You are on page 1of 96

Structural properties of MgO(100) in MgO/Fe(100)

heterostructures: a photoelectron diffraction study

Stefano Boseggia

December 17, 2010


Contents

Preface 3

1 Photoelectron diffraction 4
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Theoretical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2.1 Single scattering model theory (SC) . . . . . . . . . . . . . . . 8
1.2.2 Multiple scattering (MS) model . . . . . . . . . . . . . . . . . 17
1.2.3 Comparisons to other techniques . . . . . . . . . . . . . . . . 20
1.3 The MSCD software . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.3.1 Inelastic Scattering . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3.2 Correlated vibrational effects . . . . . . . . . . . . . . . . . . 25
1.3.3 Inner potential correction . . . . . . . . . . . . . . . . . . . . 26
1.3.4 Instrumental angular averaging . . . . . . . . . . . . . . . . . 26
1.3.5 Multiple scattering order . . . . . . . . . . . . . . . . . . . . . 26
1.3.6 R-A approximation order . . . . . . . . . . . . . . . . . . . . . 27
1.3.7 Cluster size . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.4 Calculation for MgO(100) . . . . . . . . . . . . . . . . . . . . . . . . 30
1.4.1 MgO crystal structure . . . . . . . . . . . . . . . . . . . . . . 30
1.4.2 Calculation parameters . . . . . . . . . . . . . . . . . . . . . . 31

2 Experimental setup 36
2.1 LASSE Laboratory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.1.1 Introduction Chamber (IC) . . . . . . . . . . . . . . . . . . . 36
2.1.2 Sample Preparation Chamber (SPC) . . . . . . . . . . . . . . 38
2.1.3 Measurement Chamber (MC) . . . . . . . . . . . . . . . . . . 39
2.1.4 Photocathode Preparation Chamber (PPC) and Pulsed Laser
Deposition Chamber (PLDC) . . . . . . . . . . . . . . . . . . 45
2.2 Experimental setup for XPD measurements . . . . . . . . . . . . . . 46
2.2.1 Hemispherical Energy Analyzer optimization . . . . . . . . . . 48

i
2.2.2 XPD measurements . . . . . . . . . . . . . . . . . . . . . . . . 50

3 Study of MgO/Fe(100) structure 54


3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.2 Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2.1 MgO/Fe(100)-p(1×1)O . . . . . . . . . . . . . . . . . . . . . . 58
3.2.2 MgO/Fe(100) . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.3 XPD: experimental measurements and theoretical calculations . . . . 62
3.3.1 Subtraction of the isotropic background . . . . . . . . . . . . . 64
3.3.2 Theoretical calculations with MSCD . . . . . . . . . . . . . . 67
3.3.3 Data analysis and discussion . . . . . . . . . . . . . . . . . . . 71
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

A 82
A.1 Input file for MSCD . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
A.1.1 Input file for atomic charge density of Oxigen . . . . . . . . . 82
A.1.2 MgO crystal structure input file for Muffin-tin potential cal-
culation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
A.1.3 Input file for MSCD calculation for MgO(100)-Mg KLL . . . . 83
A.2 Acquisition software for XPD measurements . . . . . . . . . . . . . . 85

Bibliography 88

ii
List of Figures

1.1 The three basic types of photoelectron diffraction measurement . . . 5


1.2 Photoelectron diffraction: (a) schematic diagram; (b) main scattering
effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Nickel plane-wave scattering factor amplitudes |fN i | . . . . . . . . . . 7
1.4 Theoretical calculations of electron scattering from a single Cu atom
at a distance of 3.5 Å from the emitter and for energies of: (a) 100
eV, (b) 300 eV, (c) 1000eV. . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Inner potential correction. . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6 Theoretical χ curves for emission from a s level as a function of
the scattering angle θN i for a single Ni scatterer at a distance of
2.49 Å from the emitter (EM). The three approximations PW, SW,
and SW(1) are shown for energies of 50-950 eV . . . . . . . . . . . . . 18
1.7 Classical trajectories for forward focusing of electrons emitted from a
point source by two atoms. . . . . . . . . . . . . . . . . . . . . . . . . 20
1.8 A scattering event leading from atom a to atom c via atom b through
the R-A scattering matrices Fλλ0 (ρ, ρ0 ) . . . . . . . . . . . . . . . . . 24
1.9 Instrumental angular averaging; Θ is the half aperture angle. . . . . . 27
1.10 Calculated Cu 3p photoelectron intensities, as a function of multiple
scattering order, from clean Cu(111) in a fixed forward-scattering
emission direction. Photoelectron energies are 156 eV and 547 eV. . . 28
1.11 Single scattering intensities from a 2 Cu-atom cluster for scattering
angles varying between forward and bacward scattering. Emission
from four different initial states: (a) s type , (b) p, (c) d, (d) f. . . . . 29
1.12 Cluster size used in MSCD . . . . . . . . . . . . . . . . . . . . . . . . 30
1.13 MgO crystal structure: (a)there are two FCC lattices which are sep-
arated by one half of the body diagonal of the unit cube, (b) polar
scan ([100] azimuth), (c) polar scan ([110] azimuth). . . . . . . . . . . 31
1.14 Muffin-Tin potential for a Fe crystal. . . . . . . . . . . . . . . . . . . 32

iii
1.15 (a) Auger electron diffraction polar scan of MgKLL electrons for a
MgO(100) crystal along the [110] azimuthal. (b) cross-section of
MgO(100) along [110] direction. The angles of the low-index direc-
tions correspondent to the most dense atomic chains are also shown. . 34
1.16 (a) Auger electron diffraction polar scan of MgKLL electrons for
a MgO(100) crystal along the [100] azimuthal. (b)cross-section of
MgO(100) along [100] direction. The angles of the low-index direc-
tions correspondent to the most dense atomic chains are also shown. . 35

2.1 Global view of the LASSE system devoted to growth and characteri-
zation of nanostructured films. High and ultra-high vacuum chambers
are interconnected by gauge valves . . . . . . . . . . . . . . . . . . . 37
2.2 Schematic view of the LEED optics . . . . . . . . . . . . . . . . . . . 39
2.3 Schematic view of the Knudsen cell. Crucible contains the compound
to be evaporated. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4 Description of a photoemission experiment: (a) definition of the an-
gles and wave vectors of the incident photon and emitted electron; (b)
representation of the photoexcitation process in the electronic band
scheme (c) conservation of the wave vector component parallel to the
surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.5 XPD basic experimental setup . . . . . . . . . . . . . . . . . . . . . . 46
2.6 Main Components of the PHOIBOS 150 c
. . . . . . . . . . . . . . . . 49
2.7 Analyzer acceptance area as a function of the polar angle θ. . . . . . 50
2.8 (a) AED da Mg KLL per un cristallo di MgO(100) lungo l’azimut
[110]: (curva nera) AED in cui le intensità dei picchi di emissione sono
state misurate con ∆E=1 eV, (curva rossa) XPD in cui le intensità
dei picchi di emissione sono state misurate con ∆E=0.2 eV. (b) Picco
Mg KLL con ∆E=0.2 eV. (c) Picco Mg KLL con ∆E=0.2 eV. . . . . 52
2.9 AED of Mg KLL along the MgO [100] direction with 5 mm iris: (red
curve) Low Magnification, (black curve) Medium Magnification. . . . 53
2.10 XPD, polar and azimuthal scan of Mg KLL for MgO(100) crystal . . 53

3.1 Spin-polarized tunnel effect in Ferromagnet/Insulator/Ferromagnet


junctions for parallel configuration (a) and antiparallel configuration
(b).The dendity of states (DOS) for the spin-up and spin-down bands
are shown at the bottom ( ∆ex is the exchange interaction). . . . . . 55
3.2 Top view of the mutual position of Fe and MgO lattice. . . . . . . . . 57
3.3 Fe(100)-p(1×1)O LEED with electron energy of 133 eV. . . . . . . . 58

iv
3.4 Fe2p3/2 peak intensity attenuation as a function of the MgO thickness
for the MgO/Fe(100)-p(1×1)O heterostructure: (black curve) the-
oretical attenuation calulated applying the Beer-Lambert law, (red
curve) experimental attenuation. . . . . . . . . . . . . . . . . . . . . . 59
3.5 LEED pattern with electron energy of 133 eV of a Fe film contami-
nated by carbon: beside typical spots of Fe lattice, the spots due to
the c(2X2) reconstruction are visible. . . . . . . . . . . . . . . . . . . 60
3.6 Fe2p3/2 peak intensity attenuation as a function of the MgO thick-
ness for the MgO/Fe(100) heterostructure: (black curve) theoretical
attenuation calulated applying the Beer-Lambert law, (red curve) ex-
perimental attenuation. . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.7 Polar scans of an MgO(100) crystal for emission from different levels:
Mg1s (black curve),O1s (red curve) e Mg KLL (blue curve) . . . . . . 63
3.8 (a) Schematic view of the XPD measurement system with angle defi-
nition used in I0 calculation, (b) schematic drawing showing the vari-
ation of the escape depth as a function of take-off angle θ. . . . . . . 65
3.9 AED of Mg KLL for an MgO/Fe(100) sample for an MgO thickness of
4ML, along the [100] azimuth. The acquired AED curve I(θ) (black
curve), the isotropic component I0 calculated using Eq. 3.8(red curve)
and the anisotropic component χ(θ) (blue curve) are shown. . . . . . 66
3.10 Strutture geometriche utilizzate nei calcoli teorici con MSCD: (a)
MgO/Fe(100)-p(1×1)O; (b) MgO/Fe(100) . . . . . . . . . . . . . . . 69
3.11 Calculated AED curves for 2ML for the MgO/Fe(100) (red curve)
and MgO/Fe(100)-p(1×1)O (black curve) (azimuth [110]) . . . . . . . 70
3.12 Calculated (blue line) and measured AED curves of 2 ML of MgO
onto Fe(100) (red line) and Fe(100)-p(1×1)O (black line) along the
[100] azimuth (a) and the [110]azimuth (b). . . . . . . . . . . . . . . . 72
3.13 Calculated (blue line) and measured AED curves of 4 ML of MgO
onto Fe(100) (red line) and Fe(100)-p(1×1)O (black line) along the
[100] azimuth (a) and the [110]azimuth (b). . . . . . . . . . . . . . . . 73
3.14 Calculated (blue line) and measured AED curves of 14 ML of MgO
onto Fe(100) (red line) and Fe(100)-p(1×1)O (black line) along the
[100] azimuth (a) and the [110]azimuth (b). . . . . . . . . . . . . . . . 74
3.15 Experimental(black dotted curve) and calcualted AED curves for
MgO/Fe(100)-p(1×1)O with a⊥ =2.105 Å (red curve)and a⊥ =2.15–
2.18 Å (blue curve) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

v
3.16 AED polar scans of 2 ML of MgO onto Fe(100) for: (a) MgO/Fe(100)
(red curve), MgO/Fe(100)-p(1×1)O (black curve) . . . . . . . . . . . 76
3.17 ∆FWHM relative to the MgO/Fe(100) and MgO/Fe(100)-p(1×1)O
samples as a function of MgO coverage along [100] (red curve) and
[110](black curve) azimuths . . . . . . . . . . . . . . . . . . . . . . . 77
3.18 Theoretical calculations of 0◦ peak broadening . . . . . . . . . . . . . 78
3.19 Gaussian fit of 0◦ peak calculated for three different structures: 0◦
tilt (green curve), 2.1◦ tilt (blue curve), 4.5◦ tilt (red curve) , 5.3◦ tilt
(purple curve) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.20 Schematic drawing for the surface roughness: (a) film with an oriented
step morphology, (b) epitaxial film. . . . . . . . . . . . . . . . . . . . 80

A.1 Interface of the software realized for XPD data acquistion. . . . . . . 85

vi
List of Tables

1.1 Normalized XPS cross section for combined emission from one, two,
and three monolayers of Cu(100), at 317 eV, 2p emission. . . . . . . . 20
1.2 Comparison of several surface-structure techniques by different criteria 22
1.3 Rehr-Albers approximation orders, dimension of the corresponding
scattering-amplitudes matrices, and allowed values of (µ, ν). . . . . . 24

2.1 Acceptance Angle as a function of the Iris diameter for the 3 magni-
fication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.1 Mg1s and Fe2p3/2 intensities for the MgO/Fe(100)-p(1×1)O and MgO/Fe(100)
heterostructures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2 FWHM for the 0◦ peak as a function of MgO coverage for the MgO/Fe(100)
and MgO/Fe(100)-p(1×1)O samples along [100] and [110] azimuths . 77

vii
Preface

Spin electronics, or spintronics, represents nowadays one of greatest scientific inter-


est fields in both solid state physics and technological research. The main aim is
to take advantage of the degree of freedom that has always been neglected in tra-
ditional electronics: electron spin. One of the most interesting results due to this
research field is the realization of Magnetic Tunnel Junction devices(MTJ). Two
ferromagnetic layers, with different coercive fields, are spaced out by a thin insulat-
ing layer in order to avoid the coupling between their magnetic moments. If at the
Fermi level the spin polarization keeps the same sign in the two ferromagnets, the
tunnelling electron current is larger for parallel configuration (P), while it is lower
for an antiparallel configuration (AP). An external magnetic field can then force one
of the two configurations P or AP and so it can tune the resistance of the device.
The efficiency of such a device can be expressed in terms of Tunnelling Magneto
Resistance (TMR):
RAP − RP
T MR = (1)
RP
where RP e RAP are the resistances related, respectively, to parrallel and antiparallel
configuration.
Possible applications of MTJ’s include memory cells for MRAM (Magnetic RAn-
dom Acces Memory) [1] and the realization of biosensors, based on the revelation of
superparamagnetic particles linked with biological markers [2].
MTJ’s based on Fe/MgO/Fe etherostructures have recently had development.
For these devices, have been measured , TMR values of 180% at room temperature
while ab initio theoretical calculations predict values of 1000% [3] [4]. The differ-
ence between the measured TMR values and those calculated can be explained in
different ways(structural disorder, interdiffusion, partial Fe oxidation,... ): specifi-
cally, MgO/Fe and Fe/MgO interfaces turn out having a key role in determining the
TMR. In fact, the interface itself plays a fundamental role in the tunnel effect: it
selects, according to symmetry rules, some electronic bands rather than others, cre-
ating a different conductivity between electrons of different symmetry states. One of
the possible problems consists in the formation of a FeO layer at MgO/Fe interface.

1
The formation of such layer at the interface MgO/Fe has been observed by Meyer-
heim et al. when MgO is deposited onto Fe(100) [5]. Oh et al. have also suggested
the coexistence of two distinct phases MgO/Fe(100) and MgO/FeO/Fe(100) at the
interface [6]. Calculation by Zhang and Butler shows that the expected TMR ratio
in presence of a FeO layer decreases down to 76% [7]. There are other studies that
do not reveal the formation of such a layer [8]and for this reason the phenomenon
is still debated.
Considering this scientifical background, we have studied the MgO growth (2-14
MonoLayer) on two different surfaces: Fe(100) and Fe-p(1×1)O. The former is the
most common surface used for the realization of MTJ’s based on MgO barrier that
can be subjected to partial oxidation after the deposition of MgO. This oxidation can
induce some disorder at the interface. The latter is instead a Fe surface that oxidizes
in a controlled manner with a chemisorption of a single monolayer of oxygen. Earlier
studies show that the coverage with exactly 1 oxygen ML stabilizes Fe surface.
Moreover , an improved magnetic behaviour has been observed for the Fe-p(1×1)O
rather than Fe. Infact the spin-dependent effects, present also in the clean surface,
are strongly amplified in the oxidize surface [9].
The purpose of this work is to understand which surface is better for MgO growth,
with particular attention to crystal structure near the interface: the relevance of this
study is due to the fact that it has been demostrated that the interface properties are
fundamental for determining the tunnel efficiency. The work done during this thesis
is a puzzle piece that stands inside an uptodate field of research that has acquired
relevant importance worldwide, the development of MTJ’s, with the particularity
that we have used an uncommon technique in this field: electron diffraction(ED).
The basic physics of electron diffraction (photoelectron and Auger electron) is
simply electron diffraction with a source of electrons inside a solid surface. If elec-
trons are emitted from an atom, the outgoing electron wavefield can be elastically
scattered by sorrounding atoms, and the coherent interference of the directly emit-
ted and elastically scattered component of this wavefield leads to variations in the
measured emission intensity outside the solid which depend on the associated scat-
tering path-lenghts and thus on the atomic location of the emitter atom relative
to the sorrounding atoms.This tecnique is an atom-specific probe of short range
order (∼20 Å). These properties make ED an outstanding instrument in order to
understand the crystal structure of nanostructure materials and the local order that
surrounds different atoms.

2
The work has been developed in L.A.S.S.E. laboratory (LAyered Structures for
Spin Electronics) situated in Como by the inter-university L-NESS(Laboratory for
Epitaxial Nanostructures on Silicon and for Spintronics).
The thesis is structured in the following way.
In chapter 1, the theory of photoelectron diffraction is presented , starting from
the most simple model, SSC-PW(Single scattering cluster-plane waves), then arriv-
ing to a more detailed model, MSC-SW (Multiple scattering cluster-spherical waves.
A software is also illustrated, based on multiple scattering calculation that per-
mit to obtain theorical curve of ED. These are very useful for the interpretation of
experimental data.
Chapter 2 describes in detail the experimental techniques used in this work and
the general structure of L.A.S.S.E. laboratory. Remarkably, it shows the experi-
mental setup for XPD (X ray photoelectron diffraction) measurement and explains
a particular software, developed in this thesis, that permits complete automatization
of data acquisition.
Finally, in the third chapter are presented the experimental data of ED(emission
from Mg KLL) measurement together with theoretical calculation for the system
MgO/Fe(100) and MgO/Fe(100)-p(1×1)O, with the purpose of determining the dif-
ference between the two interfaces and to identify which of them is the most appropi-
ate candidate to realize MTJ’s devices. Comparison between experimental data and
theoretical calculation shows that MgO grows epitaxially in both samples with a
well ordered crystal structure, but MgO/Fe(100) sample presents a higher degree
of disorder located at the interface: this agree with previous studies that show the
formation of a FeO layer, and with the fact that Fe(100)-p(1×1)O is a more stable
surface. Thus MgO/Fe(100)-p(1×1)O structure seems to have better perspectives
for MTJ’s application, even though at a MgO thickness of 14 ML, the MgO growth
seems not follow a layer by layer with a probably nucleation of islands.

3
Chapter 1

Photoelectron diffraction

1.1 Introduction
A knowledge of the atomic identities, position and bonding mechanisms within the
first 3-5 layers of a surface is essential to any quantitative understanding of sur-
face phenomena. A number of surface-structure probes has thus been developed
in an attempt to provide surface morphological properties(see Par. 1.2.3), each of
these methods has certain unique advantages and disadvantages and they are often
complementary to one another; among these, photoelectron diffraction (PD) has
a fundamentale role. The phenomenon of photoelectron diffraction 1 , discovered
almost 20 years ago, is nowadays become a powerful instrument to quantitatively
determine thin film structure and adsorbate geometry [10].
The basic physics of photoelectron diffraction is simply electron diffraction, but
unlike other tecniques like LEED (Low Energy Electron Diffraction) the source of
electrons is inside a solid surface. If electrons are photoemitted from an atom, the
outgoing photoelectron wavefield can be elastically scattered by surrounding atoms,
and the coherent interference of the directly emitted and elastically scattered com-
ponents of this wavefield leads to variations in the measured photoemission intensity
outside the solid which depend on the associated scattering path-lenghts and thus
on the atomic location of the emitter atom relative to surrounding atoms[11]. Pho-
toelectron diffraction thus carries information from which the local geometry around
the emitter atom can be extracted.
The basic experiment in PD (or AED) involves exciting a core photoelectron or
a core-like Auger transition from an atom and then observing modulations in the
resulting peak intensities that are due to final-state scattering from atoms neigh-
boring the emitter. Peak intensities can be monitored as a function either of the
1
and its close relative, Auger electron diffraction (AED)

4
emission direction or in the case of photoelectron diffraction, of the exciting photon
energy.
The three basic types of measurement possible are shown in Fig. 1.1: an az-
imuthal or φ scan, a polar or θ scan, and, for photoelectron diffraction (this is not
valid for AED where the energy of the emitted electrons is not dependent on the
exciting energy), a scan of energy in a normal or off-normal geometry.

Figure 1.1: The three basic types of photoelectron diffraction measurement: 1) scanned-
angle measurement in variable geometry: a) an azimuthal scan(φ) at constant
polar angle; b) a polar scan(θ) at constant azimuthal angle. 2) scanned-
energy measurement in fixed geometry: c) normal emission; d) off-normal
emission [11].

With soft X-ray excitation at about 1.2–1.5 keV, i.e. at the typical X-ray pho-
toelectron spectroscopy limit, scanned-angle measurements have been termed X-ray
photoelctron diffraction (XPD). Scanned-energy photoelectron measurements span-
ning the VUV to soft X-ray range have been called angle-resolved photoemission
fine structure (ARPES) to emphasize their similarity to the more familiar surface
extended X-ray absorption fine structure (SEXAFS). In these measurements the
photon energy is stepped and the photoemission intensity is measured in a specific
direction of emission as a function of the resulting photoelectron energy. To date,
scanned-angle studies are much more numerous; this is due to their greater simplic-
ity, since scanned-energy work has several requirements in addition: the sweeping of
photon energy with a synchroton radiation source and the correct normalization of
the photon fluxes and electron-analyzer trasmissions as a function of energy. How-

5
Figure 1.2: Photoelectron diffraction: (a) schematic diagram; (b) main scattering effects:
(i) backward scattering (θj =180◦ ) , (ii)forward scattering (θj =0◦ )

ever, an advantage in scanned-energy work is that Fourier transform (FT) methods


can be used to estimate the path-lenght for various strong scatterers.
In XPD experiments photoemission intensity depends on the probability that
the wavefunction of the emitted electron is scattered by neighboring atoms. This
probability is represented by elastic scattering factor f (θj ) that is a function of both
the scattering angle θj and the photoelectron kinetic energy. θj is the angle between
photoelectron wavevector k and rj which is the vector that goes from the emitter
atom to other atoms involved in scattering process(see Fig. 1.2(a)). Fig. 1.3 illus-
trates Nickel plane-wave scattering factor amplitudes (|fN i |) as a function of both
the scattering angle θN i and the photoelectron kinetic energy. For low energies of
50–200 eV, it is clear that there is a high anplitude for scattering into all angles.
For the intermediate range of about 200–500 eV, it is a reasonable approximation
to think of only forward scattering(θj = 0◦ ) and backward scattering(θj = 180◦ ) as
being important. However, at energies above 500 eV, we see that the scattering
amplitude is significant only in the forward direction, in which it is strongly peaked.
The degree of forward peaking increases as the energy is increased. Forward scatter-
ing, or zero-order diffraction, presents the strongest scattering factor f (θj ) since for
θj =0, i.e. for emission along the emitter-scatterer direction, there is no path-lenght

6
Figure 1.3: Nickel plane-wave scattering factor amplitudes |fN i | as a function of both the
scattering angle θN i and the photoelectron kinetic energy [12]

difference (∆L = rj (1 − cos θj ) = 0) between the two interfering components (see


Fig. 1.2(a)). The utility of such forward scattering at higher energies in surface-
structural studies was noted in early XPD investigations and it has been termed
” forward focusing”’. This effect turns out to be one of the most useful aspects
of higher-energy photoelectron or Auger electron diffraction to determine crystal
structure (see Fig. 1.2(b)). In fact, if an electron is emitted from an atom of a crys-
tal, the measured photoemission intensity will be higher along high-density atomic
chains since the final intensity will be the sum of the directly emitted photoelectron
wavefield (φ0 ) and all the forward scattered wavefields (φj ).
A quite different approach to photoelectron diffraction is to work under con-
ditions which ensure that backscattering is important. This effect turns out to be
useful to determine the location of adsorbed atoms which lie above a surface, possible
only with scattering at angle of about θj ∼180◦ .
The degree of modulation of intensity observed in PD or AED experiments can
∆I
be very large, with overall values of anisotropy as high as Imax = 50% − 70. Such
effects are relatively easy to measure, this is by contrst with the related surface-
strcture technique of SEXAFS, in which typical modulation are about one tenth as
large.
A final important aspect of either photoelectron or Auger electron diffraction is
that both are atom-specific probes of short-range order. Thus, each type of atom in
a sample can in principle be studied, and each will have a unique diffraction pattern

7
associated with the neighbors around it. The principal features of diffraction curves
are associated with scatterers as far as 20 Å away.
In the following paragraphs the theoretical modeling of XPD will be discussed
furthermore will be presented the MSCD (Multiple Scattering Calculation of Diffrac-
tion) package, a program that simulates the elemental and state-specific core-level
photoelectron diffraction pattern from a surface, based on multiple scattering theory,
and by fitting to experiments obtains atomic structural information.

1.2 Theoretical models


In this paragraph the theoretical models of XPD and AED will be discussed in
detail. We will begin here by presenting only the essential ingredients of the simplest
approach, the single-scattering cluster (SSC) model, and then comment toward the
end of this section on several improvements that can be made to it, as well as on
some effects expected due to multiple scattering (MS) events.

1.2.1 Single scattering model theory (SC)


The fundamental assumptions of the single-scattering cluster2 model are essentially
identical to those used in describing extended X-ray absorptiion fine structure (EX-
AFS) by P. A. Lee[13]. We consider photoelectron emission first and then discuss
the modifications required to described Auger emission.
Radiation with polarization E is incident on some atom in a cluster, from which it
ejects a core-level photoelectron. If the initial core-electron wave function is denoted
by ψc (r) and the final photoelectron wave function corresponding to emission with
wave vector k by ψ(r, k), then the observed intensity will be given in the dipole
aproximation by

I(k) ∝ | < ψ(r, k)|E · r|ψc (r) > |2 (1.1)

The final-state wave function in single scattering is further described as being the
superposition of a direct wave φ0 (r, k) and all singly scattered waves φj (r, rj → k)
that result from initial φ0 emission toward a scatter j at rj and then subsequent
scattering so as to emerge from the surface in the durection k. Thus, the overall
wave function can be written as
2
cluster is a small group of atoms

8
X
ψ(r, k) = φ0 (r, k) + φj (r, rj → k) (1.2)
j

and the observed intensity will be

I(k) ∝ |ψ(r, k)|2 (1.3)

Because the detector is situated at essentially infinity along k, all of the waves
in Eq. 1.2 can finally be taken to have the limiting spherical forms φ0 ∝ exp(ikr) r
exp(|r−r |)
or φj ∝ |r−rj |j modulated by the photoexcitation matrix element of the Eq. 1.1.
Flux conservation also dictates that the portion of φ0 which passes to the scatter j
to produce φj decays in amplitude as a spherical wave, or as r1j . This decay is the
principal reason why PD and AED are short-range probes (∼20 Å), although the
effect of inelastic scattering contributes additionally to this. If θj is the scattering
angle, the overall path length difference (PLD) between φ0 and any φj is rj (1−cos θj )
(cfr. Fig. 1.2(a)), and these PLDs provide most of the bond-length information about
atoms in the cluster.

Matrix Elements and Final-State Interference

When this model has been applied to photoelectron emission, the dipole matrix ele-
ment has usually been treated as involving a p-wave final state (i.e. the appropriate
case for emission from a s subshell). This yields a matrix-element modulation of
the form E · k̂ for an arbitrary direction of emission k̂ where k̂ is a vector unit. For
emission from other subshells with l 6= 0, more complex expressions including both
of the interfering l −1 and l +1 channels are involved. For example for emission from
a p subshell we have to considered the interference between the s and d channels.
Since the differential photoelectric crsoss section dσnldΩ
(E,k)
is proportional to in-
tensity rather than amplitude, another possible approximation might be to use a φ0
modulation of [ dσnldΩ
(E,k) 1/2
] , where n, l are the principal and orbital angular momen-
tum quantum numbers of the considered state. Although this is not strictly correct
and it also does not account for possible sign changes in the matrix element with
direction due to the photoelectron parity, it may be a reasonably adequate approx-
imation for higher-energy XPD in which the forward-dominated electron-scattering
process selects out rj choices very nearly parallel to k. That is, for the range of
rj directions near the k direction that produce significant scattering, the matrix
element varies little, so it can be considered constant.
Such final-state momentum and interference effects have been studied by Fried-

9
man and Fadley [14] and the results as a function of electron kinetic energy are
presented in Fig. 1.4. Here, a Cu emitter is 3.5 Ȧ away from a single Cu scatterer,
and three different electron kinetic energies of 100, 300 and 1000 eV are considered.
The intensity fluctuation as a function of scattering angle and for three different
initial state li are normalized to the unscattered intensity I0 as χ = I−I
I0
0
.
Several general conclusion can be drawn from the curves:

• Increasing the angular momenta in the final state is found to decrease sys-
tematically the amplitude of forward scattering, thus constituting a reason
for which calculations using the p final state may overpredict the degree of
anisotropy for emission from subshells with li ≥ 1 for θj = 0◦ .

• in the backscattering direction, the parity of the photoelctron waves is evident,


since the odd waves (li = 0, 2) exhibit the same sign of χ, and the opposite sign
is seen for the even waves (li = 1, 3). The previously discussed approximation
of using the square root of the differential cross section neglets these sign
difference.

• The smallest differences between different final-state angular momenta are for
the highest energy, where, in the dominant forward direction, the main effect
is a small reduction of amplitudes in the forward scattering direction. We see
a general suppression of the relative importance of the higher-order features,
but little change occurs in the shape of 0-th order peak. However, at the lowest
and medium energy range the differences between the curves as a function of
the final state become increasingly more significant.

Overall, these results indicate that the use of the correct final-state angular
momenta with interference will probably be important for energies below about 500
eV. For higher energies of 1000 eV , forward scattering should be reasonably well
treated by the simple p final state, although the relative intensities of higher-order
features may be overestimated.

Electron-Atom Scattering

The electron-atom scattering that produces φj is most simply described by a complex


plane-wave (PW) scattering factor

fj (θj ) = |fj (θj )| exp[iψj (θj )] (1.4)

where ψj (θj ) is the phase shift associated with the scattering.

10
Figure 1.4: Theoretical
calculations
of electron
scattering
from a single
Cu atom at
a distance of
3.5 Å from
the emitter
and for
energies of:
(a) 100 eV,
(b) 300 eV,
(c) 1000eV.
Different
final-state
assumptions
are com-
pared: li =0
(s to a single
p channel),
li =1 (p to
interfering
s + d chan-
nels), li =2
(d to p + f ),
li =3 (f to d
+ g) [14].

11
The use of this form for φj implicity assumes that the portion of φ0 incident on
the jth scatter has sufficiently low curvature compared to the scattering potential
dimensions to be treated as a plane wave (“small-atom”approximation). For large
exp(ik|r−r |)
r, the scattered wave φj is thus proportional to fj (θj ) |r−rj | j , with an overall
phase shift relative to φ0 of krj (1 − cos θj ) + ψj (θj ) that is due to both path-length
difference (see Fig. 1.2 (a)) and scattering process. The PW scattering factorfj (θj )
is thus determined by applying the partial-wave method to a suitable spherically
symmetric scattering potential (for example the “muffin tin potential ”3 ) for each
atomic type in the cluster [11]. The expression we obtain is
n
X
−1
f (θ) = (2ik) (2l + 1)[exp(2iδl ) − 1]Pl cos(θ) (1.5)
i=0

where the Pl are Legendre polynomials.


There are some useful generalization concerning the behavior of |fj | as atomic
number is varied:

• The forward scattering amplitude |fj | at higher energy decreases with atomic
number. This behavior can be rationalized by a classical argument in which
it is noted that forward scattering trajectories graze the outer reaches of the
scattering potential and so are only deflected slightly; these trajectories are
thus primarily sensitive to the outer regions of the potential (since atomic
radius decreases increasing atomic number).

• The backscattering amplitude at higher energy is by contrast found to increase


monotonically with atomic number. This is also expected from a classical
argument in which backscattering involves strongly deflected trajectories that
pass close to nucleus (since nuclear radius increase as a function of the atomic
number).

Inelastic Scattering

The effects of inelastic scattering on wave amplitudes during propagation below


surface must also be included. If intensity falls of as exp(−L/Λe ), where L is an
arbitrary path length below the surface and Λe is the inelastic attenuation lenght,
then the amplitude is expected phenomenologically to fall off as the square root
3
the muffin tin potential is a shape approximation of the potential field in an atomistic environ-
ment. In its simplest form, non-overlapping spheres are centered on the atomic positions. Within
these regions, the potential experienced by an electron is approximated to be spherically symmetric
about the given nucleus. In the remaining interstitial region, the potential is approximated as a
constant.

12
of this or exp(−L/2Λe ) = exp(−γL), where γ is the attenuation coefficient. By
contrast, some diffraction experiments yields Λe values that are about 0.50–0.75
times the literature values based upon this intensity-attenuation approach. These
reduced values of Λe are not surprising in view of several factors: uncertainties of
at least ±20% are common in measurements of attenuation lengths; the effects of
elastic scattering and diffraction on intensities can introduce additional uncertainties
of this order; finally, the effective attenuation length in a diffraction experiment
should be shorter than in a simple intensity-attenuation measurement. Fortunately,
electron diffraction features for most cases are not strongly affected by varying Λe
over its plausible range, and so its choice is in general not crucial to final structural
conclusions.

Vibrational Effects

Vibrational attenuation of interference effects is furthermore potentially important


and can be included in the simplest way by multiplying each φj by its associated
temperature-dependent Debey-Waller factor:

Wj (T ) = exp[−∆kj2 Uj2 (T )] = exp[−2k 2 (1 − cos θj )Uj2 (T )] (1.6)

where ∆kj is the magnitude of the change in wave vector produced by the scattering
and Uj2 (T ) is the temperature-dependent one-dimensional mean-squared vibrational
displacement of atom j that is a function of both the temperature and the Debye
temperature ΘD . At this level of approximation, Uj2 (T ) is assumed to be isotropic in
space, and any correlations in the movement of near-neighbor atoms are neglected.
Uj2 (T ) values or Debye temperatures can be obtained from the literature.
At high energy, the electron scattering is significant only when θj is rather close
to zero, and this acts through the (1 − cos θj ) factor of Eq. 1.6 to yield Wj (T )
very close to unity for all important scattered waves. So vibrational effects are to
first order not very significant in forward-scattering range, although they can be
very important where backscattering is the dominant diffraction mode and thus
(1 − cos θj ) is a maximum.

Single-Scattering Cluster Model

With these assumptions, the simplest SSC-PW expression for photoelectron inten-
sity I(k) can be written down from Eqs.1.1-1.3 as

13
Z 2
X E · r̂j

I(k) ∝ E · k̂e−γL + |fj (θj )|Wj e−γLj {exp i [krj (1 − cos θj ) + ψj (θj )]} dE


j
r j
XZ |fj (θj )|2
+ (E · r̂j )2 2
(1 − Wj2 )e−2γLj dE
j
rj

(1.7)

where (E·k̂) and (E·r̂j ) represent p-wave photoemission matrix-element modulations


along the k̂ and r̂j unit vectors, respectively, and e−γL and e−γLj are appropriate
inelastic attenuation factors. Vibrational effects are included in the Debye-Waller
factor Wj (T ). Thus, (E · k̂) exp(−γL) is the amplitude of the direct wave φ0 (r, k)
and (E · r̂j )|fj (θj )|Wj exp(−γLj ) is the effective amplitude of φj (r, rj → k). The
complex exponential exp {i [krj (1 − cos θj ) + ψj (θj )]} allows the total final phase
difference between φ0 and φj .
P
The second j corrects the first absolute value squared for the incorrect inclusion
of Debye-Waller attenuations in terms involving a product of a scattered wave with
itself. That is, in expanding the absolute value squared, only products involving
unlike waves like φ0 φ∗j or φj φl (j 6= l) should include Debye-Waller products of Wj or
Wj Wl , respectively. The (1 − Wj2 ) factor in the second summation is thus necessary
to yield overall correct products of the form φj φ∗j without any Wj2 factor. The
second sum has been called thermal diffused scattering, and it is often quite small
with respect to the overall modulations.
The integrals on E ~ simply sum over the different polarizations perpendicular to
the radiation propagation direction. For a totally unpolarized source the simplest
way to carry out this integration is just to sum the intensities regarding two per-
pendicular polarizations of convenient orientation in which unpolarized light can be
decomposed.
In modifying this model to described Auger emission, the usual assumption is
that the much freer mixing of angular momenta in the final state overall leads to
an outgoing wave with s character [15] [16]. Although selection rules do limit the
allowed final angular momentum states in Auger emission, for certain cases, the
l = 0 channel is dominant. Thus, assuming for Auger emission an s final state, we
simply remove all factors involving (E · k̂) and (E · r̂j ) in Eq. 1.7. We obtain that
Auger emission is indipendent of light polarization.
It is also worth noting here that the cluster sum on j in Eq. 1.7 makes no
explicit use of the translational periodicities that may be present, even though the
atomic coordinates rj used as inputs may incorporate such periodicities. Thus,

14
Figure 1.5: Inner potential correction.

neither surface nor bulk reciprocal lattice vector G are explicitly involved unlike
other technique as LEED, since Eq. 1.7 concerns only short-range order.
The last parameter of importance in Eq. 1.7 is the range of j or the choice of
a suitable cluster of atoms. This is done empirically so as to include all significant
scatterers by verifying that the predicted diffraction patterns do not change signif-
icantly with the addition of further atoms at the periphery of the cluster. Clusters
can range from a few atoms for near-normal emission from a molecule on a surface
to as many as several hundred atoms for substrate emission.
A further physical effect that has to be considered in our model is the possibility
of electron refraction at the surface in crossing the surface barrier or inner poten-
tial of height V0 . Even at relative high energies, for emission angles near grazing,
refraction effects of a few degrees can be produced. Thus, for a lower take-off angles
relative to the surface and/or lower kinetic energies, a proper allowance for refraction
is necessary. This is accomplished most simply by using a suitable inner potential V0
derived from experiment or theory to predict the internal angle of emission θin for a
given external propagation direction θout. The resulting expression for an electron
energy of EKin = EKout + V0 inside the surface is
" 1/2 #
−1 EKout
θin = cos cos θout (1.8)
EKout + V0

where θin and θout are measured with respect to the surface.

15
A final step in any realistic calculation based upon this model is to integrate the
direction of emission k over the solid angle Ω0 accepted into the electron analyzer.
We can thus integrate the measurements over a cone of ±3◦ half angle, although for
higher resolution analyzer, like ours, a smaller cone of ±1◦ has been used.

Improvements to the Model

A first possible correction to the simple SSC-PW model is to choose a more correct
form for the primary wave as it leaves the emitter. The SSC-PW result of Eq. 1.7
assumes a simple outgoing plane wave from the emitter which then scatters to pro-
duce an outgoing spherical wave from each scatter. In fact, the correct primary
wave should consists of an ingoing spherical wave plus the outgoing plane wave [13].
Such a primary wave experiences the emitter potential and represents the correct
soluton to the Schrodinger equation inside of a muffin-tin-like region centered on
the emitter. If this form of the primary wave is used, the equivalent of Eq. 1.7 with
neglect of effects due to vibration is:

Z X E · r̂j
I(k) ∝ E · k̂e−γL + |fj (θj )|e−γLj {exp i [krj (1 − cos θj ) + ψj (θj )]}


j
rj
2
X E · r̂j
+ fj (π)fem (π − θj )e−γ(L+2rj ) i2krj
e ~
dE
j
rj2
(1.9)

This equation, although still single scattering in assumption, now contains,


through the scattering of the incoming wave, a second sum of terms that are classic
double scattering events represented by fem factor. Because these added terms are
in effect double scattering and also exhibit stronger attenuation due to both 1/rj2
and e−2γrj , this sum is expected for many cases to be a small correction to Eq. 1.7.
A next important correction is the use of spherical waves (SW) scattering instead
of the asymptotic and much simpler plane-wave (PW) scattering. SW scattering
factors were first calculated by Barton and Shirley using a Taylor-series magnetic
quantum number expansion (MQNE) [17] [18]. By contrast, using separable Green’s
function approaches, Rehr et al. [19] derive an equation identical to Eq. 1.9 in form,
but in which the plane-wave scattering factors fj (θ) are replaced by three effective
spherical-wave scattering factors fj,ef f (θ, rj ), fem,ef f (π − θ, rj ) for the scatterer and
the emitter atoms. These factors depend on rj , as they must converge to the PW
result as rj −→ ∞. For example, Fig. 1.6 compares PW and SW scattering from a

16
single Ni scatterer at a distance of 2.49 Å from the emitter for the case of s emission
to a single p channel for different energies. For larger scattering angles (≥40◦ ) and
higher energies (≥200 eV) the PW and SW results are essentially identical. However,
for lower energies and in the forward scattering direction , there are significant
differences. In particular, for energies ≈100 eV, the forward scattering peak is
significantly reduced in amplitude by a factor that can be as low as 0.5. Thus,
particularly at higher energies, the much simpler PW approximation is still found to
yield results very similar in form to those with SW scattering, and it has been found
possible to draw useful structural conclusions with it. Sometimes, PW scattering at
high energy has been used together with an empirical reduction factor of forward
scattering amplitudes by a factor of 0.4–0.5 that can be largely justified as being
due to SW effects. In Fig. 1.6 is also presented curves labelled SW (1) , that represent
a first-order aproximation to the full SW scattering due to Rehr [20]. Is important
to emphasize that a full SW computation requires about 20 times the time of the
analogous PW computation. By contrast, the net increase in time for a computation
based on the Rehr’s approximation SW (1) as compared to a PW computation is less
than a factor two and the agreement with a full SW computation is essentially
perfect.
An additional important correction for some cases is the use of correlated vi-
brational motion in which atoms that are near neighbors of the emitter have lower
vibrational amplitude relative to the emitter [11]. This correction is more impor-
tant at lower energies for which large angle or backscattering events become more
important. This is done replacing Uj2 (T ) factor in Eq.1.6 by σj2 (T ) =< (∆kj · uj )2 >
factor which is a thermal average of the projection of the atomic displacement uj
as measured with respect to the emitter onto direction of the change in wave vector
produced by the scattering ∆kj .
A final aspect of the model which might be improved is more accurate allowance
for both surface refraction and attenuation due to inelastic scattering which could
also be considered as anisotropic. However, it has been observed that these elements
in any cases produce only small variations to the strong anisotropies associated with
diffraction effects.

1.2.2 Multiple scattering (MS) model


Finally, we have to consider the importance of multiple scattering (MS). First stud-
ies have been done by Tong et al. using LEED-type methods that requires full
traslational symmetry along the surface[22], and then was developed by Barton and
co-worker a cluster model with SW scattering and the Taylor series MQNE method

17
Figure 1.6: Theoretical χ curves
for emission from a
s level as a function
of the scattering an-
gle θN i for a single Ni
scatterer at a distance
of 2.49 Å from the
emitter (EM). The
three approximations
PW, SW, and SW(1)
are shown for energies
of 50-950 eV [21].

18
to simplify the calculation[18]. More recently, Rehr and Albers have proposed a
Green’s-function matrix method for such MSC-SW calculation[19], which will be
presented in details in Par. 1.3.
In general, considering multiple scattering Eq.1.2 become:
X XX
ψ(r, k) =φ0 (r, k) + φj (r, rj → k) + φjk (r, rj → rk → k)
j j k
XXX
+ φjkl (r, rj → rk → rl → k)
j k l
XXXX
+ φjklm (r, rj → rk → rl → rm → k) + ordini superiori
j k l m

(1.10)

One effect of MS, first discussed by Tong et al. [22], is a defocusing of intensity
occurring in multiple forward scattering at higher energies along a dense row of
atoms, suc that an SSC-PW or SSC-SW caslculation along such a row may oversti-
mate the intensity by a factor of two or more. In Tab. 1.1 are shown the normalized
XPS cross section calculated for combined emission from one, two, and three mono-
layers of Cu(100), at 317 eV, for 2p emission together with the experimental data
for the three SSC-PW, SSC-SW, and MSC-SW models. Comparing with the data,
spherical-wave single-scattering theory works for 2 ML, plane-wave single-scattering
theory fails even there, and only the multiple-scattering result adequately repro-
duces the emission intensities for the 3-ML case. The SSC-SW result overstimates
the 3-ML case by 48%, while the PW result overstimates it by 128%. Thus, single
scattering theory overstimates the third layer contribution.
These values illustrate an important physical effect which is correctly described
by multiple-scattering theory, but not by single-scattering theory : an electron emit-
ted from a third-layer atom, after being focused into the forward direction by a
second-layer atom, is further deflected (and hence defocused) to directions other
than forward by an atom in the first layer. In single-scattering theory, however,
the defocusing effect is neglected because it is a second-order scattering effect. As
a consequence, single-scattering theory overestimates the forward-direction peak in
case of three monolayers. In Fig.1.7 are shown the trajectories of electrons. Reading
from right to left, ray A0 is focused into the forward direction by atom 1, but is then
deflected to a new direction (ray A) by atom 2. Ray B, on the other hand, exits
atom 1 at an oblique angle, but is then focused into the forward direction by atom 2.
Thus, multiple scattering defocuses the ray chosen in single-scattering theory (ray
A0 ), and instead, introduces a new ray into the forward direction (ray B).
An additional important multiple scattering effect pointed out by Barton et al.

19
(d)
(b) (c) Single
(a) Multiple Single Scattering Scattering
Monolayers Expt. Scattering (SW) (PW)
1 1 1 1 1 1
2 2.3 2.35 2.46 2.50 3.40
3 2.7 2.97 4.0 4.15 6.16
(8 partial (8 partial (20 partial (20 partial
waves) waves) waves) wave)

Table 1.1: Normalized XPS cross section for combined emission from one, two, and three
monolayers of Cu(100), at 317 eV, 2p emission. The intensities are normalized
to unity for the 1-ML case [22].

Figure 1.7: Classical trajectories for forward focusing of electrons emitted from a point
source by two atoms [22].

is due to strong to strong nearest-neighbor backscattering at lower energies[17] [18].


This they find in certain scanned-energy cases to significally increase intensity due
to events of the type emitter→neighbor→emitter→detector.

A further important point in connection with such multiple-scattering calcu-


lation is that events up to at least the fifth order have to be included to assure
reasonable convergence. In fact, it is found that including only second-order events
can often lead to curves which are in much poorer agreement with experiment than
the corrisponding first-order calculation.

1.2.3 Comparisons to other techniques


We conclude this paragraph by comparing PD and AED to several other current
probes of surface structure. Evidently no one surface-structure probe directly and
unambigously provides all the desired information on atomic identities, positions,

20
bond distance and bond direction; by contrast, we need to use complementary in-
formation from several methods.
In Tab. 1.2 we show several techniques assesed according to a number of char-
acteristics: X-ray photoelectron diffraction (XPD) , angle resolved photoemission
fine structure (ARPEFS), Auger electron diffraction (AED), surface extended X-
ray absorption fine structure (SEXAFS), near edge X-ray absorption fine structure
(NEXAFS), low-energy electron diffraction (LEED), surface-sensitive grazing inci-
dence X-ray scattering (GIXS), and scanning tunneling microscope (STM). The
basic assumptions in SEXAFS and NEXAFS are essentially identical to those used
in describing ARPEFS. The difference is that in such experiments is measured the
absorption coefficient variation µ as a function of the incident photon energy instead
of the photoemission intensity. NEXAFS is distinguished from the closely related
SEXAFS method in that NEXAFS concentrates on fine structure within about 30
eV of the absorption edge while EXAFS considers the extended spectrum out to
much higher electron kinetic energies. LEED and GIXS are classical diffraction
techniques: the former uses electrons as source, the latter X-ray radiation. STM is
a microscope technique based on the concept of quantum tunneling highly sensitive
to the first monolayer.

21
XPD ARPEFS SEXAFS NEXAFS LEED GIXS STM
Atom YES YES YES YES NO NO NO
specific
Chemical-state YES YES NO YES NO NO NO
specific
Bond YES YES YES YES NO NO YES
directions
Bond YES YES YES NO NO NO YES
distances
Adsorption YES YES YES NO NO NO YES
site
symmetries
Coordination YES YES YES NO NO NO YES
numbers
Position ∼0.02Å ∼0.02Å ∼0.02Å ? ∼0.01Å ∼0.001Å ∼0.3Å
accuracies (Horiz.) (Horiz.)
? ∼0.05Å
(Vert.) (Vert.)
Valence-electron NO NO NO YES NO NO YES
states
Short-range YES YES YES YES NO NO YES
order
Long-range NO NO NO NO YES YES YES
order
Fourier NO YES YES NO NO YES YES
transform
analysis
Require synchrotron NO YES YES YES NO YES NO
radiation

Table 1.2: Comparison of several surface-structure techniques by different criteria

22
1.3 The MSCD software
The MSCD software (Multiple Scattering Calculation of Diffraction) was devoloped
by Y. Cheng e M.A. Van Hove at Lawrence Berkeley National Laboratory [23].
This program simulates the elemental and state-specific core level photoelectron
diffraction pattern from a surface, based on multiple scatterin theory and the Rehr-
Albers separable rapresentation of spherical-wave propagators [20]. Both energy-
scanned and angle-scanned photoelectron experiments can be simulated.
We begin using a formalism based on the Green’s function propagator instead
of a calculation based on the electron wave functions and the matrix elements.
The Green’s function represents the spherical-wave free propagation between the
scattering atoms. Thus, we prefer to express the result in terms of the Green’s
propagator in the real space instead of directly calculate the final state[24]. Then, the
scattering process can be conveniently formulated in terms of the Green’s propagator
GL and the scattering matrices Fλλ0 (ρ, ρ0 ) that depends on the atomic distances,
where ρ = krj is the interatomic vector leading from and to the site, as illustrated in
Fig.1.8. The Green’s functions and the scattering matrices are formulated in terms of
a set of spherical harmonics that are functions of the angular momentum L = (l, m)
by means of quantum numbers l and m. In the Rehr-Albers approximation we
consider a reduced set λ = (µ, ν) which represents a truncated expression of the
combinations between l e m. For µ = −lmax · · · lmax and ν = 0 · · · |µ| we return
to the exact representation. Since this equation converges relatively quickly, can
usually be truncated without significant loss of accuracy. The scattering matrix F00
is equal to the single-scattering factor f (θ, rj ).
The advantage of the R-A representation is that the aproximation leads to a
smaller matrix sizes, resulting in much reduced computation times. In Tab. 1.3 are
presented the Rehr-Albers approximation orders, the dimensions of the correspond-
ing scattering-amplitude matrices, and allowed values of (µ,ν). In most cases R-A
representation requires matrix sizes of only (6X6) (2nd order of R-A approximation).
By contrast, in the exact formalism, the propagator matrix GLL0 has the dimension
(lmax + 1)2 × (lmax + 1)2 , where lmax = kmax Rmt (Rmt is the muffin-tin radius) and for
a typical muffin-tin radius of 1.5 Å this yields matrix sizes of (36X36) to (441X441)
from low to high energies.

23
Figure 1.8: A scattering event leading from atom a to atom c via atom b through the R-A
scattering matrices Fλλ0 (ρ, ρ0 ) [23]

R-A order F matrix dimension (µ,ν)


0 1×1 (0,0)
1 3x3 (0,0),(±1,0),
2 6x6 (0,0),(±1,0),(0,1),(±2,0)
3 10x10 (0,0),(±1,0),(0,1),(±2,0),(±1,1),(±3,0)
4 15x15 (0,0),(±1,0),(0,1),(±2,0),(±1,1),(±3,0),(0,2)(±2,1),(±4,0)

Table 1.3: Rehr-Albers approximation orders, dimension of the corresponding scattering-


amplitudes matrices, and allowed values of (µ, ν) [23].

24
1.3.1 Inelastic Scattering
For including inelastic attenuation we use the common phenomenological approach
of the exponential decay described in Par.1.2.1. In MSCD for calculating the inelas-
tic mean free path Λe we can use :

• the empirical formula TPP-2 by Tanuma, Powell and Penn [25] :

E
Λe =  C D
 (1.11)
Ep2 β log(γE) − E
+ E2

This formula works for electron energies of 50-2000 eV. Λe is the inelastic mean
free path in Å, E is the electron energy in eV, Ep = 28.821(NV ρ/M )1/2 is the
free-electron plasmon energy in eV, ρ is the density of the bulk in g cm−3 , NV
is the number of valence electrons per atom (for elements) and molecule (for
compounds) and M is the atomic or molecular weight in amu. The terms β,
γ, C e D are parameters given by:

0.994
β = −0.0216 + p 2 2
+ 7.39 · 10−4 ρ
Ep + Eg
γ = 0.191ρ
(1.12)
0.91NV ρ
C = 1.97 − = 1.97 − 1.096 · 10−3 Ep2
M
20.8NV ρ
D = 53.4 − = 53.4 − 0.025Ep2
M

• Another relatively simple and convenient means proposed by Wagner, Davis e


Riggs [26]
Λe = kE m (1.13)

where m e k are material-dependent parameters. They found that m ranged


from 0.54 to 0.81, and k for different elements or compound can be derived
from Tables [23].

In the following paragraphs the formula TPP-2 will be used.

1.3.2 Correlated vibrational effects


For including correlated thermal vibrational effects we follow the Eq. 1.6 described
in Par.1.2.1. In a realistic calculation we have to account that the mean square
relative displacement σj2 is different from bulk atoms to surface atoms, because
going from bulk to surface changes the number of nearest-neighbors surrounding

25
each atom, as a consequence the effective mass changes. The relation between the
mean square relative displacement σj2 and different atomic masses has been given
by Allen, Alldredge and Wette [27]:
p p
2 2
σbulk Mbulk = σsurf Msurf,ef f (1.14)

where 
 M if T /ΘD << 1,
 surf


Msurf,ef f = Mbulk if T /ΘD > 1, (1.15)

if T /ΘD ∼

···M

M
surf bulk = 1.

1.3.3 Inner potential correction


As we have seen in Par. 1.2.1, also MSCD includes the possibility of electron refrac-
tion at the surface in crossing inner potential of height V0 . For this reason ’Eq.1.8
is still valid.
Presently there exist no generally acknowledged theoretical or experimental inner
potential data for any kind of element, hence V0 is treated as an adjustable parameter
and fit to experiment.

1.3.4 Instrumental angular averaging


As we have noticed in Par. 1.2.1, for obtain a realistic calculation we have to con-
sider that the experimental apparatus for measuring photoemission intensities has
a small but finite angular resolution characterized by half the angle subtended by
the aperture at the source, Θ.
The instrumental angular averaging due to this finite aperture of the detector is
done by summing the photoelectron intensities over a grid of points on a circular
aperture centred on the nominal emission direction as defined by wave vector k (see
Fig. 1.9). We calculate photoemission intensities Ia , Ib , Ic , Id , Ie for five different
directions, and we assume that the average intensity in each sector is the average of
the intensities at the three triangle corners. Thus, we obtain:

IT OT = (2Ia + Ib + Ic + Id + Ie )/6 (1.16)

1.3.5 Multiple scattering order


In order to investigate the convergence of the calculation as a function of the scat-
tering order was performed by authors a calculation for a default cluster of 86 atoms

26
Figure 1.9: Instrumental angular averaging; Θ is the half aperture angle [23].

representing the ideal clean Cu(111) surface for 3p level emission in a energy-scanned
geometry using the 4th order of R-A approximation. As illustrated in Fig. 1.10, the
maximum order of scattering needed to adequately simulate PD patterns is 6th
or 7th, but probably lower than this with the inclusion of vibrational effects and
angular averaging.

1.3.6 R-A approximation order


To perform a stringent test of the R-A order Chen and Van Hove calculated the
photoemission intensities for a 2 Cu-atoms cluster (a single-scattering event) with
an arbitrary 2 Å bond length as a function of the scattering angle. Fig.1.11 shows
the calculations for different R-A orders compared with exact results for diverse
initial states.It is seen that :

• for a s initial state emitting into p photoelectron waves, the first R-A order
(3X3 matrix) is sufficient

• for a p initial state emitting into s and d waves, the second R-A orders (6X6
matrix) is adequate

• for a d initial state emitting into p and f waves, the third R-A orders (10X10
matrix) might be needed

• for a f initial state emitting into d and g waves, the fourth R-A orders (15X15
matrix) is necessary to obtain results accurate within 1%.

27
Figure 1.10: Calculated Cu 3p photoelectron intensities, as a function of multiple scat-
tering order, from clean Cu(111) in a fixed forward-scattering emission di-
rection. Photoelectron energies are 156 eV and 547 eV [23].

Can be demostrated that in a larger cluster subsequent scattering can be treated


with equal or, more fequently lower, order in R-A, so that the R-A order needed
from the first scattering is an upper limit for the entire multiple-scattering problem.
Overall, we can infer a simple rule of thumb for guaranteeing adequate results: for
emission from an initial state li , use the (li + 1)-th R-A order for the first scattering
event after emission.

1.3.7 Cluster size


Cluster represents the number of atoms involved in the emission-scattering process
considered in the XPD calculations. If a cluster is to represent an infinitely extended
surface or include multilayer emission from a bulk specimen, its size must be chosen
large enough. To properly scale this problem, photoelectron waves leaving an emit-
ter in free space decay in intensity with the inverse square of the distance from the
emitter, i.e. ∝ 1/r2 : if there were no other damping effects, this would require an
infinitely large cluster, since the number of scatterers on a shell at a given distance
increases with the square of that distance, compensating the 1/r2 decay. Inelastic
scattering adds an exponential decay factor which ensures that a finite cluster suf-
fices. Vibrational effects and angular broadening act to further shrink the volume

28
Figure 1.11: Single scattering intensities from a 2 Cu-atom cluster for scattering angles
varying between forward and bacward scattering. Emission from four differ-
ent initial states: (a) s type , (b) p, (c) d, (d) f [23].

29
that is effective in producing diffraction modulations.

Figure 1.12: Cluster size used in MSCD [23].

Can be demostrated that clusters of about 100 atoms in size should be sufficient
for describing adequately most problems. In MSCD the cluster shape has been
chosen to be a half ellipsoid, as shown in Fig.1.12, and the parameters that control
its dimension are r and h: the former is the radius of the circular surface overlapping
the sample surface, the latter represents the cluster extension in the bulk.

1.4 Calculation for MgO(100)


In this paragraph will be presented an example of a PD calculation using MSCD
for the MgO(100) surface. These simulations have been done in order to test the
efficiency of the program for a well-know XPD case and serve as a subsequent com-
parison with the study of MgO/Fe(100) and Mgo/Fe(100)-p(1×1)O interfaces that
will be discussed in Chapter 3.

1.4.1 MgO crystal structure


The MgO crystal structure, shown in Fig.1.13(a), is the same of NaCl: the space
lattice is face centered cubic (FCC)and there are two such FCC lattices which are
separated by one half of the body diagonal of the unit cube. One lattice is occupied
by Mg atoms, the other by O atoms. The lattice parameter is 4.21 Å.

30
Figure 1.13: MgO crystal structure: (a)there are two FCC lattices which are separated
by one half of the body diagonal of the unit cube, (b) polar scan ([100]
azimuth), (c) polar scan ([110] azimuth).

1.4.2 Calculation parameters


First we need to calculate the Muffin-tin potential for the atoms in the cluster:
i.e. for O and Mg. Calculated Electronic Properties of Metals (by Moruzzi et
al.[28]) listed calculated muffin-tin potentials for the first 32 metal elements. For
other elements or compounds is necessary to use a specific program for calculate the
potential data. We have used the Barbieri-Van Hove Phase Shift Package (BVHPS),
a software developed by Barbieri and M.A. Van Hove that can calculate muffin-tin
potentials for all the elements in arbitrary given environment. First we have tested
the program for a Fe crystal for which the potential data calculated by Moruzzi et
al. exist. As shown in Fig.1.14 the calculations are in good agreement.
Using the BVPS we have performed a free atom self-consistent calculation for O
and Mg atoms. This is accomplished by using a self-consistent Dirack-Fock approach
(i.e. relativistic approach which computes, separately for each element, the self-
consistent atomic orbitals). The obtained orbitals can then be used to compute the
total radial charge densities associated to each element. Now one can compute the
muffin tin potential by means of the Augmented Plane Waves method. In essence,
the atomic charge densities of the different elements making up the structure that
we are interested in are superimposed to reflect the actual position of these elements
in the structure. As muffin-tin radius we adopted Rmt = 1.06 Å , which corresponds

31
Figure 1.14: Muffin-Tin potential for a Fe crystal. The radius ris in unit of Bohr radius.
The rV(r) term is in unit of Bohr-Rydberg.

to half the distance between MgO nearest-neighbors. In Appendix A are shown the
input file regarding the atomic charge densities and the imput file relative to MgO
crystal structure.
Once the muffin-tin potentials have been obtained, was possible to compute ra-
dial matrices and phase shifts by means of PSRM (phase shift and radial matrix )
package, a MSCD internal package. Radial matrices describe the emission process
from a core/Auger level— in our case the Mg KLL line — and represent the inter-
action between the atomic potential and the electromagnetic radiation. Vice versa,
phase shifts represent the shifts due to scattering process between the emitted wave
and the scattered waves.
The PD calculation was performed using the Al-Kα line as a photon source,
to simulate a typical unpolarized X-ray source that exists in common laboratories.
Since in MSCD only linearly polarized photons are allowed (for example synchrotron
radiation), in order to simulate the diffraction pattern for unpolarized radiation we
need to calculate the diffraction pattern for two perpendicular direction of photon
polarization and then to sum them, as it was analyzed in Par. 1.2.1. However, since
we consider an Auger emission, we have seen that we can remove the (E· k̂) factor in
Eq. 1.7 so the photoemission intensity will be independent of photon polarization.

32
Considering those parameters treated in detail in Par. 1.3:

• for the emission process, we chose a p initial state (li = 1) and a s final state
typical of an Auger emission(see Par.1.2.1).

• for the scattering process, the multiple scattering order was set to 8 and the
Rehr-Albers approximation order set to 2. Such R-A order, as seen before, is
sufficient for a p initial state (For the detailed input file see Appendix A.1.3).

• concerning the cluster dimension, we used r = 8 Å and h = 18 Å , which


means that both 150 MgO and O atoms are involved in scattering process..

• in order to include inelastic scattering effects Eq.1.11 was used with Nv = 8,


Eg = 7.3 eV, M = 40.304 amu4 and ρ = 3.6 gcm−3 .

• in order to include vibrational effects we used Eq.1.6 and 1.14 with ΘD =


390K. Since the values of the photoelectron kinetic energies are extremely high
(∼ 1100 eV), we know that in forward-scattering-dominated AED vibrational
effects are not very significant.

• in order to consider electron refraction at the surface, we used Eq. 1.8 with
V0 = 10 eV. However, we observed that the AED pattern was not strongly
influenced by this parameter. This probably beacause the polar scan range
from θ = 0◦ to θ = 70◦ , not including emission angles near grazing.

Thus, we calculated two polar scan with θ = 0−70◦ along two different azimuthal
directions [110] and [100]. Fig. 1.13 (b) and (c) shows the calculated diffraction
pattern. We considered X-ray radiation incident on the MgO(001) surface; the [100]
and [110] azimuthal directions are those illustrated in figure. Considering the crystal
cubic symmetry (001) and (100) surfaces are equivalent, thus, hereafter, we will use
MgO(100) instead of MgO(001).
Fig. 1.15- 1.16 show the calculated diffraction patterns together with the main
crystal directions of MgO that corresponds to the most dense atomic chains. Ob-
serving the figures it is evident that the position of photoemission intensity maxima
corresponds to the directions of most dense atomic chains of MgO. 0◦ peak is obvi-
ously the most intense since in that direction the number of atoms is the greatest.

Thus, MSCD package represents a powerful instrument to calculate theorical


diffraction patterns that are fundamental to obtain, by fitting to experiment, atomic
structural information.
4
these values can be easyly derived from literature

33
Figure 1.15: (a) Auger electron diffraction polar scan of MgKLL electrons for a MgO(100)
crystal along the [110] azimuthal. (b) cross-section of MgO(100) along [110]
direction. The angles of the low-index directions correspondent to the most
dense atomic chains are also shown.

34
Figure 1.16: (a) Auger electron diffraction polar scan of MgKLL electrons for a MgO(100)
crystal along the [100] azimuthal. (b)cross-section of MgO(100) along [100]
direction. The angles of the low-index directions correspondent to the most
dense atomic chains are also shown.

35
Chapter 2

Experimental setup

This thesis has been realized at LASSE (LAyered Structures for Spin Electronics)
laboratory which is situated at L-NESS (Laboratory for Epitaxial Nanostructures on
Silicon and Spintronics) interuniversity center of Como, Italy. The LASSE apparatus
was built with the aim of integrating preparation, growth and analysis techniques
for nanostructured materials into a unique vacuum system.

2.1 LASSE Laboratory


Figure 2.1 shows a schematic view of the complete system, consisting of several UHV
chambers, connected by gate valves:

• an Introduction Chamber (IC);

• a Pulsed Laser Deposition Chamber (PLDC);

• a Sample Preparation Chamber(SPC);

• a Measurement Chamber (MC);

• a Photocathode Preparation Chamber (PPC).

Every chamber has indipendent pumping system and samples can be transfer from
one to another by means of magnetic or mechanic transfer arms.

2.1.1 Introduction Chamber (IC)


The Introduction Chamber (IC) serves as fast entry look and as a connection be-
tween the Pulsed Laser Deposition Chamber (PLDC) and the other ultra high vac-
uum chambers: the Sample Preparation Chamber (SPC) and the Measurement

36
Figure 2.1: Global view of the LASSE system devoted to growth and characterization
of nanostructured films. High and ultra-high vacuum chambers are intercon-
nected by gauge valves

37
Chamber (MC). The pression in the chamber, pumped through rotary and turbo-
molecular pumps can range from atmospheric pressure (during sample introduction)
to 10−7 Torr (at this pressure sample can be transferred to other chambers without
contaminating the system) in few minutes.

2.1.2 Sample Preparation Chamber (SPC)


Ultra high vacuum condition are reached in SPC (10−10 Torr) via turbo, ion pump
and getter pump. Layered magnetic structures can be grown in SPC chamber by
Molecular Beam Epitaxy (MBE). MBE is an Ultra-High-Vacuum (UHV)-based tech-
nique for producing high quality epitaxial structures with monolayer control. The
principle underlying MBE growth is relatively simple: a beam of atoms or clusters
of atoms, are produced by heating up a solid source in a Knudsen cell. They then
migrate in an UHV environment and impinge on a substrate surface, where they
can diffuse and eventually incorporate into the growing film.
The chamber has a cylindrical shape with vertical axis and internal diameter of
300 mm. The manipulator is inserted from above, while the main service ports are
all arranged along a circumference perpendicular to the chamber axis. The sample
can be heated up to 700◦ C and has five degrees of freedom (three translations, two
rotations): in this configuration it can be placed in the proper position for (i) thin
films deposition (ii) substrate sputter-cleaning by means of a differentially pumped
ion gun, (iii) surface quality control via Low Energy Electron Diffraction (LEED).
Thermal evaporation is achieved by electron bombardment of high purity metal
rods, for 3d metals like Fe or Co, or of Mo/Ta crucibles (see Fig.2.3) containing the
low melting point noble metals. We used such a kind of crucible as well as for the
evaporation of stoichiometric MgO. The evaporation cells, all pointing to a common
position (the chamber center), are surrounded by a water cooled jacket so as to keep
the pressure well below 5·10−10 Torr during operation. Well defined deposition times
are achieved by operating a rotary shutter, which also allows co-evaporation from
two different cells.
A quartz crystal micro-balance placed in the focus of the Knudsen cells monitors
the evaporation rate. Typical growth rates are 0.2 Å/min (MgO) and 1 Å/min
(Fe, Co) for deposition of ultrathin films (up to some hundreds Å), ∼ 15 Å/min for
thicker Fe films (up to some thousands Å) via a supplementary evaporation cell that
can be placed closer to the sample. Oxidation treatments can be performed as well,
introducing molecular Oxygen by a leak valve.
An argon gun is available for cleaning of the surfaces via sputtering. The sput-
tering incidence angle is usually ranging from 30◦ to 90◦ with respect the surface of

38
the sample, and the flux of 1,5 keV Ar+ ions is generally kept at 1 µA/cm2 .
In Low Energy Electron Diffraction (LEED) (see 2.2) low energy electron (typ-
ically in the range 50 - 200 eV) impinge on the sample. The sample itself must be
a single crystal with a well-ordered surface structure in order to generate a back-
scattered electron diffraction pattern. The diffraction pattern is recorded and anal-
ysis of the spot positions yields information on the size, symmetry and rotational
alignment of the adsorbate unit cell with respect to the substrate unit cell. In the
SPC chamber the LEED optics is retractile and is protected by a movable shutter
from possible damages during ion sputtering and evaporation.

Figure 2.2: Schematic view of the LEED optics

2.1.3 Measurement Chamber (MC)


Ultra high vacuum condition are reached in MC (10−10 Torr) via turbo, ion pump
and getter pump. Sample characterization is performed in the Main Chamber (MC)
by means of electron spectroscopies (X-ray Phtoemission Spectroscopy (XPS), Ul-
traviolet Photoemission Spectroscopy (UPS), Spin resolved Inverse Photoemission
Spectroscopy (SPIPE)) and Magneto Optical Kerr Effect (MOKE). In the Measured
Chamber the (x, y, z, θ, φ) manipulator, which is also equipped with a liquid nitro-
gen cooling circuit, lies horizontally, i.e., along a diameter of the chamber central
body. The chamber is equipped with a hemispherical energy analyzer (HEA), while
the probing exciting sources available are a Mg/Al cathode X-ray source, an He
discharge ultraviolet (UV) lamp.

39
Figure 2.3: Schematic view of the Knudsen cell. Crucible contains the compound to be
evaporated.

Photoemission Spectroscopy

Photoemission is the most important experimental techniques for studying the sur-
face band structure of occupied and empty electronic states, respectively. Depending
on their kinetic energy the mean free-path of the electrons in a solid ranges from
about 5 Å up to some hundreds of Ångstroms [29]. In the case of photoemis-
sion for the photon energies employed in on-campus laboratories (typically 21.2eV
and 1486.67 eV) the technique is surface sensitive and thus well suited for surface
studies. In the UV, range probe depths of the order of 5÷10 Å can be obtained,
depending on the investigated sample, while in the XPS regime (hν=1253.67 eV
and hν=1486.67 eV for the most commonly used Kα lines of Mg and Al), the probe
depth is of the order of a few tens of Å, strongly depending on the binding energy
of the core levels under investigation.
Photoemission spectroscopy performed with UV photons (UPS) or with X-ray
photons (XPS) is based on well-known photoelectric effect: when a solid is irradiated
by monochromatic photons, these excite electrons from occupied states into empty
states (within the solid), hence they are released into vacuum (free-electron plane-
wave states) and detected by an electron energy analyzer. Thus the kinetic energy
of the emitted photoelectron is determined and its wave vector kex outside the solid

40
is derived from its energy and the direction of the analyzer aperture with respect to
the sample orientation [30].

Figure 2.4: Description of a photoemission experiment. (a) Definition of the angles and
wave vectors of the incident photon (~ω) and emitted electron e− . (b) Repre-
sentation of the photoexcitation process in the electronic band scheme E(k) of
a semiconductor. Only direct transition (ki 'kf ) are taken into account. The
energies of the initial state (Ei ) and final state (Ef ) are referred to the Fermi
level EF . (c) Conservation of the wave vector component k|| , (parallel to the
surface) upon transmission of the emitted electron through the surface [31].

Optical excitation by a fixed photon energy ~ω populates empty states in the


crystal above the vacuum level and the corresponding energy distribution of the
electron measured outside the crystal yields a qualitative image of the distribution
of the occupied crystal states (valence and core level states depending on the value of
~ω). The measured distribution of the sharp peaks is superimposed on the secondary
background, which arises from electrons that have lost quasi-continuos amounts of
energy due to multiple inelastic scattering events in the crystal. The sharp peaks in
the spectrum correspond to the kinetic energy EK referred to the analyzer electrodes
given by

EK = ~ω − Ei − φ (2.1)

41
where Ei is the binding energy of the initial state (to be determined) and φ the
work function of the analyzer. All energies in such photoemission experiment are
conveniently referred to the Fermi level EF of the sample.
A rigorous theoretical approach to the photoemission process requires a full
quantum-mechanical treatment of the complete coherent process in which an elec-
tron is removed from an occupied state within the solid and deposited at the detector
(one-step process). The less accurate but simpler and more instructive approach is
the so called three-step process in which the photoemission process is artificially sep-
arated into three independent parts:
(i) optical excitation of an electron from an initial into a final electron state within
the crystal;
(ii) propagation of the excited electron to the surface;
(iii) emission of the electron from the solid into vacuum crossing the surface.
While in principle these three steps are not independent of each other, in the three-
step model the independent treatment of these three contributions leads to a simple
factorization of the corresponding probabilities in the photoemission current of the
emitted electrons.
(i) The optical excitation of an electron in the first step is simply described by the
golden-rule transition probability for optical excitations

2π 2π
Wf i = | hf, k|H|i, ki |2 δ(Ef (k) − Ei (k) − ~ω) = mf i δ(Ef − Ei − ~ω) (2.2)
~ ~

here the perturbation operator H involves the momentum operator p and the
vector potential A of the incident electromagnetic wave: H = e/2m(A · p + A · p).
The δ function describes the energy conservation in the excitation of an electron from
the initial state |i, ki with energy Ei (k) into the final state hf, k| with energy Ef (k)
of the electronic band structure. The matrix elements mf i defines the selection rules
for the observability of a particular initial state |i, ki on the basis of the experimental
geometries and the symmetry of the initial and final electronic states involved in the
process.
(ii) The second step in the model is the propagation of the electrons in the solid to
the surface. A large number of electrons undergo inelastic scattering process; they
lose part of their energy by electron-plasmon or electron-phonon scattering. Such
electrons contribute to the continuous background in the photoemission spectrum
which is called the true secondary background; they have lost the information about
their initial electronic level Ei . The probability P(z) that an electron at a depth z
reaches the surface without inelastic scattering is then related to the mean-free path

42
λ as follows
P (z) = e−z/λ (2.3)

In general, λ depends on the kinetic energy of the electrons E, their wave vector
k and on the particular crystallographic direction.
(iii) The third step, i.e. the transmission of the photoexcited electrons through the
surface, can be considered as the scattering of a Bloch electron wave function from
the surface-atom potential with translational symmetry parallel, but not normal to
the surface. In this case because of the 2D translational symmetry in the plane
of the surface, but not in the direction normal to the surface (Figure ??) only the
parallel component of the wave vector is conserved. The considerations are of great
importance in the case of UPS, where the wave vectors Q of photons is negligible
with respect to the electron wave vectors. In XPS, instead, Q is not negligible
and angular effects are smeared out due to non perfect collimation of the incoming
photon beam. In the case of UPS the in plane component of k is conserved, apart
from the a vector G of the surface reciprocal lattice

kex
|| = k|| + G|| (2.4)

where k|| and kex|| are the wave vectors of the electron inside and outside the
crystal respectively. The component normal to the surface k⊥ instead, is not con-
served during transmission through the surface. For the electron emitted into the
vacuum, the kex|| value is determined by the energy conservation requirement

~2 k ex 2 ~2 ex 2
EK = = (k⊥ + k||ex 2 ) = Ef − Evac (2.5)
2m 2m
If we introduce the analyzer work function φ =Evac -EF and the (positive) binding
energy EB referred to the Fermi level EF one also has

~ω = Ef − Ei = EK + φ + EB . (2.6)

The wave vector component parallel to the surface outside the crystal, then is
determined from the knowledge of the emission angle θ with respect to the normal
to the surface
r r
2m p 2m
k||ex = 2
~ω − EB − φ sinθ = EK sinθ (2.7)
~ ~2
This gives the internal wave vector component k|| through Equation 2.4.
On the other hand, the wave vector component k⊥ of the electron inside the
crystal is changed upon transmission through the surface. The expression for this

43
component is
r
2m
k⊥ = (EK − V0 ) cosθ (2.8)
~2
where V0 is the inner potential of the crystal. There are several methods to
determine V0 (k⊥ ) in the solid. One can either try to identify the extrema of electron
bands at the critical points and compare them with band-structure calculations. If
no calculations are available, one can in some cases approximate k⊥ by assuming
free-electron-like final states, which are offset from the vacuum potential by the
inner potential V0 of the crystal. The k⊥ outside the crystal is determined by
energy conservation according to the Equation 2.5 as
r r
ex 2m 2 =
2m
k⊥ = E K − (k || + G|| ) EK cosθ (2.9)
~2 ~2
In a photoemission experiment one uses an electron analyzer that accepts electron
within an adjustable collection angle. In case this is large enough an integration for
every possible k|| in the first Brillouin zone can be performed, according to Equation
2.7.

Inverse Photoemission and MOKE

Inverse photoemission spectroscopy (IPES) can be seen as the time-reversal of pho-


toemission process. Electrons with well-defined energy impinge on the crystal, and
are thereby injected into empty electronic states [30]. They then decay into empty
states at lower energy and the corresponding excess of energy is released as a photon.
Observing such transitions allows for the investigation of the unoccupied electronic
structure. The theoretical description of the process is similar to that of the nor-
mal photoemission process (three step approximation). Spectroscopy of unoccupied
states (above EF ) can be performed by varying the primary energy of the injected
electrons and detecting photons at fixed photon energy ~ω0 (isochromat mode). In
the dispersive mode, instead, the primary beam has a fixed energy and the spectro-
scopic analysis of the emitted UV radiation is performed by means of a UV spec-
trometer. In our experimental setup the isochromat mode has been implemented,
based on a bandpass photon detector and an electron gun employing a GaAs photo-
cathode. If a polarized electron beam is used this technique is called SPIPE (Spin
Polarized Inverse Photoemission). To obtain the polarized electron beam, the spin
polarized photoelectrons are excited by circularly polarized light when photon en-
ergy just greater than the band gap of the GaAs photocathode. The electron spin
polarization is reversed by changing the polarization of the circularly polarized light

44
by a Pockels cell. The Magneto-optical Kerr Effect (MOKE) is observed as a net
rotation and elliptical polarization of incident linearly polarized light as it is re-
flected from a magnetized sample. This change in the polarization state, or ℘-state,
of an incident electromagnetic wave arises from the interaction of the electric and
magnetic fields of the waves with the spin of the electrons in the material. The
magnitude of this change in polarization is proportional to the magnetization of the
sample. Linearly polarized light can be depicted as a combination of equal amounts
of right and left circularly polarized light. Right and left circularly polarized light
effectively have different index of refraction in magnetized media so that they are
absorbed and re-emitted in a different way depending on the direction and strength
of the sample’s magnetization. The reflected light is then the sum of unequal pro-
portions of right and left circularly polarized light; that is, the reflected light is now
elliptically polarized with its axis of polarization rotated by an amount θk called the
Kerr angle. By passing this reflected light through a crossed polarizer, we can pick
off the component orthogonal to the direction of the incident light. By measuring
its magnitude as a function of an external magnetic field applied to the sample, a
hysteresis loop for the sample can be observed.

2.1.4 Photocathode Preparation Chamber (PPC) and Pulsed


Laser Deposition Chamber (PLDC)
The Photocathode Preparation Chamber (PPC) is dedicated to the preparation of
the polarized electron source for Inverse Photoemission Spectroscopy. Ultra high
vacuum condition are reached in this chamber (10−10 Torr). Materials with complex
stoichiometry, such as high TC superconductors, ternary semiconductors, ferrites,
manganites, can be more effectively grown by Pulsed Laser Deposition (PLD) in
PLDC chamber.
This is a conceptually simple technique based on vaporization induced by pulsed
laser ablation of a target surface, giving rise to the formation of a plasma (the
characteristic plume) which in turns deposits onto a suitable substrate. The layer
by layer growth is monitored by means of a doubly differentially pumped Reflection
High Energy Electron Diffraction (RHEED) apparatus, which can operate at high
pressure (P∼ 10−1 Torr) by means of a differential pumping.

45
2.2 Experimental setup for XPD measurements
During my thesis I worked to create an automated measurement system for the
acquisition of XPD scans. Fig.2.5 shows the basic experimental setup for XPD
measurements. It consists of:

• an excitation source (X-rays lamp), with Mg-Kα (E=1253.67 eV) and Al-Kα
(E=11486.67 eV) lines;

• a specimen holder with 2 axis of angular motion θ and φ to acquire polar and
azimuthal scans;

• a hemispherical electron analyzer (HEA, Hemispherical Energy Analyzer ) with


good angular resolution(up to 1◦ ).

Figure 2.5: XPD basic experimental setup

Since scanning angle obviously involves an added cost in time for any study, it is
desiderable to have an automated scanning of spectra, determing of peak intensities
by more accurate area-integration or peak-fitting procedures, and stepping of angles
under computer control. For this reason I create a dedicated sofware by means of
Labview 7.1 c
. With such a software several operations can be done:

1. the acquisition of XPS spectra in a specific energy range.

46
2. The acquisition of XPD polar (θ) scans. The software controls the sample
angular position θ and , at the same time, aquires the photoemission spectra
correspondent to a specific atomic core level. Acquiring XPD spectra relative
to different atomic core levels allows us to study the local order surrounding a
specific atomic species. The measured XPS spectra is convolved with a Gaus-
sian function of FWHM (full width at half maximum of 0.85 eV (for Mg-Kα
line) and 1 eV (for Al-Kα line). These values represents the experimental
analyzer linewidth regarding the two excitation sources. The convolution op-
eration produces a smoothing of the experimental peak acquired, reducing
experimental noise and then measurement uncertainty. After the convolution,
a linear background is subtracted to take in account the secondary electron
contribution. Thus, the photoemission peak area relative to a specific photoe-
mission direction is calculated (see Fig. 2.5). The final XPD scan will present
the polar angles θ on the x-axis and the peak area regarding different angles
on the y-axis.
Due to MC configuration, polar angle θ can range from -12◦ to 70◦ with steps
of 1◦ .

3. The acquisition of XPD azimuthal (φ) scans. The process is analogous to that
used to measured polar scans. In this case, the position of angle φ is controlled
by the software and at the same time photoemission spectra are acquired.
Due to MC configuration, azimuthal angle φ can range from -90◦ to 90◦ with
steps of 1◦ .

4. The acquisition of both polar and azimuthal scans. This is done varying
simultaneously both polar and azimuthal angles. Thus we obtain 2D data
curves that can be used to derived 3D crystal structure.

5. The possibility to set the sample center relative to the azimuthal angle φ.
This is done rotating the sample by an angle θ, to which corresponds a well
known crystal direction, and acquiring an azimuthal scan. The maximum
of photoemission intensity will be for φ = φ0 . This operation is important
for our measurement system because when the substrate is mounted on the
manipulator we don’t have an accurate knowledge of the φ0 position so we can
commit an error of few degrees. On the other hand with this procedure the
error committed can be reduced to less than 1◦ .

47
2.2.1 Hemispherical Energy Analyzer optimization
To optimize the angular resolution in XPD measurement was necessary to study
the modes in which the HEA may be operated to minimize angular acceptance
maintaining a count rate enough to ensure a good statistics in data acquired. As
hemispherical electron energy analyzer we used a PHOIBOS 150 c
with a 9 channel-
tron detector [32]. The main components are shown in Fig.2.6. X-rays illuminate
an area of a sample causing electrons to be ejected with a range of energies and
directions. The electron optics, which may be a set of electrostatic and/or magnetic
lens units, collect a proportion of these emitted electrons defined by those rays that
can be transferred through the apertures and focused onto the analyzer entrance
slit (S1) by means of electrostatic lenses (T1-T10) . Electrostatic fields within the
hemispherical capacitor of radius r0 are established to only allow electrons of a given
energy (the so called Pass Energy PE) to arrive at the detector slits (S2) and onto
the detectors themselves. After the S2 slit electrons are amplified and counted by
a 9 channel detector (C1-C9). Actually the analyzer radius ranges from 0.75 r0 to
1.25 r0 then electrons with kinetic energy of Ek = Ep ± ∆Ep can arrive at the detec-
tor. The electrons within this ∆Ep range will be counted distinctly by channeltrons,
that presented an energy dispersion, where the nominal value Ep corresponds to the
central channeltron (C5).
The analyzer acceptance area Aacc can be defined by adjusting a tunable system
of 7 different entrance slit and by a magnification factor M as:

DSLIT
Aacc = (2.10)
M

where DSLIT is the entrance slit (S1) dimension and M is a magnification factor
determined by electrostatic lenses. There are 3 lens modes that are distinguished
by different magnification factor and angular acceptance:
1. High Magnification Mode (HM) with M=10 and angular acceptance up to ±9◦

2. Medium Magnification Mode (MM) with M=5 and angular acceptance up to


±6◦

3. Low Magnification Mode (LM) with M=2 and angular acceptance up to ±3◦
In XPD measurement the acceptance area is not so important, but since it can be
written as a function of the polar angle as Aef f ∝ A/ cos θ (see Fig.2.7), we need to
avoid that for angles near grazing the analyzer accepts electrons out of the sample.
On the other hand the analyzer angular acceptance is of primary importance in
XPD measuraments, especially concerning the possibility to distinguish the differ-

48
Figure 2.6: Main Components of the PHOIBOS 150 c
:
U0 main retardation voltage
Uchannel HV -Uchannel base detector voltage
LA...LE lens potentials
T1...T10 electrostatic lenses
S1 hemispherical capacitor entrance slit
S2 hemispherical capacitor exit slit
IH inner hemisphere
OH outer hemisphere
r0 nominal capacitor radius (150 mm)
C1...C9 channeltrons

ent features of diffraction patterns: in fact, if the angular acceptance is too high,
photoemission intesity relative to different angles is mediated in the same measure-
ment producing a resolution reduction. Concerning the angular acceptance of the
measurement system, is possible to limit it up to values of ±1◦ by the use of a Iris
aperture (in the first diffraction plane of these lens modes)(see Fig.2.6). If we con-
sider the electrons beam that arrives to the analyzer like a Gaussian-profile beam,
the Iris aperture behaves suppressing intensities in the tail region of the Gaussian
beam. Moreover we have to consider that by means of the Iris aperture we reduce
not only the angular acceptance but also the analyzed sample area and then the

49
Figure 2.7: Analyzer acceptance area as a function of the polar angle θ.

Acceptance Iris Iris Iris


Angle (HM) (MM) (LM)
±1◦ 3 mm 3 mm 4 mm
±2◦ 6 mm 6 mm 8 mm
±3◦ 9 mm 9 mm 12 mm
±4◦ 12 mm 12 mm
±5◦ 15 mm 15mm
±6◦ 18 mm

Table 2.1: Acceptance Angle as a function of the Iris diameter for the 3 magnification

photoemission signal. Tab. 2.1 shows the angular acceptance as a function of both
the magnification factor and the Iris diameter. We can notice that for the same Iris
diameter the angular acceptance is lower in Low Magnification.

2.2.2 XPD measurements


The automated measurement system was tested for the case of an MgO(100) crystal
cleaned by annealing up to 600◦ C. Fig. 2.8 shows two different acquisition:

• a polar scan using the MgKLL Auger intensity peak with energy steps ∆E = 1
eV (red line);

• a polar scan using the MgKLL Auger intensity peak with energy steps of
∆E = 0.2 eV (black line).

50
Comparing the two curves we can observe that they are in good agreement. It was
estimated that the maximum error commited calculating the peak area in the case
∆E = 1 ev is only 5% compared to the case ∆E = 0.2 eV. Thus, we can conclude
that a high energy resolution is not essential in XPD measurement and using a higher
∆E step we can reduce significantly the measurement time. However, this result is
strictly correct only if we consider a high intensity photoemission peak coming from
bulk materials(in Fig2.8the MgKLL peak intesity is about 7000 counts) therefore the
error committed increasing the ∆E step is minimum. When the photoemission signal
is weak — for example acquiring the MgKLL peak intensity from few monolayer of
MgO deposited onto a Fe substrate(where this signal can be attenuated by a factor
10 compared to the value in the bulk) — is necessary to use at least energy steps
of∆E≤0.5 eV.
To the contrary the Iris values and the analizer modes play a key role for the
optimal angualr resolution. In Fig. 2.9 are shown AED spectra of MgKLL for a
MgO(100) crystal along [110] direction using a 5mm Iris for two different magni-
fication modes: LM (red curve) and MM (black curve).The corresponding angular
acceptance are: ±1◦ for the Low Magnification and ±2◦ for the Medium Magnifica-
tion. From Figure 2.9 we can observe that in general for the same photoemission
signal and iris values, the LM curve shows a higher angular resulution being able to
distinguish more accurately all the characteristic peaks of the spectrum.
Looking for a compromise between a good angular resolution and a high intensi-
tity photoemission signal we have decided to operate in “Low Magnification” mode
with a 5 mm Iris. In this manner we have a strong photoemission signal and angular
acceptance of about ∼1◦ . With this configuration a both polar and azimuthal scan
has been carried out for a MgO(100) crystal. Fig.2.10 shows experimental results
together with the main crystal directions relative to (100) surface. We acquired a
polar scan with θ = 0◦ − 70◦ for each φ = 0◦ − 360◦ angle. In this way we can
obtain the entire MgO crystal structure relative to the (100) surface. In Fig.2.10 are
clearly visible peaks for the crystallographic directions corresponding to the more
dense atomic chains: the most intense central peak is relative to [100] direction,
moreover are visible the four equivalent h110i directions, the four equivalent h111i
directions and the four equivalent h221i directions.

51
Figure 2.8: (a) AED da Mg KLL per un cristallo di MgO(100) lungo l’azimut [110]: (curva
nera) AED in cui le intensità dei picchi di emissione sono state misurate con
∆E=1 eV, (curva rossa) XPD in cui le intensità dei picchi di emissione sono
state misurate con ∆E=0.2 eV. (b) Picco Mg KLL con ∆E=0.2 eV. (c) Picco
Mg KLL con ∆E=0.2 eV.

52
Figure 2.9: AED of Mg KLL along the MgO [100] direction with 5 mm iris: (red curve)
Low Magnification, (black curve) Medium Magnification.

Figure 2.10: XPD, polar and azimuthal scan of Mg KLL for MgO(100) crystal

53
Chapter 3

Study of MgO/Fe(100) structure

3.1 Introduction
In recent years the realization of Magnetic Tunnel Junction(MTJ) devices has be-
come a major research field. In these devices two ferromagnetic layers, with different
coercive fields, are spaced out by a thin insulating layer in order to avoid the cou-
pling between their magnetic moments. If at the Fermi level the spin polarization
keeps the same sign in the two ferromagnets, the tunnelling electron current is larger
for parallel configuration (P), while it is lower for an antiparallel configuration (AP).
An external magnetic field can then force one of the two configurations P or AP
and so it can tune the resistance of the device. For a 3d ferromagnetic electrode,
the Tunneling Magneto Resistance can be understood in terms of a two-band model
in which the d-band is split (by the exchange interaction ∆ex ) into spin-up and
spin-down bands with different density of states at the Fermi energy (see Fig. 3.1).
When the magnetization of the layers is parallel, the majority-band electrons tun-
nel across to the majority band of the opposing electrode and the minority to the
minority band. When they are antiparallel, the majority/minority band electrons
are forced to tunnel into the minority/majority band of the opposing electrode. The
reduced number of states available for tunneling between the ferromagnetic layers
when the layers are antiparallel results in an increased tunneling resistance. The
reality is somewhat more complicated than this simple picture but it will not be
treated here.
The efficiency of such a device can be expressed in terms of Tunnelling Magneto
Resistance (TMR):
RAP − RP
T MR = (3.1)
RP
where R è is the resistance depending on the different configuration.

54
Figure 3.1: Spin-polarized tunnel effect in Ferromagnet/Insulator/Ferromagnet junctions
for parallel configuration (a) and antiparallel configuration (b).The dendity of
states (DOS) for the spin-up and spin-down bands are shown at the bottom
( ∆ex is the exchange interaction).

In particular MTJ based on Fe/MgO/Fe heterostructures have been recently de-


veloped; the main reason is that Fe is a ferromagnetic material with a high Curie
temperature and, using a MgO barrier, TMR values up to several 1000% were pre-
dicted theoretically [4]. However, experimentally measured TMR values at room
temperature are tipically of 180–200% [3], this is due to non ideality in the sto-
ichiometry and in the crystal structure at the ferromagnet/barrier interface, that
play a crucial role for the performance of MTJ. In fact, in tunnel effect plays a
key role the ferromagnet/barrier interface that selects by means of symmetry rules
some electronic bands rather than other, leading to a different conductivity of the
electrons that belong to states of different symmetry [4]. Recent x-ray diffraction
(XRD) experiments have suggested the formation of a substoichiometric layer of
FeO when MgO is deposited onto Fe(100) surface [5], by contrast the presence of
this layer is not observed by other experiments [8]; therefore is still an open question
not clarified in literature. The study of a possible Fe oxidation is really important
since theoretically calculations dimonstrate that an atomic layer of iron-oxide at
the interface between Fe substrate and the MgO layer greatly reduces the tunneling
magnetoconductance and consequently TMR ratio [7].

55
In the present work the MgO growth from 2 ML up to 14 ML1 onto the Fe(100)
surface was studied by means of XPD, a sensible to short-range order technique
and therefore suitable to determine the crystal structure at the interface. The
MgO growth relative to two different surfaces was studied: the Fe(100) and the
Fe(100)-p(1×1)O. The former surface used at first in the production of MTJ with
an MgO barrier, presents, as said above, instability problems due to high reactivity
of the clean Fe surface and thus is technologically diffucult to realize. The latter,
more stable, is a Fe surface oxided in a controlled way obtained by the adsorption
of 1 ML of oxigen on the Fe surface that reconstructs 1×1. This surface is less ex-
poseded to further oxidation since the monolayer of oxigen saturates the Fe dangling
bonds. The two surfaces are slightly different from a magnetic point of view. The
Fe(100)-p(1×1)O surface shows in some points of the Brillouin zone an increase of
the exchange splitting between majority and minority spin states of approximately
50% larger than in the clean surface, and globally an enhancement of spin depen-
dent effects at the interface [9]. This suggests an increase of the surface magnetic
moment that can be qualitatevely explained with the fact that the adsorption of 1
ML of oxigen gives rise to a Fe-Fe bond lenght at the surface larger than in the bulk,
with an increase of the magnetic moment towards the value of isolated Fe atoms.
The surface magnetic moment increase and a well ordered crystal structure are both
fundamental properties for the realization of MTJ with a high TMR ratio.

3.2 Sample preparation


Before being introduced in the vacuum chamber, the MgO substrate was cleaned
by a 10 minutes ultrasonic bath in isopropanol and 20 minutes in deionized water.
Subsequently the substrate was then cleaned by annealing up to 600◦ C until the
pressure was below 2 × 10−9 Torr. The low-pressure value in the chamber is enough
to avoid, once the annealing is terminated, a possible material readsorption on the
sample surface. As reported extensively in literature, the only way to obtain a carbon
free surface is grow on the top stoichiometric MgO [33]. Futhermore the MgO/MgO
omoepitaxy allows us to achieve a more ordered crystal structure (in fact spots in
LEED images are less modulated in this case). Thus 10 nm of MgO was deposited
at the slow rate of ∼0.2 Å/min. The evaporation cell of the MgO was composed by
a piece of MgO single crystals put into a Tantalum crucible ad heated by electron
beam bombardment. The Knudsen-cell parameters were: heating filament current
I=2.5 A and crucible voltage HV=2.4 kV (see Fig. 2.3). The coverage thickness,
1
12–14 ML are typical barrier thickness used for MTJ devices

56
Figure 3.2: Top view of the mutual position of Fe and MgO lattice.

of about 10 nm, was estimated by calibrating the evaporator in situ with a quartz
balance. After the deposition, the sample was treated with a 20 minutes annealing
at 300◦ C necessary to reorder the crystal structure.
A 100 nm Fe film at the growth rate of ∼8 Å/min was then deposited onto
the MgO. The Fe Knudsen-cell parameters were: heating filament current I=4.85 A
and Fe rod voltage HV=2.4 kV. During the entire deposition process the chamber
pressure was lower than 1.26×10−9 Torr.
There are several studies in literature that relate the Fe bcc 2 epitaxial growth onto
MgO(100) and vice versa. For the deposited Fe onto the MgO(100) and for the MgO
deposited onto the Fe(100), the Fe[100] grows parallel to the MgO[110](Fig.3.2). In
fact the Fe lattice parameter in the [100] direction is 2.87 Å, whereas the MgO lattice

parameter in the [110] direction is 2.98 Å (=4.21 22 Å) leading to a mismatch of only
3.8%. Futhermore LEED experiments suggest that Fe atoms sit on top of oxigen
atoms [34].
The obtained Fe(100) surface can be employed as starting point for the realization
of the two specific interfaces considered in the study.
2
body-centered cubic

57
Figure 3.3: Fe(100)-p(1×1)O LEED with electron energy of 133 eV.

3.2.1 MgO/Fe(100)-p(1×1)O
Before proceeding to the creation of the Fe-p(1×1)O surface it is necessary to obtain
a perfectly clean Fe(100) surface. XPS and LEED techniques were used in order
to detect the presence of contaminations. Initially a small quantity of oxigen was
found (¡3%). The cleaning procedure consists in several cycles of sputtering with
Ar+ ions and annealing until there is no presence of contaminations (within the XPS
accuracy of ∼1h). A single cycle consists in 45 minutes of sputtering (accelerating
voltage for ions=1.5 kV, sample current=1.5 µA) followed by a 5 minutes annealing
at 500◦ C and a flash-heating at 620◦ C for a few seconds.
Once a clean Fe surface is achieved the Fe(100)-p(1×1)O surface is obtained
exposing the sample to 30 L (1 L = 1 1 × 10−6 Torr s) of O2 at 300◦ C and then flash-
heated at 620◦ C for few seconds. This procedure gives a nicely ordered structure,
as evidence by the sharp p(1×1) LEED pattern(Fig.3.3). The coverage of oxygen
obtained with this procedure correspond to about 1 ML of oxygen.
Subsequently MgO has been evaporated step by step to obtain epitaxial films
with thickness of 2ML, 4ML and 14 ML. The MgO epitaxial growth onto Fe hap-
pens in a similar way as Fe onto MgO and is favored since surface free energy of
MgO(1.1 J/m2 )is less than the surface free energy of Fe (2.9 J/m2 ). The MgO
Knudsen-cell parameters are the same reported above. The coverage thickness was
initially estimated by calibrating the evaporator in situ with a quartz balance and
subsequently verified by measuring the Fe2p3/2 intensity attenuation and applying

58
Figure 3.4: Fe2p3/2 peak intensity attenuation as a function of the MgO thickness for
the MgO/Fe(100)-p(1×1)O heterostructure: (black curve) theoretical atten-
uation calulated applying the Beer-Lambert law, (red curve) experimental
attenuation.

the Beer-Lambert law


−dM gO
I F e2p (dM gO ) = I0F e2p e cos θλ (3.2)

where I0F e2p is the Fe2p3/2 intensity peak obtained on the Fe-p(1×1)O layer, theta
the photoemission angle (θ = 0 at normal incidence)and λ is the escape depth for
Fe2p3/2 electrons in the MgO (λ =19.75 Å). The λ value was obtained from database
National Institute of Standards and Technology(NIST) using the TPP-2 formula (see
Eq.1.11), with asymmetry photoionization coefficient β=1.47, energy gap=7.3 eV,
number of valence electrons=8, photoelectron kinetic energy Ek (F e2p3/2 )=775 eV
and MgO density ρ=3.58 g/cm3 [35]. The error in the coverage estimate is in the
order pf ±0.2 ML.
In Fig.3.4 is reported the theoretical and experimental attenuation of Fe2p3/2
signal as a function of MgO thickness. The data are in good agreement within the
experimental error and thus we can conclude that the thickness of the layers is 2ML,
4ML and 14ML and interdiffusion phenonema are not present.
After each deposition the sample was annealed for 20 minutes at 300◦ C. As exten-
sively reported in the literature [3] this treatment is the best compromise necessary
to obtain a good crystallinity of the MgO and to preserve the interfaces between the
two ferromagnet layer and the MgO barrier. For every MgO coverage XPD polar
scans were acquired along the Mg[100] and Mg[110] azimuths.

59
Figure 3.5: LEED pattern with electron energy of 133 eV of a Fe film contaminated
by carbon: beside typical spots of Fe lattice, the spots due to the c(2X2)
reconstruction are visible.

3.2.2 MgO/Fe(100)
For the preparation of a carbon and oxygen free iron surface we followed the pro-
cedure described before, i.e. several sputtering and annealing cycles. In this case,
in addition to oxigen contaminations we found carbon contaminations forming a
c(2×2) pattern well noticable in LEED(see Fig.3.5). An oxidation treatment fol-
lowed by an annealing is necessary to reduce the quantity of carbon on the surface:
in this manner the CO formed on the surface during the oxidation (a more volatile
compound) desorbs with the annealing, leaving a cleaner surface. The oxidation
conditions were the same as described before.
Once a clean Fe surface was achieved MgO has been evaporated step by step to
obtain epitaxial films with thickness of 2ML, 4ML and 12 ML. The MgO Knudsen-
cell parameters are the same as said above. The coverage thickness was verified
measuring the Fe2p3/2 intensity attenuation as in the former case and using Eq.3.2
in which I0F e2p represents the Fe2p3/2 intensity relative to the clean Fe surface. As
shown in Fig. ?? for 2ML coverage thickness the measured Fe2p3/2 intensity atten-
uation is less than the theoretical one yielding a thickness estimate of 1.5 ML. We
then calculated the Mg1s and Fe2p3/2 intensity ratio (see Tab.3.1) and we noticed
that is the same as in MgO/Fe(100)-p(1×1)O heterostructure. In fact, considering

60
IM g1s (arb.units) IF e2p3/2 (arb. units) IM g1s /IF e2p3/2
MgO/Fe(100)-p(1×1)O 256000 514000 0.499
MgO/Fe(100) 339000 767000 0.443

Table 3.1: Mg1s and Fe2p3/2 intensities for the MgO/Fe(100)-p(1×1)O and MgO/Fe(100)
heterostructures

the Mg1s and Fe2p3/2 peaks, Mg1s peak intensity will be given by:
Z d dM gO
−λ
IM g1s ∝ NM g σM g1s T (KEM g1s ) e M g (M gO)

0 d g
 (3.3)
−λ M
∝ NM g σM g1s T (KEM g1s ) 1 − e M g (M gO)

while the Fe2p3/2 intensity will be given by:


dM gO
−λ
IF e2p3/2 ∝ NF e σF e2p3/2 T (KEF e2p3/2 )e F e (M gO) (3.4)

where N is the atom density, σ is a cross-section sensible to different electron states,


T (KE) is the analyzer trasmission dependent on photoelectron kinetic energy (KE),
λM g (M gO) represents the mean free path of Mg1s photoelectrons within the MgO
layers and λF e (M gO) represents the mean free path of Fe2p3/2 photoelectrons within
the MgO layers. The exponential factor appearing in Eq. 3.3 is given by, in a discrete
model, the intensity integration over the MgO thickness, whereas the exponential
factor in Eq. 3.4 represents the Fe photoelectron attenuation within MgO layers.
Thus, obtaining the same IM g1s /IF e2p3/2 ratio, means that the deposited thickness
dM gO is essentially the same, while for the MgO/Fe(100) sample the signal coming
from the substrate is less attenuated than predicted due to a potential rearrangement
at the interface [36]. This rearrangement can partially change the stoichiometry or
MgO crystal order as we show further.
This behaviour is reasonable compared also to SXRD (Surface X-ray diffraction
studies, as reported in literature, that show the high reactivity of the clean Fe(100)
surface with the formation of a FeO layer after the MgO deposition [5]. Although
our former XPS studies do not reveal a substantial Fe oxidation therefore the case
remains open.
After each deposition the sample was annealed for 20 minutes at 300◦ C and two
XPD polar scans were acquired along the Mg[100] and Mg[110] azimuths.

61
Figure 3.6: Fe2p3/2 peak intensity attenuation as a function of the MgO thickness for the
MgO/Fe(100) heterostructure: (black curve) theoretical attenuation calulated
applying the Beer-Lambert law, (red curve) experimental attenuation.

3.3 XPD: experimental measurements and theo-


retical calculations
As said in previous paragraphs, XPD measurement was carried out in order to
determine the MgO crystal structure at the interface for several thicknesses for the
MgO/Fe and MgO/Fe-p(1×1)O samples and clarify some effects, e.g. the surface
rearrangement observed for the MgO/Fe(100) heterostructure by XPS analysis.
The excitation source was a standard unpolarized Al-Kα X-ray source (E=1486.67
eV). The lines that can be used in diffraction measurements of an MgO crystal are
the Mg1s, O1s, O KLL and Mg KLL. In Fig. 3.7 the polar scans from a clean MgO
substrate are reported for several lines: Mg1s, O1s and Mg KLL. The Mg1s line
presents a high binding energy (BE ∼1100 eV), thus Mg1s photoelectron generated
by the Al-Kα (E=1486.67 eV) source will present a low kinetic energy (EK ∼300
eV) and therefore is not possible to take advantage of the forward scattering ef-
fect to determine the main crystallographic directions of the samples (see Par. 1.1).
Oxygen features were excluded: as a matter of fact, in the case of MgO/Fe-p(1×1)O
both the Oxygen embedded in MgO and in the Fe-p(1×1)O surface will contribute
to the total signal, while in the case of MgO/Fe only the MgO contribution will be
present, so that a direct comparison of the two MgO structure would be impossible.
We then chose the Mg KLL (BE= 315 eV) line, which means that the kinetic

62
Figure 3.7: Polar scans of an MgO(100) crystal for emission from different levels: Mg1s
(black curve),O1s (red curve) e Mg KLL (blue curve)

energy of the photoelectrons considered in scattering process is 1170 eV. In this


energy range the forward scattering effect is predominant so the maximum of the
signal will be along the more dense atomic chains (see Fig.1.2(b)). Moreover the
contribution to the total diffraction signal comes only from the MgO layer. Since the
initial line is an Auger line the scattering is not dependent on the light polarization
(see Par. 1.2.1). It is noticable that in our system the x-ray source is unpolarized so
using a core level instead of an Auger level doesn’t give us further structural infor-
mation. On the other side, using an Auger line simplifies theoretical calculations, in
the sense that the unpolarized light decomposition into two perpendicular polarized
components is not necessary3 .
Two polar scans with θ=-7◦ –70◦ along the [100] and [110] azimuths for each
thickness (2, 4 and 12-14 ML) for the MgO/Fe(100) and MgO/Fe(100)-p(1×1)O
samples were carried out. The analyzer parameters used are those described in
Par. 2.2.2(Low magnification, 5 mm Iris) with an angular acceptance of ±1◦ .
3
Since we chose the Mg KLL line from now on the measurements presented will be relative to
an AED experiment.

63
3.3.1 Subtraction of the isotropic background
AED is a powerful technique to investigate short-range order and the crystal struc-
ture that surrounds a specific emitting atom. We have to consider, although, that
the angular distribution of the intensity is originated by two contributions: one
anisotropic component, which is determined by the geometry of the local atomic
structure and one, slowly varying, isotropic component which depends on both in-
strumental factors (such as sample illumination and detector angular resolution) and
material dependent factors (such as atomic differential cross-section, film thickness
and escape depth) [37]. In order to obtain an AED curve representative only of
the local structure geometry is necessary to subtract the isotropic component from
the acquired measurements. Afterwards, the obtained data can be compared with
theoretical calculations that do not present this component. In order to calculate
the isotropic component I0 we took in account the phisically relevant parameters
in our measurement system (for the angle determination see Fig. 3.8(a)). These
parameters are:

1. The differential atomic cross-section dσnl /dΩ. It depends on the angle between
the directions of the polarization vector E and the photoelectron wavevector
k. As said in the first chapter, an Auger emission is isotropic (s-type wave),
so the (E · k̂) factor can be neglected. For this reason the I0 is not dependent
on the differential atomic cross-section and consequently on the angle.

2. The analyzer acceptance area. As presented in Par. 2.2.1, it varies as a function


of θ angle as Aacc = A/cos(θ). The contribution to I0 will be:

1
Ianal (θ) ∝ (3.5)
cos(θ)

3. The photon intensity that illuminates the sample. It depends on the angle
between the direction of the excitation source and the surface normal(α) giving
rise to an intensity factor:
IIll (α) ∝ cos(α) (3.6)

4. The inelastic mean free path (IMFP). The flux of electrons emitted at a depth
z and detected at a polar angle θ from the surface normal (see Fig. 3.8(b)),
will be reduced by inelastic scattering according to the Beer-Lambert law
I(z, θ) = e−z/λ cos(θ) . Integration over the depth from the surface yields the

64
Figure 3.8: (a) Schematic view of the XPD measurement system with angle definition
used in I0 calculation, (b) schematic drawing showing the variation of the
escape depth as a function of take-off angle θ.

angular dependence expected for an emitting slab of matter of thickness D


Z D
IIM F P (D, θ) ∝ I(z, θ)dz
0
h i (3.7)
D
∝ λ cos(θ) 1 − e− λ cos(θ)

which corresponds to the well known ∼ cos(θ) behaviour in the limit case
D −→ ∞ of an homogeneous semi-infinite emmiting volume.

Combining all these factors we obtain the final expression for the isotropic com-
ponent

I0 = kIanal (θ) · IIll (α) · IIM F P (D, θ)


h D
i (3.8)
= λ cos(α) 1 − e− λ cos(θ)

where k is a scale factor to be determined by fitting the data.


Fig. 3.9 shows the acquired AED curve I(θ) (black curve), the isotropic com-
ponent I0 calculated using Eq. 3.8(red curve) and the anisotropic component χ(θ)
(blue curve).

65
Figure 3.9: AED of Mg KLL for an MgO/Fe(100) sample for an MgO thickness of 4ML,
along the [100] azimuth. The acquired AED curve I(θ) (black curve), the
isotropic component I0 calculated using Eq. 3.8(red curve) and the anisotropic
component χ(θ) (blue curve) are shown.

66
From now on the AED curves will be presented in the form χ(θ) = (I − I0 )/I0
instead of I(θ).

3.3.2 Theoretical calculations with MSCD


In order to obtain further information regarding experimental data, MSCD simula-
tions were executed relative to MgO/Fe-p(1×1)O and MgO/Fe structures. Calcu-
lation parameters are the same used in Par.1.4.2:

• For the emission process, an initial p-state (li = 1) and a final s-state (char-
acteristic of an Auger emission)(see Par.1.2.1), considering only the (li − 1)
channel, were chosen.

• for the scattering process, the multiple scattering order was set to 8 and the
Rehr-Albers approximation order set to 2. Such R-A order, as seen before, is
sufficient for a p initial state (For the detailed input file see Appendix A.1.3).

• concerning the cluster dimension, we used r = 8 Å and h = 18 Å , which


means that both 150 MgO and O atoms are involved in scattering process.

• in order to include inelastic scattering effects Eq.1.11 was used with Nv = 8,


Eg = 7.3 eV, M = 40.304 amu and ρ = 3.6 gcm−3 .

• in order to include vibrational effects we used Eq.1.6 and 1.14 with ΘD = 390K.

• in order to consider electron refraction at the surface, we used Eq. 1.8 with
V0 = 10 eV.

Muffin-tin potentials and consequently phase-shift and radial matrix deriving from
them were recalculated with a lower muffin-tin radius (Rmt =0.94 Å) in confront to
the MgO substrate case (see Par. 1.4.2) to avoid the overlap between spherosymmet-
ric charge density. Because, including the Fe layers in the calculation, the nearest-
neighbor distance decreases to 1.98 Å (muffin-tin radius must be less than half the
nearest-neighbor distance).
Concerning the geometric structure, we assumed that, for both samples, the in-
plane atoms position is as illustrated in Fig. 3.2, i.e. the MgO[110] is parallel to the
Fe[100] and MgO grow compressed in plane. The in-plane lattice parameter a|| is
the Fe lattice parameter of the bulk (2.86 Å). Concerning the out-of-plane lattice
parameters we considered two different structures (see Fig. 3.10(a)):

67
• For the MgO/Fe(100)-p(1×1)O heterostructure we use the same structure ex-
perimentally derived by Meyerheim et al. and used by Zhang et al. in the
study of TMR variation due to the presence of a FeO layer [5] [7]. This struc-
ture is shown in Fig. 3.10: the distance from the FeO layer to the first MgO
layer was observed to be about 2.15 Å, the distance between the O in the
FeO layer and the Fe atom underneath is 1.89 Å. There is also a significant
buckling in the FeO layer of about 0.2 Å. We assume a relaxation of the first
MgO layer at a distance of 2.24 Å. The further MgO layers present a distance
of alternatively 2.15 and 2.18 Å.

• For the MgO/Fe(100) heterostructure we considered the Fe layers fixed at the


experimental lattice constant for bulk Fe of a=2.86 Å. We assume no vertical
relaxations between the layers and the Fe-O distance at the interface is taken
to be 2.16 Å. The distance between MgO layers is the same as in the previous
case.
The substantial difference between the two structure is localized at the interface
where, for the MgO/Fe-p(1×1)O sample, beyond the presence of 1 ML of Oxygen
a relaxation of the first MgO and Fe layers is considered. The out-of-plane lattice
parameter of the bulk a⊥ is alternatively 2.15 Å and 2.18 Åin both the structures. It
is larger than the bulk value (a⊥ =2.105 Å) because the MgO cubic cell, contracted
in the plane due to the lattice mismatch with the Fe, is tetragonally deformed by
expanding the out-of-plane lattice constant.
The curves calculated using these structures are essentially the same with the
exception of 2 ML coverage, where the effect of the different interface is more con-
sistent. The difference between the calculated curves is due to the relaxation of
the first MgO layer considered for the MgO/Fe(100)-p(1×1)O structure that can
be attributed to a possible Fe oxidation. The presence of the FeO layer and the
relaxation of the first Fe layer can contribute to the total difraction signal only in
a backward scattering regime then their contribution will be negligible, since the
Mg KLL photoelectron energy (∼1100 eV) involved in the scattering process deter-
mines a strong forward scattering regime. At high MgO coverage the contribution
of the first MgO layer will not result significant and the AED curves relative to the
structures will tend to be identical.
In Fig. 3.11 the theoretical calculations are compared for the two structures for
a MgO thickness of 2ML (azimuth[110]). The curves are similar, but it is noti-
cable that the peak at 52◦ for the MgO/Fe(100) is shifted on the right of about
0.9◦ compared to MgO/Fe(100)-p(1×1)O (a). Vice versa, the peak at 52◦ for the
MgO/Fe(100 is shifted on the left of about 0.9◦ compared to MgO/Fe(100)-p(1×1)O

68
Figure 3.10: Strutture geometriche utilizzate nei calcoli teorici con MSCD: (a)
MgO/Fe(100)-p(1×1)O; (b) MgO/Fe(100)

(b). A similar behaviour is observed as well along the [100] azimuth. This is the
only substantial difference that can be noticed between the two structures.
From now on the calculated curves relative to the MgO/Fe(100)-p(1×1)O case
will be used as representative of both structures, taking in consideration the differ-
ences at 2ML.

69
Figure 3.11: Calculated AED curves for 2ML for the MgO/Fe(100) (red curve) and
MgO/Fe(100)-p(1×1)O (black curve) (azimuth [110])

70
3.3.3 Data analysis and discussion
In Fig. 3.12-3.14 experimental AED data togheter with theoretical calculations are
presented.
At first is noticable that AED curves relative to both the structures are compara-
ble and present similar diffraction features as the film thickness is varied for both the
two azimuthal directions. In fact all the peaks of the χ(θ) function have essentially
the same angular position and shape in both samples. This means that the orienta-
tion of the main crystallographic directions and the bond-length are the same in both
cases. Furthermore we can observe that the deposited MgO thicknesses are substan-
tially correct, since the peaks intensity and their shape as a function of the thickness
(2 ML, 4 ML , 14 ML) are in good agreement with numerical calculations. This
confirms that MgO coverage calculated by means of XPS through Fe2p3/2 intensity
attenuation were correct. Furthermore, since no characteristic peaks of theoretical
calculations are suppressed in experimental data, the deposited film presents a high
degree of crystallinity.
At high coverage (12–14 ML)intensity modulations of AED curves for both the
samples result less than in theoretical calculations. This can be eventually attributed
to the presence of a partially disordered system in addition to the crystal system,
which presents low or null diffractive effects that decrease AED modulations.
Concerning the MgO out-of-plane lattice parameter (a⊥ ) estimate, in Fig. 3.15
are compared experimental data and theoretical calculations relative to the MgO/Fe(100)-
p(1×1)O (4 ML) case obtained using a⊥ =2.105 Å (red curve) and a⊥ =2.15–
2.18 Å (blue curve). In the former the structure illustrated in Fig. 3.10(a) was
used, but choosing a fixed lattice parameter a⊥ =2.105 Å starting from the second
MgO layer to simulate a MgO that mantains a lattice parameter equal to the bulk
one.
In the latter we used the structure presented in Fig. 3.10 (a) that considers the
distance between MgO layers alternatively of 2.15 Å and 2.18 Å [5]. This structure
is used in all theoretical calculations of this work and takes in account that for the
MgO cell the out-of-plane lattice constant a⊥ is expanded as a consequence of the
contraction of the in-plane lattice parameter.
Concerning the peaks located at 32◦ and 52◦ , calculations relative to the two
structures do not perfectly match the trend of experimental data, in fact the calcu-
lated peaks are shifted of few degrees. This kind of difference between theoretical
calculations and experimental data have been already observed in literature rela-
tive to other systems [38]. On the other hand, concerning the peak located at 65◦
in experimental curve, this feature is well fitted in the calculation with a⊥ =2.15–

71
Figure 3.12: Calculated (blue line) and measured AED curves of 2 ML of MgO onto
Fe(100) (red line) and Fe(100)-p(1×1)O (black line) along the [100] azimuth
(a) and the [110]azimuth (b).

72
Figure 3.13: Calculated (blue line) and measured AED curves of 4 ML of MgO onto
Fe(100) (red line) and Fe(100)-p(1×1)O (black line) along the [100] azimuth
(a) and the [110]azimuth (b).

73
Figure 3.14: Calculated (blue line) and measured AED curves of 14 ML of MgO onto
Fe(100) (red line) and Fe(100)-p(1×1)O (black line) along the [100] azimuth
(a) and the [110]azimuth (b).

74
Figure 3.15: Experimental(black dotted curve) and calcualted AED curves for
MgO/Fe(100)-p(1×1)O with a⊥ =2.105 Å (red curve)and a⊥ =2.15–
2.18 Å (blue curve)

2.18 Å, while in the simulations with a⊥ =2.105 Å is shifted at about 63◦ . It seems
therefore, according to other studies in literature [5], that MgO, adapting the in-
plane lattice parameter to Fe lattice parameter, expands in normal direction with
a⊥ =2.15–2.18 Å. del Fe, si espanda in direzione perpendicolare alla superficie con
a⊥ =2.15–2.18 Å. The valley observed in the centre of the peak at 0◦ in the sim-
ulations is due to an imprecision of MSCD software to simulate correctly the O
potential in some cases, already noticed in other works [39].

The Fig. 3.16, which is a zoom view of the Fig. 3.13(b) shows a similar trend ob-
served in the Fig. 3.11 for the theoretical calculations. It is noticable, in fact, that
the peak at 52◦ for the MgO/Fe(100)is shifted to the right of about 0.9◦ compared
to MgO/Fe(100)-p(1×1)O (a). Vice versa, the peak at 64◦ for the MgO/Fe(100)
is shifted to the left of about 1.5◦ compared to MgO/Fe(100)-p(1×1)O (b). This
behaviour confirms that the deposited films present the structural differences con-
sidered in theoretical calculations and moreover the relaxation of the first MgO layer
due to a partial Fe oxidation in the MgO/Fe(100) sample does not happen.

75
Figure 3.16: AED polar scans of 2 ML of MgO onto Fe(100) for: (a) MgO/Fe(100) (red
curve), MgO/Fe(100)-p(1×1)O (black curve)

The difference between polar scans relative to the two samples are essentially
two:

1. the different width of the peak at 0◦

2. the intensity attenuation of peaks located at high angles (θ ≥50◦ ) for the
MgO/Fe-p(1×1)O sample compared to MgO/Fe sample and the theoretical
calculations.

Width of the peak at 0◦

Observing the width of the peaks at 0◦ relative to the two samples, substantial differ-
ences are noticable. In Tab. 3.2 is shown the FWHM (Full width at half maximum)
of the 0◦ peak relative to both samples as a function of the MgO coverage for both
the azimuthal directions. In Fig. 3.17 the difference between the FWHM (∆FWHM)
relative to the two samples as a function of the MgO coverage is presented. It is
evident that the FWHM relative to Fe/MgO system is larger for every coverage. In
fact, fig. 3.17 shows a ∆FWHM of 7◦ at 2 ML, which decreases up to about 4◦ at
4 ML and the peaks have the same FWHM at 14 ML (for less than 1◦ ), i.e. when
the MgO barrier is complete. The effect is less evident along the [100] direction,
even if yields a ∆FWHM of about 5◦ at 2 ML, that decreases up to about 1◦ at
4 ML. This effect, very significant at low coverage that decreases with increased
barrier, can be explained with the presence of a higher disorder at the interface for

76
FWHM(deg.)[100] FWHM(deg.)[110]
2 ML 4ML Barrier 2 ML 4ML Barrier
MgO/Fe(100) 17.03 18.52 11.48 22.40 19.11 12.73
MgO/Fe(100)-p(1×1)O 12.37 17.62 10.55 15.43 15.22 11.27

Table 3.2: FWHM for the 0◦ peak as a function of MgO coverage for the MgO/Fe(100)
and MgO/Fe(100)-p(1×1)O samples along [100] and [110] azimuths

Figure 3.17: ∆FWHM relative to the MgO/Fe(100) and MgO/Fe(100)-p(1×1)O samples


as a function of MgO coverage along [100] (red curve) and [110](black curve)
azimuths

the MgO/Fe sample. The disorder type that contributes to the total XPD signal
and consequently to peaks broadening can be:

• the presence of a totally disordered system that, presenting minimal diffractive


effects, decreases XPD modulation; XPD;

• the sum of inchoerent domains whose single effect is to shift the peak to two
different directions, but once added generate the peak broadening.

The presence of higher disorder in the MgO/Fe sample in respect to MgO/Fe-


p(1×1)O is reasonable compared to the surface rearrangement observed during the
deposition and the fact that the Fe(100) is less stable than Fe-p(1×1)O surface.
Through theoretical calculations the peak broadening due to disorder has been
verified for 2 ML coverage, where it seems to be more relevant. Fig. 3.18 reports
the calculated polar scan ([110] azimuth) for two different domains, one where the
O atoms are not directly placed on top of the MgO ones but are shifted to the

77
Figure 3.18: Theoretical calculations of 0◦ peak broadening

right by 0.08 Å in confront of the equilibrium position ((a) curve in Fig.3.18) and
the other one where the O atoms are shifted to the left by the same ammount ((b)
curve in Fig.3.18). This means that the normal emission direction at 0◦ is now
tilted at 2◦ for the former and -2◦ for the latter. Finally we average between the
two contributions. From Fig. 3.18 it results that the broadening concerns almost
exclusively the peak at 0◦ , meanwhile the peaks located at θ ≥40◦ are not affected.
This is in agreement with experimental data where only the 0◦ peak broadening is
observed. The same procedure was carried out for a system where the O atoms
were shifted by 0.17 Å (4.5◦ tilt) and for a system where the O atoms were shifted
by 0.2 Å (5.3◦ tilt). Fig, 3.19 shows the Gaussian fit of the 0◦ peak relative to
these 3 structures together with the non deformed structure. It was observed that
the FWHM increases of about 1◦ in case of 2◦ tilt, 4◦ in case of 4.5◦ tilt and 6◦ in
case of 5.3◦ . Therefore the structure that presents a tilt of 5.3◦ is the one that best
approximates the ∆FWHM observed in experimental data. The 0.2 Å value which
represents the lateral shift is small compared to interatomic distances. A similar
broadening effect was detected in the study of CO adsorption on Pt(111) [40] by
Wesner.
The disorder due to the presence of inchoerent domains is easily included in AED
measurement since photoelectron signal is integrated by the analyzer over a relative

78
Figure 3.19: Gaussian fit of 0◦ peak calculated for three different structures: 0◦ tilt (green
curve), 2.1◦ tilt (blue curve), 4.5◦ tilt (red curve) , 5.3◦ tilt (purple curve)

large spatial zone of the sample (few millimeters).


At 4 ML coverage the curves do not present substancial differences. This means
that the higher degree of disorder noticed for MgO/Fe(100) sample is located at
the interface and after the first monolayers MgO films present the same structure
features. In fact, at this coverage, ∆FWHM decreases appreciably uo to 4◦ along
the [110] azimuthal direction and 1◦ along the [100] direction.
At 12–14 ML coverage the width of 0◦ peak is substatially the same, in fact the
effects of the first MgO layers weigh not much at this coverage.
The presence of higher disorder located at the interface in the MgO/Fe sample in
respect to MgO/Fe-p(1×1)O is reasonable compared to the surface rearrangement
observed during the deposition. In experimental data, however, the oxidation of the
Fe(100) surface, within the accuracy of the measurement technique (few%), is not
evident.

Intensity attenuation of the peaks located at high angles

A further difference between the samples is the fact that at high coverage (12–14
ML) is observed an attenuation of the peaks at high angles (θ ≥50◦ ), along both
azimuthal directions, realative to the MgO/Fe-p(1×1)O heterostructure compared
to the MgO/Fe sample and theoretical calculations. This result can be qualitatively

79
understood within the context of a directed surface roughness model in which there is
an oriented step morphology. In AED the presence of a nearest-neighbor scatterer at
a given angle is the controlling factor in determining the intensity. If one assumes a
directed surface roughness model such as the one shown in Fig. 3.20is evidendent that
the number of nearest-neighbor scatterers in some directions corresponding to near-
grazing angles (C) is less than in direction corresponding to near-normal emission
(A, B) compared to an epitaxial film. A similar behaviour has been observed in
previous XPD studies regarding a different system [41].
Thus, this interpretation seems to suggest that, regarding the MgO/Fe-p(1×1)
sample, the initial growth is layer-by-layer but when the film reaches a higher cov-
erage it relaxes the internal stress growing in a Volmer-Weber mode (islands) gen-
erating a step morphology.

Figure 3.20: Schematic drawing for the surface roughness: (a) film with an oriented step
morphology, (b) epitaxial film.

3.4 Conclusions
In this chapter the MgO growth on two different surfaces, the Fe(100) and the
Fe(100)-p(1×1)O, was studied by means of electron diffraction at Mg KLL line for
variable coverage from 2 ML up to 14 ML. In order to obtain further information
regarding experimental data, MSCD simulations relative to both the structures were
carried out. In order to comparing calculated and experimetal data the iosotropic

80
component I0 was subtracted from AED data. Such a component was calculated
taking in account the phisically relevant parameters in our measurement system.
A good cristallinity of the MgO film was observed relative to both the struc-
tures for every coverage. The film grown onto Fe(100)-p(1×1)O shows, however, a
higher crystal order mainly located at the interface compared to that grown onto
the Fe(100). This is in agreement with the surface rearrangement observed dur-
ing deposition and with the initial hypotesis that the Fe(100)-p(1×1)O surface, a
Fe(100) surface oxided in a controlled way, is more stable from chemical point of
view. Since the structural properties of the ferromagnet/insulator interface are fun-
damentals for the spin-polarized cunduction in MTJ’s devices and, as observed in
previous studies [9], the Fe(100)-p(1×1) shows a better magnetic behaviour com-
pared to Fe(100), we can consider the Fe(100)-p(1×1) as the best starting point for
the fabrication of magnetic tunnel junctions. As an ideal presecution of this work in
LASSE laboratory we are producing MTJ devices based on Fe/MgO/Fe(100) and
Fe/MgO/Fe(100)-p(1×1)O heterostructures on which TMR ratio will be measured
to verify these conclusions.
Furthermore, in our work tthe attenuation of AED peak intensity relative to
high angles for the MgO/Fe(100)-p(1×1)O sample was observed. This effect was
explained using a directed surface roughness model. Further developments can be
the study of the MgO crystal order by means of Fe2p photoelectrons for taking in
account more the effect of the different interface and to verify the possible existence
of a step morphology with complementary techniques, e.g. the Scanning Tunneling
Microscope (STM).

81
Appendix A

A.1 Input file for MSCD


In the following we report several input files used in MSCD simulations.

A.1.1 Input file for atomic charge density of Oxigen


------------------------------------------------------------------------------
i
8 1000 Z, NR (number of points in radial grid)
d
1 1=relativistic, 0=n.r.
x
0.0d0 0.d0=HF, 1.d0=LDA, -alfa = xalfa...
a
0 4 0.5 0.0005 100 relic,levels,mixing SCF, eigen. tol,for ech.
1 0 0 -0.5 1 2.00000000 n,l,l,-j (atomic quantum numbers),1,occupation
2 0 0 -0.5 1 2.00000000
2 1 1 -0.5 1 1.33333333
2 1 1 -1.5 1 2.66666666

w
atelem1.i name of output file
q

A.1.2 MgO crystal structure input file for Muffin-tin poten-


tial calculation
------------------------------------------------------------------------------
MgO FOR MgO(001) ,RELA

82
5.625562 SPA=Mg-Mg distance (atomic units)
0.5000 0.5000 0.0000 Coordinates of unit cell(SPA units)
0.0000 0.5000 0.5000
0.5000 0.0000 0.5000 Notice the value 0.5 (bulk calculation)
2 number of ineq. atoms in this file (NINEQ)
MAGNESIUM NAME
1 12.0000 0.0000 1.9800 #ATOMS in Unit Cell,Z,Valence,Muffin Radius(a.u)
0.0000 0.0000 0.0000 COORDINATES (SPA units)
OXIGEN
1 8.0000 0.0000 1.9800
0.5000 0.5000 0.5000

2 NFORM=2,1,0(output ready for rel,wil or cav)


0.7586915 ALPHA (FOR HARTREE TYPE EXCHANGE TERM)
The value of the alpha constant can be obtained
from K. Schwarz, Phys. Rev. B 5, 2466 (1972)

10 NH (for estimating muffin tin zero)

A.1.3 Input file for MSCD calculation for MgO(100)-Mg


KLL

741 10 300 datakind begining-row linenumbers


----------------------------------------------------------------
MSCD Version 1.00 Yufeng Chen and Michel A Van Hove
Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA 94720
Copyright (c) Van Hove Group 1997. All rights reserved
----------------------------------------------------------------
Mg(001)-KLL Magnesium Oxide input file

un "Stefano" user name


sn MgO(001)KLps01 psMg.txt input phase shift data file
ps02 psO.txt input phase shift data file
rm rmMgKLL.txt input radial matrix data file
pe MgOKLL_ppol_out.txt output photo emission data file

83
122 10 0.1 scanmode,dispmode,ftolerance

1 0 8 2 linitial,lnum,msorder,raorder
24 2 0 0 layers,finals,fitmath,trymax
17.44 17.61 0.1 kmin,kmax,kstep (per angstrom)
0.0 70.0 1.0 dthetamin,dthetamax,dthetastep (degree)
45.0 45.0 0.0 dphimin,dphimax,dphistep (degree)
36.0 90.0 1 ltheta, lphi, beampol (degree)
0.0 0.0 1.0 mtheta, mphi, acceptang (degree)
8.0 18.0 4.21 radius,depth,lattice(angs)
8 7.3 3.6 40.304 valence,bandgap(eV),density(g/cm3),mweight
24.305 15.999 0.0 0.0 effective weight for kind 1-4 (amu)
0.0 0.0 0.0 0.0 magnetization amplitude for kind 1-4
10.0 743.0 300 0.001 vinner(eV),tdebye,tsample(K),pathcut
0.0 0.0 0.0 fit try for vinner, tdebye and lattice

1 1 1 0 layer, kind, emitter, lineatom


0 0 0 0 latoms(xa,xb,ya,yb)
1.0000000 0.000000 unita(len ang) (fcc (001) structure)
1.0000000 90.000000 unitb(len ang) (in unit of lattice)
0.0000000 0.000000 origin(len ang) (in unit of lattice)
0.2500000 interlayer spacing (unit lattice)
0.0 0.0 0.0 fit try for spacing, length and units

2 1 1 0 layer, kind, emitter, lineatom


0 0 0 0 latoms(xa,xb,ya,yb)
1.0000000 0.000000 unita(len ang) (fcc (001) structure)
1.0000000 90.000000 unitb(len ang) (in unit of lattice)
0.7071070 45.000000 origin(len ang) (in unit of lattice)
0.0000000 interlayer spacing (unit lattice)
0.0 0.0 0.0 fit try for spacing, length and units

3 2 0 0 layer, kind, emitter, lineatom


0 0 0 0 latoms(xa,xb,ya,yb)
0.7071070 45.000000 unita(len ang) (fcc (001) structure)
0.7071070 135.000000 unitb(len ang) (in unit of lattice)

84
0.5000000 0.000000 origin(len ang) (in unit of lattice)
0.0000000 interlayer spacing (unit lattice)
0.0 0.0 0.0 fit try for spacing, length and units

A.2 Acquisition software for XPD measurements


In Fig. A.1 is presented the interface of the software realized with Labview 7.1
c
and
used for XPD data acquistion.

Figure A.1: Interface of the software realized for XPD data acquistion.

85
Bibliography

[1] J. S. Moodera and G. Mathon. J. Magn. Magn. Mater, 200:248, 1999.

[2] W. Shen et al. Appl. Phys. Lett., 86:253901, 2005.

[3] S. Yuasa et al. Nature Materials, 3:868, 2004.

[4] W. H. Butler et al. Phys. Rev. B, 63:054416, 2001.

[5] H. L. Meyerheim et al. Phys. Rev. B, 65:144433, 2002.

[6] H. Oh et al. Appl. Phys. Lett., 82:361, 2003.

[7] X. G. Zhang et al. Phys. Rev. B, 68:092402, 2003.

[8] K.Miyokawa et al. J. Appl. Phys., 44:L9, 2005.

[9] R. Bertacco and F. Ciccacci. Surface Science, 419:265, 1999.

[10] D.P. Woodruff. Photoelectron diffraction: past, present and future. J. Elect.
Spect., 126:55–65, 2002.

[11] Charles S. Fadley. Advances in Surface And Interface Science, Volume I: Tech-
niques. Plenum Press, New York, 1992.

[12] M. Sagurton et al. Surface Science, 182:287–361, 1987.

[13] P. A. Lee. Possibility of adsorbate position determination using final-state


interference effect. Phys. Rev. B, 13:5261, 1976.

[14] D.J. Friedman and C.S. Fadley. J. Elect. Spect., 51:689, 1990.

[15] M.-Lu. Xu and M. A. Van Hove. Surface Science, 207:215–232, 1989.

[16] E. Lang and H. Hilferinf.

[17] J.J. Barton and D.A. Shirley. Curved-wave-front corrections for photoelectron
scattering. Phys. Rev. B, 32:1892, 1985.

86
[18] S.W. Robey J.J. Barton and D.A. Shirley. Phys. Rev. B, 34:778, 1986.

[19] J.J Rehr et al. New high-energy-aproximation for x-ray-absorption near-edge


structure. Phys. Rev. B, 34:4350, 1986.

[20] J.J. Rehr and R.C. Albers. Phys. Rev. B, 41:8139, 1990.

[21] M. Sagurton et al. Phys. Rev. B, 33:2207, 1985.

[22] H.C. Poon S.Y. Tong and D.R. Snider. Phys. Rev. B, 32:2096, 1985.

[23] Y.Cheng and M.A. Van Hove. MSCD photoelectron diffraction pro-
gram package. http://www.ap.cityu.edu.hk/personal-website/Van-Hove
files/mscd/mscdpack.html/.

[24] J.J.Rehr et al. J. Elect. Spect., 126:67–76, 2002.

[25] C. J. Powell S. Tanuma and D. R. Penn. Surf. Interface. Anal., 21:165, 1994.

[26] L. E. Davis C. D. Wagner and W. M. Riggs. Surf. Interface. Anal., 2:53, 1980.

[27] G. P. Alldredge R. E. Allen and F. W. de Wette. Phys. Rev. B, 6:632, 1972.

[28] J.F. Janak V.L Moruzzi and A.R. Williams. Calculated Electronic Properties
of Metals. Pergamon Press, New York, 1978.

[29] H. Lüth. Solid Surface Interface and Thin Films, volume “Low-Energy
Electron Diffraction (LEED) and Reflection High-Energy Electron Diffraction
(REED)(Panel VIII)”. Springer, Berlin, 4th edition, 2001.

[30] H. Lüth. Solid Surface Interface and Thin Films, volume “Photoemission and
Inverse Photoemission (Panel XI)”. Springer, Berlin, 4th edition, 2001.

[31] H. Lüth. Solid Surface Interface and Thin Films, volume “Aspect of Photoe-
mission theory”. Springer, Berlin, 4th edition, 2001.

[32] SPECS GmbH Surface Analysis and Computer Technology. PHOIBOS Hemi-
spherical Energy Analyzer Series. http://www.specs.de.

[33] M. Sicot et al. J. Appl. Phys., 99:08D301, 2006.

[34] T. Urano and K. Kanaji. J. Phys. Soc. Jap., 57:3043, 1988.

[35] C. J. Powell and A. Jablonski. NIST Electron Effective-Attenuation-Length


Database.

87
[36] A. Cattoni et al. to be submitted.

[37] F. Bruno et al. J. Elect. Spect., 127:85–92, 2002.

[38] G. C. Gazzadiet al. Phys. Rev. B, 61:3, 2000.

[39] A. Chassé et al. Surface Science, 602:597–606, 2008.

[40] D. A. Wesner et al. Phys. Rev. B, 33:12, 1986.

[41] Yongsup Park et al. Appl. Phys. Lett., 66:16, 1995.

88

You might also like