemical reactiv-
ity. The electronic properties of surfaces also determine the growth
of nanostructures such as dimers, chains and superlattices of atoms or
noble metal islands [29]. Controlling these properties on length scales
shorter than the diffusion lengths of the electrons and spins (some tens
of nm for metals) is a major goal in nanoelectronics and nanospintronics,
respectively.
63
64 Tuning the NaCl/Au(111) interface state
An important class of systems where control of the electronic wave

functions is of the utmost importance is that of heterointerfaces. Among
these metal-insulator interfaces are especially intriguing because of their
fascinating properties such as metal-insulator transition, band gap nar-
rowing, and superconductivity [97]. They are also technologically im-
portant, as they are used for catalysis and magnetic tunneling junc-
tions [98]. Until now, however, the local electronic wave functions at
metal-insulator interfaces have not been studied in detail, which may
be related to the fact that well-defined interfaces are difficult to prepare
due to the different natures of the chemical bonds.
Crystalline NaCl is a very promising material in this respect, be-
cause it can be grown stoichiometrically as atomically flat layers on
metal surfaces. NaCl is therefore extremely suitable to study local elec-
tronic variations of the metal support. NaCl islands have been success-
fully grown on numerous crystalline surfaces, including Cu(111) [99],
Cu(100) [10], Cu(311) [58], Ag(100) [100], Ge(100) [8], Al(111) [101],
and Au(111) [102]. For NaCl on Cu(111) Repp et al. found that the
Cu(111) surface state (SS) survives as an interface state (IS) at the
NaCl/Cu(111) interface [99]. The IS is shifted to higher energies and
the effective mass is increased when compared to the Cu(111) SS. In
this NaCl/Cu(111) system the dispersion behavior of the IS was shown
not to depend on the NaCl thickness when changing from bilayered to
trilayered NaCl.
In this chapter we report on the morphology and the electronic
properties of thin NaCl films grown on atomically flat Au(111) by ther-
mal evaporation. The evaporation of NaCl on Au(111) at room tempera-
ture leads to the growth of large polycrystalline islands of bilayer NaCl,
which can be converted into trilayer NaCl by post annealing. We mea-
sure the NaCl/Au(111) IS dispersion in a broad energy range at low
temperature and find that it varies with the number of NaCl layers. Fur-
thermore, the dispersion behavior of the contribution of bulk state (BS)
electrons to the LDOS at the NaCl/Au(111) interface is investigated
and compared with results obtained previously for the bare Au(111)
surface [103].
5.1 Morphologic Characterization 65
5.1 Morphologic characterization of thermally

grown NaCl on Au(111)
5.1.1 As deposited
Figure 5.1. (a) 150 × 150 nm2 pseudo 3D STM image of NaCl
islands on Au(111) (V = 1.5 V, I = 0.05 nA). The characteris-
tic carpet-like growth of islands over the Au(111) step edges can
be observed, as well as the straight non-polar edges of the NaCl
nanocrystals. (b) 10 × 10 nm2 close-up view of a NaCl edge.
On the left hand side the atomic corrugation of the NaCl can be
observed. (c) Height profile taken across the dashed white line
indicated in (a), together with a schematic representation of the
different NaCl layers and the Au(111) substrate. Note that the
apparent height of the NaCl layers decreases with increasing NaCl
thickness as discussed in the text.
Deposition of thermally evaporated NaCl onto clean and atomically flat

Au(111) surfaces at room temperature leads to the growth of large bi-
layer (height = 2.9 Å ± 0.2 Å) crystalline islands of (100) terminated
NaCl. The NaCl islands preferentially nucleate at and grow continu-
ously over step edges of the Au(111) [see Figs. 5.1 (a) and 5.1 (b)].
NaCl islands typically exhibit a rectangular shape, having straight step
edges and kinks. Inside the NaCl islands vacancies may be present and
on top small patches of an additional third layer are observed. Rarely, a
small patch of a single NaCl layer is observed (height = 1.8 Å ± 0.2 Å)
at the edges of bilayer islands. Remarkably, the apparent height of single
layer NaCl is not equal to half the apparent height of bilayer NaCl, as
illustrated in Fig. 5.1 (c). This can be related to the change in tunnel
barrier height when measuring on different thicknesses of NaCl [101].
The NaCl islands do not modify the herringbone reconstruction of
the Au(111) surface, which can be observed as an additional corruga-
tion of the NaCl islands [Fig 5.1 (a) and (b)]. The measured herringbone
corrugation on the bare Au(111) substrate and on the bilayer NaCl crys-
tallites are very similar, i.e. around 20 pm. The atomic corrugation of
the NaCl (measured on a large island and far away from a step edge of
the Au(111) surface) has a periodicity of 0.50 nm ± 0.04 nm. This corre-
sponds to a NaCl lattice constant of 0.25 nm ± 0.02 nm, in good agree-
ment with the in-plane NaCl distance for a monolayer and a trilayer of
non-supported NaCl derived from ab initio simulations (0.27 nm) [101].
5.1.2 After annealing
Next, we annealed the sample to a temperature of 460 K. After anneal-

ing, the diameter of an average NaCl crystallites has increased threefold,
the crystallites are nearly defect free, and their height is typically three
monolayers (4.0 Å ± 0.2 Å, see Fig. 5.2). The number of kinks in
the edges of the islands is reduced and vacancies within the islands are
filled up, reducing the number of low-coordinated atoms. The measured
herringbone and atomic corrugation of the NaCl are identical to that
before annealing, and the NaCl islands still decorate the Au(111) steps
in a carpet-like fashion.
5.2 Dispersion relations 67
Figure 5.2. (a) 150 × 150 nm2 pseudo 3D STM image of a NaCl
island on Au(111) after annealing (V = 1.0 V, I = 0.1 nA). (b)
7.5 × 7.5 nm2 close-up view of a NaCl edge after annealing. On
the left hand side the atomic corrugation of the NaCl can be
observed. (c) Height profile taken across the dashed white line
indicated in (a), together with a schematic representation of the
different NaCl layers and the Au(111) substrate. Yellow and red
indicate protrusions (a) or a high LDOS [(b) and (d)], and blue
and black indicate lower lying regions (a) or a low density of states
[(b) and (d)].
5.2 Dispersion relation of interface state and

bulk state electrons
5.2.1 Bilayer NaCl/Au(111)
Interference patterns, reflected by circular isotropic patterns, in the local
density of states (LDOS) resulting from scattering of electrons at the
Figure 5.3. (a) 66 × 66 nm2 STM image of a NaCl island on

Au(111) (V = +0.6 V, I = 0.3 nA) and corresponding LDOS
images at (b) V = -0.2 V, I = 0.1 nA, and (d) V = +0.2 V,
I = 0.1 nA, revealing the voltage dependence of the NaCl/Au(111)
IS. (c) is a FT-image of the area enclosed by the black square in
(d). The Fermi contours of both IS and BS are indicated by the
white arrows.
NaCl(2 ML)/Au(111) interface are nicely illustrated for the NaCl island
in Fig. 5.3 (a) on the corresponding LDOS map [Fig. 5.3 (d)]. Brighter
features correspond to a higher LDOS, while darker features correspond
to a lower LDOS. At -200 mV, however, the LDOS map shows for the
NaCl/Au(111) interface a high intensity at the hcp regions and a low
intensity at the fcc regions of the underlying Au(111) [Fig. 5.3 (b)], which
will be discussed in section 5.2.3. At this energy interference patterns are
observed on the clean Au(111), stemming from the undisturbed SS. From
the LDOS maps we conclude that upon adsorption of a NaCl layer on
Au(111) the Au(111) SS survives and forms an IS, with a different energy
versus wave vector dispersion, at the metal/insulator interface [99].
Figure 5.4. Experimental dispersion behavior of both

IS (black squares) and BS (black triangles) electrons at
NaCl(2 ML)/Au(111). Both dispersion relations are fitted with
a parabolic relation for a free electron gas (black solid lines). At
higher energies, the measured dispersion for BS electrons below
NaCl suddenly diverges, also fitted with a black solid line. Data
for clean Au(111) SS (open squares) and BS (open triangles),
taken from [103], are added for comparison.
The energy versus wave vector dispersion of this IS band was in-
vestigated by mapping the LDOS with the STM for a broad range of
sample voltages. By taking the Fourier transform of the maps, a direct
image of the Fermi contour of the surface state is obtained [see Fig. 5.3
(c)] [104]. This technique is called Fourier transform (FT) STM [27,103].
By taking the radial average of the FT images, a value of the parallel
part k || of the wave vector ~k = ~k⊥ + ~k|| is obtained at each applied
bias voltage V for electrons with energy E(~k⊥ , ~k|| ) = E F +eV [104].
Although it is not a priori clear whether either the inner or outer ring
should be linked to the surface or bulk contribution, this becomes clear

by plotting the dispersion relation for the inner and the outer ring over
an extended energy range (Fig. 5.4). The obtained dispersion relation
of the interface state band at the NaCl(2 ML)/Au(111) interface (before
annealing) is presented in Fig 5.4. The wave vector k || could be resolved
in FT images of LDOS maps up to 1000 meV. The experimental disper-
sion relation E (k || ) can be fitted well by the parabolic relation expected
for a quasi-2D free-electron-like gas [Eq. (1.17)]. The obtained values
for E 0 and m* /m e are reported in Table 5.1. Compared with the SS of
the the bare Au(111) surface, a slight lowering of the effective mass was
observed combined with an upward shift of the onset energy.
By mapping the LDOS of the NaCl/Au(111) surface the dispersion
of the Au BS electrons near the surface can also be determined [103,104].
The influence of an insulating layer on the dispersion of the metallic BS
has, to our knowledge, not been investigated before. The measured be-
havior of the BS dispersion below the NaCl bilayer is comparable to
the dispersion measured for BS electrons on clean Au(111) (Fig. 5.4),
though shifted in energy. At lower energies the BS exhibits a parabolic
behavior, which can be fitted by Eq. (1.17). The obtained values for
E 0 and m* /m e are reported in Table 5.1 as well. At higher energies
the measured dispersion for BS electrons below NaCl suddenly diverges,
similar to the bare Au(111) case [103]. For NaCl, however, this diver-
gence occurs at a higher energy (around 250 meV) and smaller wave
vector (k || = 0.13 Å−1 ) when compared to the clean Au(111) (at E f
and k || = 0.17 Å−1 ) [103].
E 0 (meV) m* /m e
IS BS IS BS
Au(111) -480 -720 ± 3 0.23 ± 0.03 0.16 ± 0.01
NaCl(2 ML) -269 ± 5 -460 ± 5 0.19 ± 0.03 0.08 ± 0.02
NaCl(3 ML) -160 ± 10 -250 ± 20 0.21 ± 0.03 0.12 ± 0.03
Table 5.1. Experimental values of the onset energy E 0 and the

effective mass m* of the NaCl/Au(111) interface state and bulk
state for both bilayered and trilayered NaCl. The experimental
values are determined by fitting the measured dispertion relations
using Eq. (1.17). The experimental values of the pristine Au(111)
SS and BS taken from [103] are added for comparison.
Figure 5.5. Dispersion behavior of both IS (gray squares) and BS

(grey triangles) electrons measured at NaCl(3 ML)/Au(111) (after
annealing). Both dispersion relations are fitted with a parabolic
relation for a free electron gas (grey solid line for IS and grey dot-
ted line for BS). The fits of the dispersion of NaCl(2 ML)/Au(111)
are added as a reference (IS black solid line and BS black dotted
line).
5.2.2 Trilayer NaCl/Au(111)

Next, we measured the dispersion relation of both IS and BS electrons
for NaCl(3 ML)/Au(111) (after annealing). The obtained data were
again fitted with Eq. (1.17) (Fig. 5.5) and reported in Table 5.1. For
both IS and BS E 0 is thus further shifted upwards, while the effective
mass slightly increases. The dependence of both E 0 and the E (k || ) on
the NaCl layer thickness has not been observed before [99]. It is an
unexpected result, since the induced electron density at the third NaCl
layer is negligible owing to the exponentially decaying wave function
outside the metal substrate. Consequently, the SS dispersion for three
layers of NaCl should not differ from that for two layers.
Measurements of the dispersion of the NaCl(2 ML)/Cu(111) IS by
Repp et al. [99] also revealed an upward shift of the IS onset energy, to-
gether with an increase of the effective mass relative to the SS electrons
on clean Cu(111) surfaces. This behavior was modeled by treating the
NaCl classically and by applying the static dielectric constant of NaCl
within a one-dimensional phase-accumulation model [99,105]. This mo-
del, however, can not explain the shift of the BS, nor the further shift
of E 0 for the surface and bulk electrons of NaCl(3 ML)/Au(111) when
compared to NaCl(2 ML)/Au(111).
Alternatively, the binding energy shift can be explained by Pauli
repulsion; the repulsive short-range interaction that arises as a conse-
quence of wave function overlap of the electron cloud around the sodium
or chlorine cores with the substrate electrons. In this weak-interaction
picture the orthogonality of the metal states to the nearly unperturbed
NaCl wave functions induces an oscillation of the metal wave functions in
the region of the atomic NaCl cores, and the charge from the evanescent
tail is driven back into the substrate. As a consequence, the energies of
the delocalized electrons of the Au(111), including the IS electrons, are
shifted towards the Fermi level. Because the IS are spatially confined to
the surface region, they are strongly influenced by the adsorbed NaCl.
Consequently, part of the electrons that belonged to the SS band of the
clean Au(111) are shifted through the Fermi level upon absorption of the
NaCl layer and transferred into BS. Additionally, also the BS electrons
close to the surface will feel the Pauli repulsion from the NaCl, giving
rise to the observed shift of the BS dispersion.
5.2.3 Influence of the Au(111) reconstruction

Surface states cannot only be observed via the standing waves in LDOS
maps but also in local (dI/dV )(V ) spectra. A sudden increase in tun-
neling conductance around -480 meV is the well-known signature of
the Au(111) surface state [106]. As shown by Repp et al. for the
NaCl/Cu(111) system [99], (dI/dV )(V ) measurements can also reveal
the shift in E 0 induced by the adsorption of NaCl.
The Au(111) herringbone reconstruction induces an electronic su-
perlattice at the NaCl/Au(111) interface at energies just above the on-
set energy of the IS. The formation of this superlattice is illustrated by
(dI/dV )(V ) curves, taken on the hcp and fcc regions of the Au(111)
herringbone reconstruction, for NaCl(2 ML)/Au(111) [Fig. 5.6 (b)] as
Figure 5.6. Averaged d I/d V(V) spectra taken on the hcp

(green) and fcc (black) region of the Au(111) herringbone re-
construction for (a) clean Au(111) (taken from [103]), (b)
NaCl(2 ML)/Au(111) and (c) NaCl(3 ML)/Au(111).
well as for NaCl(3 ML)/Au(111) [Fig. 5.6 (c)]. It can be seen that for
energies just above E 0 the LDOS is higher on the hcp compared to the
fcc regions. As a consequence the interface electrons are “confined” to
the hcp regions at these energies and no 2D IS is observed [Fig. 5.3 (b)].
This “confinement” prevents the development of a quasi-free 2-DEG,
thus no values can be extracted for k || for energies just above the onset
of the IS (Fig. 5.4 and Fig. 5.5). This feature stems from the differ-
ent Pauli repulsion between fcc and hcp Au(111) with the NaCl and is
further elaborated in chapter 6.
We also used recorded (dI/dV )(V ) curves to determine E 0 . On
NaCl(2 ML)/Au(111), based on an average of numerous dI/dV spec-
tra recorded at various positions of the Au(111) reconstruction, E 0 was
found to be −277 ± 10 meV, coinciding well with the value found by
fitting the experimentally obtained E (k || ) relation by the parabolic re-
lation expected for a quasi-free 2-DEG (E 0 = −269 ± 5 meV).
When we determine E 0 with a similar procedure for the IS of tri-
layered NaCl on Au(111), we obtain a value of −270 ± 10 meV. This

value does not coincide with the E 0 obtained from fitting the k || (E )
relation of the NaCl(3 ML)/Au(111) IS with the parabolic behavior of
a 2-DEG (E 0 = −160 meV). This difference is likely due to the fact
that the k || (E ) relation at low energies deviates from the used parabolic
dispersion for an ideal 2-DEG. On NaCl(2 ML)/Au(111) the difference
between hcp and fcc regions remains pronounced up to a merging point
of −30 meV (Fig. 5.6 (b)). Unfortunately, one cannot reliably map
the LDOS for voltages between −50 meV and +50 meV. This means
for NaCl(2 ML)/Au(111) that we can only measure the k || (E ) rela-
tion for relatively high energies where the IS behaves like a 2-DEG.
For NaCl(3 ML)/Au(111), however, the difference between hcp and fcc
disappears around −130 meV [Fig. 5.6 (c)], allowing us to investigate
k || (E ) just above the merging point. At these energies it is likely that
the 2D behavior of the IS is not yet recovered, explaining the differ-
ence between E 0 obtained from the 2-DEG fit and E 0 determined by
(dI/dV )(V ) curves.
The deviation of the IS dispersion from that of a 2-DEG constitutes
a large difference between NaCl/Au(111) and the previously studied
NaCl/Cu(111) [99]. This deviation is due to the interplay of the intrin-
sic reconstruction of the Au(111) surface and the NaCl, and possibly
explains the newly observed tunability of the NaCl/Au(111) IS by chang-
ing the NaCl from bi- to trilayer. The interplay between the Au(111)
surface and the NaCl will be further explored in the next chapter.
5.3 Interaction between NaCl crystallites and

the Au(111) surface
Some NaCl crystallites of smaller size can be found on Au(111) terraces,
far away from any Au(111) step edges. Islands with sizes up to 560 nm2
could be moved by scanning with the STM tip in close proximity to the
surface. In order to do this, the tunneling voltage was lowered consider-
ably and the tunneling current increased. An example of such procedure
is presented in Fig. 5.7. Similar findings have been reported for NaCl
crystallites on Cu(111) [107]. For NaCl on Cu(111) it was reported that
the maximal area of islands that are moved is 250 nm2 , wich is smaller
than the area of the island moved in Fig 5.7 [107]. On the nano-scale
friction scales with surface area, this indicating that the friction of NaCl
on Au(111) is lower than on Cu(111). The main difference between the
5.4 Summary 75
Cu(111) surface and the Au(111) surface is that the latter one has the
herringbone reconstruction.
Figure 5.7. Sequence of 45 × 45 nm2 STM images of a NaCl

island grown on Au(111). (a) is taken at larger tip-sample sep-
aration (V = +1.0 V, I = 0.1 nA). (b) and (c) are recorded at
smaller sample-tip separation (V = +0.1 V, I = 0.4 nA). Under
the latter conditions, the NaCl crystallite in the center is moved
by the STM tip.
5.4 Summary
We have investigated the growth and the properties of NaCl on Au(111)
by means of scanning tunneling microscopy and spectroscopy. NaCl
grows in polycrystalline bilayer islands if deposited onto clean Au(111)
at room temperature. By annealing the substrate we transformed the
bilayer NaCl into trilayer NaCl. Using Fourier-transform images of lo-
cal density of states maps we showed that the Au(111) SS survives as
a NaCl/Au(111) interface state. The energy versus wave vector disper-
sion of both interface state and bulk state electrons were determined for
both bilayer and trilayer NaCl/Au(111). The presence of the Au(111)
herringbone surface reconstruction is found to influence the properties
of the combined metal-insulator system and changes the energy versus
wave vector dispersion. Due to the Au(111) herringbone reconstruction
under the NaCl, the IS of the combined system can be tuned by control-
ling the thickness of the adsorbed NaCl layer. This tuning is a special
property of the NaCl/Au(111) system under study and is different from
the NaCl/Cu(111) system that has been studied earlier [99]. Another
effect of the herringbone reconstruction is the apparent lowering of the
energy required to move NaCl crystallites on the metal surface.
Chapter 6
Surface-reconstruction at
play: resolving all atoms of
an alkali halide and
modulation of the
NaCl/Au(111) quantum
well state
The results presented in this chapter are based on:
Surface-reconstruction at play: resolving all atoms of an

alkali halide and modulation of the NaCl/Au(111) quantum
well state
Chris Van Haesendonck, Peter Lievens, Mario I. Trioni,
Livia Giordano, and Gianfranco Pacchioni
Submitted
Alkali halide substrates represent the archetypical ionic insulator

material. Since they can be grown as very thin atomically flat layers
and exhibit a large band gap, they are widely used for various funda-
mental studies [99,108,109]. Ever since the first scanning probe measure-
ments with atomic resolution on alkali halides, only one of the atomic
species could be imaged [8,9]. However, simultaneously resolving both
77
78 Surface-reconstruction at play
alkali and halogen atoms is essential for quantitative scanning probe

microscopy studies of point defects [10], adsorbates [12], tip induced
manipulation [13], and other surface processes on the atomic scale [110].
In this chapter we demonstrate the imaging of NaCl with unprecedented
atomic resolution by exploiting the Au(111) herringbone reconstructed
surface to modulate the interaction between the alkali halide and the
metal surface. Ab initio calculations reveal that the adsorption of the
dielectric NaCl layer causes an upward shift of the Au(111) surface state
in quantitative agreement with the experimental spectroscopic measure-
ments. This shift depends on the local NaCl/Au(111) interaction which
is modulated by the hcp versus fcc character of the Au(111) reconstruc-
tion. Under some conditions, these local variations allow for simultane-
ous visualization of both the sodium and the chlorine component of the
thin alkali halide bilayer.
Alkali halide films have previously been studied by scanning tun-
neling microscopy (STM) and scanning force microscopy (SFM). The
resolution that can be achieved by STM is related to the higher density
of occupied metal induced gap states at the anions when compared to
the cations [10,101]. As a result only the halide atoms are imaged as
protrusions in STM topography images. SFM resolution on the other
hand is governed by the local electrostatic potential of the SFM tip in-
teracting with the sample [111]. Either the halide atoms or the alkaline
atoms can be visualized by controlling or measuring [112] the polarity
of the local electrostatic potential of the SFM tip. Imaging both species
at the same time was, however, not achieved by neither STM nor SFM
until now.
Theoretical model
We gratefully acknowledge M. I. Trioni, L. Giordano, and G. Pacchioni,
who carried out and analyzed all the theoretical calculations presented
in this chapter [113].
The complexity of the system associated with i) the long range her-
ringbone reconstruction of the Au(111) surface and with ii) the incom-
mensurate growth of the square NaCl lattice on the hexagonal Au(111)
surface, makes a full ab initio treatment of the system unfeasible. For
this reason we adopted two different strategies to describe the involved
systems. We briefly describe the two different strategies, a full descrip-
tion is given in [113] and its supplementary information.
For the analysis of the interface state shift (section 6.1) the clean
6.1 NaCl/Au(111) interface state 79
Au(111) surface and the NaCl/Au(111) system are represented via a

model potential that is constant in planes parallel to the surface. The
potential depends only on the z-coordinate, which runs normal to the
surface. The key quantities that play a role in the observed phenom-
ena, such as the projected bulk band gap at Γ, the presence and the
energy of the Shockley surface state, and the work function of the gold
metal surface, are correctly described. The interface state shift anal-
ysis involved three steps. First, a phenomenological potential is taken
from the literature to describe the clean gold surface [114]. Second, ab
initio perturbation potentials are employed to simulate the surface re-
construction. Third, the NaCl adsorption of a NaCl bilayer is simulated
by a double well potential computed ab initio and fitted by an analytical
model potential.
For the interpretation of the experimental LDOS maps presented in
section 6.2 density functional theory calculations are performed with the
exchange-correlation functional proposed by Perdew, Burke, and Ernz-
erhof [115]. A coincidence unit cell containing four NaCl formula units
per layer is used and represents a 5% residual strain, which is accommo-
dated in a five-layers thick Au slab. Due to the weak interface interaction
the estimation of van der Waals forces was included via the pair-wise
force field as implemented in the DFT-D2 method of Grimme [116].
6.1 NaCl/Au(111) interface state

When deposited on a clean Au(111) surface at room temperature, NaCl
forms bilayered (100)-terminated islands with perfect non-polar step
edges that are transparent for the Au(111) herringbone reconstruction
in STM topographical images [Fig. 6.1 (a) and Figs. 6.2 (a) and (b)].
Figures 6.1 (b) and (d) present typical close-up views of the atomically
resolved NaCl surface, where the bright protrusions correspond to chlo-
rine ions [99,101].
In Fig. 6.1 (c) we present an atomically resolved 2D visualization
of (dI /dV )(V ) curves taken above trilayer NaCl along the black line
indicated in Fig. 6.1 (b). The increased intensity around −250 meV
stems from the onset of the Au(111) surface state, which survives as an
interface state at the NaCl/Au(111) interface. Remarkably, the onset
energy E 0 of the newly formed interface state appears to be spatially
dependent on the Au(111) herringbone reconstruction. From the spec-
tra in Fig. 6.1 (e) it can be seen that the interface state is shifted by
Figure 6.1. (a) 60 × 60 nm2 pseudo 3D topographical STM im-

age of NaCl islands grown on Au(111) (V = 1.0 V, I = 0.1 nA).
One can observe the carpet-like growth of NaCl islands over the
steps of the Au(111) surface. The Au(111) herringbone recon-
struction is not influenced by the NaCl islands and its corruga-
tion is the same on NaCl/Au(111) and on the bare Au(111). (b)
30 × 4 nm2 atomically resolved STM image of trilayer NaCl grown
on Au(111) (V = 600 mV, I = 0.6 nA). Such large patches of tri-
layer NaCl were obtained by annealing the substrate to 460 K,
which turns the NaCl from being predominantly bilayered to tri-
layered (chapter 5). (c) Atomically resolved 2D visualization of
(dI /dV )(V ) spectra along the black line in (b). The color scale
in the figure reflects the density of states, see inset in (e). The
sharp increase in the DOS when going from lower (top) to higher
(bottom) energy corresponds to the onset of the interface state of
the NaCl(3 ML)/Au(111), revealing the spatial variation of the
the onset energy of the interface state induced by the Au(111)
herringbone reconstruction. (d) 1.6 × 0.9 nm2 atomic resolution
close-up view of the trilayer NaCl island presented in (b). Pro-
trusions correspond to the Cl− ions. (e) Averaged (dI /dV )(V )
spectra taken on clean Au(111) and on trilayer NaCl on the hcp
[blue cross in (b)] and fcc [black cross in (b)] region of the re-
constructed Au(111), revealing a pronounced difference in onset
energy of the NaCl/Au(111) interface state between these two re-
gions.
70 meV towards the Fermi level in the fcc region when compared to the
hcp region. A similar modulation of the interface state by the Au(111)
herringbone reconstruction is observed for bilayered NaCl (Fig. 6.2). On
Figure 6.2. Topography and electronic structure of

NaCl(2 ML)/Au(111). (a) 26 × 26 nm2 STM image of
NaCl grown on Au(111) (V = −1.1 V, I = 0.1 nA). (b) Height
profile along the white arrow indicated in (a). (c) (dI/dV )(V )
curves on the hcp region (blue) and the fcc region (black) of
the underlying Au(111). (d) 2D visualization of the (dI/dV )
(V ) spectra along the white arrow in (a) reveals the spatial
dependence of the LDOS resulting from the Au(111) herringbone
reconstruction.
pristine Au(111), however, delocalized surface electrons respond mainly

to the reconstruction by spatially transferring state density at one par-
ticular energy, but there is no difference in the onset energy of the surface
state [106].
To understand the origin of the observed shift of E0 as well as to gain
quantitative insight into the reasons for the different E0 values in the fcc
and hcp regions,ab initio calculations on the NaCl/Au(111) system are
performed as described in the beginning of the chapter. The Au(111)
surface is described using the “embedding method” approach to model
the potential along the z−coordinate normal to the surface [117]. The
model potential is able to describe well the s-p like Shockley surface
state band starting at −480 meV. In Fig. 6.3 the squared modulus of
the Shockley state wave function is presented together with the model
potential of the Au(111) surface.
$#
)*+,-+./0ï12$3,45
$"
ï#
ï!" A0,/-$=>3!!!5
B>:9/8,$B+/+,
ï% ï& ï' $" $' $& $% $( $!" $!' $!& $!%

6.7+/-8,$9:*;$+<,$9.:7+$=>$0/?,:$3@5
Figure 6.3. Squared modulus of the Shockley state wave function

together with the model potential of the Au(111) surface.
The more attractive potential in the hcp region is modeled by con-

sidering the potential change induced in the surface region upon a com-
pression of the interatomic distances. In Fig. 6.4 we present the lateral
average of the potential induced in the surface region caused by a 5%
compression of the inter-atomic distance. The induced potential is at-
tractive and it should increase the binding energy of the surface state.
In fact, we found a downward shift of 16 meV for each unit of percent-
age of compression. There is no experimental verification possible for
this quantity, but as a first approximation it can be compared to the
attractive potential of 25 meV for the hcp region proposed by Chen et
al. [106].
Next, the adsorbed bilayer of NaCl on the Au surface is modeled by
a double well potential computed ab initio and fitted by an analytical
model potential (Fig. 6.5). The NaCl layer causes the pristine Shockley
state to become an interface state that is localized in the region between
the metal and the NaCl film. Although the potential due to the overlayer
is attractive, the ionic film acts as a repulsive potential for the surface
state, so that the interface state is less bound. This can be understood
when considering the potential barrier generated by the insulating gap.
In some energy regions the electronic states cannot penetrate. Conse-
%"%
ï%"$
)*+,-.+(/01.*1234(5.67
ï%"#
ï%"'
ï%"&
ï!"%
ï!"$ 66(:;0<(3A(2*2120(;.9,419
ï!"# B2112*C(:,*-20*
ï$ (% ($ (# (' (&
82913*-.(:;0<(1=.(:2;91(>,(43?.;(5@7
Figure 6.4. Potential induced by a 5% compression of the inter-

atomic distance.
quently the potential well “felt” by the surface state is narrower and
the bound state shifts from −480 meV to −220/−270 meV. The model
potentials of pristine Au(111) and of the NaCl/Au(111) system in the
fcc region are presented in Fig. 6.5, together with the calculated square
moduli of the surface and interface states in the fcc region. By taking
the atomic potential of the adsorbed insulating layer explicitly into ac-
count we find that the interface state wave function exhibits maxima
at the positions of the NaCl layers, unlike in earlier publications where
the NaCl is taken into account only via its static dielectric constant [99].
The calculated values for the onset energy of the different interface states
are listed in Table 6.1.
The good agreement between the theoretical and experimental val-

ues of E 0 in both the fcc and hcp regions demonstrates that the proposed
model potential captures the physics involved in the interface state shift.
As the interface state wave function is already damped after two NaCl
layers, the calculated shift for the trilayer is essentially the same as for
the bilayer, suggesting that the small experimentally found difference
may be due to a difference in the structural parameters of bilayered and
trilayered NaCl films.
Experimental Theoretical
fcc hcp fcc hcp
NaCl(2 ML) −220 ± 20 −270 ± 20 −211 −267
NaCl(3 ML) −230 ± 20 −290 ± 20 −211 −268
Table 6.1. Theoretical and experimental values of the onset en-

ergy E 0 (in meV) of the NaCl/Au(111) interface state for both fcc
and hcp Au(111) reconstructed regions. The experimental values
are determined from (dI /dV )(V) curves following the geometrical
procedure given in [118]. The experimental onset energy of the
pristine Au(111) surface state is −480 ± 10 meV for both regions.
The theoretical values are obtained using the embedding method
approach.
6.2 Resolving all NaCl atoms

The above results clearly demonstrate the important influence of the
herringbone surface reconstruction on the spatially varying properties
of the NaCl/Au(111) interface state. In the following we will show that
this interface modulation also influences the properties of the NaCl over-
layer. To investigate this influence we measured maps of the local density
of states (LDOS) within a broad energy range. In Fig. 6.6 (a) and (b)
we present a STM topographical image and the corresponding LDOS
map recorded at +600 meV, respectively, of a bilayered NaCl island. It
can be seen that the LDOS map reveals only one of the atomic species
of NaCl in the fcc regions [Fig. 6.6 (d)], similar to the topography image
[Fig. 6.6 (c)]. Remarkably, on the hcp regions both atomic species can
be resolved in the LDOS map [Fig. 6.6 (e)], in contrast to the topogra-
phy image. This is the first observation of both atomic species in real
space. It is clear that the modulated interface is a key element to ob-
tain the ultimate atomic resolution when using STM on an alkali-halide
layer. Both atoms are also resolved on the “ridges” of the herringbone
reconstruction, which provide a transition region from the fcc to the
hcp configuration. This atomic resolution could only be obtained after
a non-controlled changing of the microscope tip apex. It is likely that a
Cl− ion from the NaCl surface was picked-up during scanning in close
proximity of the NaCl surface. The role of the precise termination of
the tip apex on the measurement is discussed in more detail below. On
trilayered NaCl a similar full atomic resolution could not be achieved.
With the aim of understanding the origin of the unprecedented
6.2 Resolving all NaCl atoms 85
Figure 6.5. Lower panel: model potentials for pristine Au(111)

(solid lines) and NaCl(2 ML)/Au(111) (dashed line) in the fcc
region. Upper panel: comparison between the squared moduli of
the surface state of Au(111) (red line) and the interface state for
NaCl/Au(111) (blue line) wave functions in the fcc region. The
adsorption of the NaCl layers induces a repulsive potential above
the interface. Note the compression of the interface state.
atomic contrast on the hcp regions of the NaCl(2 ML)/Au(111) sys-

tem and why this contrast is not observed on the fcc regions of this
system, density functional theory (DFT) calculations are performed. A
possible theoretical interpretation of the full atomic resolution observed
in the experimental LDOS maps at 600 mV in the hcp region of the her-
ringbone reconstruction requires realistic calculations of the NaCl(2-3
ML)/Au(111), which explicitly consider the atoms of the system treated
with ab initio DFT. The model includes dispersion forces and accounts
for the experimentally measured work function reduction of 1.1 eV with
respect to the Au(111) surface (Fig. 6.7). As the lattice mismatch and
the different symmetry of the film with respect to the substrate make the
description of this system with periodic calculations not straightforward,
we constructed a model of the system with a coincidence structure. Such
a coincidence structure was obtained by superposing the two unit cells
presented in Fig. 6.8: a (2 × 2) NaCl unit cell on a ( 31 13 ) superstructure
of the Au(111) substrate (Fig. 6.8). For details of this notation see [120].
The interface coincidence structure does not allow to model the hcp
region of the Au(111) by a compression of the interatomic distances.
Figure 6.6. (a) 23 × 19.6 nm2 STM image of NaCl(2

ML)/Au(111) (V = 0.6 V, I = 0.4 nA). (b) Correspond-
ing LDOS map of (a). (c), (d) and (e) 2.9 × 2.9 nm2 close-up
views of the regions indicated in (a) and (b). A model of the
NaCl lattice is added as a guide for the eye. (c) is Fourier filtered.
Instead, the gold surface reconstruction is taken into account by con-

sidering an artificially reduced NaCl/Au(111) interface distance in the
hcp region (by 0.02 nm). The modulation of the Au(111) surface and
the stiffness of the NaCl layer justify this approach. Due to shorter
interface distance in the hcp region the NaCl layer locally experiences
an enhanced electron density, in agreement with the measured dI /dV
curves as presented in Fig. 6.1 (c), where it can also be seen that the
ridges of the herringbone reconstruction provide a smooth transition
from one region to the next. The influence of the interface distance on
the NaCl properties is significant. First of all it is found that a NaCl
rumpling of the order of 0.0076 nm occurs in the fcc region (with the
Cl relaxing outwards), which decreases to 0.0063 nm in the hcp region
[Figs. 6.6 (f) and (g)]. Moreover, the Na signal is enhanced in simulated
LDOS maps at +600 mV with respect to the Na signal at 0 V (Fig. 6.9)
Figure 6.7. Determination of the image state shift upon adsorp-

tion of NaCl onto Au(111). (dI/dV )(V ) curves obtained by nu-
merically differentiating z(V ) curves measured on bare Au(111)
and bilayer NaCl/Au(111) with closed feedback-loop. This mea-
surement, when performed with the same tip, provides a estimate
of the change in work function of the Au(111) surface upon the
adsorption of the NaCl [119]. A downwards shift of 1.1 V of the
image state series of Au(111) is observed upon adsorption of NaCl.
due to the increased density of metal induced gap states at the Na atoms
when going from the Fermi energy towards the conduction band. How-
ever, these two effects alone are not sufficient to cause the Na atoms to
appear as protrusions in the simulated LDOS maps.
Since the full atomic resolution could only be achieved with a mod-
ified STM tip, tip-sample interactions needs to be considered. By relax-
ing the tip-sample geometry using ab initio calculations [121], Bennewitz
et al. have shown that the presence of a negative ion at the apex of a
SFM tip can induce a significant relaxation of the NaCl surface ions
on Cu(111). Their calculations reveal an outward relaxation of the Na
atoms by 0.014 nm towards the tip, while Cl atoms are displaced about
0.006 nm away from the tip [121]. In our measurements this negative
ion could be a Cl atom that is picked up during scanning.
To check if this tip modification explains both the enhanced res-
olution while maintaining a difference in the fcc and hcp regions, we
Figure 6.8. Au(111) and NaCl(100) unit cells used to build the
interface model.
Figure 6.9. Simulated STM images. Left: Simulated STM image

(V = 600 mV) for NaCl(2 ML)/Au(111) at the fcc region (1.56 ×
1.56 nm2 , 5 × 10−6 eV/Å3 electron isodensity). Right: Isodensity
contours along the white line of the image presented in the left
panel within the fcc (green curve) and hcp (pink curve) models
(V = 600 mV, 5 × 10−6 eV/Å3 electron isodensity).
calculated for both hcp and fcc NaCl/Au(111) the energy cost of lifting
a Na atom from the NaCl layer to allow its detection in LDOS maps.
Our calculations indicate that the average energy needed to lift a Na
atom to the level of the mean Cl plane is of the order of 5 meV and
is 20% larger for the fcc region compared to the hcp one. The lower
energy cost for Na lifting is related to a more efficient screening of the
Na-Cl interaction by the Au electron density at the hcp regions. These
data concur also in indicating that an enhanced Na signal is expected
in the LDOS maps at the hcp region. The combined effect of the differ-
ent structure and tip-film interaction is then likely responsible for the
Na detection only at the hcp region (Fig. 6.10), while the enhanced Na
character of the LDOS at higher energies (Fig 6.11) can explain why the
full atomic resolution is visible in LDOS maps at 600 mV [Fig. 6.6 (d)]
and not in topographical STM images [Fig. 6.6 (c)].
Figure 6.10. Schematic representation of bilayered NaCl on fcc

Au(111)[ABCABC... (a)] and on hcp Au(111)[CBCABC... (b)],
respectively. A, B and C Au layers are represented by black dots,
dark gray dots, and light gray dots, respectively. Note that the
Au(111) herringbone reconstruction leaves the bulk stacking in-
tact, while only the atoms from the top layer change from the fcc
to the hcp positions. The atoms in the top layer of the NaCl are
connected by straight lines to indicate the rumpling. The influ-
ence on the rumpling of the NaCl by a negative ion terminating
the STM tip apex is also illustrated [121]. Since the energy cost
to displace Na atoms is lower in the hcp regions, this influence
is stronger in the hcp region than in the fcc regions. For clarity
reasons the rumpling of the NaCl is not drawn to scale.
For trilayered films the theoretical NaCl rumpling is found to be

larger (0.0134 nm) and not affected by the hcp versus fcc interface dis-
tance. Moreover, the energy cost associated with raising the Na atom
is more than twice for 3ML with respect to 2ML, mainly because of
the larger distortion required for this system in order to bring the Na
at the Cl level. This can explain the lack of full atomic resolution in
experimental LDOS maps for NaCl(3 ML)/Au(111).
1
Na/(Na+Cl)
Cl/(Na+Cl)
0.8
PDOS ratio
0.6
0.4
0.2
0
-1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Energy-Ef (eV)
Figure 6.11. Ratio between the partial density of states of the

Na and the Cl at the surface.
6.3 Summary
By growing NaCl on a Au(111) surface we are able to study the inter-
action between a stiff insulating material and a reconstructed surface.
This interaction modulates the interface state, leading to an enhanced
electron density in the hcp regions of the Au(111) herringbone surface
reconstruction when compared to fcc regions of this reconstruction. The
enhanced electron density of the hcp regions reduces the rumpling of bi-
layered NaCl/Au(111) and lowers the energy cost required to lift a Na
atom out of the NaCl film, allowing to resolve for the first time the
atomic positions of both species of a alkali halide by STM.
Chapter 7
Co atoms on and F-centers

in NaCl/Au(111)
The results presented in this chapter are part of an ongoing work.

Theoretical calculations are being carried out by the group of G. Pac-
chioni [113], and additional measurements are planned as described in
the outlook.
In this chapter we give an overview of our experimental STM/STS
work done on two different types of isolated few electron systems, i.e.
F-centers created in NaCl(2 ML)/Au(111) by the STM tip (section 7.1)
and Co atoms adsorbed on NaCl(3 ML)/Au(111) (section 7.2). We
demonstrate that F-centers exhibit pronounced electronic interaction
when they are created in close proximity of each other, i.e. the forma-
tion of discrete “molecular orbitals”-like states. On the other hand, we
show that Co atoms strongly interact with the surrounding NaCl and
form combined electronic states with the NaCl(3 ML)/Au(111) system.
Furthermore, atomic resolution imaging reveals that Co atoms can oc-
cupy two positions on the NaCl, i.e. on top of the Na or on top of the
Cl.
7.1 F-Centers in NaCl(2 ML)/Au(111)

Ideal model systems for the study of tunneling through a localized elec-
tronic state decoupled by a polar insulator from the metal substrate and
the associated line shape is provided by atomic size vacancies and ad-
sorbates on NaCl films on metal surfaces. An atomic-scale vacancy or
adsorbate on an otherwise defect-free environment constitutes a fairly
91
92 Co atoms on and F-centers in NaCl/Au(111)
simple and well defined system that is ideally suited for fundamental
investigations.
The intrinsic defects in bulk NaCl and at NaCl surfaces have been
studied in detail both experimentally and theoretically. These studies
have shown that a Cl vacancy, which is commonly referred to as a color
or F-center, introduces a singly occupied defect state in the insulator
band gap [10].
Figure 7.1. (a) (2.4 nm × 2.4 nm) topographical STM image

of a F-center in NaCl(2 ML)/Au(111) (I = 0.6 nA, V = +500
mV). Only the Cl atoms of the NaCl are imaged. (b) (1.1 V ×
2.4 nm) 2D visualization of (dI /dV )(V ) spectra along the white
line in (a). (c) Averaged (dI /dV )(V ) spectra taken on the NaCl
(black cross in (a)) and the F-center (red cross in (a)), revealing an
enhancement of the onset of the NaCl(2 ML)/Au(111) interface
state by the F-center.
An example of a single Cl vacancy site created in bilayered NaCl

on Au(111) is presented in Fig. 7.1 (a). The Cl vacancy appears in the
STM image as a missing protrusion which is identified as a missing Cl
ion. The Cl ions neighboring the vacancy site appear slightly elevated in
STM topography images [10]. Using the electron-stimulated desorption
technique described in section 3.4, it was not possible to control the
amount nor the exact location of the created vacancies. Furthermore,
stable STM imaging appeared to be hampered by the tendency of the
Cl vacancies to react with the STM tip and disturb the measurement.
F-centers could only be created reliably in bilayer NaCl films due to the
high voltages required for this technique.
First, we investigated the electronic properties of a single F-center
7.1 F-Centers in NaCl(2 ML)/Au(111) 93
and its influence on the surrounding NaCl (Fig. 7.1). We find that
the defect-induced potential gives rise to an interface state localization
(ISL) [122], which can be observed as a peak in the dI /dV spectrum
around -230 meV [Figs. 7.1 (b) and (c)]. This state is split off from the
NaCl(2 ML)/Au(111) QWS near the band bottom. Since the precise ISL
energy depends on the onset energy of the QWS, which is modulated
by the underlying Au(111) herringbone reconstruction (chapter 6), it is
difficult to determine a precise value of the onset energy of the ISL. Our
results are in good agreement with earlier findings of Repp et al. [10],
who reported a small shift of -0.04 eV of the ISL onset energy with
respect to the onset energy of the NaCl(2 ML)/Cu(111) QWS.
Figure 7.2. (a) (3.3 nm × 3.3 nm), (e) (3.2 nm × 3.2 nm), and
(i) (3.2 nm × 3.2 nm) Topographical STM images of F-centers in
NaCl(2 ML)/Au(111) (I = 0.5 nA, V = +500 mV). (b), (f), and
(j) LDOS maps of the regions shown in (a), (e), and (i) taken at
+400 mV. (c), (g), and (k) LDOS maps of the regions shown in
(a), (e), and (i) taken at +500 mV. (d), (h), (l) LDOS maps of
the regions shown in (a), (e), and (i) taken at +600 mV.
In the LDOS maps presented in Fig. 7.2 we can see that F-centers
at close proximity are able to couple electronically, thereby forming dis-

crete “molecular orbital”-like states in the LDOS maps [Fig. 7.2 (f),
(g), (h) and (j), (k), and (l)] that are not observed for single F-centers
[Fig. 7.2 (a) tot (d)]. In the first row of Fig. 7.2 we see that a single
isolated F-center far from defects develops no special electronic states.
In the middle row of Fig. 7.2 we see a p-like orbital developing at the
position of the two central F-centers, and in the bottom row of Fig. 7.2
we that the F-centers [located in the middle of Fig. 7.2 (i)] form different
“molecular orbital”-like states in their LDOS. At +600 mV [Fig. 7.2 (l)]
a fifth F-center on the left hand side appears to interact as well with
the four middle ones. On the right hand side two single F-centers can
be observed, which do not reveal interaction. These findings indicate
that the F-centers interact only up to a maximum separation of one Cl
atom. To our knowledge mutual interaction of F-centers has not been
studied by STM so far. This is hence the first “real space” observation
of electronic coupling of F-centers in NaCl, which leads to the creation
of a few electron system inside an insulator.
7.2 Co atoms deposited on NaCl(3 ML)

The adsorption of noble metal atoms on NaCl(2 ML)/Cu has been
studied in detail [123,124]. It was shown that the noble metal atoms
do not interact strongly with the NaCl and can be charged stably on
NaCl(2 ML)/Cu, but not on NaCl(3 ML)/Cu. We used more reactive
atoms (Co) in our study, in order to investigate the influence of the ad-
sorbate on the local electronic properties of the NaCl. For the study of
single atoms trilayered NaCl obtained by annealing prior to the depo-
sition of the Co atoms was chosen as a support because of the nearly
defect-free surface it presents.
Before depositing the Co atoms the NaCl/Au(111) substrate was
annealed in order to obtain defect-free NaCl(3 ML)/Au(111) to minimize
any influence of imperfections of the NaCl on the Co atoms.
7.2.1 Topography
Co atoms are observed on top of both Cl and Na [Fig. 7.3 (Left)]. Co
on Cl is observed as a bright protrusion (Fig. 7.3 (Left), upper two
bright dots), while Co on Na is observed as a “bridge” between four
neighbouring Cl atoms (Fig. 7.3 (left), three atoms). No Co is adsorbed
in between the Na and the Cl. This indicates that single Co atoms
7.2 Co atoms deposited on NaCl(3 ML) 95
Figure 7.3. (Left) (5.2 nm × 8.0 nm) Topographical STM image

of 5 Co atoms adsorbed on NaCl(3 ML)/Au(111). Only the Cl
atoms of the NaCl are imaged. Co located on top of a Cl atom
is observed as a bright protrusion (2 uppermost Co atoms), while
Co located on top of a Na atom is observed as a bridge between
neighbouring Cl atoms (3 atoms). (Right) (3.5 × 5.3) Topograph-
ical STM image of Co atoms adsorbed on NaCl(3 ML)/Au(111)
after annealing the sample to 460 K. Only the Cl atoms of the
NaCl are imaged. The Co is still adsorbed on both Cl and Na.
Co located on top of a Cl atom is observed as a bright protrusion,
while Co located on top of a Na atom is observed as a bridge
between neighbouring Cl atoms.
interact differently with the NaCl than the Au and Ag atoms used in
previous studies, which adsorb only on top of Cl [123,124],
We annealed the NaCl(3 ML)/Au(111) up to 460 K [Fig. 7.3 (Right)]
and observed that the adsorbed Co atoms remain stable on both posi-
tions (Na and Cl), in contrast to Au atoms on NaCl(2 ML)/Cu(111),
which are found to be mobile above 60 K [123].
7.2.2 Local density of states

We investigated the electronic properties of the IS near the adsorbed
Co atoms. A 2D visualization of (dI /dV )(V ) spectra in the vicinity
of a Co atom adsorbed ontop of a Na atom, taken along the blue line
indicated in Fig. 7.4 (a), is presented in Fig. 7.4 (b). It can be discerned
Figure 7.4. Topographical STM image of a Co atom adsorbed

ontop of a Na atom (a) and 2D visualization of (dI /dV )(V ) spec-
tra along the blue line (a).
that the LDOS is enhanced ontop of the Co atom at the onset energy
of the IS. The chosen direction of the 2D visualization is parallel to the
Au(111) herringbone reconstruction, and not along one of the symmetry
axes of the NaCl, in order to filer out the local electronic variations
stemming from the Au(111) herringbone reconstruction (section 6.1).
For Co atoms ontop of Cl however [Fig. 7.5 (b)], the LDOS is lower at
the onset energy of the IS.
Figure 7.5. Topographical STM image of a Co atom adsorbed

ontop of a Cl atom (a) and 2D visualization of (dI /dV )(V ) spec-
tra along the blue line (a).
Additionally, when mapping the LDOS of Co adsorbed on trilayered

NaCl on Au(111) we find that “orbitals” develop around the Co atoms
at specific energies (Fig. 7.6). These orbitals are different for Co ad-
sorbed on Na when compared to Co adsorbed on Cl. This implies that
the observed orbitals are not just the atomic Co orbitals, but that the
orbitals rather stem from an interaction between the Co and the NaCl,
in contrast to Au and Ag atoms adsorbed on NaCl [123,124]. Due to
the low coverage the Co atoms are well separated and no interaction
7.3 Summary 97
between Co atoms was observed.
Figure 7.6. Topographical STM images and corresponding

LDOS maps of Co atoms adsorbed on NaCl(3 ML)/Au(111) at
the Na (upper row) and the Cl (bottom row) position.
7.3 Summary
We have measured with atomic resolution the few-electron systems for-
med by F-centers in bilayered NaCl and Co atoms adsorbed on trilayered
NaCl on Au(111).
When measuring the F-centers, we observed the formation of or-
bitals for multiple F-centers that are in close proximity.
Measurements on the Co atoms indicate that Co atoms are strongly
bound to trilayered NaCl on Au(111), since the Co remains stable on
both the Na and the Cl position upon annealing to 460 K. By recording
LDOS maps we visualized the electronic state formed by the interaction
of the Co with the NaCl.
Chapter 8
Imaging Co nanoclusters
deposited on thin NaCl
films with atomic resolution
and 3D reconstruction
The results presented in this chapter are based on a manuscript in prepa-

ration:
Imaging Co nanoclusters deposited on thin NaCl films with

atomic resolution and 3D reconstruction
There is a huge potential to build new electronic and chemically

sensitive devices using nanostructured thin layers made out of nanopar-
ticles, if the construction of the devices is based upon atomistic under-
standing of the nanoparticles [125]. Therefore, when studying nanopar-
ticles, the first important step is the determination of their structure.
This is a nontrivial task since nanoparticles can present a large variety of
structures, and due to their small size determination of these structures
is experimentally challenging. Scanning probe microscopy can play a
important role in unraveling the morphology of nanoparticles, as it has
the capabilities to obtain atomically resolved images in real space of
the outer shell of the nanoparticles. However, so far not many atomi-
cally resolved studies have been done with scanning probe microscopy
99
100 Imaging Co nanoclusters with atomic resolution
on nanoparticles. Here we report on atomically resolved imaging of Co

nanoparticles measured by scanning tunneling microscopy. We demon-
strate that in certain cases we can even reconstruct the complete 3D
structure of the nanoparticle based on the STM measurements.
For the experiments reported in this chapter, Co clusters were de-
posited onto clean trilayer NaCl on Au(111), enabling us to systemati-
cally investigate the morphology as well as the electronic properties of
individual nanoclusters in the size range of a few up to several hundreds
of atoms by means of STM and STS under controlled ultra-high vacuum
(UHV) conditions. In section 8.1, we find that the clusters have neg-
ligible mobility after deposition at room temperature on clean trilayer
NaCl on Au(111). By means of high-resolution STM measurements we
were able to observe the atomic structure of the Co clusters. From an
extensive height analysis (section 8.1) combined with the measured elec-
tronic properties of the Co clusters (section 8.2), we infer that the cluster
deformation upon impact on the surface is rather limited and that the
clusters partially penetrate into the NaCl.
Finally, in section 8.3 we reconstruct the complete 3D structure of
a Co nanoparticle based on the STM measurements.
8.1 Morphology
In order to investigate single Co clusters, their mobility on the surface
needs to be limited to keep them separated. Previously, deposited Co
clusters on Au(111) where shown to have a very low mobility, which
was related to the strong metal-metal interaction [68]. However, this
interaction is canceled out when changing the substrate from Au(111)
to NaCl. In order to ensure that the clusters are immobilized upon
landing onto the NaCl, we relied on the inherent kinetic energy of the
clusters in the cluster beam.
From previous experiments it is known that clusters can get trapped
on the surface upon landing if they have enough kinetic energy [126,
127]. Following the relation derived from the conservation of energy and
momentum in a binary elastic collision [126,128] and using the value of
0.23 eV as the potential barrier for local extraction of a single atom of
the NaCl [107], we find that the threshold energy for self-pinning is about
0.13 eV/atom. Our laser vaporization source creates Co clusters with
a low kinetic energy (≈ 0.15 eV/atom). This kinetic energy is enough
to ensure self-pinning of the Co clusters into the NaCl, but small when
8.1 Morphology 101
compared to the atomic binding energy within the cluster. Consequently,

negligible cluster fragmentation is expected upon impact.
Figure 8.1. (a) 90 × 90 nm2 pseudo 3D STM image of Co clusters

on Au(111) and on NaCl/Au(111). The clusters reveal negligible
mobility after deposition on thin NaCl films. (b) (7 × 7 nm2 ) and
(c) (9 × 9 nm2 ) close-up view of a Co cluster on NaCl/Au(111).
In Fig. 8.1 (a) we present the morphology of preformed Co clus-

ters deposited on both Au(111) and NaCl/Au(111). The clusters are
randomly distributed across the surface and reveal negligible mobility
after deposition on both the Au(111) and the NaCl, indicating that the
clusters indeed pin themselves upon landing. Occasionally, next to the
clusters a defect can be observed, which is likely created by the cluster
when it lands on the NaCl [Fig. 8.2 (a)].
When we zoom in on individual clusters in order to visualize them
with high resolution [Figs. 8.1 (b) and (c), and Fig. 8.2], we can im-
age atomic size features. STM gives only a “top-down” projected image
of the nanoparticles. In view of the expected tip convolution it is not
straightforward to extract detailed 3D crystallographic information for
the nanoparticle. Nevertheless, this result is very promising in view of
determining active catalytic sites on nanoparticles with STM. In sec-
tion 8.3 we will attempt to construct the 3D crystal structure of the
cluster presented in Fig. 8.1 (c).
Figure 8.2. (a) 10 × 6 nm2 close-up view of a Co cluster on

NaCl/Au(111). A pinning defect is visible in the NaCl at the
edge of the cluster. (b) (9 × 9 nm2 ), (c) (8 × 8 nm2 ), and
(d) (11 × 11 nm2 ) close-up views of individual Co clusters on
NaCl/Au(111).
We performed a systematic analysis of the cluster height distribu-

tion, both on the Au(111) (<h> = 1.9 nm) and on the NaCl/Au(111)
(<h> = 1.7 nm) [Fig. 8.3]. When we compare the cluster distribution
with the mass spectrum of the free cationic Co clusters in the cluster
beam [Fig. 8.3 (<d> = 2.5 nm)] we see that the apparent cluster height
after deposition is less than the cluster diameter in the gas phase. The
restricted flattening of the clusters on the Au(111) is consistent with our
previous measurements [68], where we found that the clusters become
hemispherical after deposition. However, one must keep in mind that
the similarity in height distribution of Co clusters on Au(111) and NaCl
does not prove that the clusters that landed on the NaCl are indeed
hemispherical. As the height distribution of the clusters on the NaCl
and on the Au(111) coincide well, we anyway estimated the amount of
atoms of the cluster by calculating the volume of a hemisphere with the
same height and dividing it by the volume of a Co atom.
8.2 Electronic properties 103
Figure 8.3. Height histogram of the deposited Co clusters on

Au(111) as determined by STM measurements, together with the
diameter abundance spectrum of the free clusters. Clusters expe-
rience only a rather limited flattening upon impact and interaction
with the Au surface [68]. The average cluster height on the NaCl
is lower than the average cluster height on the Au(111).
8.2 Electronic properties
Unlike Co clusters deposited on Au(111), there is a real possibility for

Co clusters deposited on NaCl to penetrate the topmost surface layers.
As we work with trilayer NaCl, this means that the Co clusters may
penetrate down to the Au(111) surface. To investigate this in more de-
tail, we measured I (V ) and dI /dV (V ) spectra on top of the Co clusters
on NaCl. If the clusters land on top of the NaCl, they form a double
tunneling junction with the STM tip and a Coulomb blockade should
be present around zero bias voltage, as explained in section 1.4. On the
other hand, if the clusters penetrate the NaCl, the Co clusters will be
in contact with the Au(111) and no such blockade will be observed [68].
Figures 8.4 (a) and (b) present both I (V ) and (dI /dV )(V ) spectra,
respectively, on top of two Co clusters on NaCl, where both cases are
visible. The majority of the clusters (90 %) reveal a metallic behavior
similar to the curves in black in Fig. 8.4. The other clusters exhibit
Figure 8.4. (a) I (V ) and (b) corresponding dI /dV (V ) spectra

taken at the center of Co clusters on NaCl (black and green) and at
the surrounding NaCl/Au(111) surface (red). Inset in (a): STM
image of the region were the curves where taken.
a Coulomb blockade around zero bias voltage, similar to the curves in

green Fig. 8.4. These results indicate that a small but significant per-
centage of the clusters (10 %) land on top of the NaCl. However, as
the majority of the clusters is able to penetrate the NaCl, these results
show that for most clusters, part of the cluster is hidden in the NaCl
and that we underestimate the cluster height by measuring the height
of the clusters with respect to the NaCl surface.
An important issue is to determine whether the clusters are flat at
the bottom, or are only in contact with the Au(111) by a few atomic
defects. As we only measure the top of the nanoparticle with STM, we
8.3 Atomic reconstruction 105
have to determine this in an indirect way. We measured the Coulomb

blockade on a number of clusters, and plotted the size of the gap as a
function of cluster height. The size of the Coulomb blockade in a sys-
tem is determined by the dominating capacitance of the two junctions,
i.e., Vgap = e/max{C1 , C2 } [129] with e the electron charge and C1 and
C2 the capacitance of cluster-tip and cluster-substrate, respectively. As
the cluster-substrate capacitance is enhanced by the presence of a dielec-
tricum and the bottom of the cluster may be flattened due to the impact
upon landing [80], it is reasonable to assume that the cluster-substrate
capacitance is the largest.
In a simple approximation we can estimate the cluster-substrate
capacitance using a parallel plate model: C = ǫ0 ǫr S/d, with S the surface
of the capacitance and d the NaCl thickness. Using ǫN aCl = 5.9 and
0.28 nm as the thickness for one NaCl layer, and calculating S from
the measured cluster height, we draw curves for the expected gap as a
function of the measured cluster height (Fig. 8.5). For the calculation of
S we assume that the clusters are hemispherical, thereby obtaining the
largest possible estimation of S. The gap size of clusters with a smaller
S should be higher than our estimate, while partial penetration of a
NaCl layer will lower the gap size. From comparison between theory
and experiment we learn that the clusters can penetrate the complete
NaCl layers, and that clusters that are in contact with the Au(111)
are likely to have a complete contact with the Au(111), and not just a
contact by a few atomic defects.
8.3 Atomic reconstruction

We attempted to reconstruct the cluster presented in Fig. 8.1 (c). An
image of the final result is presented in Fig. 8.6 (a). The process to obtain
this reconstruction contained several steps. In a first step we estimate
the number of atoms in the cluster. From the analysis mentioned in
section 8.1 we obtain an estimate of 192 Co atoms.
The second step is to determine the crystal structure and lattice di-
rection of the cluster. It has been shown that small Co particles exhibit
the fcc crystal structure [130], while previous experiments on deposited
Co clusters revealed that Co clusters in contact with Au(111) align them-
selves with the Au(111) surface. Their heights are preferentially multi-
ples of one Co monolayer (one monolayer = 0.205 nm) [68]. The cluster
seen in Fig. 8.1 (c) has metallic behavior and is thus probably in good
Figure 8.5. Measured gap size of the Coulomb blockade as a

function of cluster size.
contact with the Au(111), as argued in section 8.2. The measured height
of the cluster with respect to the NaCl is 1 nm [Fig. 8.6 (d)]. When this
height is added to the apparent height of the NaCl (0.4 nm), we infer
that the cluster is 7 monolayers high.
The third and final step consists of optimizing the color scale of
the experimental topographical STM image, in order to visualize the
contours of each atomic layer [Fig. 8.6 (b)]. We then start to stack the
atomic layers on top of each other in fcc stacking. However, in order
to get the best agreement between experiment and model the atoms
of the top atomic layer are attached in the hcp rather than the fcc
positions. This is not surprising since the topmost layer consists of a
stable hexagon, which geometrically only fits onto the layer below in the
hcp positions. Furthermore, hcp termination of certain facets of fcc Co
clusters is a known phenomena that has been studied theoretically [131],
and it is part of a growth mechanism which produces icosahedra from
smaller decahedra.
A comparison between the experimental result and the reconstructed
model can be obtained by comparing an experimental line profile with
a line profile obtained from the model. To obtain this line profile we
8.3 Atomic reconstruction 107
Figure 8.6. (a) 3D atomic model of the cluster presented in (b).

(b) Image of the cluster presented in Fig. 8.1 (c), color optimized
to visualize the individual atomic layers. (c) 3D model of the clus-
ter, with atoms remover in order to obtain the profile presented in
(d). (d) Experimental line profile of the cluster presented in (b),
taken along the indicated black line in (b), together with a side
view the direction of the arrow in (c) of the 3D model presented
in (c) in order to obtain a similar profile.
remove atoms of our model cluster [Fig. 8.6 (c) compared to Fig. 8.6
(a)]. The atoms were removed solely in order to reveal the atoms of
the cluster which constitute the height profile [Fig. 8.6 (a)], when view-
ing the cluster from the side in the direction of the arrow in Fig. 8.6
(c). The comparison is presented in Fig. 8.6 (d) and a good agreement
between experiment and model is obtained.
This reconstruction is only possible so far for clusters in contact with
the Au(111). Clusters not in contact with the Au(111), (Fig. 8.1 (b) and
Fig. 8.2) do not adjust their crystal structure to the Au(111), making
the assigning of the atomic layers less straightforward. Another point
to consider is that clusters that are not in contact with the Au(111) can
be non-crystalline, as they do not change their structure upon landing,
and it has been predicted that Co clusters can exhibit non-crystalline
shapes in the gas phase [131].
8.4 Summary
We investigated the morphology of Co clusters deposited onto trilayered
NaCl on Au(111) by means of STM and STS. Our measurements reveal
the presence of atomic size features on the clusters. Based on a com-
parison of the experimental results with realistic models, we are able
to determine the full atomic structure of one of the deposited clusters,
revealing the potential power of STM to determine the atomic structure
of nanoparticles.
By looking at their electronic properties with STS we demonstrated
that while the majority of the clusters penetrate into the NaCl, a smaller
but significant part of the clusters land on top of the thin insulating film.
We show that the particles exhibit a size-dependent Coulomb blockade,
which can be understood by treating the Co/NaCl(3 ML)/Au(111) sys-
tem as a classical parallel plate capacitance.
Conclusions and outlook
Summary
In this work we have studied the dependency of the properties of elec-
trons on their confinement conditions. In order to obtain these different
confined electron systems, we have investigated nanoparticles on both
Au(111) and NaCl, adatoms on and vacancies in NaCl, and the 2-DEG
formed at the NaCl/Au(111) interface.
The study of confined electrons necessarily consisted of two parts:
first we needed to determine the geometric structure of the entire system,
and secondly the electronic properties of the system had to be measured.
In order to address both criteria, this study was performed using scan-
ning tunneling microscopy (STM) to determine the morphology of the
systems under study, combined with scanning tunneling spectroscopy
(STS) in order to simultaneously measure their electronic properties.
Measurements were performed at low temperatures and under ultra-
high vacuum conditions, in order to ensure cleanliness and to obtain
high resolution results.
In the first experimental part of this work, we demonstrated that on
the nanoscale a system is more than the sum of its parts, as the catalytic
activity of TiOx nanoparticles deposited onto Au(111) showed. Experi-
mentally a mutual influence between the nanoparticles and the metallic
surface was measured. It was demonstrated that the deposited TiOx
nanoparticles are hemispherical and that they remained stable on the
Au(111) surface at elevated temperatures, while the study of the elec-
tronic properties revealed an electron transfer from the nanoparticles
to the Au(111). Taken together these results led us to a better under-
standing of the mechanism behind the catalytic activity of the combined
system, i.e. cooperative effects between the TiOx and the Au(111).
However, for the aim of this work, the investigation of individual
109
110 Conclusions
systems of confined electrons, the cooperative effects could mask the in-
trinsic properties of the confined electron systems under study. There-
fore, to minimize the influence of the substrate the second part of this
work consisted of investigating a means to isolate the nanosystems from
the metallic substrate.
Thin layers of NaCl were chosen to fulfill this requirement, and
their properties were studied in detail. Morphologically, the most ob-
vious findings for this part of our work were that NaCl grows in flat
stoichiometric layers on Au(111), and that the NaCl thickness could be
controlled by post-annealing. Electronically, it was found that the in-
terplay between the herringbone reconstructed Au(111) surface and the
NaCl played an important role. Firstly, because it led to a tune ability
of the NaCl/Au(111) interface state depending on the NaCl thickness.
Secondly we showed that due to the interplay between the Au(111) sur-
face reconstruction and the NaCl, both the Na and the Cl atoms of the
NaCl could be imaged, something not done before.
Finally, in the third part of this work individual nanosystems were
studied using the NaCl as a substrate. This part built up from the
smallest possible nanosystems, F-centers and adatoms, where electronic
properties akin to molecular orbitals were observed, to Co clusters de-
posited onto trilayered NaCl films. The topography of all three systems
under study was determined with atomic precision, which is essential if
one wants to build devices based on the atomic understanding of small
systems. The electronic properties of these systems were determined
with the same precision, leading to a complete experimental picture of
the confined electron systems.
General conclusions
Throughout this thesis we have striven to determine the morphology
and electronic properties of nanoscale systems with the high spatial and
energy resolution inherent to STM and STS. We investigated a ma-
jor difficulty that one faces when studying individual nanosystems with
STM, i.e. the interaction with the metallic support. By growing thin
NaCl films on top of the metallic substrate a possible solution to this
problem was obtained. As a proof of concept the properties of individual
nanoscale systems were determined.
From our study it is clear that on the nanoscale one cannot treat
the support and the nanosystems separately.
Conclusions 111
Outlook
Our studies of deposited clusters were limited to the study of non-mass-
selected clusters. As one cannot precisely count all the atoms in a single
nanoparticle with STM, a more thorough study of individual clusters
requires use of mass-selection prior to the deposition of the clusters.
With a knowledge of the exact amount of atoms, comparing theoretical
predicted structures to the experimental results will also become more
straightforward. To address this issue, an electrostatic mass selector for
the cluster deposition apparatus providing a better size resolution than
STM has been recently developed in-house (Fig. 8.7).
Figure 8.7. Schematic drawing of the in-house developed four-

fold electrostatic quadrupole mass selector [132], together with
simulated ion trajectories (shown in green). The simulation was
done with SIMION 7.0.
Another limitation is the study of clusters consisting of only one

element. Since the laser vaporization source is a very versatile cluster
source, many materials can be used and mixed [133,134]. Binary clusters
are especially well suited when one is thinking to use clusters as a kind of
“enhanced atom” that could serve for the construction of novel nanos-
tructured materials, since they offer the inherent possibility of tuning
their properties by changing their composition [135].
Recently, next to the spin-averaged STS used in this work, in the
laboratory of solid state physics and magnetism we developed the means
112 Conclusions
to perform spin-polarized STS [136]. As magnetism is observed in small

metal clusters, it would be highly interesting to perform spin-polarized
STS on such clusters, especially when combined with mass-selection.
Additionally, magnetic moments are also predicted for clusters consist-
ing of a cage of Si atoms doped with a magnetic transition metal atom
in the middle of the cage. As these clusters are predicted to be highly
stable, they could provide a way for developing novel silicon-based nan-
odevices [137].
Moreover, if individual clusters prove to be stable, it would be highly
interesting to determine what the effect is of bringing the clusters in close
proximity of each other. By studying the mutual interaction between
clusters, one starts to walk the road towards building new materials
out of nanoscale building-blocks, which may lead to unexpected new
properties.
However, our studies also point to interesting research that does not
include clusters. Part of the future work will be further completing the
research on Co atoms and F-centers of which the start is laid in chapter 7.
Spin-polarized STM measurements can give us more information on the
magnetic moments of the Co atoms, and additional measurements on F-
centers and their aggregates can help understand the observed formation
of molecular-like orbitals in further detail. By combining experiment
and theory, as is done in chapter 6, we can come to a more complete
understanding of these atomic sized systems.
Nederlandse Samenvatting
In het oude Griekenland speculeerden filosofen al over het bestaan van

“atomen”. Eén van hun grootste vragen was of materie al dan niet
eindeloos deelbaar was in kleinere deeltjes. Ze konden immers zien dat
wanneer men een stuk goud in twee stukken breekt, de twee stukken nog
altijd goud zijn (zelfde kleur, hardheid, ...). Maar, vroegen ze zich af,
hoe vaak kan men een klomp goud in twee breken zodat de helften nog
steeds goud zijn? Volgens de filosofie van Leucippus en de meer bekende
Democritus kon men materie in twee breken tot het zo klein was als
een atoom. Dit atoom was dan volgens hun redenering nog altijd goud,
maar als men het in nog kleinere stukken zou proberen te breken zou
men iets verkrijgen dat geen goud meer was.
Tegenwoordig weten we dat als we een materiaal stapsgewijs uit
atomen opbouwen, de eigenschappen van het zo gevormde deeltje sterk
afhangen van het precieze aantal atomen, van zijn vorm, evenals van de
chemische elementen die we hiervoor als bouwsteen gebruiken. De afme-
tingen van zulke deeltjes, waarvan de eigenschappen zo sterk afhangen
van hun specifieke samenstelling, is typisch één of enkele nanometer(s).
Zulke deeltjes noemen we dan ook nanodeeltjes.
De elektronische, magnetische en chemische eigenschappen van na-
nodeeltjes schalen dus niet met hun afmetingen, maar zijn uniek voor
ieder specifiek nanodeeltje. De eigenschappen van een nanodeeltje met
10 goud atomenkunnen bijvoorbeeld niet geëxtrapoleerd worden uit de
eigenschappen van nanodeeltjes met bijvoorbeeld 8 of 9 goudatomen.
De oorzaken van dit “afwijkend” gedrag zijn te vinden in het kwantum-
mechanisch karakter van de materie dat steeds meer uitgesproken wordt
bij een steeds kleiner wordende lengteschaal.
Het doel van dit doctoraat was het bepalen van de intrinsieke elek-
tronische eigenschappen van nanodeeltjes die op een oppervlak liggen.
Om deze eigenschappen accuraat te meten werd er gebruik gemaakt van
tunnelmicroscopie (Engels: Scanning Tunneling Microscopy, STM) ge-
113
114 Nederlandse Samenvatting
combineerd met tunnelspectroscopie (Engels: Scanning Tunneling Spec-

troscopy, STS) bij lage temperaturen en in ultra-hoogvacuüm (UHV).
Bij deze meettechniek wordt het te onderzoeken oppervlak afgetast met
een scherpe naald, terwijl simultaan een kwantummechanische tunnel-
stroom tussen naald en oppervlak wordt opgemeten. Met STM kan men
lokaal zowel de morfologische als de elektronische eigenschappen opme-
ten met een zeer hoge ruimtelijke en energetische resolutie.
We zijn begonnen met het onderzoeken van de eigenschappen van
TiOx nanodeeltjes die gedeponeerd werden op een atomair vlak Au(111)
oppervlak. Om deze nanodeeltjes te verkrijgen hebben we eerst zuivere
Ti nanodeeltjes aangemaakt in de gasfase en vervolgens deze Ti na-
nodeeltjes gedeponeerd op het Au(111) oppervlak bij UHV condities,
waarna we de nanodeeltjes gecontroleerd geoxideerd hebben. We heb-
ben zowel de morfologie als de elektronische eigenschappen van deze
nanodeeltjes onderzocht, zowel voor als na verwarmen tot 970 K. Met
behulp van systematische analyse van de morfologische data werd aan-
getoond dat de TiOx nanodeeltjes slechts een beetje worden afgevlakt
door de depositie op het Au(111) oppervlak, en dat ze hemisferisch zijn.
De mobiliteit van de nanodeeltjes was verwaarloosbaar, zelfs als het op-
pervlak verwarmd werd tot hoge temperaturen. Metingen van de elek-
tronische eigenschappen van de TiOx nanodeeltjes onthulden dat het
TiOx /Au(111) systeem gelijkrichting vertoonde, wat gezien werd als een
teken van de vorming van een Schottky-barrière tussen het n-type half-
geleidend TiOx en het Au(111) oppervlak. Een ladingsoverdracht van
de halfgeleider naar het Au(111) werd waargenomen. Deze ladingsover-
dracht hielp om de katalytische eigenschappen van het TiOx /Au(111)
systeem te verklaren.
Het verklaren van de katalytische eigenschappen is een belangrijk
resultaat, maar omdat de focus in dit werk op de eigenschappen van de
individuele nanodeeltjes lag en niet op het gecombineerd systeem nano-
deeltje/substraat wilden we de invloed van het substraat verminderen.
Teneinde de invloed van het substraat te minimaliseren werd ervoor
geopteerd om een dunne isolerende en chemisch inerte laag te groeien
op het Au(111) oppervlak, die dan als substraat voor de nanodeeltjes
kon dienen. Het gekozen materiaal waaruit deze laag bestaat was NaCl
(keukenzout). Het volgende deel van dit werk bestaat uit het in detail
karakteriseren van de gegroeide dunne NaCl lagen. NaCl is een isolator
met een grote bandkloof en groeit stoichiometrisch wanneer het bij de
juiste temperatuur gesublimeerd wordt. De oppervlaktetoestand van het
Nederlandse Samenvatting 115
Au(111) overleefde de groei van het NaCl en werd een intermediaire toe-
stand. Uit de STM studie bleek dat wanneer het NaCl werd gedeponeerd
op Au(111) op kamertemperatuur, er zich spontaan vlakke eilanden van
twee monolagen NaCl op het Au(111) vormden. Als het Au(111) daarna
verwarmd werd, werden deze eilanden omgezet in grotere eilanden van
drie monolagen NaCl. Een andere vondst was de belangrijke invloed van
de Au(111) “visgraat” reconstructie op de elektronische eigenschappen
van het NaCl/Au(111) systeem. Ten eerste omdat de reconstructie er-
voor zorgt dat de toestand aan het grensvlak afhangt van de dikte van
het NaCl, wat niet het geval is voor NaCl gegroeid op Cu(111). Ten
tweede omdat door de wisselwerking tussen het NaCl en de Au(111) vis-
graatreconstructie het mogelijk werd om zowel de Na als de Cl atomen
van het NaCl simultaan te visualiseren met behulp van STM, hetgeen
nooit eerder gerealiseerd was.
In het laatste deel van dit werk werden individuele nanosystemen
bestudeerd, gebruik makend van de dunne NaCl lagen als substraat.
Voor dit gedeelte werden deeltjes bestaande uit één enkel atoom tot en-
kele honderden atomen, dit zijn gedeponeerde Co clusters, bestudeerd
op NaCl. Als deeltjes met atomaire afmetingen hebben we zowel in-
dividuele Co atomen op het NaCl als ontbrekende Cl atomen in het
NaCl bestudeerd. Zowel de atomaire systemen als de Co nanodeeltjes
werden bestudeerd met atomaire precisie, essentieel als men nieuwe toe-
passingen wil ontwikkelen gebaseerd op het begrijpen van deze systemen
op atomaire schaal. De elektronische eigenschappen van beide systemen
werden met dezelfde precisie bestudeerd, waardoor we experimenteel een
volledig beeld van deze systemen verkregen hebben.
116 Nederlandse Samenvatting
Bibliography
[1] C. Kittel and P. McEuen, Introduction to Solid State Physics (Wi-

ley, New York, 2005), Vol. 7.
[2] D. A. Jefferson, J. M. Thomas, G. R. Millward, K. Tsuno, A.

Harriman, and R. D. Brydson, Atomic structure of ultrafine cat-
alyst particles resolved with a 200-keV transmission electron mi-
croscope, Nature 323, 428 (1986).
[3] J. Kiwi and M. Grätzel, Projection, size factors, and reaction dy-
namics of colloidal redox catalysts mediating light induced hydro-
gen evolution from water, J. Am. Chem. Soc. 101, 7214 (1979).
[4] H. J. Freund, Clusters and islands on oxides: from catalysis via

electronics and magnetism to optics, Surf. Sci. 500, 271 (2002).
[5] D. J. Griffiths, Introduction to quantum mechanics (Pearson Pren-

tice Hall, New York, 2005), Vol. 1.
[6] A. P. Alivisatos, Semiconductor clusters, nanocrystals, and quan-

tum dots, Science 271, 933 (1996).
[7] A. Lin, L. Hirsch, M. H. Lee, J. Barton, N. Halas, J. West, and R.

Drezek, Nanoshell-enabled photonics-based imaging and therapy
of cancer, Technol. Cancer Res. T. 3, (2004).
[8] K. Glöckler, M. Sokolowski, A. Soukopp, and E. Umbach, Initial

growth of insulating overlayers of NaCl on Ge(100) observed by
scanning tunneling microscopy with atomic resolution, Phys. Rev.
B 54, 7705 (1996).
[9] A. L. Shluger, R. M. Wilson, and R. T. Williams, Theoretical and

experimental investigation of force imaging at the atomic scale on
alkali halide crystals, Phys. Rev. B 49, 4915 (1994).
117
118 Bibliography
[10] J. Repp, G. Meyer, S. Paavilainen, F. E. Olsson, and M. Persson,

Scanning tunneling spectroscopy of Cl vacancies in NaCl films:
Strong electron-phonon coupling in double-barrier tunneling junc-
tions, Phys. Rev. Lett. 95, 225503 (2005).
[11] C. Barth and C. R. Henry, High-resolution imaging of gold clus-

ters on KBr(001) surfaces investigated by dynamic scanning force
microscopy, Nanotechnology 15, 1264 (2004).
[12] C. Bombis, N. Kalashnyk, W. Xu, E. Lægsgaard, F. Besenbacher,

and T. R. Linderoth, Hydrogen-Bonded Molecular Networks of
Melamine and Cyanuric Acid on Thin Films of NaCl on Au(111),
Small 5, 2177 (2009).
[13] O. Custance, R. Perez, and S. Morita, Atomic force microscopy as

a tool for atom manipulation, Nat. Nanotechnol. 4, 803 (2009).
[14] Z. Y. Li, N. P. Young, M. Di Vece, S. Palomba, R. E. Palmer,

A. L. Bleloch, B. C. Curley, R. L. Johnston, J. Jiang, and J.
Yuan, Three-dimensional atomic-scale structure of size-selected
gold nanoclusters, Nature 451, 46 (2007).
[15] S. Van Aert, K. J. Batenburg, M. D. Rossell, R. Erni, and

G. Van Tendeloo, Three-dimensional atomic imaging of crystalline
nanoparticles, Nature 470, 374 (2011).
[16] N. Bohr, On the Constitution of Atoms ands Molecules, Philos.

Mag. 26, (1913).
[17] N. W. Ashcroft and N. D. Mermin, Solid State Physics (Holt,

Rinehart and Winston, New York, 1976), Vol. 826.
[18] M. Brack, The physics of simple metal clusters: self-consistent

jellium model and semiclassical approaches, Rev. Mod. Phys. 65,
677 (1993).
[19] E. L. Wolf, Nanophysics and nanotechnology: an introduction to

modern concepts in nanoscience (Vch Verlagsgesellschaft Mbh,
Wienheim, 2006).
[20] W.Pauli, Über den Zusammenhang des Abschlusses der Elektro-

nengruppen im Atom mit der Komplexstruktur der Spektren, Z.
f. Physik 31, 765 (1925).
Bibliography 119
[21] F. Bloch, Heisenberg and the early days of quantum mechanics,

Phys. Today 29, 23 (1976).
[22] F. Bloch, Uber die Quantenmechanik der Elektronen in Kristall-

gittern Ph.D. thesis, University of Leipzig, 1928.
[23] W. Shockley, On the surface states associated with a periodic po-

tential, Phys. Rev. 56, 317 (1939).
[24] S. D. Kevan and R. H. Gaylord, High-resolution photoemission

study of the electronic structure of the noble-metal (111) surfaces,
Phys. Rev. B 36, 5809 (1987).
[25] M. F. Crommie, C. P. Lutz, and D. M. Eigler, Imaging standing

waves in a two-dimensional electron gas, Nature 363, 524 (1993).
[26] L. Bürgi, N. Knorr, H. Brune, M. A. Schneider, and K. Kern, Two-

dimensional electron gas at noble-metal surfaces, Appl. Phys. A
75, 141 (2002).
[27] P. T. Sprunger, L. Petersen, E. W. Plummer, E. Lægsgaard, and

F. Besenbacher, Giant Friedel oscillations on the beryllium (0001)
surface, Science 275, 1764 (1997).
[28] Y. Hasegawa and P. Avouris, Direct observation of standing wave

formation at surface steps using scanning tunneling spectroscopy,
Phys. Rev. Lett. 71, 1071 (1993).
[29] E. Bertel and N. Memmel, Promotors, poisons and surfactants:

Electronic effects of surface doping on metals, Appl. Phys. A-
Mater. 63, 523 (1996).
[30] L. Petersen, Scanning tunneling microscopy studies of the elec-

tronic structure of metal surfaces Ph.D. thesis, Institute of Physics
and Astronomy, University of Aarhus, 1999.
[31] F. K. Meier, Co on Pt(111) studied by spin-polarized scanning tun-

neling microscopy and spectroscopy Ph.D. thesis, Universitätsbib-
liothek, 2006.
[32] F. Reinert, G. Nicolay, S. Schmidt, D. Ehm, and S. Hüfner, Direct

measurements of the L-gap surface states on the (111) face of noble
metals by photoelectron spectroscopy, Phys. Rev. B 63, 115415
(2001).
120 Bibliography
[33] K. Schouteden, E. Lijnen, E. Janssens, A. Ceulemans, L. F. Chi-

botaru, P. Lievens, and C. Van Haesendonck, Confinement of sur-
face state electrons in self-organized Co islands on Au(111), New
J. Phys. 10, 043016 (2008).
[34] J. Kliewer, R. Berndt, and S. Crampin, Scanning tunnelling spec-

troscopy of electron resonators, New J. Phys. 3, 22 (2001).
[35] L. P. Kouwenhoven, C. M. Marcus, P. L. McEuen, S. Tarucha,

R. M. Westervelt, and N. S. Wingreen, in Mesoscopic Electron
Transport, edited by L. L. Sohn, L. P. Kouwenhoven, and G. Schön
(Springer, Dusseldorf, 1997), p. 105.
[36] K. Schouteden, Tunneling microscopy and spectroscopy of metal-

lic and magnetic nanoparticles on atomically flat surfaces Ph.D.
thesis, K.U.Leuven, 2008.
[37] R. Wiesendanger, Scanning Probe Microscopy and Spectroscopy:

Methods and Applications (Cambridge University Press, Cam-
bridge, 1994).
[38] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel, Surface studies

by scanning tunneling microscopy, Phys. Rev. Lett. 49, 57 (1982).
[39] G. Binnig and H. Rohrer, In touch with atoms, Rev. Mod. Phys.
71, 324 (1999).
[40] J.Tersoff. and D. R. Hamann, Theory of the scanning tunneling

microscope, Phys. Rev. B 31, 805 (1985).
[41] C. J. Chen, Origin of atomic resolution on metal surfaces in scan-

ning tunneling microscopy, Phys. Rev. Lett. 65, 448 (1990).
[42] N. D. Lang, Spectroscopy of single atoms in the scanning tunneling

microscope, Phys. Rev. B 34, 5947 (1986).
[43] G.Binnig, K. H. Frank, H. Fuchs, N. Garcia, B. Reihl, H. Rohrer,

F. Salvan, and A. R. Williams, Tunneling spectroscopy and inverse
photoemission: image and field states, Phys. Rev. Lett. 55, 991
(1985).
[44] J. Lambe and R. C. Jaklevic, Molecular vibration spectra by in-

elastic electron tunneling, Phys. Rev. 165, 821 (1968).
Bibliography 121
[45] J. Kröger, L. Limot, H. Jensen, R. Berndt, S. Crampin, and E.

Pehlke, Surface state electron dynamics of clean and adsorbate-
covered metal surfaces studied with the scanning tunnelling mi-
croscope, Prog. Surf. Sci. 80, 26 (2005).
[46] J. Klein, A. Leger, M. Belin, D. Défourneau, and M. J. L.Sangster,

Inelastic-electron-tunneling spectroscopy of metal-insulator-metal
junctions, Phys. Rev. B 7, 2336 (1973).
[47] I. Horcas, R. Fernandez, J.M. Gomez-Rodriguez, J. Colchero, J.

Gomez-Herrero, and A.M. Baro, WSXM: A software for scanning
probe microscopy and a tool for nanotechnology, Rev. Sci. Instrum.
78, 013705 (2007).
[48] N. Vandamme, E. Janssens, F. Vanhoutte, P. Lievens, and C.

Van Haesendonck, Scanning probe microscopy investigation of
gold clusters deposited on atomically flat substrates, J. Phys.: Con-
dens. Matter 15, S2983 (2003).
[49] C. Wöll, S. Chiang, R. J. Wilson, and P. H. Lippel, Determina-

tion of atom positions at stacking-fault dislocations on Au(111) by
scanning tunneling microscopy, Phys. Rev. B 39, 7988 (1989).
[50] D. D. Chambliss, R. J. Wilson, and S. Chiang, Nucleation of or-

dered Ni island arrays on Au(111) by surface-lattice dislocations,
Phys. Rev. Lett. 66, 1721 (1991).
[51] S. Narasimhan and D. Vanderbilt, Elastic stress domains and the

herringbone reconstruction on Au(111), Phys. Rev. Lett. 69, 1564
(1992).
[52] R. E. Palmer, S. Pratontep, and H.-G. Boyen, Nanostructured

surfaces from size-selected clusters, Nat. Mater. 2, 443 (2003).
[53] G. Verschoren, Transport and magnetic properties of cluster-

assembled nanogranular gold-cobalt films Ph.D. thesis, Katholieke
Universiteit Leuven, 2005.
[54] W. Bouwen, P. Thoen, F. Vanhoutte, S. Bouckaert, F. Despa, H.

Weidele, R.E. Silverans, and P. Lievens, Production of bimetallic
clusters by a dual-target dual-laser vaporization source, Rev. Sci.
Instrum. 71, 54 (2000).
122 Bibliography
[55] P. Piseri, T. Mazza, G. Bongiorno, C. Lenardi, L. Ravagnan, F.

Della Foglia, F. DiFonzo, M. Coreno, M. DeSimone, K. C. Prince,
and P. Milani, Core level spectroscopy of free titanium clusters in
supersonic beams, New J. Phys. 8, 136 (2006).
[56] I. Shyjumon, M. Gopinadhan, C. A. Helm, B. M. Smirnov, and R.

Hippler, Deposition of titanium/titanium oxide clusters produced
by magnetron sputtering, Thin Solid Films 500, 41 (2006).
[57] N. Nilius, Properties of oxide thin films and their adsorption be-
havior studied by scanning tunneling microscopy and conductance
spectroscopy, Surf. Sci. Rep. 64, 595 (2009).
[58] J. Repp, S. Fölsch, G. Meyer, and K. H. Rieder, Ionic films on vic-

inal metal surfaces: enhanced binding due to charge modulation,
Phys. Rev. Lett. 86, 252 (2001).
[59] A. Jaroenworaluck, W. Sunsaneeyametha, N. Kosachan, and R.

Stevens, Characteristics of silica-coated TiO2 and its UV absorp-
tion for sunscreen cosmetic applications, Surf. Interf. Anal. 38,
473 (2006).
[60] J. M. Herrmann, Heterogeneous photocatalysis: fundamentals and

applications to the removal of various types of aqueous pollutants,
Catal. Today 53, 115 (1999).
[61] K. Sunada, Y. Kikuchi, K. Hashimoto, and A. Fujishima, Bacte-

ricidal and detoxification effects of TiO2 thin film photocatalysts,
Env. Sci. & Tech. 32, 726 (1998).
[62] M. Castro, S. R. Liu, H. J. Zhai, and L. S. Wang, Structural

and electronic properties of small titanium clusters: A density
functional theory and anion photoelectron spectroscopy study, J.
Chem. Phys. 118, 2116 (2003).
[63] F. Della Foglia, T. Losco, P. Piseri, P. Milani, and E. Selli, Photo-

catalytic activity of nanostructured TiO2 films produced by super-
sonic cluster beam deposition, J. Nanopart. Res. 11, 1339 (2009).
[64] M. C. Barnes, A. R. Gerson, S. Kumar, and N. M. Hwang, The

mechanism of TiO2 deposition by direct current magnetron reac-
tive sputtering, Thin Solid Films 446, 29 (2004).
Bibliography 123
[65] E. Barborini, I. N. Kholmanov, A. M. Conti, P. Piseri, S. Vinati,

P. Milani, and C. Ducati, Supersonic cluster beam deposition of
nanostructured titania, Eur. Phys. J. D 24, 277 (2003).
[66] K. L. Jonas, V. Von Oeynhausen, J. Bansmann, and K. H. Meiwes-
Broer, Tunnelling spectroscopy on silver islands and large de-
posited silver clusters on Ge(001), Appl. Phys. A 82, 131 (2006).
[67] A. Kleibert, F. Bulut, R. K. Gebhardt, W. Rosellen, D. Sudfeld, J.
Passig, J. Bansmann, K. H. Meiwes-Broer, and M. Getzlaff, Corre-
lation of shape and magnetic anisotropy of supported mass-filtered
Fe and FeCo alloy nanoparticles on W(110), J. Phys.: Cond. Matt.
20, 445005 (2008).
[68] K. Schouteden, A. Lando, E. Janssens, C. Van Haesendonck, and
P. Lievens, Morphology and electron confinement properties of Co
clusters deposited on Au(111), New J. Phys. 10, 083005 (2008).
[69] Z. Song, J. Hrbek, and R. Osgood, Formation of TiO2 Nanoparti-
cles by Reactive-Layer-Assisted Deposition and Characterization
by XPS and STM, Nano Lett. 5, 1327 (2005).
[70] D. V. Potapenko, J. Hrbek, and R. M. Osgood, Scanning tunnel-
ing microscopy study of titanium oxide nanocrystals prepared on
Au(111) by reactive-layer-assisted deposition., ACS Nano 2, 1353
(2008).
[71] J. A. Rodriguez, S. Ma, P. Liu, J. Hrbek, J. Evans, and M. Perez,
Activity of CeOx and TiOx Nanoparticles Grown on Au (111) in
the Water-Gas Shift Reaction, Science 318, 1757 (2007).
[72] E. Farfan-Arribas, J. Biener, C. M. Friend, and R. J. Madix, Re-
activity of methanol on TiO2 nanoparticles supported on the Au
(111) surface, Surf. Sci. 591, 1 (2005).
[73] M. S. Chen and D. W. Goodman, Structure-activity relationships
in supported Au catalysts, Catal. Today 111, 22. (2006).
[74] J. Vangrunderbeek, C. Van Haesendonck, and Y. Bruynseraede,
Electron localization and superconducting fluctuations in quasi-
two-dimensional Ti films, Phys. Rev. B 40, 7594 (1989).
[75] C. Vlekken, De metaal-isolator overgang in zuurstofrijke Ti-lagen
Ph.D. thesis, K.U.Leuven, 1993.
124 Bibliography
[76] J. Biener, E. Farfan-Arribas, M. Biener, C. M. Friend, and R.J.

Madix, Synthesis of TiO nanoparticles on the Au(111) surface, J.
Chem. Phys. 123, 094705 (2005).
[77] S. Padovani, F. Scheurer, and J. P. Bucher, Burrowing self-

organized cobalt clusters into a gold substrate, Europhys. Lett.
45, 327 (1999).
[78] K. Schouteden, E. Lijnen, D. A. Muzychenko, A. Ceulemans,

L. F. Chibotaru, P. Lievens, and C. Van Haesendonck, A study of
the electronic properties of Au nanowires and Au nanoislands on
Au(111) surfaces, Nanotechnology 20, 395401 (2009).
[79] F. Claeyssens, S. Pratontep, C. Xirouchaki, and R. E. Palmer,

Immobilization of large size-selected silver clusters on graphite,
Nanotechnology 17, 805 (2006).
[80] H. Haberland, Z. Insepov, and M. Moseler, Molecular-dynamics

simulation of thin-film growth by energetic cluster impact, Phys.
Rev. B 51, 11061 (1995).
[81] S. J. Carroll, S. Pratontep, M. Streun, R. E. Palmer, S. Hobday,

and R. Smith, Pinning of size-selected Ag clusters on graphite
surfaces, J. Chem. Phys. 113, 7723 (2000).
[82] P. Pyykkö, Theoretical chemistry of gold. II, Inorg. Chim. Acta

358, 4113 (2005).
[83] D. R. Lide, CRC handbook of chemistry and physics (CRC press,

Florida, 1993).
[84] P. E. Viljoen and J. P. Roux, Oxidation of a Au + 4 at% Ti alloy,

Vacuum 41, 1746 (1990).
[85] D. V. Potapenko and R. M. Osgood, Preparation of TiO2

Nanocrystallites by Oxidation of Ti-Au(111) Surface Alloy, Nano
Lett. 9, 2378 (2009).
[86] R. Berndt, J. K. Gimzewski, and R. R. Schlittler, Bias-dependent

STM images of oxygen-induced structures on Ti(0001) facets, Surf.
Sci. 310, 85 (1994).
Bibliography 125
[87] E. Dupont-Ferrier, P. Mallet, L. Magaud, and J. Y. Veuillen, STM

investigation of the charge transport mechanisms to nanoscale
metallic islands on a semiconductor substrate, Phys. Rev. B 75,
205315 (2007).
[88] G. D. J. Smit, S. Rogge, and T. M. Klapwijk, Scaling of nano-

Schottky-diodes, Appl. Phys. Lett. 81, 3852 (2002).
[89] R. T. Tung, Electron transport of inhomogeneous Schottky barri-

ers, Appl. Phys. Lett. 58, 2821 (1991).
[90] Y. Yin, J. Jiang, Q. Cai, and B. Cai, Scanning tunneling mi-

croscopy and in situ spectroscopy of ultra thin Ti films and nano
sized TiOx dots induced by STM, Appl. Surf. Sci. 199, 319 (2002).
[91] D. Matthey, J. G. Wang, S. Wendt, J. Matthiesen, R. Schaub, E.

Laegsgaard, B. Hammer, and F. Besenbacher, Enhanced Bonding
of Gold Nanoparticles on Oxidized TiO2 (110), Science 315, 1692
(2007).
[92] K. Mitsuhara, Y. Kitsudo, H. Matsumoto, A. Visikovskiy, M.

Takizawa, T. Nishimura, T. Akita, and Y. Kido, Electronic
charge transfer between Au nano-particles and TiO2 -terminated
SrTiO3 (001) substrate, Surf. Sci. 604, 548 (2010).
[93] N. Lopez, T. V. W. Janssens, B. S. Clausen, Y. Xu, M. Mavrikakis,

T. Bligaard, and J. K. Nørskov, On the origin of the catalytic
activity of gold nanoparticles for low-temperature CO oxidation,
J. Catal. 223, 232 (2004).
[94] B. Yoon, H. Hakkinen, U. Landman, A. S. Worz, J. M. Antonietti,

S. Abbet, K. Judai, and U. Heiz, Charging effects on bonding and
catalyzed oxidation of CO on Au8 clusters on MgO, Science 307,
403 (2005).
[95] M. F. Crommie, C. P. Lutz, and D. M. Eigler, Confinement of

electrons to quantum corrals on a metal surface, Science 262, 218
(1993).
[96] C. R. Moon, L. S. Mattos, B. K. Foster, G. Zeltzer, and H. C.

Manoharan, Quantum holographic encoding in a two-dimensional
electron gas, Nat. Nanotechnol. 4, 167 (2009).
126 Bibliography
[97] M. Kiguchi, G. Yoshikawa, S. Ikeda, and K. Saiki, Elec-

tronic properties of metal-induced gap states formed at alkali-
halide/metal interfaces, Phys. Rev. B 71, 153401 (2005).
[98] J. Robertson, High dielectric constant gate oxides for metal oxide
Si transistors, Rep. Prog. Phys. 69, 327 (2006).
[99] J. Repp, G. Meyer, and K. H. Rieder, Snell’s Law for Surface

Electrons: Refraction of an Electron Gas Imaged in Real Space,
Phys. Rev. Lett. 92, 36803 (2004).
[100] M. Pivetta, F. Patthey, M. Stengel, A. Baldereschi, and W. D.

Schneider, Local work function Moiré pattern on ultrathin ionic
films: NaCl on Ag(100), Phys. Rev. B 72, 115404 (2005).
[101] W. Hebenstreit, J. Redinger, Z. Horozova, M. Schmid, R. Pod-

loucky, and P. Varga, Atomic resolution by STM on ultra-thin
films of alkali halides: experiment and local density calculations,
Surf. Sci. 424, L321 (1999).
[102] C. Loppacher, U. Zerweck, and L. M. Eng, Kelvin probe force mi-

croscopy of alkali chloride thin films on Au(111), Nanotechnology
15, S9 (2004).
[103] K. Schouteden, P. Lievens, and C. Van Haesendonck, Fourier-

transform scanning tunneling microscopy investigation of the en-
ergy versus wave vector dispersion of electrons at the Au(111)
surface, Phys. Rev. B 79, 195409 (2009).
[104] L. Petersen, P. Laitenberger, E. Lægsgaard, and F. Besenbacher,

Screening waves from steps and defects on Cu(111) and Au(111)
imaged with STM: Contribution from bulk electrons, Phys. Rev.
B 58, 7361 (1998).
[105] A. Hotzel, G. Moos, K. Ishioka, M. Wolf, and G. Ertl, Femtosecond

electron dynamics at adsorbate-metal interfaces and the dielectric
continuum model, Appl. Phys. B 68, 615 (1999).
[106] W. Chen, V. Madhavan, T. Jamneala, and M. F. Crommie, Scan-

ning tunneling microscopy observation of an electronic superlattice
at the surface of clean gold, Phys. Rev. Lett. 80, 1469 (1998).
Bibliography 127
[107] C. Bombis, F. Ample, J. Mielke, M. Mannsberger, C. J. Vil-

lagómez, C. Roth, C. Joachim, and L. Grill, Mechanical Behavior
of Nanocrystalline NaCl Islands on Cu(111), Phys. Rev. Lett. 104,
185502 (2010).
[108] L. Gross, F. Mohn, N. Moll, P. Liljeroth, and G. Meyer, The chem-

ical structure of a molecule resolved by atomic force microscopy,
Science 325, 1110 (2009).
[109] L. Gross, F. Mohn, P. Liljeroth, J. Repp, F. J. Giessibl, and G.

Meyer, Measuring the charge state of an adatom with noncontact
atomic force microscopy, Science 324, 1428 (2009).
[110] L. Gross, Recent advances in submolecular resolution with scan-

ning probe microscopy, Nat. Chem. 3, 273 (2011).
[111] A. S. Foster, C. Barth, A. L. Shluger, and M. Reichling, Unam-

biguous interpretation of atomically resolved force microscopy im-
ages of an insulator, Phys. Rev. Lett. 86, 2373 (2001).
[112] R. Hoffmann, L. N. Kantorovich, A. Baratoff, H. J. Hug, and H. J.

Güntherodt, Sublattice identification in scanning force microscopy
on alkali halide surfaces, Phys. Rev. Lett. 92, 146103 (2004).
[113] K. Lauwaet, K. Schouteden, E. Janssens, C. Van Haesendonck,

P. Lievens, M.I. Trioni, L. Giordano, and G. Pacchioni, Surface-
reconstruction at play: resolving all atoms of an alkali halide and
modulation of NaCl/Au(111) interface state, Submitted (2012).
[114] E. V. Chulkov, V. M. Silkin, and P. M. Echenique, Image potential

states on metal surfaces: binding energies and wave functions,
Surf. Sci. 437, 330 (1999).
[115] J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized gradient

approximation made simple, Phys. Rev. Lett. 77, 3865 (1996).
[116] S. Grimme, Semiempirical GGA-type density functional con-

structed with a long-range dispersion correction, J. Comput.
Chem. 27, 1787 (2006).
[117] J. E. Inglesfield, A method of embedding, J. Phys. C: Solid State

Phys. 14, 3795 (1981).
128 Bibliography
[118] L. Limot, T. Maroutian, P. Johansson, and R. Berndt, Surface-

state Stark shift in a scanning tunneling microscope, Phys. Rev.
Lett. 91, 196801 (2003).
[119] C. L. Lin, S. M. Lu, W. B. Su, H. T. Shih, B. F. Wu, Y. D. Yao,

C. S. Chang, and T. T. Tsong, Manifestation of Work Function
Difference in High Order Gundlach Oscillation, Phys. Rev. Lett.
99, 216103 (2007).
[120] K. Oura, Surface Science: An Introduction (Springer-Verlag, Dus-

seldorf, 2003).
[121] R. Bennewitz, A. S. Foster, L. N. Kantorovich, M. Bammerlin, C.

Loppacher, S. Schär, M. Guggisberg, E. Meyer, and A. L. Shluger,
Atomically resolved edges and kinks of NaCl islands on Cu(111):
experiment and theory, Phys. Rev. B 62, 2074 (2000).
[122] F. E. Olsson, M. Persson, A. G. Borisov, J. P. Gauyacq, J. Lagoute,

and S. Fölsch, Localization of the Cu(111) surface state by single
Cu adatoms, Phys. Rev. Lett. 93, 206803 (2004).
[123] J. Repp, G. Meyer, F. E. Olsson, and M. Persson, Controlling the

charge state of individual gold adatoms, Science 305, 493 (2004).
[124] F. E. Olsson, S. Paavilainen, M. Persson, J. Repp, and G. Meyer,

Multiple charge states of Ag atoms on ultrathin NaCl films, Phys.
Rev. Lett. 98, 176803 (2007).
[125] A. T. Bell, The impact of nanoscience on heterogeneous catalysis,

Science 299, 1688 (2003).
[126] M. Di Vece, S. Palomba, and R. E. Palmer, Pinning of size-selected

gold and nickel nanoclusters on graphite, Phys. Rev. B 72, 073407
(2005).
[127] A. Lando, K. Lauwaet, and P. Lievens, Controlled nanostructuring

of a gold film covered with alkanethiol SAM by low energy cluster
implantation, Phys. Chem. Chem. Phys. 11, 1521 (2009).
[128] S. Vučković, J. Samela, K. Nordlund, and V. N. Popok, Pinning

of size-selected Co clusters on highly ordered pyrolytic graphite,
Eur. Phys. J. D 52, 107 (2009).
Bibliography 129
[129] M. Amman, R. Wilkins, E. Ben-Jacob, P. D. Maker, and R. C.

Jaklevic, Analytic solution for the current-voltage characteristic
of two mesoscopic tunnel junctions coupled in series, Phys. Rev.
B 43, 1146 (1991).
[130] O. Kitakami, H. Sato, Y. Shimada, F.Sato, and M. Tanaka, Size

effect on the crystal phase of cobalt fine particles, Phys. Rev. B
56, 13849 (1997).
[131] S. Rives, A. Catherinot, F. Dumas-Bouchiat, C. Champeaux, A.

Videcoq, and R. Ferrando, Growth of Co isolated clusters in the
gas phase: Experiment and molecular dynamics simulations, Phys.
Rev. B 77, 085407 (2008).
[132] C. P. Romero, K. Lauwaet, D. Ievlev, E. Janssens, and P. Lievens,

A new in-line electrostatic deflector for the study of deposited
clusters, To be submitted .
[133] S. Neukermans, E. Janssens, H. Tanaka, R. E. Silverans, and P.

Lievens, Element-and Size-Dependent Electron Delocalization in
AuN X+ Clusters (X= Sc, Ti, V, Cr, Mn, Fe, Co, Ni), Phys. Rev.
Lett. 90, 33401 (2003).
[134] E. Janssens, P. Gruene, G. Meijer, L. Wöste, P. Lievens, and A.

Fielicke, Argon physisorption as structural probe for endohedrally
doped silicon clusters, Phys. Rev. Lett. 99, 63401 (2007).
[135] S. Neukermans, E. Janssens, Z. F. Chen, R. E. Silverans, P. R.

Schleyer, and P. Lievens, Extremely Stable Metal-Encapsulated
AlPb+ +
10 and AlPb12 Clusters: Mass-Spectrometric Discovery and
Density Functional Theory Study, Phys. Rev. Lett. 92, 163401
(2004).
[136] K. Schouteden, K. Lauwaet, D. A. Muzychenko, P. Lievens, and

C. Van Haesendonck, Spin-dependent electronic structure of self-
organized Co nanomagnets, New J. Phys. 13, 033030 (2011).
[137] V. Kumar and Y. Kawazoe, Magic behavior of Si15 M and Si16 M

(M= Cr, Mo, and W) clusters, Phys. Rev. B 65, 073404 (2002).
130 Bibliography
Publications
• Controlled nanostructuring of a gold film covered with alkanethiol

SAM by low energy cluster implantation,
A. Lando , K. Lauwaet, P. Lievens,
Phys. Chem. Chem. Phys. 11, 1521 (2009).
• Creation of stable TiOx nanoparticles on Au(111),

K. Lauwaet, K. Schouteden, E. Janssens, C. Van Haesendonck,
and P. Lievens,
Phys. Rev. B 83, 155433 (2011).
• Spin-resolved electronic structure of self-organized Co Nanomag-

nets
K. Schoudeten, K. Lauwaet, D.A. Muzychenko, P. Lievens, and C.
Van Haesendonck,
New J. Phys. 13, 033030 (2011).
• Atomic Scale Dynamics of Ultra-Small Ge Clusters,

S. Bals, S. Van Aert, B. Schoeters, B. Partoens, E. Yücelen, C.
Romero, K. Lauwaet, M.J. Van Bael, P. Lievens, and G. Van Ten-
derloo,
Submitted (2011).
• Surface-reconstruction at play: resolving all atoms of an alkali

halide and modulation of the NaCl/Au(111) quantum well state,
K. Lauwaet, K. Schouteden, E. Janssens, C. Van Haesendonck, P.
Lievens, M.I. Trioni, L. Giordano, and G. Pacchioni,
Submitted (2012).
• Tuning the NaCl/Au(111) interface state by the NaCl layer thick-

ness,
131
132 Publications
and P. Lievens,
To be submitted (2012).
• Band structure quantization in deposited ZnO nanoclusters,

K. Schoudeden, Y.J. Zeng, K. Lauwaet, C.P. Romero, P. Lievens,
and C. Van Haesendonck,
In Preparation (2012).
• Imaging Co nanoclusters deposited on thin NaCl films with atomic

resolution and 3D reconstruction,
and P. Lievens,
In preparation (2012).
• A new in-line electrostatic deector for mass selected deposition of

clusters,
C.P. Romero, K. Lauwaet, D. Ievlev, E. Janssens, and P. Lievens,
In preparation (2012).
Curriculum Vitae
Koen Kauwaet
Geboren op 19 juni 1985 te Lommel
1999-2003 Technisch Secundair Onderwijs, Industiële Wetenschappen,

Technisch Instituut Don Bosco Helchteren
2003-2005 Kandidaat in de Natuurkunde, L.U.C. Diepenbeek
2005-2007 Licentiaat in de Natuurkunde, K.U. Leuven, met proef-

schrift “Metaalclusters gedeponeerd op fenyl- en n-alkaanthiol zelf-
geassembleerde monolagen”
2007-2012 Doctoraat in de Natuurkunde, K.U.Leuven, met Docto-

raatsproefschrift “Scanning tunneling microscopy and spectroscopy
on surfaces, interfaces, and atomic clusters”
2012-2013 Postdoctoraal medewerker aan het “Institutos Madrileños

de Estudios Avanzados (IMDEA), Nanociencia”
133

2012 Thesis Koen Lauwaet

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2012 Thesis Koen Lauwaet

Uploaded by

Copyright:

Available Formats



     

2-DEG Two Dimensional Electron Gas

CDA Cluster Deposition Apparatus

DFT Density Functional Theory

DOS Density of States

fcc face centered cubic

hcp hexagonal close packed

LDOS Local Density of States

SFM Scanning Force Microscopy

SPM Scanning Probe Microscopy

STM Scanning Tunneling Microscopy

UHV Ultra-High Vacuum

The nanoscale: Where size does matter 5

2 Exploring the nanoworld with a scanning tunneling mi-

3 Realization of systems with reduced dimensionality 39

3.2 Depositing preformed clusters in UHV . . . . . . . . . . . 43

4 TiOx nanoclusters on Au(111) 49

5 Tuning the NaCl/Au(111) interface state 63

7 Co atoms on and F-centers in NaCl/Au(111) 91

8 Imaging Co nanoclusters with atomic resolution 99

Nederlandse Samenvatting 113

Curriculum vitae 133

optimize the activity and specificity of catalysts [2,3].

Figure 1. Schematic representation of a cubo octahedral metal

The second class of effects is due to the electronic energy levels in

Figure 2. (a) Schematic illustration of the density of states in

realized by depositing nanoparticles on the metal surface, whereas the

dividual nanosized objects. Real-space information at the atomic scale

Au(111). Deposition of NaCl on Au(111) at room temperature leads

In this era of miniaturization of electronic devices there is an increas-

However, one does not need to go to very special geometries to

After this general introduction the remainder of the chapter deals

1.1 Confinement of particles

1.1.1 The infinite square well

From the boundary conditions it follows that ψ(0) = ψ(L) = 0 [5].

1.1.2 Particles trapped in three dimensions

where ∇2 = ∂ 2 /∂x2 +∂ 2 /∂y 2 +∂ 2 /∂z 2 and ~r is a vector with components

~2 (kn2 x + kn2 y + kn2 z )

1.2 The nearly-free electron gas inside metals

Figure 1.2. Metal as a 3D box filled with non-interacting elec-

each with its own energy. An electron belonging to a particular wave

volves small energies when looking at single electrons.

Figure 1.3. Density of single-particle states as a function of en-

1.2.2 Nearly-free electron model

U (~r + ~a) = U (~r) (1.11)

for all Bravais lattice vectors ~a.

u~k (~r + ~a) = u~k (~r) . (1.13)

Electrons fulfilling these requirements are called Bloch electrons, in con-

1.3 Electrons at metal surfaces

h2 n2z ~2 (kn2 x + kn2 y )

where ~k denotes the Bloch wave vector and ~r is a spatial coordinate.

Figure 1.4. One dimensional semi-infinite lattice model poten-

E(k|| ) = E0∗ + ~2 k||2 /2m∗ . (1.17)

Here m∗ is the effective electron mass, typically given in units of the

Density of states in two dimensions

noble metal E0∗ (meV)) m∗ /me

Table 1.1. Values of the surface state of Cu(111), Ag(111), and

Figure 1.5. Array of 28 Mn atoms forming a rectangle of size

Electron transport through nanoparticles

nanoparticle. To achieve this the time it takes an electron to tunnel to

Figure 1.6. Energy band diagrams of a 0D system coupled to

Exploring the nanoworld

i.e., STS. How this is done is the topic of section 2.2.