the formation and characterisation of aperiodic ultra-thin films on the surfaces of quasicrystals
TRANSCRIPT
The formation and
characterisation of aperiodic
ultra-thin films on the surfaces
of quasicrystals
Thesis submitted in accordance with the requirements of the
University of Liverpool for the degree of Doctor in Philosophy
by
Joseph A. Smerdon
1
Contents
Abstract v
1 Introduction 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Chapter breakdown . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Quasicrystals 42.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1.1 Crystallography . . . . . . . . . . . . . . . . . . . . . . 42.1.2 Diffraction and the discovery of quasicrystals . . . . . . 5
2.2 Aperiodic order . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2.1 The Fibonacci sequence . . . . . . . . . . . . . . . . . 72.2.2 The golden ratio τ and the Penrose tiling . . . . . . . . 102.2.3 The Penrose tiling . . . . . . . . . . . . . . . . . . . . 10
2.3 Bulk structure of decagonal Al72Ni11Co17 . . . . . . . . . . . . 132.4 Bulk structure of icosahedral quasicrystals . . . . . . . . . . . 16
2.4.1 The icosahedral glass model . . . . . . . . . . . . . . . 162.4.2 The icosahedral quasicrystal model . . . . . . . . . . . 172.4.3 Consensus on physical structure . . . . . . . . . . . . . 18
2.5 Quasicrystalline surfaces . . . . . . . . . . . . . . . . . . . . . 192.5.1 Five-fold surface of i -Al70Pd21Mn9 . . . . . . . . . . . 192.5.2 Ten-fold surface of d -Al72Ni11Co17 . . . . . . . . . . . . 23
2.6 Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.6.1 The atom . . . . . . . . . . . . . . . . . . . . . . . . . 232.6.2 The solid . . . . . . . . . . . . . . . . . . . . . . . . . 252.6.3 Magnetic properties of quasicrystals . . . . . . . . . . . 25
3 Thin film deposition 273.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.2 Growth modes . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.3 Homogeneous and heterogeneous systems . . . . . . . . . . . . 30
i
3.3.1 Homogeneous systems . . . . . . . . . . . . . . . . . . 313.3.2 Heterogeneous systems . . . . . . . . . . . . . . . . . . 31
3.4 Pseudomorphism . . . . . . . . . . . . . . . . . . . . . . . . . 323.4.1 Single-species pseudomorphism . . . . . . . . . . . . . 353.4.2 Multi-layer pseudomorphism . . . . . . . . . . . . . . . 38
4 Experimental Methods 394.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394.2 Ultra High Vacuum (UHV) . . . . . . . . . . . . . . . . . . . 39
4.2.1 Vacuum chamber . . . . . . . . . . . . . . . . . . . . . 394.2.2 Sample cleaning process . . . . . . . . . . . . . . . . . 41
4.3 Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.3.1 Auger electron spectroscopy (AES) . . . . . . . . . . . 424.3.2 Medium energy ion scattering (MEIS) . . . . . . . . . 454.3.3 Scanning tunneling microscopy (STM) . . . . . . . . . 514.3.4 X-ray Magnetic Circular Dichroism (XMCD) . . . . . . 594.3.5 Beamline ID8 at the European Synchrotron Radiation
Facility . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.3.6 Low-Energy Electron Diffraction . . . . . . . . . . . . . 634.3.7 Thin film deposition . . . . . . . . . . . . . . . . . . . 68
5 Characterisation of an ultrathin Cu film formed on the five-fold surface of i -Al70Pd21Mn9 using medium-energy ion scat-tering spectroscopy 715.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 715.2 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . 725.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3.1 Two-dimensional data . . . . . . . . . . . . . . . . . . 735.3.2 One-dimensional data . . . . . . . . . . . . . . . . . . 75
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 785.4.1 The unannealed Cu film . . . . . . . . . . . . . . . . . 785.4.2 The annealed Cu film . . . . . . . . . . . . . . . . . . . 86
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6 Adsorption of cobalt on the ten-fold surface of d-Al72Ni11Co17
and on the five-fold surface of i -Al70Pd21Mn9 896.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.1.1 Practical magnetism . . . . . . . . . . . . . . . . . . . 896.1.2 Physical motivation . . . . . . . . . . . . . . . . . . . . 90
6.2 Experimental Details . . . . . . . . . . . . . . . . . . . . . . . 916.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
ii
6.3.1 Scanning tunneling microscopy and Auger electron spec-troscopy . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.3.2 Low-energy electron diffraction . . . . . . . . . . . . . 946.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 966.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
7 Surface magnetism of quasicrystals and thin films depositedthereon 997.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 997.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
7.2.1 Icosahedral Al70Pd21Mn9 . . . . . . . . . . . . . . . . . 1007.2.2 Magnetism and ordering: XMCD . . . . . . . . . . . . 1017.2.3 Easy and hard magnetization directions: XMCD . . . . 102
7.3 d -Al72Ni11Co17 . . . . . . . . . . . . . . . . . . . . . . . . . . 1027.3.1 Preparation conditions . . . . . . . . . . . . . . . . . . 1027.3.2 Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . 106
7.4 Deposited Fe film on d -Al72Ni11Co17 . . . . . . . . . . . . . . 1087.4.1 Hysteresis curves . . . . . . . . . . . . . . . . . . . . . 108
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8 Summary and suggestions for further work 113
List of Publications 116
Presentations 125
Bibliography 126
iii
Acknowledgments
The preparation of this thesis has been a long and difficult task, but has beengreatly eased by the excellent support I have had from the start, not leastfrom Dr Julian Ledieu. From the tellings off I’ve received for missing onetoo many Thursday mornings (Wednesday is karaoke night at the AugustusJohn) to the constant encouragement through to the small hours of the longexperimental runs, he has been a considerable help, though I may not alwayshave appreciated it at the time.
The folks at the Surface Science Research Centre in Liverpool have madethis journey a very enjoyable one – from the very beginning of my time hereas an undergraduate project student all the way through to my final year asa piece of the furniture, I have been made to feel welcome and important allof the time. I have made some very good friends through this place, and Iknow I’ll be friends with them for some considerable time to come – I’ve seenpeople get married and have children and move countries and careers in myshort time here, and I’ve been to visit them, and will again (when I get ajob).
The group has been a pleasure to work with; I thank Ronan McGrath forinviting me to become a student, and for allowing me to pursue the projectunder his supervision. I’m grateful for the many opportunities he has offeredme, such as trips in the States, where I made many other good friends, notleast Professor Renee Diehl, who has also been a constant help throughoutthe process, offering advice and guidance in difficult times.
But most of all, I have to thank my parents, for constantly believing inmy ability to do this, and that it’s the right thing for me to do. I thank themfor the all the help they have given and generosity they have showed, and Ihope I will ultimately be able to repay them, though I don’t know how.
Ta Mum. Ta Dad.
Joseph Anthony Smerdon 2006
iv
Abstract
This thesis is a report of work carried out in the Surface Science ResearchCentre of the University of Liverpool, the Medium Energy Ion Scatteringfacility in Daresbury, and the X-ray Magnetic Circular Dichroism beamline(ID8) at the European Synchrotron Radiation Facility in Grenoble. Thethree-year project has resulted in the successful identification of elementalcandidates for the formation of single-element quasiperiodic thin films on thesurfaces of quasicrystals, and also an elucidation of the local atomic structureof one of these films. The basic structural characteristics of the films seem tohold for all systems studied. Further to this, a study to discover the magneticproperties of selected quasicrystals and elemental films deposited thereon hasbeen undertaken, and certain results are reported.
The elucidation of the local atomic structure of a pseudomorphic film ofCu deposited on the five-fold surface of i -Al70Pd21Mn9 using medium-energyion scattering spectroscopy (MEIS) is reported. Monte Carlo calculations,using the VEGAS code, have been utilised to simulate the blocking of 100keV He+ ions scattered from the overlayer. The coordinates of the Cu atomsin the overlayer derived from this procedure are consistent with a structurecomposed of islands of material occurring in five rotational domains orientedalong high symmetry directions on the quasicrystal surface. Each islandconsists of nanoscale rows of cubic material arranged according to a one-dimensional Fibonacci sequence with long and short distances related by thegolden mean τ . Upon annealing the film transforms to an alloyed structurecomposed of five orientational domains of fcc material with the (110) axisperpendicular to the surface.
The adsorption behaviour of Co on the ten-fold surface of d -Al72Ni11Co17
and on the five-fold surface of i -Al70Pd21Mn9 has been studied using scanningtunneling microscopy (STM), Auger electron spectroscopy (AES) and low-energy electron diffraction (LEED). The analysis of a distinctive quasiperi-odic LEED pattern for coverages from 3 - 30 ML of Co on d -Al72Ni11Co17
suggests that the Co forms in a pseudomorphic row structure composed of do-mains of Fibonacci spaced rows having a periodic lattice parameter along therows of 2.5 ± 0.1 A. The same structure, though less well-ordered and with alarger lattice parameter, is formed on the five-fold surface of i -Al70Pd21Mn9.
Some discussion is made of results from a study to probe the magneticproperties of quasicrystals in different stages of preparation, and inducedinterface magnetism of d -Al72Ni11Co17 is observed for the first time.
v
Glossary of Acronyms
LEED - Low Energy Electron DiffractionSTM - Scanning Tunnelling MicroscopyAES - Auger Electron SpectroscopyXPS - X-ray Photoemission SpectroscopyUHV - Ultra High VacuumFFT - Fast Fourier TransformXMCD - X-ray Magnetic Circular DichroismMEIS - Medium Energy Ion Scattering spectroscopy
vi
Chapter 1
Introduction
Madam, of what use is a baby?
Michael Faraday’s famous response to the Queen’s enquiries regarding
the purpose of basic electrical and magnetic studies.
Or, to put it another way:
Physics is like sex; sure, it has some practical value, but that’s
not why we do it.
Richard Feynman, in a lecture.
1.1 Motivation
Since the first applications of the scientific method and the development
of the philosophy of science, it has always been the unexpected discoveries
that have led to the greatest scientific breakthroughs – for example, consider
Michelson and Morley’s experiment to determine the speed of the Earth
relative to the aether that was expected to be the medium for electromagnetic
waves [1]. The unequivocal proof that no such thing as the aether existed
eventually led to the proposal of the theory of special relativity [2].
The discovery of quasicrystals in 1984 by Dan Shechtman has not (yet)
led to anything quite as grand as a new theory of mass, space and time, but
has certainly been extremely effective in shaking the world of crystallography
to its core. Everybody knows that you can tile a floor with square tiles, or
1
graph paper with squares, or fill a room to its limit with cubic or cuboid,
or even triangular- or hexagonal-prism-shaped boxes, and this appeared to
be the route by which crystal packing may be ultimately understood. Roger
Penrose dabbled in the 70s with his ‘Penrose tilings’ that could be used to tile
a plane (or floor) in such a fashion that the tiling could be rotated five times
into itself, but these were mathematical games – even the title of his paper [3]
cited ‘aesthetics’ as the reason for pursuing such geometrical diversions.
Quasicrystals are binary, tertiary or quaternary alloys that form in ex-
tremely well-ordered, yet aperiodic structures. Studies of the ordering within
these systems indicate that although they can, and often do, exhibit ordering
as perfect as that found within conventional crystals, the order of rotational
symmetry present precludes the existence of a unit cell.
Quasicrystals exhibit five-, eight-, ten- or twelve-fold rotational symme-
try along one or more crystal directions. The surface structure of five-fold
and ten-fold planes may be understood in terms of the aesthetically pleasing
geometrical diversions of Penrose, and the icosahedral bulk structure of qua-
sicrystals exhibiting five-fold planes may be considered to be analogous to a
three-dimensional Penrose tiling.
The mechanisms and chemical interactions causing these exotic materials
to form in such unusual structures are somewhat mysterious – we can work
out mathematically that the icosahedral structure is a projection of a periodic
six-dimensional structure, but how could individual atoms and clusters of
atoms be expected to know that? An energy-minimisation curve dictates the
way the solid can condense out of the melt, and although icosahedral clusters
are common in molten alloys, how, in these special cases, can they know to
ultimately coalesce at the vertices of larger icosahedra, and these in turn at
the vertices of still larger icosahedra?
There is a need to know, and the motivation for this work lies in simpli-
fying quasiperiodic systems as far as possible in order to perhaps enable the
ultimate discovery of all of these mechanisms.
2
1.2 Chapter breakdown
Chapter 2 deals with the basics of (quasi-)crystallography, and the methods
involved in the discovery of quasicrystals. There is also an introduction to the
mathematics of aperiodic ordering, and some details of the two quasicrystals
discussed in this thesis, in terms of their bulk and surface structures.
Chapter 3 summarises the work carried out, so far, on deposition of el-
emental overlayers on quasicrystals. This work encompasses the efforts to-
wards the creation of a single-element quasicrystal.
Chapter 4 introduces the experimental techniques employed in this thesis
work. An introduction to vacuum technology is provided, as is some expla-
nation of the capabilities, and limitations, of the various techniques used,
such as diffraction (LEED), microscopy (STM), surface composition analysis
(AES), backscattering (MEIS) and a probe of magnetic properties (XMCD).
Chapter 5 represents the first time the local atomic structure of a pseudo-
morphic single-element film deposited on a quasicrystal has been determined.
Through a combination of techniques, culminating in this adaptation of a
Monte Carlo technique to model a MEIS study of a copper film deposited on
the five-fold surface of i -Al70Pd21Mn9, the three-dimensional atomic struc-
ture is solved.
Chapter 6 is an analytical chapter discussing the behaviour of cobalt
atoms deposited on the five- and ten-fold faces of an icosahedral and a
decagonal quasicrystal. The techniques used to discover the structure of
the overlayer formed are LEED and STM. Although the magnetic proper-
ties are not probed in this chapter, this study constitutes the first time a
magnetic pseudomorphic overlayer has been deposited on a quasicrystalline
substrate.
Chapter 7 deals with some aspects of the magnetic behaviour of qua-
sicrystals. Though not as analytical as the previous chapters, it presents the
first evidence obtained for induced interface magnetism in Co at the surface
of d -Al72Ni11Co17 with a thin magnetic film deposited thereon.
Chapter 8 provides a brief summary of the thesis, and indicates possible
directions for future work.
3
Chapter 2
Quasicrystals
2.1 Introduction
2.1.1 Crystallography
Periodic structures
Given the task of differentiating between different types of solid (‘condensed
matter’), the first and most easily made distinction is that between materials
in which long-range order is present and materials in which long-range order
is absent. Materials in which order is not present are known as ‘glasses’
or amorphous materials. The simplest type of ordering possible in solid
materials is that in which sets of atomic positions (collectively, the unit cell)
are related to each other by an expression of the form
ax+by+cz
where x, y and z are integer and a, b and c are unit vectors. These
unit vectors define a lattice, which is infinite in three-dimensional space and
isotropic i.e. every point in the lattice has an environment identical to that
of every other point. The other element needed to define a crystal is a basis,
which is an atom or group of atoms that is placed at every point on the
4
Figure 2.1: Only 2-, 3-, 4- and 6-fold symmetry can tile a plane.
lattice. The unit vectors a, b and c and the basis comprise the unit cell,
which defines the structure.
This definition of a crystal carries within it the stipulation that only
rotational symmetry based on 60◦ or 90◦ intervals is permitted i.e. 2-, 3-,
4- or 6-fold symmetry. This is illustrated in figure 2.1. It also implies the
existence of just fourteen types of lattice (called Bravais lattices) and 230
point groups.
2.1.2 Diffraction and the discovery of quasicrystals
Crystal structure may be found by diffraction techniques. A beam of ra-
diation with known wavelength incident on an ordered structure will be
diffracted according to the equation
nλ = 2d sin θ (2.1)
where λ is the wavelength and n = (h2 + k2 + l2)1/2
i.e. beams of diffracted radiation emerge at well-defined angles. A diffrac-
5
Figure 2.2: Shechtman’s logbook, from 1982, documenting the first discoveryof a quasicrystalline phase in the Al75Mn25 alloy. The term SAD refers toselected area diffraction, with the acronym DF referring to the use of thedark field mode of the technique, where one diffraction beam is used in theproduction of a transmission electron micrograph.
tion pattern has the same order of rotational symmetry as the structure it
results from, revealing instant information about the type of lattice giving
rise to the diffraction.
In 1982, working on sabbatical at the National Bureau of Standards in
Washington D.C., Dan Shechtman was using electron diffraction to identify
the structures of different phases in the AlMn system. These were formed
using a fast-quenching technique in which a melt is sprayed onto a spinning
wheel, resulting in cooling at the rate of 105 Ks−1 and the formation of ther-
modynamically metastable structures. Out of the many alloys he studied, he
found one that had a ten-fold diffraction pattern; surprising, considering this
is forbidden by the rules of crystallography referred to above. The relevant
page from the logbook is reproduced in figure 2.2.
He had discovered what eventually became known as quasicrystals, short
for quasi-periodic crystals. Two years later, he published his results [4] and
6
caused much controversy in the world of crystallography. The well known
chemist Linus Pauling dismissed his findings as caused by multiply-twinned
cubic crystals [5], but as the field became established and as more quasicrys-
tals — among them thermodynamically stable species — were discovered,
their reality became indisputable.
2.2 Aperiodic order
Since, in order to have a diffraction pattern at all, quasicrystals must exhibit
an extremely high degree of ordering, and the observation of forbidden orders
of rotational symmetry proves that the ordering cannot be periodic, they
must possess ordering that is aperiodic.
2.2.1 The Fibonacci sequence
The simplest kind of aperiodic ordering takes the form of a one-dimensional
sequence in which the order of the components follows rigid rules and yet
never repeats. One such sequence is the Fibonacci sequence, given by the
rule ‘the next term is given by the sum of the previous two terms’ [6]. This
results in the following sequence:
1
1
2
3
5
8
13
21
...
7
or, if two objects denoted L and S constitute terms in the sequence, then
its form is:
S
L
LS
LSL
LSLLS
LSLLSLSL
LSLLSLSLLSLLS...
This sequence fulfils the criteria for aperiodic order; it is the result of two
simple rules (L → LS ; S → L), and extends to infinity without repeating.
Any part of any term, of any length, may be found in an infinite term an
infinite number of times, but as a whole the sequence never repeats.
This sequence may also be generated in a way that at once may seem sim-
pler and more complicated, called the ‘cut-and-project’ method, illustrated
in figure 2.3. If a slice is taken through a periodic grid at an irrational angle,
it intersects the grid in a quasiperiodic fashion, in this case resulting in a
one-dimensional Fibonacci sequence. This is easily understandable in terms
of the fact that, by definition, an irrational number may not be expressed as a
fraction; therefore, a line originating from one set of integer coordinates with
an irrational gradient will never intersect another set of integer coordinates.
As discussed later in this chapter, an icosahedral quasicrystal structure may
analogously be considered to be a decoration of a three dimensional projec-
tion of a six-dimensional hypercubic lattice [7].
8
Figure 2.3: (a) The project method. (b) The cut method. In the six-dimensional analogue to this diagram, the ‘lines of finite length’ are replacedby two-dimensional ‘atomic surfaces’, and the way the three-dimensional pro-jection cuts them reveals atomic positions in the bulk structure.
9
Figure 2.4: In a pentagram, the ratio of the areas a/b is equal to τ .
2.2.2 The golden ratio τ and the Penrose tiling
The golden ratio is a solution to the equation:
x2 − x− 1 = 0 (2.2)
where x is equal to√
5+12
= (τ) and 1−√
52
= (1/τ).
It is also the ratio of L to S objects in a term in the Fibonacci sequence
as n→∞ and the value to which the ratio of consecutive Fibonacci numbers
converges as well as the inherent ratio of areas and line lengths formed in
a pentagram (figure 2.4). It also appears in nature in, for example, the
patterns of seeds observed on flowerheads and pinecones (figure 2.5), in which
the numbers of clockwise and anticlockwise spirals are consecutive terms in
the Fibonacci sequence, and so their ratio approximates to τ . Artists and
architects have been aware of the aesthetic value of τ for thousands of years
(see figure 2.6), and some of Mozart’s sonatas are divided according to the
golden section also.
2.2.3 The Penrose tiling
Roger Penrose is the Emeritus Rouse Ball Professor of Mathematics at the
University of Oxford. He is responsible for some extremely important de-
10
Figure 2.5: The daisy and pinecone are examples of plants that exhibit Fi-bonacci ordering and τ -scaling.
Figure 2.6: The Parthenon is depicted, with τ -scaling relationships high-lighted. Inset: τ and the human body.
11
Figure 2.7: The Penrose P1 tiling. The four constituent tiles are the boat,star, pentagon and rhomb tiles.
velopments in mathematical physics, particularly in general relativity and
cosmology, and holds some controversial philosophical views, for example on
the quantum mechanical nature of human thought. However, in relation
to this thesis, his most important work is on what have become known as
Penrose tilings.
Penrose tilings are a two-dimensional manifestation of the most impor-
tant aspect of the Fibonacci sequence: aperiodicity. The Penrose P1 tiling
(reprinted from ref [3] in figure 2.7) is the first tiling Penrose discovered, in
1974, that could be used to tile a plane with five-fold symmetry. As discussed
earlier, it is not possible to tile a plane in a five-fold fashion with a single
shape of tile; what Penrose discovered were some simple ways to tile a plane
in a five-fold fashion with a relatively small number of tiles. His most famous
tiling, the Penrose P3 tiling, consists of two rhombic tiles, the so-called fat
and skinny rhombs (shown in figure 2.8), with interior angles of 108◦ and
72◦ and 144◦ and 36◦ respectively. The ratio of the areas of these rhombs is
12
equal to τ . These rhombs may be tiled together to form either a periodic or
an aperiodic tiling; to enforce aperiodicity matching rules must be applied.
These are shown in the figure as arrows on the sides of the rhombs. The
matching rules in this case are equivalent to saying that no two rhombs may
come together to form a parallelogram. In an aperiodic tiling, the ratio of
numbers of fat rhombs to skinny rhombs is also equal to τ .
2.3 Bulk structure of decagonal Al72Ni11Co17
Only icosahedral quasicrystals have three-dimensional aperiodic structure
and stacking; other kinds of quasicrystal that exhibit eight- [8–10], ten- [11–
13] and twelve-fold [14, 15] rotational symmetry do so along only one axis
and are periodically stacked perpendicular to that axis. The structure of
decagonal Al-Ni-Co has been well-studied [16–26], and many bulk models
have been proposed. These fall into two broad categories: the tiling models
and the quasi-unit cell models. The tiling models largely follow a decoration
of tiles based on the Penrose P1 tiling (the four-component tiling depicted
in figure 2.7) and suffer from a lack of consistency in the prediction of the
density of the 20 A decagonal clusters that are visible in images of the surfaces
of this family of quasicrystals.
The most promising candidate for the structure of the decagonal Al72Ni11Co17
quasicrystal is based on a quasi-unit cell. This kind of cluster, two competing
candidates for which are depicted in figure 2.9, overlaps in a specific fashion
to cover a plane. A completed plane consists of two five-fold planes closely
stacked and related by inversion symmetry to provide the ten-fold symmetry.
13
Figure 2.8: The Penrose P3 tiling and the rhombs that make it up, withmatching rules shown.
14
Figure 2.9: Two competing models for the atomic decoration of the decagonal(2 nm) quasi-unit-cell for Al72Ni20Co8: (a); a model with broken ten-foldsymmetry and (b); an alternative model with unbroken ten-fold symmetrybut with accidental symmetry breaking introduced in the central region dueto chemical and occupational (vacancy) disordering. Reprinted from ref. [26].Copyright APS 2000.
15
Z
X
Y
5
2
6 3
4
1
Figure 2.10: Schematic of an icosahedron, showing the six five-fold symmetryaxes.
2.4 Bulk structure of icosahedral quasicrys-
tals
Icosahedral quasicrystals are structurally complex, in that they are fully
quasiperiodic in three dimensions. Like an icosahedron, the quasicrystal has
31 high symmetry axes:
ten 3-fold axes
six 5-fold axes
fifteen 2-fold axes
2.4.1 The icosahedral glass model
The first model proposed for the bulk structure of icosahedral quasicrys-
tal phases was the icosahedral glass model proposed by Shechtman et al. in
16
1985 [27]. This model consists of a random packing of orientationally ordered
icosahedral clusters, and has been shown to give diffraction peaks consistent
with that observed from the earliest icosahedral phases. However, the peaks
produced had an inherent width due to the large degree of disorder present
in the structure, and this was inconsistent with the extremely sharp diffrac-
tion patterns obtained from higher quality icosahedral quasicrystal samples
produced subsequently with more refined techniques.
2.4.2 The icosahedral quasicrystal model
The icosahedral quasicrystal model is a highly ordered model based on a
three-dimensional aperiodic icosahedral tiling.
Basic concepts
To fully define a pentagon, four unit vectors from the centre to the vertices
are required. Thus to uniquely define a point in five-fold symmetrical two-
dimensional space, a four-dimensional coordinate system is necessary. From
this it follows that the cut-and-project method introduced previously may
be used to produce the Penrose tiling; a particular two-dimensional section
through a four-dimensional periodic lattice will result in a set of vertices
consistent with the Penrose tiling.
Solving the structure of icosahedral quasicrystals also requires the cut-
and-project method. As mentioned earlier, an icosahedron has six five-fold
axes, by which it may be fully defined. In the same way that a five-fold two-
dimensional tiling may be defined by a periodic lattice in four-dimensional
space, a periodic lattice in six-dimensional space may be projected into three
dimensions to give an aperiodic lattice with icosahedral symmetry. A con-
sideration of higher-dimensional periodic space is always required to fully
describe any aperiodic structure.
Three dimensional diffraction patterns are modeled using the 6-D struc-
ture, with the atomic structure contained in the windows or ‘atomic surfaces’,
which are the ‘projection strips’. Different models contain different geometry
17
B C DA
Figure 2.11: A Mackay icosahedral cluster (51 atoms). The cluster is definedby a central body-centred cube (9 atoms, radius 2.57 A), or a partially oc-cupied dodecahedral shell, followed by an icosahedral shell (12 atoms, radius4.56 A), and finally an external icosidodecahedral shell (30 atoms, radius4.80 A).
windows, with different atomic content.
2.4.3 Consensus on physical structure
Weeks after the discovery of the first quasicrystal was reported, Levine and
Steinhardt proposed a structural model for an icosahedral quasicrystal by ex-
trapolating the 2D Penrose tiling to 3D by extending the rhombs to become
rhombohedrons [28]. The 3D tiling exhibits a long-range icosahedral symme-
try, and is identical to that obtained using the cut-and-project method. It
still forms the basis for models of icosahedral quasicrystals today, in which
the tiling is decorated by clusters of atoms.
The i -Al70Pd21Mn9 quasicrystal is periodic in six-dimensional space. The
Fm35 space group characterises the structure in 6D, which may be considered
to be a face-centred hypercubic lattice [29]. A neutron and X-ray diffraction
study has resulted in a generally accepted model for the 3-dimensional physi-
cal structure [30]. It is found that the structure is best described as atomically
dense planes which are slightly corrugated and stacked in a quasiperiodic way
along the five-fold axis of the quasicrystal [31].
One way to consider the structure of an icosahedral quasicrystal is that it
is based on a hierarchy of pseudo-Mackay icosahedra (PMI), in which most
18
+ +
B AC
Figure 2.12: A Bergman cluster (33 atoms). The cluster is defined by acentral atom, an icosahedral shell (12 atoms) and an outer dodecahedralshell (20 atoms).
of the vertices of a Mackay cluster are decorated by single atoms. These
PMIs then decorate the vertices of a larger (τ -cubed factor) PMI, and so on
upwards. There are essentially two types of PMI; one is a regular Mackay
icosahedron with Mn atoms on the external icosahedron vertices and Al else-
where (this occurs also with some substitution of Mn atoms with Pd atoms),
and also one containing about 20 Pd atoms and the rest Al. Interconnecting
space is filled with pieces of PMI [32, 33]. Truncated Bergman clusters are
formed at the intersection points of PMI. A Bergman cluster is depicted in
figure 2.12; a truncated Bergman cluster is a cluster without the external
dodecahedral shell. These clusters, and the cluster model in general, are dis-
cussed in greater detail by Gratias et al. [33]. This kind of model is equivalent
to the aforementioned three-dimensional Penrose tiling, with clusters deco-
rating rhombohedrons within the tiling. The most-studied surface to date is
the one formed by cutting the quasicrystal perpendicular to a five-fold axis.
2.5 Quasicrystalline surfaces
2.5.1 Five-fold surface of i -Al70Pd21Mn9
If an i -Al-Pd-Mn quasicrystal is cleaved under UHV along a five-fold plane,
a surface with an approximate roughness of 1 nm is obtained, upon which
19
Figure 2.13: Scanning tunneling microscopy of a UHV-cleaved i -Al70Pd21Mn9
surface. The clusters are clearly visible. Reprinted from ref. [34]. CopyrightAPS 1996.
clusters of a particular size are visible. This is due to the crack propagation
being deflected around these harder Bergman clusters [34].
Dark five-fold hollows
An Al-Pd-Mn five-fold surface prepared by the sputter-anneal technique,
such as those presented in this thesis, exhibits a flat terraced surface covered
with atomic-scale features, as shown in figure 2.14. The most recognisable
of these features are the dark five-fold stars, which always occur with a
similar size and orientation, and have positions relatable to each other via a
Fibonacci pentagrid and τ -scaling. The size of these stars has been shown
to be consistent with that of a Bergman cluster, and the stars are generally
accepted to be Bergman clusters that have been bisected across a particular
plane [35].
20
5 nm
Figure 2.14: Scanning tunneling microscopy of a flat i -Al70Pd21Mn9 sur-face. Many dark five-fold hollows may be observed. Reprinted from ref. [35].Copyright Elsevier 2001.
21
Figure 2.15: An STM image of a dark five-fold hollow, with LDOS (see 4.3.3)plotted on the z-scale to provide a 3D image. The distances indicated areconsistent with the size of a truncated Bergman cluster. Reprinted fromref. [35]. Copyright Elsevier 2001.
22
2.5.2 Ten-fold surface of d-Al72Ni11Co17
The ten-fold surface of Al-Ni-Co may also be understood in terms of a Penrose
tiling model. The surface may be considered to be composed of overlapping
essentially two-dimensional decagonal clusters, as previously mentioned in
Section 2.3. The clusters overlap according to specific rules, as in the match-
ing rules imposed on Penrose tilings, and when these rules are followed, a
Penrose tiling may be consistently superimposed on the surface. The surface
forms with a hierarchy of length scales related by τ . One atomic layer is
formed by two closely spaced layers, each five-fold and related by inversion
symmetry. In this way the ten-fold surface is produced.
2.6 Magnetism
2.6.1 The atom
The classical picture of the atom involves a positively charged nucleus com-
posed of protons and neutrons surrounded by shells of electrons, with each
innermost shell completed before the next starts to fill. The quantum me-
chanical nature of an electron in one of these shells can be described as a
wavefunction carrying four different properties: charge, mass, angular mo-
mentum and spin. To the wavefunction we assign four quantum numbers
(n, l, ml and ms), representing the principal, the angular momentum, the
magnetic and the spin quantum numbers respectively. Two electrons are for-
bidden to carry the same set of quantum numbers; this stipulation is known
as the Pauli exclusion principle, and implies that the relative direction of
two interacting spins cannot be changed without changing the spatial dis-
tribution of charge. This change in the Coulomb electrostatic energy of the
whole system is equivalent to the existence of a direct coupling between the
directions of the spins involved [38].
A closed shell of electrons gives a zero net magnetic moment. Systems
with an intrinsic magnetic moment may either have atoms with unpaired
electrons, or may permit the fulfilment of the Pauli principle without a com-
23
20 nm
a
Figure 2.16: (a); An STM image of the two-fold surface of d -Al-Ni-Co. (b);An STM image of the ten-fold surface of d -Al-Ni-Co, with some Fourierfiltering applied. Figure adapted from figures published in Refs [36, 37].Copyright APS 2002, 2004.
24
plete pairing of electrons throughout the system. For many-electron systems,
the net magnetic moment is determined by the vector sum of the spin and
orbital angular momentum of its electrons. The total angular momentum J
is expressed as a vector sum:
J = L + S (2.3)
2.6.2 The solid
A system of many atoms will always exhibit a magnetic response in the pres-
ence of an applied field, either from the localised magnetic moments present,
or, in the case of systems with delocalised electrons, in the motion of the
sea of electrons. As the applied magnetic field increases, the orientation of
magnetic moments within the solid changes smoothly from a random distri-
bution to total alignment with the field when a particular saturating value of
the field is reached. Thermal motion presents the randomising mechanism,
and therefore there is a temperature for each system below which any ap-
plied magnetic field causes saturation. This temperature is called the Curie
temperature. According to the ordering of the solid, there may also be easy
or hard directions of magnetization — the different directions generally lie
along different crystal planes. In an amorphous system, there is isotropy of
magnetization.
2.6.3 Magnetic properties of quasicrystals
The quasicrystals i -Al-Pd-Mn and d-Al-Ni-Co do not exhibit much mag-
netic behaviour; approximately 1% of Mn atoms in i -Al70Pd21Mn9 have an
appreciable magnetic moment [39], and orientational frustration prevents d -
Al72Ni11Co17 from exhibiting any magnetic behaviour whatsoever [40].
Charrier et al. [41] reported the existence of a long-range antiferromag-
netic ordering in the rare-earth based RE -Mg-Zn quasicrystalline phases,
observed using neutron powder diffraction. It appeared that there was long-
range quasiperiodic antiferromagnetic ordering based on a four-element Fi-
25
bonacci sequence, with the S and L segments further divided into ‘spin-up’
and ‘spin-down’ states. However, a further study on quasicrystalline pow-
ders and single-crystals using neutron diffraction [42] failed to duplicate the
result, instead revealing the presence of weak paramagnetism at tempera-
tures above the Curie temperature. The earlier result was attributed to the
presence of spurious crystalline phases in the quasicrystal samples.
The only kind of magnetism that has been observed in quasicrystals,
therefore, is a weak paramagnetism. However, as mentioned, even this does
not occur in the decagonal phases.
26
Chapter 3
Thin film deposition
3.1 Motivation
The formation of thin films on surfaces has many applications; the modifi-
cation of tribological properties of a surface to improve wear resistance or
reduce friction for mechanical applications is common. Two examples are
given in figure 3.1. Also well known is the motivation to improve thin film
magnetic properties, driven today by the desire to improve data storage ca-
pabilities for computing applications. Optical properties may be modified
by a film deposited on a lens; a film of half-λ thickness reduces reflection
by causing destructive interference of a selected frequency of light, or, in the
case of a writable CD, a molecular layer is deposited whose crystallographic
properties may be modified by a laser to cause it to either reflect or absorb
light, allowing the storage of data.
On a more basic physical level, the interaction of adsorbates with sub-
strates allows us to characterise the physics of a system. Quasicrystals, for
example, are extremely complex structurally and chemically, and one route
towards reducing the overall complexity of the system would be to some-
how manufacture a single-element quasicrystal in order to differentiate the
physical behaviour due to properties inherent in the aperiodicity from the
chemical behaviour of the system. The most direct way of doing this would
27
Figure 3.1: Left ; an aircraft engine. Right ; the head of an electric shaver.These are common examples of applications for coatings to modify mechan-ical properties.
be to use an existing quasicrystal as a template by employing it as a sub-
strate for adsorption experiments; this has the additional benefit of reducing
the complexity still further by reducing the physical size to the nanoscale.
3.2 Growth modes
The formation of epitaxial films has traditionally been classified into three
basic growth modes, shown in figure 3.2. If the affinity between adsorbate
and substrate is high, the density of nucleation points is high and islands will
grow around nucleating atoms in a two-dimensional fashion until a complete
monolayer is formed, at which point further layers will grow in a layer-by-
layer fashion. This growth mode is usually referred to as the Frank-Van der
Merwe growth mode [43]. If, on the other hand, adsorbing atoms experience
a greater attraction to other adsorbing atoms than they do to the substrate,
film growth proceeds in a three-dimensional (perpendicular) fashion, in order
to minimise the adsorbate-substrate interaction. This mode is referred to as
the Volmer-Weber growth mode [43]. The growth occurs in the Stranski-
Krastanov mode when the system is inbetween these two extremes — this is
most often the case when the substrate-adsorbate affinity is strong, but there
28
(b) Volmer-Weber growth(a) Frank-van der Merwe growth (c) Stranski-Krastanov growth
substrate
adsorbate
Figure 3.2: (a) Layer-by-layer growth; (b) Three-dimensional growth; (c)Layer-by-layer reverting to island growth
is a lattice mismatch between the film and substrate causing significant strain
in the overlayer. After a few monolayers have been deposited, the adsorbate
film reverts to three-dimensional growth to minimise the strain energy present
in the film.
Thermodynamic considerations
The thermodynamic tendency of a system to minimise its total energy pro-
vides a way of understanding the different growth modes explained above. As
surface atoms are bound less tightly than those below the surface, they have
more energy. This difference in energy between surface and bulk is referred
to as the surface free energy. The commonly known phenomena of capillary
action and droplet formation are the result of surface tension which is the
physical manifestation of the system minimising its total surface free energy
by minimising the surface area.
Surface energy is usually represented by the Greek letter γ, with sub-
scripts added to denote the surface between substrate and film (sf), sub-
strate and vapour (sv), and film and vapour (fv). Youngs equation relates
the contact angle of a droplet or film nucleus to the surface energies.
cos θ =γsv + γsfγfv
(3.1)
where θ is the angle the adsorbate droplet makes with the substrate (figure
3.3).
29
substrate
adsorbate
Figure 3.3: Illustrating Young’s equation.
Equation 3.1 can be used to express the conditions for the above growth
modes in terms of the surface free energy. For layer-by-layer growth, θ = 0
and so
γsv ≥ γsf + γfv (3.2)
For island growth, θ > 0 and so
γsv < γsf + γfv (3.3)
In general, adsorbates with low surface energies will wet substrates with
high surface energies, resulting in a Volmer-Weber growth mode; however, if
the lattice mismatch is too great between substrate and adsorbate, Stranski-
Krastanov growth will occur to relieve strain energy. In the case of qua-
sicrystal substrates, there is no lattice present and so it is possible that films
cannot adopt a Volmer-Weber growth mode, perhaps reducing the potential
candidates for growth modes to two.
3.3 Homogeneous and heterogeneous systems
An introduction to another terminology which is perhaps more transparent
than the “growth modes” referred to above is given in this section.
30
3.3.1 Homogeneous systems
Homogeneous systems are those in which the adsorbate primarily bonds to
itself, that is, adsorbing atoms slide around freely on the surface until they
encounter other adsorbate atoms, at which point island formation occurs.
The islands become immobile, usually very rapidly. There is no preferred
network of points on the substrate at which adsorbate atoms congregate; the
potential energy surface has a low enough corrugation for diffusion events to
occur practically unhindered. Hence the island growth mechanism depends
entirely on adsorbate-adsorbate bonding and there is no possibility within
this class of system for the substrate to act as a template [44].
3.3.2 Heterogeneous systems
Heterogeneous systems involve active bond formation between the substrate
and the adsorbate atoms. Either the atoms may bond at the point where
they impinge on the surface, or they may travel around until they find a
preferred site at which to bond, or, in the case of physisorption, a potential
well that is too deep to escape. These sites are called nucleation points and
form a network of sites at which island formation begins. Traditionally, this
has been expected to mean that nucleation depends on defects in the surface.
However, the increased separation and variance in depth of features on the
surfaces of complex alloys and quasicrystals may constitute the formation
of such a network of points. If nucleation points exist on a surface, then
it is possible for the surface to enforce some kind of order in the adsorbed
layer [44].
The heterogeneous class of interactions may be further divided, according
to the level of interaction between the substrate and the adsorbate. If there
is a small interaction between the substrate and the adsorbate, it is likely
that the adsorbate will form in islands of its own preferred bulk structure,
but also that these islands will be oriented along directions of high symmetry
present on the template surface. This happens in the case of Al on AlPdMn,
in which face-centred cubic (fcc) islands are formed in five orientational do-
31
Figure 3.4: Low-energy electron diffraction pattern from Al/i -Al70Pd21Mn9.Reprinted from ref. [45]. Copyright APS 2001.
mains, each with the (111) axis aligned parallel to one of the three-fold axes
of the icosahedral quasicrystal substrate [45]. The structure of the overlayer
is determined from the diffraction pattern from this system (shown in figure
3.4). At low temperatures, Xe adsorbed on d -Al72Ni11Co17 forms hcp islands
as shown in figure 3.5, each rotated according to a high symmetry direction
of the ten-fold surface. This kind of growth is called rotational epitaxial
growth [37, 46–49].
3.4 Pseudomorphism
The other kind of heterogeneous growth is called pseudomorphism. This
occurs when the substrate acts as a nanotemplate and successfully imposes
structural characteristics on adsorbed overlayers. As mentioned above, this
may only occur when a network of so-called nucleation points exists. To
understand this phenomenon, it is helpful to refer back to some of the struc-
tural properties of quasicrystals mentioned in Chapter 2 in order to highlight
possible candidates for the network of nucleation points.
C60 adsorption on i -Al70Pd21Mn9
The dark five-fold stars are an attractive first choice to be considered as
nucleation points; adsorbing species could be expected to physically fit within
32
Figure 3.5: (a); A structure model for how Xe grows on d -Al72Ni11Co17,with (b); its Fourier transform. (c); Diffraction from Xe/d -Al72Ni11Co17.Reprinted from ref. [37]. Copyright APS 2004.
them, and presumably the increased coordination that would be experienced
by an adsorbate occupying one of these hollows would indicate that they
may indeed form the network of nucleation points required for the surface to
function as a nanotemplate. Indeed, one study carried out by the Liverpool
group investigated the adsorption behaviour of C60 molecules on the five-fold
surface of i -Al70Pd21Mn9. Carbon-60 molecules are the ideal adsorbate for
initial studies of this kind of interaction, for they usually remain intact upon
adsorption and experience only fairly weak physisorption forces. Therefore,
the validity of the hypothesis that the hollows should form a potential well
may be tested. Also, the cage diameter is comparable with the diameter of
the hollows [50,51].
It was discovered that at extremely low coverages (0.065 ML), the C60
molecules often adsorb according to the locations of the five-fold hollows,
resulting in short-range τ -scaled ordering (see figure 3.6) [50]. However, due
to the non-exclusive nature of the adsorption (i.e. the C60 molecules are also
33
Figure 3.6: An STM image of C60 on i -Al70Pd21Mn9, showing the τ -scalingrelationships of the adsorption sites. Reprinted from ref. [35]. CopyrightElsevier 2001.
34
found in other sites on the surface), this does not constitute pseudomorphism.
Gold deposition on i -Al72Pd19.5Mn8.5
The first observed potential candidate for pseudomorphism was reported in
2001 after Shimoda et al. deposited approximately 10 ML of Au on a five-fold
surface of i -Al72Pd19.5Mn8.5 that had been pre-dosed with In. The indium
acts as a surfactant without which the gold grows in a three-dimensional
fashion. As deposited, the film is poly-crystalline, but x-ray photoelectron
diffraction (XPD) measurements taken (figure 3.7) after annealing the film to
400 K reveal the formation of an alloyed Au-Al film that apparently exhibits
some local icosahedral symmetry [52]. X-ray photoelectron diffraction is
a tool that measures the macroscale average of the nanoscale local order,
therefore, for a five-fold pattern to be observed, not only must there exist
local icosahedral order, but also the ordered clusters that give rise to the
XPD features must be oriented in the same direction. However, there is
no evidence that there is long-range correlated quasicrystalline order in this
film, and the x-ray photoelectron spectrum reveals a composition perfectly
consistent with an AuAl2 alloy1.
3.4.1 Single-species pseudomorphism
Single-element pseudomorphism was first reported by Franke et al. [54] on
the observation of Sb and Bi films deposited on the five-fold surface of i -
Al-Pd-Mn and the ten-fold surface of d -Al-Ni-Co. The films were charac-
terised using diffraction and helium-atom-scattering (HAS). It was found
that diffraction from the films (see figure 3.8) was τ -scaled, with peaks in
the same positions as those from the clean surface. In this way, rotational
epitaxial growth may be ruled out, and one may conclude that the films grow
in a truly quasiperiodic fashion. The films, in all cases, grew to a maximum
coverage of 1 ML, with the density of atoms in the film consistent with the
1Recent MEIS results [53] indicate that this film is amorphous or polycrystalline, andsuggests that the apparent five-fold diffraction pattern is in fact from the transition metals,which have core levels separated from those of Au by around 1 eV.
35
Figure 3.7: Stereographic projections of XPD images of the (a) Au 4f and(b) Al 2s emissions from the five-fold surface of icosahedral i -Al-Pd-Mn afterAu depositions and subsequent annealing. The region covering 0◦ to 70◦
for polar angle is observed. (c) Stereographic projection of the symmetricelements of the icosahedral structure. [35]. Copyright The Japan Society ofApplied Physics 2001.
density of aluminium atoms in the expected terminations of the quasicrys-
tals used. Thus it may be tentatively asserted that the network of unique
nucleation points for these systems lies atop the Al atoms.
The Liverpool group, in collaboration with co-workers from the French
CNRS, has observed similar behaviour when lead is deposited upon the five-
fold surface of i -Al70Pd21Mn9 [55]. Also, Sharma et al. reported the for-
mation of a similar film using tin as the adsorbate on the five-fold surface
of i -AlCuFe [56]. All monolayer films discovered to date that follow a truly
icosahedral pseudomorphic growth mode revert to a usual crystalline form
of the adsorbate if a multilayer is deposited. This is an important point, for
these films exhibit true five- or ten-fold symmetry. Mathematically, at least
two structural units are required to implement five- or ten-fold symmetry;
therefore, it would be extremely interesting if a truly five-fold or ten-fold film
may be formed with one element.
36
Figure 3.8: Low-energy electron diffraction patterns from (a); clean i -Al-Pd-Mn, (b); Bi/i -Al-Pd-Mn (c); Sb/i -Al-Pd-Mn. Reprinted from ref. [54].Copyright APS 2002.
37
3.4.2 Multi-layer pseudomorphism
The Liverpool group is responsible for the discovery of the first two systems
that exhibit a kind of pseudomorphism distinct from those cases referred to
above [57–60]. The film formed by Cu on a five-fold surface of i -Al70Pd21Mn9,
and that formed by Co on the ten-fold surface of d -Al72Ni11Co17, are exam-
ples of multi-layer pseudomorphism.
Copper deposited on the five-fold surface of i -Al70Pd21Mn9 at submono-
layer coverage results in a reduction of the density of five-fold hollows ob-
served by STM [51], implying that they constitute the network of nucleation
points for this system. However, island growth is also observed. As cover-
age increases up to 3 ML, the clean surface quasicrystalline LEED pattern
disappears, eventually to be replaced by a ten-fold pattern as the coverage
increases to around 5 ML. This pattern persists up to around 30 ML, and is
the result of five rotational domains of a copper structure [58].
When imaged by STM, certain features are immediately apparent within
this copper structure. Rows are immediately visible within the islands, and
it can be seen that the row structure persists across layers. Moreover, these
rows are spaced according to a one-dimensional Fibonacci sequence, with the
ratio of the long and short separations equal to τ .
Fibonacci ordering and τ -scaling are fundamental characteristics of qua-
sicrystalline ordering, and for a copper film to exhibit such features implies
that the substrate is imposing its order, i.e. acting as a nanotemplate. The
absence of icosahedral symmetry is apparent also, since the film forms in
orientational domains in which the local atomic structure is predominantly
fcc (see Chapter 5).
The films described in this thesis identify a particular niche within the
definition of pseudomorphism; the tendency of the adsorbates to strike a
balance between their preferred bulk ordering and that of the quasicrystalline
nanotemplate is likely to be the main characteristic that allows them to form
multi-layer films. The ability to form these films represents an opportunity
to study the implications of quasicrystalline ordering without the complexity
of the alloyed thermodynamically stable phases.
38
Chapter 4
Experimental Methods
4.1 Introduction
In this chapter the apparatus and techniques used in the preparation of this
thesis will be examined and explained in detail.
4.2 Ultra High Vacuum (UHV)
4.2.1 Vacuum chamber
The chambers used in the experiments described in this thesis were con-
structed from argon-arc welded stainless steel or mu-metal. These materials
are used because of their low rate of outgassing; they also provide shielding
from magnetic fields, which is important when techniques involving beams
of electrically charged particles are used.
The need to conduct surface science experiments in vacuum arises from
the necessity of minimising unknowns; one needs to be certain that a tech-
nique is probing the surface of a sample, and to be sure of that there must be
nothing else present to interfere with any information-gathering mechanism.
To define vacuum we must use a description of the different flow regimes
39
low medium high ultra-high extreme high>1 mbar >10−3 mbar >10−7 mbar >10−12 mbar <10−12 mbar
Table 4.1: Table describing the scale by which vacuum is classified
that may be in effect in volumes of gas. The turbulent flow regime is in effect
in pressures higher than 10−2 mbar, and describes fluid motion, with each
particle constantly interacting with those around it. In pressures lower than
10−3 mbar, the molecular flow regime persists, and the particles travel freely
between their far fewer interactions. An important idea is the mean free path
of a particle, which is the average distance it travels between interactions.
This distance increases dramatically as the pressure drops; a particle with
a mean free path of 7 cm in 10−3 mbar will have a mean free path of 7 m
in 10−5 mbar. In common applications, such as inside a cathode-ray tube
display, the only factor affecting the quality of vacuum required is the ability
of the electron beam to travel freely from the filament to the screen; hence a
vacuum of 10−5 mbar is more than sufficient [61,62].
In a chamber designed for the conduction of surface science experiments
there is another restriction on the level of vacuum required, which is imposed
by the necessity of cleanliness. If the sticking coefficient of an impurity
molecule is 1, i.e. every impurity molecule sticks to the surface, then, from
the knowledge that a monolayer is equivalent to around 1015 molecules cm−2,
and that the impingement rate of impurities is directly proportional to the
pressure, we can determine that at a pressure of 10−5 mbar a monolayer
of impurities is formed in less than a second. If the pressure is reduced to
10−10 mbar, then, for a sticking coefficient of unity, the time taken to form a
monolayer increases to around 8 hours, which is a reasonable time in which
to conduct an experiment. The pressure range system is described in table
4.1 [62].
To achieve a pressure of less than 10−10 mbar, two main types of pumps
are used. Pumps that operate in the turbulent flow regime (rotary vane
pumps) rough the pressure in the chamber down to 10−3 mbar and are used
to back pumps that operate in the molecular flow regime (turbomolecular
40
pumps), which reduce the pressure further to around 10−11 mbar. Titanium
sublimation pumps are also used; these act as a getter source and remove
particles from the vacuum when the titanium condenses onto the chamber
walls. Finally, ion pumps are used in applications which require low vibration,
for example scanning tunneling microscopy (STM).
4.2.2 Sample cleaning process
Ex-situ preparation
It has been found that ex-situ polishing of the samples has a major effect on
the quality following further in-situ preparation. For the eventual production
of micron scale terraces, it is necessary to polish the sample prior to insertion
with 6 µm, 1 µm and finally 1/4 µm paste. Following polishing, the samples
have mirror-like surfaces with some small pits, which upon further exami-
nation were observed to be pentagonal in character for the i -Al70Pd21Mn9
sample. It is thought that such pits are reflections of vacancies or voids in
the bulk structure.
In-situ preparation
Following insertion into vacuum, the samples are prepared using conventional
surface science methods. The broad process is similar for both samples used
in this thesis, with the only difference being the heating temperatures.
Once the chamber has been baked out (a procedure involving heating the
entire chamber to around 420 K for twelve hours to remove moisture from
the walls), the sample is initially degassed up to within 100 K of the anneal
temperature. To reach temperatures lower than 700 K, radiative heating is
used; sufficient current is passed through a filament about 1-2 mm behind the
sample to make it glow brightly. Degassing is carried out relatively slowly,
maintaining a pressure of below 5×108 mbar at all times.
Impurities are cleaned from the sample surface using argon ion bombard-
ment. The ions are accelerated to up to 3 keV and impinge on the surface
41
at a grazing incidence. The choice of grazing incidence flattens the surface
somewhat due to protrusions shadowing depressions and hence being prefer-
entially sputtered. When argon ions are incident on lighter elements, they
transfer more kinetic energy; this results in these elements being preferen-
tially removed from the surface.
The process of preparing the samples in vacuum consists of cycles of argon
ion bombardment (90 minutes for the first time, 45 minutes thereafter) fol-
lowed by annealing to 920 K for i -Al70Pd21Mn9 or 1070 K for d -Al72Ni11Co17
for 4 hours, with the sample ready for study after a total of 20 hours annealing
time. For STM measurements, the sample was prepared with a final shorter
sputter/anneal cycle prior to data acquisition of 20 minutes sputtering and
2 hours annealing. These annealing times were established using criteria of
surface order, cleanliness and flatness as measured using STM. Presumably
such extended annealing times are required to restore the quasicrystalline
composition at the surface, by diffusion from the bulk.
The higher temperatures required for the annealing process are reached
by applying a high voltage between the sample plate and the filament, dis-
torting the cloud of electrons and further heating the sample via electron
bombardment. Ceramic beads are used for electrical isolation; after a long
time in use, these are prone to becoming coated in metal and will form a
bridge for current if too high a voltage is applied. Tungsten is used for
filaments intended for electron emission.
4.3 Techniques
4.3.1 Auger electron spectroscopy (AES)
Auger electron spectroscopy is a technique by which surface composition and
cleanliness may be determined. The effect was discovered independently by
both Austrian physicist Lise Meitner and French physicist Pierre Auger in the
1920s. Although Meitner made the discovery and reported it in Zeitschrift
fur Physik in 1923, two years before Auger, the effect came to be named after
42
Auger in the English-speaking scientific community [63,64].
When a beam of medium-energy electrons (typically in the range between
2-5 keV) is incident on a surface, some electrons which are subsequently emit-
ted have energies that are dependent only on the element from which they
have been emitted. The physical process by which these electrons are emitted
is called the Auger effect, and these electrons are called Auger electrons.
The effect occurs because the incident electrons can remove a core state
electron from a surface atom. The core state is subsequently filled by an
outer shell electron from the same atom; this leaves the atom in an excited
state and energy must be released in some way. In some cases this energy will
be transferred to another electron in the outer shell, which is then released
as an Auger electron. The energy of this electron is given by equation:
Eelectron = Ecorestate − (ES1 + ES2) (4.1)
in which S1 and S2 are the outer shell states. Because the energies of these
orbits are determined by the proton number of the atom, the composition
of the surface may be determined by matching the features observed to a
known library of information [64]. The other process by which an atom may
lose energy is fluorescence, in which the excess energy leads to the emission
of a photon.
Due to the higher binding energies for electrons in heavier atoms, AES is
most sensitive to atoms of low mass, detecting elements with a proton number
as low as 3. An Auger signal will sit on a huge background of electrons
generated via loss processes, so the analysis usually involves differentiating
the spectrum, as shown in figure 4.1. For light elements sensitivity can be as
high as 100 ppm, and due to the low energy and therefore low mean free path
of the Auger electrons, surface sensitivity is usually about 3 ML, though may
range from 5 - 100 A.
The Omicron VT-STM system used for the scanning tunneling microscopy
experiments described in this thesis is equipped with an Omicron SPEC-
TALEED optic. This is a rear-view low-energy electron diffraction optic
which incorporates four spherical grids for retarding field Auger analysis,
43
Figure 4.1: (a); Raw spectrum. The high background is clearly evident. (b);Differentiated spectrum.
44
and an electron gun which can supply a beam of up to 3.5 keV, rather than
the usual 1 kV limit for LEED optics. The thoriated tungsten filament op-
erates at 1.6 A to produce a beam current of 30 µA with a primary electron
energy of 3 keV. The Wehnelt lens voltage is set appropriately to reduce
beam diameter to the minimum of around 250 µm. The energy resolution of
this system used in retarding field mode is around 1 eV.
4.3.2 Medium energy ion scattering (MEIS)
The great benefit of the MEIS technique is the unparalleled ability to map
both the composition and the structure of the surface and near-surface region
of a given sample simultaneously. The MEIS system at Daresbury CLRC is
shown in figure 4.2. Data is gathered in a two-dimensional fashion, with an
energy scale on the y-axis and an angular scale on the x -axis. A spectrum
is built up in stages (tiles) with the tiles subsequently joined together on
computer.
The experimental station facilities include various interconnected UHV
systems between which samples can be transferred under UHV. The scat-
tering chamber, which houses the ion analyser, sample goniometer and two-
dimensional (energy and angle) position sensitive detector is where exper-
iments are carried out. The preparation chamber, whose facilities include
LEED, Auger, sputter cleaning, evaporation sources and gas dosing, is used
for sample preparation and characterisation prior to ion scattering experi-
ments. Data acquisition is computer controlled via the MIDAS graphical
user interface originally developed for the nuclear physics EUROGAM pro-
gramme.
The ion beam itself may be accelerated to between 50 and 400 keV, and is
composed of either H+ or He+ ions. The beam has a divergence of less than
0.1◦ and a spot size at the sample of 1 mm horizontal by 0.5 mm vertical
resulting in a current at the sample of 0.1 - 1.0 µA dependent on beam energy
and species.
The toroidal electrostatic analyser has a pass energy of 0-400 keV and may
move 140◦ in the horizontal plane. The chevron array anode two-dimensional
45
Figure 4.2: The MEIS system at CCLRC Daresbury.
(energy and angle) detector with microchannelplates has an energy window
of 2% of the analyser pass energy with a resolving power dE/E = 2.4×10−3.
The angular window is 27◦ with an angular resolution of φ = 0.3◦.
Sample mounting
The MEIS technique relies on the possibility of rotating the sample about
three axes during data acquisition. The sample holder (shown in figure 4.3)
is unique to this system. Each holder is made from stainless steel and in-
corporates a filament for heating the sample, which itself is secured on the
holder with molybdenum clamps and screws.
46
Figure 4.3: Sample holder for MEIS. (a); Sample position. (b); Contacts forthe filament and high-voltage supply.
47
Two-dimensional spectra
[r]60mm Two-dimensional MEIS data obtained from an
annealed film of Cu/i -Al70Pd21Mn9.
Two-dimensional data (such as in figure 4.4) yields certain information
immediately: each diagonal streak is attributable to a surface peak corre-
sponding to one of the elements present, as shown; a subsequent constant
background of the signal constitutes a bulk signal for that element. This
data may be analysed to provide quantitative data on the thicknesses and
compositions of layers near to the surface of the sample. Also visible on
the two-dimensional spectra are vertical bands of lower intensity. These are
known as blocking dips and occur at angular positions dependent on the
structural nature of the sample. For periodic samples, Monte Carlo analysis
of this data can reveal atomic positions with picometer resolution. For the
aperiodic systems studied in this thesis, resolution is somewhat lower due to
approximations made to allow Monte Carlo methods to be used.
48
Angle-resolved data
Medium-energy ion scattering spectroscopy yields surface structural data be-
cause of an easily understood concept, illustrated in the figures shown. Es-
sentially, the device sends ions into the sample at a particular velocity, and
measures the angles along which these ions return. To explain the concepts
upon which the technique is essentially based, let us refer first to figure 4.5,
in which an atomic coordinate model consisting of 50,000 atoms of a qua-
sicrystal is visualised. In this figure, we can see the model visualised without
perspective (as a collimated beam would). Although the model is opaque
when viewed in approximate alignment with a crystal plane, it almost van-
ishes when viewed along one of these channeling directions, so named because
ions with incident angles nearly but not exactly aligned may be channeled
along due to interactions with atoms in the structure. When we visualise
the model with perspective, as in figure 4.7 at the end of this section, the
rotational symmetries present also become apparent in the channeling direc-
tions observed. Part (b) of the figure illustrates how surface sensitivity may
be achieved, since in this ideal case, only the top layer of the model scatters
the incident ions.
However, this treatment does not adequately explain the technique, as
the detector is obviously not measuring the reflected component of the ion
beam along its incident direction. A double-alignment geometry is achieved
when the beam is incident along a major crystallographic direction, and the
detector is scanned over an angular range including an opposite crystallo-
graphic direction. In this way, the impinging ions are first scattered from the
illuminated atoms in the top few layers of the sample, and then are subject to
blocking due to the atoms that may scatter them on the way out of the sur-
face. If the sample is well-ordered, blocking atoms only occur at well-defined
angles relative to the scattering atoms, and in this way a curve of counts with
blocking dips may be collected, which may then be analysed with a Monte
Carlo method to discover the structure of the surface [65–68]. The SIMNRA
code permits the modeling of systems to fit the energy-resolved data, yield-
ing quantitative information about the depth, thickness and composition of
49
Figure 4.4: (a); A view of the quasicrystal model offset about 2◦ from achanneling direction. (b); Aligned along a channeling direction.
Figure 4.5: An example of double-alignment geometry as it relates to MEIS.
50
layers in the surface and near-surface regions.
To fit the angle-resolved data in order to find structural attributes for
the surface, it is necessary to model individual interactions of ions impinging
on the surface using the VEGAS code. The code takes as input a possible
candidate for the structure of the surface, and by calculating the yield of
ions at each angular position relative to a fixed angle of incidence, a blocking
curve for the theoretical structure may be obtained. By iterating this process
until a good fit is made with the experimental data, the surface structure may
be determined to the desired accuracy.
The VEGAS code
The Vegas code was originally developed at the FOM institute, Amsterdam,
for the Monte Carlo analysis of the interaction of ion beams with crystalline
surfaces. It takes as input a candidate model for the surface structure, con-
sisting of atomic coordinates, and a set of parameters defining the beam.
The basic method of operation of VEGAS is to follow the paths of the ions
through the crystal and sum up the individual ion-atom hitting probabilities
as the ions pass the atoms in the crystal. Each ion that hits or passes near to
an atom will be subject to a deflection [69]; this deflection will influence its
ongoing path unless considered to be negligible by the code. Through time
reversibility, the ion-atom detection probabilities may then be summed for
each atom and given scattering angle. The code is optimised in several ways:
the deflection angles are listed in a lookup table, and also an auxiliary lattice
is created to aid the tracking of the ion beam through the model.
The auxiliary lattice
When a model is processed by the VEGAS code, atoms are grouped into sets
of x, y and z planes. This grouping forms the auxiliary lattice, which is a
convenient method of indexing atoms closest to each other. The auxiliary
lattice also defines the set of points that the ion beam passes through during
the simulation.
51
The process the code follows for each step in the path of an ion is as
follows:
• ion is started at some position in crystal
• determine atoms closest to ion
• sum up hitting probabilities for these atoms
• randomly move atoms according to vibrational amplitude
• check if an atom is close enough for collision
• if yes: get deflection angle and new direction
if no: move ion forward to next auxiliary lattice point
The VEGAS code is intended for the analysis of periodic systems; as each
ion passes out of one side of the input model, it enters the other side in a
fashion consistent with periodic boundary conditions and will continue to do
so until it is backscattered by an atom. However, a way to study aperiodic
systems has evolved in which the models are made large enough so that less
than 10 % of atoms are in close proximity to a boundary. This carries with
it large processing overheads and a certain sacrifice in ultimate resolution of
the technique, since, beyond a certain point, refinements to a model become
too computationally expensive to be viable.
4.3.3 Scanning tunneling microscopy (STM)
Scanning tunneling microscopy is a technique by which a two-dimensional
plot of the convolution of the local density of states (LDOS) of a surface
and of the LDOS of a conducting tip may be obtained. For many samples
this is equivalent to a topographical map of the surface, and in all cases (for
conducting systems) such a plot is in some way representative of the atomic
coordinates of atoms on the surface. Since the LDOS is often highest at
locations of atomic bonds, the regions between atoms are often the brightest
52
Figure 4.6: The model viewed with perspective, and aligned along a five-foldaxis. An inverted pentagon is clearly visible in the centre of the figure, andmany channeling directions can be easily identified.
53
on such a plot, especially when covalent bonding is present, as in the case of
graphite.
The ‘tunneling’ referred to is a quantum effect. Classically, an electron is
permitted to either be or not be in a particular region of space; if an electron
has insufficient energy to overcome a potential barrier, it cannot exist either
on the opposite side of the barrier or within it. However, a quantum mechan-
ical approach yields the result that solutions to the Schrodinger equation for
an electron inside a potential barrier do indeed exist, albeit with probability
decreasing exponentially according to the barrier size [70].
In 1971, based on this idea, Young and co-workers constructed a machine
they called a topografiner to investigate metal-vacuum-metal tunneling [71].
This device operated by maintaining a constant field emission tip-sample dis-
tance using orthogonally-mounted piezo-electric drives to manoeuvre the tip.
A piezo-electric drive is a device made of a tube of piezo-ceramic material.
During manufacture, the tube is heated to a high temperature, and exposed
to a voltage generated using tubular electrodes running outside and through
the centre of the tube. The polar molecules within the tube align themselves
somewhat with the electric field and the tube is then cooled to preserve this
radial polarization; any subsequent voltage applied causes the tube to either
contract or expand radially due to electrostatic interaction and a correspond-
ing expansion or contraction along the length of the tube occurs. By using a
tripod of these tubes, controlled movement on the nanoscale may be achieved
in three dimensions.
The precision of manipulation is extremely important. The governing
relation for STM [72,73] is
JT ∝ eAφ1/2S (4.2)
in which JT is the current, φ is the average barrier height and S the
width, and A is a constant equal to 1.025 A−1eV(−1/2) for vacuum tunneling.
This displays an exponential dependence on the width of the barrier. The
exponential character of this equation is the basis of the extreme resolution
the technique provides; given that the bias voltage and tunneling current are
54
Figure 4.7: The sample plate as used for STM, with a gold-on-glass substratemounted.
small, the difference in tunneling current may vary by up to three orders
of magnitude per atomic separation in the perpendicular direction. The
topografiner had limited resolving power by the standards of today, achieving
a vertical resolution of 30 A and a lateral resolution of 1000 A. However,
in 1982 Binning and Rohrer were awarded the Nobel Prize for Physics by
demonstrating the use of a technique based on the same principles, called
Scanning Tunneling Microscopy, to generate atomically resolved images of
(110) surfaces of CaIrSn [74]. Today we can routinely expect sub-angstrom
resolution on almost any conducting surface, due to improvements in the
dampening of thermal, vibrational and magnetic effects.
STM device and UHV chamber
Scanning tunneling microscopy data presented in this thesis were obtained
using an Omicron VT-STM (shown in figure 4.10). Tips were made from
W wire using a procedure detailed later in this section. The tip is crimped
55
in a tip holder and may then inserted into the UHV chamber and scanner
as required. The ability to remove tips from the STM in-situ allows such
procedures as the sputtering and annealing of tips to increase cleanliness
and sharpness.
The Omicron VT-STM uses a particular type of sample holder common
to all Omicron STM systems. The holder is a flat molybdenum or tantalum
plate which may be transferred to all parts of the chamber, and removed
from the chamber, using a set of transfer arms. The sample is attached to
the holder using welded molybdenum straps. Although it is very important
to attach the sample securely, care must be taken when welding close to
quasicrystal samples, as they are prone to shattering when exposed to large
temperature gradients.
A schematic of two possible modes of operation for an STM is shown in
figure 4.9. Constant current mode is most often employed; in this mode a
feedback loop is used to maintain a constant sample-tip current as the tip is
scanned across the surface. In this way a constant tip-sample separation may
be obtained, and the z-motion of the tip may be tracked using knowledge of
the voltage supplied to the z-piezo to provide the topographical information.
Constant height mode involves keeping the voltage to the z-piezo constant
and using the sample-tip current to provide the topographical information.
There is a high possibility of crashing the tip into the sample when an STM
is operated in this mode.
Together with the feedback loop controlling the z -component of the mo-
tion of the tip, there is a device providing suitable voltages to the other piezos
to allow the tip to be rastered across the surface under scrutiny. Thus an
image may be built up. The voltage response of a piezo is non-linear, and
due to the impact this may have on the lateral motion of the tip, an STM
will build an image up from just one direction of the motion, rather than
alternate scan lines in opposite directions.
56
Figure 4.8: (a); Operation of an STM in constant current mode. (b); Con-stant height mode.
The STM head shown in figure 4.10 has several components labeled.
These are:
• Sample carousel: This is a magazine capable of storing twelve sample
plates or tips.
• Suspension system: A set of springs, so the STM may be isolated from
acoustic vibrations while scanning.
• Sample stage + tip: The scanning assembly.
• Wobble stick: This is used to transfer samples and tips between the
carousel and the stage.
• Copper fins: The fins and magnets dampen any eddy currents that may
occur during scanning.
The piezoelectric scanner has a voltage sensitivity of the order of 10−10 m
V−1. Therefore, by applying a potential difference of 10 V a tip displacement
of 10 A can be made. For coarser motion there is a set of servo-motors which
can manipulate the tip in three dimensions on the centimetre scale.
57
Figure 4.9: An Omicron variable-temperature scanning tunneling microscopehead.
58
STM tip
The STM tip may be made of any conducting material. To produce images
with features well-resolved on the atomic scale, however, certain limitations
exist. As mentioned previously an increased tip-sample separation of one
atomic nearest-neighbour distance may reduce current by up to three orders
of magnitude. Therefore an atomically sharp tip presents the most effective
scanner for atomic-scale images.
The tips used in the collection of data for this thesis were chemically
etched from tungsten wire. A bias of 10 V is applied to a piece of tungsten
wire suspended in a meniscus of potassium hydroxide formed in a 2 mm hole
cut in a stainless steel plate. Immediately the wire is severed, the voltage
is switched off, leaving a tip which when viewed through a 200× optical
microscope should still appear vanishingly sharp. A tip which is observed to
be hooked will be discarded at this stage. Hydroxide residue is then rinsed
off with water and the tip is secured in the holder.
Upon introduction to the STM preparation chamber, the tip is sputtered
for half an hour with 3 keV Ar+ ions to remove the oxide layer and once
this process is complete the tip is supposedly ready. However, sometimes
asymmetric or multiple tips are formed; this kind of tip produces the effect
shown in figure 4.11, where every feature on the surface is imaged once by
each protrusion of the tip, resulting in many instances of the same motif.
Also, adsorbate or impurity atoms or molecules may be picked up by the tip
as it travels over the surface; this may have the effect of either improving the
scan or worsening it. If the scan suffers from a multiple or asymmetric tip, or
contamination by impurities, nanostructuring of the tip using high voltage
pulses (15 nA, 10 V) is known to often improve the sharpness.
59
Figure 4.10: An example of a multiple tip. The tip is imaged at each sharpprotrusion (with a smaller radius of curvature than the tip) on the samplesurface.
60
4.3.4 X-ray Magnetic Circular Dichroism (XMCD)
The importance of current magnetic research is directly related to the ra-
pidity at which developments in knowledge of magnetic behaviour can be
utilised in technological applications, such as information storage and sensor
systems.
X-ray magnetic circular dichroism is a powerful technique for determin-
ing surface magnetic properties due to its ability to differentiate between
the contributions to the magnetic susceptibility from the orbital and spin
magnetic moments of a material.
X-ray photoelectron spectroscopy (XPS)
In XPS, a beam of x-rays is directed at a sample surface. Incident photons,
each with a well-defined energy of E = hν, are absorbed by electrons with
well-defined binding energies (Eb). If hν > Eb, the electron is emitted with a
kinetic energy equal to hν−Eb. This energy is defined as Koopman’s energy;
in practice, it is never observed due to interactions with other electrons and
with the bulk, which themselves may provide information about the sample.
The relative concentrations of different elements present can be determined
from the peak intensities with an optimum accuracy of 5-10% [75]. Losses
from inelastic scattering processes prevent electrons from much deeper than
10 atomic layers being detected [76], making XPS a very surface sensitive
technique. As in AES, H and He cannot be detected as they have no core
levels.
Any light source which emits photons with an energy greater than the
work function of the solid may be used for XPS, but the most widespread use
of the technique relies on gas discharge sources or on common X-ray sources
such as Al and Mg Kα emissions. Gas discharge sources have energies in the
region of 10s of eV, whereas X-ray sources are in the 1000 eV range. However,
in relation to this thesis, the most useful light source is synchrotron radiation,
which, with suitable monochromators, can provide photons of any energy in
this range.
61
Magnetic sensitivity
The origin of the XMCD effect is the differing photon absorption cross-
sections of magnetic materials for left- and right- circularly polarised light.
An XMCD spectrum is essentially the difference between the absorption spec-
tra measured with the two opposite signs of polarisation, and yields a wealth
of information that cannot be determined using any other technique discov-
ered so far. Some of the quantities obtainable are: determination of LZ/SZ
ratios, measurement of the spin and orbital components of element-specific
and site-specific magnetic moments, measurement of element-specific hys-
teresis loops and determination of absolute local moments.
The XMCD effect is shown diagrammatically in figure 4.12. For magnetic
materials in the presence of an applied magnetic field, spin up and spin down
bands are not equally populated. For an applied field in the up direction,
there will be some empty spin up 3d states. Conservation of spin implies
that only 2p electrons with up spin can be excited into the 3d states. For
the case in which the orbital motion of the 2p states is in the same sense as
the circular motion of the incident light the transition probability is larger;
when the two motions are in opposite directions the transition probability is
smaller.
4.3.5 Beamline ID8 at the European Synchrotron Ra-
diation Facility
The ID8 beamline in Grenoble is an intense source of polarized soft x-rays
which are principally used to probe magnetism in a diverse range of systems
such as nanoclusters, superconductors, semimetals and ultrathin films. The
beamline is equipped with a low temperature 7T superconducting magnet, a
spin-polarized photoemission spectrometer and a rapid field flipping cham-
ber. The photon energy is tunable in the range 0.4-1.6 keV, with an energy
resolution close to dE/E = 5× 10−4 at 850 eV, making it ideal for studying
the magnetic properties of transition metals.
The X-ray source consists of two APPLE II undulators with a period of
62
2p states2p
2p
spin upelectron
spin downelectron
Fermilevel
Magneticfield
empty3d states
3d states
Energy
RCP
ab
c d
LCP
a, d: induce transition withlarger probability
b, c: induce transition withsmaller probability
3/2
1/2
Figure 4.11: The XMCD effect: For magnetic materials in an applied mag-netic field, spin up and spin down bands are not equally populated. For anapplied field in the up direction, there will be some empty spin up 3d states.In this case, because of the conservation of spin, only 2p electrons with upspin can be excited into 3d states. When the orbital motion of the 2p statesis in the same sense as the circular motion of the incident light the transitionprobability is larger and when the two motions are in opposite directions thetransition probability is smaller. Therefore, for light of a particular handed-ness, the L3 peak in the absorption spectrum, arising from the excitation ofthe 2p3/2 electrons, may be enhanced, whilst the L2 peak, arising from theexcitation of the 2p1/2 electrons, may be reduced. When either the magneti-sation direction or the polarisation direction are reversed the effect on thesize of the L3 and L2 peaks is also reversed.
63
Figure 4.12: A schematic of the ID8 beamline at the ESRF.
64
Figure 4.13: The sample plate as used for XMCD, illustrating the azimuthalrotation system.
88 mm and a minimum gap of 16 mm, resulting in a circular polarisation
rate of 100 %. The optical element is a spherical grating monochromator of
the Dragon type, resulting in a horizontal beam size at the sample of less
than 1 mm2, and a vertical beam size of 40µm.
Sample mounting
Due to the use of an Omicron STM to check the sample preparation prior
to insertion in the analyzing chamber, the entire system has been built to
use the standard Omicron sample plate. During the experiment described in
this thesis, efforts were made to provide for azimuthal orientation by making
a section of the plate freely rotating, as shown in figure 4.14. These efforts
were ultimately unsuccessful, due to the difficulty in manipulating a 5 mm
diameter plate at cryogenic temperatures, and the reduced robustness of the
multiple-component plate made it very difficult to obtain high quality STM
images.
4.3.6 Low-Energy Electron Diffraction
As mentioned in Chapter 2, crystal structure is almost exclusively deter-
mined using diffraction techniques. The common method employed in sur-
65
face science is the diffraction of low-energy (less than 500 eV) electrons, first
observed from a crystalline surface by Davisson and Germer [77]. When the
condition for elastic scattering (Bragg’s law) is satisfied:
nλ = 2d sin θ (4.3)
where λ is the wavelength and n = (h2 + k2 + l2)1/2, diffraction will occur
with the scattered beam having wavevector K′
[75]. In three dimensions
conservation of energy then gives
| K | =| K′ | (4.4)
where K and K′
are respectively the incident and diffracted wavevectors.
The diffraction condition may also be written
K′= K + ghkl (4.5)
where the reciprocal lattice vector ghkl is given by
ghkl = ha∗ + kb∗ + lc∗ (4.6)
with a∗, b∗ and c∗ being the primitive translation vectors of the reciprocal
lattice.
For surface diffraction, this may be simplified due to the conservation of
momentum parallel to the surface, which reduces the equation to
K′
‖ = K‖ + ghk (4.7)
where ‖ denotes components parallel to the surface and with ghk given by
ghk = ha∗ + kb∗ (4.8)
and
a∗ = 2πb× n
A, b∗ = 2π
n× a
A, A = a · b× n (4.9)
66
Gk`
k
2
Figure 4.14: The Ewald sphere. The points on the right-hand side of thefigure represent reciprocal lattice points on the crystal. The vector k isdrawn in the direction of the incident beam, and the origin is chosen suchthat k terminates at any reciprocal lattice point. A sphere is drawn of radiusk = 2π/λ about the origin of k. A diffracted beam will be observed ifthe circumference of the sphere intersects any other reciprocal lattice point.As depicted, the sphere intercepts a point connected with the end of k by areciprocal lattice vector G. The diffracted beam is in the direction k’ = k+G.The angle θ is the Bragg angle in equation 4.3.
where n is a unit vector normal to the surface.
The Ewald sphere
The Ewald sphere represented in figure 4.15 was conceived by Paul Peter
Ewald, a German physicist and crystallographer. It is a geometric reciprocal-
space construction used in crystallography which demonstrates the relation-
ship between the wavelength of incident light, the angle of diffraction for a
given reflection and the unit cell and reciprocal unit cell of the crystal.
In the two-dimensional case (for example a surface diffraction study by
LEED), the points representing the lattice above are extended in reciprocal
67
space to become rods, since the small distance probed perpendicular to the
surface becomes a very large distance in reciprocal space.
Practical details
The type of LEED system most widely used today is that shown schemat-
ically in figure 4.16. Due to the high scattering cross-section of electrons
with energies lower than 1 keV [61], and according to the ‘Universal Curve’,
the LEED technique can be considered to be surface sensitive for electron
energies in the range 20 - 300 eV. The apparatus consists of an electron
gun used to fire electrons with a specific energy, and therefore wavelength,
towards the surface of the sample, three grids, and a phosphorous screen.
The hemispherical grid G1 is earthed, as is the sample, to provide the nec-
essary field-free region. The second grid G2 is raised to a positive potential
to suppress any inelastically back-scattered electrons. The final grid G3 ac-
celerates the elastically-scattered electrons toward the phosphorous screen,
which is usually held at around 6 keV. The diffracted electron beams cause
the phosphor on the screen to fluoresce, and a diffraction pattern may then
be observed.
The SPECTALEED optics used in the production of this thesis has four
grids in total, to allow it to be used for Auger electron spectroscopy also.
Electron coherence length
The electron coherence length is a measure of the distance perpendicular to
the incident beam (i.e. across the surface) for which electrons are coherently
scattering. For this reason, the observation of a LEED pattern need not
imply that the whole surface is ordered with respect to itself. The usual
electron coherence length for a standard LEED optic is 10-20 nm, meaning
that coherent interference only occurs for patches of 10-20 nm diameter on
the surface. All of these patches produce exactly the same pattern if the
surface is ordered, so the diffraction pattern shown on the phosphorescent
screen is the combination of all the diffracted patterns. However, it should
68
-Vp (retard)
Sample
Diffracted beams
Drift tube
Fluorescent screen
Filament -Vp
~ 6 KeV
G1G2
G3
Incident energy
eVp
Figure 4.15: A schematic of a common LEED optic.
be understood that a LEED pattern is not the result of diffraction occurring
coherently for the total width of the incident beam.
69
Figure 4.16: A schematic of a simple filament evaporation source.
4.3.7 Thin film deposition
For the experiments described in this thesis, two methods of film formation
were employed. Each involves heating the adsorbate source to a temperature
at which some material evaporates, so that it may then condense on the
surface.
Filament evaporation
In this method of evaporation, a length of wire (tungsten or tantalum) is
coiled around a sample of the material to be evaporated. The ends of the
wire are spot-welded to contacts on the vacuum side of a UHV feedthrough
as shown in figure 4.17. A DC supply may then be connected to the atmo-
sphere side of the feedthrough and current passed to heat the filament. For
materials that are liquid for a large range of temperature under UHV, the
source must be formed before use. This involves heating the material until it
melts and wets the wire; the system is then mechanically stable and should
give predictable evaporation.
70
There are three main drawbacks with this type of evaporator:
• A lot of heat is generated. This is undesirable for the reason that, due
to the r−2 dependency of the radial distribution of the evaporant atoms
around the coil, the substrate should be fairly close to the evaporator.
This means that the surface temperature will rise, and this may prevent
the observation of any thermodynamically metastable films that may
otherwise have been formed.
• A material intended for evaporation may have an extremely small liquid
region on the phase diagram, or may even sublime. This makes the
formation of a mechanically stable source extremely difficult.
• Only materials which can evaporate below the melting point of the
filament are available for use as evaporants with this method.
Chapter 6 of this thesis refers to an experiment in which Co was evaporated
over quasicrystal surfaces. The experiment was attempted using filament
evaporation, but failed, due to the first two problems explained above. Hence
the experiment was repeated using e-beam evaporation.
Electron beam evaporation
Electron beam evaporation uses the principle of heating via electron bom-
bardment. An electron cloud is created by the resistive heating of a filament,
and this cloud is distorted using an electric field to cause it to impinge on
the evaporant. This results in extreme localised heating, and can be used to
evaporate any material without the need for direct heat transfer. A schematic
of an e-beam evaporation cell is shown in figure 4.18.
The cell used for experiments in this thesis is the Omicron EFM 3.
This device comprises a water-cooled copper shroud, a retractable evapo-
rant holder and a shutter to enable control of deposition time. The low
overall temperature of the evaporator allows the vacuum to be maintained
at around 10−10 mbar, and no spurious heat is produced at the sample sur-
face. The output of the evaporator is a largely collimated beam, so there is
71
Figure 4.17: A schematic illustrating the principle of e-beam evaporation.
a large degree of freedom in the positioning of the substrate. Also, the EFM
3 includes an ion flux monitor. The percentage of ions with respect to atoms
in the output beam remains essentially constant, so this flux current reading
may be used as a dosage rate indicator. Feedback loops also exist, to keep
the temperature of the evaporant and the ion flux constant.
The only potential drawback for this type of evaporator is in the un-
wanted supply of energy to the surface of the sample via the ions that are
not captured by the flux monitor.
72
Chapter 5
Characterisation of an ultrathin
Cu film formed on the five-fold
surface of i -Al70Pd21Mn9 using
medium-energy ion scattering
spectroscopy
5.1 Introduction
This chapter represents the first time a technique (MEIS) using Monte Carlo
analysis has been successfully employed in the determination of a multi-
layer pseudomorphic film deposited on an aperiodic substrate. Moreover,
the MEIS study reported in this chapter is of the first aperiodic multilayer
pseudomorphic film ever observed. Evidence from medium-energy ion scat-
tering (MEIS) studies for the structure and composition of an as-deposited
Cu film on the five-fold surface of i -Al70Pd21Mn9 and for a film which has
73
been annealed to 600 K post-deposition is presented.
Medium-energy ion scattering (MEIS) is an extremely powerful tool for
the elucidation of the structure and composition of the surface and near-
surface regions. Quantitative information on the composition as a function
of depth in the near-surface region can be found from energy-resolved spectra,
and the positions and relative intensities of blocking dips in the angle-resolved
spectra offer important information about the local atomic structure of the
surface layers.
5.2 Experimental Details
Data were obtained using the UK national MEIS facility at the CCLRC
Daresbury Laboratory [78]. In this technique, the intensity of scattered He+
ions of primary energy 100 keV is measured using a toroidal electrostatic
analyser with a position-sensitive detector, allowing simultaneous collection
of scattered ions over a range of angles and energies. For each data set, a
series of two-dimensional scans may be tiled together electronically to pro-
duce the final spectrum where the scattered ion intensity is represented by
a false colour scale. These two-dimensional spectra may then be projected
onto either an energy scale, from which compositional data may be deduced,
or onto an angle-resolved scale, from which information on the structure may
be obtained.
The i -Al70Pd21Mn9 sample was grown in the Ames laboratory using the
Bridgman method. Following alignment using back-reflection Laue diffrac-
tion, the sample was cut by spark-etching perpendicular to a five-fold (100100)
axis. Prior to insertion into UHV, the quasicrystal was polished with suc-
cessively finer grades of diamond paste (down to 0.25 µm). This procedure
gives an optimal starting point for the subsequent in-situ preparation, which
consisted of cycles of Ar ion bombardment and subsequent annealing to a
temperature of 950 K up to a total annealing time of 20 hours. These prepa-
ration conditions have been found to give a flat surface with micron-scale
terraces [79, 80].
74
Copper was evaporated with the chamber at a pressure of 1×10−9 mbar,
using a resistively heated wire wrapped around a rod of oxygen-free high pu-
rity Cu. The sample was at room temperature during Cu evaporation, and
the Cu source had been thoroughly degassed. Low-energy electron diffraction
(LEED) was used to monitor the approximate coverage of Cu on the surface,
showing the disappearance of the clean surface five-fold pattern and then the
emergence of the ten-fold pattern associated with the aperiodic Cu film [58].
Once the film had been formed on the surface, the sample was transferred to
the analysing chamber (base pressure 10−10 mbar) for the scattering exper-
iments. The two-dimensional MEIS data were taken using an incidence of
31.8◦ with respect to the surface normal, which equates to a three-fold axis
orthogonal to a (1,1,1,1,1,1) type plane. For a description of the notation
system see e.g. ref [31]. Each individual scan was collected for a fixed ion
fluence of 2.5×1015 ions cm−2, and the sample was moved between scans to
minimise beam-induced damage.
5.3 Results
5.3.1 Two-dimensional data
Figure 5.1 shows an example of a two-dimensional data set taken from the
i -Al70Pd21Mn9 sample after the deposition and subsequent annealing to 600
K of approximately 9 ML of Cu (calibrated using the energy-resolved data).
Four diagonal bands of intensity can be seen, each arising from a particular
element and separated by energy, due to the increased energy transferred
when an ion is scattered by an atom of lower mass. In addition, a feature of
lower intensity can be seen running down the data set. This is a channelling
direction through the quasicrystal characteristic of the icosahedral ordering
present.
75
Al
Mn
Cu
Pd
98
58
78
68
88
75 125 75 125
Angle (deg)
En
erg
y(k
eV
)a) b)
98
58
78
68
88
Inte
ns
ity
Channels
Figure 5.1: (a); Two-dimensional data obtained from the as-deposited Cu/i -Al70Pd21Mn9 film; (b); from the annealed Cu/i -Al70Pd21Mn9 film. The scat-tering features from the four elemental components of different mass andsome characteristic quasicrystalline channels are indicated.
76
5.3.2 One-dimensional data
Energy-resolved spectra
Energy-resolved spectra collected from the films are compared in figure 5.2.
The data were fitted using the SIMNRA5.0 energy spectrum simulation
code [81]. The fit involves components for both the surface, where com-
plete illumination of the atoms occurs, and for a near-surface region starting
directly below the surface and extending to the whole probing depth of the
technique (around 500 A). For the latter, reduced illumination is seen due to
the fact that the energy spectrum is taken in a double alignment geometry
and hence the atoms shadow each other.
In the data from the unannealed film, the existence of a strong Cu surface
peak and lack of surface peaks for the bulk components indicates that the
as-deposited film does not alloy with the quasicrystal. The Cu coverage
can be determined with a high degree of accuracy from the fitting code,
and has been determined to be 18×1015 atoms cm−2, which is equivalent to
9 ± 0.1 ML (from the density of bulk fcc Cu). Hereafter coverages will be
quoted in MLE (monolayer equivalents). Below this is an Al-rich interlayer
region, followed by an Mn-rich sub-surface region, formed by the flat surface
preparation process [82]. The data from the film after annealing to 600 K
shows strong surface peaks from all the elements present. This compositional
profile is well-matched using a two-layer model comprising the alloy film and
a bulk-like sub-surface region. Both fits also included small amounts of Ar
in the sub-surface region. Buried Ar has been observed in these materials in
a previous MEIS study [82] and arises from the sputtering procedure used
to clean the sample. The compositions of the surface, near-surface and sub-
surface regions, corrected for the visibility of each element in the sub-surface
regions, for the pre- and post-anneal films are compared in Table 5.1.
Angle-resolved spectra
The blocking curves for the as-deposited and annealed Cu films are shown in
figure 5.3 (a) and (b) respectively. The curves were obtained by integrating
77
Figure 5.2: One-dimensional energy-resolved data obtained from Cu/i -Al70Pd21Mn9. (a) From the unannealed film, showing a strong Cu signaland no surface peaks for the bulk components. (b) From the annealed film,showing surface peaks for all elements present, indicating the alloying of thecopper with the surface of the quasicrystal.
78
As-depositedThickness (×1015 atoms cm−2) 19 (9.5 MLE) 19 (9.5 MLE) -
Surface (%) Interlayer (%) Full depth (%)Cu 100 0 0
Element Al 0 84 78Pd 0 11 10Mn 0 5 12
After 10 mins @ 600 KThickness (×1015 atoms cm−2) 13 (6.5 MLE) -
Surface (%) Full depth (%)Cu 8 3
Element Al 65 77Pd 11 7Mn 16 13
Table 5.1: Table showing compositional values for the unannealed and an-nealed films.
the Cu signal from the two-dimensional data and then projecting it onto the
angle axis. Signal from just above the Cu peak was also taken and subtracted
from the data, in order to provide a background correction. In this way
features from the sub-surface Pd blocking behaviour may be removed from
the Cu data prior to fitting.
The data were simulated with the VEGAS code using a class of models
based on the following principles. Domains of material in five possible orien-
tations with respect to the substrate are known to be present from the STM
data. Each domain is composed of strips of material at separations following
a Fibonacci sequence of long and short distances. The structure of the indi-
vidual strips was modelled as conventional crystalline material using several
alternative low index planes parallel to the surface, in a number of alterna-
tive rotational orientations with respect to the substrate. Whilst fcc(111),
fcc(110), fcc(311) and hcp(0001) planes had no similarity with the data at
any of the rotational orientations evaluated, the fcc(100) plane showed good
agreement when a <110> type azimuth was aligned with the five-fold di-
rections of the quasicrystalline substrate. In this orientation Cu atoms are
arranged with the fcc nearest neighbour distances running in-plane both par-
79
allel (y) and perpendicular (x) to the strip pattern. This gives rise to 3 atomic
rows in the short strips (width 4.561 A) and 5 atomic rows in the long strips
(width 7.379 A). In order to further improve the fit, the atomic separations
were varied in all three dimensions (x, y and z ). In the best fit model the
separation of the Cu atoms parallel to the strip pattern (y) was increased
from 2.56 A to 2.71 ± 0.02 A. Whilst the separation perpendicular to the
strip pattern (x ) remained at bulk values, the separation in height (z ) was
adjusted from 1.81 A to 1.67 A to maintain film density at the bulk value.
The blocking curve from the annealed film (Fig 5.3 (b)) was best approx-
imated by simulating a curve from five domains of fcc material with (110)
planes perpendicular to the surface and <110> azimuths oriented along the
five fold directions of the substrate. In order to achieve a good fit 6.9 ± 2.1
% strain perpendicular to the surface was also required. Due to the low per-
centage of Cu in this alloy film, the statistics are less good for this fit. The
goodness of fit between any simulation and the data were determined using
IGOR macros based upon a MEIS R-factor [82].
5.4 Discussion
5.4.1 The unannealed Cu film
The angular data were modelled using the VEGAS code. Since the Cu film
forms in five structurally equivalent domains (as demonstrated by the pre-
vious STM studies) a single domain model could be constructed and sim-
ulations run in five azimuthal rotations and averaged to build up the final
spectrum. Because a pentagon has five-fold symmetry along the appropriate
axes, orientations that are opposite according to the line of symmetry can be
considered equivalent, reducing the number of simulations needed to three
per structural model. The interaction of the beam with the model could then
be simulated for three inequivalent azimuthal orientations and the blocking
curves averaged with 1:2:2 weighting to provide the final simulated spectrum.
An example of the three component blocking curves making up the best fit
80
70 80 90 100 110 120
Scattering angle (degrees)
DataSimulation
(a)
(b)
(c)
Inte
nsity
(counts
)
5400
4700
4000
600
500
400
360
325
290
5.5
5.0
4.5
1.8
1.5
1.2
5.0
4.5
4.0
Illu
min
atio
n (
laye
rs)
Figure 5.3: (a); One-dimensional angle-resolved data obtained from Cu/i -Al70Pd21Mn9 (b); annealed to 600 K for 10 minutes. (c); Clean surface Pddata for i -Al70Pd21Mn9 [82].
81
model is shown in the inset of figure 5.6.
Different dimensions of the model were considered to optimise calculation
time without sacrificing accuracy taking into account the fact that the com-
putation time directly scales with model size. A previous treatment of clean
surface MEIS data [82] involved the use of a large (∼4, 000 atom) cuboidal sec-
tion of the accepted bulk icosahedral structural model as an approximation
to the i -Al70Pd21Mn9 surface. A model of comparable size was not considered
necessary in this case, as the copper does not exhibit icosahedral structure,
but is aperiodic in one dimension only. Hence the model is extended along
only one dimension and, as explained above, rotated through pentagonal an-
gles in order to build up a spectrum for the five observed domains. The
ideal length of the model in the aperiodic direction (y) was deemed to be
around 100 A, since below this value varying the length altered the character
of the simulated curve. The height of the model in the direction perpendic-
ular to the surface (z ) was 9 MLE, according to our measured average film
thickness assuming the film density to be the same as that of bulk copper.
The length of the model parallel to the strips (y) was equal to the repeat
distance, which in the case of an unstrained model would be the bulk nearest
neighbour distance (2.56 A).
The strip structure of the copper film was apparent from the initial STM
work on the system [57], and the separations reported for the long and short
segments (7.3 A and 4.5 A) were in good agreement with separations of a
number of features in a viable termination of the ideal bulk structure, in-
cluding small pentagonal Al features and Mn atoms. Hence the positional
accuracy of the separations was extended to that of the clean surface model,
using in this case, the positions of Mn atoms. The ‘dark stars’ apparent in
STM images of the clean i -Al70Pd21Mn9 surface also have similar separations
and have been shown to be nucleation sites for other adsorbates [51]. The
theoretical work of Krajcı et al. [83] also suggests that these may be can-
didate sites for the nucleation of the copper structure. Figure 5.4 shows an
STM image of the clean surface with the Fibonacci separations of the ‘dark
star’ features indicated and a section of the proposed copper film structure
superimposed on the surface. Even if the ‘dark star’ features are responsible
82
for nucleation of the copper film, only a few atoms in the proposed structure
sit in well-defined sites and hence the MEIS data would be relatively insensi-
tive to this since no new features would be expected in the surface blocking
curves of the substrate.
Fitting of the blocking curves was carried out initially considering a struc-
ture with fcc(110) termination within the strips, since this orientation has
been observed previously in cubic films found on quasicrystal surfaces [84]
(indeed matches that of the film formed on annealing the copper structure).
Efforts in this direction were unsuccessful and hence we tried the (311) ori-
entation (also observed when sputtering i -Al70Pd21Mn9) and close packed
fcc(111) and hcp(0001) orientations. However, none of these structures gave
a reasonable fit to the experimental data, which could only be achieved us-
ing an fcc(100) termination within the strips. Thus the MEIS data provides
confirmation of the structure independent of the previous STM or LEED
studies although of course it is in agreement with these previous studies
which showed a step height of 1.9 A(STM) and nearest neighbour separation
consistent with a (100) termination (LEED).
In order to improve the fit between the simulated and experimental data
calculations were performed for more complex models including one with∼14, 000 atoms which included two domains (shown in figure 5.7). This
work demonstrated that the number of atoms near the domain boundary is
not large enough to produce any additional blocking features in the data.
In addition, the fits achieved using smaller models all required significant
amounts of disorder of around 30% to be added (using our previously estab-
lished methodology [82]), and one potential reason for this was thought to
be the effect of domain boundaries. However, these larger model simulations
indicated that domain boundaries could not account for all the disorder seen,
since these models did not produce a significantly different blocking pattern
to the smaller models.
The previous LEED analysis [58] yields a Cu nearest neighbour distance
of 2.5±0.1 A. Although not in perfect agreement with our model that yields
2.71±0.02 A, the combined findings do provide strong evidence for this struc-
tural model. It is noted that in the stereographic projection for the fcc(100)
83
Figure 5.4: Figure reproduced from Ref. [51] showing the pentagonal ‘darkstars’ imaged on the five-fold surface of i -Al70Pd21Mn9, and including a one-dimensional Fibonacci grid joining like points on the surface to form a rowstructure with separations comparable to those of the unannealed copperfilm. Superimposed is a portion of the proposed copper structure.
84
2.71 A
15.1 A
101.56 A
L S L L S L L SL SL L S L
x
y
z
S L
x
z
y(along rows)
Figure 5.5: The best-fit model for Cu/i -Al70Pd21Mn9 as run. Also shown isan extension of the model into one 100 A × 60 A × 15 A domain.
surface (reproduced in figure 5.6), there is a line of high symmetry at 71.57◦.
An extension of 2.5% in one direction of the lattice allows this line to fall at
72◦, which is a pentagonal angle. Indeed this orientation is responsible for
the primary features seen in the blocking curve as demonstrated by the three
individual curves that make up the best-fit model which are also shown in
the figure. This supports the idea of an expansion rather than a contraction,
and for copper would mean a lattice expansion from 2.56 A to 2.62 A; we
detect a lattice expansion of 5% from 2.56 A to 2.71 A. Reducing the spacing
along the strips leads to a splitting of the blocking dip at around 80◦ which
is inconsistent with the data. In order to maintain film density the spacing
parallel to the surface (z ) has been adjusted whilst the value perpendicular to
the strips remained at bulk values. The lack of strain parallel to the surface
and perpendicular to the strips is perhaps not surprising since the breaks
in the structure that occur at the strip edges could be viewed as a strain
relieving mechanism.
Parallels with Ag/GaAs(110), a previous study
In a previous study, it was found that Ag can be deposited onto GaAs(110)
to produce a film which exhibits quasiperiodic characteristics [85]. In this
study, Ag was deposited at low temperature (135 K) to form an overlayer
composed of Ag nanoclusters with a narrow size distribution ( 2 nm). Upon
85
Figure 5.6: Part of the stereographic projection for fcc, with pentagonalangles superimposed. Inset : The three components of the best fit model. Theblocking dips may be easily identified with directions on the stereographicprojection.
86
annealing to room temperature, the Ag overlayer reforms into a perfectly flat
thin film. Pits in the film extend 15 A down to the substrate, indicating that
15 A is the critical thickness for this film, as further evidenced by another
experiment in which the work shows that for a dosage significantly lower
than would be sufficient for a complete film, the film forms in interconnected
islands with only 2 characteristic heights of 11 A and 15 A.
The nominal structure for this overlayer is close-packed (111), but there
are surface modulations which cause stripes across the surface consistent with
a silver-mean quasiperiodicity. The silver mean is given by:
δS = 1 +√
2 ≈ 2.4142136... (5.1)
and the stripes are separated by spacings of ∼13 A and ∼17 A.
At first glance this system seems remarkably similar to the Cu/i -Al70Pd21Mn9
film, but there are several important differences. The first is that the length
scale of the modulation in our case is somewhat smaller (a factor of ∼3).
Also, the quasiperiodicity observed in the case of the Ag/GaAs system is
a ‘silver mean’ quasiperiodicity, which has not been observed in quasicrys-
talline systems, whereas that of the Cu/i -Al70Pd21Mn9 film is a golden mean
quasiperiodicity, which is inherent in all five-fold systems. Another difference
is the registry between the strips, which in the case of the Ag/GaAs system
is periodic, but in the current study has been modelled using random off-
sets. These offsets could be interpreted as a reflection of the pseudo-random
positioning of the ‘dark star’ nucleation sites on the surface. However, any at-
tempt to model this quasiperiodic ordering, either by using offsets generated
from positions of Mn atoms in an ideal clean model surface or interpenetrat-
ing Fibonacci grids oriented at 72◦ to each other failed to improve the fit.
The possibility of these offsets being truly random cannot therefore be ruled
out. Finally, we do not observe a critical thickness for the Cu film — it grows
in a layer-by-layer fashion with each layer less complete than those before.
87
5.4.2 The annealed Cu film
From the energy spectra presented in figure 5.2 it is clear that the total
amount of Cu in the film has been dramatically reduced after annealing to
600 K for 10 minutes. The composition and thickness data shown in Table
5.1 demonstrate that the total Cu content in the quaternary alloy film has
dropped from 18× 1015 atoms cm−2 to 1× 1015 atoms cm−2, a loss of 95%.
Since the annealing temperature was relatively low the vapour pressure of Cu
would be negligible and re-evaporation can be ruled out as a mechanism for
the loss. The fit presented in figure 5.2 (b) includes Cu running through the
near surface region at a level of approximately 3%. However, the sub-surface
signal between 75 and 85 keV has contributions from Mn, Cu and Pd. Whilst
the Pd contribution is fixed by the signal seen in front of the Cu peak there is
a great deal of uncertainty in the relative proportions of Mn and Cu present.
Nonetheless, the most likely mechanism for Cu loss from the surface film is
thought to be via diffusion into the bulk.
The angle resolved data presented in figure 5.3 (b) clearly demonstrate
the transformation of the film into five domains of cubic material with (110)
termination. In fact, this termination is commonly seen for crystalline phases
formed on the five-fold surface of i -Al70Pd21Mn9 after sputtering[10,19]. Both
cubic close packing and icosahedral structure are densely packed arrange-
ments. The (111) cubic surface and the icosahedral three-fold surface in-
volve triangular arrays of atoms which, when aligned together lead to a near
coincidence of cubic (110) planes with icosahedral five-fold axes and hence,
this is a favoured epitaxial arrangement of the two structures. This epitaxial
arrangement is discussed in considerably more detail by Shen et al. [84].
In modeling the angle resolved MEIS data of figure 5.3 (b) two parameters
were adjusted to optimise the fit. The first of these was disorder, which was
accounted for by mixing the simulation with flat signal, a procedure that
has been successfully used previously [10]. The resulting fit of 6.7 ± 5.9 %
implies a high level of crystallinity in the quaternary alloy film. The other
parameter incorporated into the fit was perpendicular strain with the best fit
arising for a value of 6.9 ± 2.1 %. The origin of this strain is not clear since
88
no detailed understanding of the interface between the film and substrate
has been achieved. However, in the case of the AlPd(110) film formed when
i -Al70Pd21Mn9 is sputtered, which has a lattice parameter of 3.03 A, no
strain is observed [82]. If the interface of this quaternary alloy phase with
the icosahedral substrate is basically similar to the sputtered phase then the
fitted strain would imply a larger lattice parameter of approximately 3.1 A.
Whilst this value is smaller than the lattice parameter of any of its constituent
pure metals this is not uncommon for metal alloys as demonstrated by the
lattice parameter of AlPd which is itself somewhat smaller than that of either
Al or Pd in elemental form.
5.5 Conclusions
The structure of a pseudomorphic thin copper film grown on the five-fold sur-
face of i -Al70Pd21Mn9 has been investigated. It has been found to consist of
five rotational domains of a one-dimensional Fibonacci array of strips whose
widths are τ -scaled distances characteristic of the quasicrystal surface (L =
7.379 A and S = 4.561 A). The strips are comprised of fcc material with the
(100) direction oriented perpendicular to the surface and nearest neighbour
distances running both parallel and perpendicular to the strip pattern. The
fcc material is distorted so that the nearest neighbour distance along the
strips is 2.71 ± 0.02 A, with a corresponding compression perpendicular to
the surface resulting in a step height of 1.67 ± 0.02 A. Whilst the period-
icity and structure type is in good agreement with that previously deduced
from two-dimensional LEED measurements, this study represents the first
full structural characterisation of this copper film.
Annealing the room temperature grown copper led to the diffusion of Cu
into the substrate leaving a quaternary alloy at the surface. This alloy had
fcc(110) structure similar to that found in the fcc AlPd film formed when
i -Al70Pd21Mn9 is sputtered.
89
Figure 5.7: The large two-domain model for Cu/i -Al70Pd21Mn9 with an STMimage [57] superimposed.
90
Chapter 6
Adsorption of cobalt on the
ten-fold surface of
d-Al72Ni11Co17 and on the
five-fold surface of
i -Al70Pd21Mn9
6.1 Introduction
6.1.1 Practical magnetism
The origin of magnetism in materials is due to effects from uncompensated
spin in atomic systems. Because there are two directions of spin, spin-up and
spin-down, and also because in some systems, called ferromagnetic systems,
these uncompensated spin orientations persevere after the initial influence of
an external magnetic field has been removed, there has been an explosion in
91
the use of this phenomenon recently as a way of storing information.
By depositing a thin layer of ferromagnetic material on a substrate, one
can create a system in which one can store information by inducing differ-
ent magnetization directions in domains present in the film. The need for
ever greater storage densities, driven by the information industry, results in
the necessity of optimizing this kind of information storage system as far as
possible. The limiting factor in magnetic storage technology is the physical
size of each bit of information on the surface of the medium, given that each
bit has to have a near perfect chance of retaining its magnetic polarization
as long as required. A few variables affect this quantity, such as the rate
at which domains become demagnetized due to the opposing magnetic fields
surrounding them and also the orientation of each bit of information. One
of the latest breakthroughs in magnetic storage technology is so called per-
pendicular recording, in which a thin diamagnetic layer is deposited between
two thin ferromagnetic layers deposited on a surface, the result of which is to
induce an opposing magnetic moment in the lower ferromagnetic film, which
serves to ‘lock’ the surface magnetic domain orientation with less surface
space needed for each domain [86].
6.1.2 Physical motivation
The future of magnetic technology depends on the extent to which we can
increase our knowledge of the magnetism phenomenon, and this in turn de-
pends on how we can test our theories against practical evidence. In this
chapter the formation of a pseudomorphic Co layer on the ten-fold surface
of d -Al72Ni11Co17 at 300 K and the formation of a less well-ordered Co thin
film on the five-fold surface of i -Al70Pd21Mn9 are reported. The quasiperiodic
structure of the films is evidenced in low-energy electron diffraction (LEED)
patterns, and in part from scanning tunneling microscopy (STM) measure-
ments taken from the films. Knowledge of the local atomic structure of these
films is an essential precursor to subsequent magnetic studies.
92
6.2 Experimental Details
Using an Omicron EFM-3 e-beam evaporator, described in Chapter 3, Co
was deposited onto the ten-fold d -Al72Ni11Co17 surface at a rate of 6×10−3
ML sec−1 and onto the five-fold surface of i -Al70Pd21Mn9 at the same rate
(as determined by STM coverage measurements and by a comparison of the
ion flux as measured by the evaporator). All deposition was carried out with
the substrates at a temperature of 300 K. Scanning tunneling microscopy,
LEED and AES measurements were taken from the resulting films. The
error on the coverages quoted for Co/d -Al72Ni11Co17 is estimated at 5% and
for Co/i -Al70Pd21Mn9 it is estimated at 20%.
6.3 Results
6.3.1 Scanning tunneling microscopy and Auger elec-
tron spectroscopy
Figure 6.1 shows images taken from the Co/d -Al72Ni11Co17 film at room tem-
perature with an Omicron variable-temperature STM. No order is discernible
at 0.4 ML coverage (determined by STM and checked with AES).
At higher coverages (5.6 ML and above) evidence for a row structure of
the film is observed. It can be seen from the emergence of straight line domain
boundaries in the STM images shown in figure 6.1 that certain orientations
are preferred for the formation of islands of Co. At higher coverages, some
one-dimensional row structuring within these islands becomes apparent. Due
to limitations in resolution, it cannot be shown conclusively from STM that
the inter-row spacings are arranged in a Fibonacci sequence, as has been
previously observed for Cu adsorption on the five-fold i -Al70Pd21Mn9 surface.
The step height of the film as determined from STM is 2.0 ± 0.1 A, which is
consistent with the interlayer separation of hcp Co(0001) previously observed
with STM to be 1.95 ± 0.10 A [87]. The fact that the Al/Co Auger peak
ratio drops to zero with increasing coverage implies that there is no alloying
93
Figure 6.1: (a-g); Comparison of the ten-fold d -Al72Ni11Co17 surface as stud-ied with increasing Co coverage. The coverage is respectively 0.4 ML, 1.9ML, 3.7 ML, 5.6 ML and, for (e-g), 7.5 ML . (f) is a detail of (e), with therow structure showing more clearly. Indications of orientational order aresuperimposed on (e-g). A graph showing the trend of Auger peak area ratiosvs coverage is also shown.
94
Figure 6.2: (a-d); Comparison of the five-fold surface of i -Al70Pd21Mn9 asstudied with increasing Co coverage. Coverage is (a) 0.4 ML; (b) as (a)with five-fold depressions highlighted; (c) 2.5 ML coverage, and (d) 13 MLcoverage.
within the film.
Scanning tunneling microscopy images taken from the Co/i -Al70Pd21Mn9
film revealed no definite order, at any coverage (determined by STM and
checked with AES), though at higher coverages there is a suggestion of linear
structures within the incomplete layers formed. Auger electron spectroscopy
again implies that there is no alloying within the film. Figure 6.2(a,b) indi-
cates that the dark five-fold hollows are not a preferred nucleation site for
this adsorbate (as they have been found to be in previous studies [51]) as they
may still be clearly observed when there is a coverage of 0.4 ML Co on the
surface, with a density of approximately half that found on the clean surface.
95
Figure 6.3: Low-energy electron diffraction patterns from 3 systems at 2beam energies for each system. From left to right, the coverages are: 20 ± 4ML Co, 7.5 ± 0.4 ML Co, 4 ± 0.2 ML Cu.
The size of the five-fold hollows (around 8 A) indicates that they belong to
the substrate. Coverages for this film are a guide only, as the growth mode
results in a film consisting of incomplete layers.
6.3.2 Low-energy electron diffraction
The LEED patterns obtained from the multilayer films show more evidence
for ordering. Figure 6.3 shows the patterns obtained at 2 different electron
energies from the Co/i -Al70Pd21Mn9 film, the Co/d -Al72Ni11Co17 film (both
at room temperature) and previously obtained patterns from the pseudomor-
phic Cu/i -Al70Pd21Mn9 film [57] (at 85 K). The patterns on the right have
been corrected for the different acceptance angle and size of the optics so
they may be compared with the others. Each coverage quoted provided the
optimum LEED pattern in terms of lowest background and clearest diffrac-
tion peaks. It can be observed, despite the lower quality of the patterns,
96
Figure 6.4: (a); LEED pattern obtained from the clean surface at 120 eV. (b);LEED pattern from a 7.5 ML Co overlayer at 120 eV. Lines are superimposedto aid identification of the periodic separation of the streaks.
that the Co/i -Al70Pd21Mn9 diffraction is qualitatively similar to that for
Cu/i -Al70Pd21Mn9, and the Co/d -Al72Ni11Co17 patterns, apart from having
a different set of momentum transfers, show the same broad characteristics
as both.
In diffraction obtained from the Cu/i -Al70Pd21Mn9 film, the periodic ar-
rangement of the streaks was attributed to a periodic ordering along the
Fibonacci rows, and a 2D LEED analysis was carried out to determine this
periodicity. Although, in this case, Fibonacci ordering may not be directly
observed from STM of the Co/d -Al72Ni11Co17 film, the similarity of the
diffraction patterns leads us to conduct a similar investigation into this sys-
tem.
Using the PhD thesis of Sharma [88], a diffraction peak of known mo-
mentum transfer (k = 2.67 A−1), indexed as (1221), was identified in the
d -Al72Ni11Co17 diffraction pattern shown in figure 6.4 (a). The momentum
transfer k′ of the streaks in the Co/d -Al72Ni11Co17 film was hence calcu-
lated from the ratio of the line length y/x using k′ = k.y/x and was found
to be 2.56 A−1. Hence the periodic spacing along the rows is found to be
97
2π/k′ = 2.5 ± 0.1 A, a figure which corresponds well to the nearest neighbour
distance in the stable hcp phase of cobalt of 2.51 A.
6.4 Discussion
On the five-fold surface of i -Al70Pd21Mn9, there is no suggestion of order in
the STM images until a relatively high coverage of Co is reached and even
then it is not well resolved. This order is also not apparent in diffraction
patterns taken from the film until similar high coverages are reached. The
growth as determined by STM is most closely approximated by the layer-by-
layer model, though the layers are rough, and the island size small (tens of
angstroms).
The Co/i -Al70Pd21Mn9 film produces a diffraction pattern that, within
the error introduced by the correction for optics, and indeed the quality of the
patterns themselves, is extremely similar to that from the Cu/i -Al70Pd21Mn9
film. Assuming the streaks observed in the pattern from the Co film indicate
periodicity, as they do for the Cu film, then it may be concluded that they
correspond to the same periodicity, as far as is permitted within the accuracy
of such an analysis. The NN distance in bulk fcc Cu is 2.56 A. Our LEED
analyses of the Cu film and of the Co films yield a periodicity along the rows
of 2.5 ± 0.1 A. This kind of analysis has been used to assist the assertion
elsewhere [58] that the structure of the Cu phase present on i -Al70Pd21Mn9
consists of five orientational domains composed of rows of atoms arranged
according to a one-dimensional Fibonacci sequence.
On the ten-fold surface of d -Al72Ni11Co17, deposited Co appears to form
an initial complete layer upon which smaller domains of row-structured ma-
terial take shape. With each subsequent layer, the lateral island size de-
creases. The growth is therefore neither strictly layer-by-layer nor 3D, though
layer-by-layer growth is the closest approximation. At the highest coverages
studied (15ML), the size of the Co islands was on the order of hundreds of
angstroms.
The streaks present in the diffraction pattern from the Co/d -Al72Ni11Co17
98
Lattice NN distance Interlayerparameter (A) (A) spacing (A) (STM)
hcp Co −− 2.51 1.95fcc Co 3.56 2.52 1.78 (001)bcc Co 2.82 2.44 1.41 (110)fcc Cu 3.62 2.56 1.81 (001)
Lattice parameter Lattice parameteron i -Al70Pd21Mn9 (A) on d -d -Al72Ni11Co17 (A)
Co 2.5(3) 2.4(6)Cu 2.5(3) −−
Table 6.1: Table showing certain structural properties of copper and cobalt[89,90]
film can immediately be seen to correspond to a smaller period (due to the sig-
nificantly larger separation of the streaks on the pattern). The NN distance
in bulk hcp Co is 2.51 A. Our LEED analysis of the Co/d -Al72Ni11Co17 film
yields a periodicity along the rows of 2.5 ± 0.1 A. The momentum transfer
extracted for the separation of the streaks on the Cu/i -Al70Pd21Mn9 was 2.48
A−1 [58]. To the extent that the values of the momentum transfers derived
from the bulk structures are correct for these surfaces, it can be seen that the
difference in the momentum transfers for the two films is (2.56/2.48 = 1.03)
and this translates to a difference in lattice parameter of 3%. It can thus be
determined that the periodicity for Co on i -Al70Pd21Mn9 is 3% larger, in one
dimension, than that on d -Al72Ni11Co17. Bulk fcc Cu has a NN distance 2%
larger than that of bulk hcp Co. Table 6.1 lists these properties.
We suggest that one reason behind the small domain size of the Co/d -
Al72Ni11Co17 film compared to the Cu/i -Al70Pd21Mn9 system is that of the
strain within the film. Although strain-relieving features were identified in
the Cu/i -Al70Pd21Mn9 film [57], no such features are observed in the STM
images from the Co film, at any coverage, perhaps due to the lower resolution.
99
6.5 Conclusions
When deposited onto the ten-fold surface of d -Al72Ni11Co17, Co grows in
a predominantly layer-by layer fashion, forming a film with a number of
orientational domains. These domains have a row structure visible in STM
images, and are shown by a LEED analysis to be consistent with a one-
dimensional Fibonacci arrangement of rows, with a periodic arrangement of
atoms perpendicular to the Fibonacci direction. The nearest neighbour (NN)
distance along these rows has been determined to be 2.5 ± 0.1 A, and has
also been shown to be smaller than the periodicity of the Cu/i -Al70Pd21Mn9
structure by around 3%. This is consistent with the NN distance in bulk hcp
and fcc Co; however, due to the interlayer spacing observed by STM, it is
suggested that the structure adopted by the Co on this surface is composed
of rows of hcp structured material with the (1000) hcp orientation along the
rows. It has also been observed that a film of Co deposited on i -Al70Pd21Mn9
appears to have the same broad structure and periodicity as a film of Cu
deposited onto the same surface. However, the Co film is much less ordered
than the Cu film on this surface, and also less ordered than the Co film
deposited on d -Al72Ni11Co17.
In the regime of pseudomorphic growth, there have been to date three
multilayer films formed on quasicrystal substrates: the copper film on i -
Al70Pd21Mn9 [57] and the studies under discussion. Although the struc-
tural properties of the two substrates are very different (i -Al70Pd21Mn9 is
an icosahedral quasicrystal with three-dimensional aperiodicity, whereas d -
Al72Ni11Co17 is periodic along one direction and decagonal and aperiodically
ordered perpendicular to that direction), common features emerge in the
films deposited: the atoms order themselves in row structures according to
orientations present within the substrate, and these row structures are them-
selves ordered according to a one-dimensional Fibonacci sequence.
100
Chapter 7
Surface magnetism of
quasicrystals and thin films
deposited thereon
7.1 Introduction
The ability to tailor the magnetic properties of a surface is very attractive
to the information industry; hard disks with an information density exceed-
ing 100 Gbits/sq.in. are now commonplace [91], and the need for greater
information density is impending. The identification of materials with novel
magnetic properties is thus a fertile research area.
It may be expected that the deposition of thin magnetic films on non-
magnetic surfaces would allow the study of low-dimensional magnetic nanos-
tructures. The interfaces between the magnetic films and the non-magnetic
surfaces may present interesting magnetic induction effects, allowing the
study of the magnetic properties of single atoms. This kind of approach
has led to breakthroughs in the understanding of, for example, the magnetic
behaviour of Ni as it varies for different kinds of film and for the bulk [92].
101
The physical structure of magnetic films may be expected to have an
influence on their magnetic properties — for example, an anisotropy in the
magnetic susceptibility may be due to the varying density of magnetic atoms
in a crystal plane, or on a larger scale, the nanostructuring of a film itself
— a large anisotropy has been observed for a magnetically soft FeTaN film
deposited using magnetron sputtering to incorporate Ta in a bcc FeN lat-
tice [93], and this has been attributed to the columnar structure of the film
obtained.
This chapter summarises non-quantitative results from an x-ray magnetic
circular dichroism (XMCD) experiment carried out in the European Syn-
chrotron Research Facility in Grenoble (ESRF). These results relate to the
anisotropy of magnetic transition metal films deposited on quasicrystals, and
also demonstrate, for the first time, that such films may induce a magnetic
response in the magnetically frustrated atoms near the quasicrystal surface.
7.2 Results
7.2.1 Icosahedral Al70Pd21Mn9
Preparation conditions: clustered and flat surfaces
In our MEIS study of various preparation conditions for clean i -Al70Pd21Mn9,
Noakes et al. provide valuable insights as to how the quasicrystal surface
evolves according to different preparation conditions [82].
The sputtering process causes preferential depletion of Al atoms in the
surface due to the increased energy transferred to lower mass target atoms,
thus shifting the surface composition away from the quasicrystalline region
in the phase diagram [31]. This may result in various types of multi-domain
cubic structure at the surface. Manganese is also depleted, and so the cubic
structure is composed of mainly an AlPd alloy with (110) planes oriented
parallel to the five-fold axes of the substrate [82].
For annealing temperatures up to 800 K, a clustered surface largely equiv-
102
alent to the cleaved surface may be obtained [34, 94–99]. This surface is
the most well-ordered surface obtainable by sputter-annealing cycles, with a
composition extremely close to that of the bulk, but with Al and Mn content
slightly reduced due to the comparatively higher vapour pressures of these
elements [100].
Annealing to higher temperatures (> 920 K)causes multiple effects. The
pseudo-Mackay icosahedron clusters at the surface break down to present
a flat terraced five-fold surface with dark five-fold hollows [35]; this is the
surface that presents the five-fold structure most clearly.
However, for multi-element systems, the differing diffusivities and vapour
pressures between elements may be expected to play a part during annealing,
and it is found that for the high temperatures and long anneal times neces-
sary to obtain the flat terraced phase of i-Al70Pd21Mn9, a good deal of Mn
segregation takes place, resulting in Mn-rich polycrystalline or amorphous
domains in the surface where the composition has shifted too far from the
quasicrystalline region in the phase diagram [82].
7.2.2 Magnetism and ordering: XMCD
Manganese is the only magnetic element in the AlPdMn quasicrystal. Pre-
vious theoretical studies [39] yield the result that of the manganese present,
only a very small amount ( 1%) has an appreciable magnetic moment, though
those atoms that do have quite a large moment. No direct Mn-Mn coordi-
nation is present in the magnetic Mn atoms, though the presence of a Curie
transition implies that there must be a correlation between Mn atoms at
larger separations [101]. Only two equivalent sites in the 3/2 approximant
studied theoretically by Krajcı et al. [39] exhibited appreciable magnetic be-
haviour.
The dichroic signal from clustered (annealed to 800 K) Al-Pd-Mn is very
small, consistent with the ideas presented above. Flat-terraced Al-Pd-Mn
exhibits a larger magnetic response. This phenomenon is illustrated in figure
7.1. The marked broadness of the peak from the flat terraced phase com-
pared to that from the clustered phase indicates the larger variation in the
103
electronic environment of the Mn atoms, due to the multiplicity of structural
environments present.
7.2.3 Easy and hard magnetization directions: XMCD
A comparison of the magnetization of the Mn atoms for semi-in-plane (60◦
to the normal) and out-of-plane (normal) fields indicates that there is easy
magnetization in the direction with a component in-plane. The dichroic
responses for the clustered phase are shown in figure 7.2.
Figure 7.3 shows that there is less anisotropy between in-plane and out-
of-plane magnetization directions for the terraced phase of i -Al70Pd21Mn9.
7.3 d-Al72Ni11Co17
7.3.1 Preparation conditions
Studies intended to quantitatively and accurately determine the quality of
the surface structure of d -Al72Ni11Co17 as a function of annealing tempera-
ture do not go so far in the temperature scale as to include the maximum
of the ‘surface quality’ vs ‘temperature’ curve, as they do in the case of
i -Al70Pd21Mn9 [82, 102]. However, it may be expected that the tradeoff
between flat terraced phases and overall surface structural fidelity is not
present to as great an extent, due partly to the fact that d -Al72Ni11Co17 is
a two-dimensional quasicrystal, and thus has no propensity for formation of
a clustered phase; or at least, a clustered phase with an inherently rough
surface. Nickel and Co are adjacent in the Periodic Table and therefore are
not preferentially sputtered, although Al is depleted from the surface dur-
ing preparation. The surface transforms to a multiple domain bcc structure
upon annealing [103], but the composition of this layer is not known.
104
Figure 7.1: Manganese x-ray magnetic circular dichroism signals for flat andclustered i -Al70Pd21Mn9.
105
Figure 7.2: Clustered i -Al70Pd21Mn9 Mn dichroism signals for in-plane andout-of-plane field.
106
Figure 7.3: Terraced i -Al70Pd21Mn9 Mn dichroism signals for in-plane andout-of-plane field.
107
Figure 7.4: (a) Monte Carlo simulations of the magnetic microstructure ina sample of finite size for pure dipolar interaction. The microstructure hasbeen obtained for a square sample of about 10,500 vector spins on the Pen-rose tiling. The spins belonging to the perimeter of decagons (marked) formclosed chains. (b) Experimental model. The perspective view of the mag-netic microstructure. The red arrows represent the orientation of dipolarmoments of magnets fixed onto the nodes of the Penrose tiling (rhombuses).The magnets can rotate in the horizontal plane. Reproduced from ref. [40],copyright APS 2003.
7.3.2 Magnetism
Although nearly 30 atomic percent of d -Al72Ni11Co17 is composed of mag-
netic species, it does not exhibit any magnetic response whatsoever. This can
be understood in terms of magnetic frustration as explained by Vedmedenko
et al. [40]. Their theoretical paper treating arrays of magnets arranged ac-
cording to a Penrose tiling indicates that magnetic moments orientate in two
structures: ordered decagonal rings and frustrated disordered spin glass-like
structures within the rings, resulting in no net moment, as shown in figure
7.4. The Co and Ni edges are shown in figure 7.5; within the resolution of
the technique, no dichroism may be observed.
108
Figure 7.5: XMCD spectra of d -Al72Ni11Co17 Cobalt and Ni edges. There isno observable dichroism.
109
7.4 Deposited Fe film on d-Al72Ni11Co17
In figure 7.6, some dichroism from the Co may be observed. This is inter-
face magnetism induced by the magnetic response of a 3 ML (determined
by a quartz crystal microbalance thickness monitor) Fe film deposited on
the surface using an Omicron EFM 3 e-beam evaporator. A corresponding
dichroism for the Ni present may also be observed, though the data for this
is not presented.
The dichroic response from the Fe film may be observed in figure 7.7.
A comparison is made for the response due to a magnetic field in the in-
plane (60◦ to the normal) and out-of-plane (normal) directions. The easy
direction of magnetization is in-plane, in accordance with the precedent set
by the reported easy magnetization direction of the columnar structure in
the FeTaN/FeN system referred to above.
7.4.1 Hysteresis curves
Diagram 7.8 shows absolute intensity hysteresis curves for the Fe/d -Al72Ni11Co17
film. Both sets of data were taken with the substrate at 5 K. The greater
hysteresis observed for the in-plane case is consistent with the larger dichroic
response observed, and suggests that the film locks into a state of magnetic
order more readily for an inducing field parallel to the surface plane. The
fact that hardly any hysteresis is observed for the out-of-plane case implies
that for an inducing field perpendicular to the surface plane, the film exhibits
almost purely paramagnetic behaviour, presumably with the randomizing in-
fluence being the state of magnetic frustration experienced at the surface/film
interface.
7.5 Summary
This study has not produced any surprises; we have found that clustered i -
Al70Pd21Mn9 has a relatively small set of possible environments of Mn atoms
compared to the terraced phase. We attribute this to the higher level of
110
Figure 7.6: Induced Co in d -Al72Ni11Co17 moment with a 3 ML adsorbed Fefilm.
111
Figure 7.7: In-plane and out-of-plane dichroism for Fe deposited on d -Al72Ni11Co17.
112
-0.3
0.0
0.3
Fe hysteresis OOPFe/AlNiCo
-0.2 0.0 0.2
-0.4
0.0
0.4
Fe hysteresis IPFe/AlNiCo
Magnetic field (T)
XM
CD
effect (a
rb. units)
Figure 7.8: In-plane and out-of-plane hysteresis curves for Fe deposited ond -Al72Ni11Co17.
113
disorder (on the macroscopic scale) present in samples that have undergone
this preparation, as previously reported in the MEIS study on the subject of
different annealing times and temperatures [82]. The larger dichroic response
for the terraced phase also lends weight to this idea, as small domains of
crystalline Mn within the sample may contribute to the magnetic behaviour.
This decrease in overall structural fidelity at the surface is due at least in part
to the use of higher annealing temperatures during preparation of the surface.
The use of higher annealing temperatures results in increased segregation of
Mn and it is also tentatively asserted that, due to the higher thermodynamic
stability of the five-fold surface, the formation of a flat terraced five-fold
surface is achieved by breaking down the two- and three-fold facets of the
natural cluster units that go to make up the structure of i -Al70Pd21Mn9.
The lack of dichroic response from the d -Al72Ni11Co17 sample is entirely
in keeping with what is known about the structure of this quasicrystal, and
also goes some way towards confirming that the flat terraced phase is the
most well-ordered phase of d -Al72Ni11Co17.
The thin transition metal films deposited on these quasicrystals exhibit
the presence of a dichroic response in all cases, though the difference in mag-
nitude of this response for different directions of inducing field shows the
anisotropy of magnetization present in these films. The response generally
appears to be greater for an inducing field with some component parallel to
the surface/film interface; behaviour in which an easy direction of magneti-
zation appears to lie along a plane of high density is not uncommon [93].
The appearance of an induced moment in the magnetic substrate atoms
for d -Al72Ni11Co17 when a film of Fe is deposited is the first time this phe-
nomenon has been observed, and, in showing the effect due to a ferromagnetic
deposited film, lends further credence to the idea of magnetic frustration of
the substrate surface suggested by Vedmedenko et al. [40].
Future work with this data could include the determination of the spin
and orbital components of the magnetization in all cases, along with a de-
termination of the moment per transition metal atom given that meaningful
approximations could be devised in order to conduct the necessary ab initio
calculations.
114
Chapter 8
Summary and suggestions for
further work
There is a common feeling within the quasicrystal field that there are now
only minor questions left relating to the structures of the two quasicrystals
treated in this thesis. However, it has been an interesting task to try to
decipher the structure of the films that we have grown on these surfaces.
The elucidation of the copper structure as a kind of balance between qua-
sicrystallinity and the natural Cu bulk form provides a stepping stone for
the understanding of how it may be easier for a phase to condense in an
aperiodic structure. Though in this case we have a template in the form of
the quasicrystal substrate, it has been demonstrated that this does indeed
have an influence that is not too weak, or too strong, to be effective. In-
terestingly, though not easy to analyze at this stage, it appears from some
of the diffraction from Co and Cu forming adsorbed pseudomorphic over-
layers that a periodicity present in such an overlayer may be enforced by
the substrate rather than the adsorbate. This is borne out in part by a talk
by Wolfgang Theis at the 9th International Conference on Quasicrystals, in
which he treats the idea that a there may indeed be a lowest energy state
associated with a particular periodicity on an aperiodic substrate. This is
extremely difficult to analyse mathematically, since in the transition from
115
one dimension to two (or three) dimensions the problem becomes non-linear
in nature. Further theoretical work in this direction could involve relaxing
a layer of copper on a large-unit-cell approximant using density functional
theory.
The successful adaptation of the MEIS technique to the analysis of com-
plex aperiodic structures is encouraging, and the knowledge gleaned from the
study presented in this thesis provides a firm stepping stone for the analysis
of future films deposited on quasicrystalline substrates. The use of LEED
to determine a lattice constant from the diffraction pattern of the aperiodic
structures formed by Co deposited on d -Al72Ni11Co17 demonstrates the use-
fulness of this technique even in the case in which an aperiodic system is
under scrutiny.
Although what has been learned from the magnetic studies reported in
Chapter 7 is largely confirmatory of what was already considered to be known
about these systems, they demonstrate that this technique is an effective
probe of the surface order. Although all the results presented in this section
are qualitative, further analysis of this data using the XMCD sum rules
could provide much more insight into the magnetic behaviour of all of these
systems.
Further experimental work in the field of simplifying quasicrystal forma-
tion needs to at least attempt to develop a way of predicting what kind
of growth mode an adsorbate will adopt. So far, we can define the follow-
ing modes: homogeneous growth, rotational epitaxy, pseudomorphism, and
possibly we may subdivide pseudomorphism into monolayer and multilayer
pseudomorphism to acknowledge the fact that the aperiodic multilayer films
we have formed do not truly conform to this definition. However, we may
now identify four observed unique growth modes. There is perhaps the pos-
sibility that a fifth single-element quasicrystal mode may exist, but it does
not seem to make much physical (or chemical) sense. We do not, however,
have any way of predicting which mode any adsorbate will adopt. One po-
tential experiment that springs to mind would be the adsorption of Pd onto
a fivefold surface of i -Al-Cu-Fe. Aluminium copper iron is an icosahedral
quasicrystal similar to Al-Pd-Mn except that Cu occupies Pd positions and
116
Fe occupies Mn positions in the structure. Therefore perhaps Pd on i -Al-
Cu-Fe will form a film similar to that formed by Cu on i -Al-Pd-Mn. This
would go some way towards confirming a dependence on the interplay of the
chemical (or valence) properties of the substrate and adsorbate.
117
List of Publications
Pseudomorphic Growth of a Single Element Quasiperiodic Ultrathin Film
on a Quasicrystal Substrate
J. Ledieu, J.-T. Hoeft, D.E. Reid, J.A. Smerdon, R.D. Diehl, T.A.
Lograsso, A.R. Ross, R.McGrath.
Phys. Rev. Lett., 92, 135507, (2004)
Compositional and structural changes in i-i-Al70Pd21Mn9 quasicrystals
induced by sputtering and annealing: A medium energy ion scattering study
T.C.Q. Noakes, P. Bailey, C.F. McConville, C.R. Parkinson, M. Draxler,
J.A. Smerdon, J. Ledieu, R. McGrath, A.R. Ross, T.A. Lograsso.
Surf. Sci., 583, 139, (2005)
Copper adsorption on the fivefold Al70Pd21Mn9 quasicrystal surface
J. Ledieu, J.T. Hoeft, D.E. Reid, J.A. Smerdon, R.D. Diehl, N. Ferralis,
T.A. Lograsso, A.R. Ross, R. McGrath.
Phys. Rev. B, 72, 035420, (2005)
118
Characterisation of thin aperiodic and periodic Cu films formed on the
five-fold surface of i-Al70Pd21Mn9 using medium-energy ion scattering
spectroscopy
J.A. Smerdon, T.C.Q. Noakes, J. Ledieu, C.F. McConville, M. Draxler,
T.A. Lograsso, A.R. Ross and R. McGrath.
Phys. Rev. B, accepted for publication, (2006)
Film growth arising from the deposition of Au onto an i-Al70Pd21Mn9
quasicrystal: a medium energy ion scattering study
T.C.Q. Noakes, P. Bailey, M. Draxler, C.F. McConville, T.A. Lograsso,
A.R. Ross, L.Leung, J.A. Smerdon and R. McGrath.
J. Phys.: Cond. Matter, 18, 5017-5027 (2006)
Adsorption of cobalt on the ten-fold surface of d-Al72Ni11Co17 and on the
five-fold surface of i-Al70Pd21Mn9
J.A. Smerdon, J. Ledieu, J.T. Hoeft, D.E. Reid, L.H. Wearing, R.D. Diehl,
T.A. Lograsso, A.R. Ross, R. McGrath.
Phil. Mag. 86, 841-847 (2006)
119
List of Figures
2.1 Only 2-, 3-, 4- and 6-fold symmetry can tile a plane. . . . . . . 5
2.2 Shechtman’s logbook, from 1982, documenting the first dis-
covery of a quasicrystalline phase in the Al75Mn25 alloy. The
term SAD refers to selected area diffraction, with the acronym
DF referring to the use of the dark field mode of the tech-
nique, where one diffraction beam is used in the production of
a transmission electron micrograph. . . . . . . . . . . . . . . . 6
2.3 (a) The project method. (b) The cut method. In the six-
dimensional analogue to this diagram, the ‘lines of finite length’
are replaced by two-dimensional ‘atomic surfaces’, and the way
the three-dimensional projection cuts them reveals atomic po-
sitions in the bulk structure. . . . . . . . . . . . . . . . . . . . 9
2.4 In a pentagram, the ratio of the areas a/b is equal to τ . . . . . 10
2.5 The daisy and pinecone are examples of plants that exhibit
Fibonacci ordering and τ -scaling. . . . . . . . . . . . . . . . . 11
2.6 The Parthenon is depicted, with τ -scaling relationships high-
lighted. Inset: τ and the human body. . . . . . . . . . . . . . 11
2.7 The Penrose P1 tiling. The four constituent tiles are the boat,
star, pentagon and rhomb tiles. . . . . . . . . . . . . . . . . . 12
2.8 The Penrose P3 tiling and the rhombs that make it up, with
matching rules shown. . . . . . . . . . . . . . . . . . . . . . . 14
120
2.9 Two competing models for the atomic decoration of the decago-
nal (2 nm) quasi-unit-cell for Al72Ni20Co8: (a); a model with
broken ten-fold symmetry and (b); an alternative model with
unbroken ten-fold symmetry but with accidental symmetry
breaking introduced in the central region due to chemical and
occupational (vacancy) disordering. Reprinted from ref. [26].
Copyright APS 2000. . . . . . . . . . . . . . . . . . . . . . . . 15
2.10 Schematic of an icosahedron, showing the six five-fold symme-
try axes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.11 A Mackay icosahedral cluster (51 atoms). The cluster is de-
fined by a central body-centred cube (9 atoms, radius 2.57
A), or a partially occupied dodecahedral shell, followed by an
icosahedral shell (12 atoms, radius 4.56 A), and finally an ex-
ternal icosidodecahedral shell (30 atoms, radius 4.80 A). . . . 18
2.12 A Bergman cluster (33 atoms). The cluster is defined by a
central atom, an icosahedral shell (12 atoms) and an outer
dodecahedral shell (20 atoms). . . . . . . . . . . . . . . . . . . 19
2.13 Scanning tunneling microscopy of a UHV-cleaved i -Al70Pd21Mn9
surface. The clusters are clearly visible. Reprinted from ref.
[34]. Copyright APS 1996. . . . . . . . . . . . . . . . . . . . . 20
2.14 Scanning tunneling microscopy of a flat i -Al70Pd21Mn9 sur-
face. Many dark five-fold hollows may be observed. Reprinted
from ref. [35]. Copyright Elsevier 2001. . . . . . . . . . . . . . 21
2.15 An STM image of a dark five-fold hollow, with LDOS (see
4.3.3) plotted on the z-scale to provide a 3D image. The dis-
tances indicated are consistent with the size of a truncated
Bergman cluster. Reprinted from ref. [35]. Copyright Elsevier
2001. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.16 (a); An STM image of the two-fold surface of d -Al-Ni-Co. (b);
An STM image of the ten-fold surface of d -Al-Ni-Co, with
some Fourier filtering applied. Figure adapted from figures
published in Refs [36,37]. Copyright APS 2002, 2004. . . . . . 24
121
3.1 Left ; an aircraft engine. Right ; the head of an electric shaver.
These are common examples of applications for coatings to
modify mechanical properties. . . . . . . . . . . . . . . . . . . 28
3.2 (a) Layer-by-layer growth; (b) Three-dimensional growth; (c)
Layer-by-layer reverting to island growth . . . . . . . . . . . . 29
3.3 Illustrating Young’s equation. . . . . . . . . . . . . . . . . . . 30
3.4 Low-energy electron diffraction pattern from Al/i -Al70Pd21Mn9.
Reprinted from ref. [45]. Copyright APS 2001. . . . . . . . . . 32
3.5 (a); A structure model for how Xe grows on d -Al72Ni11Co17,
with (b); its Fourier transform. (c); Diffraction from Xe/d -
Al72Ni11Co17. Reprinted from ref. [37]. Copyright APS 2004. . 33
3.6 An STM image of C60 on i -Al70Pd21Mn9, showing the τ -scaling
relationships of the adsorption sites. Reprinted from ref. [35].
Copyright Elsevier 2001. . . . . . . . . . . . . . . . . . . . . . 34
3.7 Stereographic projections of XPD images of the (a) Au 4f
and (b) Al 2s emissions from the five-fold surface of icosahe-
dral i -Al-Pd-Mn after Au depositions and subsequent anneal-
ing. The region covering 0◦ to 70◦ for polar angle is observed.
(c) Stereographic projection of the symmetric elements of the
icosahedral structure. [35]. Copyright The Japan Society of
Applied Physics 2001. . . . . . . . . . . . . . . . . . . . . . . 36
3.8 Low-energy electron diffraction patterns from (a); clean i -Al-
Pd-Mn, (b); Bi/i -Al-Pd-Mn (c); Sb/i -Al-Pd-Mn. Reprinted
from ref. [54]. Copyright APS 2002. . . . . . . . . . . . . . . . 37
4.1 (a); Raw spectrum. The high background is clearly evident.
(b); Differentiated spectrum. . . . . . . . . . . . . . . . . . . . 44
4.2 The MEIS system at CCLRC Daresbury. . . . . . . . . . . . . 46
4.3 Sample holder for MEIS. (a); Sample position. (b); Contacts
for the filament and high-voltage supply. . . . . . . . . . . . . 46
4.4 Two-dimensional MEIS data obtained from an annealed film
of Cu/i -Al70Pd21Mn9. . . . . . . . . . . . . . . . . . . . . . . 47
122
4.5 (a); A view of the quasicrystal model offset about 2◦ from a
channeling direction. (b); Aligned along a channeling direction. 48
4.6 An example of double-alignment geometry as it relates to MEIS. 49
4.7 The model viewed with perspective, and aligned along a five-
fold axis. An inverted pentagon is clearly visible in the centre
of the figure, and many channeling directions can be easily
identified. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.8 The sample plate as used for STM, with a gold-on-glass sub-
strate mounted. . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.9 (a); Operation of an STM in constant current mode. (b);
Constant height mode. . . . . . . . . . . . . . . . . . . . . . . 55
4.10 An Omicron variable-temperature scanning tunneling micro-
scope head. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.11 An example of a multiple tip. The tip is imaged at each sharp
protrusion (with a smaller radius of curvature than the tip)
on the sample surface. . . . . . . . . . . . . . . . . . . . . . . 58
4.12 The XMCD effect: For magnetic materials in an applied mag-
netic field, spin up and spin down bands are not equally pop-
ulated. For an applied field in the up direction, there will be
some empty spin up 3d states. In this case, because of the
conservation of spin, only 2p electrons with up spin can be ex-
cited into 3d states. When the orbital motion of the 2p states
is in the same sense as the circular motion of the incident light
the transition probability is larger and when the two motions
are in opposite directions the transition probability is smaller.
Therefore, for light of a particular handedness, the L3 peak
in the absorption spectrum, arising from the excitation of the
2p3/2 electrons, may be enhanced, whilst the L2 peak, arising
from the excitation of the 2p1/2 electrons, may be reduced.
When either the magnetisation direction or the polarisation
direction are reversed the effect on the size of the L3 and L2
peaks is also reversed. . . . . . . . . . . . . . . . . . . . . . . 61
4.13 A schematic of the ID8 beamline at the ESRF. . . . . . . . . . 62
123
4.14 The sample plate as used for XMCD, illustrating the azimuthal
rotation system. . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.15 The Ewald sphere. The points on the right-hand side of the
figure represent reciprocal lattice points on the crystal. The
vector k is drawn in the direction of the incident beam, and the
origin is chosen such that k terminates at any reciprocal lattice
point. A sphere is drawn of radius k = 2π/λ about the origin
of k. A diffracted beam will be observed if the circumference
of the sphere intersects any other reciprocal lattice point. As
depicted, the sphere intercepts a point connected with the end
of k by a reciprocal lattice vector G. The diffracted beam is
in the direction k’ = k + G. The angle θ is the Bragg angle
in equation 4.3. . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.16 A schematic of a common LEED optic. . . . . . . . . . . . . . 67
4.17 A schematic of a simple filament evaporation source. . . . . . 68
4.18 A schematic illustrating the principle of e-beam evaporation. . 70
5.1 (a); Two-dimensional data obtained from the as-deposited
Cu/i -Al70Pd21Mn9 film; (b); from the annealed Cu/i -Al70Pd21Mn9
film. The scattering features from the four elemental compo-
nents of different mass and some characteristic quasicrystalline
channels are indicated. . . . . . . . . . . . . . . . . . . . . . . 74
5.2 One-dimensional energy-resolved data obtained from Cu/i -
Al70Pd21Mn9. (a) From the unannealed film, showing a strong
Cu signal and no surface peaks for the bulk components. (b)
From the annealed film, showing surface peaks for all elements
present, indicating the alloying of the copper with the surface
of the quasicrystal. . . . . . . . . . . . . . . . . . . . . . . . . 76
5.3 (a); One-dimensional angle-resolved data obtained from Cu/i -
Al70Pd21Mn9 (b); annealed to 600 K for 10 minutes. (c); Clean
surface Pd data for i -Al70Pd21Mn9 [82]. . . . . . . . . . . . . . 79
124
5.4 Figure reproduced from Ref. [51] showing the pentagonal ‘dark
stars’ imaged on the five-fold surface of i -Al70Pd21Mn9, and
including a one-dimensional Fibonacci grid joining like points
on the surface to form a row structure with separations com-
parable to those of the unannealed copper film. Superimposed
is a portion of the proposed copper structure. . . . . . . . . . 82
5.5 The best-fit model for Cu/i -Al70Pd21Mn9 as run. Also shown
is an extension of the model into one 100 A × 60 A × 15
A domain. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.6 Part of the stereographic projection for fcc, with pentagonal
angles superimposed. Inset : The three components of the
best fit model. The blocking dips may be easily identified
with directions on the stereographic projection. . . . . . . . . 84
5.7 The large two-domain model for Cu/i -Al70Pd21Mn9 with an
STM image [57] superimposed. . . . . . . . . . . . . . . . . . . 88
6.1 (a-g); Comparison of the ten-fold d -Al72Ni11Co17 surface as
studied with increasing Co coverage. The coverage is respec-
tively 0.4 ML, 1.9 ML, 3.7 ML, 5.6 ML and, for (e-g), 7.5 ML
. (f) is a detail of (e), with the row structure showing more
clearly. Indications of orientational order are superimposed on
(e-g). A graph showing the trend of Auger peak area ratios vs
coverage is also shown. . . . . . . . . . . . . . . . . . . . . . . 92
6.2 (a-d); Comparison of the five-fold surface of i -Al70Pd21Mn9 as
studied with increasing Co coverage. Coverage is (a) 0.4 ML;
(b) as (a) with five-fold depressions highlighted; (c) 2.5 ML
coverage, and (d) 13 ML coverage. . . . . . . . . . . . . . . . 93
6.3 Low-energy electron diffraction patterns from 3 systems at 2
beam energies for each system. From left to right, the cov-
erages are: 20 ± 4 ML Co, 7.5 ± 0.4 ML Co, 4 ± 0.2 ML
Cu. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
125
6.4 (a); LEED pattern obtained from the clean surface at 120 eV.
(b); LEED pattern from a 7.5 ML Co overlayer at 120 eV.
Lines are superimposed to aid identification of the periodic
separation of the streaks. . . . . . . . . . . . . . . . . . . . . . 95
7.1 Manganese x-ray magnetic circular dichroism signals for flat
and clustered i -Al70Pd21Mn9. . . . . . . . . . . . . . . . . . . 103
7.2 Clustered i -Al70Pd21Mn9 Mn dichroism signals for in-plane
and out-of-plane field. . . . . . . . . . . . . . . . . . . . . . . 104
7.3 Terraced i -Al70Pd21Mn9 Mn dichroism signals for in-plane and
out-of-plane field. . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.4 (a) Monte Carlo simulations of the magnetic microstructure
in a sample of finite size for pure dipolar interaction. The mi-
crostructure has been obtained for a square sample of about
10,500 vector spins on the Penrose tiling. The spins belonging
to the perimeter of decagons (marked) form closed chains. (b)
Experimental model. The perspective view of the magnetic
microstructure. The red arrows represent the orientation of
dipolar moments of magnets fixed onto the nodes of the Pen-
rose tiling (rhombuses). The magnets can rotate in the hori-
zontal plane. Reproduced from ref. [40], copyright APS 2003. . 106
7.5 XMCD spectra of d -Al72Ni11Co17 Cobalt and Ni edges. There
is no observable dichroism. . . . . . . . . . . . . . . . . . . . . 107
7.6 Induced Co in d -Al72Ni11Co17 moment with a 3 ML adsorbed
Fe film. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
7.7 In-plane and out-of-plane dichroism for Fe deposited on d -
Al72Ni11Co17. . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
7.8 In-plane and out-of-plane hysteresis curves for Fe deposited on
d -Al72Ni11Co17. . . . . . . . . . . . . . . . . . . . . . . . . . . 111
126
Presentations
Oral Presentations
Formation of Aperiodic Ultrathin Films of Cu on Quasicrystal Surfaces
UK Scanning Probe Microscopy ’05 meeting, University of Warwick 2005.
Poster Presentations
Scanning Tunneling Microscopy of C60 on Graphite
Surface Science Summer School, University of Warwick 2002.
Fibonacci Films: the creation of an aperiodic film on a quasicrystal
substrate
Presented at the Young Researchers SET for Britain meeting, House of
Commons 2004.
Growth of aperiodic Cu films on an icosahedral i-Al70Pd21Mn9 quasicrystal
investigated using medium-energy ion scattering
Condensed Matter and Materials Physics ’04, University of Warwick 2004,
9th International Conference on Quasicrystals, Iowa State University 2005.
127
Forbidden Beauty
Stand at the Royal Society Summer Exhibition 2004.
STM and LEED of Co films adsorbed on d-Al72Ni11Co17 and on the
five-fold surface of i-Al70Pd21Mn9
9th International Conference on Quasicrystals, Iowa State University 2005.
128
Bibliography
[1] A.A. Michelson and E.W. Morley, Am. J. Sci. 34, 22 (1887).
[2] A. Einstein, Ann. Phys. 17, 891 (1905).
[3] R. Penrose, Bull. Inst. Math. Appl. 10, 266 (1974).
[4] D. Shechtman, I. Blech, D. Gratias, and J.W. Cahn, Phys. Rev. Lett
53, 1951 (1984).
[5] L. Pauling, Nature 317, 512 (1985).
[6] The Fibonacci sequence consists of terms such that the nth. term is
the sum of the (n − 1) and (n − 2) terms; furthermore the ratio of
successive terms approaches the Golden Ratio τ as n becomes large;
τ is an irrational number whose first few terms are τ=1.618. . . . See
e.g. R.A. Dunlop, The Golden Ratio and Fibonacci Numbers, (World
Scientific, Singapore, 1997).
[7] P. Guyot, P. Kramer, and M. de Boissieu, Rep. Prog. Phys. 54, 1373
(1991).
[8] N. Wang, H. Chen, and K.H. Kuo, Phys. Rev. Lett. 59, 1010 (1987).
[9] Z.H. Mai, L. Xu, N. Wang, K.H. Kuo, Z.C. Jin, and G. Cheng, Phys.
Rev. B 40, 12183 (1989).
[10] J.C. Jiang, K.K. Fung, and K.H. Kuo, Phys. Rev. Lett. 68, 616 (1992).
[11] S.E. Burkov, Phys. Rev. Lett. 67, 614 (1991).
129
[12] K.K. Fung, C.Y. Yang, Y.Q. Zhou, J.G. Zhou, W.S. Zhan, and B.G.
Shen, Phys. Rev. Lett. 56, 2060 (1986).
[13] L. Bendersky, Phys. Rev. Lett. 55, 1461 (1985).
[14] Q.B. Yang and W.D. Wei, Phys. Rev. Lett. 58, 1020 (1987).
[15] H. Chen, D.X. Li, and K.H. Kuo, Phys. Rev. Lett. 60, 1645 (1988).
[16] C. Soltmann and C. Beeli, Phil. Mag. Lett. 81, 877 (2001).
[17] Y. Yan, S.J. Pennycook, and A.-P. Tsai, Phys. Rev. Lett. 81, 5145
(1998).
[18] H. Takakura, A. Yamamoto, and A.P. Tsai, Acta Crystallographica
A57, 576 (2001).
[19] Y. Yan, S.J. Pennycook, and A.-P. Tsai, Mat. Res. Soc. Symp. Proc.
553, 189 (1999).
[20] Y. Yan and S.J. Pennycook, Phys. Rev. B. 61, 14291 (2000).
[21] H-C. Jeong and P.J. Steinhardt, Phys. Rev. Lett. 73, 1943 (1994).
[22] H-C. Jeong and P.J. Steinhardt, Phys. Rev. B. 55, 3520 (1997).
[23] K. Hiraga, T. Ohsuna, and S. Nishimura, Phil. Mag. Lett. 80, 653
(2000).
[24] M. Krajcı, J. Hafner, and M. Mihalkovic, Phys. Rev. B. 62, 243 (2000).
[25] M. Mihalkovic, I. Al-Lehyani, E. Cockayne, C.L. Henley, N.
Moghadam, J.A. Moriarty, Y. Wang, and M. Widom, Phys. Rev. B
65, 104205 (2002).
[26] E. Abe, K. Saitoh, H. Takakura, A.P. Tsai, P.J. Steinhardt, and H.-C.
Jeong, Phys. Rev. Lett. 84, 4609 (2000).
[27] D. Shechtman and I. Blech, Met. Trans. A 16A, 1005 (1985).
130
[28] D. Levine and P.J. Steinhardt, Phys. Rev. Lett. 53, 2477 (1984).
[29] T. Janssen, Phys. Rep. 168, 55 (1988).
[30] M. Boudard, M. de Boissieu, C. Janot, G. Heger, C. Beeli, H.-U. Nissen,
H. Vincent, R. Ibberson, M. Audier, and J.M. Dubois, J. Phys.: Cond.
Matter 4, 10149 (1992).
[31] C. Janot, Quasicrystals, A Primer (Oxford Science Publications, Ox-
ford, 1992), p. 187.
[32] C. Janot and M. de Boissieu, Phys. Rev. Lett 72, 1674 (1994).
[33] D. Gratias, F. Puyraimond, M. Quiquandon, and A. Katz, Phys. Rev.
B 63, 024202 (2000).
[34] Ph. Ebert, M. Feuerbacher, N. Tamura, M. Wollgarten, and K. Urban,
Phys. Rev. Lett. 18, 3827 (1996).
[35] J. Ledieu, R. McGrath, R.D. Diehl, T.A. Lograsso, A.R. Ross, Z. Pa-
padopolos, and G. Kasner, Surf. Sci. Lett. 492, L729 (2001).
[36] M. Kishida, Y. Kamimura, R. Tamura, K. Edagawa, S. Takeuchi, T.
Sato, Y. Yokoyama, J.Q. Guo, and A.P. Tsai, Phys. Rev. B. 65, 094208
(2002).
[37] N. Ferralis, K. Pussi, I. Fisher, R.D. Diehl, M. Lindroos, and C.J.
Jenks, Phys. Rev. B 69, 075410 (2004).
[38] W. Pauli, Z. Phys. 41, 81 (1926).
[39] M. Krajcı and J. Hafner, Phys. Rev. B. 58, 14110 (1998).
[40] E.Y. Vedmedenko, H.P. Oepen, and J. Kirschner, Phys. Rev. Lett. 90,
137203 (2003).
[41] B. Charrier, B. Ouladdiaf, and D. Schmitt, Phys. Rev. Lett. 78, 4637
(1997).
131
[42] T.J. Sato, H. Takakura, A.P. Tsai, and K. Shibata, Phys. Rev. Lett.
81, 2364 (1998).
[43] M. Ohring, Materials Science of Thin Films; Deposition and Structure,
2 ed. (San Diego; Academic Press, ADDRESS, 2002).
[44] V. Fournee and P.A. Thiel, J. Phys. D: Appl. Phys. 38, R83 (2005).
[45] B. Bolliger, V.E. Dmitrienko, M. Erbudak, R. Luscher, H.-U. Nissen,
and A.R. Kortan, Phys. Rev. B 63, 052203 (2001).
[46] V. Fournee, A.R. Ross, T.A. Lograsso, J.W. Evans, and P.A. Thiel,
Surf. Sci. 537, 5 (2003).
[47] V. Fournee, T.C. Cai, A.R. Ross, T.A. Lograsso, J.W. Evans, and P.A.
Thiel, Phys. Rev. B 67, 033406 (2003).
[48] F. Baier and P.A. Thiel, in preparation .
[49] T. Fluckiger, Y. Weisskopf, M. Erbudak, R. Luscher, and A.R. Kortan,
Nano. Lett. 3, 1717 (2003).
[50] J. Ledieu, C.A. Muryn, G. Thornton, R.D. Diehl, D.W. Delaney, T.A.
Lograsso, and R. McGrath, Surf. Sci. 472, 89 (2001).
[51] J. Ledieu and R. McGrath, J. Phys.: Condens. Matter 15, S3113
(2003).
[52] M. Shimoda, J.Q. Guo, T.J. Sato, and A.P. Tsai, Jpn. J. Appl. Phys.
40, 6073 (2001).
[53] T.C.Q. Noakes, P. Bailey, M. Draxler, C.F. McConville, L. Leung, J.A.
Smerdon, R. McGrath, A.R. Ross, and T.A. Lograsso, in preparation .
[54] K.J. Franke, H.R. Sharma, W. Theis, P. Gille, P. Ebert, and K.H.
Rieder, Phys. Rev. Lett. 89, 156104 (2002).
[55] V. Fournee, J. Ledieu, L. Leung, L.H. Wearing, R. McGrath, A.R.
Ross, and T.A. Lograsso, in preparation .
132
[56] H.R. Sharma, M. Shimoda, J.A. Barrow, A.R. Ross, T.A. Lograsso,
and A.P. Tsai, Mat. Res. Soc. Symp. Proc. 805, LL.7.10 (2004).
[57] J. Ledieu, J.-T. Hoeft, D.E. Reid, J.A. Smerdon, R.D. Diehl, T.A.
Lograsso, A.R. Ross, and R. McGrath, Phys. Rev. Lett. 92, 135507
(2004).
[58] J. Ledieu, J.T. Hoeft, D.E. Reid, J.A. Smerdon, R.D. Diehl, N. Ferralis,
T.A. Lograsso, A.R. Ross, and R. McGrath, Phys. Rev. B 72, 035420
(2005).
[59] J.A. Smerdon, J. Ledieu, J.T. Hoeft, D.E. Reid, L.H. Wearing, R.D.
Diehl, T.A. Lograsso, A.R. Ross, and R. McGrath, Phil. Mag. 86, 841
(2006).
[60] J.A. Smerdon, T.C.Q. Noakes, J. Ledieu, J.T. Hoeft, D.E. Reid, R.D.
Diehl, T.A. Lograsso, A.R. Ross, and R. McGrath, in preparation .
[61] M. Prutton, Introduction to Surface Physics (Oxford Science Publica-
tions, Oxford, 1995).
[62] A. Chambers, R.K. Fitch, and B.S. Halliday, Basic Vacuum Technology
2nd edition (Institute of Physics, Bristol, 1998).
[63] L. Meitner, Z. Physik 19, 311 (1923).
[64] A. Isaacs, A Dictionary of Physics, Third Edition (Oxford University
Press, Oxford, 1996).
[65] R.M. Tromp, in Practical Surface Analysis (Second Edition) Volume
2: Ion and Neutral Spectroscopy, edited by D. Briggs and M.P. Seah
(Wiley, USA, 1992).
[66] P.R. Watson, J. Phys. Chem. Ref. Data 19, 85 (1990).
[67] J.F. van der Veen, Surf. Sci. Rep. 5, 199 (1985).
[68] T.C.Q. Noakes and P. Bailey, Phys. Rev. B 58, 4934 (1998).
133
[69] F.J. Smith, Physica 30, 497 (1964).
[70] H.-J. Guntherodt and R. Wiesendanger, Scanning Tunneling
Microscopy 1 (Springer-Verlag, Berlin, 1992).
[71] R. Young, J. Ward, and F. Scire, Phys. Rev. Lett. 27, 922 (1971).
[72] R.H. Fowler and L. Nordheim, Proc. Roy. Soc. London Ser. A 119,
173 (1928).
[73] J. Frenkel, Phys. Rev. 36, 1604 (1930).
[74] G. Binning, H. Rohrer, Ch. Gerber, and E. Weibel, Phys. Rev. Lett
49, 57 (1982).
[75] D.P. Woodruff and T.A. Delchar, Modern Techniques of Surface Sci-
ence, Second Edition (Cambridge University Press, Cambridge, 1994).
[76] A. Perez-Rodriguez, A. Cornet, and J.R. Morante, Microelectronic En-
gineering 40, 223 (1998).
[77] C. Davisson and L.H. Germer, Phys. Rev. 30, 705 (1927).
[78] P. Bailey, T.C.Q. Noakes, and D.P. Woodruff, Surface Science 426, 358
(1999).
[79] J. Ledieu, A.W. Munz, T.M. Parker, R. McGrath, R.D. Diehl, D.W.
Delaney, and T.A. Lograsso, Surf. Sci. 433/435, 665 (1999).
[80] J. Ledieu, A.W. Munz, T.M. Parker, R. McGrath, R.D. Diehl, D.W.
Delaney, and T.A. Lograsso, Mat. Res. Soc. Symp. Proc. 553, 237
(1999).
[81] M. Mayer, ‘SIMNRA Users Guide’ Report IPP 9/113, Max-Planck-
Institute fur Plasmaphysik Garching, Germany, 1997.
[82] T.C.Q. Noakes, P. Bailey, C.F. McConville, C.R. Parkinson, M.
Draxler, J.A. Smerdon, J. Ledieu, R. McGrath, A.R. Ross, and T.A.
Lograsso, Surface Science 583, 139 (2005).
134
[83] M. Krajcı and J. Hafner, Phys. Rev. B 71, 054202 (2005).
[84] Z. Shen, M.J. Kramer, C.J. Jenks, A.I. Goldman, T.A. Lograsso, D.W.
Delaney, M. Heinzig, W. Raberg, and P.A. Thiel, Phys. Rev. B 58,
9961 (1998).
[85] A.R. Smith, K.-J. Chao, Q. Niu, and C.-K. Shih, Science 273, 226
(1996).
[86] S. Iwasaki, IEEE Transactions on Magnetics 16, 71 (1980).
[87] J. de la Figuera, J.E. Prieto, C. Ocal, and R. Miranda, Phys. Rev. B
47, R19 (1993).
[88] H.R. Sharma, Structure, Morphology, and Dynamics of Quasicrystal
Surfaces (PhD thesis, Freie Universitat Berlin, 2002).
[89] S.A. Chambers, S.B. Anderson, H.W. Chen, and J.H. Weaver, Phys.
Rev. B 35, 2592 (1987).
[90] Y.U. Idzerda, W.T. Elam, B.T. Jonker, and G.A. Prinz, Phys. Rev.
Lett. 62, 2480 (1989).
[91] M. Mallary, A. Torabi, and M. Benakli, IEEE Transactions on Mag-
netics 38, 1719 (2002).
[92] J. Tersoff and L.M. Falicov, Phys. Rev. B 26, 6186 (1982).
[93] B. Viala, M.K. Minor, and J.A. Barnard, J. Appl. Phys. 80, 3941
(1996).
[94] Ph. Ebert, F. Yue, and K. Urban, Phys. Rev. B 57, 2821 (1998).
[95] P. Ebert, F. Kluge, B. Grushko, and K. Urban, Phys. Rev. B 60, 874
(1999).
[96] C.J. Jenks, D.W. Delaney, T.E. Bloomer, S.-L. Chang, T.A. Lograsso,
Z. Shen, C.-M. Zhang, and P.A. Thiel, Applied Surface Science 103,
485 (1996).
135
[97] C.J. Jenks, S.-L. Chang, J.W. Anderegg, P.A. Thiel, and D.W. Lynch,
Phys. Rev. B 54, 6301 (1996).
[98] G. Cappello, A. Dchelette, F. Schmithusen, J. Chevrier, F. Comin, A.
Stierle, V. Formoso, M. De Boissieu, T. Lograsso, C. Jenks, and D.
Delaney, Mat. Res. Soc. Symp. Proc. 553, 243 (1999).
[99] F. Schmithusen, G. Cappello, M. de Boissieu, F. Comin, and J.
Chevrier, Surf. Sci. 444, 113 (2000).
[100] C. Jenks, D.W. Delaney, T.E. Bloomer, S.-L. Chang, T.A. Lograsso, Z.
Shen, C.-M. Zhang, and P.A. Thiel, Appl. Surf. Sci. 103, 485 (1996).
[101] M. Klanjsek, P. Jeglic, P. McGuinness, M. Feuerbacher, E.S. Zilstra,
J.M. Dubois, and J. Dolinsek, Phys. Rev. B 68, 134210 (2003).
[102] E. J. Cox, J. Ledieu, R. McGrath, R. D. Diehl, C. J. Jenks, and I.
Fisher, Mat. Res. Soc. Symp. Proc. 643, K11.3.1 (2001).
[103] M. Zurkirch, B. Bolliger, M. Erbudak, and A.R. Kortan, Phys. Rev. B
58, 14113 (1998).
136